Red Hat Bugzilla – Bug 122966
logout+restart cycles power to IDE disks unnecessarily
Last modified: 2007-11-30 17:10:42 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040505
Description of problem:
RedHat > Logout > Restart now causes IDE disk drives to be powered
down, and the reboot causes them to be powered back up. This causes
unnecessary wear and tear on the drives.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Running .358 kernel with Gnome desktop: RedHat > Logout > Restart
Actual Results: On the text console:
Shutting down: hdb ## (E)IDE harddrive
Shutting down: hda ## (E)IDE harddrive
and each drive powers down in separately (heads retract, spindle
stops). Non-harddrives hdc and hdd get no message and no apparent action.
Expected Results: sync and flush cache only; no power cycle.
The old message (from .349 kernel) was
Shutting down: hda hdb hdc hdd
but the harddrives kept spinning. hdc is a DVD-ROM, hdd is a CD-RW.
Not only wear and tear. In my case, the BIOS often fails to bring the
disks back up, such that I can't reliably reboot the boxes. I've
experienced such problems on boxes with A7V133 MoBos, one of them with
one HD on the VIA IDE controller and one HD on the Promise controller
(both ATA100); the other box with far many more disks, including a few
Sil0680 ATA133 controllers. The reboot is slow, bcause the disks are
turned on one after the other, and it takes several seconds for each,
and quite often the BIOS times out or something like that waiting for
a disk to come back up and won't complete the boot, so manual
interaction is required (Ctrl-Alt-Del). The 2.4 code only forced this
flush on shutdowns, not on reboots. Can't we tell the difference and
proceed accordingly? Sure not flushing on reboot leaves a window of
opportunity for disk corruption due to power failure after the kernel
reboots and before the disks complete the flush, but this opportunity
already exists in case of power failure before or while the kernel
forces the flush, so it's not such a big deal IMHO.
*** Bug 123608 has been marked as a duplicate of this bug. ***
This is a bug in the initscripts package and affects all platforms.
It's actually the kernel reboot/shutdown code that does this, totally
by itself. initscripts have nothing to do with it.
Rats, clicked on commit to soon.
I also meant to point out that the problem appears to be that the
kernel had no way to distinguish between the shutdown and the reboot
cases. I saw some patches in lkml floating around that were meant to
fix this, but I think they were post-2.6.6. Maybe we could merge them
into an update kernel that might be issued for FC2 at some point in
I filed a kernel bug report at http://bugme.osdl.org/show_bug.cgi?id=2732
I'd like to comment that this bug only manifests itself with ATA
drives, on my system the hard drives on my SCSI interface do not have
*** This bug has been marked as a duplicate of 123873 ***
*** Bug 123873 has been marked as a duplicate of this bug. ***
Was that actually neccessary to change what was a dup of what
Alexandre. All the important information is now in Bug 123873 which
isnt the first one people will find.
This was the one with a link to the osdl.org bug submission, that has
a link back to this one, so I thought it would be better this way. We
don't want the link to the upstream bug report buried in one of a
dozen dups that this one might get.
Seems like there's a patch available:
ChangeSet@1.1726, 2004-05-20 22:22:11-07:00, email@example.com
[PATCH] ide-disk.c: don't put disks in STANDBY mode on reboot
From: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>
From: Rene Herman <firstname.lastname@example.org>
Prevent the disks from spinning down across a reboot.
verified, works for me. I guess we can close this bug.
has a new kernel been issued to the development rpms yet?
Yes, fc-devel has new kernels (.422 last time I checked)
*** Bug 125487 has been marked as a duplicate of this bug. ***
Does the new kernel in updates, kernel-2.6.6-1.427, fix this issue? If
it does shouldn't this bug be marked as resolved, and not Rawhide?
> Does the new kernel in updates, kernel-2.6.6-1.427, fix this issue?
> If it does shouldn't this bug be marked as resolved, and not Rawhide?
This bug *is* closed. (AFAIK the Red Hat Bugzilla setup does not make
a distinction between CLOSED and RESOLVED.) And RAWHIDE is the correct
resolution -- this bug was resolved by an update in rawhide, and the
fact that the update has now been released doesn't change that.
(Perhaps my explanation is not correct, but the answer still is --
CLOSED/RAWHIDE is the correct state for this bug to be in.)