Re: lockdep false positive? -- firewire-core transaction timer vs. scsi-core host lock

[Johannes Berg <johannes@sipsolutions.net>]

On Mon, 2010-08-16 at 23:42 +0200, Peter Zijlstra wrote:

> softirq:
>   spin_lock(&t->split_timeout_timer);
> 
> IRQ:
>   spin_lock(&(shost->host_lock)->rlock);
>    spin_lock(&t->split_timeout_timer);

Actually, I think it's worried that you could have


CPU 0                         CPU 1
 softirq:                      softirq:
   spin_lock(timer)              spin_lock(rlock)
 irq:
   spin_lock(rlock)
                                 spin_lock(timer) [still in softirq]

because it has previously seen the nesting that I put on CPU 1, and now
found that the timer "lock" is used with interrupts enabled.

At least that's a scenario I could understand?

I'm convinced it's a false positive though, question is how to shut it
up :-)

johannes