[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Heads up on initramfs-tools dev


  The size of the generated initramfs of initramfs-tools
  in the case of MODULES=dep improved since 0.90a version.
  i'd like to get more tester feedback on that setting.
  -> /etc/initramfs-tools/initramfs.conf
  The ide subsys has a funky /sys usage where the sysfs
  modules string does not match the modules name. There the
  sysfs walk is presumed to fail (PIIX_IDE != piix).
  This will go away soon after the migration to pata.

  For initramfs debugging see


  This setting boots fine since 0.90a and klibc 1.5.6
  on /dev/sdaX root (UUID, LABEL usage included).
  0.91 moves busybox from Depends to Recommends.
  allowing package enforced BUSYBOX=n
  keep in mind that in the case of a boot failure
  your recovery tools in that situation will be
  very limited to echo, cat, dmesg, uname, fstype
  (see ls /usr/lib/klibc/bin/ ).

  I'd ask the d-i folks to queue busybox before i-t
  in the base install as it is needed for any but
  unusual installations. also i'll send patches out
  to different initramfs boot script to enforce
  BUSYBOX=y as they might rely on busybox tools
  (sed, awk, grep, tr, ..).

  Next steps on a real small initramfs for embedded
  usage is udev-klibc (compiles and boots fine) and
  a m-i-t against klibc.

  klibc got a stdio branch so if others want to play^W
  compile against klibc that might be the better
  start to hack on (lvm2, mdadm or even busybox comes
  to the mind) just install libklic-dev and use
  make CC=klcc

  hpa is active, patches are quickly reviewed and
  merged upstream.
  git clone git://git.kernel.org/pub/scm/libs/klibc/klibc.git
  Mailing list: http://www.zytor.com/mailman/listinfo/klibc

  The klibc linux-libc-dev build-dep change went well
  (quickly fixed mips+ia64), but gcc-4.2 seems to hit
  on sparc.  that will need attention by sparc porters
  once the kernel sparc issues are solved. #440721

initramfs races

  Push usage of scsi_wait_scan upstream (thanks a lot
  willy for the work). already implemented in qlogic,
  aic94xx, lpfc and fusion. The udev boot script just
  needs to modprobe it. Patches for ata aka sata and
  pata are in preparation.

  The suse bootscripts wait up to some seconds for
  /dev/mapper/control to appear. that seems sensible
  and easy to do in the debian lvm2 boot script too.


  Once dkt [1] is in place linux-images will ship
  parts of the modules prebuild as initramfs in order
  to reduce build time of the default initramfs.
  Also grub2 allows multiple initramfs files usage.

  The dkt will also allow to remove the corresponding
  initramfs sha1sum's in /var/lib/initramfs-tools on
  linux-image removal.

git repo
  initramfs repository moved some time ago to git [2]:
  git clone git://git.debian.org/git/kernel/initramfs-tools.git

  A special thanks goes to mika for test coverage [3]
  on grml.

happy hacking


[1] http://multibuild.org/documentation/dkt
[2] http://www.itp.tuwien.ac.at/~mattems/blog/2007/04/10#bzr_to_git
[3] http://www.mail-archive.com/grml@mur.at/msg01388.html

Reply to: