Bug#784373: jessie-pu: package ceph/0.80.9-2 (pre approval)
Package: release.debian.org
Severity: normal
Tags: jessie
User: release.debian.org@packages.debian.org
Usertags: pu
I'd like to update ceph in jessie to the latest upstream bugfix release.
The version of ceph in jessie is a long term support (LTS) release which
will receive updates at least until January 2016. Updates will be bugfix
only. New features go into new release which are developed in parallel.
See at the end of this report for the upstream changelog.
See http://ceph.com/docs/master/releases/ for the ceph release timeline
and support statement.
IMO this is in line with other packages like the Linux kernel and
postgresql which have a similar policy and are also already updated to
new upstream bugfix releases in stable.
I have uploaded the version I would like to have in stable to unstable,
but did not upload to jessie-proposed-updates yet. But this will only be
a rebuild in a clean jessie chroot. I attached the debdiff to the
current version in stable.
I'm a bit unsure about the correct versioning of the stable update.
Certainly the version should be lower than the one in unstable
(0.80.9-2). The prefered version from my POV would be 0.80.9-2~deb8u1,
but I have seen that other packages use +deb8u1 which would mean it should
be 0.80.9-1+deb8u1 to be lower than unstable. I don't care much about
this and will just use what the release team prefers.
If you think that the upstream changes are inappropriate for a stable update,
then I'd like to at least get the NEWS.Debian note about rbdmap and systemd
into stable.
Thanks for considering,
Gaudenz
The upstream changes since 0.80.7 are:
0.80.9:
ceph-fuse: flock, fcntl lock support (Yan, Zheng, Greg Farnum)
crush: fix straw bucket weight calculation, add straw_calc_version tunable (#10095 Sage Weil)
crush: fix tree bucket (Rongzu Zhu)
crush: fix underflow of tree weights (Loic Dachary, Sage Weil)
crushtool: add –reweight (Sage Weil)
librbd: complete pending operations before losing image (#10299 Jason Dillaman)
librbd: fix read caching performance regression (#9854 Jason Dillaman)
librbd: gracefully handle deleted/renamed pools (#10270 Jason Dillaman)
mon: fix dump of chooseleaf_vary_r tunable (Sage Weil)
osd: fix PG ref leak in snaptrimmer on peering (#10421 Kefu Chai)
osd: handle no-op write with snapshot (#10262 Sage Weil)
radosgw-admin: create subuser when creating user (#10103 Yehuda Sadeh)
rgw: change multipart uplaod id magic (#10271 Georgio Dimitrakakis, Yehuda Sadeh)
rgw: don’t overwrite bucket/object owner when setting ACLs (#10978 Yehuda Sadeh)
rgw: enable IPv6 for embedded civetweb (#10965 Yehuda Sadeh)
rgw: fix partial swift GET (#10553 Yehuda Sadeh)
rgw: fix quota disable (#9907 Dong Lei)
rgw: index swift keys appropriately (#10471 Hemant Burman, Yehuda Sadeh)
rgw: make setattrs update bucket index (#5595 Yehuda Sadeh)
rgw: pass civetweb configurables (#10907 Yehuda Sadeh)
rgw: remove swift user manifest (DLO) hash calculation (#9973 Yehuda Sadeh)
rgw: return correct len for 0-len objects (#9877 Yehuda Sadeh)
rgw: S3 object copy content-type fix (#9478 Yehuda Sadeh)
rgw: send ETag on S3 object copy (#9479 Yehuda Sadeh)
rgw: send HTTP status reason explicitly in fastcgi (Yehuda Sadeh)
rgw: set ulimit -n from sysvinit (el6) init script (#9587 Sage Weil)
rgw: update swift subuser permission masks when authenticating (#9918 Yehuda Sadeh)
rgw: URL decode query params correctly (#10271 Georgio Dimitrakakis, Yehuda Sadeh)
rgw: use attrs when reading object attrs (#10307 Yehuda Sadeh)
rgw: use rn for http headers (#9254 Benedikt Fraunhofer, Yehuda Sadeh)
0.80.8:
build: remove stack-execute bit from assembled code sections (#10114 Dan Mick)
ceph-disk: fix dmcrypt key permissions (#9785 Loic Dachary)
ceph-disk: fix keyring location (#9653 Loic Dachary)
ceph-disk: make partition checks more robust (#9721 #9665 Loic Dachary)
ceph: cleanly shut down librados context on shutdown (#8797 Dan Mick)
common: add $cctid config metavariable (#6228 Adam Crume)
crush: align rule and ruleset ids (#9675 Xiaoxi Chen)
crush: fix negative weight bug during create_or_move_item (#9998 Pawel Sadowski)
crush: fix potential buffer overflow in erasure rules (#9492 Johnu George)
debian: fix python-ceph -> ceph file movement (Sage Weil)
libcephfs,ceph-fuse: fix flush tid wraparound bug (#9869 Greg Farnum, Yan, Zheng)
libcephfs: close fd befure umount (#10415 Yan, Zheng)
librados: fix crash from C API when read timeout is enabled (#9582 Sage Weil)
librados: handle reply race with pool deletion (#10372 Sage Weil)
librbd: cap memory utilization for read requests (Jason Dillaman)
librbd: do not close a closed parent image on failure (#10030 Jason Dillaman)
librbd: fix diff tests (#10002 Josh Durgin)
librbd: protect list_children from invalid pools (#10123 Jason Dillaman)
make check improvemens (Loic Dachary)
mds: fix ctime updates (#9514 Greg Farnum)
mds: fix journal import tool (#10025 John Spray)
mds: fix rare NULL deref in cap flush handler (Greg Farnum)
mds: handle unknown lock messages (Yan, Zheng)
mds: store backtrace for straydir (Yan, Zheng)
mon: abort startup if disk is full (#9502 Joao Eduardo Luis)
mon: add paxos instrumentation (Sage Weil)
mon: fix double-free in rare OSD startup path (Sage Weil)
mon: fix osdmap trimming (#9987 Sage Weil)
mon: fix paxos corner cases (#9301 #9053 Sage Weil)
osd: cancel callback on blacklisted watchers (#8315 Samuel Just)
osd: cleanly abort set-alloc-hint operations during upgrade (#9419 David Zafman)
osd: clear rollback PG metadata on PG deletion (#9293 Samuel Just)
osd: do not abort deep scrub if hinfo is missing (#10018 Loic Dachary)
osd: erasure-code regression tests (Loic Dachary)
osd: fix distro metadata reporting for SUSE (#8654 Danny Al-Gaaf)
osd: fix full OSD checks during backfill (#9574 Samuel Just)
osd: fix ioprio parsing (#9677 Loic Dachary)
osd: fix journal direct-io shutdown (#9073 Mark Kirkwood, Ma Jianpeng, Somnath Roy)
osd: fix journal dump (Ma Jianpeng)
osd: fix occasional stall during peering or activation (Sage Weil)
osd: fix past_interval display bug (#9752 Loic Dachary)
osd: fix rare crash triggered by admin socket dump_ops_in_filght (#9916 Dong Lei)
osd: fix snap trimming performance issues (#9487 #9113 Samuel Just, Sage Weil, Dan van der Ster, Florian Haas)
osd: fix snapdir handling on cache eviction (#8629 Sage Weil)
osd: handle map gaps in map advance code (Sage Weil)
osd: handle undefined CRUSH results in interval check (#9718 Samuel Just)
osd: include shard in JSON dump of ghobject (#10063 Loic Dachary)
osd: make backfill reservation denial handling more robust (#9626 Samuel Just)
osd: make misdirected op checks handle EC + primary affinity (#9835 Samuel Just, Sage Weil)
osd: mount XFS with inode64 by default (Sage Weil)
osd: other misc bugs (#9821 #9875 Samuel Just)
rgw: add .log to default log path (#9353 Alexandre Marangone)
rgw: clean up fcgi request context (#10194 Yehuda Sadeh)
rgw: convet header underscores to dashes (#9206 Yehuda Sadeh)
rgw: copy object data if copy target is in different pool (#9039 Yehuda Sadeh)
rgw: don’t try to authenticate CORS peflight request (#8718 Robert Hubbard, Yehuda Sadeh)
rgw: fix civetweb URL decoding (#8621 Yehuda Sadeh)
rgw: fix hash calculation during PUT (Yehuda Sadeh)
rgw: fix misc bugs (#9089 #9201 Yehuda Sadeh)
rgw: fix object tail test (#9226 Sylvain Munaut, Yehuda Sadeh)
rgw: make sysvinit script run rgw under systemd context as needed (#10125 Loic Dachary)
rgw: separate civetweb log from rgw log (Yehuda Sadeh)
rgw: set length for keystone token validations (#7796 Mark Kirkwood, Yehuda Sadeh)
rgw: subuser creation fixes (#8587 Yehuda Sadeh)
rpm: misc packaging improvements (Sandon Van Ness, Dan Mick, Erik Logthenberg, Boris Ranto)
rpm: use standard udev rules for CentOS7/RHEL7 (#9747 Loic Dachary)
-- System Information:
Debian Release: 8.0
APT prefers testing-proposed-updates
APT policy: (500, 'testing-proposed-updates'), (500, 'testing'), (100, 'unstable'), (1, 'experimental')
Architecture: amd64 (x86_64)
Foreign Architectures: i386
Kernel: Linux 3.16.0-4-amd64 (SMP w/4 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)
diff -Nru ceph-0.80.7/ceph.spec ceph-0.80.9/ceph.spec
--- ceph-0.80.7/ceph.spec 2014-10-14 21:15:22.000000000 +0200
+++ ceph-0.80.9/ceph.spec 2015-03-09 20:54:08.000000000 +0100
@@ -9,13 +9,16 @@
# common
#################################################################################
Name: ceph
-Version: 0.80.7
+Version: 0.80.9
Release: 0%{?dist}
Summary: User space components of the Ceph file system
License: GPL-2.0
Group: System Environment/Base
URL: http://ceph.com/
Source0: http://ceph.com/download/%{name}-%{version}.tar.bz2
+%if 0%{?fedora} || 0%{?centos} || 0%{?rhel}
+Patch0: init-ceph.in-fedora.patch
+%endif
Requires: librbd1 = %{version}-%{release}
Requires: librados2 = %{version}-%{release}
Requires: libcephfs1 = %{version}-%{release}
@@ -24,13 +27,13 @@
Requires: python-argparse
Requires: python-ceph
Requires: python-requests
+Requires: python-flask
Requires: xfsprogs
Requires: cryptsetup
Requires: parted
Requires: util-linux
Requires: hdparm
Requires(post): binutils
-BuildRoot: %{_tmppath}/%{name}-%{version}-build
BuildRequires: make
BuildRequires: gcc-c++
BuildRequires: libtool
@@ -50,14 +53,14 @@
BuildRequires: leveldb-devel > 1.2
BuildRequires: xfsprogs-devel
BuildRequires: yasm
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
BuildRequires: snappy-devel
%endif
#################################################################################
# specific
#################################################################################
-%if ! 0%{?rhel}
+%if ! 0%{?rhel} || 0%{?fedora}
BuildRequires: sharutils
%endif
@@ -173,8 +176,8 @@
Summary: RADOS distributed object store client library
Group: System Environment/Libraries
License: LGPL-2.0
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
-Obsoletes: ceph-libs
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
+Obsoletes: ceph-libs < %{version}-%{release}
%endif
%description -n librados2
RADOS is a reliable, autonomic distributed object storage cluster
@@ -187,8 +190,8 @@
Group: System Environment/Libraries
License: LGPL-2.0
Requires: librados2 = %{version}-%{release}
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
-Obsoletes: ceph-libs
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
+Obsoletes: ceph-libs < %{version}-%{release}
%endif
%description -n librbd1
RBD is a block device striped across multiple distributed objects in
@@ -200,8 +203,9 @@
Summary: Ceph distributed file system client library
Group: System Environment/Libraries
License: LGPL-2.0
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
-Obsoletes: ceph-libs
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
+Obsoletes: ceph-libs < %{version}-%{release}
+Obsoletes: ceph-libcephfs
%endif
%description -n libcephfs1
Ceph is a distributed network file system designed to provide excellent
@@ -215,7 +219,6 @@
License: LGPL-2.0
Requires: librados2 = %{version}-%{release}
Requires: librbd1 = %{version}-%{release}
-Requires: python-flask
%if 0%{defined suse_version}
%py_requires
%endif
@@ -264,6 +267,23 @@
%description -n cephfs-java
This package contains the Java libraries for the Ceph File System.
+%package libs-compat
+Summary: Meta package to include ceph libraries.
+Group: System Environment/Libraries
+License: LGPL-2.0
+Obsoletes: ceph-libs
+Requires: librados2 = %{version}-%{release}
+Requires: librbd1 = %{version}-%{release}
+Requires: libcephfs1 = %{version}-%{release}
+Provides: ceph-libs
+
+%description libs-compat
+This is a meta package, that pulls in librados2, librbd1 and libcephfs1. It
+is included for backwards compatibility with distributions that depend on the
+former ceph-libs package, which is now split up into these three subpackages.
+Packages still depending on ceph-libs should be fixed to depend on librados2,
+librbd1 or libcephfs1 instead.
+
%if 0%{?opensuse} || 0%{?suse_version}
%debug_package
%endif
@@ -273,6 +293,9 @@
#################################################################################
%prep
%setup -q
+%if 0%{?fedora} || 0%{?rhel} || 0%{?centos}
+%patch0 -p1 -b .init
+%endif
%build
# Find jni.h
@@ -329,7 +352,7 @@
chmod 0644 $RPM_BUILD_ROOT%{_docdir}/ceph/sample.fetch_config
# udev rules
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
install -m 0644 -D udev/50-rbd.rules $RPM_BUILD_ROOT/usr/lib/udev/rules.d/50-rbd.rules
install -m 0644 -D udev/60-ceph-partuuid-workaround.rules $RPM_BUILD_ROOT/usr/lib/udev/rules.d/60-ceph-partuuid-workaround.rules
%else
@@ -337,13 +360,13 @@
install -m 0644 -D udev/60-ceph-partuuid-workaround.rules $RPM_BUILD_ROOT/lib/udev/rules.d/60-ceph-partuuid-workaround.rules
%endif
-%if (0%{?rhel} || 0%{?rhel} < 7)
+%if (0%{?rhel} && 0%{?rhel} < 7)
install -m 0644 -D udev/95-ceph-osd-alt.rules $RPM_BUILD_ROOT/lib/udev/rules.d/95-ceph-osd.rules
%else
install -m 0644 -D udev/95-ceph-osd.rules $RPM_BUILD_ROOT/lib/udev/rules.d/95-ceph-osd.rules
%endif
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
mv $RPM_BUILD_ROOT/lib/udev/rules.d/95-ceph-osd.rules $RPM_BUILD_ROOT/usr/lib/udev/rules.d/95-ceph-osd.rules
mv $RPM_BUILD_ROOT/sbin/mkcephfs $RPM_BUILD_ROOT/usr/sbin/mkcephfs
mv $RPM_BUILD_ROOT/sbin/mount.ceph $RPM_BUILD_ROOT/usr/sbin/mount.ceph
@@ -404,6 +427,7 @@
%{_bindir}/cephfs
%{_bindir}/ceph-clsinfo
%{_bindir}/ceph-rest-api
+%{python_sitelib}/ceph_rest_api.py*
%{_bindir}/crushtool
%{_bindir}/monmaptool
%{_bindir}/osdmaptool
@@ -424,7 +448,7 @@
%{_sbindir}/ceph-disk-udev
%{_sbindir}/ceph-create-keys
%{_sbindir}/rcceph
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
%{_sbindir}/mount.ceph
%else
/sbin/mount.ceph
@@ -451,7 +475,7 @@
%{_libdir}/ceph/erasure-code/libec_jerasure*.so*
%{_libdir}/ceph/erasure-code/libec_test_jerasure*.so*
%{_libdir}/ceph/erasure-code/libec_missing_entry_point.so*
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
/usr/lib/udev/rules.d/60-ceph-partuuid-workaround.rules
/usr/lib/udev/rules.d/95-ceph-osd.rules
%else
@@ -529,7 +553,7 @@
%defattr(-,root,root,-)
%{_bindir}/ceph-fuse
%{_mandir}/man8/ceph-fuse.8*
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
%{_sbindir}/mount.fuse.ceph
%else
/sbin/mount.fuse.ceph
@@ -624,7 +648,7 @@
%files -n librbd1
%defattr(-,root,root,-)
%{_libdir}/librbd.so.*
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
/usr/lib/udev/rules.d/50-rbd.rules
%else
/lib/udev/rules.d/50-rbd.rules
@@ -656,7 +680,6 @@
%{python_sitelib}/rbd.py*
%{python_sitelib}/cephfs.py*
%{python_sitelib}/ceph_argparse.py*
-%{python_sitelib}/ceph_rest_api.py*
#################################################################################
%files -n rest-bench
@@ -702,4 +725,8 @@
%{_javadir}/libcephfs.jar
%{_javadir}/libcephfs-test.jar
+%files libs-compat
+# We need an empty %files list for ceph-libs-compat, to tell rpmbuild to actually
+# build this meta package.
+
%changelog
diff -Nru ceph-0.80.7/ceph.spec.in ceph-0.80.9/ceph.spec.in
--- ceph-0.80.7/ceph.spec.in 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/ceph.spec.in 2015-03-09 20:51:17.000000000 +0100
@@ -16,6 +16,9 @@
Group: System Environment/Base
URL: http://ceph.com/
Source0: http://ceph.com/download/%{name}-%{version}.tar.bz2
+%if 0%{?fedora} || 0%{?centos} || 0%{?rhel}
+Patch0: init-ceph.in-fedora.patch
+%endif
Requires: librbd1 = %{version}-%{release}
Requires: librados2 = %{version}-%{release}
Requires: libcephfs1 = %{version}-%{release}
@@ -24,13 +27,13 @@
Requires: python-argparse
Requires: python-ceph
Requires: python-requests
+Requires: python-flask
Requires: xfsprogs
Requires: cryptsetup
Requires: parted
Requires: util-linux
Requires: hdparm
Requires(post): binutils
-BuildRoot: %{_tmppath}/%{name}-%{version}-build
BuildRequires: make
BuildRequires: gcc-c++
BuildRequires: libtool
@@ -50,14 +53,14 @@
BuildRequires: leveldb-devel > 1.2
BuildRequires: xfsprogs-devel
BuildRequires: yasm
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
BuildRequires: snappy-devel
%endif
#################################################################################
# specific
#################################################################################
-%if ! 0%{?rhel}
+%if ! 0%{?rhel} || 0%{?fedora}
BuildRequires: sharutils
%endif
@@ -173,8 +176,8 @@
Summary: RADOS distributed object store client library
Group: System Environment/Libraries
License: LGPL-2.0
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
-Obsoletes: ceph-libs
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
+Obsoletes: ceph-libs < %{version}-%{release}
%endif
%description -n librados2
RADOS is a reliable, autonomic distributed object storage cluster
@@ -187,8 +190,8 @@
Group: System Environment/Libraries
License: LGPL-2.0
Requires: librados2 = %{version}-%{release}
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
-Obsoletes: ceph-libs
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
+Obsoletes: ceph-libs < %{version}-%{release}
%endif
%description -n librbd1
RBD is a block device striped across multiple distributed objects in
@@ -200,8 +203,9 @@
Summary: Ceph distributed file system client library
Group: System Environment/Libraries
License: LGPL-2.0
-%if 0%{?rhel_version} || 0%{?centos_version} || 0%{?fedora}
-Obsoletes: ceph-libs
+%if 0%{?rhel} || 0%{?centos} || 0%{?fedora}
+Obsoletes: ceph-libs < %{version}-%{release}
+Obsoletes: ceph-libcephfs
%endif
%description -n libcephfs1
Ceph is a distributed network file system designed to provide excellent
@@ -215,7 +219,6 @@
License: LGPL-2.0
Requires: librados2 = %{version}-%{release}
Requires: librbd1 = %{version}-%{release}
-Requires: python-flask
%if 0%{defined suse_version}
%py_requires
%endif
@@ -264,6 +267,23 @@
%description -n cephfs-java
This package contains the Java libraries for the Ceph File System.
+%package libs-compat
+Summary: Meta package to include ceph libraries.
+Group: System Environment/Libraries
+License: LGPL-2.0
+Obsoletes: ceph-libs
+Requires: librados2 = %{version}-%{release}
+Requires: librbd1 = %{version}-%{release}
+Requires: libcephfs1 = %{version}-%{release}
+Provides: ceph-libs
+
+%description libs-compat
+This is a meta package, that pulls in librados2, librbd1 and libcephfs1. It
+is included for backwards compatibility with distributions that depend on the
+former ceph-libs package, which is now split up into these three subpackages.
+Packages still depending on ceph-libs should be fixed to depend on librados2,
+librbd1 or libcephfs1 instead.
+
%if 0%{?opensuse} || 0%{?suse_version}
%debug_package
%endif
@@ -273,6 +293,9 @@
#################################################################################
%prep
%setup -q
+%if 0%{?fedora} || 0%{?rhel} || 0%{?centos}
+%patch0 -p1 -b .init
+%endif
%build
# Find jni.h
@@ -329,7 +352,7 @@
chmod 0644 $RPM_BUILD_ROOT%{_docdir}/ceph/sample.fetch_config
# udev rules
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
install -m 0644 -D udev/50-rbd.rules $RPM_BUILD_ROOT/usr/lib/udev/rules.d/50-rbd.rules
install -m 0644 -D udev/60-ceph-partuuid-workaround.rules $RPM_BUILD_ROOT/usr/lib/udev/rules.d/60-ceph-partuuid-workaround.rules
%else
@@ -337,13 +360,13 @@
install -m 0644 -D udev/60-ceph-partuuid-workaround.rules $RPM_BUILD_ROOT/lib/udev/rules.d/60-ceph-partuuid-workaround.rules
%endif
-%if (0%{?rhel} || 0%{?rhel} < 7)
+%if (0%{?rhel} && 0%{?rhel} < 7)
install -m 0644 -D udev/95-ceph-osd-alt.rules $RPM_BUILD_ROOT/lib/udev/rules.d/95-ceph-osd.rules
%else
install -m 0644 -D udev/95-ceph-osd.rules $RPM_BUILD_ROOT/lib/udev/rules.d/95-ceph-osd.rules
%endif
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
mv $RPM_BUILD_ROOT/lib/udev/rules.d/95-ceph-osd.rules $RPM_BUILD_ROOT/usr/lib/udev/rules.d/95-ceph-osd.rules
mv $RPM_BUILD_ROOT/sbin/mkcephfs $RPM_BUILD_ROOT/usr/sbin/mkcephfs
mv $RPM_BUILD_ROOT/sbin/mount.ceph $RPM_BUILD_ROOT/usr/sbin/mount.ceph
@@ -404,6 +427,7 @@
%{_bindir}/cephfs
%{_bindir}/ceph-clsinfo
%{_bindir}/ceph-rest-api
+%{python_sitelib}/ceph_rest_api.py*
%{_bindir}/crushtool
%{_bindir}/monmaptool
%{_bindir}/osdmaptool
@@ -424,7 +448,7 @@
%{_sbindir}/ceph-disk-udev
%{_sbindir}/ceph-create-keys
%{_sbindir}/rcceph
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
%{_sbindir}/mount.ceph
%else
/sbin/mount.ceph
@@ -451,7 +475,7 @@
%{_libdir}/ceph/erasure-code/libec_jerasure*.so*
%{_libdir}/ceph/erasure-code/libec_test_jerasure*.so*
%{_libdir}/ceph/erasure-code/libec_missing_entry_point.so*
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
/usr/lib/udev/rules.d/60-ceph-partuuid-workaround.rules
/usr/lib/udev/rules.d/95-ceph-osd.rules
%else
@@ -529,7 +553,7 @@
%defattr(-,root,root,-)
%{_bindir}/ceph-fuse
%{_mandir}/man8/ceph-fuse.8*
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
%{_sbindir}/mount.fuse.ceph
%else
/sbin/mount.fuse.ceph
@@ -624,7 +648,7 @@
%files -n librbd1
%defattr(-,root,root,-)
%{_libdir}/librbd.so.*
-%if 0%{?rhel} >= 7
+%if 0%{?rhel} >= 7 || 0%{?fedora}
/usr/lib/udev/rules.d/50-rbd.rules
%else
/lib/udev/rules.d/50-rbd.rules
@@ -656,7 +680,6 @@
%{python_sitelib}/rbd.py*
%{python_sitelib}/cephfs.py*
%{python_sitelib}/ceph_argparse.py*
-%{python_sitelib}/ceph_rest_api.py*
#################################################################################
%files -n rest-bench
@@ -702,4 +725,8 @@
%{_javadir}/libcephfs.jar
%{_javadir}/libcephfs-test.jar
+%files libs-compat
+# We need an empty %files list for ceph-libs-compat, to tell rpmbuild to actually
+# build this meta package.
+
%changelog
diff -Nru ceph-0.80.7/configure ceph-0.80.9/configure
--- ceph-0.80.7/configure 2014-10-14 21:14:30.000000000 +0200
+++ ceph-0.80.9/configure 2015-03-09 20:53:15.000000000 +0100
@@ -1,6 +1,6 @@
#! /bin/sh
# Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.68 for ceph 0.80.7.
+# Generated by GNU Autoconf 2.68 for ceph 0.80.9.
#
# Report bugs to <ceph-devel@vger.kernel.org>.
#
@@ -570,8 +570,8 @@
# Identity of this package.
PACKAGE_NAME='ceph'
PACKAGE_TARNAME='ceph'
-PACKAGE_VERSION='0.80.7'
-PACKAGE_STRING='ceph 0.80.7'
+PACKAGE_VERSION='0.80.9'
+PACKAGE_STRING='ceph 0.80.9'
PACKAGE_BUGREPORT='ceph-devel@vger.kernel.org'
PACKAGE_URL=''
@@ -1441,7 +1441,7 @@
# Omit some internal or obsolete options to make the list less imposing.
# This message is too long to be a string in the A/UX 3.1 sh.
cat <<_ACEOF
-\`configure' configures ceph 0.80.7 to adapt to many kinds of systems.
+\`configure' configures ceph 0.80.9 to adapt to many kinds of systems.
Usage: $0 [OPTION]... [VAR=VALUE]...
@@ -1512,7 +1512,7 @@
if test -n "$ac_init_help"; then
case $ac_init_help in
- short | recursive ) echo "Configuration of ceph 0.80.7:";;
+ short | recursive ) echo "Configuration of ceph 0.80.9:";;
esac
cat <<\_ACEOF
@@ -1657,7 +1657,7 @@
test -n "$ac_init_help" && exit $ac_status
if $ac_init_version; then
cat <<\_ACEOF
-ceph configure 0.80.7
+ceph configure 0.80.9
generated by GNU Autoconf 2.68
Copyright (C) 2010 Free Software Foundation, Inc.
@@ -2682,7 +2682,7 @@
This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.
-It was created by ceph $as_me 0.80.7, which was
+It was created by ceph $as_me 0.80.9, which was
generated by GNU Autoconf 2.68. Invocation command line was
$ $0 $@
@@ -4682,7 +4682,7 @@
# Define the identity of the package.
PACKAGE='ceph'
- VERSION='0.80.7'
+ VERSION='0.80.9'
cat >>confdefs.h <<_ACEOF
@@ -12660,7 +12660,7 @@
# Define the identity of the package.
PACKAGE='ceph'
- VERSION='0.80.7'
+ VERSION='0.80.9'
cat >>confdefs.h <<_ACEOF
@@ -22464,7 +22464,7 @@
# report actual input values of CONFIG_FILES etc. instead of their
# values after options handling.
ac_log="
-This file was extended by ceph $as_me 0.80.7, which was
+This file was extended by ceph $as_me 0.80.9, which was
generated by GNU Autoconf 2.68. Invocation command line was
CONFIG_FILES = $CONFIG_FILES
@@ -22530,7 +22530,7 @@
cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`"
ac_cs_version="\\
-ceph config.status 0.80.7
+ceph config.status 0.80.9
configured by $0, generated by GNU Autoconf 2.68,
with options \\"\$ac_cs_config\\"
diff -Nru ceph-0.80.7/configure.ac ceph-0.80.9/configure.ac
--- ceph-0.80.7/configure.ac 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/configure.ac 2015-03-09 20:51:17.000000000 +0100
@@ -8,7 +8,7 @@
# VERSION define is not used by the code. It gets a version string
# from 'git describe'; see src/ceph_ver.[ch]
-AC_INIT([ceph], [0.80.7], [ceph-devel@vger.kernel.org])
+AC_INIT([ceph], [0.80.9], [ceph-devel@vger.kernel.org])
# Create release string. Used with VERSION for RPMs.
RPM_RELEASE=0
diff -Nru ceph-0.80.7/debian/ceph.NEWS ceph-0.80.9/debian/ceph.NEWS
--- ceph-0.80.7/debian/ceph.NEWS 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/debian/ceph.NEWS 2015-05-05 18:10:29.000000000 +0200
@@ -0,0 +1,56 @@
+ceph (0.80.9-2) unstable; urgency=medium
+
+ ## CRUSH fixes in 0.80.9
+
+ The 0.80.9 point release fixes several issues with CRUSH that trigger excessive
+ data migration when adjusting OSD weights. These are most obvious when a very
+ small weight change (e.g., a change from 0 to .01) triggers a large amount of
+ movement, but the same set of bugs can also lead to excessive (though less
+ noticeable) movement in other cases.
+
+ However, because the bug may already have affected your cluster, fixing it
+ may trigger movement back to the more correct location. For this reason, you
+ must manually opt-in to the fixed behavior.
+
+ In order to set the new tunable to correct the behavior:
+
+ ceph osd crush set-tunable straw_calc_version 1
+
+ Note that this change will have no immediate effect. However, from this
+ point forward, any ‘straw’ bucket in your CRUSH map that is adjusted will get
+ non-buggy internal weights, and that transition may trigger some rebalancing.
+
+ You can estimate how much rebalancing will eventually be necessary on your
+ cluster with:
+
+ ceph osd getcrushmap -o /tmp/cm
+ crushtool -i /tmp/cm --num-rep 3 --test --show-mappings > /tmp/a 2>&1
+ crushtool -i /tmp/cm --set-straw-calc-version 1 -o /tmp/cm2
+ crushtool -i /tmp/cm2 --reweight -o /tmp/cm2
+ crushtool -i /tmp/cm2 --num-rep 3 --test --show-mappings > /tmp/b 2>&1
+ wc -l /tmp/a # num total mappings
+ diff -u /tmp/a /tmp/b | grep -c ^+ # num changed mappings
+
+ Divide the total number of lines in /tmp/a with the number of lines
+ changed. We've found that most clusters are under 10%.
+
+ You can force all of this rebalancing to happen at once with:
+
+ ceph osd crush reweight-all
+
+ Otherwise, it will happen at some unknown point in the future when
+ CRUSH weights are next adjusted.
+
+ ## Mapping rbd devices with rbdmap on systemd systems
+
+ If you have setup rbd mappings in /etc/ceph/rbdmap and corresponding mounts
+ in /etc/fstab things might break with systemd because systemd waits for the
+ rbd device to appear before the legacy rbdmap init file has a chance to run
+ and drops into emergency mode if it times out.
+
+ This can be fixed by adding the nofail option in /etc/fstab to all rbd
+ backed mount points. With this systemd does not wait for the device and
+ proceeds with the boot process. After rbdmap mapped the device, systemd
+ detects the new device and mounts the file system.
+
+ -- Gaudenz Steinlin <gaudenz@debian.org> Mon, 04 May 2015 22:49:48 +0200
diff -Nru ceph-0.80.7/debian/changelog ceph-0.80.9/debian/changelog
--- ceph-0.80.7/debian/changelog 2014-12-11 02:55:49.000000000 +0100
+++ ceph-0.80.9/debian/changelog 2015-05-05 21:29:25.000000000 +0200
@@ -1,3 +1,18 @@
+ceph (0.80.9-2) unstable; urgency=medium
+
+ * [70fc1d] Add NEWS entry about CRUSH issues fixed in 0.80.9
+ * [f41bb6] Add NEWS entry about rbd backed filesystems and systemd
+
+ -- Gaudenz Steinlin <gaudenz@debian.org> Tue, 05 May 2015 21:29:15 +0200
+
+ceph (0.80.9-1) unstable; urgency=medium
+
+ * [4b4e] Imported Upstream version 0.80.9
+ * [7102] Remove patches firefly-latest and p2139 applied upstream
+ * [5869] Add myself to uploaders
+
+ -- Gaudenz Steinlin <gaudenz@debian.org> Mon, 04 May 2015 08:49:37 +0200
+
ceph (0.80.7-2) unstable; urgency=medium
* Team upload.
diff -Nru ceph-0.80.7/debian/control ceph-0.80.9/debian/control
--- ceph-0.80.7/debian/control 2014-12-11 02:50:22.000000000 +0100
+++ ceph-0.80.9/debian/control 2015-05-04 22:08:51.000000000 +0200
@@ -3,7 +3,8 @@
Priority: optional
Maintainer: Ceph Maintainers <ceph-maintainers@lists.ceph.com>
Uploaders: Laszlo Boszormenyi (GCS) <gcs@debian.org>,
- James Page <jamespage@debian.org>
+ James Page <jamespage@debian.org>,
+ Gaudenz Steinlin <gaudenz@debian.org>
Homepage: http://ceph.com/
Vcs-Git: git://anonscm.debian.org/pkg-ceph/ceph.git
Vcs-Browser: http://anonscm.debian.org/gitweb/?p=pkg-ceph/ceph.git
diff -Nru ceph-0.80.7/debian/patches/firefly-latest.patch ceph-0.80.9/debian/patches/firefly-latest.patch
--- ceph-0.80.7/debian/patches/firefly-latest.patch 2014-12-11 02:33:34.000000000 +0100
+++ ceph-0.80.9/debian/patches/firefly-latest.patch 1970-01-01 01:00:00.000000000 +0100
@@ -1,30 +0,0 @@
-Last-Update: 2014-10-02
-Forwarded: not-needed
-Origin: upstream
-Author: Dmitry Smirnov <onlyjob@member.fsf.org>
-Description: fixes from "firefly" branch since 0.80.6 release
-
---- a/src/common/config.cc
-+++ b/src/common/config.cc
-@@ -946,9 +946,9 @@
- return -ENOSYS;
- }
-
- static const char *CONF_METAVARIABLES[] =
-- { "cluster", "type", "name", "host", "num", "id", "pid" };
-+ { "cluster", "type", "name", "host", "num", "id", "pid", "cctid" };
- static const int NUM_CONF_METAVARIABLES =
- (sizeof(CONF_METAVARIABLES) / sizeof(CONF_METAVARIABLES[0]));
-
- void md_config_t::expand_all_meta()
-@@ -1058,8 +1058,10 @@
- else if (var == "id")
- out += name.get_id().c_str();
- else if (var == "pid")
- out += stringify(getpid());
-+ else if (var == "cctid")
-+ out += stringify((unsigned long long)this);
- else
- assert(0); // unreachable
- expanded = true;
- }
diff -Nru ceph-0.80.7/debian/patches/p2139.patch ceph-0.80.9/debian/patches/p2139.patch
--- ceph-0.80.7/debian/patches/p2139.patch 2014-12-11 02:33:34.000000000 +0100
+++ ceph-0.80.9/debian/patches/p2139.patch 1970-01-01 01:00:00.000000000 +0100
@@ -1,39 +0,0 @@
-From 4eb18dd487da4cb621dcbecfc475fc0871b356ac Mon Sep 17 00:00:00 2001
-From: Ma Jianpeng <jianpeng.ma@intel.com>
-Date: Wed, 23 Jul 2014 10:10:38 -0700
-Subject: [PATCH] os/FileJournal: Update the journal header when closing
- journal
-
-When closing journal, it should check must_write_header and update
-journal header if must_write_header alreay set.
-It can reduce the nosense journal-replay after restarting osd.
-
-Signed-off-by: Ma Jianpeng <jianpeng.ma@intel.com>
-Reviewed-by: Sage Weil <sage@redhat.com>
----
- src/os/FileJournal.cc | 3 ++-
- 1 file changed, 2 insertions(+), 1 deletion(-)
-
---- a/src/os/FileJournal.cc
-+++ b/src/os/FileJournal.cc
-@@ -543,8 +543,9 @@
- stop_writer();
-
- // close
- assert(writeq_empty());
-+ assert(!must_write_header);
- assert(fd >= 0);
- VOID_TEMP_FAILURE_RETRY(::close(fd));
- fd = -1;
- }
-@@ -1101,9 +1102,9 @@
- dout(10) << "write_thread_entry start" << dendl;
- while (1) {
- {
- Mutex::Locker locker(writeq_lock);
-- if (writeq.empty()) {
-+ if (writeq.empty() && !must_write_header) {
- if (write_stop)
- break;
- dout(20) << "write_thread_entry going to sleep" << dendl;
- writeq_cond.Wait(writeq_lock);
diff -Nru ceph-0.80.7/debian/patches/series ceph-0.80.9/debian/patches/series
--- ceph-0.80.7/debian/patches/series 2014-12-11 02:33:47.000000000 +0100
+++ ceph-0.80.9/debian/patches/series 2015-05-04 22:08:51.000000000 +0200
@@ -1,11 +1,9 @@
## Backported / Upstream
-firefly-latest.patch
bug-8342.patch
bug-8624a.patch
bug-8624b.patch
bug-8821.patch
bug-9341.patch
-p2139.patch
client-sleep1.patch
client-sleep2.patch
client-sleep3.patch
diff -Nru ceph-0.80.7/man/ceph.8 ceph-0.80.9/man/ceph.8
--- ceph-0.80.7/man/ceph.8 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/man/ceph.8 2015-03-09 20:51:17.000000000 +0100
@@ -1,8 +1,8 @@
.\" Man page generated from reStructuredText.
.
-.TH "CEPH" "8" "January 12, 2014" "dev" "Ceph"
+.TH "CEPH" "8" "December 18, 2014" "dev" "Ceph"
.SH NAME
-ceph \- ceph file system control utility
+ceph \- ceph administration tool
.
.nr rst2man-indent-level 0
.
@@ -59,103 +59,2320 @@
..
.SH SYNOPSIS
.nf
-\fBceph\fP [ \-m \fImonaddr\fP ] [ \-w | \fIcommand\fP ... ]
+\fBceph\fP \fBauth\fP \fIadd\fP \fI<entity>\fP {\fI<caps>\fP [\fI<caps>\fP\&...]}
+.fi
+.sp
+.nf
+\fBceph\fP \fBauth\fP \fIexport\fP \fI<entity>\fP
+.fi
+.sp
+.nf
+\fBceph\fP \fBconfig\-key\fP \fIget\fP \fI<key>\fP
+.fi
+.sp
+.nf
+\fBceph\fP \fBmds\fP \fIadd_data_pool\fP \fI<pool>\fP
+.fi
+.sp
+.nf
+\fBceph\fP \fBmds\fP \fIgetmap\fP {\fI<int[0\-]>\fP}
+.fi
+.sp
+.nf
+\fBceph\fP \fBmon\fP \fIadd\fP \fI<name>\fP <\fIIPaddr[:port]\fP>
+.fi
+.sp
+.nf
+\fBceph\fP \fBmon_status\fP
+.fi
+.sp
+.nf
+\fBceph\fP \fBosd\fP \fIcreate\fP {\fI<uuid>\fP}
+.fi
+.sp
+.nf
+\fBceph\fP \fBosd\fP \fBcrush\fP \fIadd\fP \fI<osdname (id|osd.id)>\fP
+.fi
+.sp
+.sp
+\fI<float[0.0\-]>\fP \fI<args>\fP [\fI<args>\fP\&...]
+.nf
+\fBceph\fP \fBpg\fP \fIforce_create_pg\fP \fI<pgid>\fP
+.fi
+.sp
+.nf
+\fBceph\fP \fBpg\fP \fIstat\fP
+.fi
+.sp
+.nf
+\fBceph\fP \fBquorum_status\fP
.fi
.sp
.SH DESCRIPTION
.sp
-\fBceph\fP is a control utility for communicating with the monitor
-cluster of a running Ceph distributed storage system.
+\fBceph\fP is a control utility which is used for manual deployment and maintenance
+of a Ceph cluster. It provides a diverse set of commands that allows deployment of
+monitors, OSDs, placement groups, MDS and overall maintenance, administration
+of the cluster.
+.SH COMMANDS
+.SS auth
.sp
-There are three basic modes of operation.
-.SS Interactive mode
+Manage authentication keys. It is used for adding, removing, exporting
+or updating of authentication keys for a particular entity such as a monitor or
+OSD. It uses some additional subcommands.
.sp
-To start in interactive mode, no arguments are necessary. Control\-d or
-\(aqquit\(aq will exit.
-.SS Watch mode
+Subcommand \fBadd\fP adds authentication info for a particular entity from input
+file, or random key if no input given and/or any caps specified in the command.
.sp
-Watch mode shows cluster state changes as they occur. For example:
+Usage:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
-ceph \-w
+ceph auth add <entity> {<caps> [<caps>...]}
.ft P
.fi
.UNINDENT
.UNINDENT
-.SS Command line mode
.sp
-Finally, to send a single instruction to the monitor cluster (and wait
-for a response), the command can be specified on the command line.
-.SH OPTIONS
+Subcommand \fBcaps\fP updates caps for \fBname\fP from caps specified in the command.
+.sp
+Usage:
.INDENT 0.0
-.TP
-.B \-i infile
-will specify an input file to be passed along as a payload with the
-command to the monitor cluster. This is only used for specific
-monitor commands.
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth caps <entity> <caps> [<caps>...]
+.ft P
+.fi
+.UNINDENT
.UNINDENT
+.sp
+Subcommand \fBdel\fP deletes all caps for \fBname\fP\&.
+.sp
+Usage:
.INDENT 0.0
-.TP
-.B \-o outfile
-will write any payload returned by the monitor cluster with its
-reply to outfile. Only specific monitor commands (e.g. osd getmap)
-return a payload.
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth del <entity>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBexport\fP writes keyring for requested entity, or master keyring if
+none given.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth export {<entity>}
+.ft P
+.fi
.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBget\fP writes keyring file with requested key.
+.sp
+Usage:
.INDENT 0.0
-.TP
-.B \-c ceph.conf, \-\-conf=ceph.conf
-Use ceph.conf configuration file instead of the default
-/etc/ceph/ceph.conf to determine monitor addresses during startup.
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth get <entity>
+.ft P
+.fi
.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBget\-key\fP displays requested key.
+.sp
+Usage:
.INDENT 0.0
-.TP
-.B \-m monaddress[:port]
-Connect to specified monitor (instead of looking through ceph.conf).
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth get\-key <entity>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBget\-or\-create\fP adds authentication info for a particular entity
+from input file, or random key if no input given and/or any caps specified in the
+command.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth get\-or\-create <entity> {<caps> [<caps>...]}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBget\-or\-create\-key\fP gets or adds key for \fBname\fP from system/caps
+pairs specified in the command. If key already exists, any given caps must match
+the existing caps for that key.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth get\-or\-create\-key <entity> {<caps> [<caps>...]}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBimport\fP reads keyring from input file.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth import
+.ft P
+.fi
+.UNINDENT
.UNINDENT
-.SH EXAMPLES
.sp
-To grab a copy of the current OSD map:
+Subcommand \fBlist\fP lists authentication state.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth list
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBprint\-key\fP displays requested key.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth print\-key <entity>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBprint_key\fP displays requested key.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph auth print_key <entity>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS compact
+.sp
+Causes compaction of monitor\(aqs leveldb storage.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph compact
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS config\-key
+.sp
+Manage configuration key. It uses some additional subcommands.
+.sp
+Subcommand \fBget\fP gets the configuration key.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph config\-key get <key>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBput\fP puts configuration key and values.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph config\-key put <key> {<val>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBexists\fP checks for configuration keys existence.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph config\-key exists <key>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBlist\fP lists configuration keys.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph config\-key list
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdel\fP deletes configuration key.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph config\-key del <key>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS df
+.sp
+Show cluster\(aqs free space status.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph df
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS fsid
+.sp
+Show cluster\(aqs FSID/UUID.
+.sp
+Usage:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
-ceph \-m 1.2.3.4:6789 osd getmap \-o osdmap
+ceph fsid
.ft P
.fi
.UNINDENT
.UNINDENT
+.SS health
.sp
-To get a dump of placement group (PG) state:
+Show cluster\(aqs health.
+.sp
+Usage:
.INDENT 0.0
.INDENT 3.5
.sp
.nf
.ft C
-ceph pg dump \-o pg.txt
+ceph health
.ft P
.fi
.UNINDENT
.UNINDENT
-.SH MONITOR COMMANDS
+.SS heap
+.sp
+Show heap usage info (available only if compiled with tcmalloc)
.sp
-A more complete summary of commands understood by the monitor cluster can be found in the
-online documentation, at
+Usage:
.INDENT 0.0
.INDENT 3.5
-\fI\%http://ceph.com/docs/master/rados/operations/control\fP
+.sp
+.nf
+.ft C
+ceph heap dump|start_profiler|stop_profiler|release|stats
+.ft P
+.fi
+.UNINDENT
.UNINDENT
+.SS injectargs
+.sp
+Inject configuration arguments into monitor.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph injectargs <injected_args> [<injected_args>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS log
+.sp
+Log supplied text to the monitor log.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph log <logtext> [<logtext>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS mds
+.sp
+Manage metadata server configuration and administration. It uses some
+additional subcommands.
+.sp
+Subcommand \fBadd_data_pool\fP adds data pool.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds add_data_pool <pool>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBcluster_down\fP takes mds cluster down.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds cluster_down
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBcluster_up\fP brings mds cluster up.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds cluster_up
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBcompat\fP manages compatible features. It uses some additional
+subcommands.
+.sp
+Subcommand \fBrm_compat\fP removes compatible feature.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds compat rm_compat <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrm_incompat\fP removes incompatible feature.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds compat rm_incompat <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBshow\fP shows mds compatibility settings.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds compat show
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdeactivate\fP stops mds.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds deactivate <who>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump\fP dumps information, optionally from epoch.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds dump {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBfail\fP forces mds to status fail.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds fail <who>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBgetmap\fP gets MDS map, optionally from epoch.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds getmap {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBnewfs\fP makes new filesystem using pools <metadata> and <data>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds newfs <int[0\-]> <int[0\-]> {\-\-yes\-i\-really\-mean\-it}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBremove_data_pool\fP removes data pool.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds remove_data_pool <pool>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrm\fP removes inactive mds.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds rm <int[0\-]> <name> (type.id)>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrmfailed\fP removes failed mds.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds rmfailed <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset_max_mds\fP sets max MDS index.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds set_max_mds <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset_state\fP sets mds state of <gid> to <numeric\-state>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds set_state <int[0\-]> <int[0\-20]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBsetmap\fP sets mds map; must supply correct epoch number.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds setmap <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBstat\fP shows MDS status.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds stat
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBstop\fP stops mds.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds stop <who>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBtell\fP sends command to particular mds.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mds tell <who> <args> [<args>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS mon
+.sp
+Manage monitor configuration and administration. It uses some additional
+subcommands.
+.sp
+Subcommand \fBadd\fP adds new monitor named <name> at <addr>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mon add <name> <IPaddr[:port]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump\fP dumps formatted monmap (optionally from epoch)
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mon dump {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBgetmap\fP gets monmap.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mon getmap {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBremove\fP removes monitor named <name>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mon remove <name>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBstat\fP summarizes monitor status.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mon stat
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBmon_status\fP reports status of monitors.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph mon_status
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS osd
+.sp
+Manage OSD configuration and administration. It uses some additional
+subcommands.
+.sp
+Subcommand \fBcreate\fP creates new osd (with optional UUID).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd create {<uuid>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBcrush\fP is used for CRUSH management. It uses some additional
+subcommands.
+.sp
+Subcommand \fBadd\fP adds or updates crushmap position and weight for <name> with
+<weight> and location <args>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush add <osdname (id|osd.id)> <float[0.0\-]> <args> [<args>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBadd\-bucket\fP adds no\-parent (probably root) crush bucket <name> of
+type <type>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush add\-bucket <name> <type>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBcreate\-or\-move\fP creates entry or moves existing entry for <name>
+<weight> at/to location <args>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush create\-or\-move <osdname (id|osd.id)> <float[0.0\-]> <args>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+[<args>...]
+.sp
+Subcommand \fBdump\fP dumps crush map.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush dump
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBlink\fP links existing entry for <name> under location <args>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush link <name> <args> [<args>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBmove\fP moves existing entry for <name> to location <args>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush move <name> <args> [<args>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBremove\fP removes <name> from crush map (everywhere, or just at
+<ancestor>).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush remove <name> {<ancestor>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBreweight\fP change <name>\(aqs weight to <weight> in crush map.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush reweight <name> <float[0.0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrm\fP removes <name> from crush map (everywhere, or just at
+<ancestor>).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush rm <name> {<ancestor>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrule\fP is used for creating crush rules. It uses some additional
+subcommands.
+.sp
+Subcommand \fBcreate\-erasure\fP creates crush rule <name> for erasure coded pool
+created with <profile> (default default).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush rule create\-erasure <name> {<profile>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBcreate\-simple\fP creates crush rule <name> to start from <root>,
+replicate across buckets of type <type>, using a choose mode of <firstn|indep>
+(default firstn; indep best for erasure pools).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush rule create\-simple <name> <root> <type> {firstn|indep}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump\fP dumps crush rule <name> (default all).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush rule dump {<name>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBlist\fP lists crush rules.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush rule list
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBls\fP lists crush rules.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush rule ls
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrm\fP removes crush rule <name>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush rule rm <name>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset\fP sets crush map from input file.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush set
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset\fP with osdname/osd.id update crushmap position and weight
+for <name> to <weight> with location <args>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush set <osdname (id|osd.id)> <float[0.0\-]> <args> [<args>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBshow\-tunables\fP shows current crush tunables.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush show\-tunables
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBtunables\fP sets crush tunables values to <profile>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush tunables legacy|argonaut|bobtail|firefly|optimal|default
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBunlink\fP unlinks <name> from crush map (everywhere, or just at
+<ancestor>).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd crush unlink <name> {<ancestor>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdeep\-scrub\fP initiates deep scrub on specified osd.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd deep\-scrub <who>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdown\fP sets osd(s) <id> [<id>...] down.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd down <ids> [<ids>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump\fP prints summary of OSD map.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd dump {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBerasure\-code\-profile\fP is used for managing the erasure code
+profiles. It uses some additional subcommands.
+.sp
+Subcommand \fBget\fP gets erasure code profile <name>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd erasure\-code\-profile get <name>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBls\fP lists all erasure code profiles.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd erasure\-code\-profile ls
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrm\fP removes erasure code profile <name>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd erasure\-code\-profile rm <name>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset\fP creates erasure code profile <name> with [<key[=value]> ...]
+pairs. Add a \-\-force at the end to override an existing profile (IT IS RISKY).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd erasure\-code\-profile set <name> {<profile> [<profile>...]}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBfind\fP find osd <id> in the CRUSH map and shows its location.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd find <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBgetcrushmap\fP gets CRUSH map.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd getcrushmap {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBgetmap\fP gets OSD map.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd getmap {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBgetmaxosd\fP shows largest OSD id.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd getmaxosd
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBin\fP sets osd(s) <id> [<id>...] in.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd in <ids> [<ids>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBlost\fP marks osd as permanently lost. THIS DESTROYS DATA IF NO
+MORE REPLICAS EXIST, BE CAREFUL.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd lost <int[0\-]> {\-\-yes\-i\-really\-mean\-it}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBls\fP shows all OSD ids.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd ls {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBlspools\fP lists pools.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd lspools {<int>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBmap\fP finds pg for <object> in <pool>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd map <poolname> <objectname>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBmetadata\fP fetches metadata for osd <id>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd metadata <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBout\fP sets osd(s) <id> [<id>...] out.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd out <ids> [<ids>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBpause\fP pauses osd.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pause
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBperf\fP prints dump of OSD perf summary stats.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd perf
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBpg\-temp\fP set pg_temp mapping pgid:[<id> [<id>...]] (developers
+only).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pg\-temp <pgid> {<id> [<id>...]}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBpool\fP is used for managing data pools. It uses some additional
+subcommands.
+.sp
+Subcommand \fBcreate\fP creates pool.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool create <poolname> <int[0\-]> {<int[0\-]>} {replicated|erasure}
+{<erasure_code_profile>} {<ruleset>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdelete\fP deletes pool.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool delete <poolname> {<poolname>} {\-\-yes\-i\-really\-really\-mean\-it}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBget\fP gets pool parameter <var>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool get <poolname> size|min_size|crash_replay_interval|pg_num|
+pgp_num|crush_ruleset|hit_set_type|hit_set_period|hit_set_count|hit_set_fpp|
+
+ceph osd pool get <poolname> auid|target_max_objects|target_max_bytes
+
+ceph osd pool get <poolname> cache_target_dirty_ratio|cache_target_full_ratio
+
+ceph osd pool get <poolname> cache_min_flush_age|cache_min_evict_age|
+erasure_code_profile
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBget\-quota\fP obtains object or byte limits for pool.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool get\-quota <poolname>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBmksnap\fP makes snapshot <snap> in <pool>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool mksnap <poolname> <snap>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrename\fP renames <srcpool> to <destpool>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool rename <poolname> <poolname>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrmsnap\fP removes snapshot <snap> from <pool>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool rmsnap <poolname> <snap>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset\fP sets pool parameter <var> to <val>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool set <poolname> size|min_size|crash_replay_interval|pg_num|
+pgp_num|crush_ruleset|hashpspool|hit_set_type|hit_set_period|
+
+ceph osd pool set <poolname> hit_set_count|hit_set_fpp|debug_fake_ec_pool
+
+ceph osd pool set <poolname> target_max_bytes|target_max_objects
+
+ceph osd pool set <poolname> cache_target_dirty_ratio|cache_target_full_ratio
+
+ceph osd pool set <poolname> cache_min_flush_age
+
+ceph osd pool set <poolname> cache_min_evict_age|auid <val>
+{\-\-yes\-i\-really\-mean\-it}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset\-quota\fP sets object or byte limit on pool.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool set\-quota <poolname> max_objects|max_bytes <val>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBstats\fP obtain stats from all pools, or from specified pool.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd pool stats {<name>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBprimary\-affinity\fP adjust osd primary\-affinity from 0.0 <=<weight>
+<= 1.0
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd primary\-affinity <osdname (id|osd.id)> <float[0.0\-1.0]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBprimary\-temp\fP sets primary_temp mapping pgid:<id>|\-1 (developers
+only).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd primary\-temp <pgid> <id>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrepair\fP initiates repair on a specified osd.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd repair <who>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBreweight\fP reweights osd to 0.0 < <weight> < 1.0.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+osd reweight <int[0\-]> <float[0.0\-1.0]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBreweight\-by\-utilization\fP reweight OSDs by utilization
+[overload\-percentage\-for\-consideration, default 120].
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd reweight\-by\-utilization {<int[100\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrm\fP removes osd(s) <id> [<id>...] in the cluster.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd rm <ids> [<ids>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBscrub\fP initiates scrub on specified osd.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd scrub <who>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset\fP sets <key>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd set pause|noup|nodown|noout|noin|nobackfill|norecover|noscrub|
+nodeep\-scrub|notieragent
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBsetcrushmap\fP sets crush map from input file.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd setcrushmap
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBsetmaxosd\fP sets new maximum osd value.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd setmaxosd <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBstat\fP prints summary of OSD map.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd stat
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBthrash\fP thrashes OSDs for <num_epochs>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd thrash <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBtier\fP is used for managing tiers. It uses some additional
+subcommands.
+.sp
+Subcommand \fBadd\fP adds the tier <tierpool> (the second one) to base pool <pool>
+(the first one).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd tier add <poolname> <poolname> {\-\-force\-nonempty}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBadd\-cache\fP adds a cache <tierpool> (the second one) of size <size>
+to existing pool <pool> (the first one).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd tier add\-cache <poolname> <poolname> <int[0\-]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBcache\-mode\fP specifies the caching mode for cache tier <pool>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd tier cache\-mode <poolname> none|writeback|forward|readonly
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBremove\fP removes the tier <tierpool> (the second one) from base pool
+<pool> (the first one).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd tier remove <poolname> <poolname>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBremove\-overlay\fP removes the overlay pool for base pool <pool>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd tier remove\-overlay <poolname>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset\-overlay\fP set the overlay pool for base pool <pool> to be
+<overlaypool>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd tier set\-overlay <poolname> <poolname>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBtree\fP prints OSD tree.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd tree {<int[0\-]>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBunpause\fP unpauses osd.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph osd unpause
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBunset\fP unsets <key>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+osd unset pause|noup|nodown|noout|noin|nobackfill|norecover|noscrub|
+nodeep\-scrub|notieragent
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS pg
+.sp
+It is used for managing the placement groups in OSDs. It uses some
+additional subcommands.
+.sp
+Subcommand \fBdebug\fP shows debug info about pgs.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg debug unfound_objects_exist|degraded_pgs_exist
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdeep\-scrub\fP starts deep\-scrub on <pgid>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg deep\-scrub <pgid>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump\fP shows human\-readable versions of pg map (only \(aqall\(aq valid
+with plain).
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg dump {all|summary|sum|delta|pools|osds|pgs|pgs_brief}
+
+ceph pg dump {all|summary|sum|delta|pools|osds|pgs|pgs_brief...}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump_json\fP shows human\-readable version of pg map in json only.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg dump_json {all|summary|sum|pools|osds|pgs[all|summary|sum|pools|
+osds|pgs...]}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump_pools_json\fP shows pg pools info in json only.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg dump_pools_json
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBdump_stuck\fP shows information about stuck pgs.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg dump_stuck {inactive|unclean|stale[inactive|unclean|stale...]}
+{<int>}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBforce_create_pg\fP forces creation of pg <pgid>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg force_create_pg <pgid>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBgetmap\fP gets binary pg map to \-o/stdout.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg getmap
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBmap\fP shows mapping of pg to osds.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg map <pgid>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBrepair\fP starts repair on <pgid>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg repair <pgid>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBscrub\fP starts scrub on <pgid>.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg scrub <pgid>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBsend_pg_creates\fP triggers pg creates to be issued.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg send_pg_creates
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset_full_ratio\fP sets ratio at which pgs are considered full.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg set_full_ratio <float[0.0\-1.0]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBset_nearfull_ratio\fP sets ratio at which pgs are considered nearly
+full.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg set_nearfull_ratio <float[0.0\-1.0]>
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.sp
+Subcommand \fBstat\fP shows placement group status.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph pg stat
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS quorum
+.sp
+Enter or exit quorum.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph quorum enter|exit
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS quorum_status
+.sp
+Reports status of monitor quorum.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph quorum_status
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS report
+.sp
+Reports full status of cluster, optional title tag strings.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph report {<tags> [<tags>...]}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS scrub
+.sp
+Scrubs the monitor stores.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph scrub
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS status
+.sp
+Shows cluster status.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph status
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS sync force
+.sp
+Forces sync of and clear monitor store.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph sync force {\-\-yes\-i\-really\-mean\-it} {\-\-i\-know\-what\-i\-am\-doing}
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SS tell
+.sp
+Sends a command to a specific daemon.
+.sp
+Usage:
+.INDENT 0.0
+.INDENT 3.5
+.sp
+.nf
+.ft C
+ceph tell <name (type.id)> <args> [<args>...]
+.ft P
+.fi
+.UNINDENT
+.UNINDENT
+.SH OPTIONS
+.INDENT 0.0
+.TP
+.B \-i infile
+will specify an input file to be passed along as a payload with the
+command to the monitor cluster. This is only used for specific
+monitor commands.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-o outfile
+will write any payload returned by the monitor cluster with its
+reply to outfile. Only specific monitor commands (e.g. osd getmap)
+return a payload.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-c ceph.conf, \-\-conf=ceph.conf
+Use ceph.conf configuration file instead of the default
+\fB/etc/ceph/ceph.conf\fP to determine monitor addresses during startup.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-id CLIENT_ID, \-\-user CLIENT_ID
+Client id for authentication.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-name CLIENT_NAME, \-n CLIENT_NAME
+Client name for authentication.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-cluster CLUSTER
+Name of the Ceph cluster.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-admin\-daemon ADMIN_SOCKET
+Submit admin\-socket commands.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-admin\-socket ADMIN_SOCKET_NOPE
+You probably mean \-\-admin\-daemon
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-s, \-\-status
+Show cluster status.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-w, \-\-watch
+Watch live cluster changes.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-watch\-debug
+Watch debug events.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-watch\-info
+Watch info events.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-watch\-sec
+Watch security events.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-watch\-warn
+Watch warning events.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-watch\-error
+Watch error events.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-version, \-v
+Display version.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-verbose
+Make verbose.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-concise
+Make less verbose.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-f {json,json\-pretty,xml,xml\-pretty,plain}, \-\-format
+Format of output.
+.UNINDENT
+.INDENT 0.0
+.TP
+.B \-\-connect\-timeout CLUSTER_TIMEOUT
+Set a timeout for connecting to the cluster.
.UNINDENT
.SH AVAILABILITY
.sp
-\fBceph\fP is part of the Ceph distributed storage system. Please refer to the Ceph documentation at
-\fI\%http://ceph.com/docs\fP for more information.
+\fBceph\fP is a part of the Ceph distributed storage system. Please refer to
+the Ceph documentation at \fI\%http://ceph.com/docs\fP for more information.
.SH SEE ALSO
.sp
-\fBceph\fP(8),
+\fBceph\-mon\fP(8),
+\fBceph\-osd\fP(8),
+\fBceph\-mds\fP(8)
.SH COPYRIGHT
2010-2014, Inktank Storage, Inc. and contributors. Licensed under Creative Commons BY-SA
.\" Generated by docutils manpage writer.
diff -Nru ceph-0.80.7/src/ceph-disk ceph-0.80.9/src/ceph-disk
--- ceph-0.80.7/src/ceph-disk 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/ceph-disk 2015-03-09 20:51:17.000000000 +0100
@@ -89,7 +89,7 @@
# issues with ext4 before the xatts-in-leveldb work, and it seemed
# that user_xattr helped
ext4='noatime,user_xattr',
- xfs='noatime',
+ xfs='noatime,inode64',
)
MKFS_ARGS = dict(
@@ -791,11 +791,13 @@
# make a new key
try:
if not os.path.exists(key_dir):
- os.makedirs(key_dir)
+ os.makedirs(key_dir, stat.S_IRUSR|stat.S_IWUSR|stat.S_IXUSR)
with file('/dev/urandom', 'rb') as i:
key = i.read(256)
- with file(path, 'wb') as key_file:
- key_file.write(key)
+ fd = os.open(path, os.O_WRONLY|os.O_CREAT,
+ stat.S_IRUSR|stat.S_IWUSR)
+ assert os.write(fd, key) == len(key)
+ os.close(fd)
return path
except:
raise Error('unable to read or create dm-crypt key', path)
@@ -968,6 +970,35 @@
return num
+def update_partition(action, dev, description):
+ # try to make sure the kernel refreshes the table. note
+ # that if this gets ebusy, we are probably racing with
+ # udev because it already updated it.. ignore failure here.
+
+ # On RHEL and CentOS distros, calling partprobe forces a reboot of the
+ # server. Since we are not resizing partitons so we rely on calling
+ # partx
+ if platform_distro().startswith(('centos', 'red', 'scientific')):
+ LOG.info('calling partx on %s device %s', description, dev)
+ LOG.info('re-reading known partitions will display errors')
+ command(
+ [
+ 'partx',
+ action,
+ dev,
+ ],
+ )
+
+ else:
+ LOG.debug('Calling partprobe on %s device %s', description, dev)
+ command(
+ [
+ 'partprobe',
+ dev,
+ ],
+ )
+
+
def zap(dev):
"""
Destroy the partition table and content of a given disk.
@@ -993,6 +1024,9 @@
dev,
],
)
+
+ update_partition('-d', dev, 'zapped')
+
except subprocess.CalledProcessError as e:
raise Error(e)
@@ -1068,32 +1102,7 @@
],
)
- # try to make sure the kernel refreshes the table. note
- # that if this gets ebusy, we are probably racing with
- # udev because it already updated it.. ignore failure here.
-
- # On RHEL and CentOS distros, calling partprobe forces a reboot of the
- # server. Since we are not resizing partitons so we rely on calling
- # partx
- if platform_distro().startswith(('centos', 'red')):
- LOG.info('calling partx on prepared device %s', journal)
- LOG.info('re-reading known partitions will display errors')
- command(
- [
- 'partx',
- '-a',
- journal,
- ],
- )
-
- else:
- LOG.debug('Calling partprobe on prepared device %s', journal)
- command(
- [
- 'partprobe',
- journal,
- ],
- )
+ update_partition('-a', journal, 'prepared')
# wait for udev event queue to clear
command(
@@ -1118,7 +1127,6 @@
except subprocess.CalledProcessError as e:
raise Error(e)
-
def prepare_journal_file(
journal):
@@ -1279,12 +1287,7 @@
data,
],
)
- command(
- [
- 'partprobe',
- data,
- ],
- )
+ update_partition('-a', data, 'created')
command(
[
# wait for udev event queue to clear
@@ -1500,33 +1503,7 @@
prepare_lock.release() # noqa
if stat.S_ISBLK(dmode):
- # try to make sure the kernel refreshes the table. note
- # that if this gets ebusy, we are probably racing with
- # udev because it already updated it.. ignore failure here.
-
- # On RHEL and CentOS distros, calling partprobe forces a reboot of
- # the server. Since we are not resizing partitons so we rely on
- # calling partx
- if platform_distro().startswith(('centos', 'red')):
- LOG.info('calling partx on prepared device %s', args.data)
- LOG.info('re-reading known partitions will display errors')
-
- command(
- [
- 'partx',
- '-a',
- args.data,
- ],
- )
-
- else:
- LOG.debug('Calling partprobe on prepared device %s', args.data)
- command(
- [
- 'partprobe',
- args.data,
- ],
- )
+ update_partition('-a', args.data, 'prepared')
except Error as e:
if journal_dm_keypath:
@@ -1918,7 +1895,8 @@
raise Error('No OSD uuid assigned.')
LOG.debug('OSD uuid is %s', fsid)
- keyring = activate_key_template.format(cluster=cluster)
+ keyring = activate_key_template.format(cluster=cluster,
+ statedir=STATEDIR)
osd_id = get_osd_id(path)
if osd_id is None:
@@ -2657,7 +2635,7 @@
help='path to block device or directory',
)
activate_parser.set_defaults(
- activate_key_template=STATEDIR + '/bootstrap-osd/{cluster}.keyring',
+ activate_key_template='{statedir}/bootstrap-osd/{cluster}.keyring',
func=main_activate,
)
@@ -2681,7 +2659,7 @@
choices=INIT_SYSTEMS,
)
activate_journal_parser.set_defaults(
- activate_key_template=STATEDIR + '/bootstrap-osd/{cluster}.keyring',
+ activate_key_template='{statedir}/bootstrap-osd/{cluster}.keyring',
func=main_activate_journal,
)
@@ -2700,7 +2678,7 @@
choices=INIT_SYSTEMS,
)
activate_all_parser.set_defaults(
- activate_key_template=STATEDIR + '/bootstrap-osd/{cluster}.keyring',
+ activate_key_template='{statedir}/bootstrap-osd/{cluster}.keyring',
func=main_activate_all,
)
diff -Nru ceph-0.80.7/src/ceph.in ceph-0.80.9/src/ceph.in
--- ceph-0.80.7/src/ceph.in 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/ceph.in 2015-03-09 20:51:17.000000000 +0100
@@ -841,4 +841,8 @@
return 0
if __name__ == '__main__':
- sys.exit(main())
+ retval = main()
+ # shutdown explicitly; Rados() does not
+ if cluster_handle:
+ cluster_handle.shutdown()
+ sys.exit(retval)
diff -Nru ceph-0.80.7/src/ceph_mon.cc ceph-0.80.9/src/ceph_mon.cc
--- ceph-0.80.7/src/ceph_mon.cc 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/ceph_mon.cc 2015-03-09 20:51:17.000000000 +0100
@@ -419,6 +419,26 @@
return 0;
}
+ {
+ // check fs stats. don't start if it's critically close to full.
+ ceph_data_stats_t stats;
+ int err = get_fs_stats(stats, g_conf->mon_data.c_str());
+ if (err < 0) {
+ cerr << "error checking monitor data's fs stats: " << cpp_strerror(err)
+ << std::endl;
+ exit(-err);
+ }
+ if (stats.avail_percent <= g_conf->mon_data_avail_crit) {
+ cerr << "error: monitor data filesystem reached concerning levels of"
+ << " available storage space (available: "
+ << stats.avail_percent << "% " << prettybyte_t(stats.byte_avail)
+ << ")\nyou may adjust 'mon data avail crit' to a lower value"
+ << " to make this go away (default: " << g_conf->mon_data_avail_crit
+ << "%)\n" << std::endl;
+ exit(ENOSPC);
+ }
+ }
+
// we fork early to prevent leveldb's environment static state from
// screwing us over
Preforker prefork;
diff -Nru ceph-0.80.7/src/civetweb/civetweb.h ceph-0.80.9/src/civetweb/civetweb.h
--- ceph-0.80.7/src/civetweb/civetweb.h 2014-10-14 21:14:06.000000000 +0200
+++ ceph-0.80.9/src/civetweb/civetweb.h 2015-03-09 20:52:38.000000000 +0100
@@ -24,7 +24,7 @@
#define CIVETWEB_HEADER_INCLUDED
#ifndef CIVETWEB_VERSION
-#define CIVETWEB_VERSION "1.6"
+#define CIVETWEB_VERSION "1.7"
#endif
#ifndef CIVETWEB_API
@@ -77,7 +77,7 @@
/* This structure needs to be passed to mg_start(), to let civetweb know
which callbacks to invoke. For detailed description, see
- https://github.com/sunsetbrew/civetweb/blob/master/docs/UserManual.md */
+ https://github.com/bel2125/civetweb/blob/master/docs/UserManual.md */
struct mg_callbacks {
/* Called when civetweb has received new HTTP request.
