[Pkg-sysvinit-devel] Bug#426224: shutdown patches - the choices

[Kel Modderman]

Package: sysvinit
Version: 2.86.ds1-38.1

Hi,

I'm following up on this ticket, since I've not received any response for the
best part of 6 months.

The patch I proposed here is, IMHO, the most complete of the ones offered
so far. Recently, I was made aware of someone also claiming [0] the patch [1]
from Werner Fink, discussed [2] on the novell bug tracker, would be great to
include into our own Debian version of sysvinit to fix the shutdown issue of
spinning down disks correctly (...or not).

However, r1067 [3] was committed to pkg-sysvinit trunk, which merges the
changes done in the "libata-fixes" branch which was for testing [4]. Tejun Heo,
an active libata/kernel hacker, replied [5] with information that Werner Fink
was also working on this issue. Tejun was also active in the development of
the patch as seen on the novell bug discussion [2].

I emailed the list [6] with discussion requested at [4], created a patch for
the debian package [7], and tested the patch thoroughly on a very wide variety
of hardware from debian installed and live-cd environments over the last 6
months. The patch works as advertised.

opensuse 10.3 also contains a sysvinit using the mentioned patch. This means
it has seen widespread testing from a variety of people on a variety of
hardware.

The patch matches the criteria set out at [8] about how hddown.c could best
handle this (except using /proc):

[quote]
	a) it uses sysfs and not /proc to locate disks, and that it locates
	   all IDE and SCSI/libata disks, or maybe do both /proc and sysfs.
	b) that it verifies the state of kernel disk spindown control for
	   each disk, and for the disks where it is unavailable:
	   b1) issue cache sync to disk
	   b2) issue spin down to disk
	c) for the disks where kernel spin down control is available, enable
	   it and skip to next disk.
[/quote]

The patch does not suffer from limitation of an alternate patch offered to
fix the situation that is discussed at [9], which seems to be similar in
limitation to what has been merged [3].

Ok, there are too many damn numbers listed below with references. But you get
the picture I hope, that I'm concerned about the direction the maintainers [10]
intend to take in order to fix this issue, and that I've laid out as much
information, patch and help that I can to try and convince them there is a
better solution.

Thanks, Kel.

[0] http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/2007-December/002141.html
[1] https://bugzilla.novell.com/attachment.cgi?id=145904
[2] https://bugzilla.novell.com/show_bug.cgi?id=229210
[3] http://lists.alioth.debian.org/pipermail/pkg-sysvinit-commits/2007-November/000955.html
[4] http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/2007-June/001961.html
[5] http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/2007-June/001963.html
[6] http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/2007-June/001972.html
[7] http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/2007-June/001974.html
[8] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=426224#10
[9] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=436703
[10] http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/2007-December/002142.html