reboot: noisy if /proc/ide doesn't exist
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
upstart |
Fix Released
|
Wishlist
|
Scott James Remnant (Canonical) | ||
upstart (Ubuntu) |
Fix Released
|
Wishlist
|
Unassigned | ||
Bug Description
Binary package hint: upstart
On a system with no IDE controller and thus no /proc/ide, I get this on shutdown:
halt: Unable to iterate IDE devices: No such file or directory
/etc/init.d/halt always uses 'halt -h' unless you're using RAID. Maybe it should avoid that if /proc/ide is missing? Or maybe this code should just be removed from upstart:
22:52 <cjwatson> ps3 doesn't seem to have an IDE controller; is upstart's warning message about that too loud maybe?
23:04 <Keybuk> err?
23:04 <Keybuk> upstart shouldn't have a warning about that
23:04 <Keybuk> what warning are you seeing?
23:04 <Keybuk> (why would upstart need an IDE controller?)
23:05 <cjwatson> "halt: Unable to iterate IDE devices: No such file or directory"
23:06 <Keybuk> oh
23:06 <Keybuk> heh
23:06 <Keybuk> interesting question of style that one
23:07 <Keybuk> since you did call halt with -h ("put disks into standby"), it seems potentially wrong to not warn that it didn't work
23:07 <Keybuk> (as apposed to default behaviour)
23:07 <Keybuk> doesn't sysvinit do the same?
23:07 <cjwatson> I have no idea, never run sysvinit on this machine
23:08 <Keybuk> I'm determined that that option is entirely pointless
23:08 <Keybuk> but too scared to find out :p
23:08 <Keybuk> both downing drives and interfaces should be things the kernel does these days
23:09 <Keybuk> file a bug plz :p
23:09 <cjwatson> maybe /etc/init.d/halt should check whether /proc/ide exists before using -h
23:10 <cjwatson> sysvinit's halt man page says that the kernel doesn't flush the write cache itself before poweroff
23:10 <cjwatson> and that that's the side-effectual purpose of putting the drive in standby
23:12 <Keybuk> kernel source disagrees
23:12 <Keybuk> ide_device_shutdown calls ide_cacheflush_p()
23:12 <cjwatson> yeah, could be a more recent change
23:13 <Keybuk> the sysvinit halt manpage is what, 5 years old?
23:13 <cjwatson> dated Nov 6, 2001
23:17 <Keybuk> I'll get Ben to tell me for sure whether that's needed anymore, I'm not 100% sure, or fix the warning on missing /proc/ide
Changed in upstart: | |
importance: | Undecided → Wishlist |
status: | Unconfirmed → Confirmed |
Changed in upstart: | |
importance: | Undecided → Wishlist |
status: | Unconfirmed → Confirmed |
Changed in upstart: | |
status: | Confirmed → Triaged |
Changed in upstart: | |
milestone: | 0.5 → none |
Changed in upstart: | |
status: | Invalid → Triaged |
Changed in upstart: | |
milestone: | none → 0.5-later |
Changed in upstart: | |
milestone: | 0.5-later → none |
summary: |
- noisy halt if /proc/ide doesn't exist + reboot: noisy if /proc/ide doesn't exist |
Changed in upstart: | |
status: | Triaged → Fix Committed |
Changed in upstart: | |
milestone: | none → 0.6.0 |
Rejecting Ubuntu task since this is an upstream problem.
Interestingly we've read through the kernel source, and are now convinced that the IDE subsystem does flush the caches and power down the disks on halt.
And we've had a long conversation with the libata maintainers, who have made the same changes to libata so it too flushes caches and powers down the disks.
So reboot needs to do none of this.