Re: [patch] voluntary-preempt-2.6.9-rc2-mm4-S7

[Lee Revell]

On Mon, 2004-09-27 at 20:05, Ingo Molnar wrote:
> i've released the -S7 VP patch:
> 
>   http://redhat.com/~mingo/voluntary-preempt/voluntary-preempt-2.6.9-rc2-mm4-S7
> 

I believe we have found a bug.  A user reported massive xruns with S7,
they turned out to be printk() overhead from tons of "using
smp_processor_id() in preemptible code" errors.  The trace below repeats
over and over.  Looks like raid0 is the problem.

using smp_processor_id() in preemptible code: kjournald/685
 [<c011892a>] smp_processor_id+0x8a/0xa0
 [<c0273f67>] raid0_make_request+0x37/0x240
 [<c02e1bde>] cond_resched+0xe/0x80
 [<c0111608>] mcount+0x14/0x18
 [<c020f4a7>] generic_make_request+0x117/0x1a0
 [<c01621e6>] submit_bh+0xe6/0x150
 [<c020f544>] submit_bio+0x14/0x120
 [<c0111608>] mcount+0x14/0x18
 [<c020f5a2>] submit_bio+0x72/0x120
 [<c0133050>] autoremove_wake_function+0x0/0x60
 [<c0111608>] mcount+0x14/0x18
 [<c0162a6a>] bio_alloc+0xea/0x1d0
 [<c01621e6>] submit_bh+0xe6/0x150
 [<c01b022b>] journal_commit_transaction+0x102b/0x1680
 [<c02e27b3>] _spin_unlock+0x13/0x40
 [<c0136914>] _metered_spin_unlock+0x14/0xa0
 [<c0117c51>] find_busiest_group+0xf1/0x310
 [<c01de306>] find_next_bit+0x16/0xa0
 [<c011815e>] load_balance_newidle+0x8e/0xc0
 [<c01178ee>] move_tasks+0xe/0x280
 [<c02e27b3>] _spin_unlock+0x13/0x40
 [<c0136914>] _metered_spin_unlock+0x14/0xa0
 [<c0135059>] sub_preempt_count+0x69/0x80
 [<c0135059>] sub_preempt_count+0x69/0x80
 [<c01263bf>] del_timer_sync+0xaf/0x160
 [<c01de306>] find_next_bit+0x16/0xa0
 [<c0111608>] mcount+0x14/0x18
 [<c01de306>] find_next_bit+0x16/0xa0
 [<c01263bf>] del_timer_sync+0xaf/0x160
 [<c0111608>] mcount+0x14/0x18
 [<c02e27bf>] _spin_unlock+0x1f/0x40
 [<c01b31f3>] kjournald+0xf3/0x270
 [<c0133050>] autoremove_wake_function+0x0/0x60
 [<c02e2b20>] _spin_unlock_irq+0x20/0x40
 [<c0133050>] autoremove_wake_function+0x0/0x60
 [<c0117223>] schedule_tail+0x23/0x70
 [<c01b30d0>] commit_timeout+0x0/0x20
 [<c01b3100>] kjournald+0x0/0x270
 [<c0102779>] kernel_thread_helper+0x5/0xc
printk: 4203 messages suppressed.
using smp_processor_id() in preemptible code: kjournald/685
 [<c011892a>] smp_processor_id+0x8a/0xa0
 [<c0273f67>] raid0_make_request+0x37/0x240

etc.

I have the .config if you need more info.  All PREEMPT_ options are
enabled.

Joel, you reported the same xruns with SCSI and IDE.  Are you running
raid0 in both cases?  Also what is your hardware configuration?  Is this
regular SMP, or HT?

Lee

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/