[ofa-general] SDP and stock kernel gets BUG?

[Jason Gunthorpe]

Hey all,

I'm trying to use SDP with stock 2.6.30.1 plus the
'drivers/infiniband/ulp/sdp' directory from
ofa_kernel-1.5-ofed20090713.src.rpm and I get this BUG ON:

BUG: scheduling while atomic: ib_cm/0/4209/0x00000003
Modules linked in: ib_sdp w83793 hwmon_vid rdma_ucm rdma_cm iw_cm
ib_addr mlx4_ib ib_ipoib ib_cm ib_sa ib_uverbs ib_umad ib_mthca e1000e
ib_mad ib_core mlx4_core i5k_amb hwmon i2c_i801 i2c_core
Pid: 4209, comm: ib_cm/0 Not tainted 2.6.30.1 #1
Call Trace:
 [<ffffffff80230bd0>] __schedule_bug+0x65/0x6a
 [<ffffffff80552875>] __schedule+0x83/0x901
 [<ffffffff80240eca>] ? del_timer_sync+0x14/0x21
 [<ffffffff805531a3>] schedule+0x13/0x31
 [<ffffffff80553499>] schedule_timeout+0x154/0x17e
 [<ffffffff80240ed7>] ? process_timeout+0x0/0xb
 [<ffffffff805526bf>] ? wait_for_common+0xb9/0x12d
 [<ffffffff805526bf>] wait_for_common+0xb9/0x12d
 [<ffffffff80231527>] ? default_wake_function+0x0/0xf
 [<ffffffff805527a3>] wait_for_completion_timeout+0xe/0x10
 [<ffffffffa00212ec>] __mlx4_cmd+0xf5/0x232 [mlx4_core]
 [<ffffffffa0028c31>] mlx4_mr_free+0x59/0xb7 [mlx4_core]
 [<ffffffffa00d17e1>] mlx4_ib_dereg_mr+0x1f/0x3b [mlx4_ib]
 [<ffffffffa0036c45>] ib_dereg_mr+0x22/0x33 [ib_core]
 [<ffffffffa00fe06f>] sdp_destroy_qp+0xcf/0x107 [ib_sdp]
 [<ffffffffa00ff1e1>] sdp_reset_sk+0x389/0x4ff [ib_sdp]
 [<ffffffffa0107f34>] ? sdp_xmit_poll+0x53/0x5c [ib_sdp]
 [<ffffffffa010468c>] ? sdp_disconnected_handler+0x9a/0xc9 [ib_sdp]
 [<ffffffffa0105f46>] sdp_cma_handler+0x188b/0x1aa5 [ib_sdp]
 [<ffffffff80555333>] ? _spin_unlock_irqrestore+0x2d/0x48
 [<ffffffff80554fd1>] ? _spin_lock_irq+0x1c/0x34
 [<ffffffff80554fd1>] ? _spin_lock_irq+0x1c/0x34
 [<ffffffff805579d2>] ? sub_preempt_count+0x9e/0xbe
 [<ffffffff80555364>] ? _spin_unlock_irq+0x16/0x2f
 [<ffffffffa00ac2d8>] ? cm_work_handler+0x0/0xbf3 [ib_cm]
 [<ffffffffa00e52b1>] cma_ib_handler+0x19b/0x209 [rdma_cm]
 [<ffffffffa00ab064>] cm_process_work+0x1c/0xaf [ib_cm]
 [<ffffffffa00ac696>] cm_work_handler+0x3be/0xbf3 [ib_cm]
 [<ffffffffa00ac2d8>] ? cm_work_handler+0x0/0xbf3 [ib_cm]
 [<ffffffffa00ac2d8>] ? cm_work_handler+0x0/0xbf3 [ib_cm]
 [<ffffffff80248149>] worker_thread+0x14b/0x1f5
 [<ffffffff8024b10f>] ? autoremove_wake_function+0x0/0x38
 [<ffffffff80247ffe>] ? worker_thread+0x0/0x1f5
 [<ffffffff8024afdd>] kthread+0x56/0x85
 [<ffffffff8020beda>] child_rip+0xa/0x20
 [<ffffffff802308a8>] ? finish_task_switch+0xb5/0xc4
 [<ffffffff8020b8ad>] ? restore_args+0x0/0x30
 [<ffffffff8024af87>] ? kthread+0x0/0x85
 [<ffffffff8020bed0>] ? child_rip+0x0/0x20

Anyone know what is up?? Does SDP work with the baseline OFED 1.5
kernel? This kernel has CONFIG_PREEMPT=y if that makes a difference.

Thanks,
Jason