On Sun, Jun 29, 2008 at 03:40:57PM +0300, Or Gerlitz wrote: > Eli, in the same setting of having troubles with the new connectx > firmware, I see also this soft lockup at ipoib which takes place when > there's an attempt to probe out the mlx4 driver. Under what condition does this happen, it not just unloading the module is it?
> I recall that some fix > was done lately to wait_for_completion in the mainline kernel, is this > hang related to that? I don't know. I would expect that to happen if the CPU is really overloaded with work. From what I see bellow, the process is in 'D' state which I think is uninterruptible and that suggests that it does not get a completion on a command. However, ofed can handle that since it checks done->done before deciding if this is a failure. What are you running? Did you wait enough time to see if the module evetually unloads? > > ipoib D ffffffff88445ea2 0 2820 105 2954 2809 > (L-TLB) > ffff8102236cbd90 0000000000000046 ffff8102238b4860 ffff8102238b4980 > 000000000000000a ffff81022c0dd100 ffff810107b94100 00000096b408c7fc > 00000000000036b9 ffff81022c0dd2e8 ffff810200000001 ffff8102238b4800 > Call Trace: > [<ffffffff88445ea2>] :ib_ipoib:ipoib_cm_tx_reap+0x0/0x21b > [<ffffffff800610e7>] wait_for_completion+0x79/0xa2 > [<ffffffff800884ac>] default_wake_function+0x0/0xe > [<ffffffff8842d18f>] :ib_cm:cm_enter_timewait+0xa5/0xb5 > [<ffffffff8842de30>] :ib_cm:cm_destroy_id+0x1cf/0x209 > [<ffffffff88445f20>] :ib_ipoib:ipoib_cm_tx_reap+0x7e/0x21b > [<ffffffff8004b2ab>] run_workqueue+0x94/0xe5 > [<ffffffff80047c09>] worker_thread+0x0/0x122 > [<ffffffff8009b283>] keventd_create_kthread+0x0/0x61 > [<ffffffff80047cf9>] worker_thread+0xf0/0x122 > [<ffffffff800884ac>] default_wake_function+0x0/0xe > [<ffffffff8009b283>] keventd_create_kthread+0x0/0x61 > [<ffffffff8009b283>] keventd_create_kthread+0x0/0x61 > [<ffffffff800321d8>] kthread+0xfe/0x132 > [<ffffffff8005bfb1>] child_rip+0xa/0x11 > [<ffffffff8009b283>] keventd_create_kthread+0x0/0x61 > [<ffffffff800320da>] kthread+0x0/0x132 > [<ffffffff8005bfa7>] child_rip+0x0/0x11 > > rmmod D 0000000000000880 0 5875 5874 > (NOTLB) > ffff810218085d88 0000000000000082 ffff81022d6961c0 ffff81022d1efc00 > 0000000000000007 ffff81022e9457e0 ffff81022fc16100 0000023f4abf6f5f > 000000000000db90 ffff81022e9459c8 ffff810200000003 0000000000000246 > Call Trace: > [<ffffffff80098556>] flush_cpu_workqueue+0x7f/0xad > [<ffffffff8009b446>] autoremove_wake_function+0x0/0x2e > [<ffffffff80098592>] flush_workqueue+0xe/0x85 > [<ffffffff88443081>] :ib_ipoib:ipoib_mcast_stop_thread+0x93/0x9a > [<ffffffff88440486>] :ib_ipoib:ipoib_ib_dev_cleanup+0x2f/0x40 > [<ffffffff8843e5d6>] :ib_ipoib:ipoib_dev_cleanup+0x78/0xaf > [<ffffffff8843e66c>] :ib_ipoib:ipoib_remove_one+0x5f/0x96 > [<ffffffff8838afa4>] :ib_core:ib_unregister_device+0x30/0xc4 > [<ffffffff883de87d>] :mlx4_ib:mlx4_ib_remove+0x47/0x7c > [<ffffffff88161280>] :mlx4_core:mlx4_remove_device+0x44/0x6a > [<ffffffff881613a7>] :mlx4_core:mlx4_unregister_interface+0x29/0x68 > [<ffffffff883e42e0>] :mlx4_ib:mlx4_ib_cleanup+0x10/0x34 > [<ffffffff800a0d35>] sys_delete_module+0x196/0x1c5 > [<ffffffff8005b28d>] tracesys+0xd5/0xe0 > > _______________________________________________ > general mailing list > [email protected] > http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general > > To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