If callback returns non-zero,
@@ -94,6 +94,10 @@
non-zero, civetweb does not log anything. */
int (*log_message)(const struct mg_connection *, const char *message);
+ /* Called when civetweb is about to log access. If callback returns
+ non-zero, civetweb does not log anything. */
+ int (*log_access)(const struct mg_connection *, const char *message);
+
/* Called when civetweb initializes SSL library. */
int (*init_ssl)(void *ssl_context, void *user_data);
@@ -176,7 +180,7 @@
};
struct mg_context *ctx = mg_start(&my_func, NULL, options);
- Refer to https://github.com/sunsetbrew/civetweb/blob/master/docs/UserManual.md
+ Refer to https://github.com/bel2125/civetweb/blob/master/docs/UserManual.md
for the list of valid option and their possible values.
Return:
@@ -330,8 +334,18 @@
Invoke this before mg_write or mg_printf when communicating with a
websocket if your code has server-initiated communication as well as
communication in direct response to a message. */
-CIVETWEB_API void mg_lock(struct mg_connection* conn);
-CIVETWEB_API void mg_unlock(struct mg_connection* conn);
+CIVETWEB_API void mg_lock_connection(struct mg_connection* conn);
+CIVETWEB_API void mg_unlock_connection(struct mg_connection* conn);
+
+#if defined(MG_LEGACY_INTERFACE)
+#define mg_lock mg_lock_connection
+#define mg_unlock mg_unlock_connection
+#endif
+
+/* Lock server context. This lock may be used to protect ressources
+ that are shared between different connection/worker threads. */
+CIVETWEB_API void mg_lock_context(struct mg_context* ctx);
+CIVETWEB_API void mg_unlock_context(struct mg_context* ctx);
/* Opcodes, from http://tools.ietf.org/html/rfc6455 */
diff -Nru ceph-0.80.7/src/civetweb/include/civetweb_conf.h ceph-0.80.9/src/civetweb/include/civetweb_conf.h
--- ceph-0.80.7/src/civetweb/include/civetweb_conf.h 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/civetweb/include/civetweb_conf.h 2015-03-09 20:52:38.000000000 +0100
@@ -0,0 +1,6 @@
+#ifndef CIVETWEB_CONF_H
+#define CIVETWEB_CONF_H
+
+#define USE_IPV6 1
+
+#endif
diff -Nru ceph-0.80.7/src/civetweb/include/civetweb.h ceph-0.80.9/src/civetweb/include/civetweb.h
--- ceph-0.80.7/src/civetweb/include/civetweb.h 2014-10-14 21:14:06.000000000 +0200
+++ ceph-0.80.9/src/civetweb/include/civetweb.h 2015-03-09 20:52:38.000000000 +0100
@@ -24,7 +24,7 @@
#define CIVETWEB_HEADER_INCLUDED
#ifndef CIVETWEB_VERSION
-#define CIVETWEB_VERSION "1.6"
+#define CIVETWEB_VERSION "1.7"
#endif
#ifndef CIVETWEB_API
@@ -77,7 +77,7 @@
/* This structure needs to be passed to mg_start(), to let civetweb know
which callbacks to invoke. For detailed description, see
- https://github.com/sunsetbrew/civetweb/blob/master/docs/UserManual.md */
+ https://github.com/bel2125/civetweb/blob/master/docs/UserManual.md */
struct mg_callbacks {
/* Called when civetweb has received new HTTP request.
If callback returns non-zero,
@@ -94,6 +94,10 @@
non-zero, civetweb does not log anything. */
int (*log_message)(const struct mg_connection *, const char *message);
+ /* Called when civetweb is about to log access. If callback returns
+ non-zero, civetweb does not log anything. */
+ int (*log_access)(const struct mg_connection *, const char *message);
+
/* Called when civetweb initializes SSL library. */
int (*init_ssl)(void *ssl_context, void *user_data);
@@ -176,7 +180,7 @@
};
struct mg_context *ctx = mg_start(&my_func, NULL, options);
- Refer to https://github.com/sunsetbrew/civetweb/blob/master/docs/UserManual.md
+ Refer to https://github.com/bel2125/civetweb/blob/master/docs/UserManual.md
for the list of valid option and their possible values.
Return:
@@ -330,8 +334,18 @@
Invoke this before mg_write or mg_printf when communicating with a
websocket if your code has server-initiated communication as well as
communication in direct response to a message. */
-CIVETWEB_API void mg_lock(struct mg_connection* conn);
-CIVETWEB_API void mg_unlock(struct mg_connection* conn);
+CIVETWEB_API void mg_lock_connection(struct mg_connection* conn);
+CIVETWEB_API void mg_unlock_connection(struct mg_connection* conn);
+
+#if defined(MG_LEGACY_INTERFACE)
+#define mg_lock mg_lock_connection
+#define mg_unlock mg_unlock_connection
+#endif
+
+/* Lock server context. This lock may be used to protect ressources
+ that are shared between different connection/worker threads. */
+CIVETWEB_API void mg_lock_context(struct mg_context* ctx);
+CIVETWEB_API void mg_unlock_context(struct mg_context* ctx);
/* Opcodes, from http://tools.ietf.org/html/rfc6455 */
diff -Nru ceph-0.80.7/src/civetweb/src/civetweb.c ceph-0.80.9/src/civetweb/src/civetweb.c
--- ceph-0.80.7/src/civetweb/src/civetweb.c 2014-10-14 21:14:06.000000000 +0200
+++ ceph-0.80.9/src/civetweb/src/civetweb.c 2015-03-09 20:52:38.000000000 +0100
@@ -20,6 +20,8 @@
* THE SOFTWARE.
*/
+#define RGW 1
+
#if defined(_WIN32)
#if !defined(_CRT_SECURE_NO_WARNINGS)
#define _CRT_SECURE_NO_WARNINGS /* Disable deprecation warning in VS2005 */
@@ -181,6 +183,10 @@
#define sleep(x) Sleep((x) * 1000)
#define rmdir(x) _rmdir(x)
+#if defined(USE_LUA) && defined(USE_WEBSOCKET)
+#define USE_TIMERS
+#endif
+
#if !defined(va_copy)
#define va_copy(x, y) x = y
#endif /* !va_copy MINGW #defines va_copy */
@@ -309,6 +315,40 @@
#endif /* End of Windows and UNIX specific includes */
+#ifdef _WIN32
+static CRITICAL_SECTION global_log_file_lock;
+static DWORD pthread_self(void)
+{
+ return GetCurrentThreadId();
+}
+
+int pthread_key_create(pthread_key_t *key, void (*_must_be_zero)(void*) /* destructor function not supported for windows */)
+{
+ assert(_must_be_zero == NULL);
+ if ((key!=0) && (_must_be_zero == NULL)) {
+ *key = TlsAlloc();
+ return (*key != TLS_OUT_OF_INDEXES) ? 0 : -1;
+ }
+ return -2;
+}
+
+int pthread_key_delete(pthread_key_t key)
+{
+ return TlsFree(key) ? 0 : 1;
+}
+
+int pthread_setspecific(pthread_key_t key, void * value)
+{
+ return TlsSetValue(key, value) ? 0 : 1;
+}
+
+void *pthread_getspecific(pthread_key_t key)
+{
+ return TlsGetValue(key);
+}
+#endif /* _WIN32 */
+
+
#include "civetweb.h"
#define PASSWORDS_FILE_NAME ".htpasswd"
@@ -320,23 +360,30 @@
#endif
#define ARRAY_SIZE(array) (sizeof(array) / sizeof(array[0]))
-#ifdef DEBUG_TRACE
-#undef DEBUG_TRACE
-#define DEBUG_TRACE(x)
-#else
+#if !defined(DEBUG_TRACE)
#if defined(DEBUG)
-#define DEBUG_TRACE(x) do { \
- flockfile(stdout); \
- printf("*** %lu.%p.%s.%d: ", \
- (unsigned long) time(NULL), (void *) pthread_self(), \
- __func__, __LINE__); \
- printf x; \
- putchar('\n'); \
- fflush(stdout); \
- funlockfile(stdout); \
-} while (0)
+
+static void DEBUG_TRACE_FUNC(const char *func, unsigned line, PRINTF_FORMAT_STRING(const char *fmt), ...) PRINTF_ARGS(3, 4);
+
+static void DEBUG_TRACE_FUNC(const char *func, unsigned line, const char *fmt, ...) {
+
+ va_list args;
+ flockfile(stdout);
+ printf("*** %lu.%p.%s.%u: ",
+ (unsigned long) time(NULL), (void *) pthread_self(),
+ func, line);
+ va_start(args, fmt);
+ vprintf(fmt, args);
+ va_end(args);
+ putchar('\n');
+ fflush(stdout);
+ funlockfile(stdout);
+}
+
+#define DEBUG_TRACE(fmt, ...) DEBUG_TRACE_FUNC(__func__, __LINE__, fmt, __VA_ARGS__)
+
#else
-#define DEBUG_TRACE(x)
+#define DEBUG_TRACE(fmt, ...)
#endif /* DEBUG */
#endif /* DEBUG_TRACE */
@@ -357,7 +404,7 @@
memory = (void *)(((char*)data)+sizeof(size_t));
}
- sprintf(mallocStr, "MEM: %p %5u alloc %7u %4u --- %s:%u\n", memory, size, totalMemUsed, blockCount, file, line);
+ sprintf(mallocStr, "MEM: %p %5lu alloc %7lu %4lu --- %s:%u\n", memory, (unsigned long)size, totalMemUsed, blockCount, file, line);
#if defined(_WIN32)
OutputDebugStringA(mallocStr);
#else
@@ -385,7 +432,7 @@
size = *(size_t*)data;
totalMemUsed -= size;
blockCount--;
- sprintf(mallocStr, "MEM: %p %5u free %7u %4u --- %s:%u\n", memory, size, totalMemUsed, blockCount, file, line);
+ sprintf(mallocStr, "MEM: %p %5lu free %7lu %4lu --- %s:%u\n", memory, (unsigned long)size, totalMemUsed, blockCount, file, line);
#if defined(_WIN32)
OutputDebugStringA(mallocStr);
#else
@@ -400,23 +447,25 @@
char mallocStr[256];
void * data;
+ void * _realloc;
size_t oldsize;
if (newsize) {
if (memory) {
data = (void *)(((char*)memory)-sizeof(size_t));
oldsize = *(size_t*)data;
- data = realloc(data, newsize+sizeof(size_t));
- if (data) {
+ _realloc = realloc(data, newsize+sizeof(size_t));
+ if (_realloc) {
+ data = _realloc;
totalMemUsed -= oldsize;
- sprintf(mallocStr, "MEM: %p %5u r-free %7u %4u --- %s:%u\n", memory, oldsize, totalMemUsed, blockCount, file, line);
+ sprintf(mallocStr, "MEM: %p %5lu r-free %7lu %4lu --- %s:%u\n", memory, (unsigned long)oldsize, totalMemUsed, blockCount, file, line);
#if defined(_WIN32)
OutputDebugStringA(mallocStr);
#else
DEBUG_TRACE("%s", mallocStr);
#endif
totalMemUsed += newsize;
- sprintf(mallocStr, "MEM: %p %5u r-alloc %7u %4u --- %s:%u\n", memory, newsize, totalMemUsed, blockCount, file, line);
+ sprintf(mallocStr, "MEM: %p %5lu r-alloc %7lu %4lu --- %s:%u\n", memory, (unsigned long)newsize, totalMemUsed, blockCount, file, line);
#if defined(_WIN32)
OutputDebugStringA(mallocStr);
#else
@@ -430,6 +479,7 @@
#else
DEBUG_TRACE("MEM: realloc failed\n");
#endif
+ return _realloc;
}
} else {
data = mg_malloc_ex(newsize, file, line);
@@ -454,43 +504,24 @@
static __inline void mg_free(void * a) {free(a);}
#endif
+/* This following lines are just meant as a reminder to use the mg-functions for memory management */
+#ifdef malloc
+ #undef malloc
+#endif
+#ifdef calloc
+ #undef calloc
+#endif
+#ifdef realloc
+ #undef realloc
+#endif
+#ifdef free
+ #undef free
+#endif
#define malloc DO_NOT_USE_THIS_FUNCTION__USE_mg_malloc
#define calloc DO_NOT_USE_THIS_FUNCTION__USE_mg_calloc
#define realloc DO_NOT_USE_THIS_FUNCTION__USE_mg_realloc
#define free DO_NOT_USE_THIS_FUNCTION__USE_mg_free
-#ifdef _WIN32
-static CRITICAL_SECTION global_log_file_lock;
-static DWORD pthread_self(void)
-{
- return GetCurrentThreadId();
-}
-
-int pthread_key_create(pthread_key_t *key, void (*_must_be_zero)(void*) /* destructor function not supported for windows */)
-{
- assert(_must_be_zero == NULL);
- if ((key!=0) && (_must_be_zero == NULL)) {
- *key = TlsAlloc();
- return (*key != TLS_OUT_OF_INDEXES) ? 0 : -1;
- }
- return -2;
-}
-
-int pthread_key_delete(pthread_key_t key)
-{
- return TlsFree(key) ? 0 : 1;
-}
-
-int pthread_setspecific(pthread_key_t key, void * value)
-{
- return TlsSetValue(key, value) ? 0 : 1;
-}
-
-void *pthread_getspecific(pthread_key_t key)
-{
- return TlsGetValue(key);
-}
-#endif /* _WIN32 */
#define MD5_STATIC static
#include "md5.h"
@@ -668,6 +699,7 @@
GLOBAL_PASSWORDS_FILE, INDEX_FILES, ENABLE_KEEP_ALIVE, ACCESS_CONTROL_LIST,
EXTRA_MIME_TYPES, LISTENING_PORTS, DOCUMENT_ROOT, SSL_CERTIFICATE,
NUM_THREADS, RUN_AS_USER, REWRITE, HIDE_FILES, REQUEST_TIMEOUT,
+ DECODE_URL,
#if defined(USE_LUA)
LUA_PRELOAD_FILE, LUA_SCRIPT_EXTENSIONS, LUA_SERVER_PAGE_EXTENSIONS,
@@ -714,6 +746,7 @@
{"url_rewrite_patterns", 12345, NULL},
{"hide_files_patterns", 12345, NULL},
{"request_timeout_ms", CONFIG_TYPE_NUMBER, "30000"},
+ {"decode_url", CONFIG_TYPE_BOOLEAN, "yes"},
#if defined(USE_LUA)
{"lua_preload_file", CONFIG_TYPE_FILE, NULL},
@@ -753,60 +786,64 @@
in_port_t *listening_ports;
int num_listening_sockets;
- volatile int num_threads; /* Number of threads */
- pthread_mutex_t mutex; /* Protects (max|num)_threads */
- pthread_cond_t cond; /* Condvar for tracking workers terminations */
+ volatile int num_threads; /* Number of threads */
+ pthread_mutex_t thread_mutex; /* Protects (max|num)_threads */
+ pthread_cond_t thread_cond; /* Condvar for tracking workers terminations */
struct socket queue[MGSQLEN]; /* Accepted sockets */
- volatile int sq_head; /* Head of the socket queue */
- volatile int sq_tail; /* Tail of the socket queue */
- pthread_cond_t sq_full; /* Signaled when socket is produced */
- pthread_cond_t sq_empty; /* Signaled when socket is consumed */
- pthread_t masterthreadid; /* The master thread ID. */
- int workerthreadcount; /* The amount of worker threads. */
- pthread_t *workerthreadids;/* The worker thread IDs. */
-
- unsigned long start_time; /* Server start time, used for authentication */
- unsigned long nonce_count; /* Used nonces, used for authentication */
+ volatile int sq_head; /* Head of the socket queue */
+ volatile int sq_tail; /* Tail of the socket queue */
+ pthread_cond_t sq_full; /* Signaled when socket is produced */
+ pthread_cond_t sq_empty; /* Signaled when socket is consumed */
+ pthread_t masterthreadid; /* The master thread ID */
+ int workerthreadcount; /* The amount of worker threads. */
+ pthread_t *workerthreadids; /* The worker thread IDs */
+
+ unsigned long start_time; /* Server start time, used for authentication */
+ pthread_mutex_t nonce_mutex; /* Protects nonce_count */
+ unsigned long nonce_count; /* Used nonces, used for authentication */
- char *systemName; /* What operating system is running */
+ char *systemName; /* What operating system is running */
/* linked list of uri handlers */
struct mg_request_handler_info *request_handlers;
#if defined(USE_LUA) && defined(USE_WEBSOCKET)
/* linked list of shared lua websockets */
- struct mg_shared_lua_websocket *shared_lua_websockets;
+ struct mg_shared_lua_websocket_list *shared_lua_websockets;
+#endif
+
+#ifdef USE_TIMERS
+ struct timers * timers;
#endif
};
struct mg_connection {
struct mg_request_info request_info;
struct mg_context *ctx;
- SSL *ssl; /* SSL descriptor */
- SSL_CTX *client_ssl_ctx; /* SSL context for client connections */
- struct socket client; /* Connected client */
- time_t birth_time; /* Time when request was received */
- int64_t num_bytes_sent; /* Total bytes sent to client */
- int64_t content_len; /* Content-Length header value */
- int64_t consumed_content; /* How many bytes of content have been read */
- char *buf; /* Buffer for received data */
- char *path_info; /* PATH_INFO part of the URL */
- int must_close; /* 1 if connection must be closed */
- int in_error_handler; /* 1 if in handler for user defined error pages */
- int buf_size; /* Buffer size */
- int request_len; /* Size of the request + headers in a buffer */
- int data_len; /* Total size of data in a buffer */
- int status_code; /* HTTP reply status code, e.g. 200 */
- int throttle; /* Throttling, bytes/sec. <= 0 means no
- throttle */
- time_t last_throttle_time; /* Last time throttled data was sent */
- int64_t last_throttle_bytes;/* Bytes sent this second */
- pthread_mutex_t mutex; /* Used by mg_lock/mg_unlock to ensure atomic
- transmissions for websockets */
+ SSL *ssl; /* SSL descriptor */
+ SSL_CTX *client_ssl_ctx; /* SSL context for client connections */
+ struct socket client; /* Connected client */
+ time_t birth_time; /* Time when request was received */
+ int64_t num_bytes_sent; /* Total bytes sent to client */
+ int64_t content_len; /* Content-Length header value */
+ int64_t consumed_content; /* How many bytes of content have been read */
+ char *buf; /* Buffer for received data */
+ char *path_info; /* PATH_INFO part of the URL */
+ int must_close; /* 1 if connection must be closed */
+ int in_error_handler; /* 1 if in handler for user defined error pages */
+ int buf_size; /* Buffer size */
+ int request_len; /* Size of the request + headers in a buffer */
+ int data_len; /* Total size of data in a buffer */
+ int status_code; /* HTTP reply status code, e.g. 200 */
+ int throttle; /* Throttling, bytes/sec. <= 0 means no throttle */
+ time_t last_throttle_time; /* Last time throttled data was sent */
+ int64_t last_throttle_bytes; /* Bytes sent this second */
+ pthread_mutex_t mutex; /* Used by mg_lock_connection/mg_unlock_connection to ensure atomic transmissions for websockets */
#if defined(USE_LUA) && defined(USE_WEBSOCKET)
- void * lua_websocket_state; /* Lua_State for a websocket connection */
+ void * lua_websocket_state; /* Lua_State for a websocket connection */
#endif
+ int is_chunked; /* transfer-encoding is chunked */
};
static pthread_key_t sTlsKey; /* Thread local storage index */
@@ -1253,7 +1290,6 @@
}
i = j = 0;
- res = -1;
for (; i < pattern_len; i++, j++) {
if (pattern[i] == '?' && str[j] != '\0') {
continue;
@@ -1298,6 +1334,11 @@
return 1;
}
+static int should_decode_url(const struct mg_connection *conn)
+{
+ return (mg_strcasecmp(conn->ctx->config[DECODE_URL], "yes") == 0);
+}
+
static const char *suggest_connection_header(const struct mg_connection *conn)
{
return should_keep_alive(conn) ? "keep-alive" : "close";
@@ -1381,7 +1422,7 @@
len += mg_vsnprintf(conn, buf + len, sizeof(buf) - len, fmt, ap);
va_end(ap);
}
- DEBUG_TRACE(("[%s]", buf));
+ DEBUG_TRACE("[%s]", buf);
mg_printf(conn, "HTTP/1.1 %d %s\r\n"
"Content-Length: %d\r\n"
@@ -1408,7 +1449,18 @@
static int pthread_mutex_lock(pthread_mutex_t *mutex)
{
- return WaitForSingleObject(*mutex, INFINITE) == WAIT_OBJECT_0? 0 : -1;
+ return WaitForSingleObject(*mutex, INFINITE) == WAIT_OBJECT_0 ? 0 : -1;
+}
+
+static int pthread_mutex_trylock(pthread_mutex_t *mutex)
+{
+ switch (WaitForSingleObject(*mutex, 0)) {
+ case WAIT_OBJECT_0:
+ return 0;
+ case WAIT_TIMEOUT:
+ return -2; /* EBUSY */
+ }
+ return -1;
}
static int pthread_mutex_unlock(pthread_mutex_t *mutex)
@@ -1853,7 +1905,7 @@
int err;
err = GetLastError();
- DEBUG_TRACE(("WaitForSingleObject() failed, error %d", err));
+ DEBUG_TRACE("WaitForSingleObject() failed, error %d", err);
} else {
if (dwevent == WAIT_OBJECT_0) {
CloseHandle(threadid);
@@ -1959,7 +2011,7 @@
mg_snprintf(conn, cmdline, sizeof(cmdline), "%s%s\"%s\\%s\"",
interp, interp[0] == '\0' ? "" : " ", full_dir, prog);
- DEBUG_TRACE(("Running [%s]", cmdline));
+ DEBUG_TRACE("Running [%s]", cmdline);
if (CreateProcessA(NULL, cmdline, NULL, NULL, TRUE,
CREATE_NEW_PROCESS_GROUP, envblk, NULL, &si, &pi) == 0) {
mg_cry(conn, "%s: CreateProcess(%s): %ld",
@@ -2002,9 +2054,9 @@
static void set_close_on_exec(int fd, struct mg_connection *conn /* may be null */)
{
if (fcntl(fd, F_SETFD, FD_CLOEXEC) != 0) {
- if (conn)
- mg_cry(conn, "%s: fcntl(F_SETFD FD_CLOEXEC) failed: %s",
- __func__, strerror(ERRNO));
+ if (conn) {
+ mg_cry(conn, "%s: fcntl(F_SETFD FD_CLOEXEC) failed: %s", __func__, strerror(ERRNO));
+ }
}
}
@@ -2031,8 +2083,7 @@
/* Start a thread storing the thread context. */
-static int mg_start_thread_with_id(mg_thread_func_t func, void *param,
- pthread_t *threadidptr)
+static int mg_start_thread_with_id(mg_thread_func_t func, void *param, pthread_t *threadidptr)
{
pthread_t thread_id;
pthread_attr_t attr;
@@ -2128,8 +2179,7 @@
/* Write data to the IO channel - opened file descriptor, socket or SSL
descriptor. Return number of bytes written. */
-static int64_t push(FILE *fp, SOCKET sock, SSL *ssl, const char *buf,
- int64_t len)
+static int64_t push(FILE *fp, SOCKET sock, SSL *ssl, const char *buf, int64_t len)
{
int64_t sent;
int n, k;
@@ -2207,7 +2257,25 @@
return nread;
}
-int mg_read(struct mg_connection *conn, void *buf, size_t len)
+static void fast_forward_request(struct mg_connection *conn)
+{
+ char buf[MG_BUF_LEN];
+ int to_read, nread;
+
+ while (conn->consumed_content < conn->content_len) {
+ to_read = sizeof(buf);
+ if ((int64_t) to_read > conn->content_len - conn->consumed_content) {
+ to_read = (int) (conn->content_len - conn->consumed_content);
+ }
+
+ nread = mg_read(conn, buf, to_read);
+ if (nread <= 0) {
+ break;
+ }
+ }
+}
+
+int mg_read_inner(struct mg_connection *conn, void *buf, size_t len)
{
int64_t n, buffered_len, nread;
const char *body;
@@ -2248,6 +2316,43 @@
return nread;
}
+static int mg_getc(struct mg_connection *conn) {
+ char c;
+ conn->content_len++;
+ if ( mg_read_inner(conn,&c,1) <= 0 ) return EOF;
+ return c;
+}
+
+int mg_read(struct mg_connection *conn, void *buf, size_t len) {
+ if ( conn->is_chunked ) {
+ if (conn->content_len <= 0 ) conn->content_len = 0;
+ if (conn->consumed_content < conn->content_len) return mg_read_inner(conn,buf,len);
+ int i = 0;
+ char str[64];
+ while (1) {
+ int c = mg_getc(conn);
+ if (c == EOF) return EOF;
+ if ( ! ( c == '\n' || c == '\r' ) ) {
+ str[i++] = c;
+ break;
+ }
+ }
+ for (; i < (int)sizeof(str); i++) {
+ int c = mg_getc(conn);
+ if ( c == EOF ) return -1;
+ str[i] = (char) c;
+ if ( i > 0 && str[i] == '\n' && str[i-1] == '\r' ) break;
+ }
+ char *end = 0;
+ long chunkSize = strtol(str,&end,16);
+ if ( end != str+(i-1) ) return -1;
+ if ( chunkSize == 0 ) return 0;
+ conn->content_len += chunkSize;
+ }
+ return mg_read_inner(conn,buf,len);
+}
+
+
int mg_write(struct mg_connection *conn, const void *buf, size_t len)
{
time_t now;
@@ -2381,7 +2486,7 @@
#define HEXTOI(x) (isdigit(x) ? x - '0' : x - 'W')
for (i = j = 0; i < src_len && j < dst_len - 1; i++, j++) {
- if (src[i] == '%' && i < src_len - 2 &&
+ if (i < src_len - 2 && src[i] == '%' &&
isxdigit(* (const unsigned char *) (src + i + 1)) &&
isxdigit(* (const unsigned char *) (src + i + 2))) {
a = tolower(* (const unsigned char *) (src + i + 1));
@@ -2656,10 +2761,11 @@
{
const char *s, *e;
int len = 0;
+ int in_content = 0;
- for (s = buf, e = s + buflen - 1; len <= 0 && s < e; s++)
+ for (s = buf, e = s + buflen - 1; len <= 0 && s < e; s++) {
/* Control characters are not allowed but >=128 is. */
- if (!isprint(* (const unsigned char *) s) && *s != '\r' &&
+ if (!in_content && !isprint(* (const unsigned char *) s) && *s != '\r' &&
*s != '\n' && * (const unsigned char *) s < 128) {
len = -1;
break; /* [i_a] abort scan as soon as one malformed character is
@@ -2670,8 +2776,14 @@
} else if (s[0] == '\n' && &s[1] < e &&
s[1] == '\r' && s[2] == '\n') {
len = (int) (s - buf) + 3;
+ in_content = 0;
}
+ if (!in_content && *s == ':') {
+ in_content = 1;
+ }
+ }
+
return len;
}
@@ -2724,6 +2836,7 @@
return result;
}
+#ifndef RGW
/* Protect against directory disclosure attack by removing '..',
excessive '/' and '\' characters */
static void remove_double_dots_and_double_slashes(char *s)
@@ -2747,6 +2860,7 @@
}
*p = '\0';
}
+#endif
static const struct {
const char *extension;
@@ -2976,9 +3090,11 @@
}
} else {
/* Try to find .htpasswd in requested directory. */
- for (p = path, e = p + strlen(p) - 1; e > p; e--)
- if (e[0] == '/')
+ for (p = path, e = p + strlen(p) - 1; e > p; e--) {
+ if (e[0] == '/') {
break;
+ }
+ }
mg_snprintf(conn, name, sizeof(name), "%.*s%c%s",
(int) (e - p), p, '/', PASSWORDS_FILE_NAME);
if (!mg_fopen(conn, name, "r", filep)) {
@@ -3108,33 +3224,98 @@
}
}
-/* Authorize against the opened passwords file. Return 1 if authorized. */
-static int authorize(struct mg_connection *conn, struct file *filep)
-{
+struct read_auth_file_struct {
+ struct mg_connection *conn;
struct ah ah;
- char line[256], f_user[256] = "", ha1[256] = "", f_domain[256] = "", buf[MG_BUF_LEN], *p;
+ char *domain;
+ char buf[256+256+40];
+ char *f_user;
+ char *f_domain;
+ char *f_ha1;
+};
- if (!parse_auth_header(conn, buf, sizeof(buf), &ah)) {
- return 0;
- }
+static int read_auth_file(struct file *filep, struct read_auth_file_struct * workdata)
+{
+ char *p;
+ int is_authorized = 0;
+ struct file fp;
+ int l;
/* Loop over passwords file */
p = (char *) filep->membuf;
- while (mg_fgets(line, sizeof(line), filep, &p) != NULL) {
- if (sscanf(line, "%255[^:]:%255[^:]:%255s", f_user, f_domain, ha1) != 3) {
+ while (mg_fgets(workdata->buf, sizeof(workdata->buf), filep, &p) != NULL) {
+
+ l = strlen(workdata->buf);
+ while (l>0) {
+ if (isspace(workdata->buf[l-1]) || iscntrl(workdata->buf[l-1])) {
+ l--;
+ workdata->buf[l] = 0;
+ } else break;
+ }
+ if (l<1) continue;
+
+ workdata->f_user = workdata->buf;
+
+ if (workdata->f_user[0]==':') {
+ /* user names may not contain a ':' and may not be empty,
+ so lines starting with ':' may be used for a special purpose */
+ if (workdata->f_user[1]=='#') {
+ /* :# is a comment */
+ continue;
+ } else if (!strncmp(workdata->f_user+1,"include=",8)) {
+ if (mg_fopen(workdata->conn, workdata->f_user+9, "r", &fp)) {
+ is_authorized = read_auth_file(&fp, workdata);
+ mg_fclose(&fp);
+ } else {
+ mg_cry(workdata->conn, "%s: cannot open authorization file: %s", __func__, workdata->buf);
+ }
+ continue;
+ }
+ /* everything is invalid for the moment (might change in the future) */
+ mg_cry(workdata->conn, "%s: syntax error in authorization file: %s", __func__, workdata->buf);
continue;
}
- f_user[255]=0;
- f_domain[255]=0;
- ha1[255]=0;
-
- if (!strcmp(ah.user, f_user) &&
- !strcmp(conn->ctx->config[AUTHENTICATION_DOMAIN], f_domain))
- return check_password(conn->request_info.request_method, ha1, ah.uri,
- ah.nonce, ah.nc, ah.cnonce, ah.qop, ah.response);
+
+ workdata->f_domain = strchr(workdata->f_user, ':');
+ if (workdata->f_domain == NULL) {
+ mg_cry(workdata->conn, "%s: syntax error in authorization file: %s", __func__, workdata->buf);
+ continue;
+ }
+ *(workdata->f_domain) = 0;
+ (workdata->f_domain)++;
+
+ workdata->f_ha1 = strchr(workdata->f_domain, ':');
+ if (workdata->f_ha1 == NULL) {
+ mg_cry(workdata->conn, "%s: syntax error in authorization file: %s", __func__, workdata->buf);
+ continue;
+ }
+ *(workdata->f_ha1) = 0;
+ (workdata->f_ha1)++;
+
+ if (!strcmp(workdata->ah.user, workdata->f_user) && !strcmp(workdata->domain, workdata->f_domain)) {
+ return check_password(workdata->conn->request_info.request_method, workdata->f_ha1, workdata->ah.uri,
+ workdata->ah.nonce, workdata->ah.nc, workdata->ah.cnonce, workdata->ah.qop, workdata->ah.response);
+ }
}
- return 0;
+ return is_authorized;
+}
+
+/* Authorize against the opened passwords file. Return 1 if authorized. */
+static int authorize(struct mg_connection *conn, struct file *filep)
+{
+ struct read_auth_file_struct workdata;
+ char buf[MG_BUF_LEN];
+
+ memset(&workdata,0,sizeof(workdata));
+ workdata.conn = conn;
+
+ if (!parse_auth_header(conn, buf, sizeof(buf), &workdata.ah)) {
+ return 0;
+ }
+ workdata.domain = conn->ctx->config[AUTHENTICATION_DOMAIN];
+
+ return read_auth_file(filep, &workdata);
}
/* Return 1 if request is authorised, 0 otherwise. */
@@ -3172,16 +3353,16 @@
static void send_authorization_request(struct mg_connection *conn)
{
- char date[64];
- time_t curtime = time(NULL);
- unsigned long nonce = (unsigned long)(conn->ctx->start_time);
-
- (void)pthread_mutex_lock(&conn->ctx->mutex);
+ char date[64];
+ time_t curtime = time(NULL);
+ unsigned long nonce = (unsigned long)(conn->ctx->start_time);
+
+ (void)pthread_mutex_lock(&conn->ctx->nonce_mutex);
nonce += conn->ctx->nonce_count;
++conn->ctx->nonce_count;
- (void)pthread_mutex_unlock(&conn->ctx->mutex);
-
- nonce ^= (unsigned long)(conn->ctx);
+ (void)pthread_mutex_unlock(&conn->ctx->nonce_mutex);
+
+ nonce ^= (unsigned long)(conn->ctx);
conn->status_code = 401;
conn->must_close = 1;
@@ -3215,8 +3396,8 @@
int mg_modify_passwords_file(const char *fname, const char *domain,
const char *user, const char *pass)
{
- int found;
- char line[512], u[512] = "", d[512] ="", ha1[33], tmp[PATH_MAX+1];
+ int found, i;
+ char line[512], u[512] = "", d[512] ="", ha1[33], tmp[PATH_MAX+8];
FILE *fp, *fp2;
found = 0;
@@ -3227,6 +3408,25 @@
pass = NULL;
}
+ /* Other arguments must not be empty */
+ if (fname == NULL || domain == NULL || user == NULL) return 0;
+
+ /* Using the given file format, user name and domain must not contain ':' */
+ if (strchr(user, ':') != NULL) return 0;
+ if (strchr(domain, ':') != NULL) return 0;
+
+ /* Do not allow control characters like newline in user name and domain.
+ Do not allow excessively long names either. */
+ for (i=0; i<255 && user[i]!=0; i++) {
+ if (iscntrl(user[i])) return 0;
+ }
+ if (user[i]) return 0;
+ for (i=0; i<255 && domain[i]!=0; i++) {
+ if (iscntrl(domain[i])) return 0;
+ }
+ if (domain[i]) return 0;
+
+ /* Create a temporary file name */
(void) snprintf(tmp, sizeof(tmp) - 1, "%s.tmp", fname);
tmp[sizeof(tmp) - 1] = 0;
@@ -3822,6 +4022,7 @@
/* HTTP message could be either HTTP request or HTTP response, e.g.
"GET / HTTP/1.0 ...." or "HTTP/1.0 200 OK ..." */
+#ifndef RGW
is_request = is_valid_http_method(ri->request_method);
if ((is_request && memcmp(ri->http_version, "HTTP/", 5) != 0) ||
(!is_request && memcmp(ri->request_method, "HTTP/", 5) != 0)) {
@@ -3832,6 +4033,17 @@
}
parse_http_headers(&buf, ri);
}
+#else
+ is_request = (memcmp(ri->http_version, "HTTP/", 5) == 0);
+ if (is_request) {
+ ri->http_version += 5;
+ }
+ if (is_request || memcmp(ri->request_method, "HTTP/", 5) == 0) {
+ parse_http_headers(&buf, ri);
+ } else {
+ request_length = -1;
+ }
+#endif
}
return request_length;
}
@@ -3928,7 +4140,7 @@
expect = mg_get_header(conn, "Expect");
assert(fp != NULL);
- if (conn->content_len == -1) {
+ if (conn->content_len == -1 && !conn->is_chunked) {
send_http_error(conn, 411, "Length Required", "%s", "");
} else if (expect != NULL && mg_strcasecmp(expect, "100-continue")) {
send_http_error(conn, 417, "Expectation Failed", "%s", "");
@@ -4235,7 +4447,7 @@
Do not send anything back to client, until we buffer in all
HTTP headers. */
data_len = 0;
- buf = mg_malloc(buflen);
+ buf = (char *)mg_malloc(buflen);
if (buf == NULL) {
send_http_error(conn, 500, http_500_error,
"Not enough memory for buffer (%u bytes)",
@@ -4343,7 +4555,7 @@
buf[len] = '\0';
/* Try to create intermediate directory */
- DEBUG_TRACE(("mkdir(%s)", buf));
+ DEBUG_TRACE("mkdir(%s)", buf);
if (!mg_stat(conn, buf, &file) && mg_mkdir(buf, 0755) != 0) {
res = -1;
break;
@@ -4362,7 +4574,7 @@
{
int rc, body_len;
struct de de;
- char date[64];
+ char date[64];
time_t curtime = time(NULL);
memset(&de.file, 0, sizeof(de.file));
@@ -4413,7 +4625,7 @@
const char *range;
int64_t r1, r2;
int rc;
- char date[64];
+ char date[64];
time_t curtime = time(NULL);
conn->status_code = mg_stat(conn, path, &file) ? 200 : 201;
@@ -4687,7 +4899,7 @@
struct file *filep)
{
const char *depth = mg_get_header(conn, "Depth");
- char date[64];
+ char date[64];
time_t curtime = time(NULL);
gmt_time_string(date, sizeof(date), &curtime);
@@ -4717,16 +4929,30 @@
conn->num_bytes_sent += mg_printf(conn, "%s\n", "</d:multistatus>");
}
-void mg_lock(struct mg_connection* conn)
+void mg_lock_connection(struct mg_connection* conn)
{
(void) pthread_mutex_lock(&conn->mutex);
}
-void mg_unlock(struct mg_connection* conn)
+void mg_unlock_connection(struct mg_connection* conn)
{
(void) pthread_mutex_unlock(&conn->mutex);
}
+void mg_lock_context(struct mg_context* ctx)
+{
+ (void) pthread_mutex_lock(&ctx->nonce_mutex);
+}
+
+void mg_unlock_context(struct mg_context* ctx)
+{
+ (void) pthread_mutex_unlock(&ctx->nonce_mutex);
+}
+
+#if defined(USE_TIMERS)
+#include "timer.inl"
+#endif /* USE_TIMERS */
+
#ifdef USE_LUA
#include "mod_lua.inl"
#endif /* USE_LUA */
@@ -5015,9 +5241,9 @@
/* Copy the mask before we shift the queue and destroy it */
if (mask_len > 0) {
- *(uint32_t*)mask = *(uint32_t*)(buf + header_len - mask_len);
+ memcpy(mask, buf + header_len - mask_len, sizeof(mask));
} else {
- *(uint32_t*)mask = 0;
+ memset(mask, 0, sizeof(mask));
}
/* Read frame payload from the first message in the queue into
@@ -5075,7 +5301,7 @@
!conn->ctx->callbacks.websocket_data(conn, mop, data, data_len)) ||
#ifdef USE_LUA
(conn->lua_websocket_state &&
- !lua_websocket_data(conn, mop, data, data_len)) ||
+ !lua_websocket_data(conn, conn->lua_websocket_state, mop, data, data_len)) ||
#endif
(buf[0] & 0xf) == WEBSOCKET_OPCODE_CONNECTION_CLOSE) { /* Opcode == 8, connection close */
break;
@@ -5130,10 +5356,10 @@
but mongoose's mg_printf/mg_write is not (because of the loop in
push(), although that is only a problem if the packet is large or
outgoing buffer is full). */
- (void) mg_lock(conn);
+ (void) mg_lock_connection(conn);
retval = mg_write(conn, header, headerLen);
retval = mg_write(conn, data, dataLen);
- mg_unlock(conn);
+ mg_unlock_connection(conn);
return retval;
}
@@ -5142,7 +5368,7 @@
{
const char *version = mg_get_header(conn, "Sec-WebSocket-Version");
#ifdef USE_LUA
- int lua_websock, shared_lua_websock = 0;
+ int lua_websock = 0;
/* TODO: A websocket script may be shared between several clients, allowing them to communicate
directly instead of writing to a data base and polling the data base. */
#endif
@@ -5155,17 +5381,17 @@
/* The C callback is called before Lua and may prevent Lua from handling the websocket. */
} else {
#ifdef USE_LUA
- lua_websock = conn->ctx->config[LUA_WEBSOCKET_EXTENSIONS] ?
- match_prefix(conn->ctx->config[LUA_WEBSOCKET_EXTENSIONS],
+ if (conn->ctx->config[LUA_WEBSOCKET_EXTENSIONS]) {
+ lua_websock = match_prefix(conn->ctx->config[LUA_WEBSOCKET_EXTENSIONS],
(int)strlen(conn->ctx->config[LUA_WEBSOCKET_EXTENSIONS]),
- path) : 0;
+ path);
+ }
- if (lua_websock || shared_lua_websock) {
- /* TODO */ shared_lua_websock = 0;
- conn->lua_websocket_state = lua_websocket_new(path, conn, !!shared_lua_websock);
+ if (lua_websock) {
+ conn->lua_websocket_state = lua_websocket_new(path, conn);
if (conn->lua_websocket_state) {
send_websocket_handshake(conn);
- if (lua_websocket_ready(conn)) {
+ if (lua_websocket_ready(conn, conn->lua_websocket_state)) {
read_websocket(conn);
}
}
@@ -5295,6 +5521,7 @@
assert(len >= 0 && len <= (int) sizeof(buf));
while ((n = mg_read(conn, buf + len, sizeof(buf) - len)) > 0) {
len += n;
+ assert(len <= (int) sizeof(buf));
}
if ((headers_len = get_request_len(buf, len)) <= 0) {
break;
@@ -5422,7 +5649,7 @@
void mg_set_request_handler(struct mg_context *ctx, const char *uri, mg_request_handler handler, void *cbdata)
{
- struct mg_request_handler_info *tmp_rh, *lastref = 0;
+ struct mg_request_handler_info *tmp_rh, *lastref = NULL;
size_t urilen = strlen(uri);
/* first see it the uri exists */
@@ -5507,10 +5734,10 @@
return tmp_rh->handler(conn, tmp_rh->cbdata);
}
-
- /* try for pattern match */
- if (match_prefix(tmp_rh->uri, tmp_rh->uri_len, uri) > 0) {
- return tmp_rh->handler(conn, tmp_rh->cbdata);
+
+ /* try for pattern match */
+ if (match_prefix(tmp_rh->uri, tmp_rh->uri_len, uri) > 0) {
+ return tmp_rh->handler(conn, tmp_rh->cbdata);
}
}
@@ -5528,21 +5755,28 @@
char path[PATH_MAX];
int uri_len, ssl_index, is_script_resource;
struct file file = STRUCT_FILE_INITIALIZER;
- char date[64];
+ char date[64];
time_t curtime = time(NULL);
if ((conn->request_info.query_string = strchr(ri->uri, '?')) != NULL) {
* ((char *) conn->request_info.query_string++) = '\0';
}
uri_len = (int) strlen(ri->uri);
- mg_url_decode(ri->uri, uri_len, (char *) ri->uri, uri_len + 1, 0);
+
+ if (should_decode_url(conn)) {
+ mg_url_decode(ri->uri, uri_len, (char *) ri->uri, uri_len + 1, 0);
+ }
+
+#ifndef RGW
remove_double_dots_and_double_slashes((char *) ri->uri);
+#endif
+
path[0] = '\0';
convert_uri_to_file_name(conn, path, sizeof(path), &file, &is_script_resource);
conn->throttle = set_throttle(conn->ctx->config[THROTTLE],
get_remote_ip(conn), ri->uri);
- DEBUG_TRACE(("%s", ri->uri));
+ DEBUG_TRACE("%s", ri->uri);
/* Perform redirect and auth checks before calling begin_request() handler.
Otherwise, begin_request() would need to perform auth checks and
redirects. */
@@ -5555,6 +5789,7 @@
} else if (conn->ctx->callbacks.begin_request != NULL &&
conn->ctx->callbacks.begin_request(conn)) {
/* Do nothing, callback has served the request */
+ fast_forward_request(conn);
#if defined(USE_WEBSOCKET)
} else if (is_websocket_request(conn)) {
handle_websocket_request(conn, path, is_script_resource);
@@ -5630,7 +5865,7 @@
(int)strlen(conn->ctx->config[LUA_SERVER_PAGE_EXTENSIONS]),
path) > 0) {
/* Lua server page: an SSI like page containing mostly plain html code plus some tags with server generated contents. */
- handle_lsp_request(conn, path, &file, NULL);
+ handle_lsp_request(conn, path, file, NULL);
} else if (match_prefix(conn->ctx->config[LUA_SCRIPT_EXTENSIONS],
(int)strlen(conn->ctx->config[LUA_SCRIPT_EXTENSIONS]),
path) > 0) {
@@ -5663,9 +5898,9 @@
ctx->listening_sockets[i].sock = INVALID_SOCKET;
}
mg_free(ctx->listening_sockets);
- ctx->listening_sockets=0;
+ ctx->listening_sockets = NULL;
mg_free(ctx->listening_ports);
- ctx->listening_ports=0;
+ ctx->listening_ports = NULL;
}
static int is_valid_port(unsigned int port)
@@ -5695,7 +5930,7 @@
so->lsa.sin.sin_addr.s_addr = htonl((a << 24) | (b << 16) | (c << 8) | d);
so->lsa.sin.sin_port = htons((uint16_t) port);
#if defined(USE_IPV6)
- } else if (sscanf(vec->ptr, "[%49[^]]]:%d%n", buf, &port, &len) == 2 &&
+ } else if (sscanf(vec->ptr, "[%49[^]]]:%u%n", buf, &port, &len) == 2 &&
inet_pton(AF_INET6, buf, &so->lsa.sin6.sin6_addr)) {
/* IPv6 address, e.g. [3ffe:2a00:100:7031::1]:8080 */
so->lsa.sin6.sin6_family = AF_INET6;
@@ -5776,6 +6011,7 @@
sizeof(ctx->listening_ports[0]))) == NULL) {
closesocket(so.sock);
so.sock = INVALID_SOCKET;
+ mg_free(ptr);
success = 0;
}
else {
@@ -5795,15 +6031,14 @@
return success;
}
-static void log_header(const struct mg_connection *conn, const char *header,
- FILE *fp)
+static const char* header_val(const struct mg_connection *conn, const char *header)
{
const char *header_value;
if ((header_value = mg_get_header(conn, header)) == NULL) {
- (void) fprintf(fp, "%s", " -");
+ return "-";
} else {
- (void) fprintf(fp, " \"%s\"", header_value);
+ return header_value;
}
}
@@ -5814,10 +6049,15 @@
char date[64], src_addr[IP_ADDR_STR_LEN];
struct tm *tm;
+ const char *referer;
+ const char *user_agent;
+
+ char buf[4096];
+
fp = conn->ctx->config[ACCESS_LOG_FILE] == NULL ? NULL :
fopen(conn->ctx->config[ACCESS_LOG_FILE], "a+");
- if (fp == NULL)
+ if (fp == NULL && conn->ctx->callbacks.log_message == NULL)
return;
tm = localtime(&conn->birth_time);
@@ -5829,21 +6069,30 @@
}
ri = &conn->request_info;
- flockfile(fp);
sockaddr_to_string(src_addr, sizeof(src_addr), &conn->client.rsa);
- fprintf(fp, "%s - %s [%s] \"%s %s HTTP/%s\" %d %" INT64_FMT,
+ referer = header_val(conn, "Referer");
+ user_agent = header_val(conn, "User-Agent");
+
+ snprintf(buf, sizeof(buf), "%s - %s [%s] \"%s %s HTTP/%s\" %d %" INT64_FMT " %s %s",
src_addr, ri->remote_user == NULL ? "-" : ri->remote_user, date,
ri->request_method ? ri->request_method : "-",
ri->uri ? ri->uri : "-", ri->http_version,
- conn->status_code, conn->num_bytes_sent);
- log_header(conn, "Referer", fp);
- log_header(conn, "User-Agent", fp);
- fputc('\n', fp);
- fflush(fp);
+ conn->status_code, conn->num_bytes_sent,
+ referer, user_agent);
- funlockfile(fp);
- fclose(fp);
+ if (conn->ctx->callbacks.log_access) {
+ conn->ctx->callbacks.log_access(conn, buf);
+ }
+
+ if (fp) {
+ flockfile(fp);
+ fprintf(fp, "%s", buf);
+ fputc('\n', fp);
+ fflush(fp);
+ funlockfile(fp);
+ fclose(fp);
+ }
}
/* Verify given socket address against the ACL.
@@ -6072,6 +6321,7 @@
conn->num_bytes_sent = conn->consumed_content = 0;
conn->status_code = -1;
conn->must_close = conn->request_len = conn->throttle = 0;
+ conn->is_chunked = 0;
}
static void close_socket_gracefully(struct mg_connection *conn)
@@ -6116,7 +6366,8 @@
{
#if defined(USE_LUA) && defined(USE_WEBSOCKET)
if (conn->lua_websocket_state) {
- lua_websocket_close(conn);
+ lua_websocket_close(conn, conn->lua_websocket_state);
+ conn->lua_websocket_state = NULL;
}
#endif
@@ -6124,7 +6375,7 @@
if (conn->ctx->callbacks.connection_close != NULL)
conn->ctx->callbacks.connection_close(conn);
- mg_lock(conn);
+ mg_lock_connection(conn);
conn->must_close = 1;
@@ -6141,7 +6392,7 @@
conn->client.sock = INVALID_SOCKET;
}
- mg_unlock(conn);
+ mg_unlock_connection(conn);
}
void mg_close_connection(struct mg_connection *conn)
@@ -6156,7 +6407,7 @@
mg_free(conn);
}
-struct mg_connection *mg_connect(const char *host, int port, int use_ssl,
+static struct mg_connection *mg_connect(const char *host, int port, int use_ssl,
char *ebuf, size_t ebuf_len)
{
static struct mg_context fake_ctx;
@@ -6169,13 +6420,11 @@
mg_calloc(1, sizeof(*conn) + MAX_REQUEST_SIZE)) == NULL) {
snprintf(ebuf, ebuf_len, "calloc(): %s", strerror(ERRNO));
closesocket(sock);
- sock = INVALID_SOCKET;
#ifndef NO_SSL
} else if (use_ssl && (conn->client_ssl_ctx =
SSL_CTX_new(SSLv23_client_method())) == NULL) {
snprintf(ebuf, ebuf_len, "SSL_CTX_new error");
closesocket(sock);
- sock = INVALID_SOCKET;
mg_free(conn);
conn = NULL;
#endif /* NO_SSL */
@@ -6213,11 +6462,12 @@
return uri[0] == '/' || (uri[0] == '*' && uri[1] == '\0');
}
-static int getreq(struct mg_connection *conn, char *ebuf, size_t ebuf_len)
+static int getreq(struct mg_connection *conn, char *ebuf, size_t ebuf_len, int *err)
{
const char *cl;
ebuf[0] = '\0';
+ *err = 0;
reset_per_request_attributes(conn);
conn->request_len = read_request(NULL, conn, conn->buf, conn->buf_size,
&conn->data_len);
@@ -6225,16 +6475,30 @@
if (conn->request_len == 0 && conn->data_len == conn->buf_size) {
snprintf(ebuf, ebuf_len, "%s", "Request Too Large");
+ *err = 400;
+ return 0;
} else if (conn->request_len <= 0) {
snprintf(ebuf, ebuf_len, "%s", "Client closed connection");
+ return 0;
} else if (parse_http_message(conn->buf, conn->buf_size,
&conn->request_info) <= 0) {
snprintf(ebuf, ebuf_len, "Bad request: [%.*s]", conn->data_len, conn->buf);
+ *err = 400;
+ return 0;
} else {
/* Message is a valid request or response */
- if ((cl = get_header(&conn->request_info, "Content-Length")) != NULL) {
+ if (( cl = get_header(&conn->request_info, "Transfer-encoding")) != NULL && strcmp(cl,"chunked") == 0) {
+ conn->is_chunked = 1;
+ conn->content_len = 0;
+ } else if ((cl = get_header(&conn->request_info, "Content-Length")) != NULL) {
/* Request/response has content length set */
- conn->content_len = strtoll(cl, NULL, 10);
+ char *endptr;
+ conn->content_len = strtoll(cl, &endptr, 10);
+ if (endptr == cl) {
+ snprintf(ebuf, ebuf_len, "%s", "Bad Request");
+ *err = 400;
+ return 0;
+ }
} else if (!mg_strcasecmp(conn->request_info.request_method, "POST") ||
!mg_strcasecmp(conn->request_info.request_method, "PUT")) {
/* POST or PUT request without content length set */
@@ -6248,7 +6512,7 @@
}
conn->birth_time = time(NULL);
}
- return ebuf[0] == '\0';
+ return 1;
}
struct mg_connection *mg_download(const char *host, int port, int use_ssl,
@@ -6264,7 +6528,8 @@
} else if (mg_vprintf(conn, fmt, ap) <= 0) {
snprintf(ebuf, ebuf_len, "%s", "Error sending request");
} else {
- getreq(conn, ebuf, ebuf_len);
+ int err;
+ getreq(conn, ebuf, ebuf_len, &err);
}
if (ebuf[0] != '\0' && conn != NULL) {
mg_close_connection(conn);
@@ -6282,14 +6547,16 @@
char ebuf[100];
keep_alive_enabled = !strcmp(conn->ctx->config[ENABLE_KEEP_ALIVE], "yes");
- keep_alive = 0;
/* Important: on new connection, reset the receiving buffer. Credit goes
to crule42. */
conn->data_len = 0;
do {
- if (!getreq(conn, ebuf, sizeof(ebuf))) {
- send_http_error(conn, 500, "Server Error", "%s", ebuf);
+ int err;
+ if (!getreq(conn, ebuf, sizeof(ebuf), &err)) {
+ if (err > 0) {
+ send_http_error(conn, err, "Bad Request", "%s", ebuf);
+ }
conn->must_close = 1;
} else if (!is_valid_uri(conn->request_info.uri)) {
snprintf(ebuf, sizeof(ebuf), "Invalid URI: [%s]", ri->uri);
@@ -6336,12 +6603,12 @@
/* Worker threads take accepted socket from the queue */
static int consume_socket(struct mg_context *ctx, struct socket *sp)
{
- (void) pthread_mutex_lock(&ctx->mutex);
- DEBUG_TRACE(("going idle"));
+ (void) pthread_mutex_lock(&ctx->thread_mutex);
+ DEBUG_TRACE("going idle");
/* If the queue is empty, wait. We're idle at this point. */
while (ctx->sq_head == ctx->sq_tail && ctx->stop_flag == 0) {
- pthread_cond_wait(&ctx->sq_full, &ctx->mutex);
+ pthread_cond_wait(&ctx->sq_full, &ctx->thread_mutex);
}
/* If we're stopping, sq_head may be equal to sq_tail. */
@@ -6349,7 +6616,7 @@
/* Copy socket from the queue and increment tail */
*sp = ctx->queue[ctx->sq_tail % ARRAY_SIZE(ctx->queue)];
ctx->sq_tail++;
- DEBUG_TRACE(("grabbed socket %d, going busy", sp->sock));
+ DEBUG_TRACE("grabbed socket %d, going busy", sp->sock);
/* Wrap pointers if needed */
while (ctx->sq_tail > (int) ARRAY_SIZE(ctx->queue)) {
@@ -6359,7 +6626,7 @@
}
(void) pthread_cond_signal(&ctx->sq_empty);
- (void) pthread_mutex_unlock(&ctx->mutex);
+ (void) pthread_mutex_unlock(&ctx->thread_mutex);
return !ctx->stop_flag;
}
@@ -6417,19 +6684,19 @@
}
/* Signal master that we're done with connection and exiting */
- (void) pthread_mutex_lock(&ctx->mutex);
+ (void) pthread_mutex_lock(&ctx->thread_mutex);
ctx->num_threads--;
- (void) pthread_cond_signal(&ctx->cond);
+ (void) pthread_cond_signal(&ctx->thread_cond);
assert(ctx->num_threads >= 0);
- (void) pthread_mutex_unlock(&ctx->mutex);
+ (void) pthread_mutex_unlock(&ctx->thread_mutex);
- pthread_setspecific(sTlsKey, 0);
+ pthread_setspecific(sTlsKey, NULL);
#if defined(_WIN32) && !defined(__SYMBIAN32__)
CloseHandle(tls.pthread_cond_helper_mutex);
#endif
mg_free(conn);
- DEBUG_TRACE(("exiting"));
+ DEBUG_TRACE("exiting");
return NULL;
}
@@ -6452,23 +6719,23 @@
/* Master thread adds accepted socket to a queue */
static void produce_socket(struct mg_context *ctx, const struct socket *sp)
{
- (void) pthread_mutex_lock(&ctx->mutex);
+ (void) pthread_mutex_lock(&ctx->thread_mutex);
/* If the queue is full, wait */
while (ctx->stop_flag == 0 &&
ctx->sq_head - ctx->sq_tail >= (int) ARRAY_SIZE(ctx->queue)) {
- (void) pthread_cond_wait(&ctx->sq_empty, &ctx->mutex);
+ (void) pthread_cond_wait(&ctx->sq_empty, &ctx->thread_mutex);
}
if (ctx->sq_head - ctx->sq_tail < (int) ARRAY_SIZE(ctx->queue)) {
/* Copy socket to the queue and increment head */
ctx->queue[ctx->sq_head % ARRAY_SIZE(ctx->queue)] = *sp;
ctx->sq_head++;
- DEBUG_TRACE(("queued socket %d", sp->sock));
+ DEBUG_TRACE("queued socket %d", sp->sock);
}
(void) pthread_cond_signal(&ctx->sq_full);
- (void) pthread_mutex_unlock(&ctx->mutex);
+ (void) pthread_mutex_unlock(&ctx->thread_mutex);
}
static int set_sock_timeout(SOCKET sock, int milliseconds)
@@ -6500,7 +6767,7 @@
so.sock = INVALID_SOCKET;
} else {
/* Put so socket structure into the queue */
- DEBUG_TRACE(("Accepted socket %d", (int) so.sock));
+ DEBUG_TRACE("Accepted socket %d", (int) so.sock);
set_close_on_exec(so.sock, fc(ctx));
so.is_ssl = listener->is_ssl;
so.ssl_redir = listener->ssl_redir;
@@ -6581,7 +6848,7 @@
}
}
mg_free(pfd);
- DEBUG_TRACE(("stopping workers"));
+ DEBUG_TRACE("stopping workers");
/* Stop signal received: somebody called mg_stop. Quit. */
close_all_listening_sockets(ctx);
@@ -6590,11 +6857,11 @@
pthread_cond_broadcast(&ctx->sq_full);
/* Wait until all threads finish */
- (void) pthread_mutex_lock(&ctx->mutex);
+ (void) pthread_mutex_lock(&ctx->thread_mutex);
while (ctx->num_threads > 0) {
- (void) pthread_cond_wait(&ctx->cond, &ctx->mutex);
+ (void) pthread_cond_wait(&ctx->thread_cond, &ctx->thread_mutex);
}
- (void) pthread_mutex_unlock(&ctx->mutex);
+ (void) pthread_mutex_unlock(&ctx->thread_mutex);
/* Join all worker threads to avoid leaking threads. */
workerthreadcount = ctx->workerthreadcount;
@@ -6605,12 +6872,12 @@
#if !defined(NO_SSL)
uninitialize_ssl(ctx);
#endif
- DEBUG_TRACE(("exiting"));
+ DEBUG_TRACE("exiting");
#if defined(_WIN32) && !defined(__SYMBIAN32__)
CloseHandle(tls.pthread_cond_helper_mutex);
#endif
- pthread_setspecific(sTlsKey, 0);
+ pthread_setspecific(sTlsKey, NULL);
/* Signal mg_stop() that we're done.
WARNING: This must be the very last thing this
@@ -6619,7 +6886,6 @@
}
/* Threads have different return types on Windows and Unix. */
-
#ifdef _WIN32
static unsigned __stdcall master_thread(void *thread_func_param)
{
@@ -6642,12 +6908,19 @@
if (ctx == NULL)
return;
- /* All threads exited, no sync is needed. Destroy mutex and condvars */
- (void) pthread_mutex_destroy(&ctx->mutex);
- (void) pthread_cond_destroy(&ctx->cond);
+ /* All threads exited, no sync is needed. Destroy thread mutex and condvars */
+ (void) pthread_mutex_destroy(&ctx->thread_mutex);
+ (void) pthread_cond_destroy(&ctx->thread_cond);
(void) pthread_cond_destroy(&ctx->sq_empty);
(void) pthread_cond_destroy(&ctx->sq_full);
+ /* Destroy other context global data structures mutex */
+ (void) pthread_mutex_destroy(&ctx->nonce_mutex);
+
+#if defined(USE_TIMERS)
+ timers_exit(ctx);
+#endif
+
/* Deallocate config parameters */
for (i = 0; i < NUM_OPTIONS; i++) {
if (ctx->config[i] != NULL)
@@ -6710,20 +6983,20 @@
#endif /* _WIN32 && !__SYMBIAN32__ */
}
-void get_system_name(char **sysName)
+static void get_system_name(char **sysName)
{
#if defined(_WIN32)
#if !defined(__SYMBIAN32__)
char name[128];
- DWORD dwVersion = 0;
- DWORD dwMajorVersion = 0;
- DWORD dwMinorVersion = 0;
+ DWORD dwVersion = 0;
+ DWORD dwMajorVersion = 0;
+ DWORD dwMinorVersion = 0;
DWORD dwBuild = 0;
- dwVersion = GetVersion();
-
- dwMajorVersion = (DWORD)(LOBYTE(LOWORD(dwVersion)));
- dwMinorVersion = (DWORD)(HIBYTE(LOWORD(dwVersion)));
+ dwVersion = GetVersion();
+
+ dwMajorVersion = (DWORD)(LOBYTE(LOWORD(dwVersion)));
+ dwMinorVersion = (DWORD)(HIBYTE(LOWORD(dwVersion)));
dwBuild = ((dwVersion < 0x80000000) ? (DWORD)(HIWORD(dwVersion)) : 0);
sprintf(name, "Windows %d.%d", dwMajorVersion, dwMinorVersion);
@@ -6745,7 +7018,7 @@
{
struct mg_context *ctx;
const char *name, *value, *default_value;
- int i;
+ int i, ok;
int workerthreadcount;
#if defined(_WIN32) && !defined(__SYMBIAN32__)
@@ -6767,6 +7040,7 @@
if (sTlsInit==0) {
if (0 != pthread_key_create(&sTlsKey, NULL)) {
+ /* Fatal error - abort start. However, this situation should never occur in practice. */
mg_cry(fc(ctx), "Cannot initialize thread local storage");
mg_free(ctx);
return NULL;
@@ -6774,11 +7048,23 @@
sTlsInit++;
}
+ ok = 0==pthread_mutex_init(&ctx->thread_mutex, NULL);
+ ok &= 0==pthread_cond_init(&ctx->thread_cond, NULL);
+ ok &= 0==pthread_cond_init(&ctx->sq_empty, NULL);
+ ok &= 0==pthread_cond_init(&ctx->sq_full, NULL);
+ ok &= 0==pthread_mutex_init(&ctx->nonce_mutex, NULL);
+ if (!ok) {
+ /* Fatal error - abort start. However, this situation should never occur in practice. */
+ mg_cry(fc(ctx), "Cannot initialize thread synchronization objects");
+ mg_free(ctx);
+ return NULL;
+ }
+
if (callbacks) {
ctx->callbacks = *callbacks;
}
ctx->user_data = user_data;
- ctx->request_handlers = 0;
+ ctx->request_handlers = NULL;
#if defined(USE_LUA) && defined(USE_WEBSOCKET)
ctx->shared_lua_websockets = 0;
@@ -6799,7 +7085,7 @@
mg_free(ctx->config[i]);
}
ctx->config[i] = mg_strdup(value);
- DEBUG_TRACE(("[%s] -> [%s]", name, value));
+ DEBUG_TRACE("[%s] -> [%s]", name, value);
}
/* Set default value if needed */
@@ -6833,11 +7119,6 @@
(void) signal(SIGPIPE, SIG_IGN);
#endif /* !_WIN32 && !__SYMBIAN32__ */
- (void) pthread_mutex_init(&ctx->mutex, NULL);
- (void) pthread_cond_init(&ctx->cond, NULL);
- (void) pthread_cond_init(&ctx->sq_empty, NULL);
- (void) pthread_cond_init(&ctx->sq_full, NULL);
-
workerthreadcount = atoi(ctx->config[NUM_THREADS]);
if (workerthreadcount > MAX_WORKER_THREADS) {
@@ -6848,7 +7129,7 @@
if (workerthreadcount > 0) {
ctx->workerthreadcount = workerthreadcount;
- ctx->workerthreadids = mg_calloc(workerthreadcount, sizeof(pthread_t));
+ ctx->workerthreadids = (pthread_t *)mg_calloc(workerthreadcount, sizeof(pthread_t));
if (ctx->workerthreadids == NULL) {
mg_cry(fc(ctx), "Not enough memory for worker thread ID array");
free_context(ctx);
@@ -6856,19 +7137,27 @@
}
}
+#if defined(USE_TIMERS)
+ if (timers_init(ctx) != 0) {
+ mg_cry(fc(ctx), "Error creating timers");
+ free_context(ctx);
+ return NULL;
+ }
+#endif
+
/* Start master (listening) thread */
mg_start_thread_with_id(master_thread, ctx, &ctx->masterthreadid);
/* Start worker threads */
for (i = 0; i < workerthreadcount; i++) {
- (void) pthread_mutex_lock(&ctx->mutex);
+ (void) pthread_mutex_lock(&ctx->thread_mutex);
ctx->num_threads++;
- (void) pthread_mutex_unlock(&ctx->mutex);
+ (void) pthread_mutex_unlock(&ctx->thread_mutex);
if (mg_start_thread_with_id(worker_thread, ctx,
&ctx->workerthreadids[i]) != 0) {
- (void) pthread_mutex_lock(&ctx->mutex);
+ (void) pthread_mutex_lock(&ctx->thread_mutex);
ctx->num_threads--;
- (void) pthread_mutex_unlock(&ctx->mutex);
+ (void) pthread_mutex_unlock(&ctx->thread_mutex);
mg_cry(fc(ctx), "Cannot start worker thread: %ld", (long) ERRNO);
}
}
diff -Nru ceph-0.80.7/src/client/Client.cc ceph-0.80.9/src/client/Client.cc
--- ceph-0.80.7/src/client/Client.cc 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/client/Client.cc 2015-03-09 20:51:17.000000000 +0100
@@ -53,7 +53,7 @@
#include "mon/MonClient.h"
-#include "mds/MDSMap.h"
+#include "mds/flock.h"
#include "osd/OSDMap.h"
#include "mon/MonMap.h"
@@ -149,6 +149,7 @@
logger(NULL),
m_command_hook(this),
timer(m->cct, client_lock),
+ switch_interrupt_cb(NULL),
ino_invalidate_cb(NULL),
ino_invalidate_cb_handle(NULL),
dentry_invalidate_cb(NULL),
@@ -157,6 +158,7 @@
getgroups_cb_handle(NULL),
async_ino_invalidator(m->cct),
async_dentry_invalidator(m->cct),
+ interrupt_finisher(m->cct),
tick_event(NULL),
monclient(mc), messenger(m), whoami(m->get_myname().num()),
initialized(false), mounted(false), unmounting(false),
@@ -443,6 +445,12 @@
async_dentry_invalidator.stop();
}
+ if (switch_interrupt_cb) {
+ ldout(cct, 10) << "shutdown stopping interrupt finisher" << dendl;
+ interrupt_finisher.wait_for_empty();
+ interrupt_finisher.stop();
+ }
+
objectcacher->stop(); // outside of client_lock! this does a join.
client_lock.Lock();
@@ -1437,15 +1445,15 @@
void Client::put_request(MetaRequest *request)
{
- if (request->get_num_ref() == 1) {
+ if (request->_put()) {
if (request->inode())
put_inode(request->take_inode());
if (request->old_inode())
put_inode(request->take_old_inode());
if (request->other_inode())
put_inode(request->take_other_inode());
+ delete request;
}
- request->_put();
}
int Client::encode_inode_release(Inode *in, MetaRequest *req,
@@ -2034,6 +2042,9 @@
in->make_long_path(path);
ldout(cct, 10) << " path " << path << dendl;
+ bufferlist flockbl;
+ _encode_filelocks(in, flockbl);
+
in->caps[mds]->seq = 0; // reset seq.
in->caps[mds]->issue_seq = 0; // reset seq.
in->caps[mds]->mseq = 0; // reset seq.
@@ -2042,7 +2053,8 @@
path.get_ino(), path.get_path(), // ino
in->caps_wanted(), // wanted
in->caps[mds]->issued, // issued
- in->snaprealm->ino);
+ in->snaprealm->ino,
+ flockbl);
if (did_snaprealm.count(in->snaprealm->ino) == 0) {
ldout(cct, 10) << " snaprealm " << *in->snaprealm << dendl;
@@ -2182,6 +2194,8 @@
in->snaprealm_item.remove_myself();
if (in == root)
root = 0;
+ delete in->fcntl_locks;
+ delete in->flock_locks;
delete in;
}
}
@@ -3659,9 +3673,10 @@
int mds = session->mds_num;
int dirty = m->get_dirty();
int cleaned = 0;
+ uint16_t flush_ack_tid = static_cast<uint16_t>(m->get_client_tid());
for (int i = 0; i < CEPH_CAP_BITS; ++i) {
if ((dirty & (1 << i)) &&
- (m->get_client_tid() == in->flushing_cap_tid[i]))
+ (flush_ack_tid == in->flushing_cap_tid[i]))
cleaned |= 1 << i;
}
@@ -5869,6 +5884,8 @@
in->snap_cap_refs--;
}
+ _release_filelocks(f);
+
put_inode(in);
delete f;
@@ -6857,6 +6874,290 @@
return rval;
}
+int Client::_do_filelock(Inode *in, Fh *fh, int lock_type, int op, int sleep,
+ struct flock *fl, uint64_t owner, void *fuse_req)
+{
+ ldout(cct, 10) << "_do_filelock ino " << in->ino
+ << (lock_type == CEPH_LOCK_FCNTL ? " fcntl" : " flock")
+ << " type " << fl->l_type << " owner " << owner
+ << " " << fl->l_start << "~" << fl->l_len << dendl;
+
+ int lock_cmd;
+ if (F_RDLCK == fl->l_type)
+ lock_cmd = CEPH_LOCK_SHARED;
+ else if (F_WRLCK == fl->l_type)
+ lock_cmd = CEPH_LOCK_EXCL;
+ else if (F_UNLCK == fl->l_type)
+ lock_cmd = CEPH_LOCK_UNLOCK;
+ else
+ return -EIO;
+
+ if (op != CEPH_MDS_OP_SETFILELOCK || lock_cmd == CEPH_LOCK_UNLOCK)
+ sleep = 0;
+
+ /*
+ * Set the most significant bit, so that MDS knows the 'owner'
+ * is sufficient to identify the owner of lock. (old code uses
+ * both 'owner' and 'pid')
+ */
+ owner |= (1ULL << 63);
+
+ MetaRequest *req = new MetaRequest(op);
+ filepath path;
+ in->make_nosnap_relative_path(path);
+ req->set_filepath(path);
+ req->set_inode(in);
+
+ req->head.args.filelock_change.rule = lock_type;
+ req->head.args.filelock_change.type = lock_cmd;
+ req->head.args.filelock_change.owner = owner;
+ req->head.args.filelock_change.pid = fl->l_pid;
+ req->head.args.filelock_change.start = fl->l_start;
+ req->head.args.filelock_change.length = fl->l_len;
+ req->head.args.filelock_change.wait = sleep;
+
+ int ret;
+ bufferlist bl;
+
+ if (sleep && switch_interrupt_cb && fuse_req) {
+ // enable interrupt
+ switch_interrupt_cb(fuse_req, req->get());
+
+ ret = make_request(req, -1, -1, NULL, NULL, -1, &bl);
+
+ // disable interrupt
+ switch_interrupt_cb(fuse_req, NULL);
+ put_request(req);
+ } else {
+ ret = make_request(req, -1, -1, NULL, NULL, -1, &bl);
+ }
+
+ if (ret == 0) {
+ if (op == CEPH_MDS_OP_GETFILELOCK) {
+ ceph_filelock filelock;
+ bufferlist::iterator p = bl.begin();
+ ::decode(filelock, p);
+
+ if (CEPH_LOCK_SHARED == filelock.type)
+ fl->l_type = F_RDLCK;
+ else if (CEPH_LOCK_EXCL == filelock.type)
+ fl->l_type = F_WRLCK;
+ else
+ fl->l_type = F_UNLCK;
+
+ fl->l_whence = SEEK_SET;
+ fl->l_start = filelock.start;
+ fl->l_len = filelock.length;
+ fl->l_pid = filelock.pid;
+ } else if (op == CEPH_MDS_OP_SETFILELOCK) {
+ ceph_lock_state_t *lock_state;
+ if (lock_type == CEPH_LOCK_FCNTL) {
+ if (!in->fcntl_locks)
+ in->fcntl_locks = new ceph_lock_state_t(cct);
+ lock_state = in->fcntl_locks;
+ } else if (lock_type == CEPH_LOCK_FLOCK) {
+ if (!in->flock_locks)
+ in->flock_locks = new ceph_lock_state_t(cct);
+ lock_state = in->flock_locks;
+ } else
+ assert(0);
+ _update_lock_state(fl, owner, lock_state);
+
+ if (fh) {
+ if (lock_type == CEPH_LOCK_FCNTL) {
+ if (!fh->fcntl_locks)
+ fh->fcntl_locks = new ceph_lock_state_t(cct);
+ lock_state = fh->fcntl_locks;
+ } else {
+ if (!fh->flock_locks)
+ fh->flock_locks = new ceph_lock_state_t(cct);
+ lock_state = fh->flock_locks;
+ }
+ _update_lock_state(fl, owner, lock_state);
+ }
+ } else
+ assert(0);
+ }
+ return ret;
+}
+
+int Client::_interrupt_filelock(MetaRequest *req)
+{
+ Inode *in = req->inode();
+
+ int lock_type;
+ if (req->head.args.filelock_change.rule == CEPH_LOCK_FLOCK)
+ lock_type = CEPH_LOCK_FLOCK_INTR;
+ else if (req->head.args.filelock_change.rule == CEPH_LOCK_FCNTL)
+ lock_type = CEPH_LOCK_FCNTL_INTR;
+ else
+ assert(0);
+
+ MetaRequest *intr_req = new MetaRequest(CEPH_MDS_OP_SETFILELOCK);
+ filepath path;
+ in->make_nosnap_relative_path(path);
+ intr_req->set_filepath(path);
+ intr_req->set_inode(in);
+ intr_req->head.args.filelock_change = req->head.args.filelock_change;
+ intr_req->head.args.filelock_change.rule = lock_type;
+ intr_req->head.args.filelock_change.type = CEPH_LOCK_UNLOCK;
+
+ return make_request(intr_req, -1, -1, NULL, NULL, -1);
+}
+
+void Client::_encode_filelocks(Inode *in, bufferlist& bl)
+{
+ if (!in->fcntl_locks && !in->flock_locks)
+ return;
+
+ unsigned nr_fcntl_locks = in->fcntl_locks ? in->fcntl_locks->held_locks.size() : 0;
+ ::encode(nr_fcntl_locks, bl);
+ if (nr_fcntl_locks) {
+ ceph_lock_state_t* lock_state = in->fcntl_locks;
+ for(multimap<uint64_t, ceph_filelock>::iterator p = lock_state->held_locks.begin();
+ p != lock_state->held_locks.end();
+ ++p)
+ ::encode(p->second, bl);
+ }
+
+ unsigned nr_flock_locks = in->flock_locks ? in->flock_locks->held_locks.size() : 0;
+ ::encode(nr_flock_locks, bl);
+ if (nr_flock_locks) {
+ ceph_lock_state_t* lock_state = in->flock_locks;
+ for(multimap<uint64_t, ceph_filelock>::iterator p = lock_state->held_locks.begin();
+ p != lock_state->held_locks.end();
+ ++p)
+ ::encode(p->second, bl);
+ }
+
+ ldout(cct, 10) << "_encode_filelocks ino " << in->ino << ", " << nr_fcntl_locks
+ << " fcntl locks, " << nr_flock_locks << " flock locks" << dendl;
+}
+
+void Client::_release_filelocks(Fh *fh)
+{
+ if (!fh->fcntl_locks && !fh->flock_locks)
+ return;
+
+ Inode *in = fh->inode;
+ ldout(cct, 10) << "_release_filelocks " << fh << " ino " << in->ino << dendl;
+
+ list<pair<int, ceph_filelock> > to_release;
+
+ if (fh->fcntl_locks) {
+ ceph_lock_state_t* lock_state = fh->fcntl_locks;
+ for(multimap<uint64_t, ceph_filelock>::iterator p = lock_state->held_locks.begin();
+ p != lock_state->held_locks.end();
+ ++p)
+ to_release.push_back(pair<int, ceph_filelock>(CEPH_LOCK_FCNTL, p->second));
+ delete fh->fcntl_locks;
+ }
+ if (fh->flock_locks) {
+ ceph_lock_state_t* lock_state = fh->flock_locks;
+ for(multimap<uint64_t, ceph_filelock>::iterator p = lock_state->held_locks.begin();
+ p != lock_state->held_locks.end();
+ ++p)
+ to_release.push_back(pair<int, ceph_filelock>(CEPH_LOCK_FLOCK, p->second));
+ delete fh->flock_locks;
+ }
+
+ if (to_release.empty())
+ return;
+
+ struct flock fl;
+ memset(&fl, 0, sizeof(fl));
+ fl.l_whence = SEEK_SET;
+ fl.l_type = F_UNLCK;
+
+ for (list<pair<int, ceph_filelock> >::iterator p = to_release.begin();
+ p != to_release.end();
+ ++p) {
+ fl.l_start = p->second.start;
+ fl.l_len = p->second.length;
+ fl.l_pid = p->second.pid;
+ _do_filelock(in, NULL, p->first, CEPH_MDS_OP_SETFILELOCK, 0, &fl, p->second.owner);
+ }
+}
+
+void Client::_update_lock_state(struct flock *fl, uint64_t owner,
+ ceph_lock_state_t *lock_state)
+{
+ int lock_cmd;
+ if (F_RDLCK == fl->l_type)
+ lock_cmd = CEPH_LOCK_SHARED;
+ else if (F_WRLCK == fl->l_type)
+ lock_cmd = CEPH_LOCK_EXCL;
+ else
+ lock_cmd = CEPH_LOCK_UNLOCK;;
+
+ ceph_filelock filelock;
+ filelock.start = fl->l_start;
+ filelock.length = fl->l_len;
+ filelock.client = 0;
+ // see comment in _do_filelock()
+ filelock.owner = owner | (1ULL << 63);
+ filelock.pid = fl->l_pid;
+ filelock.type = lock_cmd;
+
+ if (filelock.type == CEPH_LOCK_UNLOCK) {
+ list<ceph_filelock> activated_locks;
+ lock_state->remove_lock(filelock, activated_locks);
+ } else {
+ bool r = lock_state->add_lock(filelock, false, false);
+ assert(r);
+ }
+}
+
+int Client::_getlk(Fh *fh, struct flock *fl, uint64_t owner)
+{
+ Inode *in = fh->inode;
+ ldout(cct, 10) << "_getlk " << fh << " ino " << in->ino << dendl;
+ int ret = _do_filelock(in, fh, CEPH_LOCK_FCNTL, CEPH_MDS_OP_GETFILELOCK, 0, fl, owner);
+ return ret;
+}
+
+int Client::_setlk(Fh *fh, struct flock *fl, uint64_t owner, int sleep, void *fuse_req)
+{
+ Inode *in = fh->inode;
+ ldout(cct, 10) << "_setlk " << fh << " ino " << in->ino << dendl;
+ int ret = _do_filelock(in, fh, CEPH_LOCK_FCNTL, CEPH_MDS_OP_SETFILELOCK, sleep, fl, owner, fuse_req);
+ ldout(cct, 10) << "_setlk " << fh << " ino " << in->ino << " result=" << ret << dendl;
+ return ret;
+}
+
+int Client::_flock(Fh *fh, int cmd, uint64_t owner, void *fuse_req)
+{
+ Inode *in = fh->inode;
+ ldout(cct, 10) << "_flock " << fh << " ino " << in->ino << dendl;
+
+ int sleep = !(cmd & LOCK_NB);
+ cmd &= ~LOCK_NB;
+
+ int type;
+ switch (cmd) {
+ case LOCK_SH:
+ type = F_RDLCK;
+ break;
+ case LOCK_EX:
+ type = F_WRLCK;
+ break;
+ case LOCK_UN:
+ type = F_UNLCK;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ struct flock fl;
+ memset(&fl, 0, sizeof(fl));
+ fl.l_type = type;
+ fl.l_whence = SEEK_SET;
+
+ int ret = _do_filelock(in, fh, CEPH_LOCK_FLOCK, CEPH_MDS_OP_SETFILELOCK, sleep, &fl, owner, fuse_req);
+ ldout(cct, 10) << "_flock " << fh << " ino " << in->ino << " result=" << ret << dendl;
+ return ret;
+}
+
int Client::ll_statfs(Inode *in, struct statvfs *stbuf)
{
/* Since the only thing this does is wrap a call to statfs, and
@@ -6887,6 +7188,16 @@
async_dentry_invalidator.start();
}
+void Client::ll_register_switch_interrupt_cb(client_switch_interrupt_callback_t cb)
+{
+ Mutex::Locker l(client_lock);
+ ldout(cct, 10) << "ll_register_switch_interrupt_cb cb " << (void*)cb << dendl;
+ if (cb == NULL)
+ return;
+ switch_interrupt_cb = cb;
+ interrupt_finisher.start();
+}
+
void Client::ll_register_getgroups_cb(client_getgroups_callback_t cb, void *handle)
{
Mutex::Locker l(client_lock);
@@ -8739,6 +9050,59 @@
return 0;
}
+int Client::ll_getlk(Fh *fh, struct flock *fl, uint64_t owner)
+{
+ Mutex::Locker lock(client_lock);
+
+ ldout(cct, 3) << "ll_getlk (fh)" << fh << " " << fh->inode->ino << dendl;
+ tout(cct) << "ll_getk (fh)" << (unsigned long)fh << std::endl;
+
+ return _getlk(fh, fl, owner);
+}
+
+int Client::ll_setlk(Fh *fh, struct flock *fl, uint64_t owner, int sleep, void *fuse_req)
+{
+ Mutex::Locker lock(client_lock);
+
+ ldout(cct, 3) << "ll_setlk (fh) " << fh << " " << fh->inode->ino << dendl;
+ tout(cct) << "ll_setk (fh)" << (unsigned long)fh << std::endl;
+
+ return _setlk(fh, fl, owner, sleep, fuse_req);
+}
+
+int Client::ll_flock(Fh *fh, int cmd, uint64_t owner, void *fuse_req)
+{
+ Mutex::Locker lock(client_lock);
+
+ ldout(cct, 3) << "ll_flock (fh) " << fh << " " << fh->inode->ino << dendl;
+ tout(cct) << "ll_flock (fh)" << (unsigned long)fh << std::endl;
+
+ return _flock(fh, cmd, owner, fuse_req);
+}
+
+class C_Client_RequestInterrupt : public Context {
+private:
+ Client *client;
+ MetaRequest *req;
+public:
+ C_Client_RequestInterrupt(Client *c, MetaRequest *r) : client(c), req(r) {
+ req->get();
+ }
+ void finish(int r) {
+ Mutex::Locker l(client->client_lock);
+ assert(req->head.op == CEPH_MDS_OP_SETFILELOCK);
+ client->_interrupt_filelock(req);
+ client->put_request(req);
+ }
+};
+
+void Client::ll_interrupt(void *d)
+{
+ MetaRequest *req = static_cast<MetaRequest*>(d);
+ ldout(cct, 3) << "ll_interrupt tid " << req->get_tid() << dendl;
+ tout(cct) << "ll_interrupt tid " << req->get_tid() << std::endl;
+ interrupt_finisher.queue(new C_Client_RequestInterrupt(this, req));
+}
// =========================================
// layout
diff -Nru ceph-0.80.7/src/client/Client.h ceph-0.80.9/src/client/Client.h
--- ceph-0.80.7/src/client/Client.h 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/client/Client.h 2015-03-09 20:51:17.000000000 +0100
@@ -119,6 +119,7 @@
struct MetaSession;
struct MetaRequest;
+class ceph_lock_state_t;
typedef void (*client_ino_callback_t)(void *handle, vinodeno_t ino, int64_t off, int64_t len);
@@ -127,6 +128,7 @@
vinodeno_t ino, string& name);
typedef int (*client_getgroups_callback_t)(void *handle, uid_t uid, gid_t **sgids);
+typedef void(*client_switch_interrupt_callback_t)(void *req, void *data);
// ========================================================
// client interface
@@ -214,6 +216,8 @@
SafeTimer timer;
+ client_switch_interrupt_callback_t switch_interrupt_cb;
+
client_ino_callback_t ino_invalidate_cb;
void *ino_invalidate_cb_handle;
@@ -225,6 +229,7 @@
Finisher async_ino_invalidator;
Finisher async_dentry_invalidator;
+ Finisher interrupt_finisher;
Context *tick_event;
utime_t last_cap_renew;
@@ -374,6 +379,7 @@
friend class C_Client_CacheInvalidate; // calls ino_invalidate_cb
friend class C_Client_DentryInvalidate; // calls dentry_invalidate_cb
friend class C_Block_Sync; // Calls block map and protected helpers
+ friend class C_Client_RequestInterrupt;
//int get_cache_size() { return lru.lru_get_size(); }
//void set_cache_size(int m) { lru.lru_set_max(m); }
@@ -604,6 +610,9 @@
int _fsync(Fh *fh, bool syncdataonly);
int _sync_fs();
int _fallocate(Fh *fh, int mode, int64_t offset, int64_t length);
+ int _getlk(Fh *fh, struct flock *fl, uint64_t owner);
+ int _setlk(Fh *fh, struct flock *fl, uint64_t owner, int sleep, void *fuse_req=NULL);
+ int _flock(Fh *fh, int cmd, uint64_t owner, void *fuse_req=NULL);
int get_or_create(Inode *dir, const char* name,
Dentry **pdn, bool expect_null=false);
@@ -613,6 +622,12 @@
vinodeno_t _get_vino(Inode *in);
inodeno_t _get_inodeno(Inode *in);
+ int _do_filelock(Inode *in, Fh *fh, int lock_type, int op, int sleep,
+ struct flock *fl, uint64_t owner, void *fuse_req=NULL);
+ int _interrupt_filelock(MetaRequest *req);
+ void _encode_filelocks(Inode *in, bufferlist& bl);
+ void _release_filelocks(Fh *fh);
+ void _update_lock_state(struct flock *fl, uint64_t owner, ceph_lock_state_t *lock_state);
public:
int mount(const std::string &mount_root);
void unmount();
@@ -818,6 +833,10 @@
int ll_fsync(Fh *fh, bool syncdataonly);
int ll_fallocate(Fh *fh, int mode, loff_t offset, loff_t length);
int ll_release(Fh *fh);
+ int ll_getlk(Fh *fh, struct flock *fl, uint64_t owner);
+ int ll_setlk(Fh *fh, struct flock *fl, uint64_t owner, int sleep, void *fuse_req);
+ int ll_flock(Fh *fh, int cmd, uint64_t owner, void *fuse_req);
+ void ll_interrupt(void *d);
int ll_get_stripe_osd(struct Inode *in, uint64_t blockno,
ceph_file_layout* layout);
uint64_t ll_get_internal_offset(struct Inode *in, uint64_t blockno);
@@ -825,11 +844,11 @@
int ll_num_osds(void);
int ll_osdaddr(int osd, uint32_t *addr);
int ll_osdaddr(int osd, char* buf, size_t size);
- void ll_register_ino_invalidate_cb(client_ino_callback_t cb, void *handle);
+ void ll_register_ino_invalidate_cb(client_ino_callback_t cb, void *handle);
void ll_register_dentry_invalidate_cb(client_dentry_callback_t cb, void *handle);
-
void ll_register_getgroups_cb(client_getgroups_callback_t cb, void *handle);
+ void ll_register_switch_interrupt_cb(client_switch_interrupt_callback_t cb);
};
#endif
diff -Nru ceph-0.80.7/src/client/Fh.h ceph-0.80.9/src/client/Fh.h
--- ceph-0.80.7/src/client/Fh.h 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/client/Fh.h 2015-03-09 20:51:17.000000000 +0100
@@ -5,6 +5,7 @@
class Inode;
class Cond;
+class ceph_lock_state_t;
// file handle for any open file state
@@ -23,8 +24,13 @@
loff_t consec_read_bytes;
int nr_consec_read;
+ // file lock
+ ceph_lock_state_t *fcntl_locks;
+ ceph_lock_state_t *flock_locks;
+
Fh() : inode(0), pos(0), mds(0), mode(0), flags(0), pos_locked(false),
- last_pos(0), consec_read_bytes(0), nr_consec_read(0) {}
+ last_pos(0), consec_read_bytes(0), nr_consec_read(0),
+ fcntl_locks(NULL), flock_locks(NULL) {}
};
diff -Nru ceph-0.80.7/src/client/fuse_ll.cc ceph-0.80.9/src/client/fuse_ll.cc
--- ceph-0.80.7/src/client/fuse_ll.cc 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/client/fuse_ll.cc 2015-03-09 20:51:17.000000000 +0100
@@ -636,6 +636,69 @@
cfuse->iput(in); // iput required
}
+static void fuse_ll_getlk(fuse_req_t req, fuse_ino_t ino,
+ struct fuse_file_info *fi, struct flock *lock)
+{
+ CephFuse::Handle *cfuse = (CephFuse::Handle *)fuse_req_userdata(req);
+ Fh *fh = (Fh*)fi->fh;
+
+ int r = cfuse->client->ll_getlk(fh, lock, fi->lock_owner);
+ if (r == 0)
+ fuse_reply_lock(req, lock);
+ else
+ fuse_reply_err(req, -r);
+}
+
+static void fuse_ll_setlk(fuse_req_t req, fuse_ino_t ino,
+ struct fuse_file_info *fi, struct flock *lock, int sleep)
+{
+ CephFuse::Handle *cfuse = (CephFuse::Handle *)fuse_req_userdata(req);
+ Fh *fh = (Fh*)fi->fh;
+
+ // must use multithread if operation may block
+ if (!cfuse->client->cct->_conf->fuse_multithreaded &&
+ sleep && lock->l_type != F_UNLCK) {
+ fuse_reply_err(req, EDEADLK);
+ return;
+ }
+
+ int r = cfuse->client->ll_setlk(fh, lock, fi->lock_owner, sleep, req);
+ fuse_reply_err(req, -r);
+}
+
+static void fuse_ll_interrupt(fuse_req_t req, void* data)
+{
+ CephFuse::Handle *cfuse = (CephFuse::Handle *)fuse_req_userdata(req);
+ cfuse->client->ll_interrupt(data);
+}
+
+static void switch_interrupt_cb(void *req, void* data)
+{
+ if (data)
+ fuse_req_interrupt_func((fuse_req_t)req, fuse_ll_interrupt, data);
+ else
+ fuse_req_interrupt_func((fuse_req_t)req, NULL, NULL);
+}
+
+#if FUSE_VERSION >= FUSE_MAKE_VERSION(2, 9)
+static void fuse_ll_flock(fuse_req_t req, fuse_ino_t ino,
+ struct fuse_file_info *fi, int cmd)
+{
+ CephFuse::Handle *cfuse = (CephFuse::Handle *)fuse_req_userdata(req);
+ Fh *fh = (Fh*)fi->fh;
+
+ // must use multithread if operation may block
+ if (!cfuse->client->cct->_conf->fuse_multithreaded &&
+ !(cmd & (LOCK_NB | LOCK_UN))) {
+ fuse_reply_err(req, EDEADLK);
+ return;
+ }
+
+ int r = cfuse->client->ll_flock(fh, cmd, fi->lock_owner, req);
+ fuse_reply_err(req, -r);
+}
+#endif
+
#if 0
static int getgroups_cb(void *handle, uid_t uid, gid_t **sgids)
{
@@ -742,8 +805,8 @@
removexattr: fuse_ll_removexattr,
access: fuse_ll_access,
create: fuse_ll_create,
- getlk: 0,
- setlk: 0,
+ getlk: fuse_ll_getlk,
+ setlk: fuse_ll_setlk,
bmap: 0,
#if FUSE_VERSION >= FUSE_MAKE_VERSION(2, 8)
#ifdef FUSE_IOCTL_COMPAT
@@ -752,13 +815,15 @@
ioctl: 0,
#endif
poll: 0,
-#if FUSE_VERSION > FUSE_MAKE_VERSION(2, 9)
+#endif
+#if FUSE_VERSION >= FUSE_MAKE_VERSION(2, 9)
write_buf: 0,
retrieve_reply: 0,
forget_multi: 0,
- flock: 0,
- fallocate: fuse_ll_fallocate
+ flock: fuse_ll_flock,
#endif
+#if FUSE_VERSION > FUSE_MAKE_VERSION(2, 9)
+ fallocate: fuse_ll_fallocate
#endif
};
@@ -859,6 +924,8 @@
fuse_session_add_chan(se, ch);
+ client->ll_register_switch_interrupt_cb(switch_interrupt_cb);
+
/*
* this is broken:
*
diff -Nru ceph-0.80.7/src/client/Inode.h ceph-0.80.9/src/client/Inode.h
--- ceph-0.80.7/src/client/Inode.h 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/client/Inode.h 2015-03-09 20:51:17.000000000 +0100
@@ -17,7 +17,8 @@
class Dentry;
class Dir;
struct SnapRealm;
-class Inode;
+struct Inode;
+class ceph_lock_state_t;
struct Cap {
MetaSession *session;
@@ -210,6 +211,10 @@
ll_ref -= n;
}
+ // file locks
+ ceph_lock_state_t *fcntl_locks;
+ ceph_lock_state_t *flock_locks;
+
Inode(CephContext *cct_, vinodeno_t vino, ceph_file_layout *newlayout)
: cct(cct_), ino(vino.ino), snapid(vino.snapid),
rdev(0), mode(0), uid(0), gid(0), nlink(0),
@@ -224,8 +229,8 @@
snaprealm(0), snaprealm_item(this), snapdir_parent(0),
oset((void *)this, newlayout->fl_pg_pool, ino),
reported_size(0), wanted_max_size(0), requested_max_size(0),
- _ref(0), ll_ref(0),
- dir(0), dn_set()
+ _ref(0), ll_ref(0), dir(0), dn_set(),
+ fcntl_locks(NULL), flock_locks(NULL)
{
memset(&dir_layout, 0, sizeof(dir_layout));
memset(&layout, 0, sizeof(layout));
diff -Nru ceph-0.80.7/src/client/MetaRequest.h ceph-0.80.9/src/client/MetaRequest.h
--- ceph-0.80.7/src/client/MetaRequest.h 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/client/MetaRequest.h 2015-03-09 20:51:17.000000000 +0100
@@ -9,6 +9,7 @@
#include "msg/msg_types.h"
#include "include/xlist.h"
#include "include/filepath.h"
+#include "include/atomic.h"
#include "mds/mdstypes.h"
#include "common/Mutex.h"
@@ -47,7 +48,7 @@
__u32 sent_on_mseq; // mseq at last submission of this request
int num_fwd; // # of times i've been forwarded
int retry_attempt;
- int ref;
+ atomic_t ref;
MClientReply *reply; // the reply
bool kick;
@@ -126,17 +127,14 @@
Dentry *old_dentry();
MetaRequest* get() {
- ++ref;
+ ref.inc();
return this;
}
/// psuedo-private put method; use Client::put_request()
- void _put() {
- if (--ref == 0)
- delete this;
- }
- int get_num_ref() {
- return ref;
+ bool _put() {
+ int v = ref.dec();
+ return v == 0;
}
// normal fields
diff -Nru ceph-0.80.7/src/common/ceph_crypto.h ceph-0.80.9/src/common/ceph_crypto.h
--- ceph-0.80.7/src/common/ceph_crypto.h 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/ceph_crypto.h 2015-03-09 20:51:17.000000000 +0100
@@ -78,9 +78,11 @@
assert(s == SECSuccess);
}
void Update (const byte *input, size_t length) {
- SECStatus s;
- s = PK11_DigestOp(ctx, input, length);
- assert(s == SECSuccess);
+ if (length) {
+ SECStatus s;
+ s = PK11_DigestOp(ctx, input, length);
+ assert(s == SECSuccess);
+ }
}
void Final (byte *digest) {
SECStatus s;
diff -Nru ceph-0.80.7/src/common/config.cc ceph-0.80.9/src/common/config.cc
--- ceph-0.80.7/src/common/config.cc 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/config.cc 2015-03-09 20:51:17.000000000 +0100
@@ -947,7 +947,7 @@
}
static const char *CONF_METAVARIABLES[] =
- { "cluster", "type", "name", "host", "num", "id", "pid" };
+ { "cluster", "type", "name", "host", "num", "id", "pid", "cctid" };
static const int NUM_CONF_METAVARIABLES =
(sizeof(CONF_METAVARIABLES) / sizeof(CONF_METAVARIABLES[0]));
@@ -1059,6 +1059,8 @@
out += name.get_id().c_str();
else if (var == "pid")
out += stringify(getpid());
+ else if (var == "cctid")
+ out += stringify((unsigned long long)this);
else
assert(0); // unreachable
expanded = true;
diff -Nru ceph-0.80.7/src/common/config_opts.h ceph-0.80.9/src/common/config_opts.h
--- ceph-0.80.7/src/common/config_opts.h 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/config_opts.h 2015-03-09 20:51:17.000000000 +0100
@@ -92,6 +92,7 @@
SUBSYS(heartbeatmap, 1, 5)
SUBSYS(perfcounter, 1, 5)
SUBSYS(rgw, 1, 5) // log level for the Rados gateway
+SUBSYS(civetweb, 1, 10)
SUBSYS(javaclient, 1, 5)
SUBSYS(asok, 1, 5)
SUBSYS(throttle, 1, 1)
@@ -283,7 +284,7 @@
OPTION(fuse_big_writes, OPT_BOOL, true)
OPTION(fuse_atomic_o_trunc, OPT_BOOL, true)
OPTION(fuse_debug, OPT_BOOL, false)
-OPTION(fuse_multithreaded, OPT_BOOL, false)
+OPTION(fuse_multithreaded, OPT_BOOL, true)
OPTION(crush_location, OPT_STR, "") // whitespace-separated list of key=value pairs describing crush location
@@ -489,6 +490,9 @@
OPTION(osd_heartbeat_grace, OPT_INT, 20) // (seconds) how long before we decide a peer has failed
OPTION(osd_heartbeat_min_peers, OPT_INT, 10) // minimum number of peers
+// max number of parallel snap trims/pg
+OPTION(osd_pg_max_concurrent_snap_trims, OPT_U64, 2)
+
// minimum number of peers tha tmust be reachable to mark ourselves
// back up after being wrongly marked down.
OPTION(osd_heartbeat_min_healthy_ratio, OPT_FLOAT, .33)
@@ -766,6 +770,7 @@
OPTION(rgw_socket_path, OPT_STR, "") // path to unix domain socket, if not specified, rgw will not run as external fcgi
OPTION(rgw_host, OPT_STR, "") // host for radosgw, can be an IP, default is 0.0.0.0
OPTION(rgw_port, OPT_STR, "") // port to listen, format as "8080" "5000", if not specified, rgw will not run external fcgi
+OPTION(rgw_fcgi_explicit_free, OPT_BOOL, true) // whether to call FCGX_Free explicitly on every complete request
OPTION(rgw_dns_name, OPT_STR, "")
OPTION(rgw_script_uri, OPT_STR, "") // alternative value for SCRIPT_URI if not set in request
OPTION(rgw_request_uri, OPT_STR, "") // alternative value for REQUEST_URI if not set in request
diff -Nru ceph-0.80.7/src/common/crc32c_intel_fast_asm.S ceph-0.80.9/src/common/crc32c_intel_fast_asm.S
--- ceph-0.80.7/src/common/crc32c_intel_fast_asm.S 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/common/crc32c_intel_fast_asm.S 2015-03-09 20:51:17.000000000 +0100
@@ -662,3 +662,5 @@
%endmacro
;;; func core, ver, snum
slversion crc32_iscsi_00, 00, 02, 0014
+; inform linker that this doesn't require executable stack
+section .note.GNU-stack noalloc noexec nowrite progbits
diff -Nru ceph-0.80.7/src/common/crc32c_intel_fast_zero_asm.S ceph-0.80.9/src/common/crc32c_intel_fast_zero_asm.S
--- ceph-0.80.7/src/common/crc32c_intel_fast_zero_asm.S 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/common/crc32c_intel_fast_zero_asm.S 2015-03-09 20:51:17.000000000 +0100
@@ -644,3 +644,5 @@
%endmacro
;;; func core, ver, snum
slversion crc32_iscsi_zero_00, 00, 02, 0014
+; inform linker that this doesn't require executable stack
+section .note.GNU-stack noalloc noexec nowrite progbits
diff -Nru ceph-0.80.7/src/common/hobject.cc ceph-0.80.9/src/common/hobject.cc
--- ceph-0.80.7/src/common/hobject.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/common/hobject.cc 2015-03-09 20:51:17.000000000 +0100
@@ -238,10 +238,10 @@
void ghobject_t::dump(Formatter *f) const
{
hobj.dump(f);
- if (generation != NO_GEN) {
+ if (generation != NO_GEN)
f->dump_int("generation", generation);
+ if (shard_id != ghobject_t::NO_SHARD)
f->dump_int("shard_id", shard_id);
- }
}
void ghobject_t::generate_test_instances(list<ghobject_t*>& o)
diff -Nru ceph-0.80.7/src/common/io_priority.cc ceph-0.80.9/src/common/io_priority.cc
--- ceph-0.80.7/src/common/io_priority.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/common/io_priority.cc 2015-03-09 20:51:17.000000000 +0100
@@ -41,8 +41,8 @@
int ceph_ioprio_string_to_class(const std::string& s)
{
- std::string l;
- std::transform(s.begin(), s.end(), l.begin(), ::tolower);
+ std::string l = s;
+ std::transform(l.begin(), l.end(), l.begin(), ::tolower);
if (l == "idle")
return IOPRIO_CLASS_IDLE;
diff -Nru ceph-0.80.7/src/common/Makefile.am ceph-0.80.9/src/common/Makefile.am
--- ceph-0.80.7/src/common/Makefile.am 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/Makefile.am 2015-03-09 20:51:17.000000000 +0100
@@ -83,7 +83,8 @@
osd/HitSet.cc \
mds/MDSMap.cc \
mds/inode_backtrace.cc \
- mds/mdstypes.cc
+ mds/mdstypes.cc \
+ mds/flock.cc
# inject crc in common
libcommon_crc_la_SOURCES = \
diff -Nru ceph-0.80.7/src/common/Thread.cc ceph-0.80.9/src/common/Thread.cc
--- ceph-0.80.7/src/common/Thread.cc 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/Thread.cc 2015-03-09 20:51:17.000000000 +0100
@@ -51,7 +51,8 @@
int p = ceph_gettid(); // may return -ENOSYS on other platforms
if (p > 0)
pid = p;
- if (ioprio_class >= 0 &&
+ if (pid &&
+ ioprio_class >= 0 &&
ioprio_priority >= 0) {
ceph_ioprio_set(IOPRIO_WHO_PROCESS,
pid,
diff -Nru ceph-0.80.7/src/common/Thread.h ceph-0.80.9/src/common/Thread.h
--- ceph-0.80.7/src/common/Thread.h 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/Thread.h 2015-03-09 20:51:17.000000000 +0100
@@ -41,6 +41,7 @@
public:
const pthread_t &get_thread_id();
+ pid_t get_pid() const { return pid; }
bool is_started();
bool am_self();
int kill(int signal);
diff -Nru ceph-0.80.7/src/common/TrackedOp.cc ceph-0.80.9/src/common/TrackedOp.cc
--- ceph-0.80.7/src/common/TrackedOp.cc 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/TrackedOp.cc 2015-03-09 20:51:17.000000000 +0100
@@ -121,10 +121,10 @@
// caller checks;
assert(tracking_enabled);
+ Mutex::Locker locker(ops_in_flight_lock);
i->request->clear_data();
i->request->clear_payload();
- Mutex::Locker locker(ops_in_flight_lock);
assert(i->xitem.get_list() == &ops_in_flight);
utime_t now = ceph_clock_now(cct);
i->xitem.remove_myself();
diff -Nru ceph-0.80.7/src/common/util.cc ceph-0.80.9/src/common/util.cc
--- ceph-0.80.7/src/common/util.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/common/util.cc 2015-03-09 20:51:17.000000000 +0100
@@ -18,6 +18,10 @@
#include "common/errno.h"
#include "common/strtol.h"
+#ifdef HAVE_SYS_VFS_H
+#include <sys/vfs.h>
+#endif
+
// test if an entire buf is zero in 8-byte chunks
bool buf_is_zero(const char *buf, size_t len)
{
@@ -104,3 +108,21 @@
}
return (r * (1LL << modifier));
}
+
+int get_fs_stats(ceph_data_stats_t &stats, const char *path)
+{
+ if (!path)
+ return -EINVAL;
+
+ struct statfs stbuf;
+ int err = ::statfs(path, &stbuf);
+ if (err < 0) {
+ return -errno;
+ }
+
+ stats.byte_total = stbuf.f_blocks * stbuf.f_bsize;
+ stats.byte_used = (stbuf.f_blocks - stbuf.f_bfree) * stbuf.f_bsize;
+ stats.byte_avail = stbuf.f_bavail * stbuf.f_bsize;
+ stats.avail_percent = (((float)stats.byte_avail/stats.byte_total)*100);
+ return 0;
+}
diff -Nru ceph-0.80.7/src/common/WorkQueue.cc ceph-0.80.9/src/common/WorkQueue.cc
--- ceph-0.80.7/src/common/WorkQueue.cc 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/common/WorkQueue.cc 2015-03-09 20:51:17.000000000 +0100
@@ -271,6 +271,10 @@
for (set<WorkThread*>::iterator p = _threads.begin();
p != _threads.end();
++p) {
+ ldout(cct,10) << __func__
+ << " class " << cls << " priority " << priority
+ << " pid " << (*p)->get_pid()
+ << dendl;
int r = (*p)->set_ioprio(cls, priority);
if (r < 0)
lderr(cct) << " set_ioprio got " << cpp_strerror(r) << dendl;
diff -Nru ceph-0.80.7/src/crush/builder.c ceph-0.80.9/src/crush/builder.c
--- ceph-0.80.7/src/crush/builder.c 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/crush/builder.c 2015-03-09 20:51:17.000000000 +0100
@@ -11,6 +11,8 @@
#include "builder.h"
#include "hash.h"
+#define dprintk(args...) /* printf(args) */
+
#define BUG_ON(x) assert(!(x))
struct crush_map *crush_create()
@@ -27,6 +29,7 @@
m->choose_total_tries = 19;
m->chooseleaf_descend_once = 0;
m->chooseleaf_vary_r = 0;
+ m->straw_calc_version = 0;
return m;
}
@@ -63,7 +66,7 @@
for (r=0; r < map->max_rules; r++)
if (map->rules[r] == 0)
break;
- assert(r <= INT_MAX);
+ assert(r < CRUSH_MAX_RULES);
}
else
r = ruleno;
@@ -72,6 +75,8 @@
/* expand array */
int oldsize;
void *_realloc = NULL;
+ if (map->max_rules +1 > CRUSH_MAX_RULES)
+ return -ENOSPC;
oldsize = map->max_rules;
map->max_rules = r+1;
if ((_realloc = realloc(map->rules, map->max_rules * sizeof(map->rules[0]))) == NULL) {
@@ -263,7 +268,7 @@
w += weights[i];
bucket->sum_weights[i] = w;
- /*printf("pos %d item %d weight %d sum %d\n",
+ /*dprintk("pos %d item %d weight %d sum %d\n",
i, items[i], weights[i], bucket->sum_weights[i]);*/
}
@@ -304,6 +309,10 @@
static int calc_depth(int size)
{
+ if (size == 0) {
+ return 0;
+ }
+
int depth = 1;
int t = size - 1;
while (t) {
@@ -332,6 +341,16 @@
bucket->h.type = type;
bucket->h.size = size;
+ if (size == 0) {
+ bucket->h.items = NULL;
+ bucket->h.perm = NULL;
+ bucket->h.weight = 0;
+ bucket->node_weights = NULL;
+ bucket->num_nodes = 0;
+ /* printf("size 0 depth 0 nodes 0\n"); */
+ return bucket;
+ }
+
bucket->h.items = malloc(sizeof(__s32)*size);
if (!bucket->h.items)
goto err;
@@ -342,7 +361,7 @@
/* calc tree depth */
depth = calc_depth(size);
bucket->num_nodes = 1 << depth;
- printf("size %d depth %d nodes %d\n", size, depth, bucket->num_nodes);
+ dprintk("size %d depth %d nodes %d\n", size, depth, bucket->num_nodes);
bucket->node_weights = malloc(sizeof(__u32)*bucket->num_nodes);
if (!bucket->node_weights)
@@ -354,7 +373,7 @@
for (i=0; i<size; i++) {
bucket->h.items[i] = items[i];
node = crush_calc_tree_node(i);
- printf("item %d node %d weight %d\n", i, node, weights[i]);
+ dprintk("item %d node %d weight %d\n", i, node, weights[i]);
bucket->node_weights[node] = weights[i];
if (crush_addition_is_unsafe(bucket->h.weight, weights[i]))
@@ -368,7 +387,7 @@
goto err;
bucket->node_weights[node] += weights[i];
- printf(" node %d weight %d\n", node, bucket->node_weights[node]);
+ dprintk(" node %d weight %d\n", node, bucket->node_weights[node]);
}
}
BUG_ON(bucket->node_weights[bucket->num_nodes/2] != bucket->h.weight);
@@ -386,7 +405,34 @@
/* straw bucket */
-int crush_calc_straw(struct crush_bucket_straw *bucket)
+/*
+ * this code was written 8 years ago. i have a vague recollection of
+ * drawing boxes underneath bars of different lengths, where the bar
+ * length represented the probability/weight, and that there was some
+ * trial and error involved in arriving at this implementation.
+ * however, reading the code now after all this time, the intuition
+ * that motivated is lost on me. lame. my only excuse is that I now
+ * know that the approach is fundamentally flawed and am not
+ * particularly motivated to reconstruct the flawed reasoning.
+ *
+ * as best as i can remember, the idea is: sort the weights, and start
+ * with the smallest. arbitrarily scale it at 1.0 (16-bit fixed
+ * point). look at the next larger weight, and calculate the scaling
+ * factor for that straw based on the relative difference in weight so
+ * far. what's not clear to me now is why we are looking at wnext
+ * (the delta to the next bigger weight) for all remaining weights,
+ * and slicing things horizontally instead of considering just the
+ * next item or set of items. or why pow() is used the way it is.
+ *
+ * note that the original version 1 of this function made special
+ * accomodation for the case where straw lengths were identical. this
+ * is also flawed in a non-obvious way; version 2 drops the special
+ * handling and appears to work just as well.
+ *
+ * moral of the story: if you do something clever, write down why it
+ * works.
+ */
+int crush_calc_straw(struct crush_map *map, struct crush_bucket_straw *bucket)
{
int *reverse;
int i, j, k;
@@ -422,41 +468,82 @@
i=0;
while (i < size) {
- /* zero weight items get 0 length straws! */
- if (weights[reverse[i]] == 0) {
- bucket->straws[reverse[i]] = 0;
+ if (map->straw_calc_version == 0) {
+ /* zero weight items get 0 length straws! */
+ if (weights[reverse[i]] == 0) {
+ bucket->straws[reverse[i]] = 0;
+ i++;
+ continue;
+ }
+
+ /* set this item's straw */
+ bucket->straws[reverse[i]] = straw * 0x10000;
+ dprintk("item %d at %d weight %d straw %d (%lf)\n",
+ bucket->h.items[reverse[i]],
+ reverse[i], weights[reverse[i]],
+ bucket->straws[reverse[i]], straw);
i++;
- continue;
- }
+ if (i == size)
+ break;
- /* set this item's straw */
- bucket->straws[reverse[i]] = straw * 0x10000;
- /*printf("item %d at %d weight %d straw %d (%lf)\n",
- items[reverse[i]],
- reverse[i], weights[reverse[i]], bucket->straws[reverse[i]], straw);*/
- i++;
- if (i == size) break;
-
- /* same weight as previous? */
- if (weights[reverse[i]] == weights[reverse[i-1]]) {
- /*printf("same as previous\n");*/
- continue;
- }
+ /* same weight as previous? */
+ if (weights[reverse[i]] == weights[reverse[i-1]]) {
+ dprintk("same as previous\n");
+ continue;
+ }
- /* adjust straw for next guy */
- wbelow += ((double)weights[reverse[i-1]] - lastw) * numleft;
- for (j=i; j<size; j++)
- if (weights[reverse[j]] == weights[reverse[i]])
+ /* adjust straw for next guy */
+ wbelow += ((double)weights[reverse[i-1]] - lastw) *
+ numleft;
+ for (j=i; j<size; j++)
+ if (weights[reverse[j]] == weights[reverse[i]])
+ numleft--;
+ else
+ break;
+ wnext = numleft * (weights[reverse[i]] -
+ weights[reverse[i-1]]);
+ pbelow = wbelow / (wbelow + wnext);
+ dprintk("wbelow %lf wnext %lf pbelow %lf numleft %d\n",
+ wbelow, wnext, pbelow, numleft);
+
+ straw *= pow((double)1.0 / pbelow, (double)1.0 /
+ (double)numleft);
+
+ lastw = weights[reverse[i-1]];
+ } else if (map->straw_calc_version >= 1) {
+ /* zero weight items get 0 length straws! */
+ if (weights[reverse[i]] == 0) {
+ bucket->straws[reverse[i]] = 0;
+ i++;
numleft--;
- else
+ continue;
+ }
+
+ /* set this item's straw */
+ bucket->straws[reverse[i]] = straw * 0x10000;
+ dprintk("item %d at %d weight %d straw %d (%lf)\n",
+ bucket->h.items[reverse[i]],
+ reverse[i], weights[reverse[i]],
+ bucket->straws[reverse[i]], straw);
+ i++;
+ if (i == size)
break;
- wnext = numleft * (weights[reverse[i]] - weights[reverse[i-1]]);
- pbelow = wbelow / (wbelow + wnext);
- /*printf("wbelow %lf wnext %lf pbelow %lf\n", wbelow, wnext, pbelow);*/
- straw *= pow((double)1.0 / pbelow, (double)1.0 / (double)numleft);
+ /* adjust straw for next guy */
+ wbelow += ((double)weights[reverse[i-1]] - lastw) *
+ numleft;
+ numleft--;
+ wnext = numleft * (weights[reverse[i]] -
+ weights[reverse[i-1]]);
+ pbelow = wbelow / (wbelow + wnext);
+ dprintk("wbelow %lf wnext %lf pbelow %lf numleft %d\n",
+ wbelow, wnext, pbelow, numleft);
+
+ straw *= pow((double)1.0 / pbelow, (double)1.0 /
+ (double)numleft);
- lastw = weights[reverse[i-1]];
+ lastw = weights[reverse[i-1]];
+ }
}
free(reverse);
@@ -464,7 +551,8 @@
}
struct crush_bucket_straw *
-crush_make_straw_bucket(int hash,
+crush_make_straw_bucket(struct crush_map *map,
+ int hash,
int type,
int size,
int *items,
@@ -502,7 +590,7 @@
bucket->item_weights[i] = weights[i];
}
- if (crush_calc_straw(bucket) < 0)
+ if (crush_calc_straw(map, bucket) < 0)
goto err;
return bucket;
@@ -518,7 +606,8 @@
struct crush_bucket*
-crush_make_bucket(int alg, int hash, int type, int size,
+crush_make_bucket(struct crush_map *map,
+ int alg, int hash, int type, int size,
int *items,
int *weights)
{
@@ -539,7 +628,7 @@
return (struct crush_bucket *)crush_make_tree_bucket(hash, type, size, items, weights);
case CRUSH_BUCKET_STRAW:
- return (struct crush_bucket *)crush_make_straw_bucket(hash, type, size, items, weights);
+ return (struct crush_bucket *)crush_make_straw_bucket(map, hash, type, size, items, weights);
}
return 0;
}
@@ -648,27 +737,39 @@
node = crush_calc_tree_node(newsize-1);
bucket->node_weights[node] = weight;
+ /* if the depth increase, we need to initialize the new root node's weight before add bucket item */
+ int root = bucket->num_nodes/2;
+ if (depth >= 2 && (node - 1) == root) {
+ /* if the new item is the first node in right sub tree, so
+ * the root node initial weight is left sub tree's weight
+ */
+ bucket->node_weights[root] = bucket->node_weights[root/2];
+ }
+
for (j=1; j<depth; j++) {
node = parent(node);
- if (!crush_addition_is_unsafe(bucket->node_weights[node], weight))
+ if (crush_addition_is_unsafe(bucket->node_weights[node], weight))
return -ERANGE;
bucket->node_weights[node] += weight;
- printf(" node %d weight %d\n", node, bucket->node_weights[node]);
+ dprintk(" node %d weight %d\n", node, bucket->node_weights[node]);
}
if (crush_addition_is_unsafe(bucket->h.weight, weight))
return -ERANGE;
+ bucket->h.items[newsize-1] = item;
bucket->h.weight += weight;
bucket->h.size++;
return 0;
}
-int crush_add_straw_bucket_item(struct crush_bucket_straw *bucket, int item, int weight)
+int crush_add_straw_bucket_item(struct crush_map *map,
+ struct crush_bucket_straw *bucket,
+ int item, int weight)
{
int newsize = bucket->h.size + 1;
@@ -701,13 +802,14 @@
if (crush_addition_is_unsafe(bucket->h.weight, weight))
return -ERANGE;
- bucket->h.weight += weight;
- bucket->h.size++;
+ bucket->h.weight += weight;
+ bucket->h.size++;
- return crush_calc_straw(bucket);
+ return crush_calc_straw(map, bucket);
}
-int crush_bucket_add_item(struct crush_bucket *b, int item, int weight)
+int crush_bucket_add_item(struct crush_map *map,
+ struct crush_bucket *b, int item, int weight)
{
/* invalidate perm cache */
b->perm_n = 0;
@@ -720,7 +822,7 @@
case CRUSH_BUCKET_TREE:
return crush_add_tree_bucket_item((struct crush_bucket_tree *)b, item, weight);
case CRUSH_BUCKET_STRAW:
- return crush_add_straw_bucket_item((struct crush_bucket_straw *)b, item, weight);
+ return crush_add_straw_bucket_item(map, (struct crush_bucket_straw *)b, item, weight);
default:
return -1;
}
@@ -744,7 +846,10 @@
for (j = i; j < bucket->h.size; j++)
bucket->h.items[j] = bucket->h.items[j+1];
newsize = --bucket->h.size;
- bucket->h.weight -= bucket->item_weight;
+ if (bucket->item_weight < bucket->h.weight)
+ bucket->h.weight -= bucket->item_weight;
+ else
+ bucket->h.weight = 0;
if ((_realloc = realloc(bucket->h.items, sizeof(__s32)*newsize)) == NULL) {
return -ENOMEM;
@@ -763,7 +868,7 @@
{
unsigned i, j;
int newsize;
- int weight;
+ unsigned weight;
for (i = 0; i < bucket->h.size; i++)
if (bucket->h.items[i] == item)
@@ -777,7 +882,10 @@
bucket->item_weights[j] = bucket->item_weights[j+1];
bucket->sum_weights[j] = bucket->sum_weights[j+1] - weight;
}
- bucket->h.weight -= weight;
+ if (weight < bucket->h.weight)
+ bucket->h.weight -= weight;
+ else
+ bucket->h.weight = 0;
newsize = --bucket->h.size;
void *_realloc = NULL;
@@ -812,7 +920,7 @@
for (i = 0; i < bucket->h.size; i++) {
int node;
- int weight;
+ unsigned weight;
int j;
int depth = calc_depth(bucket->h.size);
@@ -826,9 +934,12 @@
for (j = 1; j < depth; j++) {
node = parent(node);
bucket->node_weights[node] -= weight;
- printf(" node %d weight %d\n", node, bucket->node_weights[node]);
+ dprintk(" node %d weight %d\n", node, bucket->node_weights[node]);
}
- bucket->h.weight -= weight;
+ if (weight < bucket->h.weight)
+ bucket->h.weight -= weight;
+ else
+ bucket->h.weight = 0;
break;
}
if (i == bucket->h.size)
@@ -875,7 +986,8 @@
return 0;
}
-int crush_remove_straw_bucket_item(struct crush_bucket_straw *bucket, int item)
+int crush_remove_straw_bucket_item(struct crush_map *map,
+ struct crush_bucket_straw *bucket, int item)
{
int newsize = bucket->h.size - 1;
unsigned i, j;
@@ -883,7 +995,10 @@
for (i = 0; i < bucket->h.size; i++) {
if (bucket->h.items[i] == item) {
bucket->h.size--;
- bucket->h.weight -= bucket->item_weights[i];
+ if (bucket->item_weights[i] < bucket->h.weight)
+ bucket->h.weight -= bucket->item_weights[i];
+ else
+ bucket->h.weight = 0;
for (j = i; j < bucket->h.size; j++) {
bucket->h.items[j] = bucket->h.items[j+1];
bucket->item_weights[j] = bucket->item_weights[j+1];
@@ -917,10 +1032,10 @@
bucket->straws = _realloc;
}
- return crush_calc_straw(bucket);
+ return crush_calc_straw(map, bucket);
}
-int crush_bucket_remove_item(struct crush_bucket *b, int item)
+int crush_bucket_remove_item(struct crush_map *map, struct crush_bucket *b, int item)
{
/* invalidate perm cache */
b->perm_n = 0;
@@ -933,7 +1048,7 @@
case CRUSH_BUCKET_TREE:
return crush_remove_tree_bucket_item((struct crush_bucket_tree *)b, item);
case CRUSH_BUCKET_STRAW:
- return crush_remove_straw_bucket_item((struct crush_bucket_straw *)b, item);
+ return crush_remove_straw_bucket_item(map, (struct crush_bucket_straw *)b, item);
default:
return -1;
}
@@ -1002,7 +1117,9 @@
return diff;
}
-int crush_adjust_straw_bucket_item_weight(struct crush_bucket_straw *bucket, int item, int weight)
+int crush_adjust_straw_bucket_item_weight(struct crush_map *map,
+ struct crush_bucket_straw *bucket,
+ int item, int weight)
{
unsigned idx;
int diff;
@@ -1018,14 +1135,16 @@
bucket->item_weights[idx] = weight;
bucket->h.weight += diff;
- r = crush_calc_straw(bucket);
+ r = crush_calc_straw(map, bucket);
if (r < 0)
return r;
return diff;
}
-int crush_bucket_adjust_item_weight(struct crush_bucket *b, int item, int weight)
+int crush_bucket_adjust_item_weight(struct crush_map *map,
+ struct crush_bucket *b,
+ int item, int weight)
{
switch (b->alg) {
case CRUSH_BUCKET_UNIFORM:
@@ -1038,7 +1157,8 @@
return crush_adjust_tree_bucket_item_weight((struct crush_bucket_tree *)b,
item, weight);
case CRUSH_BUCKET_STRAW:
- return crush_adjust_straw_bucket_item_weight((struct crush_bucket_straw *)b,
+ return crush_adjust_straw_bucket_item_weight(map,
+ (struct crush_bucket_straw *)b,
item, weight);
default:
return -1;
@@ -1141,6 +1261,7 @@
bucket->h.weight += bucket->item_weights[i];
}
+ crush_calc_straw(crush, bucket);
return 0;
}
diff -Nru ceph-0.80.7/src/crush/builder.h ceph-0.80.9/src/crush/builder.h
--- ceph-0.80.7/src/crush/builder.h 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/crush/builder.h 2015-03-09 20:51:17.000000000 +0100
@@ -16,12 +16,12 @@
extern int crush_add_bucket(struct crush_map *map,
int bucketno,
struct crush_bucket *bucket, int *idout);
-struct crush_bucket *crush_make_bucket(int alg, int hash, int type, int size, int *items, int *weights);
-extern int crush_bucket_add_item(struct crush_bucket *bucket, int item, int weight);
-extern int crush_bucket_adjust_item_weight(struct crush_bucket *bucket, int item, int weight);
+struct crush_bucket *crush_make_bucket(struct crush_map *map, int alg, int hash, int type, int size, int *items, int *weights);
+extern int crush_bucket_add_item(struct crush_map *map, struct crush_bucket *bucket, int item, int weight);
+extern int crush_bucket_adjust_item_weight(struct crush_map *map, struct crush_bucket *bucket, int item, int weight);
extern int crush_reweight_bucket(struct crush_map *crush, struct crush_bucket *bucket);
extern int crush_remove_bucket(struct crush_map *map, struct crush_bucket *bucket);
-extern int crush_bucket_remove_item(struct crush_bucket *bucket, int item);
+extern int crush_bucket_remove_item(struct crush_map *map, struct crush_bucket *bucket, int item);
struct crush_bucket_uniform *
crush_make_uniform_bucket(int hash, int type, int size,
@@ -36,7 +36,8 @@
int *items, /* in leaf order */
int *weights);
struct crush_bucket_straw *
-crush_make_straw_bucket(int hash, int type, int size,
+crush_make_straw_bucket(struct crush_map *map,
+ int hash, int type, int size,
int *items,
int *weights);
diff -Nru ceph-0.80.7/src/crush/CrushCompiler.cc ceph-0.80.9/src/crush/CrushCompiler.cc
--- ceph-0.80.7/src/crush/CrushCompiler.cc 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/crush/CrushCompiler.cc 2015-03-09 20:51:17.000000000 +0100
@@ -191,6 +191,8 @@
out << "tunable chooseleaf_descend_once " << crush.get_chooseleaf_descend_once() << "\n";
if (crush.get_chooseleaf_vary_r() != 0)
out << "tunable chooseleaf_vary_r " << crush.get_chooseleaf_vary_r() << "\n";
+ if (crush.get_straw_calc_version() != 0)
+ out << "tunable straw_calc_version " << crush.get_straw_calc_version() << "\n";
out << "\n# devices\n";
for (int i=0; i<crush.get_max_devices(); i++) {
@@ -368,6 +370,8 @@
crush.set_chooseleaf_descend_once(val);
else if (name == "chooseleaf_vary_r")
crush.set_chooseleaf_vary_r(val);
+ else if (name == "straw_calc_version")
+ crush.set_straw_calc_version(val);
else {
err << "tunable " << name << " not recognized" << std::endl;
return -1;
diff -Nru ceph-0.80.7/src/crush/crush.h ceph-0.80.9/src/crush/crush.h
--- ceph-0.80.7/src/crush/crush.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/crush/crush.h 2015-03-09 20:51:17.000000000 +0100
@@ -26,6 +26,8 @@
#define CRUSH_MAGIC 0x00010000ul /* for detecting algorithm revisions */
#define CRUSH_MAX_DEPTH 10 /* max crush hierarchy depth */
+#define CRUSH_MAX_RULESET (1<<8) /*max crush ruleset number*/
+#define CRUSH_MAX_RULES CRUSH_MAX_RULESET /*max crush rules, shold be the same as max rulesets*/
#define CRUSH_MAX_DEVICE_WEIGHT (100u * 0x10000u)
#define CRUSH_MAX_BUCKET_WEIGHT (65535u * 0x10000u)
@@ -189,6 +191,12 @@
* mappings line up a bit better with previous mappings. */
__u8 chooseleaf_vary_r;
+ /*
+ * version 0 (original) of straw_calc has various flaws. version 1
+ * fixes a few of them.
+ */
+ __u8 straw_calc_version;
+
__u32 *choose_tries;
};
diff -Nru ceph-0.80.7/src/crush/CrushTester.cc ceph-0.80.9/src/crush/CrushTester.cc
--- ceph-0.80.7/src/crush/CrushTester.cc 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/crush/CrushTester.cc 2015-03-09 20:51:17.000000000 +0100
@@ -487,18 +487,18 @@
vector<int> out;
if (use_crush) {
- if (output_statistics)
- err << "CRUSH"; // prepend CRUSH to placement output
+ if (output_mappings)
+ err << "CRUSH"; // prepend CRUSH to placement output
crush.do_rule(r, x, out, nr, weight);
} else {
- if (output_statistics)
- err << "RNG"; // prepend RNG to placement output to denote simulation
+ if (output_mappings)
+ err << "RNG"; // prepend RNG to placement output to denote simulation
// test our new monte carlo placement generator
random_placement(r, out, nr, weight);
}
- if (output_statistics)
- err << " rule " << r << " x " << x << " " << out << std::endl;
+ if (output_mappings)
+ err << " rule " << r << " x " << x << " " << out << std::endl;
if (output_data_file)
write_integer_indexed_vector_data_string(tester_data.placement_information, x, out);
@@ -539,14 +539,14 @@
if (output_statistics)
for (unsigned i = 0; i < per.size(); i++) {
- if (output_utilization && num_batches > 1){
+ if (output_utilization) {
if (num_objects_expected[i] > 0 && per[i] > 0) {
err << " device " << i << ":\t"
<< "\t" << " stored " << ": " << per[i]
<< "\t" << " expected " << ": " << num_objects_expected[i]
<< std::endl;
}
- } else if (output_utilization_all && num_batches > 1) {
+ } else if (output_utilization_all) {
err << " device " << i << ":\t"
<< "\t" << " stored " << ": " << per[i]
<< "\t" << " expected " << ": " << num_objects_expected[i]
diff -Nru ceph-0.80.7/src/crush/CrushTester.h ceph-0.80.9/src/crush/CrushTester.h
--- ceph-0.80.7/src/crush/CrushTester.h 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/crush/CrushTester.h 2015-03-09 20:51:17.000000000 +0100
@@ -27,6 +27,7 @@
bool output_utilization;
bool output_utilization_all;
bool output_statistics;
+ bool output_mappings;
bool output_bad_mappings;
bool output_choose_tries;
@@ -176,6 +177,7 @@
output_utilization(false),
output_utilization_all(false),
output_statistics(false),
+ output_mappings(false),
output_bad_mappings(false),
output_choose_tries(false),
output_data_file(false),
@@ -226,6 +228,13 @@
return output_statistics;
}
+ void set_output_mappings(bool b) {
+ output_mappings = b;
+ }
+ bool get_output_mappings() const {
+ return output_mappings;
+ }
+
void set_output_bad_mappings(bool b) {
output_bad_mappings = b;
}
diff -Nru ceph-0.80.7/src/crush/CrushWrapper.cc ceph-0.80.9/src/crush/CrushWrapper.cc
--- ceph-0.80.7/src/crush/CrushWrapper.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/crush/CrushWrapper.cc 2015-03-09 20:51:17.000000000 +0100
@@ -150,10 +150,10 @@
for (unsigned i=0; i<b->size; ++i) {
int id = b->items[i];
if (id == item) {
- adjust_item_weight(cct, item, 0);
ldout(cct, 5) << "remove_item removing item " << item
<< " from bucket " << b->id << dendl;
- crush_bucket_remove_item(b, item);
+ crush_bucket_remove_item(crush, b, item);
+ adjust_item_weight(cct, b->id, b->weight);
ret = 0;
}
}
@@ -171,8 +171,8 @@
if (!crush->buckets[i])
continue;
crush_bucket *b = crush->buckets[i];
- for (unsigned i=0; i<b->size; ++i) {
- if (b->items[i] == item)
+ for (unsigned j=0; j<b->size; ++j) {
+ if (b->items[j] == item)
return true;
}
}
@@ -197,9 +197,9 @@
for (unsigned i=0; i<b->size; ++i) {
int id = b->items[i];
if (id == item) {
- adjust_item_weight(cct, item, 0);
ldout(cct, 5) << "_remove_item_under removing item " << item << " from bucket " << b->id << dendl;
- crush_bucket_remove_item(b, item);
+ crush_bucket_remove_item(crush, b, item);
+ adjust_item_weight(cct, b->id, b->weight);
ret = 0;
} else if (id < 0) {
int r = remove_item_under(cct, item, id, unlink_only);
@@ -459,6 +459,8 @@
int cur = item;
+ // create locations if locations don't exist and add child in location with 0 weight
+ // the more detail in the insert_item method declaration in CrushWrapper.h
for (map<int,string>::iterator p = type_map.begin(); p != type_map.end(); ++p) {
// ignore device type
if (p->first == 0)
@@ -518,17 +520,17 @@
ldout(cct, 5) << "insert_item adding " << cur << " weight " << weight
<< " to bucket " << id << dendl;
- int r = crush_bucket_add_item(b, cur, 0);
+ int r = crush_bucket_add_item(crush, b, cur, 0);
assert (!r);
+ break;
+ }
- // now that we've added the (0-weighted) item and any parent buckets, adjust the weight.
- adjust_item_weightf(cct, item, weight);
-
+ // adjust the item's weight in location
+ if(adjust_item_weightf_in_loc(cct, item, weight, loc) > 0) {
if (item >= crush->max_devices) {
crush->max_devices = item + 1;
ldout(cct, 5) << "insert_item max_devices now " << crush->max_devices << dendl;
}
-
return 0;
}
@@ -585,7 +587,7 @@
if (check_item_loc(cct, item, loc, &old_iweight)) {
ldout(cct, 5) << "create_or_move_item " << item << " already at " << loc << dendl;
} else {
- if (item_exists(item)) {
+ if (_search_item_exists(item)) {
weight = get_item_weightf(item);
ldout(cct, 10) << "create_or_move_item " << item << " exists with weight " << weight << dendl;
remove_item(cct, item, true);
@@ -620,7 +622,7 @@
if (old_iweight != iweight) {
ldout(cct, 5) << "update_item " << item << " adjusting weight "
<< ((float)old_iweight/(float)0x10000) << " -> " << weight << dendl;
- adjust_item_weight(cct, item, iweight);
+ adjust_item_weight_in_loc(cct, item, iweight, loc);
ret = 1;
}
if (get_item_name(item) != name) {
@@ -641,7 +643,7 @@
return ret;
}
-int CrushWrapper::get_item_weight(int id)
+int CrushWrapper::get_item_weight(int id) const
{
for (int bidx = 0; bidx < crush->max_buckets; bidx++) {
crush_bucket *b = crush->buckets[bidx];
@@ -654,6 +656,24 @@
return -ENOENT;
}
+int CrushWrapper::get_item_weight_in_loc(int id, const map<string,string> &loc)
+{
+ for (map<string,string>::const_iterator l = loc.begin(); l != loc.end(); l++) {
+ int bid = get_item_id(l->second);
+ if (!bucket_exists(bid))
+ continue;
+ crush_bucket *b = get_bucket(bid);
+ if ( b == NULL)
+ continue;
+ for (unsigned int i = 0; i < b->size; i++) {
+ if (b->items[i] == id) {
+ return crush_get_bucket_item_weight(b, i);
+ }
+ }
+ }
+ return -ENOENT;
+}
+
int CrushWrapper::adjust_item_weight(CephContext *cct, int id, int weight)
{
ldout(cct, 5) << "adjust_item_weight " << id << " weight " << weight << dendl;
@@ -664,7 +684,7 @@
continue;
for (unsigned i = 0; i < b->size; i++) {
if (b->items[i] == id) {
- int diff = crush_bucket_adjust_item_weight(b, id, weight);
+ int diff = crush_bucket_adjust_item_weight(crush, b, id, weight);
ldout(cct, 5) << "adjust_item_weight " << id << " diff " << diff << " in bucket " << bidx << dendl;
adjust_item_weight(cct, -1 - bidx, b->weight);
changed++;
@@ -676,7 +696,33 @@
return changed;
}
-bool CrushWrapper::check_item_present(int id)
+int CrushWrapper::adjust_item_weight_in_loc(CephContext *cct, int id, int weight, const map<string,string>& loc)
+{
+ ldout(cct, 5) << "adjust_item_weight_in_loc " << id << " weight " << weight << " in " << loc << dendl;
+ int changed = 0;
+
+ for (map<string,string>::const_iterator l = loc.begin(); l != loc.end(); l++) {
+ int bid = get_item_id(l->second);
+ if (!bucket_exists(bid))
+ continue;
+ crush_bucket *b = get_bucket(bid);
+ if ( b == NULL)
+ continue;
+ for (unsigned int i = 0; i < b->size; i++) {
+ if (b->items[i] == id) {
+ int diff = crush_bucket_adjust_item_weight(crush, b, id, weight);
+ ldout(cct, 5) << "adjust_item_weight_in_loc " << id << " diff " << diff << " in bucket " << bid << dendl;
+ adjust_item_weight(cct, bid, b->weight);
+ changed++;
+ }
+ }
+ }
+ if (!changed)
+ return -ENOENT;
+ return changed;
+}
+
+bool CrushWrapper::check_item_present(int id) const
{
bool found = false;
@@ -778,20 +824,18 @@
return -EINVAL;
}
- int ruleset = 0;
- for (int i = 0; i < get_max_rules(); i++) {
- if (rule_exists(i) &&
- get_rule_mask_ruleset(i) >= ruleset) {
- ruleset = get_rule_mask_ruleset(i) + 1;
- }
+ int rno = -1;
+ for (rno = 0; rno < get_max_rules(); rno++) {
+ if (!rule_exists(rno) && !ruleset_exists(rno))
+ break;
}
-
int steps = 3;
if (mode == "indep")
steps = 4;
int min_rep = mode == "firstn" ? 1 : 3;
int max_rep = mode == "firstn" ? 10 : 20;
- crush_rule *rule = crush_make_rule(steps, ruleset, rule_type, min_rep, max_rep);
+ //set the ruleset the same as rule_id(rno)
+ crush_rule *rule = crush_make_rule(steps, rno, rule_type, min_rep, max_rep);
assert(rule);
int step = 0;
if (mode == "indep")
@@ -810,7 +854,12 @@
CRUSH_CHOOSE_N,
0);
crush_rule_set_step(rule, step++, CRUSH_RULE_EMIT, 0, 0);
- int rno = crush_add_rule(crush, rule, -1);
+
+ int ret = crush_add_rule(crush, rule, rno);
+ if(ret < 0) {
+ *err << "failed to add rule " << rno << " because " << cpp_strerror(ret);
+ return ret;
+ }
set_rule_name(rno, name);
have_rmaps = false;
return rno;
@@ -965,6 +1014,7 @@
::encode(crush->choose_total_tries, bl);
::encode(crush->chooseleaf_descend_once, bl);
::encode(crush->chooseleaf_vary_r, bl);
+ ::encode(crush->straw_calc_version, bl);
}
static void decode_32_or_64_string_map(map<int32_t,string>& m, bufferlist::iterator& blp)
@@ -1048,6 +1098,9 @@
if (!blp.end()) {
::decode(crush->chooseleaf_vary_r, blp);
}
+ if (!blp.end()) {
+ ::decode(crush->straw_calc_version, blp);
+ }
finalize();
}
catch (...) {
@@ -1231,6 +1284,8 @@
f->dump_int("choose_local_fallback_tries", get_choose_local_fallback_tries());
f->dump_int("choose_total_tries", get_choose_total_tries());
f->dump_int("chooseleaf_descend_once", get_chooseleaf_descend_once());
+ f->dump_int("chooseleaf_vary_r", get_chooseleaf_vary_r());
+ f->dump_int("straw_calc_version", get_straw_calc_version());
// be helpful about it
if (has_firefly_tunables())
@@ -1246,6 +1301,9 @@
f->dump_int("require_feature_tunables", (int)has_nondefault_tunables());
f->dump_int("require_feature_tunables2", (int)has_nondefault_tunables2());
+ f->dump_int("require_feature_tunables3", (int)has_nondefault_tunables3());
+ f->dump_int("has_v2_rules", (int)has_v2_rules());
+ f->dump_int("has_v3_rules", (int)has_v3_rules());
}
void CrushWrapper::dump_rules(Formatter *f) const
diff -Nru ceph-0.80.7/src/crush/CrushWrapper.h ceph-0.80.9/src/crush/CrushWrapper.h
--- ceph-0.80.7/src/crush/CrushWrapper.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/crush/CrushWrapper.h 2015-03-09 20:51:17.000000000 +0100
@@ -52,24 +52,23 @@
class CrushWrapper {
mutable Mutex mapper_lock;
public:
- struct crush_map *crush;
std::map<int32_t, string> type_map; /* bucket/device type names */
std::map<int32_t, string> name_map; /* bucket/device names */
std::map<int32_t, string> rule_name_map;
- /* reverse maps */
- bool have_rmaps;
- std::map<string, int> type_rmap, name_rmap, rule_name_rmap;
-
private:
- void build_rmaps() {
+ struct crush_map *crush;
+ /* reverse maps */
+ mutable bool have_rmaps;
+ mutable std::map<string, int> type_rmap, name_rmap, rule_name_rmap;
+ void build_rmaps() const {
if (have_rmaps) return;
build_rmap(type_map, type_rmap);
build_rmap(name_map, name_rmap);
build_rmap(rule_name_map, rule_name_rmap);
have_rmaps = true;
}
- void build_rmap(const map<int, string> &f, std::map<string, int> &r) {
+ void build_rmap(const map<int, string> &f, std::map<string, int> &r) const {
r.clear();
for (std::map<int, string>::const_iterator p = f.begin(); p != f.end(); ++p)
r[p->second] = p->first;
@@ -88,6 +87,8 @@
crush_destroy(crush);
}
+ crush_map *get_crush_map() { return crush; }
+
/* building */
void create() {
if (crush)
@@ -124,12 +125,15 @@
void set_tunables_legacy() {
set_tunables_argonaut();
+ crush->straw_calc_version = 0;
}
void set_tunables_optimal() {
set_tunables_firefly();
+ crush->straw_calc_version = 1;
}
void set_tunables_default() {
set_tunables_bobtail();
+ crush->straw_calc_version = 1;
}
int get_choose_local_tries() const {
@@ -167,13 +171,21 @@
crush->chooseleaf_vary_r = n;
}
+ int get_straw_calc_version() const {
+ return crush->straw_calc_version;
+ }
+ void set_straw_calc_version(int n) {
+ crush->straw_calc_version = n;
+ }
+
bool has_argonaut_tunables() const {
return
crush->choose_local_tries == 2 &&
crush->choose_local_fallback_tries == 5 &&
crush->choose_total_tries == 19 &&
crush->chooseleaf_descend_once == 0 &&
- crush->chooseleaf_vary_r == 0;
+ crush->chooseleaf_vary_r == 0 &&
+ crush->straw_calc_version == 0;
}
bool has_bobtail_tunables() const {
return
@@ -181,7 +193,8 @@
crush->choose_local_fallback_tries == 0 &&
crush->choose_total_tries == 50 &&
crush->chooseleaf_descend_once == 1 &&
- crush->chooseleaf_vary_r == 0;
+ crush->chooseleaf_vary_r == 0 &&
+ crush->straw_calc_version == 0;
}
bool has_firefly_tunables() const {
return
@@ -189,7 +202,8 @@
crush->choose_local_fallback_tries == 0 &&
crush->choose_total_tries == 50 &&
crush->chooseleaf_descend_once == 1 &&
- crush->chooseleaf_vary_r == 1;
+ crush->chooseleaf_vary_r == 1 &&
+ crush->straw_calc_version == 0;
}
bool has_optimal_tunables() const {
@@ -223,7 +237,7 @@
int get_num_type_names() const {
return type_map.size();
}
- int get_type_id(const string& name) {
+ int get_type_id(const string& name) const {
build_rmaps();
if (type_rmap.count(name))
return type_rmap[name];
@@ -242,14 +256,14 @@
}
// item/bucket names
- bool name_exists(const string& name) {
+ bool name_exists(const string& name) const {
build_rmaps();
return name_rmap.count(name);
}
bool item_exists(int i) {
return name_map.count(i);
}
- int get_item_id(const string& name) {
+ int get_item_id(const string& name) const {
build_rmaps();
if (name_rmap.count(name))
return name_rmap[name];
@@ -271,11 +285,11 @@
}
// rule names
- bool rule_exists(string name) {
+ bool rule_exists(string name) const {
build_rmaps();
return rule_name_rmap.count(name);
}
- int get_rule_id(string name) {
+ int get_rule_id(string name) const {
build_rmaps();
if (rule_name_rmap.count(name))
return rule_name_rmap[name];
@@ -542,19 +556,27 @@
* @param id item id to check
* @return weight of item
*/
- int get_item_weight(int id);
- float get_item_weightf(int id) {
+ int get_item_weight(int id) const;
+ float get_item_weightf(int id) const {
return (float)get_item_weight(id) / (float)0x10000;
}
+ int get_item_weight_in_loc(int id, const map<string,string> &loc);
+ float get_item_weightf_in_loc(int id, const map<string,string> &loc) {
+ return (float)get_item_weight_in_loc(id, loc) / (float)0x10000;
+ }
int adjust_item_weight(CephContext *cct, int id, int weight);
int adjust_item_weightf(CephContext *cct, int id, float weight) {
return adjust_item_weight(cct, id, (int)(weight * (float)0x10000));
}
+ int adjust_item_weight_in_loc(CephContext *cct, int id, int weight, const map<string,string>& loc);
+ int adjust_item_weightf_in_loc(CephContext *cct, int id, float weight, const map<string,string>& loc) {
+ return adjust_item_weight_in_loc(cct, id, (int)(weight * (float)0x10000), loc);
+ }
void reweight(CephContext *cct);
/// check if item id is present in the map hierarchy
- bool check_item_present(int id);
+ bool check_item_present(int id) const;
/*** devices ***/
@@ -745,9 +767,6 @@
crush_bucket *b = get_bucket(item);
unsigned bucket_weight = b->weight;
- // zero out the bucket weight
- adjust_item_weight(cct, item, 0);
-
// get where the bucket is located
pair<string, string> bucket_location = get_immediate_parent(item);
@@ -758,8 +777,12 @@
crush_bucket *parent_bucket = get_bucket(parent_id);
if (!IS_ERR(parent_bucket)) {
+ // zero out the bucket weight
+ crush_bucket_adjust_item_weight(crush, parent_bucket, item, 0);
+ adjust_item_weight(cct, parent_bucket->id, parent_bucket->weight);
+
// remove the bucket from the parent
- crush_bucket_remove_item(parent_bucket, item);
+ crush_bucket_remove_item(crush, parent_bucket, item);
} else if (PTR_ERR(parent_bucket) != -ENOENT) {
return PTR_ERR(parent_bucket);
}
@@ -839,7 +862,7 @@
int *items, int *weights, int *idout) {
if (type == 0)
return -EINVAL;
- crush_bucket *b = crush_make_bucket(alg, hash, type, size, items, weights);
+ crush_bucket *b = crush_make_bucket(crush, alg, hash, type, size, items, weights);
assert(b);
return crush_add_bucket(crush, bucketno, b, idout);
}
@@ -880,9 +903,9 @@
bool ruleset_exists(int ruleset) const {
for (size_t i = 0; i < crush->max_rules; ++i) {
- if (crush->rules[i]->mask.ruleset == ruleset) {
- return true;
- }
+ if (rule_exists(i) && crush->rules[i]->mask.ruleset == ruleset) {
+ return true;
+ }
}
return false;
diff -Nru ceph-0.80.7/src/crush/mapper.c ceph-0.80.9/src/crush/mapper.c
--- ceph-0.80.7/src/crush/mapper.c 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/crush/mapper.c 2015-03-09 20:51:17.000000000 +0100
@@ -291,6 +291,7 @@
* @type: the type of item to choose
* @out: pointer to output vector
* @outpos: our position in that vector
+ * @out_size: size of the out vector
* @tries: number of attempts to make
* @recurse_tries: number of attempts to have recursive chooseleaf make
* @local_retries: localized retries
@@ -305,6 +306,7 @@
const __u32 *weight, int weight_max,
int x, int numrep, int type,
int *out, int outpos,
+ int out_size,
unsigned int tries,
unsigned int recurse_tries,
unsigned int local_retries,
@@ -323,6 +325,7 @@
int item = 0;
int itemtype;
int collide, reject;
+ int count = out_size;
dprintk("CHOOSE%s bucket %d x %d outpos %d numrep %d tries %d recurse_tries %d local_retries %d local_fallback_retries %d parent_r %d\n",
recurse_to_leaf ? "_LEAF" : "",
@@ -330,7 +333,7 @@
tries, recurse_tries, local_retries, local_fallback_retries,
parent_r);
- for (rep = outpos; rep < numrep; rep++) {
+ for (rep = outpos; rep < numrep && count > 0 ; rep++) {
/* keep trying until we get a non-out, non-colliding item */
ftotal = 0;
skip_rep = 0;
@@ -404,7 +407,7 @@
map->buckets[-1-item],
weight, weight_max,
x, outpos+1, 0,
- out2, outpos,
+ out2, outpos, count,
recurse_tries, 0,
local_retries,
local_fallback_retries,
@@ -464,6 +467,7 @@
dprintk("CHOOSE got %d\n", item);
out[outpos] = item;
outpos++;
+ count--;
if (map->choose_tries && ftotal <= map->choose_total_tries)
map->choose_tries[ftotal]++;
@@ -686,6 +690,7 @@
__u32 step;
int i, j;
int numrep;
+ int out_size;
/*
* the original choose_total_tries value was off by one (it
* counted "retries" and not "tries"). add one.
@@ -793,6 +798,7 @@
x, numrep,
curstep->arg2,
o+osize, j,
+ result_max-osize,
choose_tries,
recurse_tries,
choose_local_retries,
@@ -802,11 +808,13 @@
c+osize,
0);
} else {
+ out_size = ((numrep < (result_max-osize)) ?
+ numrep : (result_max-osize));
crush_choose_indep(
map,
map->buckets[-1-w[i]],
weight, weight_max,
- x, numrep, numrep,
+ x, out_size, numrep,
curstep->arg2,
o+osize, j,
choose_tries,
@@ -815,7 +823,7 @@
recurse_to_leaf,
c+osize,
0);
- osize += numrep;
+ osize += out_size;
}
}
diff -Nru ceph-0.80.7/src/.git_version ceph-0.80.9/src/.git_version
--- ceph-0.80.7/src/.git_version 2014-10-14 21:15:38.000000000 +0200
+++ ceph-0.80.9/src/.git_version 2015-03-09 20:54:24.000000000 +0100
@@ -1,2 +1,2 @@
-6c0127fcb58008793d3c8b62d925bc91963672a3
-v0.80.7
+b5a67f0e1d15385bc0d60a6da6e7fc810bde6047
+v0.80.9
diff -Nru ceph-0.80.7/src/include/ceph_features.h ceph-0.80.9/src/include/ceph_features.h
--- ceph-0.80.7/src/include/ceph_features.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/include/ceph_features.h 2015-03-09 20:51:17.000000000 +0100
@@ -52,6 +52,7 @@
#define CEPH_FEATURE_OSD_PRIMARY_AFFINITY (1ULL<<41) /* overlap w/ tunables3 */
#define CEPH_FEATURE_MSGR_KEEPALIVE2 (1ULL<<42)
#define CEPH_FEATURE_OSD_POOLRESEND (1ULL<<43)
+#define CEPH_FEATURE_OSD_SET_ALLOC_HINT (1ULL<<45)
/*
* The introduction of CEPH_FEATURE_OSD_SNAPMAPPER caused the feature
@@ -124,6 +125,7 @@
CEPH_FEATURE_OSD_PRIMARY_AFFINITY | \
CEPH_FEATURE_MSGR_KEEPALIVE2 | \
CEPH_FEATURE_OSD_POOLRESEND | \
+ CEPH_FEATURE_OSD_SET_ALLOC_HINT | \
0ULL)
#define CEPH_FEATURES_SUPPORTED_DEFAULT CEPH_FEATURES_ALL
diff -Nru ceph-0.80.7/src/include/ceph_fs.h ceph-0.80.9/src/include/ceph_fs.h
--- ceph-0.80.7/src/include/ceph_fs.h 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/include/ceph_fs.h 2015-03-09 20:51:17.000000000 +0100
@@ -505,8 +505,10 @@
__le32 dist[];
} __attribute__ ((packed));
-#define CEPH_LOCK_FCNTL 1
-#define CEPH_LOCK_FLOCK 2
+#define CEPH_LOCK_FCNTL 1
+#define CEPH_LOCK_FLOCK 2
+#define CEPH_LOCK_FCNTL_INTR 3
+#define CEPH_LOCK_FLOCK_INTR 4
#define CEPH_LOCK_SHARED 1
#define CEPH_LOCK_EXCL 2
diff -Nru ceph-0.80.7/src/include/util.h ceph-0.80.9/src/include/util.h
--- ceph-0.80.7/src/include/util.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/include/util.h 2015-03-09 20:51:17.000000000 +0100
@@ -4,17 +4,41 @@
* Ceph - scalable distributed file system
*
* Copyright (C) 2012 Inktank Storage, Inc.
+ * Copyright (C) 2014 Red Hat <contact@redhat.com>
*
* This is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License version 2.1, as published by the Free Software
* Foundation. See file COPYING.
*/
+#ifndef CEPH_UTIL_H
+#define CEPH_UTIL_H
// is buf~len completely zero (in 8-byte chunks)
+#include "common/Formatter.h"
#include "include/types.h"
bool buf_is_zero(const char *buf, size_t len);
int64_t unit_to_bytesize(string val, ostream *pss);
+
+struct ceph_data_stats
+{
+ uint64_t byte_total;
+ uint64_t byte_used;
+ uint64_t byte_avail;
+ int avail_percent;
+
+ void dump(Formatter *f) const {
+ assert(f != NULL);
+ f->dump_int("total", byte_total);
+ f->dump_int("used", byte_used);
+ f->dump_int("avail", byte_avail);
+ f->dump_int("avail_percent", avail_percent);
+ }
+};
+typedef struct ceph_data_stats ceph_data_stats_t;
+
+int get_fs_stats(ceph_data_stats_t &stats, const char *path);
+#endif /* CEPH_UTIL_H */
diff -Nru ceph-0.80.7/src/init-radosgw.sysv ceph-0.80.9/src/init-radosgw.sysv
--- ceph-0.80.7/src/init-radosgw.sysv 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/init-radosgw.sysv 2015-03-09 20:51:17.000000000 +0100
@@ -85,10 +85,10 @@
fi
if [ $SYSTEMD -eq 1 ]; then
- systemd-run -r bash -c "ulimit -n 32768; $RADOSGW -n $name"
+ systemd-run -r sudo -u "$user" bash -c "ulimit -n 32768; $RADOSGW -n $name"
else
- #start-stop-daemon --start -u $user -x $RADOSGW -- -n $name
- daemon --user="$user" "ulimit -n 32768; $RADOSGW -n $name"
+ ulimit -n 32768
+ daemon --user="$user" "$RADOSGW -n $name"
fi
echo "Starting $name..."
done
diff -Nru ceph-0.80.7/src/librbd/AioCompletion.h ceph-0.80.9/src/librbd/AioCompletion.h
--- ceph-0.80.7/src/librbd/AioCompletion.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/librbd/AioCompletion.h 2015-03-09 20:51:17.000000000 +0100
@@ -93,6 +93,10 @@
void init_time(ImageCtx *i, aio_type_t t) {
ictx = i;
+ {
+ Mutex::Locker l(ictx->aio_lock);
+ ++ictx->pending_aio;
+ }
aio_type = t;
start_time = ceph_clock_now(ictx->cct);
}
@@ -114,6 +118,14 @@
lderr(ictx->cct) << "completed invalid aio_type: " << aio_type << dendl;
break;
}
+
+ {
+ Mutex::Locker l(ictx->aio_lock);
+ assert(ictx->pending_aio != 0);
+ --ictx->pending_aio;
+ ictx->pending_aio_cond.Signal();
+ }
+
if (complete_cb) {
complete_cb(rbd_comp, complete_arg);
}
diff -Nru ceph-0.80.7/src/librbd/ImageCtx.cc ceph-0.80.9/src/librbd/ImageCtx.cc
--- ceph-0.80.7/src/librbd/ImageCtx.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/librbd/ImageCtx.cc 2015-03-09 20:51:17.000000000 +0100
@@ -45,13 +45,15 @@
snap_lock("librbd::ImageCtx::snap_lock"),
parent_lock("librbd::ImageCtx::parent_lock"),
refresh_lock("librbd::ImageCtx::refresh_lock"),
+ aio_lock("librbd::ImageCtx::aio_lock"),
extra_read_flags(0),
old_format(true),
order(0), size(0), features(0),
format_string(NULL),
id(image_id), parent(NULL),
stripe_unit(0), stripe_count(0),
- object_cacher(NULL), writeback_handler(NULL), object_set(NULL)
+ object_cacher(NULL), writeback_handler(NULL), object_set(NULL),
+ pending_aio(0)
{
md_ctx.dup(p);
data_ctx.dup(p);
@@ -586,6 +588,7 @@
int r = flush_cache();
if (r)
lderr(cct) << "flush_cache returned " << r << dendl;
+ wait_for_pending_aio();
cache_lock.Lock();
bool unclean = object_cacher->release_set(object_set);
cache_lock.Unlock();
@@ -655,5 +658,12 @@
<< ", object overlap " << len
<< " from image extents " << objectx << dendl;
return len;
- }
+ }
+
+ void ImageCtx::wait_for_pending_aio() {
+ Mutex::Locker l(aio_lock);
+ while (pending_aio > 0) {
+ pending_aio_cond.Wait(aio_lock);
+ }
+ }
}
diff -Nru ceph-0.80.7/src/librbd/ImageCtx.h ceph-0.80.9/src/librbd/ImageCtx.h
--- ceph-0.80.7/src/librbd/ImageCtx.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/librbd/ImageCtx.h 2015-03-09 20:51:17.000000000 +0100
@@ -10,6 +10,7 @@
#include <string>
#include <vector>
+#include "common/Cond.h"
#include "common/Mutex.h"
#include "common/RWLock.h"
#include "common/snap_types.h"
@@ -59,7 +60,8 @@
/**
* Lock ordering:
- * md_lock, cache_lock, snap_lock, parent_lock, refresh_lock
+ * md_lock, cache_lock, snap_lock, parent_lock, refresh_lock,
+ * aio_lock
*/
RWLock md_lock; // protects access to the mutable image metadata that
// isn't guarded by other locks below
@@ -68,6 +70,7 @@
RWLock snap_lock; // protects snapshot-related member variables:
RWLock parent_lock; // protects parent_md and parent
Mutex refresh_lock; // protects refresh_seq and last_refresh
+ Mutex aio_lock; // protects pending_aio and pending_aio_cond
unsigned extra_read_flags;
@@ -89,6 +92,9 @@
LibrbdWriteback *writeback_handler;
ObjectCacher::ObjectSet *object_set;
+ Cond pending_aio_cond;
+ uint64_t pending_aio;
+
/**
* Either image_name or image_id must be set.
* If id is not known, pass the empty std::string,
@@ -147,7 +153,7 @@
librados::snap_t in_snap_id);
uint64_t prune_parent_extents(vector<pair<uint64_t,uint64_t> >& objectx,
uint64_t overlap);
-
+ void wait_for_pending_aio();
};
}
diff -Nru ceph-0.80.7/src/librbd/internal.cc ceph-0.80.9/src/librbd/internal.cc
--- ceph-0.80.7/src/librbd/internal.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/librbd/internal.cc 2015-03-09 20:51:17.000000000 +0100
@@ -419,7 +419,15 @@
for (std::list<string>::const_iterator it = pools.begin();
it != pools.end(); ++it) {
IoCtx ioctx;
- rados.ioctx_create(it->c_str(), ioctx);
+ r = rados.ioctx_create(it->c_str(), ioctx);
+ if (r == -ENOENT) {
+ ldout(cct, 1) << "pool " << *it << " no longer exists" << dendl;
+ continue;
+ } else if (r < 0) {
+ lderr(cct) << "Error accessing child image pool " << *it << dendl;
+ return r;
+ }
+
set<string> image_ids;
int r = cls_client::get_children(&ioctx, RBD_CHILDREN,
parent_spec, image_ids);
@@ -633,32 +641,46 @@
parent_spec pspec(ictx->md_ctx.get_id(), ictx->id, snap_id);
// search all pools for children depending on this snapshot
Rados rados(ictx->md_ctx);
- std::list<std::string> pools;
- rados.pool_list(pools);
- std::set<std::string> children;
- for (std::list<std::string>::const_iterator it = pools.begin(); it != pools.end(); ++it) {
- IoCtx pool_ioctx;
- r = rados.ioctx_create(it->c_str(), pool_ioctx);
- if (r < 0) {
- lderr(ictx->cct) << "snap_unprotect: can't create ioctx for pool "
- << *it << dendl;
- goto reprotect_and_return_err;
- }
- r = cls_client::get_children(&pool_ioctx, RBD_CHILDREN, pspec, children);
- // key should not exist for this parent if there is no entry
- if (((r < 0) && (r != -ENOENT))) {
- lderr(ictx->cct) << "can't get children for pool " << *it << dendl;
- goto reprotect_and_return_err;
- }
- // if we found a child, can't unprotect
- if (r == 0) {
- lderr(ictx->cct) << "snap_unprotect: can't unprotect; at least "
- << children.size() << " child(ren) in pool " << it->c_str() << dendl;
- r = -EBUSY;
- goto reprotect_and_return_err;
+
+ // protect against pools being renamed/deleted
+ bool retry_pool_check;
+ do {
+ retry_pool_check = false;
+
+ std::list<std::string> pools;
+ rados.pool_list(pools);
+ for (std::list<std::string>::const_iterator it = pools.begin(); it != pools.end(); ++it) {
+ IoCtx pool_ioctx;
+ r = rados.ioctx_create(it->c_str(), pool_ioctx);
+ if (r == -ENOENT) {
+ ldout(ictx->cct, 1) << "pool " << *it << " no longer exists" << dendl;
+ retry_pool_check = true;
+ break;
+ } else if (r < 0) {
+ lderr(ictx->cct) << "snap_unprotect: can't create ioctx for pool "
+ << *it << dendl;
+ goto reprotect_and_return_err;
+ }
+
+ std::set<std::string> children;
+ r = cls_client::get_children(&pool_ioctx, RBD_CHILDREN, pspec, children);
+ // key should not exist for this parent if there is no entry
+ if (((r < 0) && (r != -ENOENT))) {
+ lderr(ictx->cct) << "can't get children for pool " << *it << dendl;
+ goto reprotect_and_return_err;
+ }
+ // if we found a child, can't unprotect
+ if (r == 0) {
+ lderr(ictx->cct) << "snap_unprotect: can't unprotect; at least "
+ << children.size() << " child(ren) in pool "
+ << it->c_str() << dendl;
+ r = -EBUSY;
+ goto reprotect_and_return_err;
+ }
+ pool_ioctx.close(); // last one out will self-destruct
}
- pool_ioctx.close(); // last one out will self-destruct
- }
+ } while(retry_pool_check);
+
// didn't find any child in any pool, go ahead with unprotect
r = cls_client::set_protection_status(&ictx->md_ctx,
ictx->header_oid,
@@ -1260,7 +1282,6 @@
if (r < 0) {
lderr(ictx->cct) << "error opening parent image: " << cpp_strerror(r)
<< dendl;
- close_image(ictx->parent);
ictx->parent = NULL;
return r;
}
@@ -2118,10 +2139,12 @@
void close_image(ImageCtx *ictx)
{
ldout(ictx->cct, 20) << "close_image " << ictx << dendl;
- if (ictx->object_cacher)
+ if (ictx->object_cacher) {
ictx->shutdown_cache(); // implicitly flushes
- else
+ } else {
flush(ictx);
+ ictx->wait_for_pending_aio();
+ }
if (ictx->parent) {
close_image(ictx->parent);
diff -Nru ceph-0.80.7/src/Makefile.am ceph-0.80.9/src/Makefile.am
--- ceph-0.80.7/src/Makefile.am 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/Makefile.am 2015-03-09 20:51:17.000000000 +0100
@@ -258,6 +258,7 @@
$(check_SCRIPTS)
check-local:
+ $(top_srcdir)/qa/workunits/erasure-code/encode-decode-non-regression.sh
$(srcdir)/test/encoding/readable.sh ../ceph-object-corpus
diff -Nru ceph-0.80.7/src/Makefile-env.am ceph-0.80.9/src/Makefile-env.am
--- ceph-0.80.7/src/Makefile-env.am 2014-10-14 21:12:53.000000000 +0200
+++ ceph-0.80.9/src/Makefile-env.am 2015-03-09 20:51:17.000000000 +0100
@@ -150,6 +150,7 @@
LIBCLIENT_FUSE = libclient_fuse.la
LIBRADOS = librados.la
LIBRGW = librgw.la
+LIBCIVETWEB = libcivetweb.la
LIBRBD = librbd.la
LIBCEPHFS = libcephfs.la
LIBERASURE_CODE = liberasure_code.la
diff -Nru ceph-0.80.7/src/Makefile.in ceph-0.80.9/src/Makefile.in
--- ceph-0.80.7/src/Makefile.in 2014-10-14 21:14:40.000000000 +0200
+++ ceph-0.80.9/src/Makefile.in 2015-03-09 20:53:25.000000000 +0100
@@ -69,7 +69,8 @@
ceph-mon$(EXEEXT) ceph-osd$(EXEEXT) ceph-mds$(EXEEXT) \
cephfs$(EXEEXT) librados-config$(EXEEXT) ceph-syn$(EXEEXT) \
$(am__EXEEXT_12) $(am__EXEEXT_13)
-noinst_PROGRAMS = get_command_descriptions$(EXEEXT)
+noinst_PROGRAMS = ceph_erasure_code_non_regression$(EXEEXT) \
+ get_command_descriptions$(EXEEXT)
sbin_PROGRAMS =
su_sbin_PROGRAMS = $(am__EXEEXT_14)
check_PROGRAMS = unittest_erasure_code_plugin$(EXEEXT) \
@@ -88,20 +89,21 @@
unittest_base64$(EXEEXT) unittest_ceph_argparse$(EXEEXT) \
unittest_ceph_compatset$(EXEEXT) unittest_osd_types$(EXEEXT) \
unittest_pglog$(EXEEXT) unittest_ecbackend$(EXEEXT) \
- unittest_hitset$(EXEEXT) unittest_gather$(EXEEXT) \
- unittest_run_cmd$(EXEEXT) unittest_signals$(EXEEXT) \
- unittest_simple_spin$(EXEEXT) unittest_librados$(EXEEXT) \
- unittest_bufferlist$(EXEEXT) unittest_crc32c$(EXEEXT) \
- unittest_arch$(EXEEXT) unittest_crypto$(EXEEXT) \
- unittest_crypto_init$(EXEEXT) unittest_perf_counters$(EXEEXT) \
- unittest_admin_socket$(EXEEXT) unittest_ceph_crypto$(EXEEXT) \
- unittest_utf8$(EXEEXT) unittest_mime$(EXEEXT) \
- unittest_escape$(EXEEXT) unittest_chain_xattr$(EXEEXT) \
- unittest_flatindex$(EXEEXT) unittest_strtol$(EXEEXT) \
- unittest_confutils$(EXEEXT) unittest_config$(EXEEXT) \
- unittest_context$(EXEEXT) unittest_heartbeatmap$(EXEEXT) \
- unittest_formatter$(EXEEXT) unittest_libcephfs_config$(EXEEXT) \
- unittest_lfnindex$(EXEEXT) unittest_librados_config$(EXEEXT) \
+ unittest_hitset$(EXEEXT) unittest_io_priority$(EXEEXT) \
+ unittest_gather$(EXEEXT) unittest_run_cmd$(EXEEXT) \
+ unittest_signals$(EXEEXT) unittest_simple_spin$(EXEEXT) \
+ unittest_librados$(EXEEXT) unittest_bufferlist$(EXEEXT) \
+ unittest_crc32c$(EXEEXT) unittest_arch$(EXEEXT) \
+ unittest_crypto$(EXEEXT) unittest_crypto_init$(EXEEXT) \
+ unittest_perf_counters$(EXEEXT) unittest_admin_socket$(EXEEXT) \
+ unittest_ceph_crypto$(EXEEXT) unittest_utf8$(EXEEXT) \
+ unittest_mime$(EXEEXT) unittest_escape$(EXEEXT) \
+ unittest_chain_xattr$(EXEEXT) unittest_flatindex$(EXEEXT) \
+ unittest_strtol$(EXEEXT) unittest_confutils$(EXEEXT) \
+ unittest_config$(EXEEXT) unittest_context$(EXEEXT) \
+ unittest_heartbeatmap$(EXEEXT) unittest_formatter$(EXEEXT) \
+ unittest_libcephfs_config$(EXEEXT) unittest_lfnindex$(EXEEXT) \
+ unittest_librados_config$(EXEEXT) \
unittest_daemon_config$(EXEEXT) unittest_osd_osdcap$(EXEEXT) \
unittest_mon_moncap$(EXEEXT) unittest_mon_pgmap$(EXEEXT) \
unittest_ipaddr$(EXEEXT) unittest_texttable$(EXEEXT) \
@@ -142,7 +144,7 @@
@LINUX_TRUE@am__append_32 = -lrt
@LINUX_TRUE@am__append_33 = -export-symbols-regex '^rados_.*'
@LINUX_TRUE@am__append_34 = -export-symbols-regex '^rbd_.*'
-@WITH_RADOSGW_TRUE@am__append_35 = librgw.la
+@WITH_RADOSGW_TRUE@am__append_35 = librgw.la libcivetweb.la
@WITH_RADOSGW_TRUE@am__append_36 = \
@WITH_RADOSGW_TRUE@ $(LIBRADOS) \
@WITH_RADOSGW_TRUE@ libcls_rgw_client.la \
@@ -177,36 +179,37 @@
@LINUX_TRUE@am__append_43 = -ldl
@LINUX_TRUE@am__append_44 = -ldl
@LINUX_TRUE@am__append_45 = -ldl
-@COMPILER_HAS_VTA_TRUE@am__append_46 = -fno-var-tracking-assignments
+@LINUX_TRUE@am__append_46 = -ldl
@COMPILER_HAS_VTA_TRUE@am__append_47 = -fno-var-tracking-assignments
-@WITH_BUILD_TESTS_TRUE@am__append_48 = test_build_libcommon \
+@COMPILER_HAS_VTA_TRUE@am__append_48 = -fno-var-tracking-assignments
+@WITH_BUILD_TESTS_TRUE@am__append_49 = test_build_libcommon \
@WITH_BUILD_TESTS_TRUE@ test_build_librados test_build_librgw \
@WITH_BUILD_TESTS_TRUE@ test_build_libcephfs
-@LINUX_TRUE@am__append_49 = ceph_kvstorebench \
+@LINUX_TRUE@am__append_50 = ceph_kvstorebench \
@LINUX_TRUE@ ceph_test_rados_list_parallel \
@LINUX_TRUE@ ceph_test_rados_open_pools_parallel \
@LINUX_TRUE@ ceph_test_rados_delete_pools_parallel \
@LINUX_TRUE@ ceph_test_rados_watch_notify
-@LINUX_TRUE@am__append_50 = libsystest.la
-@LINUX_TRUE@am__append_51 = -ldl
-@WITH_RADOSGW_TRUE@am__append_52 = ceph_test_cors \
+@LINUX_TRUE@am__append_51 = libsystest.la
+@LINUX_TRUE@am__append_52 = -ldl
+@WITH_RADOSGW_TRUE@am__append_53 = ceph_test_cors \
@WITH_RADOSGW_TRUE@ ceph_test_rgw_manifest \
@WITH_RADOSGW_TRUE@ ceph_test_cls_rgw_meta \
@WITH_RADOSGW_TRUE@ ceph_test_cls_rgw_log \
@WITH_RADOSGW_TRUE@ ceph_test_cls_rgw_opstate
-@LINUX_TRUE@am__append_53 = ceph_test_librbd_fsx
-@WITH_RADOSGW_TRUE@am__append_54 = ceph_test_cls_rgw
-@LINUX_TRUE@am__append_55 = ceph_test_objectstore
-@LINUX_TRUE@am__append_56 = -ldl
+@LINUX_TRUE@am__append_54 = ceph_test_librbd_fsx
+@WITH_RADOSGW_TRUE@am__append_55 = ceph_test_cls_rgw
+@LINUX_TRUE@am__append_56 = ceph_test_objectstore
@LINUX_TRUE@am__append_57 = -ldl
-@WITH_REST_BENCH_TRUE@am__append_58 = rest-bench
-@WITH_REST_BENCH_TRUE@@WITH_SYSTEM_LIBS3_TRUE@am__append_59 = -ls3
-@WITH_REST_BENCH_TRUE@@WITH_SYSTEM_LIBS3_FALSE@am__append_60 = libs3/build/lib/libs3.a -lcurl -lxml2
-@WITH_REST_BENCH_TRUE@@WITH_SYSTEM_LIBS3_FALSE@am__append_61 = libs3
-@LINUX_TRUE@am__append_62 = mount.ceph
-@LINUX_TRUE@am__append_63 = rbd
-@WITH_FUSE_TRUE@am__append_64 = ceph-fuse rbd-fuse
-@ENABLE_CEPHFS_JAVA_TRUE@am__append_65 = libcephfs_jni.la
+@LINUX_TRUE@am__append_58 = -ldl
+@WITH_REST_BENCH_TRUE@am__append_59 = rest-bench
+@WITH_REST_BENCH_TRUE@@WITH_SYSTEM_LIBS3_TRUE@am__append_60 = -ls3
+@WITH_REST_BENCH_TRUE@@WITH_SYSTEM_LIBS3_FALSE@am__append_61 = libs3/build/lib/libs3.a -lcurl -lxml2
+@WITH_REST_BENCH_TRUE@@WITH_SYSTEM_LIBS3_FALSE@am__append_62 = libs3
+@LINUX_TRUE@am__append_63 = mount.ceph
+@LINUX_TRUE@am__append_64 = rbd
+@WITH_FUSE_TRUE@am__append_65 = ceph-fuse rbd-fuse
+@ENABLE_CEPHFS_JAVA_TRUE@am__append_66 = libcephfs_jni.la
subdir = src
ACLOCAL_M4 = $(top_srcdir)/aclocal.m4
am__aclocal_m4_deps = $(top_srcdir)/m4/ac_check_classpath.m4 \
@@ -357,6 +360,19 @@
$(AM_CXXFLAGS) $(CXXFLAGS) $(libcephfs_jni_la_LDFLAGS) \
$(LDFLAGS) -o $@
@ENABLE_CEPHFS_JAVA_TRUE@am_libcephfs_jni_la_rpath = -rpath $(libdir)
+libcivetweb_la_LIBADD =
+am__libcivetweb_la_SOURCES_DIST = rgw/rgw_civetweb.cc \
+ rgw/rgw_civetweb_log.cc civetweb/src/civetweb.c
+@WITH_RADOSGW_TRUE@am_libcivetweb_la_OBJECTS = \
+@WITH_RADOSGW_TRUE@ rgw/libcivetweb_la-rgw_civetweb.lo \
+@WITH_RADOSGW_TRUE@ rgw/libcivetweb_la-rgw_civetweb_log.lo \
+@WITH_RADOSGW_TRUE@ civetweb/src/libcivetweb_la-civetweb.lo
+libcivetweb_la_OBJECTS = $(am_libcivetweb_la_OBJECTS)
+libcivetweb_la_LINK = $(LIBTOOL) $(AM_V_lt) --tag=CXX \
+ $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=link $(CXXLD) \
+ $(libcivetweb_la_CXXFLAGS) $(CXXFLAGS) $(AM_LDFLAGS) \
+ $(LDFLAGS) -o $@
+@WITH_RADOSGW_TRUE@am_libcivetweb_la_rpath =
libclient_la_DEPENDENCIES = $(LIBOSDC) $(am__DEPENDENCIES_1)
am_libclient_la_OBJECTS = client/Client.lo client/Inode.lo \
client/Dentry.lo client/MetaRequest.lo \
@@ -509,7 +525,7 @@
common/bloom_filter.lo common/linux_version.lo mon/MonCap.lo \
mon/MonClient.lo mon/MonMap.lo osd/OSDMap.lo osd/osd_types.lo \
osd/ECMsgTypes.lo osd/HitSet.lo mds/MDSMap.lo \
- mds/inode_backtrace.lo mds/mdstypes.lo
+ mds/inode_backtrace.lo mds/mdstypes.lo mds/flock.lo
libcommon_la_OBJECTS = $(am_libcommon_la_OBJECTS)
libcommon_crc_la_LIBADD =
am__libcommon_crc_la_SOURCES_DIST = common/sctp_crc32.c \
@@ -707,11 +723,11 @@
liblog_la_OBJECTS = $(am_liblog_la_OBJECTS)
libmds_la_DEPENDENCIES = $(LIBOSDC)
am_libmds_la_OBJECTS = mds/Anchor.lo mds/Capability.lo mds/Dumper.lo \
- mds/Resetter.lo mds/MDS.lo mds/flock.lo mds/locks.lo \
- mds/journal.lo mds/Server.lo mds/Mutation.lo mds/MDCache.lo \
- mds/Locker.lo mds/Migrator.lo mds/MDBalancer.lo mds/CDentry.lo \
- mds/CDir.lo mds/CInode.lo mds/LogEvent.lo mds/MDSTable.lo \
- mds/InoTable.lo mds/MDSTableClient.lo mds/MDSTableServer.lo \
+ mds/Resetter.lo mds/MDS.lo mds/locks.lo mds/journal.lo \
+ mds/Server.lo mds/Mutation.lo mds/MDCache.lo mds/Locker.lo \
+ mds/Migrator.lo mds/MDBalancer.lo mds/CDentry.lo mds/CDir.lo \
+ mds/CInode.lo mds/LogEvent.lo mds/MDSTable.lo mds/InoTable.lo \
+ mds/MDSTableClient.lo mds/MDSTableServer.lo \
mds/AnchorServer.lo mds/AnchorClient.lo mds/SnapRealm.lo \
mds/SnapServer.lo mds/snap.lo mds/SessionMap.lo mds/MDLog.lo \
mds/MDSUtility.lo
@@ -1073,6 +1089,13 @@
ceph_erasure_code_benchmark_DEPENDENCIES = $(am__DEPENDENCIES_10) \
$(LIBCOMMON) $(am__DEPENDENCIES_1) $(am__DEPENDENCIES_6) \
$(am__DEPENDENCIES_1)
+am_ceph_erasure_code_non_regression_OBJECTS = \
+ test/erasure-code/ceph_erasure_code_non_regression.$(OBJEXT)
+ceph_erasure_code_non_regression_OBJECTS = \
+ $(am_ceph_erasure_code_non_regression_OBJECTS)
+ceph_erasure_code_non_regression_DEPENDENCIES = \
+ $(am__DEPENDENCIES_10) $(LIBCOMMON) $(am__DEPENDENCIES_1) \
+ $(am__DEPENDENCIES_6) $(am__DEPENDENCIES_1)
am_ceph_filestore_dump_OBJECTS = tools/ceph_filestore_dump.$(OBJEXT)
ceph_filestore_dump_OBJECTS = $(am_ceph_filestore_dump_OBJECTS)
ceph_filestore_dump_DEPENDENCIES = $(am__DEPENDENCIES_10) \
@@ -1774,8 +1797,7 @@
rgw/rgw_rest_log.cc rgw/rgw_rest_opstate.cc \
rgw/rgw_rest_replica_log.cc rgw/rgw_rest_config.cc \
rgw/rgw_http_client.cc rgw/rgw_swift.cc rgw/rgw_swift_auth.cc \
- rgw/rgw_loadgen.cc rgw/rgw_civetweb.cc civetweb/src/civetweb.c \
- rgw/rgw_main.cc
+ rgw/rgw_loadgen.cc rgw/rgw_main.cc
@WITH_RADOSGW_TRUE@am_radosgw_OBJECTS = rgw/rgw_resolve.$(OBJEXT) \
@WITH_RADOSGW_TRUE@ rgw/rgw_rest.$(OBJEXT) \
@WITH_RADOSGW_TRUE@ rgw/rgw_rest_swift.$(OBJEXT) \
@@ -1793,11 +1815,9 @@
@WITH_RADOSGW_TRUE@ rgw/rgw_swift.$(OBJEXT) \
@WITH_RADOSGW_TRUE@ rgw/rgw_swift_auth.$(OBJEXT) \
@WITH_RADOSGW_TRUE@ rgw/rgw_loadgen.$(OBJEXT) \
-@WITH_RADOSGW_TRUE@ rgw/rgw_civetweb.$(OBJEXT) \
-@WITH_RADOSGW_TRUE@ civetweb/src/radosgw-civetweb.$(OBJEXT) \
@WITH_RADOSGW_TRUE@ rgw/rgw_main.$(OBJEXT)
radosgw_OBJECTS = $(am_radosgw_OBJECTS)
-@WITH_RADOSGW_TRUE@radosgw_DEPENDENCIES = $(LIBRGW) \
+@WITH_RADOSGW_TRUE@radosgw_DEPENDENCIES = $(LIBRGW) $(LIBCIVETWEB) \
@WITH_RADOSGW_TRUE@ $(am__DEPENDENCIES_12) \
@WITH_RADOSGW_TRUE@ $(am__DEPENDENCIES_1) $(am__DEPENDENCIES_6)
am__radosgw_admin_SOURCES_DIST = rgw/rgw_admin.cc
@@ -1876,7 +1896,7 @@
common/bloom_filter.cc common/linux_version.c mon/MonCap.cc \
mon/MonClient.cc mon/MonMap.cc osd/OSDMap.cc osd/osd_types.cc \
osd/ECMsgTypes.cc osd/HitSet.cc mds/MDSMap.cc \
- mds/inode_backtrace.cc mds/mdstypes.cc
+ mds/inode_backtrace.cc mds/mdstypes.cc mds/flock.cc
am__objects_15 = test_build_libcommon-ceph_ver.$(OBJEXT) \
common/test_build_libcommon-DecayCounter.$(OBJEXT) \
common/test_build_libcommon-LogClient.$(OBJEXT) \
@@ -1958,7 +1978,8 @@
osd/test_build_libcommon-HitSet.$(OBJEXT) \
mds/test_build_libcommon-MDSMap.$(OBJEXT) \
mds/test_build_libcommon-inode_backtrace.$(OBJEXT) \
- mds/test_build_libcommon-mdstypes.$(OBJEXT)
+ mds/test_build_libcommon-mdstypes.$(OBJEXT) \
+ mds/test_build_libcommon-flock.$(OBJEXT)
@WITH_BUILD_TESTS_TRUE@am_test_build_libcommon_OBJECTS = test/test_build_libcommon-buildtest_skeleton.$(OBJEXT) \
@WITH_BUILD_TESTS_TRUE@ $(am__objects_15)
test_build_libcommon_OBJECTS = $(am_test_build_libcommon_OBJECTS)
@@ -2356,6 +2377,15 @@
$(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=link $(CXXLD) \
$(unittest_hitset_CXXFLAGS) $(CXXFLAGS) $(AM_LDFLAGS) \
$(LDFLAGS) -o $@
+am_unittest_io_priority_OBJECTS = \
+ test/common/unittest_io_priority-test_io_priority.$(OBJEXT)
+unittest_io_priority_OBJECTS = $(am_unittest_io_priority_OBJECTS)
+unittest_io_priority_DEPENDENCIES = $(am__DEPENDENCIES_13) \
+ $(am__DEPENDENCIES_6)
+unittest_io_priority_LINK = $(LIBTOOL) $(AM_V_lt) --tag=CXX \
+ $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=link $(CXXLD) \
+ $(unittest_io_priority_CXXFLAGS) $(CXXFLAGS) $(AM_LDFLAGS) \
+ $(LDFLAGS) -o $@
am_unittest_ipaddr_OBJECTS = \
test/unittest_ipaddr-test_ipaddr.$(OBJEXT)
unittest_ipaddr_OBJECTS = $(am_unittest_ipaddr_OBJECTS)
@@ -2668,11 +2698,12 @@
$(libcls_version_client_a_SOURCES) $(libos_zfs_a_SOURCES) \
$(libarch_la_SOURCES) $(libauth_la_SOURCES) \
$(libcephfs_la_SOURCES) $(libcephfs_jni_la_SOURCES) \
- $(libclient_la_SOURCES) $(libclient_fuse_la_SOURCES) \
- $(libcls_hello_la_SOURCES) $(libcls_kvs_la_SOURCES) \
- $(libcls_lock_la_SOURCES) $(libcls_lock_client_la_SOURCES) \
- $(libcls_log_la_SOURCES) $(libcls_rbd_la_SOURCES) \
- $(libcls_rbd_client_la_SOURCES) $(libcls_refcount_la_SOURCES) \
+ $(libcivetweb_la_SOURCES) $(libclient_la_SOURCES) \
+ $(libclient_fuse_la_SOURCES) $(libcls_hello_la_SOURCES) \
+ $(libcls_kvs_la_SOURCES) $(libcls_lock_la_SOURCES) \
+ $(libcls_lock_client_la_SOURCES) $(libcls_log_la_SOURCES) \
+ $(libcls_rbd_la_SOURCES) $(libcls_rbd_client_la_SOURCES) \
+ $(libcls_refcount_la_SOURCES) \
$(libcls_refcount_client_la_SOURCES) \
$(libcls_replica_log_la_SOURCES) $(libcls_rgw_la_SOURCES) \
$(libcls_rgw_client_la_SOURCES) $(libcls_statelog_la_SOURCES) \
@@ -2707,6 +2738,7 @@
$(ceph_syn_SOURCES) $(ceph_bench_log_SOURCES) \
$(ceph_dupstore_SOURCES) $(ceph_erasure_code_SOURCES) \
$(ceph_erasure_code_benchmark_SOURCES) \
+ $(ceph_erasure_code_non_regression_SOURCES) \
$(ceph_filestore_dump_SOURCES) $(ceph_filestore_tool_SOURCES) \
$(ceph_kvstorebench_SOURCES) \
$(ceph_mon_store_converter_SOURCES) \
@@ -2794,8 +2826,8 @@
$(unittest_escape_SOURCES) $(unittest_flatindex_SOURCES) \
$(unittest_formatter_SOURCES) $(unittest_gather_SOURCES) \
$(unittest_heartbeatmap_SOURCES) $(unittest_histogram_SOURCES) \
- $(unittest_hitset_SOURCES) $(unittest_ipaddr_SOURCES) \
- $(unittest_lfnindex_SOURCES) \
+ $(unittest_hitset_SOURCES) $(unittest_io_priority_SOURCES) \
+ $(unittest_ipaddr_SOURCES) $(unittest_lfnindex_SOURCES) \
$(unittest_libcephfs_config_SOURCES) \
$(unittest_librados_SOURCES) \
$(unittest_librados_config_SOURCES) $(unittest_log_SOURCES) \
@@ -2821,7 +2853,8 @@
$(libcls_version_client_a_SOURCES) \
$(am__libos_zfs_a_SOURCES_DIST) $(libarch_la_SOURCES) \
$(libauth_la_SOURCES) $(libcephfs_la_SOURCES) \
- $(am__libcephfs_jni_la_SOURCES_DIST) $(libclient_la_SOURCES) \
+ $(am__libcephfs_jni_la_SOURCES_DIST) \
+ $(am__libcivetweb_la_SOURCES_DIST) $(libclient_la_SOURCES) \
$(am__libclient_fuse_la_SOURCES_DIST) \
$(libcls_hello_la_SOURCES) $(am__libcls_kvs_la_SOURCES_DIST) \
$(libcls_lock_la_SOURCES) $(libcls_lock_client_la_SOURCES) \
@@ -2862,6 +2895,7 @@
$(ceph_bench_log_SOURCES) $(ceph_dupstore_SOURCES) \
$(ceph_erasure_code_SOURCES) \
$(ceph_erasure_code_benchmark_SOURCES) \
+ $(ceph_erasure_code_non_regression_SOURCES) \
$(ceph_filestore_dump_SOURCES) $(ceph_filestore_tool_SOURCES) \
$(am__ceph_kvstorebench_SOURCES_DIST) \
$(ceph_mon_store_converter_SOURCES) \
@@ -2955,8 +2989,8 @@
$(unittest_escape_SOURCES) $(unittest_flatindex_SOURCES) \
$(unittest_formatter_SOURCES) $(unittest_gather_SOURCES) \
$(unittest_heartbeatmap_SOURCES) $(unittest_histogram_SOURCES) \
- $(unittest_hitset_SOURCES) $(unittest_ipaddr_SOURCES) \
- $(unittest_lfnindex_SOURCES) \
+ $(unittest_hitset_SOURCES) $(unittest_io_priority_SOURCES) \
+ $(unittest_ipaddr_SOURCES) $(unittest_lfnindex_SOURCES) \
$(unittest_libcephfs_config_SOURCES) \
$(unittest_librados_SOURCES) \
$(unittest_librados_config_SOURCES) $(unittest_log_SOURCES) \
@@ -3219,8 +3253,9 @@
rgw/rgw_rest_log.h rgw/rgw_rest_opstate.h \
rgw/rgw_rest_replica_log.h rgw/rgw_rest_config.h \
rgw/rgw_usage.h rgw/rgw_user.h rgw/rgw_bucket.h \
- rgw/rgw_keystone.h rgw/rgw_civetweb.h civetweb/civetweb.h \
- civetweb/include/civetweb.h civetweb/src/md5.h \
+ rgw/rgw_keystone.h rgw/rgw_civetweb.h rgw/rgw_civetweb_log.h \
+ civetweb/civetweb.h civetweb/include/civetweb.h \
+ civetweb/include/civetweb_conf.h civetweb/src/md5.h \
cls/lock/cls_lock_types.h cls/lock/cls_lock_ops.h \
cls/lock/cls_lock_client.h cls/rbd/cls_rbd.h \
cls/rbd/cls_rbd_client.h cls/refcount/cls_refcount_ops.h \
@@ -3485,7 +3520,7 @@
top_builddir = @top_builddir@
top_srcdir = @top_srcdir@
AUTOMAKE_OPTIONS = gnu subdir-objects
-SUBDIRS = ocf java $(am__append_61)
+SUBDIRS = ocf java $(am__append_62)
DIST_SUBDIRS = gtest ocf libs3 java
BUILT_SOURCES = init-ceph
@@ -3763,8 +3798,9 @@
rgw/rgw_rest_log.h rgw/rgw_rest_opstate.h \
rgw/rgw_rest_replica_log.h rgw/rgw_rest_config.h \
rgw/rgw_usage.h rgw/rgw_user.h rgw/rgw_bucket.h \
- rgw/rgw_keystone.h rgw/rgw_civetweb.h civetweb/civetweb.h \
- civetweb/include/civetweb.h civetweb/src/md5.h \
+ rgw/rgw_keystone.h rgw/rgw_civetweb.h rgw/rgw_civetweb_log.h \
+ civetweb/civetweb.h civetweb/include/civetweb.h \
+ civetweb/include/civetweb_conf.h civetweb/src/md5.h \
cls/lock/cls_lock_types.h cls/lock/cls_lock_ops.h \
cls/lock/cls_lock_client.h cls/rbd/cls_rbd.h \
cls/rbd/cls_rbd_client.h cls/refcount/cls_refcount_ops.h \
@@ -3819,7 +3855,7 @@
sbin_SCRIPTS =
su_sbin_SCRIPTS = mount.fuse.ceph mkcephfs
dist_bin_SCRIPTS =
-lib_LTLIBRARIES = librados.la librbd.la libcephfs.la $(am__append_65)
+lib_LTLIBRARIES = librados.la librbd.la libcephfs.la $(am__append_66)
noinst_LTLIBRARIES = libarch.la libauth.la libcrush.la libmon_types.la \
libmon.la libmds.la libos_types.la libos.la libosd_types.la \
libosd.la liberasure_code.la libosdc.la libclient.la \
@@ -3827,7 +3863,7 @@
libperfglue.la libcommon_crc.la libcommon.la libmsg.la \
$(am__append_35) libcls_lock_client.la \
libcls_refcount_client.la libcls_rgw_client.la \
- libcls_rbd_client.la $(am__append_50) libradostest.la
+ libcls_rbd_client.la $(am__append_51) libradostest.la
noinst_LIBRARIES = $(am__append_17) libcls_version_client.a \
libcls_log_client.a libcls_statelog_client.a \
libcls_replica_log_client.a libcls_user_client.a
@@ -3842,14 +3878,14 @@
ceph_test_signal_handlers ceph_test_rados ceph_test_mutate \
ceph_test_rewrite_latency ceph_test_msgr ceph_streamtest \
ceph_test_trans ceph_test_crypto ceph_test_keys \
- $(am__append_48) ceph_smalliobench ceph_smalliobenchfs \
+ $(am__append_49) ceph_smalliobench ceph_smalliobenchfs \
ceph_smalliobenchdumb ceph_smalliobenchrbd ceph_tpbench \
- ceph_omapbench $(am__append_49) ceph_bench_log \
- $(am__append_52) ceph_multi_stress_watch ceph_test_librbd \
- $(am__append_53) ceph_test_cls_rbd ceph_test_cls_refcount \
+ ceph_omapbench $(am__append_50) ceph_bench_log \
+ $(am__append_53) ceph_multi_stress_watch ceph_test_librbd \
+ $(am__append_54) ceph_test_cls_rbd ceph_test_cls_refcount \
ceph_test_cls_version ceph_test_cls_log ceph_test_cls_statelog \
ceph_test_cls_replica_log ceph_test_cls_lock \
- ceph_test_cls_hello $(am__append_54) ceph_test_mon_workloadgen \
+ ceph_test_cls_hello $(am__append_55) ceph_test_mon_workloadgen \
ceph_test_rados_api_cmd ceph_test_rados_api_io \
ceph_test_rados_api_c_write_operations \
ceph_test_rados_api_c_read_operations ceph_test_rados_api_aio \
@@ -3857,7 +3893,7 @@
ceph_test_rados_api_stat ceph_test_rados_api_watch_notify \
ceph_test_rados_api_snapshots ceph_test_rados_api_cls \
ceph_test_rados_api_misc ceph_test_rados_api_tier \
- ceph_test_rados_api_lock ceph_test_libcephfs $(am__append_55) \
+ ceph_test_rados_api_lock ceph_test_libcephfs $(am__append_56) \
ceph_test_objectstore_workloadgen \
ceph_test_filestore_idempotent \
ceph_test_filestore_idempotent_sequence ceph_xattr_bench \
@@ -3951,6 +3987,7 @@
LIBCLIENT_FUSE = libclient_fuse.la
LIBRADOS = librados.la
LIBRGW = librgw.la
+LIBCIVETWEB = libcivetweb.la
LIBRBD = librbd.la
LIBCEPHFS = libcephfs.la
LIBERASURE_CODE = liberasure_code.la
@@ -4031,7 +4068,6 @@
mds/Dumper.cc \
mds/Resetter.cc \
mds/MDS.cc \
- mds/flock.cc \
mds/locks.c \
mds/journal.cc \
mds/Server.cc \
@@ -4265,7 +4301,7 @@
common/bloom_filter.cc common/linux_version.c mon/MonCap.cc \
mon/MonClient.cc mon/MonMap.cc osd/OSDMap.cc osd/osd_types.cc \
osd/ECMsgTypes.cc osd/HitSet.cc mds/MDSMap.cc \
- mds/inode_backtrace.cc mds/mdstypes.cc
+ mds/inode_backtrace.cc mds/mdstypes.cc mds/flock.cc
# inject crc in common
libcommon_crc_la_SOURCES = common/sctp_crc32.c common/crc32c.cc \
@@ -4365,6 +4401,14 @@
@WITH_RADOSGW_TRUE@ rgw/rgw_dencoder.cc
@WITH_RADOSGW_TRUE@librgw_la_CXXFLAGS = -Woverloaded-virtual ${AM_CXXFLAGS}
+@WITH_RADOSGW_TRUE@CIVETWEB_INCLUDE = --include civetweb/include/civetweb_conf.h
+@WITH_RADOSGW_TRUE@libcivetweb_la_SOURCES = \
+@WITH_RADOSGW_TRUE@ rgw/rgw_civetweb.cc \
+@WITH_RADOSGW_TRUE@ rgw/rgw_civetweb_log.cc \
+@WITH_RADOSGW_TRUE@ civetweb/src/civetweb.c
+
+@WITH_RADOSGW_TRUE@libcivetweb_la_CXXFLAGS = ${CIVETWEB_INCLUDE} -Woverloaded-virtual ${AM_CXXFLAGS}
+@WITH_RADOSGW_TRUE@libcivetweb_la_CFLAGS = -Icivetweb/include ${CIVETWEB_INCLUDE}
@WITH_RADOSGW_TRUE@radosgw_SOURCES = \
@WITH_RADOSGW_TRUE@ rgw/rgw_resolve.cc \
@WITH_RADOSGW_TRUE@ rgw/rgw_rest.cc \
@@ -4383,12 +4427,10 @@
@WITH_RADOSGW_TRUE@ rgw/rgw_swift.cc \
@WITH_RADOSGW_TRUE@ rgw/rgw_swift_auth.cc \
@WITH_RADOSGW_TRUE@ rgw/rgw_loadgen.cc \
-@WITH_RADOSGW_TRUE@ rgw/rgw_civetweb.cc \
-@WITH_RADOSGW_TRUE@ civetweb/src/civetweb.c \
@WITH_RADOSGW_TRUE@ rgw/rgw_main.cc
-@WITH_RADOSGW_TRUE@radosgw_CFLAGS = -Icivetweb/include
-@WITH_RADOSGW_TRUE@radosgw_LDADD = $(LIBRGW) $(LIBRGW_DEPS) $(RESOLV_LIBS) $(CEPH_GLOBAL)
+@WITH_RADOSGW_TRUE@radosgw_CFLAGS = -I$(srcdir)/civetweb/include
+@WITH_RADOSGW_TRUE@radosgw_LDADD = $(LIBRGW) $(LIBCIVETWEB) $(LIBRGW_DEPS) $(RESOLV_LIBS) $(CEPH_GLOBAL)
@WITH_RADOSGW_TRUE@radosgw_admin_SOURCES = rgw/rgw_admin.cc
@WITH_RADOSGW_TRUE@radosgw_admin_LDADD = $(LIBRGW) $(LIBRGW_DEPS) $(CEPH_GLOBAL)
@WITH_RADOSGW_TRUE@ceph_rgw_multiparser_SOURCES = rgw/rgw_multiparser.cc
@@ -4477,11 +4519,16 @@
ceph_erasure_code_benchmark_LDADD = $(LIBOSD) $(LIBCOMMON) \
$(BOOST_PROGRAM_OPTIONS_LIBS) $(CEPH_GLOBAL) $(am__append_41)
+ceph_erasure_code_non_regression_SOURCES = \
+ test/erasure-code/ceph_erasure_code_non_regression.cc
+
+ceph_erasure_code_non_regression_LDADD = $(LIBOSD) $(LIBCOMMON) \
+ $(BOOST_PROGRAM_OPTIONS_LIBS) $(CEPH_GLOBAL) $(am__append_42)
ceph_erasure_code_SOURCES = \
test/erasure-code/ceph_erasure_code.cc
ceph_erasure_code_LDADD = $(LIBOSD) $(LIBCOMMON) \
- $(BOOST_PROGRAM_OPTIONS_LIBS) $(CEPH_GLOBAL) $(am__append_42)
+ $(BOOST_PROGRAM_OPTIONS_LIBS) $(CEPH_GLOBAL) $(am__append_43)
libec_example_la_SOURCES = test/erasure-code/ErasureCodePluginExample.cc
libec_example_la_CFLAGS = ${AM_CFLAGS}
libec_example_la_CXXFLAGS = ${AM_CXXFLAGS}
@@ -4525,7 +4572,7 @@
unittest_erasure_code_plugin_SOURCES = test/erasure-code/TestErasureCodePlugin.cc
unittest_erasure_code_plugin_CXXFLAGS = $(UNITTEST_CXXFLAGS)
unittest_erasure_code_plugin_LDADD = $(LIBOSD) $(LIBCOMMON) \
- $(UNITTEST_LDADD) $(CEPH_GLOBAL) $(am__append_43)
+ $(UNITTEST_LDADD) $(CEPH_GLOBAL) $(am__append_44)
unittest_erasure_code_jerasure_SOURCES = \
test/erasure-code/TestErasureCodeJerasure.cc \
${jerasure_sources}
@@ -4539,13 +4586,13 @@
-Ierasure-code/jerasure/jerasure/include
unittest_erasure_code_jerasure_LDADD = $(LIBOSD) $(LIBCOMMON) \
- $(UNITTEST_LDADD) $(CEPH_GLOBAL) $(am__append_44)
+ $(UNITTEST_LDADD) $(CEPH_GLOBAL) $(am__append_45)
unittest_erasure_code_plugin_jerasure_SOURCES = \
test/erasure-code/TestErasureCodePluginJerasure.cc
unittest_erasure_code_plugin_jerasure_CXXFLAGS = ${AM_CXXFLAGS} ${UNITTEST_CXXFLAGS}
unittest_erasure_code_plugin_jerasure_LDADD = $(LIBOSD) $(LIBCOMMON) \
- $(UNITTEST_LDADD) $(CEPH_GLOBAL) $(am__append_45)
+ $(UNITTEST_LDADD) $(CEPH_GLOBAL) $(am__append_46)
unittest_erasure_code_example_SOURCES = test/erasure-code/TestErasureCodeExample.cc
unittest_erasure_code_example_CXXFLAGS = $(UNITTEST_CXXFLAGS)
unittest_erasure_code_example_LDADD = $(LIBOSD) $(LIBCOMMON) $(UNITTEST_LDADD) $(CEPH_GLOBAL)
@@ -4588,8 +4635,8 @@
# These should always use explicit _CFLAGS/_CXXFLAGS so avoid basename conflicts
-ceph_dencoder_CFLAGS = ${AM_CFLAGS} $(am__append_46)
-ceph_dencoder_CXXFLAGS = ${AM_CXXFLAGS} $(am__append_47)
+ceph_dencoder_CFLAGS = ${AM_CFLAGS} $(am__append_47)
+ceph_dencoder_CXXFLAGS = ${AM_CXXFLAGS} $(am__append_48)
get_command_descriptions_SOURCES = test/common/get_command_descriptions.cc
get_command_descriptions_LDADD = $(LIBMON) $(LIBCOMMON) $(CEPH_GLOBAL)
@@ -4792,13 +4839,16 @@
unittest_pglog_SOURCES = test/osd/TestPGLog.cc
unittest_pglog_CXXFLAGS = $(UNITTEST_CXXFLAGS)
unittest_pglog_LDADD = $(LIBOSD) $(UNITTEST_LDADD) $(CEPH_GLOBAL) \
- $(am__append_51)
+ $(am__append_52)
unittest_ecbackend_SOURCES = test/osd/TestECBackend.cc
unittest_ecbackend_CXXFLAGS = $(UNITTEST_CXXFLAGS)
unittest_ecbackend_LDADD = $(LIBOSD) $(UNITTEST_LDADD) $(CEPH_GLOBAL)
unittest_hitset_SOURCES = test/osd/hitset.cc
unittest_hitset_CXXFLAGS = $(UNITTEST_CXXFLAGS)
unittest_hitset_LDADD = $(LIBOSD) $(UNITTEST_LDADD) $(CEPH_GLOBAL)
+unittest_io_priority_SOURCES = test/common/test_io_priority.cc
+unittest_io_priority_CXXFLAGS = $(UNITTEST_CXXFLAGS)
+unittest_io_priority_LDADD = $(UNITTEST_LDADD) $(CEPH_GLOBAL)
unittest_gather_SOURCES = test/gather.cc
unittest_gather_LDADD = $(UNITTEST_LDADD) $(CEPH_GLOBAL)
unittest_gather_CXXFLAGS = $(UNITTEST_CXXFLAGS)
@@ -5136,10 +5186,10 @@
ceph_kvstore_tool_CXXFLAGS = $(UNITTEST_CXXFLAGS)
ceph_filestore_tool_SOURCES = tools/ceph_filestore_tool.cc
ceph_filestore_tool_LDADD = $(LIBOSD) $(LIBOS) $(CEPH_GLOBAL) \
- -lboost_program_options $(am__append_56)
+ -lboost_program_options $(am__append_57)
ceph_filestore_dump_SOURCES = tools/ceph_filestore_dump.cc
ceph_filestore_dump_LDADD = $(LIBOSD) $(LIBOS) $(CEPH_GLOBAL) \
- $(BOOST_PROGRAM_OPTIONS_LIBS) $(am__append_57)
+ $(BOOST_PROGRAM_OPTIONS_LIBS) $(am__append_58)
monmaptool_SOURCES = tools/monmaptool.cc
monmaptool_LDADD = $(CEPH_GLOBAL) $(LIBCOMMON)
crushtool_SOURCES = tools/crushtool.cc
@@ -5167,7 +5217,7 @@
@WITH_REST_BENCH_TRUE@ common/obj_bencher.cc # needs cleanup so \
@WITH_REST_BENCH_TRUE@ it can go in libcommon.la
@WITH_REST_BENCH_TRUE@rest_bench_LDADD = $(CEPH_GLOBAL) \
-@WITH_REST_BENCH_TRUE@ $(am__append_59) $(am__append_60)
+@WITH_REST_BENCH_TRUE@ $(am__append_60) $(am__append_61)
@WITH_REST_BENCH_TRUE@@WITH_SYSTEM_LIBS3_FALSE@rest_bench_CXXFLAGS = ${AM_CXXFLAGS} -I$(top_srcdir)/src/libs3/inc
ceph_conf_SOURCES = tools/ceph_conf.cc
ceph_conf_LDADD = $(CEPH_GLOBAL) $(LIBCOMMON)
@@ -5600,6 +5650,26 @@
java/native/$(DEPDIR)/$(am__dirstamp)
libcephfs_jni.la: $(libcephfs_jni_la_OBJECTS) $(libcephfs_jni_la_DEPENDENCIES) $(EXTRA_libcephfs_jni_la_DEPENDENCIES)
$(AM_V_CXXLD)$(libcephfs_jni_la_LINK) $(am_libcephfs_jni_la_rpath) $(libcephfs_jni_la_OBJECTS) $(libcephfs_jni_la_LIBADD) $(LIBS)
+rgw/$(am__dirstamp):
+ @$(MKDIR_P) rgw
+ @: > rgw/$(am__dirstamp)
+rgw/$(DEPDIR)/$(am__dirstamp):
+ @$(MKDIR_P) rgw/$(DEPDIR)
+ @: > rgw/$(DEPDIR)/$(am__dirstamp)
+rgw/libcivetweb_la-rgw_civetweb.lo: rgw/$(am__dirstamp) \
+ rgw/$(DEPDIR)/$(am__dirstamp)
+rgw/libcivetweb_la-rgw_civetweb_log.lo: rgw/$(am__dirstamp) \
+ rgw/$(DEPDIR)/$(am__dirstamp)
+civetweb/src/$(am__dirstamp):
+ @$(MKDIR_P) civetweb/src
+ @: > civetweb/src/$(am__dirstamp)
+civetweb/src/$(DEPDIR)/$(am__dirstamp):
+ @$(MKDIR_P) civetweb/src/$(DEPDIR)
+ @: > civetweb/src/$(DEPDIR)/$(am__dirstamp)
+civetweb/src/libcivetweb_la-civetweb.lo: civetweb/src/$(am__dirstamp) \
+ civetweb/src/$(DEPDIR)/$(am__dirstamp)
+libcivetweb.la: $(libcivetweb_la_OBJECTS) $(libcivetweb_la_DEPENDENCIES) $(EXTRA_libcivetweb_la_DEPENDENCIES)
+ $(AM_V_CXXLD)$(libcivetweb_la_LINK) $(am_libcivetweb_la_rpath) $(libcivetweb_la_OBJECTS) $(libcivetweb_la_LIBADD) $(LIBS)
client/$(am__dirstamp):
@$(MKDIR_P) client
@: > client/$(am__dirstamp)
@@ -5907,6 +5977,7 @@
mds/inode_backtrace.lo: mds/$(am__dirstamp) \
mds/$(DEPDIR)/$(am__dirstamp)
mds/mdstypes.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
+mds/flock.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
libcommon.la: $(libcommon_la_OBJECTS) $(libcommon_la_DEPENDENCIES) $(EXTRA_libcommon_la_DEPENDENCIES)
$(AM_V_CXXLD)$(CXXLINK) $(libcommon_la_OBJECTS) $(libcommon_la_LIBADD) $(LIBS)
common/libcommon_crc_la-sctp_crc32.lo: common/$(am__dirstamp) \
@@ -6232,7 +6303,6 @@
mds/Dumper.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
mds/Resetter.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
mds/MDS.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
-mds/flock.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
mds/locks.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
mds/journal.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
mds/Server.lo: mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
@@ -6477,12 +6547,6 @@
librbd/$(DEPDIR)/$(am__dirstamp)
librbd.la: $(librbd_la_OBJECTS) $(librbd_la_DEPENDENCIES) $(EXTRA_librbd_la_DEPENDENCIES)
$(AM_V_CXXLD)$(librbd_la_LINK) -rpath $(libdir) $(librbd_la_OBJECTS) $(librbd_la_LIBADD) $(LIBS)
-rgw/$(am__dirstamp):
- @$(MKDIR_P) rgw
- @: > rgw/$(am__dirstamp)
-rgw/$(DEPDIR)/$(am__dirstamp):
- @$(MKDIR_P) rgw/$(DEPDIR)
- @: > rgw/$(DEPDIR)/$(am__dirstamp)
rgw/librgw_la-librgw.lo: rgw/$(am__dirstamp) \
rgw/$(DEPDIR)/$(am__dirstamp)
rgw/librgw_la-rgw_acl.lo: rgw/$(am__dirstamp) \
@@ -6817,6 +6881,12 @@
ceph_erasure_code_benchmark$(EXEEXT): $(ceph_erasure_code_benchmark_OBJECTS) $(ceph_erasure_code_benchmark_DEPENDENCIES) $(EXTRA_ceph_erasure_code_benchmark_DEPENDENCIES)
@rm -f ceph_erasure_code_benchmark$(EXEEXT)
$(AM_V_CXXLD)$(CXXLINK) $(ceph_erasure_code_benchmark_OBJECTS) $(ceph_erasure_code_benchmark_LDADD) $(LIBS)
+test/erasure-code/ceph_erasure_code_non_regression.$(OBJEXT): \
+ test/erasure-code/$(am__dirstamp) \
+ test/erasure-code/$(DEPDIR)/$(am__dirstamp)
+ceph_erasure_code_non_regression$(EXEEXT): $(ceph_erasure_code_non_regression_OBJECTS) $(ceph_erasure_code_non_regression_DEPENDENCIES) $(EXTRA_ceph_erasure_code_non_regression_DEPENDENCIES)
+ @rm -f ceph_erasure_code_non_regression$(EXEEXT)
+ $(AM_V_CXXLD)$(CXXLINK) $(ceph_erasure_code_non_regression_OBJECTS) $(ceph_erasure_code_non_regression_LDADD) $(LIBS)
tools/ceph_filestore_dump.$(OBJEXT): tools/$(am__dirstamp) \
tools/$(DEPDIR)/$(am__dirstamp)
ceph_filestore_dump$(EXEEXT): $(ceph_filestore_dump_OBJECTS) $(ceph_filestore_dump_DEPENDENCIES) $(EXTRA_ceph_filestore_dump_DEPENDENCIES)
@@ -7533,16 +7603,6 @@
rgw/$(DEPDIR)/$(am__dirstamp)
rgw/rgw_loadgen.$(OBJEXT): rgw/$(am__dirstamp) \
rgw/$(DEPDIR)/$(am__dirstamp)
-rgw/rgw_civetweb.$(OBJEXT): rgw/$(am__dirstamp) \
- rgw/$(DEPDIR)/$(am__dirstamp)
-civetweb/src/$(am__dirstamp):
- @$(MKDIR_P) civetweb/src
- @: > civetweb/src/$(am__dirstamp)
-civetweb/src/$(DEPDIR)/$(am__dirstamp):
- @$(MKDIR_P) civetweb/src/$(DEPDIR)
- @: > civetweb/src/$(DEPDIR)/$(am__dirstamp)
-civetweb/src/radosgw-civetweb.$(OBJEXT): civetweb/src/$(am__dirstamp) \
- civetweb/src/$(DEPDIR)/$(am__dirstamp)
rgw/rgw_main.$(OBJEXT): rgw/$(am__dirstamp) \
rgw/$(DEPDIR)/$(am__dirstamp)
radosgw$(EXEEXT): $(radosgw_OBJECTS) $(radosgw_DEPENDENCIES) $(EXTRA_radosgw_DEPENDENCIES)
@@ -7753,6 +7813,8 @@
mds/$(am__dirstamp) mds/$(DEPDIR)/$(am__dirstamp)
mds/test_build_libcommon-mdstypes.$(OBJEXT): mds/$(am__dirstamp) \
mds/$(DEPDIR)/$(am__dirstamp)
+mds/test_build_libcommon-flock.$(OBJEXT): mds/$(am__dirstamp) \
+ mds/$(DEPDIR)/$(am__dirstamp)
test_build_libcommon$(EXEEXT): $(test_build_libcommon_OBJECTS) $(test_build_libcommon_DEPENDENCIES) $(EXTRA_test_build_libcommon_DEPENDENCIES)
@rm -f test_build_libcommon$(EXEEXT)
$(AM_V_CXXLD)$(test_build_libcommon_LINK) $(test_build_libcommon_OBJECTS) $(test_build_libcommon_LDADD) $(LIBS)
@@ -8082,6 +8144,12 @@
unittest_hitset$(EXEEXT): $(unittest_hitset_OBJECTS) $(unittest_hitset_DEPENDENCIES) $(EXTRA_unittest_hitset_DEPENDENCIES)
@rm -f unittest_hitset$(EXEEXT)
$(AM_V_CXXLD)$(unittest_hitset_LINK) $(unittest_hitset_OBJECTS) $(unittest_hitset_LDADD) $(LIBS)
+test/common/unittest_io_priority-test_io_priority.$(OBJEXT): \
+ test/common/$(am__dirstamp) \
+ test/common/$(DEPDIR)/$(am__dirstamp)
+unittest_io_priority$(EXEEXT): $(unittest_io_priority_OBJECTS) $(unittest_io_priority_DEPENDENCIES) $(EXTRA_unittest_io_priority_DEPENDENCIES)
+ @rm -f unittest_io_priority$(EXEEXT)
+ $(AM_V_CXXLD)$(unittest_io_priority_LINK) $(unittest_io_priority_OBJECTS) $(unittest_io_priority_LDADD) $(LIBS)
test/unittest_ipaddr-test_ipaddr.$(OBJEXT): test/$(am__dirstamp) \
test/$(DEPDIR)/$(am__dirstamp)
unittest_ipaddr$(EXEEXT): $(unittest_ipaddr_OBJECTS) $(unittest_ipaddr_DEPENDENCIES) $(EXTRA_unittest_ipaddr_DEPENDENCIES)
@@ -8472,7 +8540,8 @@
-rm -f auth/none/AuthNoneAuthorizeHandler.lo
-rm -f auth/unknown/AuthUnknownAuthorizeHandler.$(OBJEXT)
-rm -f auth/unknown/AuthUnknownAuthorizeHandler.lo
- -rm -f civetweb/src/radosgw-civetweb.$(OBJEXT)
+ -rm -f civetweb/src/libcivetweb_la-civetweb.$(OBJEXT)
+ -rm -f civetweb/src/libcivetweb_la-civetweb.lo
-rm -f client/Client.$(OBJEXT)
-rm -f client/Client.lo
-rm -f client/ClientSnapRealm.$(OBJEXT)
@@ -9032,6 +9101,7 @@
-rm -f mds/snap.$(OBJEXT)
-rm -f mds/snap.lo
-rm -f mds/test_build_libcommon-MDSMap.$(OBJEXT)
+ -rm -f mds/test_build_libcommon-flock.$(OBJEXT)
-rm -f mds/test_build_libcommon-inode_backtrace.$(OBJEXT)
-rm -f mds/test_build_libcommon-mdstypes.$(OBJEXT)
-rm -f mon/AuthMonitor.$(OBJEXT)
@@ -9209,6 +9279,10 @@
-rm -f rgw/ceph_dencoder-rgw_dencoder.$(OBJEXT)
-rm -f rgw/ceph_dencoder-rgw_env.$(OBJEXT)
-rm -f rgw/ceph_dencoder-rgw_json_enc.$(OBJEXT)
+ -rm -f rgw/libcivetweb_la-rgw_civetweb.$(OBJEXT)
+ -rm -f rgw/libcivetweb_la-rgw_civetweb.lo
+ -rm -f rgw/libcivetweb_la-rgw_civetweb_log.$(OBJEXT)
+ -rm -f rgw/libcivetweb_la-rgw_civetweb_log.lo
-rm -f rgw/librgw_la-librgw.$(OBJEXT)
-rm -f rgw/librgw_la-librgw.lo
-rm -f rgw/librgw_la-rgw_acl.$(OBJEXT)
@@ -9278,7 +9352,6 @@
-rm -f rgw/librgw_la-rgw_xml.$(OBJEXT)
-rm -f rgw/librgw_la-rgw_xml.lo
-rm -f rgw/rgw_admin.$(OBJEXT)
- -rm -f rgw/rgw_civetweb.$(OBJEXT)
-rm -f rgw/rgw_common.$(OBJEXT)
-rm -f rgw/rgw_env.$(OBJEXT)
-rm -f rgw/rgw_http_client.$(OBJEXT)
@@ -9380,6 +9453,7 @@
-rm -f test/common/unittest_context-test_context.$(OBJEXT)
-rm -f test/common/unittest_crc32c-test_crc32c.$(OBJEXT)
-rm -f test/common/unittest_histogram-histogram.$(OBJEXT)
+ -rm -f test/common/unittest_io_priority-test_io_priority.$(OBJEXT)
-rm -f test/common/unittest_sharedptr_registry-test_sharedptr_registry.$(OBJEXT)
-rm -f test/common/unittest_sloppy_crc_map-test_sloppy_crc_map.$(OBJEXT)
-rm -f test/common/unittest_str_map-test_str_map.$(OBJEXT)
@@ -9390,6 +9464,7 @@
-rm -f test/encoding/ceph_dencoder-ceph_dencoder.$(OBJEXT)
-rm -f test/erasure-code/ceph_erasure_code.$(OBJEXT)
-rm -f test/erasure-code/ceph_erasure_code_benchmark.$(OBJEXT)
+ -rm -f test/erasure-code/ceph_erasure_code_non_regression.$(OBJEXT)
-rm -f test/erasure-code/libec_example_la-ErasureCodePluginExample.$(OBJEXT)
-rm -f test/erasure-code/libec_example_la-ErasureCodePluginExample.lo
-rm -f test/erasure-code/libec_fail_to_initialize_la-ErasureCodePluginFailToInitialize.$(OBJEXT)
@@ -9584,7 +9659,7 @@
@AMDEP_TRUE@@am__include@ @am__quote@auth/cephx/$(DEPDIR)/CephxSessionHandler.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@auth/none/$(DEPDIR)/AuthNoneAuthorizeHandler.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@auth/unknown/$(DEPDIR)/AuthUnknownAuthorizeHandler.Plo@am__quote@
-@AMDEP_TRUE@@am__include@ @am__quote@civetweb/src/$(DEPDIR)/radosgw-civetweb.Po@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@civetweb/src/$(DEPDIR)/libcivetweb_la-civetweb.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@client/$(DEPDIR)/Client.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@client/$(DEPDIR)/ClientSnapRealm.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@client/$(DEPDIR)/Dentry.Plo@am__quote@
@@ -9920,6 +9995,7 @@
@AMDEP_TRUE@@am__include@ @am__quote@mds/$(DEPDIR)/mdstypes.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@mds/$(DEPDIR)/snap.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@mds/$(DEPDIR)/test_build_libcommon-MDSMap.Po@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@mds/$(DEPDIR)/test_build_libcommon-flock.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@mds/$(DEPDIR)/test_build_libcommon-inode_backtrace.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@mds/$(DEPDIR)/test_build_libcommon-mdstypes.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@mon/$(DEPDIR)/AuthMonitor.Plo@am__quote@
@@ -10020,6 +10096,8 @@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/ceph_dencoder-rgw_dencoder.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/ceph_dencoder-rgw_env.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/ceph_dencoder-rgw_json_enc.Po@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb_log.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/librgw_la-librgw.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/librgw_la-rgw_acl.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/librgw_la-rgw_acl_s3.Plo@am__quote@
@@ -10055,7 +10133,6 @@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/librgw_la-rgw_user.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/librgw_la-rgw_xml.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/rgw_admin.Po@am__quote@
-@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/rgw_civetweb.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/rgw_common.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/rgw_env.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@rgw/$(DEPDIR)/rgw_http_client.Po@am__quote@
@@ -10206,6 +10283,7 @@
@AMDEP_TRUE@@am__include@ @am__quote@test/common/$(DEPDIR)/unittest_context-test_context.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/common/$(DEPDIR)/unittest_crc32c-test_crc32c.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/common/$(DEPDIR)/unittest_histogram-histogram.Po@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@test/common/$(DEPDIR)/unittest_io_priority-test_io_priority.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/common/$(DEPDIR)/unittest_sharedptr_registry-test_sharedptr_registry.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/common/$(DEPDIR)/unittest_sloppy_crc_map-test_sloppy_crc_map.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/common/$(DEPDIR)/unittest_str_map-test_str_map.Po@am__quote@
@@ -10216,6 +10294,7 @@
@AMDEP_TRUE@@am__include@ @am__quote@test/encoding/$(DEPDIR)/ceph_dencoder-ceph_dencoder.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/erasure-code/$(DEPDIR)/ceph_erasure_code.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/erasure-code/$(DEPDIR)/ceph_erasure_code_benchmark.Po@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@test/erasure-code/$(DEPDIR)/ceph_erasure_code_non_regression.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/erasure-code/$(DEPDIR)/libec_example_la-ErasureCodePluginExample.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/erasure-code/$(DEPDIR)/libec_fail_to_initialize_la-ErasureCodePluginFailToInitialize.Plo@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@test/erasure-code/$(DEPDIR)/libec_fail_to_register_la-ErasureCodePluginFailToRegister.Plo@am__quote@
@@ -10377,6 +10456,13 @@
@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
@am__fastdepCC_FALSE@ $(AM_V_CC@am__nodep@)$(LTCOMPILE) -c -o $@ $<
+civetweb/src/libcivetweb_la-civetweb.lo: civetweb/src/civetweb.c
+@am__fastdepCC_TRUE@ $(AM_V_CC)$(LIBTOOL) $(AM_V_lt) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(libcivetweb_la_CFLAGS) $(CFLAGS) -MT civetweb/src/libcivetweb_la-civetweb.lo -MD -MP -MF civetweb/src/$(DEPDIR)/libcivetweb_la-civetweb.Tpo -c -o civetweb/src/libcivetweb_la-civetweb.lo `test -f 'civetweb/src/civetweb.c' || echo '$(srcdir)/'`civetweb/src/civetweb.c
+@am__fastdepCC_TRUE@ $(AM_V_at)$(am__mv) civetweb/src/$(DEPDIR)/libcivetweb_la-civetweb.Tpo civetweb/src/$(DEPDIR)/libcivetweb_la-civetweb.Plo
+@AMDEP_TRUE@@am__fastdepCC_FALSE@ $(AM_V_CC)source='civetweb/src/civetweb.c' object='civetweb/src/libcivetweb_la-civetweb.lo' libtool=yes @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCC_FALSE@ $(AM_V_CC@am__nodep@)$(LIBTOOL) $(AM_V_lt) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(libcivetweb_la_CFLAGS) $(CFLAGS) -c -o civetweb/src/libcivetweb_la-civetweb.lo `test -f 'civetweb/src/civetweb.c' || echo '$(srcdir)/'`civetweb/src/civetweb.c
+
common/libcommon_crc_la-sctp_crc32.lo: common/sctp_crc32.c
@am__fastdepCC_TRUE@ $(AM_V_CC)$(LIBTOOL) $(AM_V_lt) --tag=CC $(libcommon_crc_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT common/libcommon_crc_la-sctp_crc32.lo -MD -MP -MF common/$(DEPDIR)/libcommon_crc_la-sctp_crc32.Tpo -c -o common/libcommon_crc_la-sctp_crc32.lo `test -f 'common/sctp_crc32.c' || echo '$(srcdir)/'`common/sctp_crc32.c
@am__fastdepCC_TRUE@ $(AM_V_at)$(am__mv) common/$(DEPDIR)/libcommon_crc_la-sctp_crc32.Tpo common/$(DEPDIR)/libcommon_crc_la-sctp_crc32.Plo
@@ -10748,20 +10834,6 @@
@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
@am__fastdepCC_FALSE@ $(AM_V_CC@am__nodep@)$(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(ceph_test_librbd_fsx_CFLAGS) $(CFLAGS) -c -o test/librbd/ceph_test_librbd_fsx-fsx.obj `if test -f 'test/librbd/fsx.c'; then $(CYGPATH_W) 'test/librbd/fsx.c'; else $(CYGPATH_W) '$(srcdir)/test/librbd/fsx.c'; fi`
-civetweb/src/radosgw-civetweb.o: civetweb/src/civetweb.c
-@am__fastdepCC_TRUE@ $(AM_V_CC)$(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(radosgw_CFLAGS) $(CFLAGS) -MT civetweb/src/radosgw-civetweb.o -MD -MP -MF civetweb/src/$(DEPDIR)/radosgw-civetweb.Tpo -c -o civetweb/src/radosgw-civetweb.o `test -f 'civetweb/src/civetweb.c' || echo '$(srcdir)/'`civetweb/src/civetweb.c
-@am__fastdepCC_TRUE@ $(AM_V_at)$(am__mv) civetweb/src/$(DEPDIR)/radosgw-civetweb.Tpo civetweb/src/$(DEPDIR)/radosgw-civetweb.Po
-@AMDEP_TRUE@@am__fastdepCC_FALSE@ $(AM_V_CC)source='civetweb/src/civetweb.c' object='civetweb/src/radosgw-civetweb.o' libtool=no @AMDEPBACKSLASH@
-@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
-@am__fastdepCC_FALSE@ $(AM_V_CC@am__nodep@)$(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(radosgw_CFLAGS) $(CFLAGS) -c -o civetweb/src/radosgw-civetweb.o `test -f 'civetweb/src/civetweb.c' || echo '$(srcdir)/'`civetweb/src/civetweb.c
-
-civetweb/src/radosgw-civetweb.obj: civetweb/src/civetweb.c
-@am__fastdepCC_TRUE@ $(AM_V_CC)$(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(radosgw_CFLAGS) $(CFLAGS) -MT civetweb/src/radosgw-civetweb.obj -MD -MP -MF civetweb/src/$(DEPDIR)/radosgw-civetweb.Tpo -c -o civetweb/src/radosgw-civetweb.obj `if test -f 'civetweb/src/civetweb.c'; then $(CYGPATH_W) 'civetweb/src/civetweb.c'; else $(CYGPATH_W) '$(srcdir)/civetweb/src/civetweb.c'; fi`
-@am__fastdepCC_TRUE@ $(AM_V_at)$(am__mv) civetweb/src/$(DEPDIR)/radosgw-civetweb.Tpo civetweb/src/$(DEPDIR)/radosgw-civetweb.Po
-@AMDEP_TRUE@@am__fastdepCC_FALSE@ $(AM_V_CC)source='civetweb/src/civetweb.c' object='civetweb/src/radosgw-civetweb.obj' libtool=no @AMDEPBACKSLASH@
-@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@
-@am__fastdepCC_FALSE@ $(AM_V_CC@am__nodep@)$(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(radosgw_CFLAGS) $(CFLAGS) -c -o civetweb/src/radosgw-civetweb.obj `if test -f 'civetweb/src/civetweb.c'; then $(CYGPATH_W) 'civetweb/src/civetweb.c'; else $(CYGPATH_W) '$(srcdir)/civetweb/src/civetweb.c'; fi`
-
test_build_libcommon-ceph_ver.o: ceph_ver.c
@am__fastdepCC_TRUE@ $(AM_V_CC)$(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(test_build_libcommon_CFLAGS) $(CFLAGS) -MT test_build_libcommon-ceph_ver.o -MD -MP -MF $(DEPDIR)/test_build_libcommon-ceph_ver.Tpo -c -o test_build_libcommon-ceph_ver.o `test -f 'ceph_ver.c' || echo '$(srcdir)/'`ceph_ver.c
@am__fastdepCC_TRUE@ $(AM_V_at)$(am__mv) $(DEPDIR)/test_build_libcommon-ceph_ver.Tpo $(DEPDIR)/test_build_libcommon-ceph_ver.Po
@@ -11178,6 +11250,20 @@
@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(LIBTOOL) $(AM_V_lt) --tag=CXX $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(libcephfs_jni_la_CPPFLAGS) $(CPPFLAGS) $(AM_CXXFLAGS) $(CXXFLAGS) -c -o java/native/libcephfs_jni_la-JniConstants.lo `test -f 'java/native/JniConstants.cpp' || echo '$(srcdir)/'`java/native/JniConstants.cpp
+rgw/libcivetweb_la-rgw_civetweb.lo: rgw/rgw_civetweb.cc
+@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(LIBTOOL) $(AM_V_lt) --tag=CXX $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(libcivetweb_la_CXXFLAGS) $(CXXFLAGS) -MT rgw/libcivetweb_la-rgw_civetweb.lo -MD -MP -MF rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb.Tpo -c -o rgw/libcivetweb_la-rgw_civetweb.lo `test -f 'rgw/rgw_civetweb.cc' || echo '$(srcdir)/'`rgw/rgw_civetweb.cc
+@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb.Tpo rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb.Plo
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ $(AM_V_CXX)source='rgw/rgw_civetweb.cc' object='rgw/libcivetweb_la-rgw_civetweb.lo' libtool=yes @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(LIBTOOL) $(AM_V_lt) --tag=CXX $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(libcivetweb_la_CXXFLAGS) $(CXXFLAGS) -c -o rgw/libcivetweb_la-rgw_civetweb.lo `test -f 'rgw/rgw_civetweb.cc' || echo '$(srcdir)/'`rgw/rgw_civetweb.cc
+
+rgw/libcivetweb_la-rgw_civetweb_log.lo: rgw/rgw_civetweb_log.cc
+@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(LIBTOOL) $(AM_V_lt) --tag=CXX $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(libcivetweb_la_CXXFLAGS) $(CXXFLAGS) -MT rgw/libcivetweb_la-rgw_civetweb_log.lo -MD -MP -MF rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb_log.Tpo -c -o rgw/libcivetweb_la-rgw_civetweb_log.lo `test -f 'rgw/rgw_civetweb_log.cc' || echo '$(srcdir)/'`rgw/rgw_civetweb_log.cc
+@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb_log.Tpo rgw/$(DEPDIR)/libcivetweb_la-rgw_civetweb_log.Plo
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ $(AM_V_CXX)source='rgw/rgw_civetweb_log.cc' object='rgw/libcivetweb_la-rgw_civetweb_log.lo' libtool=yes @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(LIBTOOL) $(AM_V_lt) --tag=CXX $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(libcivetweb_la_CXXFLAGS) $(CXXFLAGS) -c -o rgw/libcivetweb_la-rgw_civetweb_log.lo `test -f 'rgw/rgw_civetweb_log.cc' || echo '$(srcdir)/'`rgw/rgw_civetweb_log.cc
+
common/libcommon_crc_la-crc32c.lo: common/crc32c.cc
@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(LIBTOOL) $(AM_V_lt) --tag=CXX $(libcommon_crc_la_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CXXFLAGS) $(CXXFLAGS) -MT common/libcommon_crc_la-crc32c.lo -MD -MP -MF common/$(DEPDIR)/libcommon_crc_la-crc32c.Tpo -c -o common/libcommon_crc_la-crc32c.lo `test -f 'common/crc32c.cc' || echo '$(srcdir)/'`common/crc32c.cc
@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) common/$(DEPDIR)/libcommon_crc_la-crc32c.Tpo common/$(DEPDIR)/libcommon_crc_la-crc32c.Plo
@@ -13705,6 +13791,20 @@
@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(test_build_libcommon_CXXFLAGS) $(CXXFLAGS) -c -o mds/test_build_libcommon-mdstypes.obj `if test -f 'mds/mdstypes.cc'; then $(CYGPATH_W) 'mds/mdstypes.cc'; else $(CYGPATH_W) '$(srcdir)/mds/mdstypes.cc'; fi`
+mds/test_build_libcommon-flock.o: mds/flock.cc
+@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(test_build_libcommon_CXXFLAGS) $(CXXFLAGS) -MT mds/test_build_libcommon-flock.o -MD -MP -MF mds/$(DEPDIR)/test_build_libcommon-flock.Tpo -c -o mds/test_build_libcommon-flock.o `test -f 'mds/flock.cc' || echo '$(srcdir)/'`mds/flock.cc
+@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) mds/$(DEPDIR)/test_build_libcommon-flock.Tpo mds/$(DEPDIR)/test_build_libcommon-flock.Po
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ $(AM_V_CXX)source='mds/flock.cc' object='mds/test_build_libcommon-flock.o' libtool=no @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(test_build_libcommon_CXXFLAGS) $(CXXFLAGS) -c -o mds/test_build_libcommon-flock.o `test -f 'mds/flock.cc' || echo '$(srcdir)/'`mds/flock.cc
+
+mds/test_build_libcommon-flock.obj: mds/flock.cc
+@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(test_build_libcommon_CXXFLAGS) $(CXXFLAGS) -MT mds/test_build_libcommon-flock.obj -MD -MP -MF mds/$(DEPDIR)/test_build_libcommon-flock.Tpo -c -o mds/test_build_libcommon-flock.obj `if test -f 'mds/flock.cc'; then $(CYGPATH_W) 'mds/flock.cc'; else $(CYGPATH_W) '$(srcdir)/mds/flock.cc'; fi`
+@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) mds/$(DEPDIR)/test_build_libcommon-flock.Tpo mds/$(DEPDIR)/test_build_libcommon-flock.Po
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ $(AM_V_CXX)source='mds/flock.cc' object='mds/test_build_libcommon-flock.obj' libtool=no @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(test_build_libcommon_CXXFLAGS) $(CXXFLAGS) -c -o mds/test_build_libcommon-flock.obj `if test -f 'mds/flock.cc'; then $(CYGPATH_W) 'mds/flock.cc'; else $(CYGPATH_W) '$(srcdir)/mds/flock.cc'; fi`
+
test/test_build_librados-buildtest_skeleton.o: test/buildtest_skeleton.cc
@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(test_build_librados_CXXFLAGS) $(CXXFLAGS) -MT test/test_build_librados-buildtest_skeleton.o -MD -MP -MF test/$(DEPDIR)/test_build_librados-buildtest_skeleton.Tpo -c -o test/test_build_librados-buildtest_skeleton.o `test -f 'test/buildtest_skeleton.cc' || echo '$(srcdir)/'`test/buildtest_skeleton.cc
@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) test/$(DEPDIR)/test_build_librados-buildtest_skeleton.Tpo test/$(DEPDIR)/test_build_librados-buildtest_skeleton.Po
@@ -14755,6 +14855,20 @@
@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(unittest_hitset_CXXFLAGS) $(CXXFLAGS) -c -o test/osd/unittest_hitset-hitset.obj `if test -f 'test/osd/hitset.cc'; then $(CYGPATH_W) 'test/osd/hitset.cc'; else $(CYGPATH_W) '$(srcdir)/test/osd/hitset.cc'; fi`
+test/common/unittest_io_priority-test_io_priority.o: test/common/test_io_priority.cc
+@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(unittest_io_priority_CXXFLAGS) $(CXXFLAGS) -MT test/common/unittest_io_priority-test_io_priority.o -MD -MP -MF test/common/$(DEPDIR)/unittest_io_priority-test_io_priority.Tpo -c -o test/common/unittest_io_priority-test_io_priority.o `test -f 'test/common/test_io_priority.cc' || echo '$(srcdir)/'`test/common/test_io_priority.cc
+@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) test/common/$(DEPDIR)/unittest_io_priority-test_io_priority.Tpo test/common/$(DEPDIR)/unittest_io_priority-test_io_priority.Po
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ $(AM_V_CXX)source='test/common/test_io_priority.cc' object='test/common/unittest_io_priority-test_io_priority.o' libtool=no @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(unittest_io_priority_CXXFLAGS) $(CXXFLAGS) -c -o test/common/unittest_io_priority-test_io_priority.o `test -f 'test/common/test_io_priority.cc' || echo '$(srcdir)/'`test/common/test_io_priority.cc
+
+test/common/unittest_io_priority-test_io_priority.obj: test/common/test_io_priority.cc
+@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(unittest_io_priority_CXXFLAGS) $(CXXFLAGS) -MT test/common/unittest_io_priority-test_io_priority.obj -MD -MP -MF test/common/$(DEPDIR)/unittest_io_priority-test_io_priority.Tpo -c -o test/common/unittest_io_priority-test_io_priority.obj `if test -f 'test/common/test_io_priority.cc'; then $(CYGPATH_W) 'test/common/test_io_priority.cc'; else $(CYGPATH_W) '$(srcdir)/test/common/test_io_priority.cc'; fi`
+@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) test/common/$(DEPDIR)/unittest_io_priority-test_io_priority.Tpo test/common/$(DEPDIR)/unittest_io_priority-test_io_priority.Po
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ $(AM_V_CXX)source='test/common/test_io_priority.cc' object='test/common/unittest_io_priority-test_io_priority.obj' libtool=no @AMDEPBACKSLASH@
+@AMDEP_TRUE@@am__fastdepCXX_FALSE@ DEPDIR=$(DEPDIR) $(CXXDEPMODE) $(depcomp) @AMDEPBACKSLASH@
+@am__fastdepCXX_FALSE@ $(AM_V_CXX@am__nodep@)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(unittest_io_priority_CXXFLAGS) $(CXXFLAGS) -c -o test/common/unittest_io_priority-test_io_priority.obj `if test -f 'test/common/test_io_priority.cc'; then $(CYGPATH_W) 'test/common/test_io_priority.cc'; else $(CYGPATH_W) '$(srcdir)/test/common/test_io_priority.cc'; fi`
+
test/unittest_ipaddr-test_ipaddr.o: test/test_ipaddr.cc
@am__fastdepCXX_TRUE@ $(AM_V_CXX)$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(unittest_ipaddr_CXXFLAGS) $(CXXFLAGS) -MT test/unittest_ipaddr-test_ipaddr.o -MD -MP -MF test/$(DEPDIR)/unittest_ipaddr-test_ipaddr.Tpo -c -o test/unittest_ipaddr-test_ipaddr.o `test -f 'test/test_ipaddr.cc' || echo '$(srcdir)/'`test/test_ipaddr.cc
@am__fastdepCXX_TRUE@ $(AM_V_at)$(am__mv) test/$(DEPDIR)/unittest_ipaddr-test_ipaddr.Tpo test/$(DEPDIR)/unittest_ipaddr-test_ipaddr.Po
@@ -15195,6 +15309,7 @@
-rm -rf auth/cephx/.libs auth/cephx/_libs
-rm -rf auth/none/.libs auth/none/_libs
-rm -rf auth/unknown/.libs auth/unknown/_libs
+ -rm -rf civetweb/src/.libs civetweb/src/_libs
-rm -rf client/.libs client/_libs
-rm -rf cls/hello/.libs cls/hello/_libs
-rm -rf cls/lock/.libs cls/lock/_libs
@@ -15994,6 +16109,7 @@
docdir ?= ${datadir}/doc/ceph
check-local:
+ $(top_srcdir)/qa/workunits/erasure-code/encode-decode-non-regression.sh
$(srcdir)/test/encoding/readable.sh ../ceph-object-corpus
# base targets
diff -Nru ceph-0.80.7/src/mds/CInode.h ceph-0.80.9/src/mds/CInode.h
--- ceph-0.80.7/src/mds/CInode.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/CInode.h 2015-03-09 20:51:17.000000000 +0100
@@ -442,6 +442,7 @@
parent(0),
inode_auth(CDIR_AUTH_DEFAULT),
replica_caps_wanted(0),
+ fcntl_locks(g_ceph_context), flock_locks(g_ceph_context),
item_dirty(this), item_caps(this), item_open_file(this), item_dirty_parent(this),
item_dirty_dirfrag_dir(this),
item_dirty_dirfrag_nest(this),
diff -Nru ceph-0.80.7/src/mds/Dumper.cc ceph-0.80.9/src/mds/Dumper.cc
--- ceph-0.80.7/src/mds/Dumper.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/Dumper.cc 2015-03-09 20:51:17.000000000 +0100
@@ -160,7 +160,7 @@
inodeno_t ino = MDS_INO_LOG_OFFSET + rank;
Journaler::Header h;
- h.trimmed_pos = start;
+ h.trimmed_pos = start - (start % g_default_file_layout.fl_object_size);
h.expire_pos = start;
h.write_pos = start+len;
h.magic = CEPH_FS_ONDISK_MAGIC;
@@ -175,18 +175,14 @@
object_locator_t oloc(mdsmap->get_metadata_pool());
SnapContext snapc;
- bool done = false;
- Cond cond;
-
cout << "writing header " << oid << std::endl;
+ C_SaferCond header_cond;
+ lock.Lock();
objecter->write_full(oid, oloc, snapc, hbl, ceph_clock_now(g_ceph_context), 0,
NULL,
- new C_SafeCond(&lock, &cond, &done));
-
- lock.Lock();
- while (!done)
- cond.Wait(lock);
+ &header_cond);
lock.Unlock();
+ header_cond.wait();
// read
Filer filer(objecter);
@@ -198,13 +194,12 @@
uint64_t l = MIN(left, 1024*1024);
j.read_fd(fd, l);
cout << " writing " << pos << "~" << l << std::endl;
- filer.write(ino, &h.layout, snapc, pos, l, j, ceph_clock_now(g_ceph_context), 0, NULL, new C_SafeCond(&lock, &cond, &done));
-
+ C_SaferCond body_cond;
lock.Lock();
- while (!done)
- cond.Wait(lock);
+ filer.write(ino, &h.layout, snapc, pos, l, j, ceph_clock_now(g_ceph_context), 0, NULL, &body_cond);
lock.Unlock();
-
+ body_cond.wait();
+
pos += l;
left -= l;
}
diff -Nru ceph-0.80.7/src/mds/flock.cc ceph-0.80.9/src/mds/flock.cc
--- ceph-0.80.7/src/mds/flock.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/flock.cc 2015-03-09 20:51:18.000000000 +0100
@@ -44,33 +44,33 @@
bool ceph_lock_state_t::add_lock(ceph_filelock& new_lock,
bool wait_on_fail, bool replay)
{
- dout(15) << "add_lock " << new_lock << dendl;
+ ldout(cct,15) << "add_lock " << new_lock << dendl;
bool ret = false;
list<multimap<uint64_t, ceph_filelock>::iterator>
overlapping_locks, self_overlapping_locks, neighbor_locks;
// first, get any overlapping locks and split them into owned-by-us and not
if (get_overlapping_locks(new_lock, overlapping_locks, &neighbor_locks)) {
- dout(15) << "got overlapping lock, splitting by owner" << dendl;
+ ldout(cct,15) << "got overlapping lock, splitting by owner" << dendl;
split_by_owner(new_lock, overlapping_locks, self_overlapping_locks);
}
if (!overlapping_locks.empty()) { //overlapping locks owned by others :(
if (CEPH_LOCK_EXCL == new_lock.type) {
//can't set, we want an exclusive
- dout(15) << "overlapping lock, and this lock is exclusive, can't set"
+ ldout(cct,15) << "overlapping lock, and this lock is exclusive, can't set"
<< dendl;
if (wait_on_fail && !replay) {
waiting_locks.insert(pair<uint64_t, ceph_filelock>(new_lock.start, new_lock));
}
} else { //shared lock, check for any exclusive locks blocking us
if (contains_exclusive_lock(overlapping_locks)) { //blocked :(
- dout(15) << " blocked by exclusive lock in overlapping_locks" << dendl;
+ ldout(cct,15) << " blocked by exclusive lock in overlapping_locks" << dendl;
if (wait_on_fail && !replay) {
waiting_locks.insert(pair<uint64_t, ceph_filelock>(new_lock.start, new_lock));
}
} else {
//yay, we can insert a shared lock
- dout(15) << "inserting shared lock" << dendl;
+ ldout(cct,15) << "inserting shared lock" << dendl;
remove_waiting(new_lock);
adjust_locks(self_overlapping_locks, new_lock, neighbor_locks);
held_locks.insert(pair<uint64_t, ceph_filelock>(new_lock.start, new_lock));
@@ -80,7 +80,7 @@
} else { //no overlapping locks except our own
remove_waiting(new_lock);
adjust_locks(self_overlapping_locks, new_lock, neighbor_locks);
- dout(15) << "no conflicts, inserting " << new_lock << dendl;
+ ldout(cct,15) << "no conflicts, inserting " << new_lock << dendl;
held_locks.insert(pair<uint64_t, ceph_filelock>
(new_lock.start, new_lock));
ret = true;
@@ -123,9 +123,9 @@
list<multimap<uint64_t, ceph_filelock>::iterator> overlapping_locks,
self_overlapping_locks;
if (get_overlapping_locks(removal_lock, overlapping_locks)) {
- dout(15) << "splitting by owner" << dendl;
+ ldout(cct,15) << "splitting by owner" << dendl;
split_by_owner(removal_lock, overlapping_locks, self_overlapping_locks);
- } else dout(15) << "attempt to remove lock at " << removal_lock.start
+ } else ldout(cct,15) << "attempt to remove lock at " << removal_lock.start
<< " but no locks there!" << dendl;
bool remove_to_end = (0 == removal_lock.length);
uint64_t removal_start = removal_lock.start;
@@ -134,13 +134,13 @@
__s64 old_lock_client = 0;
ceph_filelock *old_lock;
- dout(15) << "examining " << self_overlapping_locks.size()
+ ldout(cct,15) << "examining " << self_overlapping_locks.size()
<< " self-overlapping locks for removal" << dendl;
for (list<multimap<uint64_t, ceph_filelock>::iterator>::iterator
iter = self_overlapping_locks.begin();
iter != self_overlapping_locks.end();
++iter) {
- dout(15) << "self overlapping lock " << (*iter)->second << dendl;
+ ldout(cct,15) << "self overlapping lock " << (*iter)->second << dendl;
old_lock = &(*iter)->second;
bool old_lock_to_end = (0 == old_lock->length);
old_lock_end = old_lock->start + old_lock->length - 1;
@@ -149,7 +149,7 @@
if (old_lock->start < removal_start) {
old_lock->length = removal_start - old_lock->start;
} else {
- dout(15) << "erasing " << (*iter)->second << dendl;
+ ldout(cct,15) << "erasing " << (*iter)->second << dendl;
held_locks.erase(*iter);
--client_held_lock_counts[old_lock_client];
}
@@ -160,7 +160,7 @@
(append_lock.start, append_lock));
++client_held_lock_counts[(client_t)old_lock->client];
if (old_lock->start >= removal_start) {
- dout(15) << "erasing " << (*iter)->second << dendl;
+ ldout(cct,15) << "erasing " << (*iter)->second << dendl;
held_locks.erase(*iter);
--client_held_lock_counts[old_lock_client];
} else old_lock->length = removal_start - old_lock->start;
@@ -176,7 +176,7 @@
if (old_lock->start < removal_start) {
old_lock->length = removal_start - old_lock->start;
} else {
- dout(15) << "erasing " << (*iter)->second << dendl;
+ ldout(cct,15) << "erasing " << (*iter)->second << dendl;
held_locks.erase(*iter);
--client_held_lock_counts[old_lock_client];
}
@@ -207,7 +207,7 @@
list<multimap<uint64_t, ceph_filelock>::iterator>
neighbor_locks)
{
- dout(15) << "adjust_locks" << dendl;
+ ldout(cct,15) << "adjust_locks" << dendl;
bool new_lock_to_end = (0 == new_lock.length);
uint64_t new_lock_start = new_lock.start;
uint64_t new_lock_end = new_lock.start + new_lock.length - 1;
@@ -219,7 +219,7 @@
iter != old_locks.end();
++iter) {
old_lock = &(*iter)->second;
- dout(15) << "adjusting lock: " << *old_lock << dendl;
+ ldout(cct,15) << "adjusting lock: " << *old_lock << dendl;
bool old_lock_to_end = (0 == old_lock->length);
old_lock_start = old_lock->start;
old_lock_end = old_lock->start + old_lock->length - 1;
@@ -228,17 +228,17 @@
old_lock_client = old_lock->client;
if (new_lock_to_end || old_lock_to_end) {
//special code path to deal with a length set at 0
- dout(15) << "one lock extends forever" << dendl;
+ ldout(cct,15) << "one lock extends forever" << dendl;
if (old_lock->type == new_lock.type) {
//just unify them in new lock, remove old lock
- dout(15) << "same lock type, unifying" << dendl;
+ ldout(cct,15) << "same lock type, unifying" << dendl;
new_lock.start = (new_lock_start < old_lock_start) ? new_lock_start :
old_lock_start;
new_lock.length = 0;
held_locks.erase(*iter);
--client_held_lock_counts[old_lock_client];
} else { //not same type, have to keep any remains of old lock around
- dout(15) << "shrinking old lock" << dendl;
+ ldout(cct,15) << "shrinking old lock" << dendl;
if (new_lock_to_end) {
if (old_lock_start < new_lock_start) {
old_lock->length = new_lock_start - old_lock_start;
@@ -262,17 +262,17 @@
}
} else {
if (old_lock->type == new_lock.type) { //just merge them!
- dout(15) << "merging locks, they're the same type" << dendl;
+ ldout(cct,15) << "merging locks, they're the same type" << dendl;
new_lock.start = (old_lock_start < new_lock_start ) ? old_lock_start :
new_lock_start;
int new_end = (new_lock_end > old_lock_end) ? new_lock_end :
old_lock_end;
new_lock.length = new_end - new_lock.start + 1;
- dout(15) << "erasing lock " << (*iter)->second << dendl;
+ ldout(cct,15) << "erasing lock " << (*iter)->second << dendl;
held_locks.erase(*iter);
--client_held_lock_counts[old_lock_client];
} else { //we'll have to update sizes and maybe make new locks
- dout(15) << "locks aren't same type, changing sizes" << dendl;
+ ldout(cct,15) << "locks aren't same type, changing sizes" << dendl;
if (old_lock_end > new_lock_end) { //add extra lock after new_lock
ceph_filelock appended_lock = *old_lock;
appended_lock.start = new_lock_end + 1;
@@ -302,7 +302,7 @@
++iter) {
old_lock = &(*iter)->second;
old_lock_client = old_lock->client;
- dout(15) << "lock to coalesce: " << *old_lock << dendl;
+ ldout(cct,15) << "lock to coalesce: " << *old_lock << dendl;
/* because if it's a neighboring lock there can't be any self-overlapping
locks that covered it */
if (old_lock->type == new_lock.type) { //merge them
@@ -354,8 +354,8 @@
&& (start != 0)
&& (lower_bound != lock_map.begin())) --lower_bound;
if (lock_map.end() == lower_bound)
- dout(15) << "get_lower_dout(15)eturning end()" << dendl;
- else dout(15) << "get_lower_bound returning iterator pointing to "
+ ldout(cct,15) << "get_lower_dout(15)eturning end()" << dendl;
+ else ldout(cct,15) << "get_lower_bound returning iterator pointing to "
<< lower_bound->second << dendl;
return lower_bound;
}
@@ -368,8 +368,8 @@
lock_map.upper_bound(end);
if (last != lock_map.begin()) --last;
if (lock_map.end() == last)
- dout(15) << "get_last_before returning end()" << dendl;
- else dout(15) << "get_last_before returning iterator pointing to "
+ ldout(cct,15) << "get_last_before returning end()" << dendl;
+ else ldout(cct,15) << "get_last_before returning iterator pointing to "
<< last->second << dendl;
return last;
}
@@ -382,7 +382,7 @@
((iter->first < start) &&
(((iter->first + iter->second.length - 1) >= start) ||
(0 == iter->second.length))));
- dout(15) << "share_space got start: " << start << ", end: " << end
+ ldout(cct,15) << "share_space got start: " << start << ", end: " << end
<< ", lock: " << iter->second << ", returning " << ret << dendl;
return ret;
}
@@ -393,7 +393,7 @@
list<multimap<uint64_t,
ceph_filelock>::iterator> *self_neighbors)
{
- dout(15) << "get_overlapping_locks" << dendl;
+ ldout(cct,15) << "get_overlapping_locks" << dendl;
// create a lock starting one earlier and ending one later
// to check for neighbors
ceph_filelock neighbor_check_lock = lock;
@@ -419,8 +419,7 @@
if (share_space(iter, lock)) {
overlaps.push_front(iter);
} else if (self_neighbors &&
- (neighbor_check_lock.client == iter->second.client) &&
- (neighbor_check_lock.pid == iter->second.pid) &&
+ ceph_filelock_owner_equal(neighbor_check_lock, iter->second) &&
share_space(iter, neighbor_check_lock)) {
self_neighbors->push_front(iter);
}
@@ -438,7 +437,7 @@
ceph_filelock>::iterator>&
overlaps)
{
- dout(15) << "get_waiting_overlaps" << dendl;
+ ldout(cct,15) << "get_waiting_overlaps" << dendl;
multimap<uint64_t, ceph_filelock>::iterator iter =
get_last_before(lock.start + lock.length - 1, waiting_locks);
bool cont = iter != waiting_locks.end();
@@ -459,15 +458,15 @@
{
list<multimap<uint64_t, ceph_filelock>::iterator>::iterator
iter = locks.begin();
- dout(15) << "owner lock: " << owner << dendl;
+ ldout(cct,15) << "owner lock: " << owner << dendl;
while (iter != locks.end()) {
- dout(15) << "comparing to " << (*iter)->second << dendl;
+ ldout(cct,15) << "comparing to " << (*iter)->second << dendl;
if (ceph_filelock_owner_equal((*iter)->second, owner)) {
- dout(15) << "success, pushing to owned_locks" << dendl;
+ ldout(cct,15) << "success, pushing to owned_locks" << dendl;
owned_locks.push_back(*iter);
iter = locks.erase(iter);
} else {
- dout(15) << "failure, something not equal in this group "
+ ldout(cct,15) << "failure, something not equal in this group "
<< (*iter)->second.client << ":" << owner.client << ","
<< (*iter)->second.owner << ":" << owner.owner << ","
<< (*iter)->second.pid << ":" << owner.pid << dendl;
diff -Nru ceph-0.80.7/src/mds/flock.h ceph-0.80.9/src/mds/flock.h
--- ceph-0.80.7/src/mds/flock.h 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/mds/flock.h 2015-03-09 20:51:18.000000000 +0100
@@ -37,7 +37,9 @@
}
class ceph_lock_state_t {
+ CephContext *cct;
public:
+ ceph_lock_state_t(CephContext *cct_) : cct(cct_) {}
multimap<uint64_t, ceph_filelock> held_locks; // current locks
multimap<uint64_t, ceph_filelock> waiting_locks; // locks waiting for other locks
// both of the above are keyed by starting offset
diff -Nru ceph-0.80.7/src/mds/Locker.cc ceph-0.80.9/src/mds/Locker.cc
--- ceph-0.80.7/src/mds/Locker.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/Locker.cc 2015-03-09 20:51:18.000000000 +0100
@@ -2689,6 +2689,9 @@
return t + 1;
}
+/**
+ * m and ack might be NULL, so don't dereference them unless dirty != 0
+ */
void Locker::_do_snap_update(CInode *in, snapid_t snap, int dirty, snapid_t follows, client_t client, MClientCaps *m, MClientCaps *ack)
{
dout(10) << "_do_snap_update dirty " << ccap_string(dirty)
@@ -2766,14 +2769,22 @@
client, NULL, ack));
}
-
+/**
+ * m might be NULL, so don't dereference it unless dirty != 0.
+ */
void Locker::_update_cap_fields(CInode *in, int dirty, MClientCaps *m, inode_t *pi)
{
+
+ if (dirty && m->get_ctime() > pi->ctime) {
+ dout(7) << " ctime " << pi->ctime << " -> " << m->get_ctime()
+ << " for " << *in << dendl;
+ pi->ctime = m->get_ctime();
+ }
+
// file
if (dirty & (CEPH_CAP_FILE_EXCL|CEPH_CAP_FILE_WR)) {
utime_t atime = m->get_atime();
utime_t mtime = m->get_mtime();
- utime_t ctime = m->get_ctime();
uint64_t size = m->get_size();
version_t inline_version = m->inline_version;
@@ -2783,11 +2794,6 @@
<< " for " << *in << dendl;
pi->mtime = mtime;
}
- if (ctime > pi->ctime) {
- dout(7) << " ctime " << pi->ctime << " -> " << ctime
- << " for " << *in << dendl;
- pi->ctime = ctime;
- }
if (in->inode.is_file() && // ONLY if regular file
size > pi->size) {
dout(7) << " size " << pi->size << " -> " << size
diff -Nru ceph-0.80.7/src/mds/Makefile.am ceph-0.80.9/src/mds/Makefile.am
--- ceph-0.80.7/src/mds/Makefile.am 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/Makefile.am 2015-03-09 20:51:18.000000000 +0100
@@ -4,7 +4,6 @@
mds/Dumper.cc \
mds/Resetter.cc \
mds/MDS.cc \
- mds/flock.cc \
mds/locks.c \
mds/journal.cc \
mds/Server.cc \
diff -Nru ceph-0.80.7/src/mds/MDCache.cc ceph-0.80.9/src/mds/MDCache.cc
--- ceph-0.80.7/src/mds/MDCache.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/MDCache.cc 2015-03-09 20:51:18.000000000 +0100
@@ -369,6 +369,8 @@
CDir *mydir = my->get_or_open_dirfrag(this, frag_t());
adjust_subtree_auth(mydir, mds->whoami);
+ LogSegment *ls = mds->mdlog->get_current_segment();
+
// stray dir
for (int i = 0; i < NUM_STRAY; ++i) {
CInode *stray = create_system_inode(MDS_INO_STRAY(mds->whoami, i), S_IFDIR);
@@ -384,8 +386,10 @@
mydir->fnode.fragstat.nsubdirs++;
// save them
straydir->mark_complete();
- straydir->mark_dirty(straydir->pre_dirty(), mds->mdlog->get_current_segment());
+ straydir->mark_dirty(straydir->pre_dirty(), ls);
straydir->commit(0, gather->new_sub());
+ stray->_mark_dirty_parent(ls, true);
+ stray->store_backtrace(gather->new_sub());
}
CInode *journal = create_system_inode(MDS_INO_LOG_OFFSET + mds->whoami, S_IFREG);
@@ -405,7 +409,7 @@
mydir->mark_complete();
- mydir->mark_dirty(mydir->pre_dirty(), mds->mdlog->get_current_segment());
+ mydir->mark_dirty(mydir->pre_dirty(), ls);
mydir->commit(0, gather->new_sub());
myin->store(gather->new_sub());
diff -Nru ceph-0.80.7/src/mds/mdstypes.h ceph-0.80.9/src/mds/mdstypes.h
--- ceph-0.80.7/src/mds/mdstypes.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/mdstypes.h 2015-03-09 20:51:18.000000000 +0100
@@ -703,7 +703,8 @@
cap_reconnect_t() {
memset(&capinfo, 0, sizeof(capinfo));
}
- cap_reconnect_t(uint64_t cap_id, inodeno_t pino, const string& p, int w, int i, inodeno_t sr) :
+ cap_reconnect_t(uint64_t cap_id, inodeno_t pino, const string& p, int w, int i,
+ inodeno_t sr, bufferlist& lb) :
path(p) {
capinfo.cap_id = cap_id;
capinfo.wanted = w;
@@ -711,6 +712,7 @@
capinfo.snaprealm = sr;
capinfo.pathbase = pino;
capinfo.flock_len = 0;
+ flockbl.claim(lb);
}
void encode(bufferlist& bl) const;
void decode(bufferlist::iterator& bl);
diff -Nru ceph-0.80.7/src/mds/Server.cc ceph-0.80.9/src/mds/Server.cc
--- ceph-0.80.7/src/mds/Server.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mds/Server.cc 2015-03-09 20:51:18.000000000 +0100
@@ -3085,20 +3085,28 @@
dout(0) << "handle_client_file_setlock: " << set_lock << dendl;
ceph_lock_state_t *lock_state = NULL;
+ bool interrupt = false;
// get the appropriate lock state
switch (req->head.args.filelock_change.rule) {
+ case CEPH_LOCK_FLOCK_INTR:
+ interrupt = true;
+ // fall-thru
case CEPH_LOCK_FLOCK:
lock_state = &cur->flock_locks;
break;
+ case CEPH_LOCK_FCNTL_INTR:
+ interrupt = true;
+ // fall-thru
case CEPH_LOCK_FCNTL:
lock_state = &cur->fcntl_locks;
break;
default:
- dout(0) << "got unknown lock type " << set_lock.type
- << ", dropping request!" << dendl;
+ dout(10) << "got unknown lock type " << set_lock.type
+ << ", dropping request!" << dendl;
+ reply_request(mdr, -EOPNOTSUPP);
return;
}
@@ -3109,16 +3117,15 @@
if (lock_state->is_waiting(set_lock)) {
dout(10) << " unlock removing waiting lock " << set_lock << dendl;
lock_state->remove_waiting(set_lock);
- } else {
+ cur->take_waiting(CInode::WAIT_FLOCK, waiters);
+ } else if (!interrupt) {
dout(10) << " unlock attempt on " << set_lock << dendl;
lock_state->remove_lock(set_lock, activated_locks);
cur->take_waiting(CInode::WAIT_FLOCK, waiters);
}
- reply_request(mdr, 0);
- /* For now we're ignoring the activated locks because their responses
- * will be sent when the lock comes up again in rotation by the MDS.
- * It's a cheap hack, but it's easy to code. */
mds->queue_waiters(waiters);
+
+ reply_request(mdr, 0);
} else {
dout(10) << " lock attempt on " << set_lock << dendl;
if (mdr->more()->flock_was_waiting &&
diff -Nru ceph-0.80.7/src/messages/MClientReconnect.h ceph-0.80.9/src/messages/MClientReconnect.h
--- ceph-0.80.7/src/messages/MClientReconnect.h 2014-10-14 21:12:50.000000000 +0200
+++ ceph-0.80.9/src/messages/MClientReconnect.h 2015-03-09 20:51:18.000000000 +0100
@@ -40,9 +40,9 @@
}
void add_cap(inodeno_t ino, uint64_t cap_id, inodeno_t pathbase, const string& path,
- int wanted, int issued,
- inodeno_t sr) {
- caps[ino] = cap_reconnect_t(cap_id, pathbase, path, wanted, issued, sr);
+ int wanted, int issued, inodeno_t sr, bufferlist& lb)
+ {
+ caps[ino] = cap_reconnect_t(cap_id, pathbase, path, wanted, issued, sr, lb);
}
void add_snaprealm(inodeno_t ino, snapid_t seq, inodeno_t parent) {
ceph_mds_snaprealm_reconnect r;
diff -Nru ceph-0.80.7/src/mon/DataHealthService.cc ceph-0.80.9/src/mon/DataHealthService.cc
--- ceph-0.80.7/src/mon/DataHealthService.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/DataHealthService.cc 2015-03-09 20:51:18.000000000 +0100
@@ -86,10 +86,10 @@
health_status_t health_status = HEALTH_OK;
string health_detail;
- if (stats.latest_avail_percent <= g_conf->mon_data_avail_crit) {
+ if (stats.fs_stats.avail_percent <= g_conf->mon_data_avail_crit) {
health_status = HEALTH_ERR;
health_detail = "low disk space, shutdown imminent";
- } else if (stats.latest_avail_percent <= g_conf->mon_data_avail_warn) {
+ } else if (stats.fs_stats.avail_percent <= g_conf->mon_data_avail_warn) {
health_status = HEALTH_WARN;
health_detail = "low disk space";
}
@@ -110,7 +110,7 @@
stringstream ss;
ss << "mon." << mon_name << " " << health_detail;
summary.push_back(make_pair(health_status, ss.str()));
- ss << " -- " << stats.latest_avail_percent << "% avail";
+ ss << " -- " << stats.fs_stats.avail_percent << "% avail";
if (detail)
detail->push_back(make_pair(health_status, ss.str()));
}
@@ -151,23 +151,18 @@
int DataHealthService::update_stats()
{
- struct statfs stbuf;
- int err = ::statfs(g_conf->mon_data.c_str(), &stbuf);
- if (err < 0) {
- derr << __func__ << " statfs error: " << cpp_strerror(errno) << dendl;
- return -errno;
- }
-
entity_inst_t our_inst = mon->messenger->get_myinst();
DataStats& ours = stats[our_inst];
- ours.kb_total = stbuf.f_blocks * stbuf.f_bsize / 1024;
- ours.kb_used = (stbuf.f_blocks - stbuf.f_bfree) * stbuf.f_bsize / 1024;
- ours.kb_avail = stbuf.f_bavail * stbuf.f_bsize / 1024;
- ours.latest_avail_percent = (((float)ours.kb_avail/ours.kb_total)*100);
- dout(0) << __func__ << " avail " << ours.latest_avail_percent << "%"
- << " total " << ours.kb_total << " used " << ours.kb_used << " avail " << ours.kb_avail
- << dendl;
+ int err = get_fs_stats(ours.fs_stats, g_conf->mon_data.c_str());
+ if (err < 0) {
+ derr << __func__ << " get_fs_stats error: " << cpp_strerror(err) << dendl;
+ return err;
+ }
+ dout(0) << __func__ << " avail " << ours.fs_stats.avail_percent << "%"
+ << " total " << prettybyte_t(ours.fs_stats.byte_total)
+ << ", used " << prettybyte_t(ours.fs_stats.byte_used)
+ << ", avail " << prettybyte_t(ours.fs_stats.byte_avail) << dendl;
ours.last_update = ceph_clock_now(g_ceph_context);
return update_store_stats(ours);
@@ -213,7 +208,7 @@
DataStats &ours = stats[mon->messenger->get_myinst()];
- if (ours.latest_avail_percent <= g_conf->mon_data_avail_crit) {
+ if (ours.fs_stats.avail_percent <= g_conf->mon_data_avail_crit) {
derr << "reached critical levels of available space on local monitor storage"
<< " -- shutdown!" << dendl;
force_shutdown();
@@ -224,12 +219,12 @@
// consumed in-between reports to assess if it's worth to log this info,
// otherwise we may very well contribute to the consumption of the
// already low available disk space.
- if (ours.latest_avail_percent <= g_conf->mon_data_avail_warn) {
- if (ours.latest_avail_percent != last_warned_percent)
+ if (ours.fs_stats.avail_percent <= g_conf->mon_data_avail_warn) {
+ if (ours.fs_stats.avail_percent != last_warned_percent)
mon->clog.warn()
<< "reached concerning levels of available space on local monitor storage"
- << " (" << ours.latest_avail_percent << "% free)\n";
- last_warned_percent = ours.latest_avail_percent;
+ << " (" << ours.fs_stats.avail_percent << "% free)\n";
+ last_warned_percent = ours.fs_stats.avail_percent;
} else {
last_warned_percent = 0;
}
diff -Nru ceph-0.80.7/src/mon/MonCommands.h ceph-0.80.9/src/mon/MonCommands.h
--- ceph-0.80.7/src/mon/MonCommands.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/MonCommands.h 2015-03-09 20:51:18.000000000 +0100
@@ -425,6 +425,9 @@
"name=ancestor,type=CephString,req=false,goodchars=[A-Za-z0-9-_.]", \
"unlink <name> from crush map (everywhere, or just at <ancestor>)", \
"osd", "rw", "cli,rest")
+COMMAND("osd crush reweight-all",
+ "recalculate the weights for the tree to ensure they sum correctly",
+ "osd", "rw", "cli,rest")
COMMAND("osd crush reweight " \
"name=name,type=CephString,goodchars=[A-Za-z0-9-_.] " \
"name=weight,type=CephFloat,range=0.0", \
@@ -433,6 +436,15 @@
COMMAND("osd crush tunables " \
"name=profile,type=CephChoices,strings=legacy|argonaut|bobtail|firefly|optimal|default", \
"set crush tunables values to <profile>", "osd", "rw", "cli,rest")
+COMMAND("osd crush set-tunable " \
+ "name=tunable,type=CephChoices,strings=straw_calc_version " \
+ "name=value,type=CephInt",
+ "set crush tunable <tunable> to <value>",
+ "osd", "rw", "cli,rest")
+COMMAND("osd crush get-tunable " \
+ "name=tunable,type=CephChoices,strings=straw_calc_version",
+ "get crush tunable <tunable>",
+ "osd", "rw", "cli,rest")
COMMAND("osd crush show-tunables", \
"show current crush tunables", "osd", "r", "cli,rest")
COMMAND("osd crush rule create-simple " \
diff -Nru ceph-0.80.7/src/mon/Monitor.cc ceph-0.80.9/src/mon/Monitor.cc
--- ceph-0.80.7/src/mon/Monitor.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/Monitor.cc 2015-03-09 20:51:18.000000000 +0100
@@ -428,6 +428,8 @@
cluster_logger = pcb.create_perf_counters();
}
+ paxos->init_logger();
+
// verify cluster_uuid
{
int r = check_fsid();
@@ -1406,6 +1408,8 @@
*/
void Monitor::handle_probe_probe(MMonProbe *m)
{
+ MMonProbe *r;
+
dout(10) << "handle_probe_probe " << m->get_source_inst() << *m
<< " features " << m->get_connection()->get_features() << dendl;
uint64_t missing = required_features & ~m->get_connection()->get_features();
@@ -1418,12 +1422,26 @@
m->required_features = required_features;
messenger->send_message(r, m->get_connection());
}
- m->put();
- return;
+ goto out;
+ }
+
+ if (!is_probing() && !is_synchronizing()) {
+ // If the probing mon is way ahead of us, we need to re-bootstrap.
+ // Normally we capture this case when we initially bootstrap, but
+ // it is possible we pass those checks (we overlap with
+ // quorum-to-be) but fail to join a quorum before it moves past
+ // us. We need to be kicked back to bootstrap so we can
+ // synchonize, not keep calling elections.
+ if (paxos->get_version() + 1 < m->paxos_first_version) {
+ dout(1) << " peer " << m->get_source_addr() << " has first_committed "
+ << "ahead of us, re-bootstrapping" << dendl;
+ bootstrap();
+ goto out;
+
+ }
}
- MMonProbe *r = new MMonProbe(monmap->fsid, MMonProbe::OP_REPLY,
- name, has_ever_joined);
+ r = new MMonProbe(monmap->fsid, MMonProbe::OP_REPLY, name, has_ever_joined);
r->name = name;
r->quorum = quorum;
monmap->encode(r->monmap_bl, m->get_connection()->get_features());
@@ -1438,6 +1456,7 @@
extra_probe_peers.insert(m->get_source_addr());
}
+ out:
m->put();
}
diff -Nru ceph-0.80.7/src/mon/MonitorDBStore.h ceph-0.80.9/src/mon/MonitorDBStore.h
--- ceph-0.80.7/src/mon/MonitorDBStore.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/MonitorDBStore.h 2015-03-09 20:51:18.000000000 +0100
@@ -87,6 +87,9 @@
struct Transaction {
list<Op> ops;
+ uint64_t bytes, keys;
+
+ Transaction() : bytes(0), keys(0) {}
enum {
OP_PUT = 1,
@@ -96,6 +99,8 @@
void put(string prefix, string key, bufferlist& bl) {
ops.push_back(Op(OP_PUT, prefix, key, bl));
+ ++keys;
+ bytes += prefix.length() + key.length() + bl.length();
}
void put(string prefix, version_t ver, bufferlist& bl) {
@@ -112,6 +117,8 @@
void erase(string prefix, string key) {
ops.push_back(Op(OP_ERASE, prefix, key));
+ ++keys;
+ bytes += prefix.length() + key.length();
}
void erase(string prefix, version_t ver) {
@@ -129,14 +136,20 @@
}
void encode(bufferlist& bl) const {
- ENCODE_START(1, 1, bl);
+ ENCODE_START(2, 1, bl);
::encode(ops, bl);
+ ::encode(bytes, bl);
+ ::encode(keys, bl);
ENCODE_FINISH(bl);
}
void decode(bufferlist::iterator& bl) {
- DECODE_START(1, bl);
+ DECODE_START(2, bl);
::decode(ops, bl);
+ if (struct_v >= 2) {
+ ::decode(bytes, bl);
+ ::decode(keys, bl);
+ }
DECODE_FINISH(bl);
}
@@ -153,6 +166,8 @@
void append(Transaction& other) {
ops.splice(ops.end(), other.ops);
+ keys += other.keys;
+ bytes += other.bytes;
}
void append_from_encoded(bufferlist& bl) {
@@ -169,6 +184,12 @@
bool size() {
return ops.size();
}
+ uint64_t get_keys() const {
+ return keys;
+ }
+ uint64_t get_bytes() const {
+ return bytes;
+ }
void dump(ceph::Formatter *f, bool dump_val=false) const {
f->open_object_section("transaction");
@@ -218,6 +239,8 @@
f->close_section();
}
f->close_section();
+ f->dump_unsigned("num_keys", keys);
+ f->dump_unsigned("num_bytes", bytes);
f->close_section();
}
};
diff -Nru ceph-0.80.7/src/mon/mon_types.h ceph-0.80.9/src/mon/mon_types.h
--- ceph-0.80.7/src/mon/mon_types.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/mon_types.h 2015-03-09 20:51:18.000000000 +0100
@@ -16,6 +16,7 @@
#define CEPH_MON_TYPES_H
#include "include/utime.h"
+#include "include/util.h"
#include "common/Formatter.h"
#define PAXOS_PGMAP 0 // before osd, for pg kick to behave
@@ -89,44 +90,50 @@
// data stats
struct DataStats {
+ ceph_data_stats_t fs_stats;
// data dir
- uint64_t kb_total;
- uint64_t kb_used;
- uint64_t kb_avail;
- int latest_avail_percent;
utime_t last_update;
-
LevelDBStoreStats store_stats;
void dump(Formatter *f) const {
assert(f != NULL);
- f->dump_int("kb_total", kb_total);
- f->dump_int("kb_used", kb_used);
- f->dump_int("kb_avail", kb_avail);
- f->dump_int("avail_percent", latest_avail_percent);
+ f->dump_int("kb_total", (fs_stats.byte_total/1024));
+ f->dump_int("kb_used", (fs_stats.byte_used/1024));
+ f->dump_int("kb_avail", (fs_stats.byte_avail/1024));
+ f->dump_int("avail_percent", fs_stats.avail_percent);
f->dump_stream("last_updated") << last_update;
-
f->open_object_section("store_stats");
store_stats.dump(f);
f->close_section();
}
void encode(bufferlist &bl) const {
- ENCODE_START(2, 1, bl);
- ::encode(kb_total, bl);
- ::encode(kb_used, bl);
- ::encode(kb_avail, bl);
- ::encode(latest_avail_percent, bl);
+ ENCODE_START(3, 1, bl);
+ ::encode(fs_stats.byte_total, bl);
+ ::encode(fs_stats.byte_used, bl);
+ ::encode(fs_stats.byte_avail, bl);
+ ::encode(fs_stats.avail_percent, bl);
::encode(last_update, bl);
::encode(store_stats, bl);
ENCODE_FINISH(bl);
}
void decode(bufferlist::iterator &p) {
DECODE_START(1, p);
- ::decode(kb_total, p);
- ::decode(kb_used, p);
- ::decode(kb_avail, p);
- ::decode(latest_avail_percent, p);
+ // we moved from having fields in kb to fields in byte
+ if (struct_v > 2) {
+ ::decode(fs_stats.byte_total, p);
+ ::decode(fs_stats.byte_used, p);
+ ::decode(fs_stats.byte_avail, p);
+ } else {
+ uint64_t t;
+ ::decode(t, p);
+ fs_stats.byte_total = t*1024;
+ ::decode(t, p);
+ fs_stats.byte_used = t*1024;
+ ::decode(t, p);
+ fs_stats.byte_avail = t*1024;
+ }
+ ::decode(fs_stats.avail_percent, p);
::decode(last_update, p);
if (struct_v > 1)
::decode(store_stats, p);
diff -Nru ceph-0.80.7/src/mon/OSDMonitor.cc ceph-0.80.9/src/mon/OSDMonitor.cc
--- ceph-0.80.7/src/mon/OSDMonitor.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/OSDMonitor.cc 2015-03-09 20:51:18.000000000 +0100
@@ -1211,7 +1211,7 @@
osdmap.get_info(from).up_from > m->version) {
dout(7) << "prepare_boot msg from before last up_from, ignoring" << dendl;
send_latest(m, m->sb.current_epoch+1);
- goto ignore;
+ return true;
}
// noup?
