Re: Unreliable hibernation on Lenovo x230 (regression)

[joeyli <jlee@suse.com>]

Hi Rafael, 

On Sat, Apr 04, 2015 at 10:12:43AM +0200, rhn wrote:
> On Fri, 03 Apr 2015 23:43:30 +0200
> "Rafael J. Wysocki" <rjw@rjwysocki.net> wrote:
> 
> > On Friday, April 03, 2015 05:58:25 PM rhn wrote:
> > > On Thu, 2 Apr 2015 17:28:05 +0200
> > > Pavel Machek <pavel@ucw.cz> wrote:
> > > 
> > > > On Wed 2015-04-01 21:47:43, rhn wrote:
> > > > > Hello,
> > > > > 
> > > > > Between kernel 3.16 and 3.17, a regression has been introduced where the first hibernation after regular shutdown always fails to resume. Subsequent hibernations succeed.
> > > > > 
> > > > > The system is a Lenovo x230 with Intel i5, booting with EFI, with the hibernate partition located on a secondary SSD drive. Installed system is Fedora 20, hibernation and reboots were issued using the KDE shutdown dialog.
> > > > > 
> > > > > I have tracked the problem to first appear in the commit
> > > > > e67ee10190e69332f929bdd6594a312363321a66	Merge branches 'pm-sleep', 'pm-cpufreq' and 'pm-cpuidle'
> > > > > 
> > > > > The problem itself manifests in dmesg as follows (system was first
> > > > > restarted, then hibernated - this log is from the subsequent
> > > > resume):
> > > > 
> > > > Ok, can you try to disable cpufreq and cpuidle, and then try if it
> > > > reproduces?
> > > > 
> > > > At that point, this is the candidate:
> > > > 
> > > > commit e67ee10190e69332f929bdd6594a312363321a66
> > > > Merge: 21c806d 84c91b7 39c8bba 372ba8c
> > > > Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > > > Date:   Mon Aug 11 23:19:48 2014 +0200
> > > > 
> > > >     Merge branches 'pm-sleep', 'pm-cpufreq' and 'pm-cpuidle'
> > > > 
> > > >     * pm-sleep:
> > > >           PM / hibernate: avoid unsafe pages in e820 reserved regions
> > > > 
> > > > ...
> > > > Alternatively, you can just try to revert
> > > > 
> > > > commit 84c91b7ae07c62cf6dee7fde3277f4be21331f85
> > > > Author: Lee, Chun-Yi <joeyli.kernel@gmail.com>
> > > > Date:   Mon Aug 4 23:23:21 2014 +0800
> > > > 
> > > >     PM / hibernate: avoid unsafe pages in e820 reserved regions
> > > > 
> > > >     When the machine doesn't well handle the e820 persistent when
> > > >     hibernate
> > > >         resuming, then it may cause page fault when writing image to
> > > >     snapshot
> > > >         buffer:
> > > > 
> > > > 
> > > > ...
> > > > 
> > > > Thanks,
> > > > 									Pavel
> > > 
> > > I tried to disable CONFIG_CPU_IDLE and CONFIG_CPU_FREQ, however for some reason I could only disable CONFIG_CPU_FREQ.
> > > 
> > > The bug persisted.
> > > 
> > > Reverting the commit 84c91b7 on top of e67ee10 fixes the problem.
> > > 
> > > I created a copy of the bug report here: https://bugzilla.kernel.org/show_bug.cgi?id=96111
> > 
> > Please check if 4.0-rc6 still has the problem and if reverting the commit in
> > question on top of it fixes the problem too.
> > 
> > 
> 
> I took the commit 8f778bbc542ddf8f6243b21d6aca087e709cabdc as the base for further checking (I started building before I read your message). It's a descendant of 4.0-rc6, so I hope it's not going to make a difference.
> 
> Results:
> 8f778bb : bad
> 8f778bb + reverted 84c91b7 : good
> 8f778bb + patch [1] : good

Thanks for your dmesg on bko#96111.
I checked and confirm there have the situation of setup_data reserved as E820_RESERVED_KERN.
I will add comment on bugzilla.

> 
> Thanks!
> 
> [1]:
> x86: Kill E820_RESERVED_KERN  https://lkml.org/lkml/2015/3/4/434 as suggested in joeyli's other email.

I think just revert 84c91b7ae until Yinghai Lu's patches merged to v4.1.
I will resend 84c91b7ae patch until Yinghai Lu's patches merged.


Regards
Joey Lee