RE: [linux-pm] [BUG] sleeping function called from invalid context during resume

[Alan Stern <stern@rowland.harvard.edu>]

On Mon, 10 Jul 2006, Brown, Len wrote:

> >... a similar fix needs to be made in
> drivers/acpi/osl.c:acpi_os_wait_semaphore().
> >If interrupts are disabled the timeout argument should be set to 0, so
> that the 
> >routine will call down_trylock() instead of down() or
> schedule_timeout_interruptible().
> 
> We used to have a hack in acpi_os_wait_semaphore():
> 
> if (in_atomic())
> 	timeout = 0;
> 
> But we deleted it upon ACPICA 20060608 when the
> ACPICA locks that were used at interrupt-time were
> converted to be Linux spin-locks.
> 
> Now it is still conceivable that during resume before interrutps are
> re-enabled,
> the PCI interrupt link devices run AML and go to acquire an AML mutex
> with
> a timeout.  However, we are single threaded at that point, so it isn't
> possible for them to acquire the mutex -- timeout or not.
> 
> I don't like the looks of the "workaround" above -- it makes the code
> confusing.
> 
> I'd be open to putting a BUG_ON() in the sleep case if interrupts are
> not enabled.
> 
> Is there another case that you can think of?

I can't give you an answer now; this case was triggered during 
resume-from-disk on my computer at home.

Tonight I'll attach a stack trace to the bug report.  Let's continue 
further discussion there, off-list.

Alan Stern

-
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/