On Sat, Dec 26, 2009 at 9:21 PM, Rafael J. Wysocki <rjw@xxxxxxx> wrote: > On Saturday 26 December 2009, Pedro Ribeiro wrote: >> On Sat, Dec 26, 2009 at 1:26 AM, Tejun Heo <tj@xxxxxxxxxx> wrote: >> > (cc'ing Rafael) >> > >> > On 12/26/2009 03:10 AM, Pedro Ribeiro wrote: >> >>> Oh... Can you please try this one then? >> >> >> >> Sorry, still hangs with this patch :( >> >> >> >> This only happens in hibernation... suspend is fine. >> > >> > Hmmm... the patch does apply to both suspend and hibernation. Rafael, >> > Pedro's ThinkPad T400 burns cpu cycle for 30secs on "Atomic >> > copy/restore" step of hibernation if the ultrabay is powered off. The >> > original report can be found at... >> > >> > http://thread.gmane.org/gmane.linux.ide/44190 >> > >> > Please note that it's with TuxOnIce patches. Anyways, ata_piix on >> > certain configurations have similar issues where after the OS is done >> > suspending and powering off the piix controller, the BIOS tries to >> > access it and ends up burning cpu cycles which can be worked around by >> > leaving the controller on after suspend. Applying the same workaround >> > didn't resolve the issue. Do you know what can make cpu burn for >> > 30secs on the atomic copy/restore step? > > No idea. > > Is this reproducible with /sys/power/pm_test = core? > > [If you echo "core" to /sys/power/pm_test and then attempt to hibernate, > it should just simulate snapshotting the system, sleep for 5 seconds and go > back to your command prompt (or wherever you started the "hibernation").] > >> > Pedro, can you please try to reproduce the problem without TuxOnIce >> > patches? >> > >> > Thanks. >> > >> > -- >> > tejun >> > >> >> Hi Rafael and Tejun, >> >> I am experiencing the same delay with the kernel swsusp, so this is >> definitely not a TuxOnIce issue. > > I don't think this can be a TuxOnIce. Most likely, ACPI is trying to do > something stupid in _PTS. > > Rafael > Hi, The issue also occurs with pm_test. Attached are two files, the one ending with "ok" shows a good pm_test cycle, with the ultrabay powered on. The one ending in "bad" shows the bad pm_test cycle with a 30 second delay between: [ 545.764159] ACPI: Waking up from system sleep state S4 [ 575.844179] ACPI: \_SB_.PCI0.SATA.PRT1 - docking (lines 48/49) Please let me know how I can assist you. Thanks for everything, Pedro
Attachment:
dmesg_pmTest_ok
Description: Binary data
Attachment:
dmesg_pmTest_bad
Description: Binary data
