[ofa-general] Re: OFED 1.3 hang in cm_destroy_id()

[Or Gerlitz]

akepner at sgi.com wrote:
> I've gotten a report of a hang very similar to one reported in:
> http://lists.openfabrics.org/pipermail/general/2008-June/052275.html
>
> Here's the backtrace of the hung ipoib task:
>
> STACK TRACE FOR TASK: 0xe00003600b070000 (ipoib)
>
>  0 schedule+0x26ec [0xa0000001005a12ac]
>  1 wait_for_completion+0x14c [0xa0000001005a198c]
>  2 cm_destroy_id+0x66c [0xa00000021531e72c]
>  3 ib_destroy_cm_id+0x2c [0xa0000002153210cc]
>  4 ipoib_cm_tx_reap+0x17c [0xa000000215719abc]
>  5 run_workqueue+0x1dc [0xa0000001000c7f1c]
>  6 worker_thread+0x1bc [0xa0000001000c963c]
>  7 kthread+0x23c [0xa0000001000d39dc]
>  8 kernel_thread_helper+0xcc [0xa0000001000133ec]
>  9 start_kernel_thread+0x1c [0xa0000001000094bc]
>
>
> The cm_id->state is IB_CM_TIMEWAIT, and the refcount is 1.
>
> This is an ia64 system with an MT23108, running OFED 1.3.
>
> Haven't yet been able to reproduce this, however.
>
Hi Arthur,

As of the large set of patches which found their way into ofed 1.3 
without passing through kernel acceptance and the lack of support for 
ofed by the Linux IB maintainer,  I truly believe that the most 
constructive approach to debug  ipoib issues is to try and reproduce the 
bug on the mainline kernel code and work with the mainstream kernel IB 
maintainer.

Or.