@@ -2465,6 +2465,31 @@
}
ss << "listed " << osdmap.blacklist.size() << " entries";
+ } else if (prefix == "osd crush get-tunable") {
+ string tunable;
+ cmd_getval(g_ceph_context, cmdmap, "tunable", tunable);
+ int value;
+ cmd_getval(g_ceph_context, cmdmap, "value", value);
+ ostringstream rss;
+ if (f)
+ f->open_object_section("tunable");
+ if (tunable == "straw_calc_version") {
+ if (f)
+ f->dump_int(tunable.c_str(), osdmap.crush->get_straw_calc_version());
+ else
+ rss << osdmap.crush->get_straw_calc_version() << "\n";
+ } else {
+ r = -EINVAL;
+ goto reply;
+ }
+ if (f) {
+ f->close_section();
+ f->flush(rdata);
+ } else {
+ rdata.append(rss.str());
+ }
+ r = 0;
+
} else if (prefix == "osd pool get") {
string poolstr;
cmd_getval(g_ceph_context, cmdmap, "pool", poolstr);
@@ -3279,11 +3304,38 @@
if (*crush_ruleset < 0) {
switch (pool_type) {
case pg_pool_t::TYPE_REPLICATED:
- *crush_ruleset = osdmap.crush->get_osd_pool_default_crush_replicated_ruleset(g_ceph_context);
- if (*crush_ruleset < 0) {
- // Errors may happen e.g. if no valid ruleset is available
- ss << "No suitable CRUSH ruleset exists";
- return *crush_ruleset;
+ {
+ if (ruleset_name == "") {
+ //Use default ruleset
+ *crush_ruleset = osdmap.crush->get_osd_pool_default_crush_replicated_ruleset(g_ceph_context);
+ if (*crush_ruleset < 0) {
+ // Errors may happen e.g. if no valid ruleset is available
+ ss << "No suitable CRUSH ruleset exists";
+ return *crush_ruleset;
+ }
+ } else {
+ int ret;
+ ret = osdmap.crush->get_rule_id(ruleset_name);
+ if (ret != -ENOENT) {
+ // found it, use it
+ *crush_ruleset = ret;
+ } else {
+ CrushWrapper newcrush;
+ _get_pending_crush(newcrush);
+
+ ret = newcrush.get_rule_id(ruleset_name);
+ if (ret != -ENOENT) {
+ // found it, wait for it to be proposed
+ dout(20) << "prepare_pool_crush_ruleset: ruleset "
+ << ruleset_name << " is pending, try again" << dendl;
+ return -EAGAIN;
+ } else {
+ //Cannot find it , return error
+ ss << "Specified ruleset " << ruleset_name << " doesn't exist";
+ return ret;
+ }
+ }
+ }
}
break;
case pg_pool_t::TYPE_ERASURE:
@@ -4115,6 +4167,19 @@
}
} while (false);
+ } else if (prefix == "osd crush reweight-all") {
+ // osd crush reweight <name> <weight>
+ CrushWrapper newcrush;
+ _get_pending_crush(newcrush);
+
+ newcrush.reweight(g_ceph_context);
+ pending_inc.crush.clear();
+ newcrush.encode(pending_inc.crush);
+ ss << "reweighted crush hierarchy";
+ getline(ss, rs);
+ wait_for_finished_proposal(new Monitor::C_Command(mon, m, 0, rs,
+ get_last_committed() + 1));
+ return true;
} else if (prefix == "osd crush reweight") {
do {
// osd crush reweight <name> <weight>
@@ -4190,6 +4255,46 @@
wait_for_finished_proposal(new Monitor::C_Command(mon, m, 0, rs,
get_last_committed() + 1));
return true;
+ } else if (prefix == "osd crush set-tunable") {
+ CrushWrapper newcrush;
+ _get_pending_crush(newcrush);
+
+ err = 0;
+ string tunable;
+ cmd_getval(g_ceph_context, cmdmap, "tunable", tunable);
+
+ int64_t value = -1;
+ if (!cmd_getval(g_ceph_context, cmdmap, "value", value)) {
+ err = -EINVAL;
+ ss << "failed to parse integer value " << cmd_vartype_stringify(cmdmap["value"]);
+ goto reply;
+ }
+
+ if (tunable == "straw_calc_version") {
+ if (value < 0 || value > 2) {
+ ss << "value must be 0 or 1; got " << value;
+ err = -EINVAL;
+ goto reply;
+ }
+ newcrush.set_straw_calc_version(value);
+ } else {
+ ss << "unrecognized tunable '" << tunable << "'";
+ err = -EINVAL;
+ goto reply;
+ }
+
+ if (!validate_crush_against_features(&newcrush, ss)) {
+ err = -EINVAL;
+ goto reply;
+ }
+
+ pending_inc.crush.clear();
+ newcrush.encode(pending_inc.crush);
+ ss << "adjusted tunable " << tunable << " to " << value;
+ getline(ss, rs);
+ wait_for_finished_proposal(new Monitor::C_Command(mon, m, 0, rs,
+ get_last_committed() + 1));
+ return true;
} else if (prefix == "osd crush rule create-simple") {
string name, root, type, mode;
@@ -5004,35 +5109,41 @@
cmd_getval(g_ceph_context, cmdmap, "ruleset", ruleset_name);
string erasure_code_profile;
cmd_getval(g_ceph_context, cmdmap, "erasure_code_profile", erasure_code_profile);
- if (erasure_code_profile == "")
- erasure_code_profile = "default";
- if (erasure_code_profile == "default") {
- if (!osdmap.has_erasure_code_profile(erasure_code_profile)) {
- if (pending_inc.has_erasure_code_profile(erasure_code_profile)) {
- dout(20) << "erasure code profile " << erasure_code_profile << " already pending" << dendl;
- goto wait;
- }
- map<string,string> profile_map;
- err = osdmap.get_erasure_code_profile_default(g_ceph_context,
+ if (pool_type == pg_pool_t::TYPE_ERASURE) {
+ if (erasure_code_profile == "")
+ erasure_code_profile = "default";
+ //handle the erasure code profile
+ if (erasure_code_profile == "default") {
+ if (!osdmap.has_erasure_code_profile(erasure_code_profile)) {
+ if (pending_inc.has_erasure_code_profile(erasure_code_profile)) {
+ dout(20) << "erasure code profile " << erasure_code_profile << " already pending" << dendl;
+ goto wait;
+ }
+
+ map<string,string> profile_map;
+ err = osdmap.get_erasure_code_profile_default(g_ceph_context,
profile_map,
&ss);
- if (err)
- goto reply;
- dout(20) << "erasure code profile " << erasure_code_profile << " set" << dendl;
- pending_inc.set_erasure_code_profile(erasure_code_profile, profile_map);
- goto wait;
+ if (err)
+ goto reply;
+ dout(20) << "erasure code profile " << erasure_code_profile << " set" << dendl;
+ pending_inc.set_erasure_code_profile(erasure_code_profile, profile_map);
+ goto wait;
+ }
}
- }
-
- if (ruleset_name == "") {
- if (erasure_code_profile == "default") {
- ruleset_name = "erasure-code";
- } else {
- dout(1) << "implicitly use ruleset named after the pool: "
+ if (ruleset_name == "") {
+ if (erasure_code_profile == "default") {
+ ruleset_name = "erasure-code";
+ } else {
+ dout(1) << "implicitly use ruleset named after the pool: "
<< poolstr << dendl;
- ruleset_name = poolstr;
+ ruleset_name = poolstr;
+ }
}
+ } else {
+ //NOTE:for replicated pool,cmd_map will put ruleset_name to erasure_code_profile field
+ ruleset_name = erasure_code_profile;
}
err = prepare_new_pool(poolstr, 0, // auid=0 for admin created pool
diff -Nru ceph-0.80.7/src/mon/Paxos.cc ceph-0.80.9/src/mon/Paxos.cc
--- ceph-0.80.7/src/mon/Paxos.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/Paxos.cc 2015-03-09 20:51:18.000000000 +0100
@@ -73,6 +73,44 @@
assert(is_consistent());
}
+void Paxos::init_logger()
+{
+ PerfCountersBuilder pcb(g_ceph_context, "paxos", l_paxos_first, l_paxos_last);
+ pcb.add_u64_counter(l_paxos_start_leader, "start_leader");
+ pcb.add_u64_counter(l_paxos_start_peon, "start_peon");
+ pcb.add_u64_counter(l_paxos_restart, "restart");
+ pcb.add_u64_counter(l_paxos_refresh, "refresh");
+ pcb.add_time_avg(l_paxos_refresh_latency, "refresh_latency");
+ pcb.add_u64_counter(l_paxos_begin, "begin");
+ pcb.add_u64_avg(l_paxos_begin_keys, "begin_keys");
+ pcb.add_u64_avg(l_paxos_begin_bytes, "begin_bytes");
+ pcb.add_time_avg(l_paxos_begin_latency, "begin_latency");
+ pcb.add_u64_counter(l_paxos_commit, "commit");
+ pcb.add_u64_avg(l_paxos_commit_keys, "commit_keys");
+ pcb.add_u64_avg(l_paxos_commit_bytes, "commit_bytes");
+ pcb.add_time_avg(l_paxos_commit_latency, "commit_latency");
+ pcb.add_u64_counter(l_paxos_collect, "collect");
+ pcb.add_u64_avg(l_paxos_collect_keys, "collect_keys");
+ pcb.add_u64_avg(l_paxos_collect_bytes, "collect_bytes");
+ pcb.add_time_avg(l_paxos_collect_latency, "collect_latency");
+ pcb.add_u64_counter(l_paxos_collect_uncommitted, "collect_uncommitted");
+ pcb.add_u64_counter(l_paxos_collect_timeout, "collect_timeout");
+ pcb.add_u64_counter(l_paxos_accept_timeout, "accept_timeout");
+ pcb.add_u64_counter(l_paxos_lease_ack_timeout, "lease_ack_timeout");
+ pcb.add_u64_counter(l_paxos_lease_timeout, "lease_timeout");
+ pcb.add_u64_counter(l_paxos_store_state, "store_state");
+ pcb.add_u64_avg(l_paxos_store_state_keys, "store_state_keys");
+ pcb.add_u64_avg(l_paxos_store_state_bytes, "store_state_bytes");
+ pcb.add_time_avg(l_paxos_store_state_latency, "store_state_latency");
+ pcb.add_u64_counter(l_paxos_share_state, "share_state");
+ pcb.add_u64_avg(l_paxos_share_state_keys, "share_state_keys");
+ pcb.add_u64_avg(l_paxos_share_state_bytes, "share_state_bytes");
+ pcb.add_u64_counter(l_paxos_new_pn, "new_pn");
+ pcb.add_time_avg(l_paxos_new_pn_latency, "new_pn_latency");
+ logger = pcb.create_perf_counters();
+ g_ceph_context->get_perfcounters_collection()->add(logger);
+}
+
void Paxos::dump_info(Formatter *f)
{
f->open_object_section("paxos");
@@ -120,6 +158,8 @@
<< " pn " << uncommitted_pn
<< " (" << uncommitted_value.length() << " bytes) from myself"
<< dendl;
+
+ logger->inc(l_paxos_collect_uncommitted);
}
// pick new pn
@@ -193,7 +233,15 @@
f.flush(*_dout);
*_dout << dendl;
+ logger->inc(l_paxos_collect);
+ logger->inc(l_paxos_collect_keys, t.get_keys());
+ logger->inc(l_paxos_collect_bytes, t.get_bytes());
+ utime_t start = ceph_clock_now(NULL);
+
get_store()->apply_transaction(t);
+
+ utime_t end = ceph_clock_now(NULL);
+ logger->tinc(l_paxos_collect_latency, end - start);
} else {
// don't accept!
dout(10) << "NOT accepting pn " << collect->pn << " from " << collect->pn_from
@@ -229,6 +277,8 @@
<< " and crossing our fingers" << dendl;
last->uncommitted_pn = previous_pn;
}
+
+ logger->inc(l_paxos_collect_uncommitted);
}
// send reply
@@ -258,14 +308,19 @@
version_t v = peer_last_committed + 1;
// include incrementals
+ uint64_t bytes = 0;
for ( ; v <= last_committed; v++) {
if (get_store()->exists(get_name(), v)) {
get_store()->get(get_name(), v, m->values[v]);
assert(m->values[v].length());
dout(10) << " sharing " << v << " ("
<< m->values[v].length() << " bytes)" << dendl;
+ bytes += m->values[v].length() + 16; // paxos_ + 10 digits = 16
}
}
+ logger->inc(l_paxos_share_state);
+ logger->inc(l_paxos_share_state_keys, m->values.size());
+ logger->inc(l_paxos_share_state_bytes, bytes);
m->last_committed = last_committed;
}
@@ -318,6 +373,7 @@
dout(10) << "store_state [" << start->first << ".."
<< last_committed << "]" << dendl;
t.put(get_name(), "last_committed", last_committed);
+
// we should apply the state here -- decode every single bufferlist in the
// map and append the transactions to 't'.
map<version_t,bufferlist>::iterator it;
@@ -345,8 +401,16 @@
f.flush(*_dout);
*_dout << dendl;
+ logger->inc(l_paxos_store_state);
+ logger->inc(l_paxos_store_state_bytes, t.get_bytes());
+ logger->inc(l_paxos_store_state_keys, t.get_keys());
+ utime_t start = ceph_clock_now(NULL);
+
get_store()->apply_transaction(t);
+ utime_t end = ceph_clock_now(NULL);
+ logger->tinc(l_paxos_store_state_latency, end - start);
+
// refresh first_committed; this txn may have trimmed.
first_committed = get_store()->get(get_name(), "first_committed");
@@ -385,6 +449,7 @@
void Paxos::handle_last(MMonPaxos *last)
{
bool need_refresh = false;
+ int from = last->get_source().num();
dout(10) << "handle_last " << *last << dendl;
@@ -396,12 +461,13 @@
// note peer's first_ and last_committed, in case we learn a new
// commit and need to push it to them.
- peer_first_committed[last->get_source().num()] = last->first_committed;
- peer_last_committed[last->get_source().num()] = last->last_committed;
+ peer_first_committed[from] = last->first_committed;
+ peer_last_committed[from] = last->last_committed;
- if (last->first_committed > last_committed+1) {
+ if (last->first_committed > last_committed + 1) {
dout(5) << __func__
- << " peon's lowest version is too high for our last committed"
+ << " mon." << from
+ << " lowest version is too high for our last committed"
<< " (theirs: " << last->first_committed
<< "; ours: " << last_committed << ") -- bootstrap!" << dendl;
last->put();
@@ -416,6 +482,31 @@
assert(g_conf->paxos_kill_at != 2);
+ // is everyone contiguous and up to date?
+ for (map<int,version_t>::iterator p = peer_last_committed.begin();
+ p != peer_last_committed.end();
+ ++p) {
+ if (p->second + 1 < first_committed && first_committed > 1) {
+ dout(5) << __func__
+ << " peon " << p->first
+ << " last_committed (" << p->second
+ << ") is too low for our first_committed (" << first_committed
+ << ") -- bootstrap!" << dendl;
+ last->put();
+ mon->bootstrap();
+ return;
+ }
+ if (p->second < last_committed) {
+ // share committed values
+ dout(10) << " sending commit to mon." << p->first << dendl;
+ MMonPaxos *commit = new MMonPaxos(mon->get_epoch(),
+ MMonPaxos::OP_COMMIT,
+ ceph_clock_now(g_ceph_context));
+ share_state(commit, peer_first_committed[p->first], p->second);
+ mon->messenger->send_message(commit, mon->monmap->get_inst(p->first));
+ }
+ }
+
// do they accept your pn?
if (last->pn > accepted_pn) {
// no, try again.
@@ -457,21 +548,6 @@
// cancel timeout event
mon->timer.cancel_event(collect_timeout_event);
collect_timeout_event = 0;
-
- // share committed values?
- for (map<int,version_t>::iterator p = peer_last_committed.begin();
- p != peer_last_committed.end();
- ++p) {
- if (p->second < last_committed) {
- // share committed values
- dout(10) << " sending commit to mon." << p->first << dendl;
- MMonPaxos *commit = new MMonPaxos(mon->get_epoch(),
- MMonPaxos::OP_COMMIT,
- ceph_clock_now(g_ceph_context));
- share_state(commit, peer_first_committed[p->first], p->second);
- mon->messenger->send_message(commit, mon->monmap->get_inst(p->first));
- }
- }
peer_first_committed.clear();
peer_last_committed.clear();
@@ -513,6 +589,7 @@
{
dout(1) << "collect timeout, calling fresh election" << dendl;
collect_timeout_event = 0;
+ logger->inc(l_paxos_collect_timeout);
assert(mon->is_leader());
mon->bootstrap();
}
@@ -534,7 +611,7 @@
// and no value, yet.
assert(new_value.length() == 0);
-
+
// accept it ourselves
accepted.clear();
accepted.insert(mon->rank);
@@ -573,8 +650,16 @@
f.flush(*_dout);
*_dout << dendl;
+ logger->inc(l_paxos_begin);
+ logger->inc(l_paxos_begin_keys, t.get_keys());
+ logger->inc(l_paxos_begin_bytes, t.get_bytes());
+ utime_t start = ceph_clock_now(NULL);
+
get_store()->apply_transaction(t);
+ utime_t end = ceph_clock_now(NULL);
+ logger->tinc(l_paxos_begin_latency, end - start);
+
assert(g_conf->paxos_kill_at != 3);
if (mon->get_quorum().size() == 1) {
@@ -629,6 +714,8 @@
assert(g_conf->paxos_kill_at != 4);
+ logger->inc(l_paxos_begin);
+
// set state.
state = STATE_UPDATING;
lease_expire = utime_t(); // cancel lease
@@ -651,8 +738,14 @@
f.flush(*_dout);
*_dout << dendl;
+ logger->inc(l_paxos_begin_bytes, t.get_bytes());
+ utime_t start = ceph_clock_now(NULL);
+
get_store()->apply_transaction(t);
+ utime_t end = ceph_clock_now(NULL);
+ logger->tinc(l_paxos_begin_latency, end - start);
+
assert(g_conf->paxos_kill_at != 5);
// reply
@@ -733,6 +826,7 @@
accept_timeout_event = 0;
assert(mon->is_leader());
assert(is_updating() || is_updating_previous());
+ logger->inc(l_paxos_accept_timeout);
mon->bootstrap();
}
@@ -764,8 +858,16 @@
f.flush(*_dout);
*_dout << dendl;
+ logger->inc(l_paxos_commit);
+ logger->inc(l_paxos_commit_keys, t.get_keys());
+ logger->inc(l_paxos_commit_bytes, t.get_bytes());
+ utime_t start = ceph_clock_now(NULL);
+
get_store()->apply_transaction(t);
+ utime_t end = ceph_clock_now(NULL);
+ logger->tinc(l_paxos_commit_latency, end - start);
+
assert(g_conf->paxos_kill_at != 8);
// refresh first_committed; this txn may have trimmed.
@@ -802,6 +904,8 @@
{
dout(10) << "handle_commit on " << commit->last_committed << dendl;
+ logger->inc(l_paxos_commit);
+
if (!mon->is_peon()) {
dout(10) << "not a peon, dropping" << dendl;
assert(0);
@@ -883,9 +987,15 @@
{
bool need_bootstrap = false;
+ utime_t start = ceph_clock_now(NULL);
+
// make sure we have the latest state loaded up
mon->refresh_from_paxos(&need_bootstrap);
+ utime_t end = ceph_clock_now(NULL);
+ logger->inc(l_paxos_refresh);
+ logger->tinc(l_paxos_refresh_latency, end - start);
+
if (need_bootstrap) {
dout(10) << " doing requested bootstrap" << dendl;
mon->bootstrap();
@@ -1019,7 +1129,7 @@
dout(1) << "lease_ack_timeout -- calling new election" << dendl;
assert(mon->is_leader());
assert(is_active());
-
+ logger->inc(l_paxos_lease_ack_timeout);
lease_ack_timeout_event = 0;
mon->bootstrap();
}
@@ -1037,7 +1147,7 @@
{
dout(1) << "lease_timeout -- calling new election" << dendl;
assert(mon->is_peon());
-
+ logger->inc(l_paxos_lease_timeout);
lease_timeout_event = 0;
mon->bootstrap();
}
@@ -1112,8 +1222,14 @@
f.flush(*_dout);
*_dout << dendl;
+ logger->inc(l_paxos_new_pn);
+ utime_t start = ceph_clock_now(NULL);
+
get_store()->apply_transaction(t);
+ utime_t end = ceph_clock_now(NULL);
+ logger->tinc(l_paxos_new_pn_latency, end - start);
+
dout(10) << "get_new_proposal_number = " << last_pn << dendl;
return last_pn;
}
@@ -1150,6 +1266,9 @@
finish_contexts(g_ceph_context, waiting_for_readable, -ECANCELED);
finish_contexts(g_ceph_context, waiting_for_active, -ECANCELED);
finish_contexts(g_ceph_context, proposals, -ECANCELED);
+ if (logger)
+ g_ceph_context->get_perfcounters_collection()->remove(logger);
+ delete logger;
}
void Paxos::leader_init()
@@ -1159,6 +1278,8 @@
finish_contexts(g_ceph_context, proposals, -EAGAIN);
+ logger->inc(l_paxos_start_leader);
+
if (mon->get_quorum().size() == 1) {
state = STATE_ACTIVE;
return;
@@ -1186,6 +1307,8 @@
finish_contexts(g_ceph_context, waiting_for_writeable, -EAGAIN);
finish_contexts(g_ceph_context, waiting_for_commit, -EAGAIN);
finish_contexts(g_ceph_context, proposals, -EAGAIN);
+
+ logger->inc(l_paxos_start_peon);
}
void Paxos::restart()
@@ -1199,6 +1322,8 @@
finish_contexts(g_ceph_context, proposals, -EAGAIN);
finish_contexts(g_ceph_context, waiting_for_commit, -EAGAIN);
finish_contexts(g_ceph_context, waiting_for_active, -EAGAIN);
+
+ logger->inc(l_paxos_restart);
}
diff -Nru ceph-0.80.7/src/mon/Paxos.h ceph-0.80.9/src/mon/Paxos.h
--- ceph-0.80.7/src/mon/Paxos.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/Paxos.h 2015-03-09 20:51:18.000000000 +0100
@@ -118,6 +118,7 @@
#include "include/Context.h"
#include "common/Timer.h"
+#include "common/perf_counters.h"
#include <errno.h>
#include "MonitorDBStore.h"
@@ -126,6 +127,43 @@
class MMonPaxos;
class Paxos;
+enum {
+ l_paxos_first = 45800,
+ l_paxos_start_leader,
+ l_paxos_start_peon,
+ l_paxos_restart,
+ l_paxos_refresh,
+ l_paxos_refresh_latency,
+ l_paxos_begin,
+ l_paxos_begin_keys,
+ l_paxos_begin_bytes,
+ l_paxos_begin_latency,
+ l_paxos_commit,
+ l_paxos_commit_keys,
+ l_paxos_commit_bytes,
+ l_paxos_commit_latency,
+ l_paxos_collect,
+ l_paxos_collect_keys,
+ l_paxos_collect_bytes,
+ l_paxos_collect_latency,
+ l_paxos_collect_uncommitted,
+ l_paxos_collect_timeout,
+ l_paxos_accept_timeout,
+ l_paxos_lease_ack_timeout,
+ l_paxos_lease_timeout,
+ l_paxos_store_state,
+ l_paxos_store_state_keys,
+ l_paxos_store_state_bytes,
+ l_paxos_store_state_latency,
+ l_paxos_share_state,
+ l_paxos_share_state_keys,
+ l_paxos_share_state_bytes,
+ l_paxos_new_pn,
+ l_paxos_new_pn_latency,
+ l_paxos_last,
+};
+
+
// i am one state machine.
/**
* This libary is based on the Paxos algorithm, but varies in a few key ways:
@@ -147,6 +185,11 @@
*/
Monitor *mon;
+ /// perf counter for internal instrumentations
+ PerfCounters *logger;
+
+ void init_logger();
+
// my state machine info
const string paxos_name;
@@ -1004,6 +1047,7 @@
*/
Paxos(Monitor *m, const string &name)
: mon(m),
+ logger(NULL),
paxos_name(name),
state(STATE_RECOVERING),
first_committed(0),
diff -Nru ceph-0.80.7/src/mon/PGMap.cc ceph-0.80.9/src/mon/PGMap.cc
--- ceph-0.80.7/src/mon/PGMap.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/PGMap.cc 2015-03-09 20:51:18.000000000 +0100
@@ -379,17 +379,31 @@
{
bufferlist::iterator p = bl.begin();
ceph::unordered_map<pg_t,pg_stat_t>::iterator s = pg_stat.find(pgid);
- if (s != pg_stat.end())
+ epoch_t old_lec = 0;
+ if (s != pg_stat.end()) {
+ old_lec = s->second.get_effective_last_epoch_clean();
stat_pg_sub(pgid, s->second);
+ }
pg_stat_t& r = pg_stat[pgid];
::decode(r, p);
stat_pg_add(pgid, r);
+
+ epoch_t lec = r.get_effective_last_epoch_clean();
+ if (min_last_epoch_clean &&
+ (lec < min_last_epoch_clean || // we did
+ (lec > min_last_epoch_clean && // we might
+ old_lec == min_last_epoch_clean)
+ ))
+ min_last_epoch_clean = 0;
}
void PGMap::remove_pg(pg_t pgid)
{
ceph::unordered_map<pg_t,pg_stat_t>::iterator s = pg_stat.find(pgid);
if (s != pg_stat.end()) {
+ if (min_last_epoch_clean &&
+ s->second.get_effective_last_epoch_clean() == min_last_epoch_clean)
+ min_last_epoch_clean = 0;
stat_pg_sub(pgid, s->second);
pg_stat.erase(s);
}
@@ -399,14 +413,33 @@
{
bufferlist::iterator p = bl.begin();
ceph::unordered_map<int32_t,osd_stat_t>::iterator o = osd_stat.find(osd);
- if (o != osd_stat.end())
+ epoch_t old_lec = 0;
+ if (o != osd_stat.end()) {
+ ceph::unordered_map<int32_t,epoch_t>::iterator i = osd_epochs.find(osd);
+ if (i != osd_epochs.end())
+ old_lec = i->second;
stat_osd_sub(o->second);
+ }
osd_stat_t& r = osd_stat[osd];
::decode(r, p);
stat_osd_add(r);
// adjust [near]full status
register_nearfull_status(osd, r);
+
+ // epoch?
+ if (!p.end()) {
+ epoch_t e;
+ ::decode(e, p);
+
+ if (e < min_last_epoch_clean ||
+ (e > min_last_epoch_clean &&
+ old_lec == min_last_epoch_clean))
+ min_last_epoch_clean = 0;
+ } else {
+ // WARNING: we are not refreshing min_last_epoch_clean! must be old store
+ // or old mon running.
+ }
}
void PGMap::remove_osd(int osd)
diff -Nru ceph-0.80.7/src/mon/PGMonitor.cc ceph-0.80.9/src/mon/PGMonitor.cc
--- ceph-0.80.7/src/mon/PGMonitor.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/mon/PGMonitor.cc 2015-03-09 20:51:18.000000000 +0100
@@ -545,6 +545,7 @@
::encode(p->first, dirty);
bufferlist bl;
::encode(p->second, bl, features);
+ ::encode(pending_inc.get_osd_epochs().find(p->first)->second, bl);
t->put(prefix, stringify(p->first), bl);
}
for (set<int32_t>::const_iterator p =
diff -Nru ceph-0.80.7/src/os/FileJournal.cc ceph-0.80.9/src/os/FileJournal.cc
--- ceph-0.80.7/src/os/FileJournal.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/os/FileJournal.cc 2015-03-09 20:51:18.000000000 +0100
@@ -103,12 +103,14 @@
goto out_fd;
#ifdef HAVE_LIBAIO
- aio_ctx = 0;
- ret = io_setup(128, &aio_ctx);
- if (ret < 0) {
- ret = errno;
- derr << "FileJournal::_open: unable to setup io_context " << cpp_strerror(ret) << dendl;
- goto out_fd;
+ if (aio) {
+ aio_ctx = 0;
+ ret = io_setup(128, &aio_ctx);
+ if (ret < 0) {
+ ret = errno;
+ derr << "FileJournal::_open: unable to setup io_context " << cpp_strerror(ret) << dendl;
+ goto out_fd;
+ }
}
#endif
@@ -544,6 +546,7 @@
// close
assert(writeq_empty());
+ assert(!must_write_header);
assert(fd >= 0);
VOID_TEMP_FAILURE_RETRY(::close(fd));
fd = -1;
@@ -564,9 +567,9 @@
JSONFormatter f(true);
f.open_array_section("journal");
+ uint64_t seq = 0;
while (1) {
bufferlist bl;
- uint64_t seq = 0;
uint64_t pos = read_pos;
if (!read_entry(bl, seq)) {
dout(3) << "journal_replay: end of journal, done." << dendl;
@@ -604,7 +607,8 @@
write_stop = false;
write_thread.create();
#ifdef HAVE_LIBAIO
- write_finish_thread.create();
+ if (aio)
+ write_finish_thread.create();
#endif
}
@@ -613,19 +617,25 @@
{
Mutex::Locker l(write_lock);
#ifdef HAVE_LIBAIO
- Mutex::Locker q(aio_lock);
+ if (aio)
+ aio_lock.Lock();
#endif
Mutex::Locker p(writeq_lock);
write_stop = true;
writeq_cond.Signal();
#ifdef HAVE_LIBAIO
- aio_cond.Signal();
- write_finish_cond.Signal();
+ if (aio) {
+ aio_cond.Signal();
+ write_finish_cond.Signal();
+ aio_lock.Unlock();
+ }
#endif
}
write_thread.join();
#ifdef HAVE_LIBAIO
- write_finish_thread.join();
+ if (aio) {
+ write_finish_thread.join();
+ }
#endif
}
@@ -649,6 +659,13 @@
buffer::ptr bp = buffer::create_page_aligned(block_size);
bp.zero();
int r = ::pread(fd, bp.c_str(), bp.length(), 0);
+
+ if (r < 0) {
+ int err = errno;
+ dout(0) << "read_header got " << cpp_strerror(err) << dendl;
+ return -err;
+ }
+
bl.push_back(bp);
try {
@@ -660,11 +677,6 @@
return -EINVAL;
}
- if (r < 0) {
- int err = errno;
- dout(0) << "read_header got " << cpp_strerror(err) << dendl;
- return -err;
- }
/*
* Unfortunately we weren't initializing the flags field for new
@@ -794,7 +806,8 @@
}
dout(20) << "prepare_multi_write queue_pos now " << queue_pos << dendl;
- //assert(write_pos + bl.length() == queue_pos);
+ assert((write_pos + bl.length() == queue_pos) ||
+ (write_pos + bl.length() - header.max_size + get_top() == queue_pos));
return 0;
}
@@ -993,22 +1006,32 @@
dout(10) << "do_write wrapping, first bit at " << pos << " len " << first.length()
<< " second bit len " << second.length() << " (orig len " << bl.length() << ")" << dendl;
- if (write_bl(pos, first)) {
- derr << "FileJournal::do_write: write_bl(pos=" << pos
- << ") failed" << dendl;
- ceph_abort();
- }
- assert(pos == get_top());
+ //Save pos to write first piece second
+ off64_t first_pos = pos;
+ off64_t orig_pos;
+ pos = get_top();
+ // header too?
if (hbp.length()) {
// be sneaky: include the header in the second fragment
second.push_front(hbp);
pos = 0; // we included the header
}
+ // Write the second portion first possible with the header, so
+ // do_read_entry() won't even get a valid entry_header_t if there
+ // is a crash between the two writes.
+ orig_pos = pos;
if (write_bl(pos, second)) {
- derr << "FileJournal::do_write: write_bl(pos=" << pos
+ derr << "FileJournal::do_write: write_bl(pos=" << orig_pos
+ << ") failed" << dendl;
+ ceph_abort();
+ }
+ orig_pos = first_pos;
+ if (write_bl(first_pos, first)) {
+ derr << "FileJournal::do_write: write_bl(pos=" << orig_pos
<< ") failed" << dendl;
ceph_abort();
}
+ assert(first_pos == get_top());
} else {
// header too?
if (hbp.length()) {
@@ -1102,7 +1125,7 @@
while (1) {
{
Mutex::Locker locker(writeq_lock);
- if (writeq.empty()) {
+ if (writeq.empty() && !must_write_header) {
if (write_stop)
break;
dout(20) << "write_thread_entry going to sleep" << dendl;
@@ -1113,7 +1136,9 @@
}
#ifdef HAVE_LIBAIO
- if (aio) {
+ //We hope write_finish_thread_entry return until the last aios complete
+ //when set write_stop. But it can't. So don't use aio mode when shutdown.
+ if (aio && !write_stop) {
Mutex::Locker locker(aio_lock);
// should we back off to limit aios in flight? try to do this
// adaptively so that we submit larger aios once we have lots of
@@ -1164,7 +1189,7 @@
}
#ifdef HAVE_LIBAIO
- if (aio)
+ if (aio && !write_stop)
do_aio_write(bl);
else
do_write(bl);
@@ -1353,7 +1378,7 @@
aio_info *ai = (aio_info *)event[i].obj;
if (event[i].res != ai->len) {
derr << "aio to " << ai->off << "~" << ai->len
- << " got " << cpp_strerror(event[i].res) << dendl;
+ << " wrote " << event[i].res << dendl;
assert(0 == "unexpected aio error");
}
dout(10) << "write_finish_thread_entry aio " << ai->off
@@ -1376,7 +1401,7 @@
assert(aio_lock.is_locked());
dout(20) << "check_aio_completion" << dendl;
- bool completed_something = false;
+ bool completed_something = false, signal = false;
uint64_t new_journaled_seq = 0;
list<aio_info>::iterator p = aio_queue.begin();
@@ -1390,6 +1415,7 @@
aio_num--;
aio_bytes -= p->len;
aio_queue.erase(p++);
+ signal = true;
}
if (completed_something) {
@@ -1409,7 +1435,8 @@
queue_completions_thru(journaled_seq);
}
}
-
+ }
+ if (signal) {
// maybe write queue was waiting for aio count to drop?
aio_cond.Signal();
}
diff -Nru ceph-0.80.7/src/osd/ECBackend.cc ceph-0.80.9/src/osd/ECBackend.cc
--- ceph-0.80.7/src/osd/ECBackend.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/ECBackend.cc 2015-03-09 20:51:18.000000000 +0100
@@ -1245,10 +1245,18 @@
for (set<hobject_t>::iterator i = need_hinfos.begin();
i != need_hinfos.end();
++i) {
+ ECUtil::HashInfoRef ref = get_hash_info(*i);
+ if (!ref) {
+ derr << __func__ << ": get_hash_info(" << *i << ")"
+ << " returned a null pointer and there is no "
+ << " way to recover from such an error in this "
+ << " context" << dendl;
+ assert(0);
+ }
op->unstable_hash_infos.insert(
make_pair(
*i,
- get_hash_info(*i)));
+ ref));
}
for (vector<pg_log_entry_t>::iterator i = op->log_entries.begin();
@@ -1458,7 +1466,7 @@
::decode(hinfo, bp);
assert(hinfo.get_total_chunk_size() == (unsigned)st.st_size);
} else {
- assert(0 == "missing hash attr");
+ return ECUtil::HashInfoRef();
}
}
ref = unstable_hashinfo_registry.lookup_or_create(hoid, hinfo);
@@ -1754,32 +1762,38 @@
break;
}
- ECUtil::HashInfoRef hinfo = get_hash_info(poid);
if (r == -EIO) {
dout(0) << "_scan_list " << poid << " got "
<< r << " on read, read_error" << dendl;
o.read_error = true;
}
- if (hinfo->get_chunk_hash(get_parent()->whoami_shard().shard) != h.digest()) {
- dout(0) << "_scan_list " << poid << " got incorrect hash on read" << dendl;
- o.read_error = true;
- }
-
- if (hinfo->get_total_chunk_size() != pos) {
- dout(0) << "_scan_list " << poid << " got incorrect size on read" << dendl;
+ ECUtil::HashInfoRef hinfo = get_hash_info(poid);
+ if (!hinfo) {
+ dout(0) << "_scan_list " << poid << " could not retrieve hash info" << dendl;
o.read_error = true;
+ o.digest_present = false;
+ } else {
+ if (hinfo->get_chunk_hash(get_parent()->whoami_shard().shard) != h.digest()) {
+ dout(0) << "_scan_list " << poid << " got incorrect hash on read" << dendl;
+ o.read_error = true;
+ }
+
+ if (hinfo->get_total_chunk_size() != pos) {
+ dout(0) << "_scan_list " << poid << " got incorrect size on read" << dendl;
+ o.read_error = true;
+ }
+
+ /* We checked above that we match our own stored hash. We cannot
+ * send a hash of the actual object, so instead we simply send
+ * our locally stored hash of shard 0 on the assumption that if
+ * we match our chunk hash and our recollection of the hash for
+ * chunk 0 matches that of our peers, there is likely no corruption.
+ */
+ o.digest = hinfo->get_chunk_hash(0);
+ o.digest_present = true;
}
- /* We checked above that we match our own stored hash. We cannot
- * send a hash of the actual object, so instead we simply send
- * our locally stored hash of shard 0 on the assumption that if
- * we match our chunk hash and our recollection of the hash for
- * chunk 0 matches that of our peers, there is likely no corruption.
- */
- o.digest = hinfo->get_chunk_hash(0);
- o.digest_present = true;
-
o.omap_digest = 0;
o.omap_digest_present = true;
}
diff -Nru ceph-0.80.7/src/osd/OSD.cc ceph-0.80.9/src/osd/OSD.cc
--- ceph-0.80.7/src/osd/OSD.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/OSD.cc 2015-03-09 20:51:18.000000000 +0100
@@ -3762,6 +3762,53 @@
monc->send_mon_message(mboot);
}
+bool OSD::_lsb_release_set (char *buf, const char *str, map<string,string> *pm, const char *key)
+{
+ if (strncmp (buf, str, strlen (str)) == 0) {
+ char *value;
+
+ if (buf[strlen(buf)-1] == '\n')
+ buf[strlen(buf)-1] = '\0';
+
+ value = buf + strlen (str) + 1;
+ (*pm)[key] = value;
+
+ return true;
+ }
+ return false;
+}
+
+void OSD::_lsb_release_parse (map<string,string> *pm)
+{
+ FILE *fp = NULL;
+ char buf[512];
+
+ fp = popen("lsb_release -idrc", "r");
+ if (!fp) {
+ int ret = -errno;
+ derr << "lsb_release_parse - failed to call lsb_release binary with error: " << cpp_strerror(ret) << dendl;
+ return;
+ }
+
+ while (fgets(buf, sizeof(buf) - 1, fp) != NULL) {
+ if (_lsb_release_set(buf, "Distributor ID:", pm, "distro"))
+ continue;
+ if (_lsb_release_set(buf, "Description:", pm, "distro_description"))
+ continue;
+ if (_lsb_release_set(buf, "Release:", pm, "distro_version"))
+ continue;
+ if (_lsb_release_set(buf, "Codename:", pm, "distro_codename"))
+ continue;
+
+ derr << "unhandled output: " << buf << dendl;
+ }
+
+ if (pclose(fp)) {
+ int ret = -errno;
+ derr << "lsb_release_parse - pclose failed: " << cpp_strerror(ret) << dendl;
+ }
+}
+
void OSD::_collect_metadata(map<string,string> *pm)
{
(*pm)["ceph_version"] = pretty_version_to_str();
@@ -3831,34 +3878,7 @@
}
// distro info
- f = fopen("/etc/lsb-release", "r");
- if (f) {
- char buf[100];
- while (!feof(f)) {
- char *line = fgets(buf, sizeof(buf), f);
- if (!line)
- break;
- char *eq = strchr(buf, '=');
- if (!eq)
- break;
- *eq = '\0';
- ++eq;
- while (*eq == '\"')
- ++eq;
- while (*eq && (eq[strlen(eq)-1] == '\n' ||
- eq[strlen(eq)-1] == '\"'))
- eq[strlen(eq)-1] = '\0';
- if (strcmp(buf, "DISTRIB_ID") == 0)
- (*pm)["distro"] = eq;
- else if (strcmp(buf, "DISTRIB_RELEASE") == 0)
- (*pm)["distro_version"] = eq;
- else if (strcmp(buf, "DISTRIB_CODENAME") == 0)
- (*pm)["distro_codename"] = eq;
- else if (strcmp(buf, "DISTRIB_DESCRIPTION") == 0)
- (*pm)["distro_description"] = eq;
- }
- fclose(f);
- }
+ _lsb_release_parse(pm);
dout(10) << __func__ << " " << *pm << dendl;
}
@@ -5784,8 +5804,13 @@
next_epoch <= osd_epoch && next_epoch <= max;
++next_epoch) {
OSDMapRef nextmap = service.try_get_map(next_epoch);
- if (!nextmap)
+ if (!nextmap) {
+ dout(20) << __func__ << " missing map " << next_epoch << dendl;
+ // make sure max is bumped up so that we can get past any
+ // gap in maps
+ max = MAX(max, next_epoch + g_conf->osd_map_max_advance);
continue;
+ }
vector<int> newup, newacting;
int up_primary, acting_primary;
@@ -5816,7 +5841,7 @@
service.pg_update_epoch(pg->info.pgid, lastmap->get_epoch());
pg->handle_activate_map(rctx);
if (next_epoch <= osd_epoch) {
- dout(10) << __func__ << " advanced by max " << g_conf->osd_map_max_advance
+ dout(10) << __func__ << " advanced to max " << max
<< " past min epoch " << min_epoch
<< " ... will requeue " << *pg << dendl;
return false;
@@ -5874,10 +5899,7 @@
while (p != waiting_for_pg.end()) {
spg_t pgid = p->first;
- vector<int> acting;
- int nrep = osdmap->pg_to_acting_osds(pgid.pgid, acting);
- int role = osdmap->calc_pg_role(whoami, acting, nrep);
- if (role >= 0) {
+ if (osdmap->osd_is_valid_op_target(pgid.pgid, whoami)) {
++p; // still me
} else {
dout(10) << " discarding waiting ops for " << pgid << dendl;
@@ -6732,7 +6754,8 @@
PG::CephPeeringEvtRef(
new PG::CephPeeringEvt(
it->first.epoch_sent, it->first.query_epoch,
- PG::MNotifyRec(pg_shard_t(from, it->first.from), it->first)))
+ PG::MNotifyRec(pg_shard_t(from, it->first.from), it->first,
+ op->get_req()->get_connection()->get_features())))
);
}
}
@@ -7564,7 +7587,7 @@
if (!pg) {
dout(7) << "hit non-existent pg " << pgid << dendl;
- if (osdmap->get_pg_acting_role(pgid.pgid, whoami) >= 0) {
+ if (osdmap->osd_is_valid_op_target(pgid.pgid, whoami)) {
dout(7) << "we are valid target for op, waiting" << dendl;
waiting_for_pg[pgid].push_back(op);
op->mark_delayed("waiting for pg to exist locally");
@@ -7578,7 +7601,7 @@
}
OSDMapRef send_map = get_map(m->get_map_epoch());
- if (send_map->get_pg_acting_role(pgid.pgid, whoami) >= 0) {
+ if (send_map->osd_is_valid_op_target(pgid.pgid, whoami)) {
dout(7) << "dropping request; client will resend when they get new map" << dendl;
} else if (!send_map->have_pg_pool(pgid.pool())) {
dout(7) << "dropping request; pool did not exist" << dendl;
@@ -7835,7 +7858,9 @@
continue;
}
if (!advance_pg(curmap->get_epoch(), pg, handle, &rctx, &split_pgs)) {
- pg->queue_null(curmap->get_epoch(), curmap->get_epoch());
+ // we need to requeue the PG explicitly since we didn't actually
+ // handle an event
+ peering_wq.queue(pg);
} else if (!pg->peering_queue.empty()) {
PG::CephPeeringEvtRef evt = pg->peering_queue.front();
pg->peering_queue.pop_front();
@@ -7931,7 +7956,12 @@
<< dendl;
int cls =
ceph_ioprio_string_to_class(cct->_conf->osd_disk_thread_ioprio_class);
- disk_tp.set_ioprio(cls, cct->_conf->osd_disk_thread_ioprio_priority);
+ if (cls < 0)
+ derr << __func__ << cpp_strerror(cls) << ": "
+ << "osd_disk_thread_ioprio_class is " << cct->_conf->osd_disk_thread_ioprio_class
+ << " but only the following values are allowed: idle, be or rt" << dendl;
+ else
+ disk_tp.set_ioprio(cls, cct->_conf->osd_disk_thread_ioprio_priority);
}
// --------------------------------
diff -Nru ceph-0.80.7/src/osd/OSD.h ceph-0.80.9/src/osd/OSD.h
--- ceph-0.80.7/src/osd/OSD.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/OSD.h 2015-03-09 20:51:18.000000000 +0100
@@ -1472,6 +1472,8 @@
void _maybe_boot(epoch_t oldest, epoch_t newest);
void _send_boot();
void _collect_metadata(map<string,string> *pmeta);
+ bool _lsb_release_set(char *buf, const char *str, map<string,string> *pm, const char *key);
+ void _lsb_release_parse (map<string,string> *pm);
void start_waiting_for_healthy();
bool _is_healthy();
@@ -1738,7 +1740,9 @@
pg->put("SnapTrimWQ");
}
void _clear() {
- osd->snap_trim_queue.clear();
+ while (PG *pg = _dequeue()) {
+ pg->put("SnapTrimWQ");
+ }
}
} snap_trim_wq;
diff -Nru ceph-0.80.7/src/osd/OSDMap.h ceph-0.80.9/src/osd/OSDMap.h
--- ceph-0.80.7/src/osd/OSDMap.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/OSDMap.h 2015-03-09 20:51:18.000000000 +0100
@@ -777,6 +777,18 @@
return calc_pg_role(osd, group, nrep);
}
+ bool osd_is_valid_op_target(pg_t pg, int osd) const {
+ int primary;
+ vector<int> group;
+ int nrep = pg_to_acting_osds(pg, &group, &primary);
+ if (osd == primary)
+ return true;
+ if (pg_is_ec(pg))
+ return false;
+
+ return calc_pg_role(osd, group, nrep) >= 0;
+ }
+
/*
* handy helpers to build simple maps...
diff -Nru ceph-0.80.7/src/osd/osd_types.cc ceph-0.80.9/src/osd/osd_types.cc
--- ceph-0.80.7/src/osd/osd_types.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/osd_types.cc 2015-03-09 20:51:18.000000000 +0100
@@ -2179,9 +2179,9 @@
f->open_array_section("acting");
for (vector<int>::const_iterator p = acting.begin(); p != acting.end(); ++p)
f->dump_int("osd", *p);
+ f->close_section();
f->dump_int("primary", primary);
f->dump_int("up_primary", up_primary);
- f->close_section();
}
void pg_interval_t::generate_test_instances(list<pg_interval_t*>& o)
@@ -2235,9 +2235,15 @@
i.primary = old_acting_primary;
i.up_primary = old_up_primary;
- if (!i.acting.empty() && i.primary != -1 &&
- i.acting.size() >=
- lastmap->get_pools().find(pool_id)->second.min_size) {
+ unsigned num_acting = 0;
+ for (vector<int>::const_iterator p = i.acting.begin(); p != i.acting.end();
+ ++p)
+ if (*p != CRUSH_ITEM_NONE)
+ ++num_acting;
+
+ if (num_acting &&
+ i.primary != -1 &&
+ num_acting >= lastmap->get_pools().find(pgid.pool())->second.min_size) {
if (out)
*out << "generate_past_intervals " << i
<< ": not rw,"
diff -Nru ceph-0.80.7/src/osd/PG.cc ceph-0.80.9/src/osd/PG.cc
--- ceph-0.80.7/src/osd/PG.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/PG.cc 2015-03-09 20:51:18.000000000 +0100
@@ -194,7 +194,8 @@
finish_sync_event(NULL),
scrub_after_recovery(false),
active_pushes(0),
- recovery_state(this)
+ recovery_state(this),
+ peer_features((uint64_t)-1)
{
#ifdef PG_DEBUG_REFS
osd->add_pgid(p, this);
@@ -1552,6 +1553,9 @@
pi.hit_set = info.hit_set;
pi.stats.stats.clear();
+ // initialize peer with our purged_snaps.
+ pi.purged_snaps = info.purged_snaps;
+
m = new MOSDPGLog(
i->shard, pg_whoami.shard,
get_osdmap()->get_epoch(), pi);
@@ -4685,7 +4689,10 @@
void PG::set_last_peering_reset()
{
dout(20) << "set_last_peering_reset " << get_osdmap()->get_epoch() << dendl;
- last_peering_reset = get_osdmap()->get_epoch();
+ if (last_peering_reset != get_osdmap()->get_epoch()) {
+ last_peering_reset = get_osdmap()->get_epoch();
+ reset_interval_flush();
+ }
}
struct FlushState {
@@ -4739,7 +4746,6 @@
const OSDMapRef osdmap = get_osdmap();
set_last_peering_reset();
- reset_interval_flush();
vector<int> oldacting, oldup;
int oldrole = get_role();
@@ -5277,37 +5283,6 @@
osd->queue_for_peering(this);
}
-void PG::queue_notify(epoch_t msg_epoch,
- epoch_t query_epoch,
- pg_shard_t from, pg_notify_t& i)
-{
- dout(10) << "notify " << i << " from replica " << from << dendl;
- queue_peering_event(
- CephPeeringEvtRef(new CephPeeringEvt(msg_epoch, query_epoch,
- MNotifyRec(from, i))));
-}
-
-void PG::queue_info(epoch_t msg_epoch,
- epoch_t query_epoch,
- pg_shard_t from, pg_info_t& i)
-{
- dout(10) << "info " << i << " from replica " << from << dendl;
- queue_peering_event(
- CephPeeringEvtRef(new CephPeeringEvt(msg_epoch, query_epoch,
- MInfoRec(from, i, msg_epoch))));
-}
-
-void PG::queue_log(epoch_t msg_epoch,
- epoch_t query_epoch,
- pg_shard_t from,
- MOSDPGLog *msg)
-{
- dout(10) << "log " << *msg << " from replica " << from << dendl;
- queue_peering_event(
- CephPeeringEvtRef(new CephPeeringEvt(msg_epoch, query_epoch,
- MLogRec(from, msg))));
-}
-
void PG::queue_null(epoch_t msg_epoch,
epoch_t query_epoch)
{
@@ -5810,8 +5785,29 @@
pg->osd->local_reserver.cancel_reservation(pg->info.pgid);
pg->state_set(PG_STATE_BACKFILL_TOOFULL);
+ for (set<pg_shard_t>::iterator it = pg->backfill_targets.begin();
+ it != pg->backfill_targets.end();
+ ++it) {
+ assert(*it != pg->pg_whoami);
+ ConnectionRef con = pg->osd->get_con_osd_cluster(
+ it->osd, pg->get_osdmap()->get_epoch());
+ if (con) {
+ if (con->has_feature(CEPH_FEATURE_BACKFILL_RESERVATION)) {
+ pg->osd->send_message_osd_cluster(
+ new MBackfillReserve(
+ MBackfillReserve::REJECT,
+ spg_t(pg->info.pgid.pgid, it->shard),
+ pg->get_osdmap()->get_epoch()),
+ con.get());
+ }
+ }
+ }
+
pg->osd->recovery_wq.dequeue(pg);
+ pg->waiting_on_backfill.clear();
+ pg->finish_recovery_op(hobject_t::get_max());
+
pg->schedule_backfill_full_retry();
return transit<NotBackfilling>();
}
@@ -6066,14 +6062,33 @@
PG::RecoveryState::RepWaitBackfillReserved::react(const RemoteBackfillReserved &evt)
{
PG *pg = context< RecoveryMachine >().pg;
- pg->osd->send_message_osd_cluster(
- pg->primary.osd,
- new MBackfillReserve(
- MBackfillReserve::GRANT,
- spg_t(pg->info.pgid.pgid, pg->primary.shard),
- pg->get_osdmap()->get_epoch()),
- pg->get_osdmap()->get_epoch());
- return transit<RepRecovering>();
+
+ double ratio, max_ratio;
+ if (g_conf->osd_debug_reject_backfill_probability > 0 &&
+ (rand()%1000 < (g_conf->osd_debug_reject_backfill_probability*1000.0))) {
+ dout(10) << "backfill reservation rejected after reservation: "
+ << "failure injection" << dendl;
+ pg->osd->remote_reserver.cancel_reservation(pg->info.pgid);
+ post_event(RemoteReservationRejected());
+ return discard_event();
+ } else if (pg->osd->too_full_for_backfill(&ratio, &max_ratio) &&
+ !pg->cct->_conf->osd_debug_skip_full_check_in_backfill_reservation) {
+ dout(10) << "backfill reservation rejected after reservation: full ratio is "
+ << ratio << ", which is greater than max allowed ratio "
+ << max_ratio << dendl;
+ pg->osd->remote_reserver.cancel_reservation(pg->info.pgid);
+ post_event(RemoteReservationRejected());
+ return discard_event();
+ } else {
+ pg->osd->send_message_osd_cluster(
+ pg->primary.osd,
+ new MBackfillReserve(
+ MBackfillReserve::GRANT,
+ spg_t(pg->info.pgid.pgid, pg->primary.shard),
+ pg->get_osdmap()->get_epoch()),
+ pg->get_osdmap()->get_epoch());
+ return transit<RepRecovering>();
+ }
}
boost::statechart::result
@@ -6097,7 +6112,7 @@
{
PG *pg = context< RecoveryMachine >().pg;
pg->reject_reservation();
- return transit<RepNotRecovering>();
+ return discard_event();
}
void PG::RecoveryState::RepRecovering::exit()
@@ -6839,6 +6854,7 @@
pg->publish_stats_to_osd();
+ pg->reset_peer_features();
get_infos();
if (peer_info_requested.empty() && !prior_set->pg_down) {
post_event(GotInfo());
@@ -6906,6 +6922,9 @@
}
get_infos();
}
+ dout(20) << "Adding osd: " << infoevt.from.osd << " features: "
+ << hex << infoevt.features << dec << dendl;
+ pg->apply_peer_features(infoevt.features);
// are we done getting everything?
if (peer_info_requested.empty() && !prior_set->pg_down) {
@@ -6964,6 +6983,7 @@
break;
}
}
+ dout(20) << "Common features: " << hex << pg->get_min_peer_features() << dec << dendl;
post_event(GotInfo());
}
}
diff -Nru ceph-0.80.7/src/osd/PG.h ceph-0.80.9/src/osd/PG.h
--- ceph-0.80.7/src/osd/PG.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/PG.h 2015-03-09 20:51:18.000000000 +0100
@@ -1308,10 +1308,12 @@
struct MNotifyRec : boost::statechart::event< MNotifyRec > {
pg_shard_t from;
pg_notify_t notify;
- MNotifyRec(pg_shard_t from, pg_notify_t ¬ify) :
- from(from), notify(notify) {}
+ uint64_t features;
+ MNotifyRec(pg_shard_t from, pg_notify_t ¬ify, uint64_t f) :
+ from(from), notify(notify), features(f) {}
void print(std::ostream *out) const {
- *out << "MNotifyRec from " << from << " notify: " << notify;
+ *out << "MNotifyRec from " << from << " notify: " << notify
+ << " features: 0x" << hex << features << dec;
}
};
@@ -1993,11 +1995,16 @@
// Prevent copying
PG(const PG& rhs);
PG& operator=(const PG& rhs);
+ uint64_t peer_features;
public:
spg_t get_pgid() const { return info.pgid; }
int get_nrep() const { return acting.size(); }
+ void reset_peer_features() { peer_features = (uint64_t)-1; }
+ uint64_t get_min_peer_features() { return peer_features; }
+ void apply_peer_features(uint64_t f) { peer_features &= f; }
+
void init_primary_up_acting(
const vector<int> &newup,
const vector<int> &newacting,
@@ -2189,12 +2196,6 @@
void take_waiters();
void queue_peering_event(CephPeeringEvtRef evt);
void handle_peering_event(CephPeeringEvtRef evt, RecoveryCtx *rctx);
- void queue_notify(epoch_t msg_epoch, epoch_t query_epoch,
- pg_shard_t from, pg_notify_t& i);
- void queue_info(epoch_t msg_epoch, epoch_t query_epoch,
- pg_shard_t from, pg_info_t& i);
- void queue_log(epoch_t msg_epoch, epoch_t query_epoch, pg_shard_t from,
- MOSDPGLog *msg);
void queue_query(epoch_t msg_epoch, epoch_t query_epoch,
pg_shard_t from, const pg_query_t& q);
void queue_null(epoch_t msg_epoch, epoch_t query_epoch);
diff -Nru ceph-0.80.7/src/osd/ReplicatedPG.cc ceph-0.80.9/src/osd/ReplicatedPG.cc
--- ceph-0.80.7/src/osd/ReplicatedPG.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/ReplicatedPG.cc 2015-03-09 20:51:18.000000000 +0100
@@ -595,6 +595,7 @@
if (command == "query") {
f->open_object_section("pg");
f->dump_string("state", pg_state_string(get_state()));
+ f->dump_stream("snap_trimq") << snap_trimq;
f->dump_unsigned("epoch", get_osdmap()->get_epoch());
f->open_array_section("up");
for (vector<int>::iterator p = up.begin(); p != up.end(); ++p)
@@ -2072,12 +2073,16 @@
}
peer_backfill_info[from] = bi;
- assert(waiting_on_backfill.find(from) != waiting_on_backfill.end());
- waiting_on_backfill.erase(from);
+ if (waiting_on_backfill.find(from) != waiting_on_backfill.end()) {
+ waiting_on_backfill.erase(from);
- if (waiting_on_backfill.empty()) {
- assert(peer_backfill_info.size() == backfill_targets.size());
- finish_recovery_op(hobject_t::get_max());
+ if (waiting_on_backfill.empty()) {
+ assert(peer_backfill_info.size() == backfill_targets.size());
+ finish_recovery_op(hobject_t::get_max());
+ }
+ } else {
+ // we canceled backfill for a while due to a too full, and this
+ // is an extra response from a non-too-full peer
}
}
break;
@@ -2560,10 +2565,6 @@
// replica collection trimming
snap_trimmer_machine.process_event(SnapTrim());
}
- if (snap_trimmer_machine.requeue) {
- dout(10) << "snap_trimmer requeue" << dendl;
- queue_snap_trim();
- }
unlock();
return;
}
@@ -3353,6 +3354,11 @@
break;
}
result = _delete_oid(ctx, true);
+ if (result >= 0) {
+ // mark that this is a cache eviction to avoid triggering normal
+ // make_writeable() clone or snapdir object creation in finish_ctx()
+ ctx->cache_evict = true;
+ }
osd->logger->inc(l_osd_tier_evict);
}
break;
@@ -3645,6 +3651,10 @@
case CEPH_OSD_OP_SETALLOCHINT:
++ctx->num_write;
{
+ if (!(get_min_peer_features() & CEPH_FEATURE_OSD_SET_ALLOC_HINT)) {
+ result = -EOPNOTSUPP;
+ break;
+ }
if (!obs.exists) {
ctx->mod_desc.create();
t->touch(soid);
@@ -4836,6 +4846,7 @@
if ((ctx->obs->exists && !ctx->obs->oi.is_whiteout()) && // head exist(ed)
snapc.snaps.size() && // there are snaps
+ !ctx->cache_evict &&
snapc.snaps[0] > ctx->new_snapset.seq) { // existing object is old
// clone
hobject_t coid = soid;
@@ -5156,7 +5167,9 @@
ctx->snapset_obc->obs.exists = false;
}
}
- } else if (ctx->new_snapset.clones.size()) {
+ } else if (ctx->new_snapset.clones.size() &&
+ !ctx->cache_evict &&
+ (!ctx->snapset_obc || !ctx->snapset_obc->obs.exists)) {
// save snapset on _snap
hobject_t snapoid(soid.oid, soid.get_key(), CEPH_SNAPDIR, soid.hash,
info.pgid.pool(), soid.get_namespace());
@@ -5167,7 +5180,8 @@
eversion_t(),
0, osd_reqid_t(), ctx->mtime));
- ctx->snapset_obc = get_object_context(snapoid, true);
+ if (!ctx->snapset_obc)
+ ctx->snapset_obc = get_object_context(snapoid, true);
bool got = ctx->snapset_obc->get_write_greedy(ctx->op);
assert(got);
dout(20) << " got greedy write on snapset_obc " << *ctx->snapset_obc << dendl;
@@ -6968,6 +6982,7 @@
if (get_osdmap()->is_blacklisted(ea)) {
dout(10) << "watch: Found blacklisted watcher for " << ea << dendl;
assert(j->second->get_pg() == this);
+ j->second->unregister_cb();
handle_watch_timeout(j->second);
}
}
@@ -9334,6 +9349,13 @@
// adjust info to backfill
info.last_backfill = hobject_t();
dirty_info = true;
+
+
+ // clear log
+ PGLogEntryHandler rollbacker;
+ pg_log.clear_can_rollback_to(&rollbacker);
+ rollbacker.apply(this, t);
+
write_if_dirty(*t);
on_shutdown();
@@ -10952,7 +10974,7 @@
pg_log_entry_t::MODIFY,
oid,
ctx->at_version,
- ctx->obs->oi.version,
+ eversion_t(),
0,
osd_reqid_t(),
ctx->mtime)
@@ -11974,13 +11996,11 @@
: my_base(ctx),
NamedState(context< SnapTrimmer >().pg->cct, "NotTrimming")
{
- context< SnapTrimmer >().requeue = false;
context< SnapTrimmer >().log_enter(state_name);
}
void ReplicatedPG::NotTrimming::exit()
{
- context< SnapTrimmer >().requeue = true;
context< SnapTrimmer >().log_exit(state_name, enter_time);
}
@@ -12040,32 +12060,45 @@
dout(10) << "TrimmingObjects: trimming snap " << snap_to_trim << dendl;
- // Get next
- hobject_t old_pos = pos;
- int r = pg->snap_mapper.get_next_object_to_trim(snap_to_trim, &pos);
- if (r != 0 && r != -ENOENT) {
- derr << __func__ << ": get_next returned " << cpp_strerror(r) << dendl;
- assert(0);
- } else if (r == -ENOENT) {
- // Done!
- dout(10) << "TrimmingObjects: got ENOENT" << dendl;
- post_event(SnapTrim());
- return transit< WaitingOnReplicas >();
+ for (set<RepGather *>::iterator i = repops.begin();
+ i != repops.end();
+ ) {
+ if ((*i)->all_applied && (*i)->all_committed) {
+ (*i)->put();
+ repops.erase(i++);
+ } else {
+ ++i;
+ }
}
- dout(10) << "TrimmingObjects react trimming " << pos << dendl;
- RepGather *repop = pg->trim_object(pos);
- if (!repop) {
- dout(10) << __func__ << " could not get write lock on obj "
- << pos << dendl;
- pos = old_pos;
- return discard_event();
- }
- assert(repop);
- repop->queue_snap_trimmer = true;
+ while (repops.size() < g_conf->osd_pg_max_concurrent_snap_trims) {
+ // Get next
+ hobject_t old_pos = pos;
+ int r = pg->snap_mapper.get_next_object_to_trim(snap_to_trim, &pos);
+ if (r != 0 && r != -ENOENT) {
+ derr << __func__ << ": get_next returned " << cpp_strerror(r) << dendl;
+ assert(0);
+ } else if (r == -ENOENT) {
+ // Done!
+ dout(10) << "TrimmingObjects: got ENOENT" << dendl;
+ post_event(SnapTrim());
+ return transit< WaitingOnReplicas >();
+ }
+
+ dout(10) << "TrimmingObjects react trimming " << pos << dendl;
+ RepGather *repop = pg->trim_object(pos);
+ if (!repop) {
+ dout(10) << __func__ << " could not get write lock on obj "
+ << pos << dendl;
+ pos = old_pos;
+ return discard_event();
+ }
+ assert(repop);
+ repop->queue_snap_trimmer = true;
- repops.insert(repop->get());
- pg->simple_repop_submit(repop);
+ repops.insert(repop->get());
+ pg->simple_repop_submit(repop);
+ }
return discard_event();
}
/* WaitingOnReplicasObjects */
@@ -12074,7 +12107,6 @@
NamedState(context< SnapTrimmer >().pg->cct, "Trimming/WaitingOnReplicas")
{
context< SnapTrimmer >().log_enter(state_name);
- context< SnapTrimmer >().requeue = false;
}
void ReplicatedPG::WaitingOnReplicas::exit()
@@ -12099,7 +12131,7 @@
for (set<RepGather *>::iterator i = repops.begin();
i != repops.end();
repops.erase(i++)) {
- if (!(*i)->all_applied) {
+ if (!(*i)->all_applied || !(*i)->all_committed) {
return discard_event();
} else {
(*i)->put();
@@ -12124,7 +12156,7 @@
context<SnapTrimmer>().need_share_pg_info = true;
// Back to the start
- post_event(SnapTrim());
+ pg->queue_snap_trim();
return transit< NotTrimming >();
}
diff -Nru ceph-0.80.7/src/osd/ReplicatedPG.h ceph-0.80.9/src/osd/ReplicatedPG.h
--- ceph-0.80.7/src/osd/ReplicatedPG.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/ReplicatedPG.h 2015-03-09 20:51:18.000000000 +0100
@@ -438,6 +438,7 @@
bool modify; // (force) modification (even if op_t is empty)
bool user_modify; // user-visible modification
bool undirty; // user explicitly un-dirtying this object
+ bool cache_evict; ///< true if this is a cache eviction
// side effects
list<watch_info_t> watch_connects;
@@ -539,7 +540,7 @@
ReplicatedPG *_pg) :
op(_op), reqid(_reqid), ops(_ops), obs(_obs), snapset(0),
new_obs(_obs->oi, _obs->exists),
- modify(false), user_modify(false), undirty(false),
+ modify(false), user_modify(false), undirty(false), cache_evict(false),
bytes_written(0), bytes_read(0), user_at_version(0),
current_osd_subop_num(0),
op_t(NULL),
@@ -1331,8 +1332,7 @@
set<RepGather *> repops;
snapid_t snap_to_trim;
bool need_share_pg_info;
- bool requeue;
- SnapTrimmer(ReplicatedPG *pg) : pg(pg), need_share_pg_info(false), requeue(false) {}
+ SnapTrimmer(ReplicatedPG *pg) : pg(pg), need_share_pg_info(false) {}
~SnapTrimmer();
void log_enter(const char *state_name);
void log_exit(const char *state_name, utime_t duration);
diff -Nru ceph-0.80.7/src/osd/Watch.h ceph-0.80.9/src/osd/Watch.h
--- ceph-0.80.7/src/osd/Watch.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osd/Watch.h 2015-03-09 20:51:18.000000000 +0100
@@ -98,6 +98,7 @@
/// removes the timeout callback, called on completion or cancellation
void unregister_cb();
public:
+
string gen_dbg_prefix() {
stringstream ss;
ss << "Notify(" << make_pair(cookie, notify_id) << " "
@@ -172,15 +173,15 @@
/// Registers the timeout callback with watch_timer
void register_cb();
- /// Unregisters the timeout callback
- void unregister_cb();
-
/// send a Notify message when connected for notif
void send_notify(NotifyRef notif);
/// Cleans up state on discard or remove (including Connection state, obc)
void discard_state();
public:
+ /// Unregisters the timeout callback
+ void unregister_cb();
+
/// NOTE: must be called with pg lock held
~Watch();
diff -Nru ceph-0.80.7/src/osdc/ObjectCacher.cc ceph-0.80.9/src/osdc/ObjectCacher.cc
--- ceph-0.80.7/src/osdc/ObjectCacher.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osdc/ObjectCacher.cc 2015-03-09 20:51:18.000000000 +0100
@@ -796,6 +796,8 @@
ldout(cct, 20) << "finishing waiters " << ls << dendl;
finish_contexts(cct, ls, err);
+ retry_waiting_reads();
+
--reads_outstanding;
read_cond.Signal();
}
@@ -1105,18 +1107,35 @@
// TODO: make read path not call _readx for every completion
hits.insert(errors.begin(), errors.end());
}
-
+
if (!missing.empty() || !rx.empty()) {
// read missing
for (map<loff_t, BufferHead*>::iterator bh_it = missing.begin();
bh_it != missing.end();
++bh_it) {
- bh_read(bh_it->second);
- if (success && onfinish) {
- ldout(cct, 10) << "readx missed, waiting on " << *bh_it->second
- << " off " << bh_it->first << dendl;
- bh_it->second->waitfor_read[bh_it->first].push_back( new C_RetryRead(this, rd, oset, onfinish) );
- }
+ uint64_t rx_bytes = static_cast<uint64_t>(
+ stat_rx + bh_it->second->length());
+ if (!waitfor_read.empty() || rx_bytes > max_size) {
+ // cache is full with concurrent reads -- wait for rx's to complete
+ // to constrain memory growth (especially during copy-ups)
+ if (success) {
+ ldout(cct, 10) << "readx missed, waiting on cache to complete "
+ << waitfor_read.size() << " blocked reads, "
+ << (MAX(rx_bytes, max_size) - max_size)
+ << " read bytes" << dendl;
+ waitfor_read.push_back(new C_RetryRead(this, rd, oset, onfinish));
+ }
+
+ bh_remove(o, bh_it->second);
+ delete bh_it->second;
+ } else {
+ bh_read(bh_it->second);
+ if (success && onfinish) {
+ ldout(cct, 10) << "readx missed, waiting on " << *bh_it->second
+ << " off " << bh_it->first << dendl;
+ bh_it->second->waitfor_read[bh_it->first].push_back( new C_RetryRead(this, rd, oset, onfinish) );
+ }
+ }
bytes_not_in_cache += bh_it->second->length();
success = false;
}
@@ -1230,7 +1249,7 @@
// no misses... success! do the read.
assert(!hit_ls.empty());
ldout(cct, 10) << "readx has all buffers" << dendl;
-
+
// ok, assemble into result buffer.
uint64_t pos = 0;
if (rd->bl && !error) {
@@ -1263,6 +1282,18 @@
return ret;
}
+void ObjectCacher::retry_waiting_reads()
+{
+ list<Context *> ls;
+ ls.swap(waitfor_read);
+
+ while (!ls.empty() && waitfor_read.empty()) {
+ Context *ctx = ls.front();
+ ls.pop_front();
+ ctx->complete(0);
+ }
+ waitfor_read.splice(waitfor_read.end(), ls);
+}
int ObjectCacher::writex(OSDWrite *wr, ObjectSet *oset, Mutex& wait_on_lock,
Context *onfreespace)
diff -Nru ceph-0.80.7/src/osdc/ObjectCacher.h ceph-0.80.9/src/osdc/ObjectCacher.h
--- ceph-0.80.7/src/osdc/ObjectCacher.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osdc/ObjectCacher.h 2015-03-09 20:51:18.000000000 +0100
@@ -341,6 +341,8 @@
vector<ceph::unordered_map<sobject_t, Object*> > objects; // indexed by pool_id
+ list<Context*> waitfor_read;
+
ceph_tid_t last_read_tid;
set<BufferHead*> dirty_bh;
@@ -457,6 +459,7 @@
int _readx(OSDRead *rd, ObjectSet *oset, Context *onfinish,
bool external_call);
+ void retry_waiting_reads();
public:
void bh_read_finish(int64_t poolid, sobject_t oid, ceph_tid_t tid,
diff -Nru ceph-0.80.7/src/osdc/Objecter.cc ceph-0.80.9/src/osdc/Objecter.cc
--- ceph-0.80.7/src/osdc/Objecter.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/osdc/Objecter.cc 2015-03-09 20:51:18.000000000 +0100
@@ -1615,7 +1615,10 @@
ops.erase(op->tid);
logger->set(l_osdc_op_active, ops.size());
- assert(check_latest_map_ops.find(op->tid) == check_latest_map_ops.end());
+
+ // our reply may have raced with pool deletion resulting in a map
+ // check in flight.
+ op_cancel_map_check(op);
if (op->ontimeout)
timer.cancel_event(op->ontimeout);
@@ -1640,7 +1643,9 @@
ldout(cct, 20) << " revoking rx buffer for " << op->tid << " on " << op->con << dendl;
op->con->revoke_rx_buffer(op->tid);
}
- if (op->outbl && op->outbl->length()) {
+ if (op->outbl &&
+ op->ontimeout == NULL && // only post rx_buffer if no timeout; see #9582
+ op->outbl->length()) {
ldout(cct, 20) << " posting rx buffer for " << op->tid << " on " << op->session->con << dendl;
op->con = op->session->con;
op->con->post_rx_buffer(op->tid, *op->outbl);
diff -Nru ceph-0.80.7/src/pybind/rados.py ceph-0.80.9/src/pybind/rados.py
--- ceph-0.80.7/src/pybind/rados.py 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/pybind/rados.py 2015-03-09 20:51:18.000000000 +0100
@@ -246,7 +246,8 @@
def shutdown(self):
"""
- Disconnects from the cluster.
+ Disconnects from the cluster. Call this explicitly when a
+ Rados.connect()ed object is no longer used.
"""
if (self.__dict__.has_key("state") and self.state != "shutdown"):
run_in_thread(self.librados.rados_shutdown, (self.cluster,))
@@ -260,9 +261,6 @@
self.shutdown()
return False
- def __del__(self):
- self.shutdown()
-
def version(self):
"""
Get the version number of the ``librados`` C library.
@@ -410,7 +408,7 @@
def connect(self, timeout=0):
"""
- Connect to the cluster.
+ Connect to the cluster. Use shutdown() to release resources.
