Lock issue with 2a7cec53816 ("RDMA/cma: Fix locking for the RDMA_CM_CONNECT state")

From: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
To: jgg@ziepe.ca, leon@kernel.org
Cc: linux-rdma@vger.kernel.org, linux-nvme@lists.infradead.org,
	Jinpu Wang <jinpu.wang@cloud.ionos.com>
Subject: Lock issue with 2a7cec53816 ("RDMA/cma: Fix locking for the RDMA_CM_CONNECT state")
Date: Mon, 26 Oct 2020 09:19:37 +0100	[thread overview]
Message-ID: <3b1f7767-98e2-93e0-b718-16d1c5346140@cloud.ionos.com> (raw)

Hi Jason & Leon,

[Sorry to send again since the previous mail didn't hit the list]

With the mentioned commit, I got the below calltraces.

[ 1209.544359] INFO: task kworker/u8:3:615 blocked for more than 120 
seconds.
[ 1209.545170]       Tainted: G           O      5.9.0-rc8+ #30
[ 1209.545341] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" 
disables this message.
[ 1209.545814] task:kworker/u8:3    state:D stack:    0 pid:  615 
ppid:     2 flags:0x00004000
[ 1209.547601] Workqueue: rdma_cm cma_work_handler [rdma_cm]
[ 1209.548645] Call Trace:
[ 1209.552120]  __schedule+0x363/0x7b0
[ 1209.552903]  schedule+0x44/0xb0
[ 1209.553093]  schedule_preempt_disabled+0xe/0x10
[ 1209.553253]  __mutex_lock.isra.10+0x26d/0x4e0
[ 1209.553421]  __mutex_lock_slowpath+0x13/0x20
[ 1209.553596]  ? __mutex_lock_slowpath+0x13/0x20
[ 1209.553751]  mutex_lock+0x2f/0x40
[ 1209.554045]  rdma_connect+0x52/0x770 [rdma_cm]
[ 1209.554215]  ? vprintk_default+0x1f/0x30
[ 1209.554359]  ? vprintk_func+0x62/0x100
[ 1209.554532]  ? cma_cm_event_handler+0x2b/0xe0 [rdma_cm]
[ 1209.554688]  ? printk+0x52/0x6e
[ 1209.554852]  rtrs_clt_rdma_cm_handler+0x393/0x8b0 [rtrs_client]
[ 1209.555056]  cma_cm_event_handler+0x2b/0xe0 [rdma_cm]
[ 1209.561823]  ? cma_comp_exch+0x4e/0x60 [rdma_cm]
[ 1209.561883]  ? cma_cm_event_handler+0x2b/0xe0 [rdma_cm]
[ 1209.561940]  cma_work_handler+0x89/0xc0 [rdma_cm]
[ 1209.561988]  process_one_work+0x20c/0x400
[...]
[ 1209.562323] INFO: task bash:630 blocked for more than 120 seconds.
[ 1209.562368]       Tainted: G           O      5.9.0-rc8+ #30
[ 1209.562404] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" 
disables this message.
[ 1209.562456] task:bash            state:D stack:    0 pid:  630 
ppid:   618 flags:0x00004000
[ 1209.562515] Call Trace:
[ 1209.562559]  __schedule+0x363/0x7b0
[ 1209.562603]  schedule+0x44/0xb0
[ 1209.562642]  schedule_preempt_disabled+0xe/0x10
[ 1209.562685]  __mutex_lock.isra.10+0x26d/0x4e0
[ 1209.562727]  ? vprintk_func+0x62/0x100
[ 1209.562765]  ? vprintk_default+0x1f/0x30
[ 1209.562806]  __mutex_lock_slowpath+0x13/0x20
[ 1209.562847]  ? __mutex_lock_slowpath+0x13/0x20
[ 1209.562887]  mutex_lock+0x2f/0x40
[ 1209.562936]  rdma_destroy_id+0x31/0x60 [rdma_cm]
[ 1209.562983]  destroy_cm.isra.24+0x2d/0x50 [rtrs_client]
[ 1209.563586]  init_sess+0x23b/0xac0 [rtrs_client]
[ 1209.563691]  ? wait_woken+0x80/0x80
[ 1209.563742]  rtrs_clt_open+0x27f/0x540 [rtrs_client]
[ 1209.563791]  ? remap_devs+0x150/0x150 [rnbd_client]
[ 1209.563867]  find_and_get_or_create_sess+0x4a1/0x5e0 [rnbd_client]
[ 1209.563921]  ? remap_devs+0x150/0x150 [rnbd_client]
[ 1209.563966]  ? atomic_notifier_call_chain+0x1a/0x20
[ 1209.564011]  ? vt_console_print+0x203/0x3d0
[ 1209.564051]  ? up+0x32/0x50
[ 1209.564093]  ? __irq_work_queue_local+0x4f/0x60
[ 1209.564132]  ? irq_work_queue+0x1a/0x30
[ 1209.564169]  ? wake_up_klogd.part.28+0x34/0x40
[ 1209.564208]  ? vprintk_emit+0x126/0x2b0
[ 1209.564249]  ? vprintk_default+0x1f/0x30
[ 1209.564292]  rnbd_clt_map_device+0x74/0x7b0 [rnbd_client]
[ 1209.564341]  rnbd_clt_map_device_store+0x39e/0x6c0 [rnbd_client]

 From my understanding, it is caused by the rdma_connect need to hold 
the handler_mutex
again in the path.

cma_ib_handler -> mutex_lock(&id_priv->handler_mutex)
-> cma_cm_event_handler
                     -> id_priv->id.event_handler = rtrs_clt_rdma_cm_handler
-> rdma_connect
-> mutex_lock(&id_priv->handler_mutex)

And seems nvme rdma could have the same issue since rdma_connect is 
called by
nvme_rdma_cm_handler -> nvme_rdma_route_resolved.

Thanks,
Guoqing