Messages in this thread | | | Date | Wed, 11 Feb 2009 10:27:45 +0100 | From | Ingo Molnar <> | Subject | Re: 2.6.29-rc3: BUG: scheduling while atomic: udevd/917/0x10000100 |
| |
* Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> On Wed, 2009-02-11 at 09:32 +0100, Ingo Molnar wrote: > > > There's a fourth possibility: > > > > - Given that it's udevd that does it - maybe we leaked a softirq preempt > > count in some rare sysfs file, and it did not get discovered until the > > next innocent piece of kernel code preempted? > > > > But i thought lockdep would already warn if we exited a syscall with locks > > held or with a preempt count elevated - Peter? > > From a quick look it only checks task->lockdep_depth, in lockdep_sys_exit().
That could be broadened then i guess - there's never a good excuse for exiting a syscall with an elevated preempt count.
Ingo
| |