Hello! Am Dienstag, 27. Januar 2015, 15:43:01 schrieb chrysn: > Package: src:linux > Version: 3.16.7-ckt2-1 > Severity: normal > > under conditions i can not narrow down further[1], all processes that > try to access a given mounted btrfs file system freeze. this affects > even processes like `ps u` called by a user not at home in the affected > file system inside the first read to an opened /proc/8119/cmdline file. I suggest you to report this upstream. There are at least two known BTRFS hang issues even with 3.19 kernel being reported on the BTRFS upstream mailing list at vger.kernel.org. See this thread: 3.19-rc5: Bug 91911: [REGRESSION] rm command hangs big time with deleting a lot of files at once (and bugzilla.kernel.org bug with that number) 3.15 and 3.16 had another hang issue that should be fixed with 3.17 and hopefully also with a 3.16 update see: fix for the infamous deadlock [1] https://btrfs.wiki.kernel.org/index.php/Changelog#v3.17_.28Oct_2014.29 [1] https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=9e0af23764344f7f1b68e4eefbe7dc865018b63d (I think there are some other related commits fixing hanging along compressed write, like with compress=lzo mount option, I think they may be already in the 3.16 kernel you are using from the package.) Also a general recommendation for me, leave enough free space. Ideally so much that BTRFS can still allocate new chunks from unreserved space, see: [Bug 90401] New: btrfs kworker thread uses up 100% of a Sandybridge core for minutes on random write into big file I am not sure whether any of the mentioned bugs are similar to what you see as I didn´t compare the backtraces. If the backtrace looks different than any of those in the bug reports and related mail threads, I strongly recommend that you report a bug upstream with bugzilla.kernel.org I think it can help the debian kernel team tremendously to be able to cherry-pick an upstream patch, but for that one needs to be written or exist already :). I am not a member of the Debian kernel team, I just monitor the debian-kernel mailinglist and use BTRFS since several years already on an increasing amount of systems. That just as a few pointers that you may use to find your way to move along with the bug you have. If you choose to report upstream, do some echo "t" > /proc/sysrq-trigger and cut attach the correspondending output of /var/log/kern.log to the upstream bug report. Ciao,, -- Martin 'Helios' Steigerwald - http://www.Lichtvoll.de GPG: 03B0 0D6C 0040 0710 4AFA B82F 991B EAAC A599 84C7
Attachment:
signature.asc
Description: This is a digitally signed message part.