Comment # 12 on bug 1100541 from Michal Hocko

(In reply to Miroslav ��pehar from comment #11)
> Created attachment 776716 [details]
> journal-new
> 
> Here's a new journal dump.
> 
> I turned the pc in the morning, then went away. Attached journal contains
> everything since turning it on until it got stuck.

Hmm, these all are spinning waiting for IPI acknowledgment. I do not really see
why the IPI is not delivered and handled properly. This is really strange
because if there was a code to prevent IPI running on a particular CPU then we
would see a hard rather than soft lockup reports.

I am not really sure whether this is related to the report pointed by Takashi
in comment 9. A kernel dump would tell us the cpu we are waiting for and what
it is doing right now. If this is a random memory corruption I can imagine that
somebody was "lucky" to overwrite the ack bit in the mask we are waiting for.