"""
self.require_state("configuring")
ret = run_in_thread(self.librados.rados_connect, (self.cluster,),
diff -Nru ceph-0.80.7/src/rgw/Makefile.am ceph-0.80.9/src/rgw/Makefile.am
--- ceph-0.80.7/src/rgw/Makefile.am 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/Makefile.am 2015-03-09 20:51:18.000000000 +0100
@@ -53,6 +53,18 @@
-lfcgi \
-ldl
+CIVETWEB_INCLUDE = --include civetweb/include/civetweb_conf.h
+
+libcivetweb_la_SOURCES = \
+ rgw/rgw_civetweb.cc \
+ rgw/rgw_civetweb_log.cc \
+ civetweb/src/civetweb.c
+
+libcivetweb_la_CXXFLAGS = ${CIVETWEB_INCLUDE} -Woverloaded-virtual ${AM_CXXFLAGS}
+libcivetweb_la_CFLAGS = -Icivetweb/include ${CIVETWEB_INCLUDE}
+
+noinst_LTLIBRARIES += libcivetweb.la
+
radosgw_SOURCES = \
rgw/rgw_resolve.cc \
rgw/rgw_rest.cc \
@@ -71,11 +83,9 @@
rgw/rgw_swift.cc \
rgw/rgw_swift_auth.cc \
rgw/rgw_loadgen.cc \
- rgw/rgw_civetweb.cc \
- civetweb/src/civetweb.c \
rgw/rgw_main.cc
-radosgw_CFLAGS = -Icivetweb/include
-radosgw_LDADD = $(LIBRGW) $(LIBRGW_DEPS) $(RESOLV_LIBS) $(CEPH_GLOBAL)
+radosgw_CFLAGS = -I$(srcdir)/civetweb/include
+radosgw_LDADD = $(LIBRGW) $(LIBCIVETWEB) $(LIBRGW_DEPS) $(RESOLV_LIBS) $(CEPH_GLOBAL)
bin_PROGRAMS += radosgw
radosgw_admin_SOURCES = rgw/rgw_admin.cc
@@ -158,7 +168,9 @@
rgw/rgw_bucket.h \
rgw/rgw_keystone.h \
rgw/rgw_civetweb.h \
+ rgw/rgw_civetweb_log.h \
civetweb/civetweb.h \
civetweb/include/civetweb.h \
+ civetweb/include/civetweb_conf.h \
civetweb/src/md5.h
diff -Nru ceph-0.80.7/src/rgw/rgw_admin.cc ceph-0.80.9/src/rgw/rgw_admin.cc
--- ceph-0.80.7/src/rgw/rgw_admin.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_admin.cc 2015-03-09 20:51:18.000000000 +0100
@@ -679,7 +679,11 @@
quota.max_objects = max_objects;
}
if (have_max_size) {
- quota.max_size_kb = rgw_rounded_kb(max_size);
+ if (max_size < 0) {
+ quota.max_size_kb = -1;
+ } else {
+ quota.max_size_kb = rgw_rounded_kb(max_size);
+ }
}
break;
case OPT_QUOTA_DISABLE:
@@ -1364,7 +1368,13 @@
cerr << "could not create user: " << err_msg << std::endl;
return -ret;
}
-
+ if (!subuser.empty()) {
+ ret = user.subusers.add(user_op, &err_msg);
+ if (ret < 0) {
+ cerr << "could not create subuser: " << err_msg << std::endl;
+ return -ret;
+ }
+ }
break;
case OPT_USER_RM:
ret = user.remove(user_op, &err_msg);
diff -Nru ceph-0.80.7/src/rgw/rgw_civetweb.cc ceph-0.80.9/src/rgw/rgw_civetweb.cc
--- ceph-0.80.7/src/rgw/rgw_civetweb.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_civetweb.cc 2015-03-09 20:51:18.000000000 +0100
@@ -42,7 +42,7 @@
if (0 && data.length() == 0) {
has_content_length = true;
- print("Transfer-Enconding: %s\n", "chunked");
+ print("Transfer-Enconding: %s\r\n", "chunked");
data.append("0\r\n\r\n", sizeof("0\r\n\r\n")-1);
} else {
int r = send_content_length(data.length());
@@ -128,7 +128,7 @@
if (!status_name)
status_name = "";
- snprintf(buf, sizeof(buf), "HTTP/1.1 %s %s\n", status, status_name);
+ snprintf(buf, sizeof(buf), "HTTP/1.1 %s %s\r\n", status, status_name);
bufferlist bl;
bl.append(buf);
@@ -168,5 +168,5 @@
has_content_length = true;
char buf[21];
snprintf(buf, sizeof(buf), "%"PRIu64, len);
- return print("Content-Length: %s\n", buf);
+ return print("Content-Length: %s\r\n", buf);
}
diff -Nru ceph-0.80.7/src/rgw/rgw_civetweb_log.cc ceph-0.80.9/src/rgw/rgw_civetweb_log.cc
--- ceph-0.80.7/src/rgw/rgw_civetweb_log.cc 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/rgw/rgw_civetweb_log.cc 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,14 @@
+#include "common/config.h"
+#include "rgw_common.h"
+
+#include "civetweb/civetweb.h"
+
+#define dout_subsys ceph_subsys_civetweb
+
+
+int rgw_civetweb_log_callback(const struct mg_connection *conn, const char *buf) {
+ dout(10) << "civetweb: " << (void *)conn << ": " << buf << dendl;
+ return 0;
+}
+
+
diff -Nru ceph-0.80.7/src/rgw/rgw_civetweb_log.h ceph-0.80.9/src/rgw/rgw_civetweb_log.h
--- ceph-0.80.7/src/rgw/rgw_civetweb_log.h 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/rgw/rgw_civetweb_log.h 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,6 @@
+#ifndef CEPH_RGW_CIVETWEB_LOG_H
+#define CEPH_RGW_CIVETWEB_LOG_H
+
+int rgw_civetweb_log_callback(const struct mg_connection *conn, const char *buf);
+
+#endif
diff -Nru ceph-0.80.7/src/rgw/rgw_common.cc ceph-0.80.9/src/rgw/rgw_common.cc
--- ceph-0.80.7/src/rgw/rgw_common.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_common.cc 2015-03-09 20:51:18.000000000 +0100
@@ -506,7 +506,7 @@
}
string substr, nameval;
substr = str.substr(pos, fpos - pos);
- url_decode(substr, nameval);
+ url_decode(substr, nameval, true);
NameVal nv(nameval);
int ret = nv.parse();
if (ret >= 0) {
@@ -690,14 +690,13 @@
return hex_table.to_num(c);
}
-bool url_decode(string& src_str, string& dest_str)
+bool url_decode(string& src_str, string& dest_str, bool in_query)
{
const char *src = src_str.c_str();
char dest[src_str.size() + 1];
int pos = 0;
char c;
- bool in_query = false;
while (*src) {
if (*src != '%') {
if (!in_query || *src != '+') {
diff -Nru ceph-0.80.7/src/rgw/rgw_common.h ceph-0.80.9/src/rgw/rgw_common.h
--- ceph-0.80.7/src/rgw/rgw_common.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_common.h 2015-03-09 20:51:18.000000000 +0100
@@ -1343,7 +1343,7 @@
extern bool verify_object_permission(struct req_state *s, int perm);
/** Convert an input URL into a sane object name
* by converting %-escaped strings into characters, etc*/
-extern bool url_decode(string& src_str, string& dest_str);
+extern bool url_decode(string& src_str, string& dest_str, bool in_query = false);
extern void url_encode(const string& src, string& dst);
extern void calc_hmac_sha1(const char *key, int key_len,
diff -Nru ceph-0.80.7/src/rgw/rgw_fcgi.cc ceph-0.80.9/src/rgw/rgw_fcgi.cc
--- ceph-0.80.7/src/rgw/rgw_fcgi.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_fcgi.cc 2015-03-09 20:51:18.000000000 +0100
@@ -32,7 +32,7 @@
int RGWFCGX::send_status(const char *status, const char *status_name)
{
- return print("Status: %s\n", status);
+ return print("Status: %s %s\r\n", status, status_name);
}
int RGWFCGX::send_100_continue()
@@ -48,7 +48,7 @@
{
char buf[21];
snprintf(buf, sizeof(buf), "%"PRIu64, len);
- return print("Content-Length: %s\n", buf);
+ return print("Content-Length: %s\r\n", buf);
}
int RGWFCGX::complete_header()
diff -Nru ceph-0.80.7/src/rgw/rgw_http_client.cc ceph-0.80.9/src/rgw/rgw_http_client.cc
--- ceph-0.80.7/src/rgw/rgw_http_client.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_http_client.cc 2015-03-09 20:51:18.000000000 +0100
@@ -42,17 +42,8 @@
return ret;
}
-int RGWHTTPClient::process(const char *method, const char *url)
+static curl_slist *headers_to_slist(list<pair<string, string> >& headers)
{
- int ret = 0;
- CURL *curl_handle;
-
- char error_buf[CURL_ERROR_SIZE];
-
- curl_handle = curl_easy_init();
-
- dout(20) << "sending request to " << url << dendl;
-
curl_slist *h = NULL;
list<pair<string, string> >::iterator iter;
@@ -63,11 +54,37 @@
if (strncmp(val.c_str(), "HTTP_", 5) == 0) {
val = val.substr(5);
}
+
+ /* we need to convert all underscores into dashes as some web servers forbid them
+ * in the http header field names
+ */
+ for (size_t i = 0; i < val.size(); i++) {
+ if (val[i] == '_') {
+ val[i] = '-';
+ }
+ }
+
val.append(": ");
val.append(p.second);
h = curl_slist_append(h, val.c_str());
}
+ return h;
+}
+
+int RGWHTTPClient::process(const char *method, const char *url)
+{
+ int ret = 0;
+ CURL *curl_handle;
+
+ char error_buf[CURL_ERROR_SIZE];
+
+ curl_handle = curl_easy_init();
+
+ dout(20) << "sending request to " << url << dendl;
+
+ curl_slist *h = headers_to_slist(headers);
+
curl_easy_setopt(curl_handle, CURLOPT_CUSTOMREQUEST, method);
curl_easy_setopt(curl_handle, CURLOPT_URL, url);
curl_easy_setopt(curl_handle, CURLOPT_NOPROGRESS, 1L);
@@ -139,20 +156,7 @@
dout(20) << "sending request to " << url << dendl;
- curl_slist *h = NULL;
-
- list<pair<string, string> >::iterator iter;
- for (iter = headers.begin(); iter != headers.end(); ++iter) {
- pair<string, string>& p = *iter;
- string val = p.first;
-
- if (strncmp(val.c_str(), "HTTP_", 5) == 0) {
- val = val.substr(5);
- }
- val.append(": ");
- val.append(p.second);
- h = curl_slist_append(h, val.c_str());
- }
+ curl_slist *h = headers_to_slist(headers);
req_data->h = h;
diff -Nru ceph-0.80.7/src/rgw/rgw_json_enc.cc ceph-0.80.9/src/rgw/rgw_json_enc.cc
--- ceph-0.80.7/src/rgw/rgw_json_enc.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_json_enc.cc 2015-03-09 20:51:18.000000000 +0100
@@ -429,7 +429,7 @@
{
RGWAccessKey k;
k.decode_json(o, true);
- m[k.subuser] = k;
+ m[k.id] = k;
}
static void decode_subusers(map<string, RGWSubUser>& m, JSONObj *o)
diff -Nru ceph-0.80.7/src/rgw/rgw_main.cc ceph-0.80.9/src/rgw/rgw_main.cc
--- ceph-0.80.7/src/rgw/rgw_main.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_main.cc 2015-03-09 20:51:18.000000000 +0100
@@ -54,6 +54,7 @@
#include "rgw_resolve.h"
#include "rgw_loadgen.h"
#include "rgw_civetweb.h"
+#include "rgw_civetweb_log.h"
#include "civetweb/civetweb.h"
@@ -93,6 +94,8 @@
RGWRequest() : id(0), s(NULL), op(NULL) {
}
+ virtual ~RGWRequest() {}
+
void init_state(req_state *_s) {
s = _s;
}
@@ -141,6 +144,8 @@
bool get_val(const string& key, const string& def_val, string *out);
bool get_val(const string& key, int def_val, int *out);
+ map<string, string>& get_config_map() { return config_map; }
+
string get_framework() { return framework; }
};
@@ -636,6 +641,10 @@
FCGX_Finish_r(fcgx);
+ if (store->ctx()->_conf->rgw_fcgi_explicit_free) {
+ FCGX_Free(fcgx, 1);
+ }
+
delete req;
}
@@ -909,6 +918,12 @@
struct mg_context *ctx;
RGWProcessEnv env;
+ void set_conf_default(map<string, string>& m, const string& key, const string& def_val) {
+ if (m.find(key) == m.end()) {
+ m[key] = def_val;
+ }
+ }
+
public:
RGWMongooseFrontend(RGWProcessEnv& pe, RGWFrontendConfig *_conf) : conf(_conf), ctx(NULL), env(pe) {
}
@@ -921,12 +936,28 @@
char thread_pool_buf[32];
snprintf(thread_pool_buf, sizeof(thread_pool_buf), "%d", (int)g_conf->rgw_thread_pool_size);
string port_str;
+ map<string, string> conf_map = conf->get_config_map();
conf->get_val("port", "80", &port_str);
- const char *options[] = {"listening_ports", port_str.c_str(), "enable_keep_alive", "yes", "num_threads", thread_pool_buf, NULL};
+ conf_map.erase("port");
+ conf_map["listening_ports"] = port_str;
+ set_conf_default(conf_map, "enable_keep_alive", "yes");
+ set_conf_default(conf_map, "num_threads", thread_pool_buf);
+ set_conf_default(conf_map, "decode_url", "no");
+
+ const char *options[conf_map.size() * 2 + 1];
+ int i = 0;
+ for (map<string, string>::iterator iter = conf_map.begin(); iter != conf_map.end(); ++iter) {
+ options[i] = iter->first.c_str();
+ options[i + 1] = iter->second.c_str();
+ dout(20)<< "civetweb config: " << options[i] << ": " << (options[i + 1] ? options[i + 1] : "<null>") << dendl;
+ i += 2;
+ }
+ options[i] = NULL;
struct mg_callbacks cb;
memset((void *)&cb, 0, sizeof(cb));
cb.begin_request = civetweb_callback;
+ cb.log_message = rgw_civetweb_log_callback;
ctx = mg_start(&cb, &env, (const char **)&options);
if (!ctx) {
@@ -965,7 +996,7 @@
vector<const char *> def_args;
def_args.push_back("--debug-rgw=1/5");
def_args.push_back("--keyring=$rgw_data/keyring");
- def_args.push_back("--log-file=/var/log/radosgw/$cluster-$name");
+ def_args.push_back("--log-file=/var/log/radosgw/$cluster-$name.log");
vector<const char*> args;
argv_to_vec(argc, argv, args);
diff -Nru ceph-0.80.7/src/rgw/rgw_op.cc ceph-0.80.9/src/rgw/rgw_op.cc
--- ceph-0.80.7/src/rgw/rgw_op.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_op.cc 2015-03-09 20:51:18.000000000 +0100
@@ -33,7 +33,8 @@
static string mp_ns = RGW_OBJ_NS_MULTIPART;
static string shadow_ns = RGW_OBJ_NS_SHADOW;
-#define MULTIPART_UPLOAD_ID_PREFIX "2/" // must contain a unique char that may not come up in gen_rand_alpha()
+#define MULTIPART_UPLOAD_ID_PREFIX_LEGACY "2/"
+#define MULTIPART_UPLOAD_ID_PREFIX "2~" // must contain a unique char that may not come up in gen_rand_alpha()
class MultipartMetaFilter : public RGWAccessListFilter {
public:
@@ -1438,7 +1439,8 @@
{
const char *uid = upload_id.c_str();
- return (strncmp(uid, MULTIPART_UPLOAD_ID_PREFIX, sizeof(MULTIPART_UPLOAD_ID_PREFIX) - 1) == 0);
+ return (strncmp(uid, MULTIPART_UPLOAD_ID_PREFIX, sizeof(MULTIPART_UPLOAD_ID_PREFIX) - 1) == 0) ||
+ (strncmp(uid, MULTIPART_UPLOAD_ID_PREFIX_LEGACY, sizeof(MULTIPART_UPLOAD_ID_PREFIX_LEGACY) - 1) == 0);
}
int RGWPutObjProcessor_Multipart::do_complete(string& etag, time_t *mtime, time_t set_mtime, map<string, bufferlist>& attrs)
@@ -1524,64 +1526,18 @@
rgw_bucket_object_pre_exec(s);
}
-static int put_obj_user_manifest_iterate_cb(rgw_bucket& bucket, RGWObjEnt& ent, RGWAccessControlPolicy *bucket_policy, off_t start_ofs, off_t end_ofs,
- void *param)
-{
- RGWPutObj *op = (RGWPutObj *)param;
- return op->user_manifest_iterate_cb(bucket, ent, bucket_policy, start_ofs, end_ofs);
-}
-
-int RGWPutObj::user_manifest_iterate_cb(rgw_bucket& bucket, RGWObjEnt& ent, RGWAccessControlPolicy *bucket_policy, off_t start_ofs, off_t end_ofs)
-{
- rgw_obj part(bucket, ent.name);
-
- map<string, bufferlist> attrs;
-
- int ret = get_obj_attrs(store, s, part, attrs, NULL, NULL);
- if (ret < 0) {
- return ret;
- }
- map<string, bufferlist>::iterator iter = attrs.find(RGW_ATTR_ETAG);
- if (iter == attrs.end()) {
- return 0;
- }
- bufferlist& bl = iter->second;
- const char *buf = bl.c_str();
- int len = bl.length();
- while (len > 0 && buf[len - 1] == '\0') {
- len--;
- }
- if (len > 0) {
- user_manifest_parts_hash->Update((const byte *)bl.c_str(), len);
- }
-
- if (s->cct->_conf->subsys.should_gather(ceph_subsys_rgw, 20)) {
- string e(bl.c_str(), bl.length());
- ldout(s->cct, 20) << __func__ << ": appending user manifest etag: " << e << dendl;
- }
-
- return 0;
-}
-
static int put_data_and_throttle(RGWPutObjProcessor *processor, bufferlist& data, off_t ofs,
MD5 *hash, bool need_to_wait)
{
- const unsigned char *data_ptr = (hash ? (const unsigned char *)data.c_str() : NULL);
bool again;
- uint64_t len = data.length();
do {
void *handle;
- int ret = processor->handle_data(data, ofs, &handle, &again);
+ int ret = processor->handle_data(data, ofs, hash, &handle, &again);
if (ret < 0)
return ret;
- if (hash) {
- hash->Update(data_ptr, len);
- hash = NULL; /* only calculate hash once */
- }
-
ret = processor->throttle_data(handle, need_to_wait);
if (ret < 0)
return ret;
@@ -1719,6 +1675,7 @@
}
if (need_calc_md5) {
+ processor->complete_hash(&hash);
hash.Final(m);
buf_to_hex(m, CEPH_CRYPTO_MD5_DIGESTSIZE, calc_md5);
@@ -1737,7 +1694,6 @@
bufferlist manifest_bl;
string manifest_obj_prefix;
string manifest_bucket;
- RGWBucketInfo bucket_info;
char etag_buf[CEPH_CRYPTO_MD5_DIGESTSIZE];
char etag_buf_str[CEPH_CRYPTO_MD5_DIGESTSIZE * 2 + 16];
@@ -1755,16 +1711,6 @@
manifest_bucket = prefix_str.substr(0, pos);
manifest_obj_prefix = prefix_str.substr(pos + 1);
- ret = store->get_bucket_info(NULL, manifest_bucket, bucket_info, NULL, NULL);
- if (ret < 0) {
- ldout(s->cct, 0) << "could not get bucket info for bucket=" << manifest_bucket << dendl;
- }
- ret = iterate_user_manifest_parts(s->cct, store, 0, -1, bucket_info.bucket, manifest_obj_prefix,
- NULL, NULL, put_obj_user_manifest_iterate_cb, (void *)this);
- if (ret < 0) {
- goto done;
- }
-
hash.Final((byte *)etag_buf);
buf_to_hex((const unsigned char *)etag_buf, CEPH_CRYPTO_MD5_DIGESTSIZE, etag_buf_str);
@@ -1940,10 +1886,14 @@
/* no need to track object versioning, need it for bucket's data only */
RGWObjVersionTracker *ptracker = (s->object ? NULL : &s->bucket_info.objv_tracker);
- /* check if obj exists, read orig attrs */
- ret = get_obj_attrs(store, s, obj, orig_attrs, NULL, ptracker);
- if (ret < 0)
- return;
+ if (s->object) {
+ /* check if obj exists, read orig attrs */
+ ret = get_obj_attrs(store, s, obj, orig_attrs, NULL, ptracker);
+ if (ret < 0)
+ return;
+ } else {
+ orig_attrs = s->bucket_attrs;
+ }
/* only remove meta attrs */
for (iter = orig_attrs.begin(); iter != orig_attrs.end(); ++iter) {
@@ -2214,6 +2164,7 @@
replace_attrs,
attrs, RGW_OBJ_CATEGORY_MAIN,
&s->req_id, /* use req_id as tag */
+ &etag,
&s->err,
copy_obj_progress_cb, (void *)this
);
@@ -2277,7 +2228,6 @@
RGWAccessControlPolicy_S3 new_policy(s->cct);
stringstream ss;
char *new_data = NULL;
- ACLOwner owner;
rgw_obj obj;
ret = 0;
@@ -2287,8 +2237,10 @@
return;
}
- owner.set_id(s->user.user_id);
- owner.set_name(s->user.display_name);
+
+ RGWAccessControlPolicy *existing_policy = (s->object == NULL? s->bucket_acl : s->object_acl);
+
+ owner = existing_policy->get_owner();
ret = get_params();
if (ret < 0)
@@ -2536,7 +2488,7 @@
do {
char buf[33];
gen_rand_alphanumeric(s->cct, buf, sizeof(buf) - 1);
- upload_id = "2/"; /* v2 upload id */
+ upload_id = MULTIPART_UPLOAD_ID_PREFIX; /* v2 upload id */
upload_id.append(buf);
string tmp_obj_name;
diff -Nru ceph-0.80.7/src/rgw/rgw_op.h ceph-0.80.9/src/rgw/rgw_op.h
--- ceph-0.80.7/src/rgw/rgw_op.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_op.h 2015-03-09 20:51:18.000000000 +0100
@@ -345,8 +345,6 @@
RGWPutObjProcessor *select_processor(bool *is_multipart);
void dispose_processor(RGWPutObjProcessor *processor);
- int user_manifest_iterate_cb(rgw_bucket& bucket, RGWObjEnt& ent, RGWAccessControlPolicy *bucket_policy, off_t start_ofs, off_t end_ofs);
-
int verify_permission();
void pre_exec();
void execute();
@@ -490,6 +488,7 @@
string source_zone;
string client_id;
string op_id;
+ string etag;
off_t last_ofs;
@@ -556,6 +555,7 @@
int ret;
size_t len;
char *data;
+ ACLOwner owner;
public:
RGWPutACLs() {
diff -Nru ceph-0.80.7/src/rgw/rgw_rados.cc ceph-0.80.9/src/rgw/rgw_rados.cc
--- ceph-0.80.7/src/rgw/rgw_rados.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_rados.cc 2015-03-09 20:51:18.000000000 +0100
@@ -876,6 +876,11 @@
return 0;
}
+CephContext *RGWPutObjProcessor::ctx()
+{
+ return store->ctx();
+}
+
RGWPutObjProcessor::~RGWPutObjProcessor()
{
if (is_complete)
@@ -900,8 +905,10 @@
return 0;
};
-int RGWPutObjProcessor_Plain::handle_data(bufferlist& bl, off_t _ofs, void **phandle, bool *again)
+int RGWPutObjProcessor_Plain::handle_data(bufferlist& bl, off_t _ofs, MD5 *hash, void **phandle, bool *again)
{
+ assert(!hash);
+
*again = false;
if (ofs != _ofs)
@@ -1028,7 +1035,7 @@
return RGWPutObjProcessor_Aio::handle_obj_data(cur_obj, bl, ofs - cur_part_ofs, ofs, phandle, exclusive);
}
-int RGWPutObjProcessor_Atomic::handle_data(bufferlist& bl, off_t ofs, void **phandle, bool *again)
+int RGWPutObjProcessor_Atomic::handle_data(bufferlist& bl, off_t ofs, MD5 *hash, void **phandle, bool *again)
{
*again = false;
@@ -1062,7 +1069,10 @@
if (!data_ofs && !immutable_head()) {
first_chunk.claim(bl);
obj_len = (uint64_t)first_chunk.length();
- int r = prepare_next_part(first_chunk.length());
+ if (hash) {
+ hash->Update((const byte *)first_chunk.c_str(), obj_len);
+ }
+ int r = prepare_next_part(obj_len);
if (r < 0) {
return r;
}
@@ -1074,7 +1084,19 @@
bool exclusive = (!write_ofs && immutable_head()); /* immutable head object, need to verify nothing exists there
we could be racing with another upload, to the same
object and cleanup can be messy */
- return write_data(bl, write_ofs, phandle, exclusive);
+ int ret = write_data(bl, write_ofs, phandle, exclusive);
+ if (ret >= 0) { /* we might return, need to clear bl as it was already sent */
+ if (hash) {
+ hash->Update((const byte *)bl.c_str(), bl.length());
+ }
+ bl.clear();
+ }
+ return ret;
+}
+
+void RGWPutObjProcessor_Atomic::complete_hash(MD5 *hash)
+{
+ hash->Update((const byte *)pending_data_bl.c_str(), pending_data_bl.length());
}
@@ -3019,7 +3041,7 @@
do {
void *handle;
- int ret = processor->handle_data(bl, ofs, &handle, &again);
+ int ret = processor->handle_data(bl, ofs, NULL, &handle, &again);
if (ret < 0)
return ret;
@@ -3029,6 +3051,11 @@
*/
ret = opstate->renew_state();
if (ret < 0) {
+ ldout(processor->ctx(), 0) << "ERROR: RGWRadosPutObj::handle_data(): failed to renew op state ret=" << ret << dendl;
+ int r = processor->throttle_data(handle, false);
+ if (r < 0) {
+ ldout(processor->ctx(), 0) << "ERROR: RGWRadosPutObj::handle_data(): processor->throttle_data() returned " << r << dendl;
+ }
/* could not renew state! might have been marked as cancelled */
return ret;
}
@@ -3114,6 +3141,7 @@
map<string, bufferlist>& attrs,
RGWObjCategory category,
string *ptag,
+ string *petag,
struct rgw_err *err,
void (*progress_cb)(off_t, void *),
void *progress_data)
@@ -3210,6 +3238,10 @@
if (ret < 0)
goto set_err_state;
+ if (petag) {
+ *petag = etag;
+ }
+
{ /* opening scope so that we can do goto, sorry */
bufferlist& extra_data_bl = processor.get_extra_data();
if (extra_data_bl.length()) {
@@ -3275,6 +3307,10 @@
if (ret < 0)
return ret;
+ if (petag) {
+ *petag = etag;
+ }
+
return 0;
}
@@ -3286,7 +3322,7 @@
return ret;
}
- bool copy_data = !astate->has_manifest;
+ bool copy_data = !astate->has_manifest || (src_obj.bucket.data_pool != dest_obj.bucket.data_pool);
bool copy_first = false;
if (astate->has_manifest) {
if (!astate->manifest.has_tail()) {
@@ -3304,7 +3340,7 @@
}
if (copy_data) { /* refcounting tail wouldn't work here, just copy the data */
- return copy_obj_data(ctx, dest_bucket_info.owner, &handle, end, dest_obj, src_obj, max_chunk_size, mtime, src_attrs, category, ptag, err);
+ return copy_obj_data(ctx, dest_bucket_info.owner, &handle, end, dest_obj, src_obj, max_chunk_size, mtime, src_attrs, category, ptag, petag, err);
}
RGWObjManifest::obj_iterator miter = astate->manifest.obj_begin();
@@ -3383,6 +3419,14 @@
if (mtime)
obj_stat(ctx, dest_obj, NULL, mtime, NULL, NULL, NULL, NULL);
+ if (petag) {
+ map<string, bufferlist>::iterator iter = src_attrs.find(RGW_ATTR_ETAG);
+ if (iter != src_attrs.end()) {
+ bufferlist& etagbl = iter->second;
+ *petag = string(etagbl.c_str(), etagbl.length());
+ }
+ }
+
return 0;
done_ret:
@@ -3419,23 +3463,23 @@
map<string, bufferlist>& attrs,
RGWObjCategory category,
string *ptag,
+ string *petag,
struct rgw_err *err)
{
bufferlist first_chunk;
RGWObjManifest manifest;
map<uint64_t, RGWObjManifestPart> objs;
- RGWObjManifestPart *first_part;
- map<string, bufferlist>::iterator iter;
- rgw_obj shadow_obj = dest_obj;
- string shadow_oid;
+ string tag;
+ append_rand_alpha(cct, tag, tag, 32);
- append_rand_alpha(cct, dest_obj.object, shadow_oid, 32);
- shadow_obj.init_ns(dest_obj.bucket, shadow_oid, shadow_ns);
+ RGWPutObjProcessor_Atomic processor(owner, dest_obj.bucket, dest_obj.object,
+ cct->_conf->rgw_obj_stripe_size, tag);
+ int ret = processor.prepare(this, ctx, NULL);
+ if (ret < 0)
+ return ret;
- int ret, r;
off_t ofs = 0;
- PutObjMetaExtraParams ep;
do {
bufferlist bl;
@@ -3443,55 +3487,40 @@
if (ret < 0)
return ret;
- const char *data = bl.c_str();
+ uint64_t read_len = ret;
+ bool again;
- if ((uint64_t)ofs < max_chunk_size) {
- uint64_t len = min(max_chunk_size - ofs, (uint64_t)ret);
- first_chunk.append(data, len);
- ofs += len;
- ret -= len;
- data += len;
- }
+ do {
+ void *handle;
- // In the first call to put_obj_data, we pass ofs == -1 so that it will do
- // a write_full, wiping out whatever was in the object before this
- r = 0;
- if (ret > 0) {
- r = put_obj_data(ctx, shadow_obj, data, ((ofs == 0) ? -1 : ofs), ret, false);
- }
- if (r < 0)
- goto done_err;
+ ret = processor.handle_data(bl, ofs, NULL, &handle, &again);
+ if (ret < 0) {
+ return ret;
+ }
+ ret = processor.throttle_data(handle, false);
+ if (ret < 0)
+ return ret;
+ } while (again);
- ofs += ret;
+ ofs += read_len;
} while (ofs <= end);
- first_part = &objs[0];
- first_part->loc = dest_obj;
- first_part->loc_ofs = 0;
- first_part->size = first_chunk.length();
-
- if ((uint64_t)ofs > max_chunk_size) {
- RGWObjManifestPart& tail = objs[max_chunk_size];
- tail.loc = shadow_obj;
- tail.loc_ofs = max_chunk_size;
- tail.size = ofs - max_chunk_size;
+ string etag;
+ map<string, bufferlist>::iterator iter = attrs.find(RGW_ATTR_ETAG);
+ if (iter != attrs.end()) {
+ bufferlist& bl = iter->second;
+ etag = string(bl.c_str(), bl.length());
+ if (petag) {
+ *petag = etag;
+ }
}
- manifest.set_explicit(ofs, objs);
-
- ep.data = &first_chunk;
- ep.manifest = &manifest;
- ep.ptag = ptag;
- ep.owner = owner;
+ ret = processor.complete(etag, NULL, 0, attrs);
- ret = put_obj_meta(ctx, dest_obj, end + 1, attrs, category, PUT_OBJ_CREATE, ep);
if (mtime)
obj_stat(ctx, dest_obj, NULL, mtime, NULL, NULL, NULL, NULL);
return ret;
-done_err:
- delete_obj(ctx, owner, shadow_obj);
- return r;
}
/**
@@ -4132,7 +4161,33 @@
if (!op.size())
return 0;
+ string tag;
+ if (state) {
+ r = prepare_update_index(state, bucket, CLS_RGW_OP_ADD, obj, tag);
+ if (r < 0)
+ return r;
+ }
+
r = ref.ioctx.operate(ref.oid, &op);
+ if (state) {
+ if (r >= 0) {
+ bufferlist acl_bl = attrs[RGW_ATTR_ACL];
+ bufferlist etag_bl = attrs[RGW_ATTR_ETAG];
+ bufferlist content_type_bl = attrs[RGW_ATTR_CONTENT_TYPE];
+ string etag(etag_bl.c_str(), etag_bl.length());
+ string content_type(content_type_bl.c_str(), content_type_bl.length());
+ uint64_t epoch = ref.ioctx.get_last_version();
+ int64_t poolid = ref.ioctx.get_id();
+ utime_t mtime = ceph_clock_now(cct);
+ r = complete_update_index(bucket, obj.object, tag, poolid, epoch, state->size,
+ mtime, etag, content_type, &acl_bl, RGW_OBJ_CATEGORY_MAIN, NULL);
+ } else {
+ int ret = complete_update_index_cancel(bucket, obj.object, tag);
+ if (ret < 0) {
+ ldout(cct, 0) << "ERROR: comlete_update_index_cancel() returned r=" << r << dendl;
+ }
+ }
+ }
if (r < 0)
return r;
@@ -4636,7 +4691,7 @@
bl.append(read_bl);
done:
- if (bl.length() > 0) {
+ if (r >= 0) {
r = bl.length();
}
if (r < 0 || !len || ((off_t)(ofs + len - 1) == end)) {
diff -Nru ceph-0.80.7/src/rgw/rgw_rados.h ceph-0.80.9/src/rgw/rgw_rados.h
--- ceph-0.80.7/src/rgw/rgw_rados.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_rados.h 2015-03-09 20:51:18.000000000 +0100
@@ -298,6 +298,11 @@
bool has_tail() {
if (explicit_objs) {
+ if (objs.size() == 1) {
+ map<uint64_t, RGWObjManifestPart>::iterator iter = objs.begin();
+ rgw_obj& obj = iter->second.loc;
+ return head_obj.object != obj.object;
+ }
return (objs.size() >= 2);
}
return (obj_size > head_size);
@@ -548,9 +553,14 @@
obj_ctx = _o;
return 0;
};
- virtual int handle_data(bufferlist& bl, off_t ofs, void **phandle, bool *again) = 0;
+ virtual int handle_data(bufferlist& bl, off_t ofs, MD5 *hash, void **phandle, bool *again) = 0;
virtual int throttle_data(void *handle, bool need_to_wait) = 0;
+ virtual void complete_hash(MD5 *hash) {
+ assert(0);
+ }
virtual int complete(string& etag, time_t *mtime, time_t set_mtime, map<string, bufferlist>& attrs);
+
+ CephContext *ctx();
};
class RGWPutObjProcessor_Plain : public RGWPutObjProcessor
@@ -564,7 +574,7 @@
protected:
int prepare(RGWRados *store, void *obj_ctx, string *oid_rand);
- int handle_data(bufferlist& bl, off_t ofs, void **phandle, bool *again);
+ int handle_data(bufferlist& bl, off_t ofs, MD5 *hash /* NULL expected */, void **phandle, bool *again);
int do_complete(string& etag, time_t *mtime, time_t set_mtime, map<string, bufferlist>& attrs);
public:
@@ -654,7 +664,8 @@
void set_extra_data_len(uint64_t len) {
extra_data_len = len;
}
- virtual int handle_data(bufferlist& bl, off_t ofs, void **phandle, bool *again);
+ virtual int handle_data(bufferlist& bl, off_t ofs, MD5 *hash, void **phandle, bool *again);
+ virtual void complete_hash(MD5 *hash);
bufferlist& get_extra_data() { return extra_data_bl; }
};
@@ -1555,6 +1566,7 @@
map<std::string, bufferlist>& attrs,
RGWObjCategory category,
string *ptag,
+ string *petag,
struct rgw_err *err,
void (*progress_cb)(off_t, void *),
void *progress_data);
@@ -1569,6 +1581,7 @@
map<string, bufferlist>& attrs,
RGWObjCategory category,
string *ptag,
+ string *petag,
struct rgw_err *err);
/**
* Delete a bucket.
diff -Nru ceph-0.80.7/src/rgw/rgw_rest.cc ceph-0.80.9/src/rgw/rgw_rest.cc
--- ceph-0.80.7/src/rgw/rgw_rest.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_rest.cc 2015-03-09 20:51:18.000000000 +0100
@@ -244,7 +244,7 @@
if (r < 0) {
ldout(s->cct, 0) << "ERROR: s->cio->print() returned err=" << r << dendl;
}
- r = s->cio->print("Accept-Ranges: %s\n", "bytes");
+ r = s->cio->print("Accept-Ranges: %s\r\n", "bytes");
if (r < 0) {
ldout(s->cct, 0) << "ERROR: s->cio->print() returned err=" << r << dendl;
}
@@ -254,9 +254,9 @@
{
int r;
if (s->prot_flags & RGW_REST_SWIFT)
- r = s->cio->print("etag: %s\n", etag);
+ r = s->cio->print("etag: %s\r\n", etag);
else
- r = s->cio->print("ETag: \"%s\"\n", etag);
+ r = s->cio->print("ETag: \"%s\"\r\n", etag);
if (r < 0) {
ldout(s->cct, 0) << "ERROR: s->cio->print() returned err=" << r << dendl;
}
@@ -265,7 +265,7 @@
void dump_pair(struct req_state *s, const char *key, const char *value)
{
if ( (strlen(key) > 0) && (strlen(value) > 0))
- s->cio->print("%s: %s\n", key, value);
+ s->cio->print("%s: %s\r\n", key, value);
}
void dump_bucket_from_state(struct req_state *s)
@@ -273,16 +273,10 @@
int expose_bucket = g_conf->rgw_expose_bucket;
if (expose_bucket) {
if (!s->bucket_name_str.empty())
- s->cio->print("Bucket: \"%s\"\n", s->bucket_name_str.c_str());
+ s->cio->print("Bucket: \"%s\"\r\n", s->bucket_name_str.c_str());
}
}
-void dump_object_from_state(struct req_state *s)
-{
- if (!s->object_str.empty())
- s->cio->print("Key: \"%s\"\n", s->object_str.c_str());
-}
-
void dump_uri_from_state(struct req_state *s)
{
if (strcmp(s->info.request_uri.c_str(), "/") == 0) {
@@ -296,12 +290,12 @@
location += "/";
if (!s->object_str.empty()) {
location += s->object_str;
- s->cio->print("Location: %s\n", location.c_str());
+ s->cio->print("Location: %s\r\n", location.c_str());
}
}
}
else {
- s->cio->print("Location: \"%s\"\n", s->info.request_uri.c_str());
+ s->cio->print("Location: \"%s\"\r\n", s->info.request_uri.c_str());
}
}
@@ -310,7 +304,7 @@
if (redirect.empty())
return;
- s->cio->print("Location: %s\n", redirect.c_str());
+ s->cio->print("Location: %s\r\n", redirect.c_str());
}
static void dump_time_header(struct req_state *s, const char *name, time_t t)
@@ -325,7 +319,7 @@
if (strftime(timestr, sizeof(timestr), "%a, %d %b %Y %H:%M:%S %Z", tmp) == 0)
return;
- int r = s->cio->print("%s: %s\n", name, timestr);
+ int r = s->cio->print("%s: %s\r\n", name, timestr);
if (r < 0) {
ldout(s->cct, 0) << "ERROR: s->cio->print() returned err=" << r << dendl;
}
@@ -341,7 +335,7 @@
char buf[32];
snprintf(buf, sizeof(buf), "%lld", (long long)t);
- int r = s->cio->print("%s: %s\n", name, buf);
+ int r = s->cio->print("%s: %s\r\n", name, buf);
if (r < 0) {
ldout(s->cct, 0) << "ERROR: s->cio->print() returned err=" << r << dendl;
}
@@ -374,16 +368,16 @@
void dump_access_control(struct req_state *s, const char *origin, const char *meth,
const char *hdr, const char *exp_hdr, uint32_t max_age) {
if (origin && (origin[0] != '\0')) {
- s->cio->print("Access-Control-Allow-Origin: %s\n", origin);
+ s->cio->print("Access-Control-Allow-Origin: %s\r\n", origin);
if (meth && (meth[0] != '\0'))
- s->cio->print("Access-Control-Allow-Methods: %s\n", meth);
+ s->cio->print("Access-Control-Allow-Methods: %s\r\n", meth);
if (hdr && (hdr[0] != '\0'))
- s->cio->print("Access-Control-Allow-Headers: %s\n", hdr);
+ s->cio->print("Access-Control-Allow-Headers: %s\r\n", hdr);
if (exp_hdr && (exp_hdr[0] != '\0')) {
- s->cio->print("Access-Control-Expose-Headers: %s\n", exp_hdr);
+ s->cio->print("Access-Control-Expose-Headers: %s\r\n", exp_hdr);
}
if (max_age != CORS_MAX_AGE_INVALID) {
- s->cio->print("Access-Control-Max-Age: %d\n", max_age);
+ s->cio->print("Access-Control-Max-Age: %d\r\n", max_age);
}
}
}
@@ -483,7 +477,7 @@
/* dumping range into temp buffer first, as libfcgi will fail to digest %lld */
snprintf(range_buf, sizeof(range_buf), "%lld-%lld/%lld", (long long)ofs, (long long)end, (long long)total);
- int r = s->cio->print("Content-Range: bytes %s\n", range_buf);
+ int r = s->cio->print("Content-Range: bytes %s\r\n", range_buf);
if (r < 0) {
ldout(s->cct, 0) << "ERROR: s->cio->print() returned err=" << r << dendl;
}
diff -Nru ceph-0.80.7/src/rgw/rgw_rest.h ceph-0.80.9/src/rgw/rgw_rest.h
--- ceph-0.80.7/src/rgw/rgw_rest.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_rest.h 2015-03-09 20:51:18.000000000 +0100
@@ -369,7 +369,6 @@
extern void list_all_buckets_end(struct req_state *s);
extern void dump_time(struct req_state *s, const char *name, time_t *t);
extern void dump_bucket_from_state(struct req_state *s);
-extern void dump_object_from_state(struct req_state *s);
extern void dump_uri_from_state(struct req_state *s);
extern void dump_redirect(struct req_state *s, const string& redirect);
extern void dump_pair(struct req_state *s, const char *key, const char *value);
diff -Nru ceph-0.80.7/src/rgw/rgw_rest_s3.cc ceph-0.80.9/src/rgw/rgw_rest_s3.cc
--- ceph-0.80.7/src/rgw/rgw_rest_s3.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_rest_s3.cc 2015-03-09 20:51:18.000000000 +0100
@@ -161,7 +161,7 @@
dump_errno(s);
for (riter = response_attrs.begin(); riter != response_attrs.end(); ++riter) {
- s->cio->print("%s: %s\n", riter->first.c_str(), riter->second.c_str());
+ s->cio->print("%s: %s\r\n", riter->first.c_str(), riter->second.c_str());
}
if (!content_type)
@@ -303,9 +303,9 @@
{
char buf[32];
snprintf(buf, sizeof(buf), "%lld", (long long)bucket.count);
- s->cio->print("X-RGW-Object-Count: %s\n", buf);
+ s->cio->print("X-RGW-Object-Count: %s\r\n", buf);
snprintf(buf, sizeof(buf), "%lld", (long long)bucket.size);
- s->cio->print("X-RGW-Bytes-Used: %s\n", buf);
+ s->cio->print("X-RGW-Bytes-Used: %s\r\n", buf);
}
void RGWStatBucket_ObjStore_S3::send_response()
@@ -321,16 +321,16 @@
dump_start(s);
}
-static int create_s3_policy(struct req_state *s, RGWRados *store, RGWAccessControlPolicy_S3& s3policy)
+static int create_s3_policy(struct req_state *s, RGWRados *store, RGWAccessControlPolicy_S3& s3policy, ACLOwner& owner)
{
if (s->has_acl_header) {
if (!s->canned_acl.empty())
return -ERR_INVALID_REQUEST;
- return s3policy.create_from_headers(store, s->info.env, s->owner);
+ return s3policy.create_from_headers(store, s->info.env, owner);
}
- return s3policy.create_canned(s->owner, s->bucket_owner, s->canned_acl);
+ return s3policy.create_canned(owner, s->bucket_owner, s->canned_acl);
}
class RGWLocationConstraint : public XMLObj
@@ -386,7 +386,7 @@
{
RGWAccessControlPolicy_S3 s3policy(s->cct);
- int r = create_s3_policy(s, store, s3policy);
+ int r = create_s3_policy(s, store, s3policy, s->owner);
if (r < 0)
return r;
@@ -487,7 +487,7 @@
if (!s->length)
return -ERR_LENGTH_REQUIRED;
- int r = create_s3_policy(s, store, s3policy);
+ int r = create_s3_policy(s, store, s3policy, s->owner);
if (r < 0)
return r;
@@ -1198,7 +1198,7 @@
RGWAccessControlPolicy_S3 s3policy(s->cct);
/* build a policy for the target object */
- int r = create_s3_policy(s, store, s3policy);
+ int r = create_s3_policy(s, store, s3policy, s->owner);
if (r < 0)
return r;
@@ -1264,7 +1264,7 @@
set_req_state_err(s, ret);
dump_errno(s);
- end_header(s, this, "binary/octet-stream");
+ end_header(s, this, "application/xml");
if (ret == 0) {
s->formatter->open_object_section("CopyObjectResult");
}
@@ -1285,13 +1285,8 @@
if (ret == 0) {
dump_time(s, "LastModified", &mtime);
- map<string, bufferlist>::iterator iter = attrs.find(RGW_ATTR_ETAG);
- if (iter != attrs.end()) {
- bufferlist& bl = iter->second;
- if (bl.length()) {
- char *etag = bl.c_str();
- s->formatter->dump_string("ETag", etag);
- }
+ if (!etag.empty()) {
+ s->formatter->dump_string("ETag", etag);
}
s->formatter->close_section();
rgw_flush_formatter_and_reset(s, s->formatter);
@@ -1318,7 +1313,7 @@
s->canned_acl.clear();
}
- int r = create_s3_policy(s, store, s3policy);
+ int r = create_s3_policy(s, store, s3policy, owner);
if (r < 0)
return r;
@@ -1460,7 +1455,7 @@
int RGWInitMultipart_ObjStore_S3::get_params()
{
RGWAccessControlPolicy_S3 s3policy(s->cct);
- ret = create_s3_policy(s, store, s3policy);
+ ret = create_s3_policy(s, store, s3policy, s->owner);
if (ret < 0)
return ret;
@@ -2047,6 +2042,12 @@
return 0;
}
+static void init_anon_user(struct req_state *s)
+{
+ rgw_get_anon_user(s->user);
+ s->perm_mask = RGW_PERM_FULL_CONTROL;
+}
+
/*
* verify that a signed request comes from the keyholder
* by checking the signature against our locally-computed version
@@ -2067,6 +2068,11 @@
return -EPERM;
}
+ if (s->op == OP_OPTIONS) {
+ init_anon_user(s);
+ return 0;
+ }
+
if (!s->http_auth || !(*s->http_auth)) {
auth_id = s->info.args.get("AWSAccessKeyId");
if (auth_id.size()) {
@@ -2080,8 +2086,7 @@
qsr = true;
} else {
/* anonymous access */
- rgw_get_anon_user(s->user);
- s->perm_mask = RGW_PERM_FULL_CONTROL;
+ init_anon_user(s);
return 0;
}
} else {
diff -Nru ceph-0.80.7/src/rgw/rgw_rest_swift.cc ceph-0.80.9/src/rgw/rgw_rest_swift.cc
--- ceph-0.80.7/src/rgw/rgw_rest_swift.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_rest_swift.cc 2015-03-09 20:51:18.000000000 +0100
@@ -223,11 +223,11 @@
{
char buf[32];
snprintf(buf, sizeof(buf), "%lld", (long long)bucket.count);
- s->cio->print("X-Container-Object-Count: %s\n", buf);
+ s->cio->print("X-Container-Object-Count: %s\r\n", buf);
snprintf(buf, sizeof(buf), "%lld", (long long)bucket.size);
- s->cio->print("X-Container-Bytes-Used: %s\n", buf);
+ s->cio->print("X-Container-Bytes-Used: %s\r\n", buf);
snprintf(buf, sizeof(buf), "%lld", (long long)bucket.size_rounded);
- s->cio->print("X-Container-Bytes-Used-Actual: %s\n", buf);
+ s->cio->print("X-Container-Bytes-Used-Actual: %s\r\n", buf);
if (!s->object) {
RGWAccessControlPolicy_SWIFT *swift_policy = static_cast<RGWAccessControlPolicy_SWIFT *>(s->bucket_acl);
@@ -247,13 +247,13 @@
{
char buf[32];
snprintf(buf, sizeof(buf), "%lld", (long long)buckets_count);
- s->cio->print("X-Account-Container-Count: %s\n", buf);
+ s->cio->print("X-Account-Container-Count: %s\r\n", buf);
snprintf(buf, sizeof(buf), "%lld", (long long)buckets_object_count);
- s->cio->print("X-Account-Object-Count: %s\n", buf);
+ s->cio->print("X-Account-Object-Count: %s\r\n", buf);
snprintf(buf, sizeof(buf), "%lld", (long long)buckets_size);
- s->cio->print("X-Account-Bytes-Used: %s\n", buf);
+ s->cio->print("X-Account-Bytes-Used: %s\r\n", buf);
snprintf(buf, sizeof(buf), "%lld", (long long)buckets_size_rounded);
- s->cio->print("X-Account-Bytes-Used-Actual: %s\n", buf);
+ s->cio->print("X-Account-Bytes-Used-Actual: %s\r\n", buf);
}
void RGWStatAccount_ObjStore_SWIFT::send_response()
@@ -552,7 +552,6 @@
int RGWGetObj_ObjStore_SWIFT::send_response_data(bufferlist& bl, off_t bl_ofs, off_t bl_len)
{
const char *content_type = NULL;
- int orig_ret = ret;
map<string, string> response_attrs;
map<string, string>::iterator riter;
@@ -593,15 +592,11 @@
}
}
- if (partial_content && !ret)
- ret = -STATUS_PARTIAL_CONTENT;
-
- if (ret)
- set_req_state_err(s, ret);
+ set_req_state_err(s, (partial_content && !ret) ? STATUS_PARTIAL_CONTENT : ret);
dump_errno(s);
for (riter = response_attrs.begin(); riter != response_attrs.end(); ++riter) {
- s->cio->print("%s: %s\n", riter->first.c_str(), riter->second.c_str());
+ s->cio->print("%s: %s\r\n", riter->first.c_str(), riter->second.c_str());
}
if (!content_type)
@@ -611,7 +606,7 @@
sent_header = true;
send_data:
- if (get_data && !orig_ret) {
+ if (get_data && !ret) {
int r = s->cio->write(bl.c_str() + bl_ofs, bl_len);
if (r < 0)
return r;
@@ -776,7 +771,8 @@
int RGWHandler_ObjStore_SWIFT::authorize()
{
- if (!s->os_auth_token && s->info.args.get("temp_url_sig").empty()) {
+ if ((!s->os_auth_token && s->info.args.get("temp_url_sig").empty()) ||
+ (s->op == OP_OPTIONS)) {
/* anonymous access */
rgw_get_anon_user(s->user);
s->perm_mask = RGW_PERM_FULL_CONTROL;
@@ -787,8 +783,6 @@
if (!authorized)
return -EPERM;
- s->perm_mask = RGW_PERM_FULL_CONTROL;
-
return 0;
}
diff -Nru ceph-0.80.7/src/rgw/rgw_swift_auth.cc ceph-0.80.9/src/rgw/rgw_swift_auth.cc
--- ceph-0.80.7/src/rgw/rgw_swift_auth.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_swift_auth.cc 2015-03-09 20:51:18.000000000 +0100
@@ -56,7 +56,7 @@
return ret;
}
-int rgw_swift_verify_signed_token(CephContext *cct, RGWRados *store, const char *token, RGWUserInfo& info)
+int rgw_swift_verify_signed_token(CephContext *cct, RGWRados *store, const char *token, RGWUserInfo& info, string *pswift_user)
{
if (strncmp(token, "AUTH_rgwtk", 10) != 0)
return -EINVAL;
@@ -123,6 +123,7 @@
dout(0) << "NOTICE: tokens mismatch tok=" << buf << dendl;
return -EPERM;
}
+ *pswift_user = swift_user;
return 0;
}
@@ -205,7 +206,7 @@
tenant_path.append(g_conf->rgw_swift_tenant_name);
}
- s->cio->print("X-Storage-Url: %s/%s/v1%s\n", swift_url.c_str(),
+ s->cio->print("X-Storage-Url: %s/%s/v1%s\r\n", swift_url.c_str(),
swift_prefix.c_str(), tenant_path.c_str());
if ((ret = encode_token(s->cct, swift_key->id, swift_key->key, bl)) < 0)
@@ -215,8 +216,8 @@
char buf[bl.length() * 2 + 1];
buf_to_hex((const unsigned char *)bl.c_str(), bl.length(), buf);
- s->cio->print("X-Storage-Token: AUTH_rgwtk%s\n", buf);
- s->cio->print("X-Auth-Token: AUTH_rgwtk%s\n", buf);
+ s->cio->print("X-Storage-Token: AUTH_rgwtk%s\r\n", buf);
+ s->cio->print("X-Auth-Token: AUTH_rgwtk%s\r\n", buf);
}
ret = STATUS_NO_CONTENT;
diff -Nru ceph-0.80.7/src/rgw/rgw_swift_auth.h ceph-0.80.9/src/rgw/rgw_swift_auth.h
--- ceph-0.80.7/src/rgw/rgw_swift_auth.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_swift_auth.h 2015-03-09 20:51:18.000000000 +0100
@@ -6,7 +6,7 @@
#define RGW_SWIFT_TOKEN_EXPIRATION (15 * 60)
-extern int rgw_swift_verify_signed_token(CephContext *cct, RGWRados *store, const char *token, RGWUserInfo& info);
+extern int rgw_swift_verify_signed_token(CephContext *cct, RGWRados *store, const char *token, RGWUserInfo& info, string *pswift_user);
class RGW_SWIFT_Auth_Get : public RGWOp {
public:
diff -Nru ceph-0.80.7/src/rgw/rgw_swift.cc ceph-0.80.9/src/rgw/rgw_swift.cc
--- ceph-0.80.7/src/rgw/rgw_swift.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_swift.cc 2015-03-09 20:51:18.000000000 +0100
@@ -505,6 +505,8 @@
validate.append_header("X-Auth-Token", admin_token);
+ validate.set_send_length(0);
+
int ret = validate.process(url.c_str());
if (ret < 0)
return ret;
@@ -609,13 +611,41 @@
bool RGWSwift::verify_swift_token(RGWRados *store, req_state *s)
{
+ if (!do_verify_swift_token(store, s)) {
+ return false;
+ }
+
+ if (!s->swift_user.empty()) {
+ string subuser;
+ ssize_t pos = s->swift_user.find(':');
+ if (pos < 0) {
+ subuser = s->swift_user;
+ } else {
+ subuser = s->swift_user.substr(pos + 1);
+ }
+ s->perm_mask = 0;
+ map<string, RGWSubUser>::iterator iter = s->user.subusers.find(subuser);
+ if (iter != s->user.subusers.end()) {
+ RGWSubUser& subuser = iter->second;
+ s->perm_mask = subuser.perm_mask;
+ }
+ } else {
+ s->perm_mask = RGW_PERM_FULL_CONTROL;
+ }
+
+ return true;
+
+}
+
+bool RGWSwift::do_verify_swift_token(RGWRados *store, req_state *s)
+{
if (!s->os_auth_token) {
int ret = authenticate_temp_url(store, s);
return (ret >= 0);
}
if (strncmp(s->os_auth_token, "AUTH_rgwtk", 10) == 0) {
- int ret = rgw_swift_verify_signed_token(s->cct, store, s->os_auth_token, s->user);
+ int ret = rgw_swift_verify_signed_token(s->cct, store, s->os_auth_token, s->user, &s->swift_user);
if (ret < 0)
return false;
diff -Nru ceph-0.80.7/src/rgw/rgw_swift.h ceph-0.80.9/src/rgw/rgw_swift.h
--- ceph-0.80.7/src/rgw/rgw_swift.h 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_swift.h 2015-03-09 20:51:18.000000000 +0100
@@ -53,6 +53,7 @@
bool supports_keystone() {
return !cct->_conf->rgw_keystone_url.empty();
}
+ bool do_verify_swift_token(RGWRados *store, req_state *s);
protected:
int check_revoked();
public:
diff -Nru ceph-0.80.7/src/rgw/rgw_user.cc ceph-0.80.9/src/rgw/rgw_user.cc
--- ceph-0.80.7/src/rgw/rgw_user.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/rgw/rgw_user.cc 2015-03-09 20:51:18.000000000 +0100
@@ -666,15 +666,6 @@
switch (key_type) {
case KEY_TYPE_SWIFT:
- kiter = swift_keys->find(kid);
-
- existing_key = (kiter != swift_keys->end());
- if (existing_key)
- break;
-
- if (swift_kid.empty())
- return false;
-
kiter = swift_keys->find(swift_kid);
existing_key = (kiter != swift_keys->end());
@@ -845,7 +836,7 @@
} while (!rgw_get_user_info_by_access_key(store, id, duplicate_check));
}
- if (key_type == KEY_TYPE_SWIFT && gen_access) {
+ if (key_type == KEY_TYPE_SWIFT) {
id = op_state.build_default_swift_kid();
if (id.empty()) {
set_err_msg(err_msg, "empty swift access key");
diff -Nru ceph-0.80.7/src/test/cli/crushtool/add-item-in-tree.t ceph-0.80.9/src/test/cli/crushtool/add-item-in-tree.t
--- ceph-0.80.7/src/test/cli/crushtool/add-item-in-tree.t 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/add-item-in-tree.t 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,10 @@
+ $ crushtool -i "$TESTDIR/tree.template" --add-item 0 1.0 device0 --loc host host0 --loc cluster cluster0 -o one > /dev/null
+ $ crushtool -i one --add-item 1 1.0 device1 --loc host host0 --loc cluster cluster0 -o two > /dev/null
+ $ crushtool -i two --add-item 2 1.0 device2 --loc host host0 --loc cluster cluster0 -o tree > /dev/null
+ $ crushtool -i tree --add-item 3 1.0 device3 --loc host host0 --loc cluster cluster0 -o four > /dev/null
+ $ crushtool -i four --add-item 4 1.0 device4 --loc host host0 --loc cluster cluster0 -o five > /dev/null
+ $ crushtool -i five --add-item 5 1.0 device5 --loc host host0 --loc cluster cluster0 -o six > /dev/null
+ $ crushtool -i six --add-item 6 1.0 device6 --loc host host0 --loc cluster cluster0 -o seven > /dev/null
+ $ crushtool -i seven --add-item 7 1.0 device7 --loc host host0 --loc cluster cluster0 -o eight > /dev/null
+ $ crushtool -d eight -o final
+ $ cmp final "$TESTDIR/tree.template.final"
diff -Nru ceph-0.80.7/src/test/cli/crushtool/adjust-item-weight.t ceph-0.80.9/src/test/cli/crushtool/adjust-item-weight.t
--- ceph-0.80.7/src/test/cli/crushtool/adjust-item-weight.t 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/adjust-item-weight.t 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,17 @@
+ $ crushtool -i "$TESTDIR/simple.template" --add-item 0 1.0 device0 --loc host host0 --loc cluster cluster0 -o one > /dev/null
+
+#
+# add device0 into host=fake, the weight of device0 in host=host0 is 1.0, the weight of device0 in host=fake is 2.0
+#
+
+ $ crushtool -i one --add-item 0 2.0 device0 --loc host fake --loc cluster cluster0 -o two > /dev/null
+ $ crushtool -d two -o final
+ $ cmp final "$TESTDIR/simple.template.adj.two"
+
+#
+# update the weight of device0 in host=host0, it will not affect the weight of device0 in host=fake
+#
+
+ $ crushtool -i two --update-item 0 3.0 device0 --loc host host0 --loc cluster cluster0 -o three > /dev/null
+ $ crushtool -d three -o final
+ $ cmp final "$TESTDIR/simple.template.adj.three"
diff -Nru ceph-0.80.7/src/test/cli/crushtool/build.t ceph-0.80.9/src/test/cli/crushtool/build.t
--- ceph-0.80.7/src/test/cli/crushtool/build.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/build.t 2015-03-09 20:51:18.000000000 +0100
@@ -52,7 +52,7 @@
#
# crush rulesets are generated using the OSDMap helpers
#
- $ CEPH_ARGS="--debug-crush 0" crushtool --outfn "$map" --build --num_osds 1 root straw 0
+ $ CEPH_ARGS="--debug-crush 0" crushtool --outfn "$map" --set-straw-calc-version 0 --build --num_osds 1 root straw 0
$ crushtool -o "$map.txt" -d "$map"
$ cat "$map.txt"
# begin crush map
diff -Nru ceph-0.80.7/src/test/cli/crushtool/help.t ceph-0.80.9/src/test/cli/crushtool/help.t
--- ceph-0.80.7/src/test/cli/crushtool/help.t 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/help.t 2015-03-09 20:51:18.000000000 +0100
@@ -33,6 +33,7 @@
--show utilization-all
include zero weight items
--show-statistics show chi squared statistics
+ --show-mappings show mappings
--show-bad-mappings show bad mappings
--show-choose-tries show choose tries histogram
--set-choose-local-tries N
diff -Nru ceph-0.80.7/src/test/cli/crushtool/set-choose.t ceph-0.80.9/src/test/cli/crushtool/set-choose.t
--- ceph-0.80.7/src/test/cli/crushtool/set-choose.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/set-choose.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,5 +1,6 @@
$ crushtool -c "$TESTDIR/set-choose.crushmap.txt" -o set-choose.crushmap
- $ crushtool -i set-choose.crushmap --test --show-statistics
+ $ crushtool -i set-choose.crushmap --test --show-mappings --show-statistics --set-straw-calc-version 0
+ crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 0 (choose), x = 0..1023, numrep = 2..3
CRUSH rule 0 x 0 [0,3]
CRUSH rule 0 x 1 [0,8]
@@ -12306,7 +12307,8 @@
CRUSH rule 5 x 1022 [1,6,4]
CRUSH rule 5 x 1023 [3,2,8]
rule 5 (chooseleaf-set) num_rep 3 result size == 3:\t1024/1024 (esc)
- $ crushtool -i set-choose.crushmap --test --show-statistics --weight 0 0 --weight 1 0 --weight 3 0 --weight 4 0
+ $ crushtool -i set-choose.crushmap --test --show-mappings --show-statistics --weight 0 0 --weight 1 0 --weight 3 0 --weight 4 0 --set-straw-calc-version 0
+ crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 0 (choose), x = 0..1023, numrep = 2..3
CRUSH rule 0 x 0 [2,5]
CRUSH rule 0 x 1 [2,8]
@@ -24618,7 +24620,8 @@
CRUSH rule 5 x 1022 [2,6,5]
CRUSH rule 5 x 1023 [5,2,8]
rule 5 (chooseleaf-set) num_rep 3 result size == 3:\t1024/1024 (esc)
- $ crushtool -i set-choose.crushmap --test --show-statistics --weight 0 0 --weight 3 0 --weight 4 .5 --weight 5 0 --weight 6 .1 --weight 7 0
+ $ crushtool -i set-choose.crushmap --test --show-mappings --show-statistics --weight 0 0 --weight 3 0 --weight 4 .5 --weight 5 0 --weight 6 .1 --weight 7 0 --set-straw-calc-version 0
+ crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 0 (choose), x = 0..1023, numrep = 2..3
CRUSH rule 0 x 0 [2,4]
CRUSH rule 0 x 1 [2,8]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/simple.template.adj.one ceph-0.80.9/src/test/cli/crushtool/simple.template.adj.one
--- ceph-0.80.7/src/test/cli/crushtool/simple.template.adj.one 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/simple.template.adj.one 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,56 @@
+# begin crush map
+
+# devices
+device 0 device0
+
+# types
+type 0 device
+type 1 host
+type 2 cluster
+
+# buckets
+host host0 {
+ id -2 # do not change unnecessarily
+ # weight 1.000
+ alg straw
+ hash 0 # rjenkins1
+ item device0 weight 1.000
+}
+cluster cluster0 {
+ id -1 # do not change unnecessarily
+ # weight 1.000
+ alg straw
+ hash 0 # rjenkins1
+ item host0 weight 1.000
+}
+
+# rules
+rule data {
+ ruleset 0
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule metadata {
+ ruleset 1
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule rbd {
+ ruleset 2
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+
+# end crush map
diff -Nru ceph-0.80.7/src/test/cli/crushtool/simple.template.adj.three ceph-0.80.9/src/test/cli/crushtool/simple.template.adj.three
--- ceph-0.80.7/src/test/cli/crushtool/simple.template.adj.three 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/simple.template.adj.three 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,64 @@
+# begin crush map
+
+# devices
+device 0 device0
+
+# types
+type 0 device
+type 1 host
+type 2 cluster
+
+# buckets
+host host0 {
+ id -2 # do not change unnecessarily
+ # weight 3.000
+ alg straw
+ hash 0 # rjenkins1
+ item device0 weight 3.000
+}
+host fake {
+ id -3 # do not change unnecessarily
+ # weight 2.000
+ alg straw
+ hash 0 # rjenkins1
+ item device0 weight 2.000
+}
+cluster cluster0 {
+ id -1 # do not change unnecessarily
+ # weight 5.000
+ alg straw
+ hash 0 # rjenkins1
+ item host0 weight 3.000
+ item fake weight 2.000
+}
+
+# rules
+rule data {
+ ruleset 0
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule metadata {
+ ruleset 1
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule rbd {
+ ruleset 2
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+
+# end crush map
diff -Nru ceph-0.80.7/src/test/cli/crushtool/simple.template.adj.two ceph-0.80.9/src/test/cli/crushtool/simple.template.adj.two
--- ceph-0.80.7/src/test/cli/crushtool/simple.template.adj.two 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/simple.template.adj.two 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,64 @@
+# begin crush map
+
+# devices
+device 0 device0
+
+# types
+type 0 device
+type 1 host
+type 2 cluster
+
+# buckets
+host host0 {
+ id -2 # do not change unnecessarily
+ # weight 1.000
+ alg straw
+ hash 0 # rjenkins1
+ item device0 weight 1.000
+}
+host fake {
+ id -3 # do not change unnecessarily
+ # weight 2.000
+ alg straw
+ hash 0 # rjenkins1
+ item device0 weight 2.000
+}
+cluster cluster0 {
+ id -1 # do not change unnecessarily
+ # weight 3.000
+ alg straw
+ hash 0 # rjenkins1
+ item host0 weight 1.000
+ item fake weight 2.000
+}
+
+# rules
+rule data {
+ ruleset 0
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule metadata {
+ ruleset 1
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule rbd {
+ ruleset 2
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+
+# end crush map
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-bobtail-tunables.t ceph-0.80.9/src/test/cli/crushtool/test-map-bobtail-tunables.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-bobtail-tunables.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-bobtail-tunables.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-a.crushmap" --test --show-statistics --rule 0 --set-choose-local-tries 0 --set-choose-local-fallback-tries 0 --set-choose-total-tries 50 --set-chooseleaf-descend-once 1
+ $ crushtool -i "$TESTDIR/test-map-a.crushmap" --test --show-mappings --show-statistics --rule 0 --set-choose-local-tries 0 --set-choose-local-fallback-tries 0 --set-choose-total-tries 50 --set-chooseleaf-descend-once 1
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 0 (data), x = 0..1023, numrep = 1..10
CRUSH rule 0 x 0 [36]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-firefly-tunables.t ceph-0.80.9/src/test/cli/crushtool/test-map-firefly-tunables.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-firefly-tunables.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-firefly-tunables.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-statistics --rule 0 --set-choose-local-tries 0 --set-choose-local-fallback-tries 0 --set-choose-total-tries 50 --set-chooseleaf-descend-once 1 --set-chooseleaf-vary-r 1 --weight 12 0 --weight 20 0 --weight 30 0
+ $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-mappings --show-statistics --rule 0 --set-choose-local-tries 0 --set-choose-local-fallback-tries 0 --set-choose-total-tries 50 --set-chooseleaf-descend-once 1 --set-chooseleaf-vary-r 1 --weight 12 0 --weight 20 0 --weight 30 0
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 0 (data), x = 0..1023, numrep = 1..10
CRUSH rule 0 x 0 [101]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-firstn-indep.t ceph-0.80.9/src/test/cli/crushtool/test-map-firstn-indep.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-firstn-indep.t 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-firstn-indep.t 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,14 @@
+ $ crushtool -c "$TESTDIR/test-map-firstn-indep.txt" -o "$TESTDIR/test-map-firstn-indep.crushmap"
+ $ crushtool -i "$TESTDIR/test-map-firstn-indep.crushmap" --test --rule 0 --x 1 --show-bad-mappings
+ bad mapping rule 0 x 1 num_rep 9 result [93,80,88,87,56,50,53,72]
+ bad mapping rule 0 x 1 num_rep 10 result [93,80,88,87,56,50,53,72]
+ $ crushtool -i "$TESTDIR/test-map-firstn-indep.crushmap" --test --rule 1 --x 1 --show-bad-mappings
+ bad mapping rule 1 x 1 num_rep 3 result [93,56]
+ bad mapping rule 1 x 1 num_rep 4 result [93,56]
+ bad mapping rule 1 x 1 num_rep 5 result [93,56]
+ bad mapping rule 1 x 1 num_rep 6 result [93,56]
+ bad mapping rule 1 x 1 num_rep 7 result [93,56]
+ bad mapping rule 1 x 1 num_rep 8 result [93,56]
+ bad mapping rule 1 x 1 num_rep 9 result [93,56]
+ bad mapping rule 1 x 1 num_rep 10 result [93,56]
+ $ rm -f "$TESTDIR/test-map-firstn-indep.crushmap"
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-firstn-indep.txt ceph-0.80.9/src/test/cli/crushtool/test-map-firstn-indep.txt
--- ceph-0.80.7/src/test/cli/crushtool/test-map-firstn-indep.txt 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-firstn-indep.txt 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,443 @@
+# begin crush map
+tunable choose_local_tries 0
+tunable choose_local_fallback_tries 0
+tunable choose_total_tries 50
+tunable chooseleaf_descend_once 1
+
+# devices
+device 0 device0
+device 1 device1
+device 2 device2
+device 3 device3
+device 4 device4
+device 5 device5
+device 6 device6
+device 7 device7
+device 8 device8
+device 9 device9
+device 10 device10
+device 11 device11
+device 12 device12
+device 13 device13
+device 14 device14
+device 15 device15
+device 16 device16
+device 17 device17
+device 18 device18
+device 19 device19
+device 20 device20
+device 21 device21
+device 22 device22
+device 23 device23
+device 24 device24
+device 25 device25
+device 26 device26
+device 27 device27
+device 28 device28
+device 29 device29
+device 30 device30
+device 31 device31
+device 32 device32
+device 33 device33
+device 34 device34
+device 35 device35
+device 36 device36
+device 37 device37
+device 38 device38
+device 39 device39
+device 40 device40
+device 41 device41
+device 42 device42
+device 43 device43
+device 44 device44
+device 45 device45
+device 46 device46
+device 47 device47
+device 48 device48
+device 49 device49
+device 50 device50
+device 51 device51
+device 52 device52
+device 53 device53
+device 54 device54
+device 55 device55
+device 56 device56
+device 57 device57
+device 58 device58
+device 59 device59
+device 60 device60
+device 61 device61
+device 62 device62
+device 63 device63
+device 64 device64
+device 65 device65
+device 66 device66
+device 67 device67
+device 68 device68
+device 69 device69
+device 70 device70
+device 71 device71
+device 72 device72
+device 73 device73
+device 74 device74
+device 75 device75
+device 76 device76
+device 77 device77
+device 78 device78
+device 79 device79
+device 80 device80
+device 81 device81
+device 82 device82
+device 83 device83
+device 84 device84
+device 85 device85
+device 86 device86
+device 87 device87
+device 88 device88
+device 89 device89
+device 90 device90
+device 91 device91
+device 92 device92
+device 93 device93
+device 94 device94
+device 95 device95
+device 96 device96
+device 97 device97
+device 98 device98
+device 99 device99
+
+# types
+type 0 device
+type 1 host
+type 2 rack
+type 3 default
+
+# buckets
+host host0 {
+ id -1 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device0 weight 1.000
+ item device1 weight 1.000
+ item device2 weight 1.000
+ item device3 weight 1.000
+}
+host host1 {
+ id -2 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device4 weight 1.000
+ item device5 weight 1.000
+ item device6 weight 1.000
+ item device7 weight 1.000
+}
+host host2 {
+ id -3 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device8 weight 1.000
+ item device9 weight 1.000
+ item device10 weight 1.000
+ item device11 weight 1.000
+}
+host host3 {
+ id -4 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device12 weight 1.000
+ item device13 weight 1.000
+ item device14 weight 1.000
+ item device15 weight 1.000
+}
+host host4 {
+ id -5 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device16 weight 1.000
+ item device17 weight 1.000
+ item device18 weight 1.000
+ item device19 weight 1.000
+}
+host host5 {
+ id -6 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device20 weight 1.000
+ item device21 weight 1.000
+ item device22 weight 1.000
+ item device23 weight 1.000
+}
+host host6 {
+ id -7 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device24 weight 1.000
+ item device25 weight 1.000
+ item device26 weight 1.000
+ item device27 weight 1.000
+}
+host host7 {
+ id -8 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device28 weight 1.000
+ item device29 weight 1.000
+ item device30 weight 1.000
+ item device31 weight 1.000
+}
+host host8 {
+ id -9 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device32 weight 1.000
+ item device33 weight 1.000
+ item device34 weight 1.000
+ item device35 weight 1.000
+}
+host host9 {
+ id -10 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device36 weight 1.000
+ item device37 weight 1.000
+ item device38 weight 1.000
+ item device39 weight 1.000
+}
+host host10 {
+ id -11 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device40 weight 1.000
+ item device41 weight 1.000
+ item device42 weight 1.000
+ item device43 weight 1.000
+}
+host host11 {
+ id -12 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device44 weight 1.000
+ item device45 weight 1.000
+ item device46 weight 1.000
+ item device47 weight 1.000
+}
+host host12 {
+ id -13 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device48 weight 1.000
+ item device49 weight 1.000
+ item device50 weight 1.000
+ item device51 weight 1.000
+}
+host host13 {
+ id -14 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device52 weight 1.000
+ item device53 weight 1.000
+ item device54 weight 1.000
+ item device55 weight 1.000
+}
+host host14 {
+ id -15 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device56 weight 1.000
+ item device57 weight 1.000
+ item device58 weight 1.000
+ item device59 weight 1.000
+}
+host host15 {
+ id -16 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device60 weight 1.000
+ item device61 weight 1.000
+ item device62 weight 1.000
+ item device63 weight 1.000
+}
+host host16 {
+ id -17 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device64 weight 1.000
+ item device65 weight 1.000
+ item device66 weight 1.000
+ item device67 weight 1.000
+}
+host host17 {
+ id -18 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device68 weight 1.000
+ item device69 weight 1.000
+ item device70 weight 1.000
+ item device71 weight 1.000
+}
+host host18 {
+ id -19 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device72 weight 1.000
+ item device73 weight 1.000
+ item device74 weight 1.000
+ item device75 weight 1.000
+}
+host host19 {
+ id -20 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device76 weight 1.000
+ item device77 weight 1.000
+ item device78 weight 1.000
+ item device79 weight 1.000
+}
+host host20 {
+ id -21 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device80 weight 1.000
+ item device81 weight 1.000
+ item device82 weight 1.000
+ item device83 weight 1.000
+}
+host host21 {
+ id -22 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device84 weight 1.000
+ item device85 weight 1.000
+ item device86 weight 1.000
+ item device87 weight 1.000
+}
+host host22 {
+ id -23 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device88 weight 1.000
+ item device89 weight 1.000
+ item device90 weight 1.000
+ item device91 weight 1.000
+}
+host host23 {
+ id -24 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device92 weight 1.000
+ item device93 weight 1.000
+ item device94 weight 1.000
+ item device95 weight 1.000
+}
+host host24 {
+ id -25 # do not change unnecessarily
+ # weight 4.000
+ alg straw
+ hash 0 # rjenkins1
+ item device96 weight 1.000
+ item device97 weight 1.000
+ item device98 weight 1.000
+ item device99 weight 1.000
+}
+rack rack0 {
+ id -26 # do not change unnecessarily
+ # weight 40.000
+ alg straw
+ hash 0 # rjenkins1
+ item host0 weight 4.000
+ item host1 weight 4.000
+ item host2 weight 4.000
+ item host3 weight 4.000
+ item host4 weight 4.000
+ item host5 weight 4.000
+ item host6 weight 4.000
+ item host7 weight 4.000
+ item host8 weight 4.000
+ item host9 weight 4.000
+}
+rack rack1 {
+ id -27 # do not change unnecessarily
+ # weight 40.000
+ alg straw
+ hash 0 # rjenkins1
+ item host10 weight 4.000
+ item host11 weight 4.000
+ item host12 weight 4.000
+ item host13 weight 4.000
+ item host14 weight 4.000
+ item host15 weight 4.000
+ item host16 weight 4.000
+ item host17 weight 4.000
+ item host18 weight 4.000
+ item host19 weight 4.000
+}
+rack rack2 {
+ id -28 # do not change unnecessarily
+ # weight 20.000
+ alg straw
+ hash 0 # rjenkins1
+ item host20 weight 4.000
+ item host21 weight 4.000
+ item host22 weight 4.000
+ item host23 weight 4.000
+ item host24 weight 4.000
+}
+
+default root {
+ id -31 # do not change unnecessarily
+ # weight 100.000
+ alg straw
+ hash 0 # rjenkins1
+ item rack1 weight 40.000
+ item rack1 weight 40.000
+ item rack2 weight 20.000
+}
+
+# rules
+rule myrule {
+ ruleset 0
+ type replicated
+ min_size 1
+ max_size 10
+ step take root
+ step choose firstn 2 type rack
+ step chooseleaf indep 4 type host
+ step emit
+}
+
+rule myrule1 {
+ ruleset 1
+ type replicated
+ min_size 1
+ max_size 10
+ step take root
+ step choose firstn 2 type rack
+ step chooseleaf indep 1 type host
+ step emit
+}
+
+# end crush map
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-indep.t ceph-0.80.9/src/test/cli/crushtool/test-map-indep.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-indep.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-indep.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-indep.crushmap" --test --show-statistics --rule 1 --set-choose-local-tries 0 --set-choose-local-fallback-tries 0 --set-choose-total-tries 50 --set-chooseleaf-descend-once 2
+ $ crushtool -i "$TESTDIR/test-map-indep.crushmap" --test --show-mappings --show-statistics --rule 1 --set-choose-local-tries 0 --set-choose-local-fallback-tries 0 --set-choose-total-tries 50 --set-chooseleaf-descend-once 2
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 1 (metadata), x = 0..1023, numrep = 1..10
CRUSH rule 1 x 0 [36]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-legacy-tunables.t ceph-0.80.9/src/test/cli/crushtool/test-map-legacy-tunables.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-legacy-tunables.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-legacy-tunables.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-a.crushmap" --test --show-statistics --rule 0
+ $ crushtool -i "$TESTDIR/test-map-a.crushmap" --test --show-mappings --show-statistics --rule 0
rule 0 (data), x = 0..1023, numrep = 1..10
CRUSH rule 0 x 0 [36]
CRUSH rule 0 x 1 [876]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-tries-vs-retries.t ceph-0.80.9/src/test/cli/crushtool/test-map-tries-vs-retries.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-tries-vs-retries.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-tries-vs-retries.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-tries-vs-retries.crushmap" --test --show-statistics --weight 0 0 --weight 8 0
+ $ crushtool -i "$TESTDIR/test-map-tries-vs-retries.crushmap" --test --show-mappings --show-statistics --weight 0 0 --weight 8 0
rule 0 (replicated_ruleset), x = 0..1023, numrep = 1..10
CRUSH rule 0 x 0 [7]
CRUSH rule 0 x 1 [10]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-0.t ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-0.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-0.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-0.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-statistics --rule 3 --set-chooseleaf-vary-r 0 --weight 0 0 --weight 4 0 --weight 9 0
+ $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-mappings --show-statistics --rule 3 --set-chooseleaf-vary-r 0 --weight 0 0 --weight 4 0 --weight 9 0
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 3 (delltestrule), x = 0..1023, numrep = 2..4
CRUSH rule 3 x 0 [94,85]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-1.t ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-1.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-1.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-1.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-statistics --rule 3 --set-chooseleaf-vary-r 1 --weight 0 0 --weight 4 0 --weight 9 0
+ $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-mappings --show-statistics --rule 3 --set-chooseleaf-vary-r 1 --weight 0 0 --weight 4 0 --weight 9 0
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 3 (delltestrule), x = 0..1023, numrep = 2..4
CRUSH rule 3 x 0 [94,6]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-2.t ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-2.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-2.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-2.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-statistics --rule 3 --set-chooseleaf-vary-r 2 --weight 0 0 --weight 4 0 --weight 9 0
+ $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-mappings --show-statistics --rule 3 --set-chooseleaf-vary-r 2 --weight 0 0 --weight 4 0 --weight 9 0
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 3 (delltestrule), x = 0..1023, numrep = 2..4
CRUSH rule 3 x 0 [94,45]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-3.t ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-3.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-3.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-3.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-statistics --rule 3 --set-chooseleaf-vary-r 3 --weight 0 0 --weight 4 0 --weight 9 0
+ $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-mappings --show-statistics --rule 3 --set-chooseleaf-vary-r 3 --weight 0 0 --weight 4 0 --weight 9 0
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 3 (delltestrule), x = 0..1023, numrep = 2..4
CRUSH rule 3 x 0 [94,85]
diff -Nru ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-4.t ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-4.t
--- ceph-0.80.7/src/test/cli/crushtool/test-map-vary-r-4.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/crushtool/test-map-vary-r-4.t 2015-03-09 20:51:18.000000000 +0100
@@ -1,4 +1,4 @@
- $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-statistics --rule 3 --set-chooseleaf-vary-r 4 --weight 0 0 --weight 4 0 --weight 9 0
+ $ crushtool -i "$TESTDIR/test-map-vary-r.crushmap" --test --show-mappings --show-statistics --rule 3 --set-chooseleaf-vary-r 4 --weight 0 0 --weight 4 0 --weight 9 0
crushtool successfully built or modified map. Use '-o <file>' to write it out.
