Re: PROBLEM: oops

[Peter Zijlstra <peterz@infradead.org>]

On Tue, 2009-08-25 at 23:23 +0200, Pawel Golaszewski wrote:
> On Tue, 25 Aug 2009, Peter Zijlstra wrote:
> > > > could you try to reproduce without that?
> > > > 
> > > > CONFIG_GROUP_SCHED=n
> > > I'll try. 
> 
> It seems that problem still exists - system has crashed too. From 
> netconsole:

> Any ideas?
> 
> Last kernel I was using is 2.6.27.13 - works fine. None between 13 and 31 
> tested...

# git log --format=oneline v2.6.27.13..v2.6.27.31 kernel/sched*
2b46f3769896dc04e1e49144d282e4655677105a wait: prevent exclusive waiter starvation

Nothing changed anywhere near the code that is falling apart..

Also:

[20621.263297] BUG: unable to handle kernel paging request at ffffffd8
[20621.263396] IP: [<c081cc50>] hrtick_start_fair+0x0/0x30


#ifndef CONFIG_FAIR_GROUP_SCHED

static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
{
        struct task_struct *p = task_of(se);
        struct rq *rq = task_rq(p);

        return &rq->cfs;
}

#endif

SCHED_FEAT(HRTICK, 0)

static void hrtick_start_fair(struct rq *rq, struct task_struct *p)
{
        struct sched_entity *se = &p->se;
        struct cfs_rq *cfs_rq = cfs_rq_of(se);

        WARN_ON(task_rq(p) != rq);

        if (hrtick_enabled(rq) && cfs_rq->nr_running > 1) {

		/* won't get here since HRTICK is disabled by default */

        }
}

Suggests something went terribly wrong and corrupted a fundamental data
structure.

Maybe turning on each and every debug feature related to memory
debugging might help.