On (07/07/09 14:22), Catalin Marinas wrote:
> > > > kernel: [ 1917.133154] INFO: RCU detected CPU 0 stall (t=485140/3000 jiffies)
> [...]
> > > What I think happens is that the kmemleak thread runs for several
> > > seconds for scanning the memory and there may not be any context
> > > switches. I have a patch to add more cond_resched() calls throughout the
> > > kmemleak_scan() function which I hope will get merged. 
> [...]
> > > I don't get any  of these messages with CONFIG_PREEMPT enabled.
> > 
> > It started with rc2-git1 (may be). Almost every scan ends with RCU pending.
> 
> Should I assume that CONFIG_PREEMPT is disabled on your system?
Yeah, sorry. 
# CONFIG_PREEMPT is not set
/* My .config is in attach. */

> 
> The branch with the pending kmemleak patches is below (I sent Linus a
> pull request):
> 
> http://www.linux-arm.org/git?p=linux-2.6.git;a=shortlog;h=kmemleak
> 
> You can try the "Add more cond_resched() calls..." patch and see if it
> makes any difference.
> 
Ok. I'll try.

> > Hm.. Something is broken...
> > cat /.../kmemleak
> > [ 7933.537868] ================================================
> > [ 7933.537873] [ BUG: lock held when returning to user space! ]
> > [ 7933.537876] ------------------------------------------------
> > [ 7933.537880] cat/2897 is leaving the kernel with locks still held!
> > [ 7933.537884] 1 lock held by cat/2897:
> > [ 7933.537887]  #0:  (scan_mutex){+.+.+.}, at: [<c10f717c>] kmemleak_open+0x4c/0x80
> 
> That the "Do not acquire scan_mutex in kmemleak_open()" patch in the
> same branch.
> 
> -- 
> Catalin
> 


Thanks,
	Sergey