Re: Corrupted files after suspend to disk

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Tuesday, March 13, 2012, Andreas Hartmann wrote:
> Rafael J. Wysocki schrieb:
> > On Tuesday, March 13, 2012, Andreas Hartmann wrote:
> >> Dave Jones wrote:
> >>> On Mon, Mar 12, 2012 at 02:24:00PM +0100, Andreas Hartmann wrote:
> >>>  
> >>>  > >>>>> This is happening to me as well. Something like 1 resume out of 5 goes
> >>>  > >>>>> wrong this very same way.
> >>>  > >>>>>
> >>>  > >>>>> This is thinkpad x200s.
> >>>  > >>>>>
> >>>  > >>>>> All the userspace is segfaulting all over the place (most frequently in
> >>>  > >>>>> libselinux for some reason).
> >>>  > >>>>>
> >>>  > >>>>> I am not able to verify the 'drop_caches' theory, as I can't invoke a
> >>>  > >>>>> single command that wouldn't crash.
> >>>  > >>>>
> >>>  > >>>> The question is how should we proceed?
> >>>  > >>>> I've reported this issue one year (!!!) ago.
> >>>  > >>>
> >>>  > >>> Hmm, 3.3-rcX seems to be the first version when it started to happen to
> >>>  > >>> me. I take it that you have seen this also with 3.2? 3.1?
> >>>  > >>
> >>>  > >> Quote from my very first email:
> >>>  > >> "I'm facing a very strange problem on my netbook (Lenovo Ideapad S10)
> >>>  > >> running Linux 2.6.37.4."
> >>>  > > 
> >>>  > > So we both seem to have Lenovos at least. I thus wanted to verify whether 
> >>>  > > the problem will trigger with thinkpad_acpi removed, but it oopsed while 
> >>>  > > rmmoding :) I will start looking into this right away.
> >>>  > > 
> >>>  > > Is your system using thinkpad_acpi as well?
> >>>  > 
> >>>  > I dont't think, that it is lenovo related as I'm having a MSI machine.
> >>>  > 
> >>>  > https://bugzilla.novell.com/show_bug.cgi?id=732908
> >>>  > 
> >>>  > Following the link, you are able to compare the used chips - maybe there
> >>>  > are some equal components?
> >>>
> >>> This looks like the i915 corruption problem mentioned in a few other threads.
> >>>
> >>> if you compare the hexdump of the good/bad files, you find that the corruption
> >>> happens in 8x 4 byte writes of either 0x00000000 or 0x00aaaaaa.
> >>>
> >>> KeithP clued me in last week that that looks like an ARGB pixel quad, so these
> >>> writes are likely 8 pixel strips.
> >>
> >> Thanks Dave. I disabled i915 (with nomodeset) and voila, the problem
> >> disappears. As I already know, that the problem isn't X-related (I saw
> >> it even without any X involved, only with runlevel 3 and nothing more),
> >> the problem seems to be now narrowed down to the relevant component.
> > 
> > I wonder what's the kernel command line you can reproduce the problem with?
> 
> It's the standard kernel command line of openSUSE:
> root=/dev/system/root resume=/dev/system/swap splash=silent quiet vga=791 3

Is it reproducible with vga=0 (or no vga= option at all)?

Rafael