[TuxOnIce-users] unexpected resume after normal reboot -> filesystem corruption

[Nix]

On 17 Oct 2010, Nigel Cunningham outgrape:

> On 18/10/10 05:02, Nix wrote:
>> I just had a rather nasty problem with 2.6.35.6+ToI. I resumed, then for
>> other reasons decided to shutdown -rf and reboot. Post-reboot I was
>> surpised to see it resuming again! Ohcrap, I thought, shutdown -rf and
>> reboot again... and again it resumed!
>>
>> I rebooted with noresume and found (unsurprisingly) a pile of FS
>> corruption on /var. One fsck later and the machine was back up.
>>
>> So I'm left wondering why on earth it tried to resume after a normal
>> reboot. All I can imagine is that if the system has so much memory that
>> it doesn't swap at all after resumption, perhaps the ToI signature is
>> somehow left hanging around on the swap partition, so it mistakenly
>> thinks that we're in the middle of a suspend?
>
> When you finish resuming, TuxOnIce normally resets the signature in
> the swap header to the normal one. The only time it doesn't reset the
> signature is when you enable the 'keep image' mode at compile time and
> then via the sysfs entry. Is there any chance that you did that?

No chance at all. It's off, and I've never turned it on. (Also, I've
suspended and resumed perfectly well with this ToI over and over again
before now.)

Like I said, this was a one-off weirdness. If it happens again, I'll
kick on every debugging option I can find and run it again (and also try
to figure out what's gone wrong with netconsole logging so I can get
decent logs from that, rather than logs that terminate the instant
syslogd comes up).