Re: soft lockup detected in acpi_processor_idle() -- false positive?

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

On Thursday, 12 January 2006 19:43, Dominik Brodowski wrote:
> Latest git, fresh after resuming from suspend-to-disk (in-kernel variant):
> 
> [4294914.586000] Restarting tasks... done
> [4294922.657000] BUG: soft lockup detected on CPU#0!
> [4294922.657000] 
> [4294922.657000] Pid: 0, comm:              swapper
> [4294922.657000] EIP: 0060:[<f003084c>] CPU: 0
> [4294922.657000] EIP is at acpi_processor_idle+0x1f3/0x2d5 [processor]
> [4294922.657000]  EFLAGS: 00000282    Not tainted  (2.6.15)
> [4294922.657000] EAX: fffff000 EBX: 005543a8 ECX: 00000000 EDX: 00000000
> [4294922.657000] ESI: edcc3064 EDI: edcc2f60 EBP: c041cfdc DS: 007b ES: 007b
> [4294922.657000] CR0: 8005003b CR2: 080c3000 CR3: 2d530000 CR4: 000006d0
> 
> 
> As acpi_processor_idle doesn't take any locks AFAIK, it seems to me to be a
> false positive -- or do I miss something obvious?

I think it's a false-positive.

This "soft lockup" message has been appearing for me for quite some time now
(actually since the softlockup patch made it into -mm ;-)), in a
non-reproducible manner, but I haven't been able to nail it down.

Still, I thought it was x86-64-specific, but your machine is an i386,
so there's more to it, apparently.  Probably there's missing
touch_softlockup_watchdog() somewhere, or the timer .suspend()/.resume()
routines need some additional review.

Greetings,
Rafael