Re: Inconsistent lock state warning for rxe_poll_cq()

[Guoqing Jiang <guoqing.jiang@linux.dev>]

On 2/15/22 2:01 AM, Bart Van Assche wrote:
> Hi Bob,
>
> If I run the SRP tests against Jason's rdma/for-next branch then these
> tests pass if I use the siw driver but not if I use the rdma_rxe driver.
> Can you please take a look at the output triggered by running blktests?
> If I run blktests as follows: ./check -q srp, the following output
> appears:
>
> WARNING: CPU: 1 PID: 1052 at kernel/softirq.c:363 
> __local_bh_enable_ip+0xa4/0xf0
>  _raw_write_unlock_bh+0x31/0x40
>  __rxe_add_index+0x38/0x50 [rdma_rxe]
>  rxe_create_ah+0xce/0x1b0 [rdma_rxe]
>  _rdma_create_ah+0x2c8/0x2f0 [ib_core]
>  rdma_create_ah+0xfd/0x1c0 [ib_core]
>  cm_alloc_msg+0xbc/0x280 [ib_cm]
>  cm_alloc_priv_msg+0x2d/0x70 [ib_cm]
>  ib_send_cm_req+0x4fe/0x830 [ib_cm]
>  cma_connect_ib+0x3c4/0x600 [rdma_cm]
>  rdma_connect_locked+0x145/0x490 [rdma_cm]
>  rdma_connect+0x31/0x50 [rdma_cm]
>  srp_send_req+0x58a/0x830 [ib_srp]
>  srp_connect_ch+0x9f/0x1d0 [ib_srp]
>  add_target_store+0xa6b/0xf20 [ib_srp]
>  dev_attr_store+0x3e/0x60
>  sysfs_kf_write+0x87/0xa0
>  kernfs_fop_write_iter+0x1c7/0x270
>  new_sync_write+0x296/0x3c0
>  vfs_write+0x43c/0x580
>  ksys_write+0xd9/0x180
>  __x64_sys_write+0x42/0x50
>  do_syscall_64+0x35/0x80
>  entry_SYSCALL_64_after_hwframe+0x44/0xae

I think it is exactly the same which I encountered with 5.17-rc3.

https://lore.kernel.org/linux-rdma/20220210073655.42281-1-guoqing.jiang@linux.dev/T/#m0c0ce0745078095ea931f61d2b1c6ce0fdd5403b

Thanks,
Guoqing