Re: Corrupted files after suspend to disk

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

On Thursday, March 24, 2011, richard -rw- weinberger wrote:
> On Thu, Mar 24, 2011 at 11:16 AM, richard -rw- weinberger
> <richard.weinberger@gmail.com> wrote:
> > On Thu, Mar 24, 2011 at 12:00 AM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> >> On Wednesday, March 23, 2011, richard -rw- weinberger wrote:
> >>> On Wed, Mar 23, 2011 at 11:22 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> >>> > On Wednesday, March 23, 2011, richard -rw- weinberger wrote:
> >>> >> On Wed, Mar 23, 2011 at 11:11 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> >>> >> > On Wednesday, March 23, 2011, richard -rw- weinberger wrote:
> >>> >> >> 2011/3/23 Rafael J. Wysocki <rjw@sisk.pl>:
> >>> >> >> > On Wednesday, March 23, 2011, richard -rw- weinberger wrote:
> >>> >> >> >> Hi,
> >>> >> >> >>
> >>> >> >> >> I'm facing a very strange problem on my netbook (Lenovo Ideapad S10)
> >>> >> >> >> running Linux 2.6.37.4.
> >>> >> >> >> After resuming from s2disk some files are corrupted.
> >>> >> >> >> But when I reboot my netbook everything seems good again.
> >>> >> >> >>
> >>> >> >> >> When I saw the problem the first time the ls command segfaulted always.
> >>> >> >> >> I did a reboot and it worked again.
> >>> >> >> >>
> >>> >> >> >> A few days later zypper crashed. After a reboot it worked again.
> >>> >> >> >> And today ssh crashed. I looked a bit closer and saw it crashed
> >>> >> >> >> somewhere within libcrypto.
> >>> >> >> >> So I made copy libcrypto and rebooted.
> >>> >> >> >> After the reboot ssh worked again but libcrypto and the copy of it hat
> >>> >> >> >> a different sha1 sum!
> >>> >> >> >> WTF?!
> >>> >> >> >>
> >>> >> >> >> Is this a known issue?
> >>> >> >> >
> >>> >> >> > No.
> >>> >> >> >
> >>> >> >> >> dmesgs and config are attached.
> >>> >> >> >>
> >>> >> >> >> The used distribution is openSUSE 11.4 with suspend-0.80.20100129-7.1
> >>> >> >> >> (default from suse).
> >>> >> >> >> I'm using ext3 as root filesystem.
> >>> >> >> >> What else do you need?
> >>> >> >> >
> >>> >> >> > Whatever you can do to narrow down the problem.  At the moment I only know
> >>> >> >> > that it's there.
> >>> >> >>
> >>> >> >> I can reproduce the problem now.
> >>> >> >> After ~20 suspend and resume iterations aide finds corrupted files in /lib/.
> >>> >> >> It's always a very basic lib like libcrypto, libglib which is used all
> >>> >> >> the time on my system.
> >>> >> >
> >>> >> > Those files are never intentionally modified, right?
> >>> >> >
> >>> >> >> Maybe it's an issue like this one?
> >>> >> >> https://lkml.org/lkml/2010/12/2/339
> >>> >> >
> >>> >> > It might have if that patch hadn't been merged before 2.6.37.
> >>> >> >
> >>> >> > Is the system 32-bit or 64-bit?
> >>> >>
> >>> >> It's a 32-bit system.
> >>> >> cmp shows that the corrupted files differ in many bytes (not scattered).
> >>> >> The corrupted bytes are always 0 or 252.
> >>> >
> >>> > Do I understand correctly that the files apparently corrupted after resume
> >>> > are not corrupted any more when you reboot?
> >>>
> >>> Yes.
> >>> Seems like a cache issue.
> >>
> >> There's a couple things you can check before we start asking other people for
> >> help.
> >>
> >> First, it would be good to know if things change when you save the image
> >> into a swap file instead of the swap partition you've been using so far
> >> (I believe it's documented quite well how to do that).
> >>
> >> Second, please verify if using the built-in save/load hibernate code leads
> >> to the same issue (you can hibernate by doing "echo disk > /sys/power/state"
> >> to verify that).
> >>
> >> Of course, please test the above separately. :-)
> >
> > Ok, I'll test this when I'm at home.
> >
> > BTW: dropping the caches helps, when some files seem corrupted.
> > Today /usr/bin/okular was broken.
> > After setting vm.drop_caches=1 it worked again.
> >
> >> Thanks,
> >> Rafael
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> >> the body of a message to majordomo@vger.kernel.org
> >> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >> Please read the FAQ at  http://www.tux.org/lkml/
> >>
> >
> > --
> > Thanks,
> > //richard
> >
> 
> On Linux 2.6.38 I'm unable to reproduce the issue.
> Only 2.6.37 seems to be affected.
> So, I'm moving over to 2.6.38. :)

OK, thanks for the report. :-)

Rafael