Re: 2.6.7-mm7

[Paul Mackerras <paulus@samba.org>]

Andrew Morton writes:

> hm, OK.  It could be that the debug patch is a bit too aggressive, or that
> ppc got lucky and happens to always be in state TASK_RUNNING when these
> calls to schedule() occur.
> 
> Maybe this task incorrectly has _TIF_NEED_RESCHED set?

Is CONFIG_PREEMPT enabled?

> Anyway, ppc guys: please take a look at the results from
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.7/2.6.7-mm7/broken-out/detect-too-early-schedule-attempts.patch
> and check that the kernel really should be calling schedule() at this time
> and place, let us know?
> 
> Thanks.
> 
> >  The first one looks like:
> >
> > Calibrating delay loop... 1064.96 BogoMIPS
> > Mount-cache hash table entries: 512 (order: 0, 4096 bytes)
> > Badness in schedule at kernel/sched.c:2153
> > Call trace:
> >  [c00099e4] dump_stack+0x18/0x28
> >  [c0006bac] check_bug_trap+0x84/0xac
> >  [c0006d38] ProgramCheckException+0x164/0x1a4
> >  [c0006240] ret_from_except_full+0x0/0x4c
> >  [c02021bc] schedule+0x24/0x684
> >  [c0005e80] syscall_exit_work+0x108/0x10c
> >  [c02e0ad0] proc_root_init+0x14c/0x158
> >  [00000000] 0x0
> >  [c02ce5a0] start_kernel+0x158/0x184
> >  [000035fc] 0x35fc

This looks like CONFIG_PREEMPT is enabled and _TIF_NEED_RESCHED is set
at the end of handling a system call.  AFAICS i386 will also call
schedule in these circumstances.  Does this mean we shouldn't do
system calls until the scheduler is running?

Paul.