Bug#642409: linux-2.6: "suspend" from gnome shutdown fails when krb5/nfs4 user logged in
On Mon, 26 Sep 2011 14:31:32 -0400
Jeff Layton <jlayton@redhat.com> wrote:
> On Mon, 26 Sep 2011 17:52:58 +0200
> John Hughes <john@Calva.COM> wrote:
>
> > On 23/09/11 17:09, Jeff Layton wrote:
> > > I went ahead and attached the latest one to the RHBZ above. If you can
> > > grab it from there and test it, that would be great.
> > >
> >
> > Ok, I'm now testing the current patch on my Dell GX270 (Pentium-4, ugh)
> > system.
> >
> > Things are not working too well.
> >
> > Sometimes the system suspends with no problems, but sometimes it seems
> > to suspend and wake up instantly:
> >
> > Sep 26 17:27:06 bovic kernel: [ 410.158736] PM: Syncing filesystems ... done.
> > Sep 26 17:27:06 bovic kernel: [ 410.161113] PM: Preparing system for mem sleep
> > Sep 26 17:27:22 bovic kernel: [ 410.213835] Freezing user space processes ... (elapsed 0.01 seconds) done.
> > Sep 26 17:27:22 bovic kernel: [ 410.228074] Freezing remaining freezable tasks ... (elapsed 0.01 seconds) done.
>
> Looks like the freeze worked...
>
> > Sep 26 17:27:22 bovic kernel: [ 410.244045] PM: Entering mem sleep
> > Sep 26 17:27:22 bovic kernel: [ 410.244065] Suspending console(s) (use no_console_suspend to debug)
> > [...]
> > Sep 26 17:27:22 bovic kernel: [ 410.926483] PM: suspend of devices complete after 682.208 msecs
> > Sep 26 17:27:22 bovic kernel: [ 410.940182] PM: late suspend of devices complete after 13.691 msecs
> > Sep 26 17:27:22 bovic kernel: [ 410.940248] ACPI: Preparing to enter system sleep state S3
> > Sep 26 17:27:22 bovic kernel: [ 410.940935] PM: Saving platform NVS memory
> > Sep 26 17:27:22 bovic kernel: [ 410.940972] Disabling non-boot CPUs ...
> > Sep 26 17:27:22 bovic kernel: [ 411.044037] CPU 1 is now offline
> > Sep 26 17:27:22 bovic kernel: [ 411.044557] ACPI: Low-level resume complete
> > Sep 26 17:27:22 bovic kernel: [ 411.044557] PM: Restoring platform NVS memory
> > Sep 26 17:27:22 bovic kernel: [ 411.044557] Enabling non-boot CPUs ...
> > Sep 26 17:27:22 bovic kernel: [ 411.044557] Booting Node 0 Processor 1 APIC 0x1
> > Sep 26 17:27:22 bovic kernel: [ 411.044557] smpboot cpu 1: start_ip = 9b000
> > Sep 26 17:27:22 bovic kernel: [ 410.942509] Initializing CPU#1
> > Sep 26 17:27:22 bovic kernel: [ 410.942509] Calibrating delay loop (skipped) already calibrated this CPU
> > Sep 26 17:27:22 bovic kernel: [ 411.076635] NMI watchdog enabled, takes one hw-pmu counter.
> > Sep 26 17:27:22 bovic kernel: [ 411.080018] Switched to NOHz mode on CPU #1
> > Sep 26 17:27:22 bovic kernel: [ 411.080077] CPU1 is up
> > Sep 26 17:27:22 bovic kernel: [ 411.080453] ACPI: Waking up from system sleep state S3
> >
> > And after the wake up things seem not to work right - doing a "dmesg" on
> > a serial console hangs for example.
> >
>
> Not sure what happened here. Looks like the "freeze phase" of the
> suspend worked. My inclination is to think that this is not due to my
> patch.
>
> What might be helpful is to try suspending without NFS being mounted.
> If that doesn't work either, then maybe back out the patch and try
> again without any NFS mounts.
Yeah, after looking over the code I really think the instant wakeup
thing is an unrelated problem. This patchset is really intended to fix
the problem where the freezer is unable to freeze tasks that are stuck
waiting on replies from NFS and CIFS servers.
If you aren't seeing these sorts of errors when trying to suspend, then
this is probably a different problem:
https://bugzilla.redhat.com/show_bug.cgi?id=717735#c0
It would be nice to have some successful tests with this set though
before I propose it, but that might not be possible given the general
fragility of the userspace suspend code :-/
John would you be able to test the things I mentioned in my prior
reply? In particular, seeing whether you're able to suspend
consistently without nfs mounts and without those patches in place? If
that doesn't work either, then I'll be inclined to think that this is
unrelated.
--
Jeff Layton <jlayton@redhat.com>
Reply to: