[Devel] Re: utrace regressions (was: -mm merge plans for 2.6.21)

[Alexey Dobriyan]

On Tue, Feb 13, 2007 at 04:05:30PM +0300, Alexey Dobriyan wrote:
> On Mon, Feb 12, 2007 at 01:36:34PM -0800, Roland McGrath wrote:
> > > 2. The following proggie renders box unusable in ~10 seconds (but not
> > >    mainline kernel where Ctrl+C will kill process).
> >
> > I haven't been able to reproduce this so far on my test machine.  I got
> > bored after about 10 minutes ("passed: 119200000").  The machine remains
> > responsive and the test dies immediately on hitting C-c.  I'll try it on
> > some other machines.  It would help to know what kind of unusable (crashy,
> > leaky, wedgey?).
> 
> This is fresh pull 2.6.20-ec2f9d1331f658433411c58077871e1eef4ee1b4 with
> linux-2.6.20-current-utrace.patch slapped on top (without rejects)
> utrace is on and ptrace is on too.
> 
> the behaviour is
> 
> 	$ expl_ptratt
> 	passed: 100000
> 	passed: 200000
> 		...
> 	passed: [always different number, but in several seconds]
> 		[pause]
> 	[OOM killer wakes up steadily killing processes	]
> 	[messages scroll away				]
> 
> I'll turn more debugging options now.
> 
> Just in case, .config:

Turning preempt off makes OOM-killings dissapear.

> # CONFIG_PREEMPT_NONE is not set
> CONFIG_PREEMPT_VOLUNTARY=y
> # CONFIG_PREEMPT is not set
> # CONFIG_X86_UP_APIC is not set