rule 3 (delltestrule), x = 0..1023, numrep = 2..4
CRUSH rule 3 x 0 [94,85]
Binary files /tmp/uijd9k0H_0/ceph-0.80.7/src/test/cli/crushtool/tree.template and /tmp/eOaGzSVj4X/ceph-0.80.9/src/test/cli/crushtool/tree.template differ
diff -Nru ceph-0.80.7/src/test/cli/crushtool/tree.template.final ceph-0.80.9/src/test/cli/crushtool/tree.template.final
--- ceph-0.80.7/src/test/cli/crushtool/tree.template.final 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/cli/crushtool/tree.template.final 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,70 @@
+# begin crush map
+
+# devices
+device 0 device0
+device 1 device1
+device 2 device2
+device 3 device3
+device 4 device4
+device 5 device5
+device 6 device6
+device 7 device7
+
+# types
+type 0 device
+type 1 host
+type 2 cluster
+
+# buckets
+host host0 {
+ id -2 # do not change unnecessarily
+ # weight 8.000
+ alg tree # do not change pos for existing items unnecessarily
+ hash 0 # rjenkins1
+ item device0 weight 1.000 pos 0
+ item device1 weight 1.000 pos 1
+ item device2 weight 1.000 pos 2
+ item device3 weight 1.000 pos 3
+ item device4 weight 1.000 pos 4
+ item device5 weight 1.000 pos 5
+ item device6 weight 1.000 pos 6
+ item device7 weight 1.000 pos 7
+}
+cluster cluster0 {
+ id -1 # do not change unnecessarily
+ # weight 8.000
+ alg tree # do not change pos for existing items unnecessarily
+ hash 0 # rjenkins1
+ item host0 weight 8.000 pos 0
+}
+
+# rules
+rule data {
+ ruleset 0
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule metadata {
+ ruleset 1
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+rule rbd {
+ ruleset 2
+ type replicated
+ min_size 1
+ max_size 10
+ step take cluster0
+ step chooseleaf firstn 0 type host
+ step emit
+}
+
+# end crush map
diff -Nru ceph-0.80.7/src/test/cli/osdmaptool/create-print.t ceph-0.80.9/src/test/cli/osdmaptool/create-print.t
--- ceph-0.80.7/src/test/cli/osdmaptool/create-print.t 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/cli/osdmaptool/create-print.t 2015-03-09 20:51:18.000000000 +0100
@@ -11,6 +11,7 @@
tunable choose_local_fallback_tries 0
tunable choose_total_tries 50
tunable chooseleaf_descend_once 1
+ tunable straw_calc_version 1
# devices
device 0 osd.0
diff -Nru ceph-0.80.7/src/test/cli/osdmaptool/create-racks.t ceph-0.80.9/src/test/cli/osdmaptool/create-racks.t
--- ceph-0.80.7/src/test/cli/osdmaptool/create-racks.t 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/cli/osdmaptool/create-racks.t 2015-03-09 20:51:18.000000000 +0100
@@ -10,6 +10,7 @@
tunable choose_local_fallback_tries 0
tunable choose_total_tries 50
tunable chooseleaf_descend_once 1
+ tunable straw_calc_version 1
# devices
device 0 device0
diff -Nru ceph-0.80.7/src/test/cli/osdmaptool/crush.t ceph-0.80.9/src/test/cli/osdmaptool/crush.t
--- ceph-0.80.7/src/test/cli/osdmaptool/crush.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/osdmaptool/crush.t 2015-03-09 20:51:18.000000000 +0100
@@ -6,5 +6,5 @@
osdmaptool: exported crush map to oc
$ osdmaptool --import-crush oc myosdmap
osdmaptool: osdmap file 'myosdmap'
- osdmaptool: imported 486 byte crush map from oc
+ osdmaptool: imported 487 byte crush map from oc
osdmaptool: writing epoch 3 to myosdmap
diff -Nru ceph-0.80.7/src/test/cli/osdmaptool/help.t ceph-0.80.9/src/test/cli/osdmaptool/help.t
--- ceph-0.80.7/src/test/cli/osdmaptool/help.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/osdmaptool/help.t 2015-03-09 20:51:18.000000000 +0100
@@ -4,6 +4,7 @@
--export-crush <file> write osdmap's crush map to <file>
--import-crush <file> replace osdmap's crush map with <file>
--test-map-pgs [--pool <poolid>] map all pgs
+ --test-map-pgs-dump [--pool <poolid>] map all pgs
--mark-up-in mark osds up and in (but do not persist)
--clear-temp clear pg_temp and primary_temp
--test-random do random placements
diff -Nru ceph-0.80.7/src/test/cli/osdmaptool/missing-argument.t ceph-0.80.9/src/test/cli/osdmaptool/missing-argument.t
--- ceph-0.80.7/src/test/cli/osdmaptool/missing-argument.t 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/cli/osdmaptool/missing-argument.t 2015-03-09 20:51:18.000000000 +0100
@@ -4,6 +4,7 @@
--export-crush <file> write osdmap's crush map to <file>
--import-crush <file> replace osdmap's crush map with <file>
--test-map-pgs [--pool <poolid>] map all pgs
+ --test-map-pgs-dump [--pool <poolid>] map all pgs
--mark-up-in mark osds up and in (but do not persist)
--clear-temp clear pg_temp and primary_temp
--test-random do random placements
diff -Nru ceph-0.80.7/src/test/cli/osdmaptool/test-map-pgs.t ceph-0.80.9/src/test/cli/osdmaptool/test-map-pgs.t
--- ceph-0.80.7/src/test/cli/osdmaptool/test-map-pgs.t 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/cli/osdmaptool/test-map-pgs.t 2015-03-09 20:51:18.000000000 +0100
@@ -24,7 +24,7 @@
pool 1 pg_num 8000
pool 2 pg_num 8000
$ TOTAL=$((POOL_COUNT * $PG_NUM))
- $ PATTERN=$(echo "size $SIZE\t$TOTAL")
+ $ PATTERN=$(echo "size $SIZE.$TOTAL")
$ grep "$PATTERN" $OUT || cat "$OUT"
size 3\t24000 (esc)
$ STATS_CRUSH=$(grep '^ avg ' "$OUT")
@@ -39,7 +39,7 @@
pool 1 pg_num 8000
pool 2 pg_num 8000
$ TOTAL=$((POOL_COUNT * $PG_NUM))
- $ PATTERN=$(echo "size $SIZE\t$TOTAL")
+ $ PATTERN=$(echo "size $SIZE.$TOTAL")
$ grep "$PATTERN" $OUT || cat "$OUT"
size 3\t24000 (esc)
$ STATS_RANDOM=$(grep '^ avg ' "$OUT")
diff -Nru ceph-0.80.7/src/test/common/histogram.cc ceph-0.80.9/src/test/common/histogram.cc
--- ceph-0.80.7/src/test/common/histogram.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/common/histogram.cc 2015-03-09 20:51:18.000000000 +0100
@@ -47,61 +47,64 @@
}
TEST(Histogram, Position) {
- {
- pow2_hist_t h;
- uint64_t lb, ub;
- h.add(0);
- ASSERT_EQ(-1, h.get_position_micro(-20, &lb, &ub));
- }
- {
- pow2_hist_t h;
- h.add(0);
- uint64_t lb, ub;
- h.get_position_micro(0, &lb, &ub);
- ASSERT_EQ(0u, lb);
- ASSERT_EQ(1000000u, ub);
- h.add(0);
- h.add(0);
- h.add(0);
- h.get_position_micro(0, &lb, &ub);
- ASSERT_EQ(0u, lb);
- ASSERT_EQ(1000000u, ub);
- }
- {
- pow2_hist_t h;
- h.add(1);
- h.add(1);
- uint64_t lb, ub;
- h.get_position_micro(0, &lb, &ub);
- ASSERT_EQ(0u, lb);
- ASSERT_EQ(0u, ub);
- h.add(0);
- h.get_position_micro(0, &lb, &ub);
- ASSERT_EQ(0u, lb);
- ASSERT_EQ(333333u, ub);
- h.get_position_micro(1, &lb, &ub);
- ASSERT_EQ(333333u, lb);
- ASSERT_EQ(1000000u, ub);
- }
- {
- pow2_hist_t h;
- h.h.resize(10, 0);
- h.h[0] = 1;
- h.h[5] = 1;
- uint64_t lb, ub;
- h.get_position_micro(4, &lb, &ub);
- ASSERT_EQ(500000u, lb);
- ASSERT_EQ(500000u, ub);
- }
- {
- pow2_hist_t h;
- h.h.resize(10, 0);
- h.h[0] = UINT_MAX;
- h.h[5] = UINT_MAX;
- uint64_t lb, ub;
- ASSERT_EQ(500000u, lb);
- ASSERT_EQ(500000u, ub);
- }
+ pow2_hist_t h;
+ uint64_t lb, ub;
+ h.add(0);
+ ASSERT_EQ(-1, h.get_position_micro(-20, &lb, &ub));
+}
+
+TEST(Histogram, Position1) {
+ pow2_hist_t h;
+ h.add(0);
+ uint64_t lb, ub;
+ h.get_position_micro(0, &lb, &ub);
+ ASSERT_EQ(0u, lb);
+ ASSERT_EQ(1000000u, ub);
+ h.add(0);
+ h.add(0);
+ h.add(0);
+ h.get_position_micro(0, &lb, &ub);
+ ASSERT_EQ(0u, lb);
+ ASSERT_EQ(1000000u, ub);
+}
+
+TEST(Histogram, Position2) {
+ pow2_hist_t h;
+ h.add(1);
+ h.add(1);
+ uint64_t lb, ub;
+ h.get_position_micro(0, &lb, &ub);
+ ASSERT_EQ(0u, lb);
+ ASSERT_EQ(0u, ub);
+ h.add(0);
+ h.get_position_micro(0, &lb, &ub);
+ ASSERT_EQ(0u, lb);
+ ASSERT_EQ(333333u, ub);
+ h.get_position_micro(1, &lb, &ub);
+ ASSERT_EQ(333333u, lb);
+ ASSERT_EQ(1000000u, ub);
+}
+
+TEST(Histogram, Position3) {
+ pow2_hist_t h;
+ h.h.resize(10, 0);
+ h.h[0] = 1;
+ h.h[5] = 1;
+ uint64_t lb, ub;
+ h.get_position_micro(4, &lb, &ub);
+ ASSERT_EQ(500000u, lb);
+ ASSERT_EQ(500000u, ub);
+}
+
+TEST(Histogram, Position4) {
+ pow2_hist_t h;
+ h.h.resize(10, 0);
+ h.h[0] = UINT_MAX;
+ h.h[5] = UINT_MAX;
+ uint64_t lb, ub;
+ h.get_position_micro(4, &lb, &ub);
+ ASSERT_EQ(0u, lb);
+ ASSERT_EQ(0u, ub);
}
TEST(Histogram, Decay) {
diff -Nru ceph-0.80.7/src/test/common/test_io_priority.cc ceph-0.80.9/src/test/common/test_io_priority.cc
--- ceph-0.80.7/src/test/common/test_io_priority.cc 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/common/test_io_priority.cc 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,51 @@
+// -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
+// vim: ts=8 sw=2 smarttab
+/*
+ * Ceph - scalable distributed file system
+ *
+ * Copyright (C) 2014 Red Hat <contact@redhat.com>
+ *
+ * Author: Loic Dachary <loic@dachary.org>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ */
+
+#include <errno.h>
+#include <gtest/gtest.h>
+
+#include "common/io_priority.h"
+
+TEST(io_priority, ceph_ioprio_string_to_class) {
+ ASSERT_EQ(IOPRIO_CLASS_IDLE, ceph_ioprio_string_to_class("idle"));
+ ASSERT_EQ(IOPRIO_CLASS_IDLE, ceph_ioprio_string_to_class("IDLE"));
+
+ ASSERT_EQ(IOPRIO_CLASS_BE, ceph_ioprio_string_to_class("be"));
+ ASSERT_EQ(IOPRIO_CLASS_BE, ceph_ioprio_string_to_class("BE"));
+ ASSERT_EQ(IOPRIO_CLASS_BE, ceph_ioprio_string_to_class("besteffort"));
+ ASSERT_EQ(IOPRIO_CLASS_BE, ceph_ioprio_string_to_class("BESTEFFORT"));
+ ASSERT_EQ(IOPRIO_CLASS_BE, ceph_ioprio_string_to_class("best effort"));
+ ASSERT_EQ(IOPRIO_CLASS_BE, ceph_ioprio_string_to_class("BEST EFFORT"));
+
+ ASSERT_EQ(IOPRIO_CLASS_RT, ceph_ioprio_string_to_class("rt"));
+ ASSERT_EQ(IOPRIO_CLASS_RT, ceph_ioprio_string_to_class("RT"));
+ ASSERT_EQ(IOPRIO_CLASS_RT, ceph_ioprio_string_to_class("realtime"));
+ ASSERT_EQ(IOPRIO_CLASS_RT, ceph_ioprio_string_to_class("REALTIME"));
+ ASSERT_EQ(IOPRIO_CLASS_RT, ceph_ioprio_string_to_class("real time"));
+ ASSERT_EQ(IOPRIO_CLASS_RT, ceph_ioprio_string_to_class("REAL TIME"));
+
+ ASSERT_EQ(-EINVAL, ceph_ioprio_string_to_class("invalid"));
+}
+
+/*
+ * Local Variables:
+ * compile-command: "cd ../.. ;
+ * make -j4 unittest_io_priority &&
+ * libtool --mode=execute valgrind --tool=memcheck --leak-check=full \
+ * ./unittest_io_priority
+ * "
+ * End:
+ */
diff -Nru ceph-0.80.7/src/test/crush/indep.cc ceph-0.80.9/src/test/crush/indep.cc
--- ceph-0.80.7/src/test/crush/indep.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/crush/indep.cc 2015-03-09 20:51:18.000000000 +0100
@@ -51,18 +51,21 @@
}
}
}
-
- crush_rule *rule = crush_make_rule(4, 0, 123, 1, 20);
- assert(rule);
- crush_rule_set_step(rule, 0, CRUSH_RULE_SET_CHOOSELEAF_TRIES, 10, 0);
- crush_rule_set_step(rule, 1, CRUSH_RULE_TAKE, rootno, 0);
- crush_rule_set_step(rule, 2,
- CRUSH_RULE_CHOOSELEAF_INDEP,
- CRUSH_CHOOSE_N,
- 1);
- crush_rule_set_step(rule, 3, CRUSH_RULE_EMIT, 0, 0);
- int rno = crush_add_rule(c->crush, rule, -1);
- c->set_rule_name(rno, "data");
+ int ret;
+ int ruleno = 0;
+ int ruleset = 0;
+ ruleno = ruleset;
+ ret = c->add_rule(4, ruleset, 123, 1, 20, ruleno);
+ assert(ret == ruleno);
+ ret = c->set_rule_step(ruleno, 0, CRUSH_RULE_SET_CHOOSELEAF_TRIES, 10, 0);
+ assert(ret == 0);
+ ret = c->set_rule_step(ruleno, 1, CRUSH_RULE_TAKE, rootno, 0);
+ assert(ret == 0);
+ ret = c->set_rule_step(ruleno, 2, CRUSH_RULE_CHOOSELEAF_INDEP, CRUSH_CHOOSE_N, 1);
+ assert(ret == 0);
+ ret = c->set_rule_step(ruleno, 3, CRUSH_RULE_EMIT, 0, 0);
+ assert(ret == 0);
+ c->set_rule_name(ruleno, "data");
if (false) {
Formatter *f = new_formatter("json-pretty");
@@ -140,7 +143,7 @@
c->dump_tree(weight, &cout, NULL);
// need more retries to get 9/9 hosts for x in 0..99
- c->crush->choose_total_tries = 100;
+ c->set_choose_total_tries(100);
for (int x = 0; x < 100; ++x) {
vector<int> out;
c->do_rule(0, x, out, 9, weight);
@@ -166,7 +169,7 @@
weight[i] = 0;
c->dump_tree(weight, &cout, NULL);
- c->crush->choose_total_tries = 100;
+ c->set_choose_total_tries(100);
for (int x = 0; x < 100; ++x) {
vector<int> out;
c->do_rule(0, x, out, 7, weight);
@@ -185,7 +188,7 @@
TEST(CRUSH, indep_out_progressive) {
CrushWrapper *c = build_indep_map(g_ceph_context, 3, 3, 3);
- c->crush->choose_total_tries = 100;
+ c->set_choose_total_tries(100);
vector<__u32> tweight(c->get_max_devices(), 0x10000);
c->dump_tree(tweight, &cout, NULL);
diff -Nru ceph-0.80.7/src/test/crush/TestCrushWrapper.cc ceph-0.80.9/src/test/crush/TestCrushWrapper.cc
--- ceph-0.80.7/src/test/crush/TestCrushWrapper.cc 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/crush/TestCrushWrapper.cc 2015-03-09 20:51:18.000000000 +0100
@@ -67,6 +67,166 @@
delete c;
}
+TEST(CrushWrapper, straw_zero) {
+ // zero weight items should have no effect on placement.
+
+ CrushWrapper *c = new CrushWrapper;
+ const int ROOT_TYPE = 1;
+ c->set_type_name(ROOT_TYPE, "root");
+ const int OSD_TYPE = 0;
+ c->set_type_name(OSD_TYPE, "osd");
+
+ int n = 5;
+ int items[n], weights[n];
+ for (int i=0; i <n; ++i) {
+ items[i] = i;
+ weights[i] = 0x10000 * (n-i-1);
+ }
+
+ c->set_max_devices(n);
+
+ string root_name0("root0");
+ int root0;
+ EXPECT_EQ(0, c->add_bucket(0, CRUSH_BUCKET_STRAW, CRUSH_HASH_RJENKINS1,
+ ROOT_TYPE, n, items, weights, &root0));
+ EXPECT_EQ(0, c->set_item_name(root0, root_name0));
+
+ string name0("rule0");
+ int ruleset0 = c->add_simple_ruleset(name0, root_name0, "osd",
+ "firstn", pg_pool_t::TYPE_REPLICATED);
+ EXPECT_EQ(0, ruleset0);
+
+ string root_name1("root1");
+ int root1;
+ EXPECT_EQ(0, c->add_bucket(0, CRUSH_BUCKET_STRAW, CRUSH_HASH_RJENKINS1,
+ ROOT_TYPE, n-1, items, weights, &root1));
+ EXPECT_EQ(0, c->set_item_name(root1, root_name1));
+
+ string name1("rule1");
+ int ruleset1 = c->add_simple_ruleset(name1, root_name1, "osd",
+ "firstn", pg_pool_t::TYPE_REPLICATED);
+ EXPECT_EQ(1, ruleset1);
+
+ vector<unsigned> reweight(n, 0x10000);
+ for (int i=0; i<10000; ++i) {
+ vector<int> out0, out1;
+ c->do_rule(ruleset0, i, out0, 1, reweight);
+ ASSERT_EQ(1u, out0.size());
+ c->do_rule(ruleset1, i, out1, 1, reweight);
+ ASSERT_EQ(1u, out1.size());
+ ASSERT_EQ(out0[0], out1[0]);
+ //cout << i << "\t" << out0 << "\t" << out1 << std::endl;
+ }
+}
+
+TEST(CrushWrapper, straw_same) {
+ // items with the same weight should map about the same as items
+ // with very similar weights.
+ //
+ // give the 0 vector a paired stair pattern, with dup weights. note
+ // that the original straw flaw does not appear when there are 2 of
+ // the initial weight, but it does when there is just 1.
+ //
+ // give the 1 vector a similar stair pattern, but make the same
+ // steps weights slightly different (no dups). this works.
+ //
+ // compare the result and verify that the resulting mapping is
+ // almost identical.
+
+ CrushWrapper *c = new CrushWrapper;
+ const int ROOT_TYPE = 1;
+ c->set_type_name(ROOT_TYPE, "root");
+ const int OSD_TYPE = 0;
+ c->set_type_name(OSD_TYPE, "osd");
+
+ int n = 10;
+ int items[n], weights[n];
+ for (int i=0; i <n; ++i) {
+ items[i] = i;
+ weights[i] = 0x10000 * ((i+1)/2 + 1);
+ }
+
+ c->set_max_devices(n);
+
+ string root_name0("root0");
+ int root0;
+ EXPECT_EQ(0, c->add_bucket(0, CRUSH_BUCKET_STRAW, CRUSH_HASH_RJENKINS1,
+ ROOT_TYPE, n, items, weights, &root0));
+ EXPECT_EQ(0, c->set_item_name(root0, root_name0));
+
+ string name0("rule0");
+ int ruleset0 = c->add_simple_ruleset(name0, root_name0, "osd",
+ "firstn", pg_pool_t::TYPE_REPLICATED);
+ EXPECT_EQ(0, ruleset0);
+
+ for (int i=0; i <n; ++i) {
+ items[i] = i;
+ weights[i] = 0x10000 * ((i+1)/2 + 1) + (i%2)*100;
+ }
+
+ string root_name1("root1");
+ int root1;
+ EXPECT_EQ(0, c->add_bucket(0, CRUSH_BUCKET_STRAW, CRUSH_HASH_RJENKINS1,
+ ROOT_TYPE, n, items, weights, &root1));
+ EXPECT_EQ(0, c->set_item_name(root1, root_name1));
+
+ string name1("rule1");
+ int ruleset1 = c->add_simple_ruleset(name1, root_name1, "osd",
+ "firstn", pg_pool_t::TYPE_REPLICATED);
+ EXPECT_EQ(1, ruleset1);
+
+ if (0) {
+ crush_bucket_straw *sb0 = reinterpret_cast<crush_bucket_straw*>(c->get_crush_map()->buckets[-1-root0]);
+ crush_bucket_straw *sb1 = reinterpret_cast<crush_bucket_straw*>(c->get_crush_map()->buckets[-1-root1]);
+
+ for (int i=0; i<n; ++i) {
+ cout << i
+ << "\t" << sb0->item_weights[i]
+ << "\t" << sb1->item_weights[i]
+ << "\t"
+ << "\t" << sb0->straws[i]
+ << "\t" << sb1->straws[i]
+ << std::endl;
+ }
+ }
+
+ if (0) {
+ JSONFormatter jf(true);
+ jf.open_object_section("crush");
+ c->dump(&jf);
+ jf.close_section();
+ jf.flush(cout);
+ }
+
+ vector<int> sum0(n, 0), sum1(n, 0);
+ vector<unsigned> reweight(n, 0x10000);
+ int different = 0;
+ int max = 100000;
+ for (int i=0; i<max; ++i) {
+ vector<int> out0, out1;
+ c->do_rule(ruleset0, i, out0, 1, reweight);
+ ASSERT_EQ(1u, out0.size());
+ c->do_rule(ruleset1, i, out1, 1, reweight);
+ ASSERT_EQ(1u, out1.size());
+ sum0[out0[0]]++;
+ sum1[out1[0]]++;
+ if (out0[0] != out1[0])
+ different++;
+ }
+ for (int i=0; i<n; ++i) {
+ cout << i
+ << "\t" << ((double)weights[i] / (double)weights[0])
+ << "\t" << sum0[i] << "\t" << ((double)sum0[i]/(double)sum0[0])
+ << "\t" << sum1[i] << "\t" << ((double)sum1[i]/(double)sum1[0])
+ << std::endl;
+ }
+ double ratio = ((double)different / (double)max);
+ cout << different << " of " << max << " = "
+ << ratio
+ << " different" << std::endl;
+ ASSERT_LT(ratio, .001);
+}
+
TEST(CrushWrapper, move_bucket) {
CrushWrapper *c = new CrushWrapper;
@@ -290,6 +450,116 @@
delete c;
}
+TEST(CrushWrapper, adjust_item_weight) {
+ CrushWrapper *c = new CrushWrapper;
+
+ const int ROOT_TYPE = 2;
+ c->set_type_name(ROOT_TYPE, "root");
+ const int HOST_TYPE = 1;
+ c->set_type_name(HOST_TYPE, "host");
+ const int OSD_TYPE = 0;
+ c->set_type_name(OSD_TYPE, "osd");
+
+ int rootno;
+ c->add_bucket(0, CRUSH_BUCKET_STRAW, CRUSH_HASH_RJENKINS1,
+ ROOT_TYPE, 0, NULL, NULL, &rootno);
+ c->set_item_name(rootno, "default");
+
+ const string HOST0("host0");
+ int host0;
+ c->add_bucket(0, CRUSH_BUCKET_STRAW, CRUSH_HASH_RJENKINS1,
+ HOST_TYPE, 0, NULL, NULL, &host0);
+ c->set_item_name(host0, HOST0);
+
+ const string FAKE("fake");
+ int hostfake;
+ c->add_bucket(0, CRUSH_BUCKET_STRAW, CRUSH_HASH_RJENKINS1,
+ HOST_TYPE, 0, NULL, NULL, &hostfake);
+ c->set_item_name(hostfake, FAKE);
+
+ int item = 0;
+
+ // construct crush map
+
+ {
+ map<string,string> loc;
+ loc["host"] = "host0";
+ float host_weight = 2.0;
+ int bucket_id = 0;
+
+ item = 0;
+ EXPECT_EQ(0, c->insert_item(g_ceph_context, item, 1.0,
+ "osd." + stringify(item), loc));
+ item = 1;
+ EXPECT_EQ(0, c->insert_item(g_ceph_context, item, 1.0,
+ "osd." + stringify(item), loc));
+
+ bucket_id = c->get_item_id("host0");
+ EXPECT_EQ(true, c->bucket_exists(bucket_id));
+ EXPECT_EQ(host_weight, c->get_bucket_weightf(bucket_id));
+
+ }
+
+ {
+ map<string,string> loc;
+ loc["host"] = "fake";
+ float host_weight = 2.0;
+ int bucket_id = 0;
+
+ item = 0;
+ EXPECT_EQ(0, c->insert_item(g_ceph_context, item, 1.0,
+ "osd." + stringify(item), loc));
+ item = 1;
+ EXPECT_EQ(0, c->insert_item(g_ceph_context, item, 1.0,
+ "osd." + stringify(item), loc));
+
+ bucket_id = c->get_item_id("fake");
+ EXPECT_EQ(true, c->bucket_exists(bucket_id));
+ EXPECT_EQ(host_weight, c->get_bucket_weightf(bucket_id));
+ }
+
+ //
+ // When there is:
+ //
+ // default --> host0 --> osd.0 1.0
+ // | |
+ // | +-> osd.1 1.0
+ // |
+ // +-> fake --> osd.0 1.0
+ // |
+ // +-> osd.1 1.0
+ //
+ // Trying to adjust osd.0 weight to 2.0 in all buckets
+ // Trying to adjust osd.1 weight to 2.0 in host=fake
+ //
+ // So the crush map will be:
+ //
+ // default --> host0 --> osd.0 2.0
+ // | |
+ // | +-> osd.1 1.0
+ // |
+ // +-> fake --> osd.0 2.0
+ // |
+ // +-> osd.1 2.0
+ //
+
+ float original_weight = 1.0;
+ float modified_weight = 2.0;
+ map<string,string> loc_one, loc_two;
+ loc_one["host"] = "host0";
+ loc_two["host"] = "fake";
+
+ item = 0;
+ EXPECT_EQ(2, c->adjust_item_weightf(g_ceph_context, item, modified_weight));
+ EXPECT_EQ(modified_weight, c->get_item_weightf_in_loc(item, loc_one));
+ EXPECT_EQ(modified_weight, c->get_item_weightf_in_loc(item, loc_two));
+
+ item = 1;
+ EXPECT_EQ(1, c->adjust_item_weightf_in_loc(g_ceph_context, item, modified_weight, loc_two));
+ EXPECT_EQ(original_weight, c->get_item_weightf_in_loc(item, loc_one));
+ EXPECT_EQ(modified_weight, c->get_item_weightf_in_loc(item, loc_two));
+}
+
TEST(CrushWrapper, insert_item) {
CrushWrapper *c = new CrushWrapper;
diff -Nru ceph-0.80.7/src/test/erasure-code/ceph_erasure_code_non_regression.cc ceph-0.80.9/src/test/erasure-code/ceph_erasure_code_non_regression.cc
--- ceph-0.80.7/src/test/erasure-code/ceph_erasure_code_non_regression.cc 1970-01-01 01:00:00.000000000 +0100
+++ ceph-0.80.9/src/test/erasure-code/ceph_erasure_code_non_regression.cc 2015-03-09 20:51:18.000000000 +0100
@@ -0,0 +1,325 @@
+// -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
+// vim: ts=8 sw=2 smarttab
+/*
+ * Ceph distributed storage system
+ *
+ * Red Hat (C) 2014 Red Hat <contact@redhat.com>
+ *
+ * Author: Loic Dachary <loic@dachary.org>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ */
+
+#include <errno.h>
+#include <boost/scoped_ptr.hpp>
+#include <boost/lexical_cast.hpp>
+#include <boost/program_options/option.hpp>
+#include <boost/program_options/options_description.hpp>
+#include <boost/program_options/variables_map.hpp>
+#include <boost/program_options/cmdline.hpp>
+#include <boost/program_options/parsers.hpp>
+#include <boost/algorithm/string.hpp>
+
+#include "global/global_context.h"
+#include "global/global_init.h"
+#include "common/errno.h"
+#include "common/ceph_argparse.h"
+#include "common/config.h"
+#include "erasure-code/ErasureCodePlugin.h"
+
+namespace po = boost::program_options;
+using namespace std;
+
+class ErasureCodeNonRegression {
+ unsigned stripe_width;
+ string plugin;
+ bool create;
+ bool check;
+ string base;
+ string directory;
+ map<string,string> parameters;
+public:
+ int setup(int argc, char** argv);
+ int run();
+ int run_create();
+ int run_check();
+ int decode_erasures(ErasureCodeInterfaceRef erasure_code,
+ set<int> erasures,
+ map<int,bufferlist> chunks);
+ string content_path();
+ string chunk_path(unsigned int chunk);
+};
+
+int ErasureCodeNonRegression::setup(int argc, char** argv) {
+
+ po::options_description desc("Allowed options");
+ desc.add_options()
+ ("help,h", "produce help message")
+ ("stripe-width,s", po::value<int>()->default_value(4 * 1024),
+ "stripe_width, i.e. the size of the buffer to be encoded")
+ ("plugin,p", po::value<string>()->default_value("jerasure"),
+ "erasure code plugin name")
+ ("base", po::value<string>()->default_value("."),
+ "prefix all paths with base")
+ ("parameter,P", po::value<vector<string> >(),
+ "parameters")
+ ("create", "create the erasure coded content in the directory")
+ ("check", "check the content in the directory matches the chunks and vice versa")
+ ;
+
+ po::variables_map vm;
+ po::parsed_options parsed =
+ po::command_line_parser(argc, argv).options(desc).allow_unregistered().run();
+ po::store(
+ parsed,
+ vm);
+ po::notify(vm);
+
+ vector<const char *> ceph_options, def_args;
+ vector<string> ceph_option_strings = po::collect_unrecognized(
+ parsed.options, po::include_positional);
+ ceph_options.reserve(ceph_option_strings.size());
+ for (vector<string>::iterator i = ceph_option_strings.begin();
+ i != ceph_option_strings.end();
+ ++i) {
+ ceph_options.push_back(i->c_str());
+ }
+
+ global_init(
+ &def_args, ceph_options, CEPH_ENTITY_TYPE_CLIENT,
+ CODE_ENVIRONMENT_UTILITY,
+ CINIT_FLAG_NO_DEFAULT_CONFIG_FILE);
+ common_init_finish(g_ceph_context);
+ g_ceph_context->_conf->apply_changes(NULL);
+
+ if (vm.count("help")) {
+ cout << desc << std::endl;
+ return 1;
+ }
+
+ stripe_width = vm["stripe-width"].as<int>();
+ plugin = vm["plugin"].as<string>();
+ base = vm["base"].as<string>();
+ check = vm.count("check") > 0;
+ create = vm.count("create") > 0;
+
+ if (!check && !create) {
+ cerr << "must specifify either --check or --create" << endl;
+ return 1;
+ }
+
+ {
+ stringstream path;
+ path << base << "/" << "plugin=" << plugin << " stipe-width=" << stripe_width;
+ directory = path.str();
+ }
+
+ if (vm.count("parameter")) {
+ const vector<string> &p = vm["parameter"].as< vector<string> >();
+ for (vector<string>::const_iterator i = p.begin();
+ i != p.end();
+ ++i) {
+ std::vector<std::string> strs;
+ boost::split(strs, *i, boost::is_any_of("="));
+ if (strs.size() != 2) {
+ cerr << "--parameter " << *i << " ignored because it does not contain exactly one =" << endl;
+ } else {
+ parameters[strs[0]] = strs[1];
+ }
+ if (strs[0] != "directory")
+ directory += " " + *i;
+ }
+ }
+ if (parameters.count("directory") == 0)
+ parameters["directory"] = ".libs";
+
+ return 0;
+}
+
+int ErasureCodeNonRegression::run()
+ {
+ int ret = 0;
+ if(create && (ret = run_create()))
+ return ret;
+ if(check && (ret = run_check()))
+ return ret;
+ return ret;
+}
+
+int ErasureCodeNonRegression::run_create()
+{
+ ErasureCodePluginRegistry &instance = ErasureCodePluginRegistry::instance();
+ ErasureCodeInterfaceRef erasure_code;
+ stringstream messages;
+ int code = instance.factory(plugin, parameters, &erasure_code, messages);
+ if (code) {
+ cerr << messages.str() << endl;
+ return code;
+ }
+
+ if (::mkdir(directory.c_str(), 0755)) {
+ cerr << "mkdir(" << directory << "): " << cpp_strerror(errno) << endl;
+ return 1;
+ }
+ unsigned payload_chunk_size = 37;
+ string payload;
+ for (unsigned j = 0; j < payload_chunk_size; ++j)
+ payload.push_back('a' + (rand() % 26));
+ bufferlist in;
+ for (unsigned j = 0; j < stripe_width; j += payload_chunk_size)
+ in.append(payload);
+ if (stripe_width < in.length())
+ in.splice(stripe_width, in.length() - stripe_width);
+ if (in.write_file(content_path().c_str()))
+ return 1;
+ set<int> want_to_encode;
+ for (unsigned int i = 0; i < erasure_code->get_chunk_count(); i++) {
+ want_to_encode.insert(i);
+ }
+ map<int,bufferlist> encoded;
+ code = erasure_code->encode(want_to_encode, in, &encoded);
+ if (code)
+ return code;
+ for (map<int,bufferlist>::iterator chunk = encoded.begin();
+ chunk != encoded.end();
+ chunk++) {
+ if (chunk->second.write_file(chunk_path(chunk->first).c_str()))
+ return 1;
+ }
+ return 0;
+}
+
+int ErasureCodeNonRegression::decode_erasures(ErasureCodeInterfaceRef erasure_code,
+ set<int> erasures,
+ map<int,bufferlist> chunks)
+{
+ map<int,bufferlist> available;
+ for (map<int,bufferlist>::iterator chunk = chunks.begin();
+ chunk != chunks.end();
+ ++chunk) {
+ if (erasures.count(chunk->first) == 0)
+ available[chunk->first] = chunk->second;
+
+ }
+ map<int,bufferlist> decoded;
+ int code = erasure_code->decode(erasures, available, &decoded);
+ if (code)
+ return code;
+ for (set<int>::iterator erasure = erasures.begin();
+ erasure != erasures.end();
+ ++erasure) {
+ if (!chunks[*erasure].contents_equal(decoded[*erasure])) {
+ cerr << "chunk " << *erasure << " incorrectly recovered" << endl;
+ return 1;
+ }
+ }
+ return 0;
+}
+
+int ErasureCodeNonRegression::run_check()
+{
+ ErasureCodePluginRegistry &instance = ErasureCodePluginRegistry::instance();
+ ErasureCodeInterfaceRef erasure_code;
+ stringstream messages;
+ int code = instance.factory(plugin, parameters, &erasure_code, messages);
+ if (code) {
+ cerr << messages.str() << endl;
+ return code;
+ }
+ string errors;
+ bufferlist in;
+ if (in.read_file(content_path().c_str(), &errors)) {
+ cerr << errors << endl;
+ return 1;
+ }
+ set<int> want_to_encode;
+ for (unsigned int i = 0; i < erasure_code->get_chunk_count(); i++) {
+ want_to_encode.insert(i);
+ }
+
+ map<int,bufferlist> encoded;
+ code = erasure_code->encode(want_to_encode, in, &encoded);
+ if (code)
+ return code;
+
+ for (map<int,bufferlist>::iterator chunk = encoded.begin();
+ chunk != encoded.end();
+ chunk++) {
+ bufferlist existing;
+ if (existing.read_file(chunk_path(chunk->first).c_str(), &errors)) {
+ cerr << errors << endl;
+ return 1;
+ }
+ bufferlist &old = chunk->second;
+ if (existing.length() != old.length() ||
+ memcmp(existing.c_str(), old.c_str(), old.length())) {
+ cerr << "chunk " << chunk->first << " encodes differently" << endl;
+ return 1;
+ }
+ }
+
+ // erasing a single chunk is likely to use a specific code path in every plugin
+ set<int> erasures;
+ erasures.clear();
+ erasures.insert(0);
+ code = decode_erasures(erasure_code, erasures, encoded);
+ if (code)
+ return code;
+
+ if (erasure_code->get_chunk_count() - erasure_code->get_data_chunk_count() > 1) {
+ // erasing two chunks is likely to be the general case
+ erasures.clear();
+ erasures.insert(0);
+ erasures.insert(erasure_code->get_chunk_count() - 1);
+ code = decode_erasures(erasure_code, erasures, encoded);
+ if (code)
+ return code;
+ }
+
+ return 0;
+}
+
+string ErasureCodeNonRegression::content_path()
+{
+ stringstream path;
+ path << directory << "/content";
+ return path.str();
+}
+
+string ErasureCodeNonRegression::chunk_path(unsigned int chunk)
+{
+ stringstream path;
+ path << directory << "/" << chunk;
+ return path.str();
+}
+
+int main(int argc, char** argv) {
+ ErasureCodeNonRegression non_regression;
+ int err = non_regression.setup(argc, argv);
+ if (err)
+ return err;
+ return non_regression.run();
+}
+
+/*
+ * Local Variables:
+ * compile-command: "cd ../.. ; make -j4 &&
+ * make ceph_erasure_code_non_regression &&
+ * libtool --mode=execute valgrind --tool=memcheck --leak-check=full \
+ * ./ceph_erasure_code_non_regression \
+ * --plugin jerasure \
+ * --parameter directory=.libs \
+ * --parameter technique=reed_sol_van \
+ * --parameter k=2 \
+ * --parameter m=2 \
+ * --directory /tmp/ceph_erasure_code_non_regression \
+ * --stripe-width 3181 \
+ * --create \
+ * --check
+ * "
+ * End:
+ */
diff -Nru ceph-0.80.7/src/test/erasure-code/Makefile.am ceph-0.80.9/src/test/erasure-code/Makefile.am
--- ceph-0.80.7/src/test/erasure-code/Makefile.am 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/erasure-code/Makefile.am 2015-03-09 20:51:18.000000000 +0100
@@ -9,6 +9,14 @@
endif
bin_DEBUGPROGRAMS += ceph_erasure_code_benchmark
+ceph_erasure_code_non_regression_SOURCES = \
+ test/erasure-code/ceph_erasure_code_non_regression.cc
+ceph_erasure_code_non_regression_LDADD = $(LIBOSD) $(LIBCOMMON) $(BOOST_PROGRAM_OPTIONS_LIBS) $(CEPH_GLOBAL)
+if LINUX
+ceph_erasure_code_non_regression_LDADD += -ldl
+endif
+noinst_PROGRAMS += ceph_erasure_code_non_regression
+
ceph_erasure_code_SOURCES = \
test/erasure-code/ceph_erasure_code.cc
ceph_erasure_code_LDADD = $(LIBOSD) $(LIBCOMMON) $(BOOST_PROGRAM_OPTIONS_LIBS) $(CEPH_GLOBAL)
diff -Nru ceph-0.80.7/src/test/erasure-code/TestErasureCodeJerasure.cc ceph-0.80.9/src/test/erasure-code/TestErasureCodeJerasure.cc
--- ceph-0.80.7/src/test/erasure-code/TestErasureCodeJerasure.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/erasure-code/TestErasureCodeJerasure.cc 2015-03-09 20:51:18.000000000 +0100
@@ -288,36 +288,6 @@
}
}
- //
- // The ruleid may be different from the ruleset when a crush rule is
- // removed because the removed ruleid will be reused but the removed
- // ruleset will not be reused.
- //
- // This also asserts that the create_ruleset() method returns a
- // ruleset and not a ruleid http://tracker.ceph.com/issues/9044
- //
- {
- stringstream ss;
- ErasureCodeJerasureReedSolomonVandermonde jerasure;
- map<std::string,std::string> parameters;
- parameters["k"] = "2";
- parameters["m"] = "2";
- parameters["w"] = "8";
- jerasure.init(parameters);
- int FIRST = jerasure.create_ruleset("FIRST", *c, &ss);
- int SECOND = jerasure.create_ruleset("SECOND", *c, &ss);
- int FIRST_ruleid = c->get_rule_id("FIRST");
- EXPECT_EQ(0, c->remove_rule(FIRST_ruleid));
- int ruleset = jerasure.create_ruleset("myrule", *c, &ss);
- EXPECT_NE(FIRST, ruleset);
- EXPECT_NE(SECOND, ruleset);
- EXPECT_NE(ruleset, c->get_rule_id("myrule"));
- int SECOND_ruleid = c->get_rule_id("SECOND");
- EXPECT_EQ(0, c->remove_rule(SECOND_ruleid));
- int myrule_ruleid = c->get_rule_id("myrule");
- EXPECT_EQ(0, c->remove_rule(myrule_ruleid));
- }
-
{
stringstream ss;
ErasureCodeJerasureReedSolomonVandermonde jerasure;
diff -Nru ceph-0.80.7/src/test/libcephfs/test.cc ceph-0.80.9/src/test/libcephfs/test.cc
--- ceph-0.80.7/src/test/libcephfs/test.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/libcephfs/test.cc 2015-03-09 20:51:18.000000000 +0100
@@ -1138,6 +1138,7 @@
}
}
+ ceph_close(cmount, fd);
ceph_shutdown(cmount);
}
diff -Nru ceph-0.80.7/src/test/librados/misc.cc ceph-0.80.9/src/test/librados/misc.cc
--- ceph-0.80.7/src/test/librados/misc.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/librados/misc.cc 2015-03-09 20:51:18.000000000 +0100
@@ -299,7 +299,8 @@
bufferlist::iterator iter = bl.begin();
uint64_t all_features;
::decode(all_features, iter);
- ASSERT_EQ(all_features, (uint64_t)RBD_FEATURES_ALL);
+ // make sure *some* features are specified; don't care which ones
+ ASSERT_NE(all_features, 0);
}
TEST_F(LibRadosMiscPP, ExecPP) {
@@ -311,7 +312,8 @@
bufferlist::iterator iter = out.begin();
uint64_t all_features;
::decode(all_features, iter);
- ASSERT_EQ(all_features, (uint64_t)RBD_FEATURES_ALL);
+ // make sure *some* features are specified; don't care which ones
+ ASSERT_NE(all_features, 0);
}
TEST_F(LibRadosMiscPP, Operate1PP) {
diff -Nru ceph-0.80.7/src/test/librados/snapshots.cc ceph-0.80.9/src/test/librados/snapshots.cc
--- ceph-0.80.7/src/test/librados/snapshots.cc 2014-10-14 21:12:51.000000000 +0200
+++ ceph-0.80.9/src/test/librados/snapshots.cc 2015-03-09 20:51:18.000000000 +0100
@@ -145,6 +145,24 @@
EXPECT_EQ(0, ioctx.snap_remove("snapfoo"));
}
+TEST_F(LibRadosSnapshotsPP, SnapCreateRemovePP) {
+ // reproduces http://tracker.ceph.com/issues/10262
+ bufferlist bl;
+ bl.append("foo");
+ ASSERT_EQ(0, ioctx.write("foo", bl, bl.length(), 0));
+ ASSERT_EQ(0, ioctx.snap_create("snapfoo"));
+ ASSERT_EQ(0, ioctx.remove("foo"));
+ ASSERT_EQ(0, ioctx.snap_create("snapbar"));
+
+ librados::ObjectWriteOperation *op = new librados::ObjectWriteOperation();
+ op->create(false);
+ op->remove();
+ ASSERT_EQ(0, ioctx.operate("foo", op));
+
+ EXPECT_EQ(0, ioctx.snap_remove("snapfoo"));
+ EXPECT_EQ(0, ioctx.snap_remove("snapbar"));
+}
+
TEST_F(LibRadosSnapshotsSelfManaged, Snap) {
std::vector<uint64_t> my_snaps;
my_snaps.push_back(-2);
diff -Nru ceph-0.80.7/src/test/librados/tier.cc ceph-0.80.9/src/test/librados/tier.cc
--- ceph-0.80.7/src/test/librados/tier.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/librados/tier.cc 2015-03-09 20:51:18.000000000 +0100
@@ -966,6 +966,82 @@
}
}
+// this test case reproduces http://tracker.ceph.com/issues/8629
+TEST_F(LibRadosTwoPoolsPP, EvictSnap2) {
+ // create object
+ {
+ bufferlist bl;
+ bl.append("hi there");
+ ObjectWriteOperation op;
+ op.write_full(bl);
+ ASSERT_EQ(0, ioctx.operate("foo", &op));
+ }
+ // create a snapshot, clone
+ vector<uint64_t> my_snaps(1);
+ ASSERT_EQ(0, ioctx.selfmanaged_snap_create(&my_snaps[0]));
+ ASSERT_EQ(0, ioctx.selfmanaged_snap_set_write_ctx(my_snaps[0],
+ my_snaps));
+ {
+ bufferlist bl;
+ bl.append("ciao!");
+ ObjectWriteOperation op;
+ op.write_full(bl);
+ ASSERT_EQ(0, ioctx.operate("foo", &op));
+ }
+ // configure cache
+ bufferlist inbl;
+ ASSERT_EQ(0, cluster.mon_command(
+ "{\"prefix\": \"osd tier add\", \"pool\": \"" + pool_name +
+ "\", \"tierpool\": \"" + cache_pool_name +
+ "\", \"force_nonempty\": \"--force-nonempty\" }",
+ inbl, NULL, NULL));
+ ASSERT_EQ(0, cluster.mon_command(
+ "{\"prefix\": \"osd tier set-overlay\", \"pool\": \"" + pool_name +
+ "\", \"overlaypool\": \"" + cache_pool_name + "\"}",
+ inbl, NULL, NULL));
+ ASSERT_EQ(0, cluster.mon_command(
+ "{\"prefix\": \"osd tier cache-mode\", \"pool\": \"" + cache_pool_name +
+ "\", \"mode\": \"writeback\"}",
+ inbl, NULL, NULL));
+
+ // wait for maps to settle
+ cluster.wait_for_latest_osdmap();
+
+ // read, trigger a promote on the head
+ {
+ bufferlist bl;
+ ASSERT_EQ(1, ioctx.read("foo", bl, 1, 0));
+ ASSERT_EQ('c', bl[0]);
+ }
+
+ // evict
+ {
+ ObjectReadOperation op;
+ op.cache_evict();
+ librados::AioCompletion *completion = cluster.aio_create_completion();
+ ASSERT_EQ(0, cache_ioctx.aio_operate(
+ "foo", completion, &op,
+ librados::OPERATION_IGNORE_CACHE, NULL));
+ completion->wait_for_safe();
+ ASSERT_EQ(0, completion->get_return_value());
+ completion->release();
+ }
+
+ // verify the snapdir is not present in the cache pool
+ {
+ ObjectReadOperation op;
+ librados::snap_set_t snapset;
+ op.list_snaps(&snapset, NULL);
+ ioctx.snap_set_read(librados::SNAP_DIR);
+ librados::AioCompletion *completion = cluster.aio_create_completion();
+ ASSERT_EQ(0, ioctx.aio_operate("foo", completion, &op,
+ librados::OPERATION_IGNORE_CACHE, NULL));
+ completion->wait_for_safe();
+ ASSERT_EQ(-ENOENT, completion->get_return_value());
+ completion->release();
+ }
+}
+
TEST_F(LibRadosTwoPoolsPP, TryFlush) {
// configure cache
bufferlist inbl;
diff -Nru ceph-0.80.7/src/test/librbd/test_librbd.cc ceph-0.80.9/src/test/librbd/test_librbd.cc
--- ceph-0.80.7/src/test/librbd/test_librbd.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/librbd/test_librbd.cc 2015-03-09 20:51:18.000000000 +0100
@@ -21,6 +21,7 @@
#include "global/global_context.h"
#include "global/global_init.h"
#include "common/ceph_argparse.h"
+#include "common/config.h"
#include "gtest/gtest.h"
@@ -40,6 +41,8 @@
#include "include/interval_set.h"
#include "include/stringify.h"
+#include <boost/scope_exit.hpp>
+
using namespace std;
static int get_features(bool *old_format, uint64_t *features)
@@ -67,6 +70,8 @@
{
if (old_format) {
return rbd_create(ioctx, name, size, order);
+ } else if ((features & RBD_FEATURE_STRIPINGV2) != 0) {
+ return rbd_create3(ioctx, name, size, features, order, 65536, 16);
} else {
return rbd_create2(ioctx, name, size, features, order);
}
@@ -1857,6 +1862,107 @@
rados_ioctx_destroy(ioctx);
ASSERT_EQ(0, destroy_one_pool(pool_name, &cluster));
+}
+
+TEST(LibRBD, LargeCacheRead)
+{
+ if (!g_conf->rbd_cache) {
+ std::cout << "SKIPPING due to disabled cache" << std::endl;
+ return;
+ }
+
+ rados_t cluster;
+ rados_ioctx_t ioctx;
+ string pool_name = get_temp_pool_name();
+ ASSERT_EQ("", create_one_pool(pool_name, &cluster));
+ rados_ioctx_create(cluster, pool_name.c_str(), &ioctx);
+
+ uint64_t orig_cache_size = g_conf->rbd_cache_size;
+ g_conf->set_val("rbd_cache_size", "16777216");
+ BOOST_SCOPE_EXIT( (orig_cache_size) ) {
+ g_conf->set_val("rbd_cache_size", stringify(orig_cache_size).c_str());
+ } BOOST_SCOPE_EXIT_END;
+ ASSERT_EQ(16777216, g_conf->rbd_cache_size);
+
+ rbd_image_t image;
+ int order = 0;
+ const char *name = "testimg";
+ uint64_t size = g_conf->rbd_cache_size + 1;
+
+ ASSERT_EQ(0, create_image(ioctx, name, size, &order));
+ ASSERT_EQ(0, rbd_open(ioctx, name, &image, NULL));
+
+ std::string buffer(1 << order, '1');
+ for (size_t offs = 0; offs < size; offs += buffer.size()) {
+ size_t len = std::min<uint64_t>(buffer.size(), size - offs);
+ ASSERT_EQ(static_cast<ssize_t>(len),
+ rbd_write(image, offs, len, buffer.c_str()));
+ }
+
+ ASSERT_EQ(0, rbd_invalidate_cache(image));
+
+ buffer.resize(size);
+ ASSERT_EQ(static_cast<ssize_t>(size-1024), rbd_read(image, 1024, size, &buffer[0]));
+
+ ASSERT_EQ(0, rbd_close(image));
+
+ rados_ioctx_destroy(ioctx);
+ ASSERT_EQ(0, destroy_one_pool(pool_name, &cluster));
+}
+
+TEST(LibRBD, TestPendingAio)
+{
+ rados_t cluster;
+ rados_ioctx_t ioctx;
+ string pool_name = get_temp_pool_name();
+ ASSERT_EQ("", create_one_pool(pool_name, &cluster));
+ rados_ioctx_create(cluster, pool_name.c_str(), &ioctx);
+
+ int features = RBD_FEATURE_LAYERING;
+ rbd_image_t image;
+ int order = 0;
+
+ std::string name = "testimg";
+
+ uint64_t size = 4 << 20;
+ ASSERT_EQ(0, create_image_full(ioctx, name.c_str(), size, &order,
+ false, features));
+ ASSERT_EQ(0, rbd_open(ioctx, name.c_str(), &image, NULL));
+
+ char test_data[TEST_IO_SIZE];
+ for (size_t i = 0; i < TEST_IO_SIZE; ++i) {
+ test_data[i] = (char) (rand() % (126 - 33) + 33);
+ }
+
+ size_t num_aios = 256;
+ rbd_completion_t comps[num_aios];
+ for (size_t i = 0; i < num_aios; ++i) {
+ ASSERT_EQ(0, rbd_aio_create_completion(NULL, NULL, &comps[i]));
+ uint64_t offset = rand() % (size - TEST_IO_SIZE);
+ ASSERT_EQ(0, rbd_aio_write(image, offset, TEST_IO_SIZE, test_data,
+ comps[i]));
+ }
+ for (size_t i = 0; i < num_aios; ++i) {
+ ASSERT_EQ(0, rbd_aio_wait_for_complete(comps[i]));
+ rbd_aio_release(comps[i]);
+ }
+ ASSERT_EQ(0, rbd_invalidate_cache(image));
+
+ for (size_t i = 0; i < num_aios; ++i) {
+ ASSERT_EQ(0, rbd_aio_create_completion(NULL, NULL, &comps[i]));
+ uint64_t offset = rand() % (size - TEST_IO_SIZE);
+ ASSERT_LE(0, rbd_aio_read(image, offset, TEST_IO_SIZE, test_data,
+ comps[i]));
+ }
+
+ ASSERT_EQ(0, rbd_close(image));
+ for (size_t i = 0; i < num_aios; ++i) {
+ ASSERT_EQ(1, rbd_aio_is_complete(comps[i]));
+ rbd_aio_release(comps[i]);
+ }
+
+ rados_ioctx_destroy(ioctx);
+ ASSERT_EQ(0, destroy_one_pool(pool_name, &cluster));
}
int main(int argc, char **argv)
diff -Nru ceph-0.80.7/src/test/Makefile.am ceph-0.80.9/src/test/Makefile.am
--- ceph-0.80.7/src/test/Makefile.am 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/Makefile.am 2015-03-09 20:51:18.000000000 +0100
@@ -368,6 +368,9 @@
unittest_pglog_CXXFLAGS = $(UNITTEST_CXXFLAGS)
unittest_pglog_LDADD = $(LIBOSD) $(UNITTEST_LDADD) $(CEPH_GLOBAL)
check_PROGRAMS += unittest_pglog
+if LINUX
+unittest_pglog_LDADD += -ldl
+endif # LINUX
unittest_ecbackend_SOURCES = test/osd/TestECBackend.cc
unittest_ecbackend_CXXFLAGS = $(UNITTEST_CXXFLAGS)
@@ -379,9 +382,10 @@
unittest_hitset_LDADD = $(LIBOSD) $(UNITTEST_LDADD) $(CEPH_GLOBAL)
check_PROGRAMS += unittest_hitset
-if LINUX
-unittest_pglog_LDADD += -ldl
-endif # LINUX
+unittest_io_priority_SOURCES = test/common/test_io_priority.cc
+unittest_io_priority_CXXFLAGS = $(UNITTEST_CXXFLAGS)
+unittest_io_priority_LDADD = $(UNITTEST_LDADD) $(CEPH_GLOBAL)
+check_PROGRAMS += unittest_io_priority
unittest_gather_SOURCES = test/gather.cc
unittest_gather_LDADD = $(UNITTEST_LDADD) $(CEPH_GLOBAL)
diff -Nru ceph-0.80.7/src/test/mon/mon-test-helpers.sh ceph-0.80.9/src/test/mon/mon-test-helpers.sh
--- ceph-0.80.7/src/test/mon/mon-test-helpers.sh 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/test/mon/mon-test-helpers.sh 2015-03-09 20:51:18.000000000 +0100
@@ -59,8 +59,9 @@
function kill_daemons() {
local dir=$1
for pidfile in $(find $dir | grep pidfile) ; do
+ pid=$(cat $pidfile)
for try in 0 1 1 1 2 3 ; do
- kill -9 $(cat $pidfile 2> /dev/null) 2> /dev/null || break
+ kill -9 $pid 2> /dev/null || break
sleep $try
done
done
diff -Nru ceph-0.80.7/src/tools/crushtool.cc ceph-0.80.9/src/tools/crushtool.cc
--- ceph-0.80.7/src/tools/crushtool.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/tools/crushtool.cc 2015-03-09 20:51:18.000000000 +0100
@@ -118,6 +118,7 @@
cout << " --show utilization-all\n";
cout << " include zero weight items\n";
cout << " --show-statistics show chi squared statistics\n";
+ cout << " --show-mappings show mappings\n";
cout << " --show-bad-mappings show bad mappings\n";
cout << " --show-choose-tries show choose tries histogram\n";
cout << " --set-choose-local-tries N\n";
@@ -190,6 +191,7 @@
int choose_total_tries = -1;
int chooseleaf_descend_once = -1;
int chooseleaf_vary_r = -1;
+ int straw_calc_version = -1;
CrushWrapper crush;
@@ -233,6 +235,9 @@
} else if (ceph_argparse_flag(args, i, "--show_statistics", (char*)NULL)) {
display = true;
tester.set_output_statistics(true);
+ } else if (ceph_argparse_flag(args, i, "--show_mappings", (char*)NULL)) {
+ display = true;
+ tester.set_output_mappings(true);
} else if (ceph_argparse_flag(args, i, "--show_bad_mappings", (char*)NULL)) {
display = true;
tester.set_output_bad_mappings(true);
@@ -263,6 +268,9 @@
} else if (ceph_argparse_withint(args, i, &chooseleaf_vary_r, &err,
"--set_chooseleaf_vary_r", (char*)NULL)) {
adjust = true;
+ } else if (ceph_argparse_withint(args, i, &straw_calc_version, &err,
+ "--set_straw_calc_version", (char*)NULL)) {
+ adjust = true;
} else if (ceph_argparse_flag(args, i, "--reweight", (char*)NULL)) {
reweight = true;
} else if (ceph_argparse_withint(args, i, &add_item, &err, "--add_item", (char*)NULL)) {
@@ -581,10 +589,8 @@
dout(2) << " item " << items[j] << " weight " << weights[j] << dendl;
}
- crush_bucket *b = crush_make_bucket(buckettype, CRUSH_HASH_DEFAULT, type, j, items, weights);
- assert(b);
int id;
- int r = crush_add_bucket(crush.crush, 0, b, &id);
+ int r = crush.add_bucket(0, buckettype, CRUSH_HASH_DEFAULT, type, j, items, weights, &id);
if (r < 0) {
dout(2) << "Couldn't add bucket: " << cpp_strerror(r) << dendl;
}
@@ -712,6 +718,10 @@
crush.set_chooseleaf_vary_r(chooseleaf_vary_r);
modified = true;
}
+ if (straw_calc_version >= 0) {
+ crush.set_straw_calc_version(straw_calc_version);
+ modified = true;
+ }
if (modified) {
crush.finalize();
diff -Nru ceph-0.80.7/src/tools/osdmaptool.cc ceph-0.80.9/src/tools/osdmaptool.cc
--- ceph-0.80.7/src/tools/osdmaptool.cc 2014-10-14 21:12:54.000000000 +0200
+++ ceph-0.80.9/src/tools/osdmaptool.cc 2015-03-09 20:51:18.000000000 +0100
@@ -35,6 +35,7 @@
cout << " --export-crush <file> write osdmap's crush map to <file>" << std::endl;
cout << " --import-crush <file> replace osdmap's crush map with <file>" << std::endl;
cout << " --test-map-pgs [--pool <poolid>] map all pgs" << std::endl;
+ cout << " --test-map-pgs-dump [--pool <poolid>] map all pgs" << std::endl;
cout << " --mark-up-in mark osds up and in (but do not persist)" << std::endl;
cout << " --clear-temp clear pg_temp and primary_temp" << std::endl;
cout << " --test-random do random placements" << std::endl;
@@ -75,6 +76,7 @@
bool mark_up_in = false;
bool clear_temp = false;
bool test_map_pgs = false;
+ bool test_map_pgs_dump = false;
bool test_random = false;
std::string val;
@@ -104,6 +106,8 @@
clear_temp = true;
} else if (ceph_argparse_flag(args, i, "--test-map-pgs", (char*)NULL)) {
test_map_pgs = true;
+ } else if (ceph_argparse_flag(args, i, "--test-map-pgs-dump", (char*)NULL)) {
+ test_map_pgs_dump = true;
} else if (ceph_argparse_flag(args, i, "--test-random", (char*)NULL)) {
test_random = true;
} else if (ceph_argparse_flag(args, i, "--clobber", (char*)NULL)) {
@@ -313,7 +317,7 @@
<< ") acting (" << acting << ", p" << acting_primary << ")"
<< std::endl;
}
- if (test_map_pgs) {
+ if (test_map_pgs || test_map_pgs_dump) {
if (pool != -1 && !osdmap.have_pg_pool(pool)) {
cerr << "There is no pool " << pool << std::endl;
exit(1);
@@ -348,6 +352,9 @@
}
size[osds.size()]++;
+ if (test_map_pgs_dump)
+ cout << pgid << "\t" << osds << "\t" << primary << std::endl;
+
for (unsigned i=0; i<osds.size(); i++) {
//cout << " rep " << i << " on " << osds[i] << std::endl;
count[osds[i]]++;
@@ -452,7 +459,7 @@
if (!print && !print_json && !tree && !modified &&
export_crush.empty() && import_crush.empty() &&
test_map_pg.empty() && test_map_object.empty() &&
- !test_map_pgs) {
+ !test_map_pgs && !test_map_pgs_dump) {
cerr << me << ": no action specified?" << std::endl;
usage();
}
Reply to: