Re: [BUG RT] 2.6.25.8-rt7 on 8-way

[Noah Watkins]

While I have disabled RCU tracing to suppress the messages, I am
getting them in another context:

BUG: using smp_processor_id() in preemptible [00000000] code: sshd/3114
caller is __qdisc_run+0x12e/0x1b5
Pid: 3114, comm: sshd Not tainted 2.6.25.8-rt7 #3

Call Trace:
 [<ffffffff8112da39>] debug_smp_processor_id+0xad/0xb8
 [<ffffffff8120e264>] __qdisc_run+0x12e/0x1b5
 [<ffffffff811ff1c8>] dev_queue_xmit+0x140/0x278
 [<ffffffff8122239e>] ip_queue_xmit+0x2c2/0x350
 [<ffffffff8127f8a6>] add_preempt_count+0x12/0x94
 [<ffffffff8108263e>] __inc_zone_state+0x6a/0x88
 [<ffffffff81231ffe>] tcp_transmit_skb+0x77e/0x7ba
 [<ffffffff8109aa60>] kmem_cache_alloc_node+0x129/0x152
 [<ffffffff81233a75>] __tcp_push_pending_frames+0x713/0x7df
 [<ffffffff81228da2>] tcp_sendmsg+0x93a/0xa52
 [<ffffffff811f0e97>] sock_aio_write+0xf8/0x110
 [<ffffffff8109feeb>] do_sync_write+0xc9/0x10c
 [<ffffffff81046da6>] autoremove_wake_function+0x0/0x2e
 [<ffffffff8103868b>] current_fs_time+0x1e/0x24
 [<ffffffff810a06ae>] vfs_write+0xc0/0x156
 [<ffffffff810a0c67>] sys_write+0x48/0x74
 [<ffffffff8100bf6b>] tracesys+0xdc/0xe1


On Mon, Jun 30, 2008 at 7:37 AM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Sun, 29 Jun 2008, Sven-Thorsten Dietrich wrote:
>
>>
>> On Sun, 2008-06-29 at 16:32 -0500, Noah Watkins wrote:
>> > My kernel is very, very noisy with the following bug:
>> >
>> > - Full dmesg below
>> > - Please CC me
>> >
>>
>> I found this as well (adding Paul).
>>
>> If you disable CONFIG_PREEMPT_RCU_BOOST, does it go away?
>>
>
> Turning off CONFIG_RCU_TRACE will also make it go away. The bug is that
> the tracing of the RCU boost code uses smp_processor_id while the RCU
> boost code is not in an atomic section.
>
> Paul has a fix for this already, and will be out in the next releases.
>
> -- Steve
>
--
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/