[bug report] workqueue: WQ_MEM_RECLAIM nvmet-wq:nvmet_rdma_release_queue_work [nvmet_rdma] is flushing !WQ_MEM_RECLAIM events:nvmet_keep_alive_timer [nvmet]

Chaitanya Kulkarni chaitanyak at nvidia.com
Fri Sep 23 10:34:35 PDT 2022


Thanks for reporting this Yi.

> [  350.946051] nvme nvme0: mapped 8/0/0 default/read/poll queues.
> [  350.946073] nvmet_rdma:nvmet_rdma_cm_handler: nvmet_rdma:
> established (9): status 0 id 00000000969b9167
> [  350.946285] nvmet:nvmet_execute_io_connect: nvmet: adding queue 1 to ctrl 1.
> [  350.946337] nvmet:nvmet_execute_io_connect: nvmet: adding queue 2 to ctrl 1.
> [  350.946378] nvmet:nvmet_execute_io_connect: nvmet: adding queue 3 to ctrl 1.
> [  350.946427] nvmet:nvmet_execute_io_connect: nvmet: adding queue 4 to ctrl 1.
> [  350.946479] nvmet:nvmet_execute_io_connect: nvmet: adding queue 5 to ctrl 1.
> [  350.946540] nvmet:nvmet_execute_io_connect: nvmet: adding queue 6 to ctrl 1.
> [  350.946595] nvmet:nvmet_execute_io_connect: nvmet: adding queue 7 to ctrl 1.
> [  350.946650] nvmet:nvmet_execute_io_connect: nvmet: adding queue 8 to ctrl 1.
> [  350.946708] nvme nvme0: new ctrl: NQN "nvme-test", addr 10.0.2.222:7777
> [  350.946739] nvmet:nvmet_req_cns_error_complete: nvmet: unhandled
> identify cns 6 on qid 0

admin-identify with cns failed I wonder why ?

hence it triggered following sequence of the errors

> [  350.960438] device-mapper: table: 253:0: multipath: error getting
> device (-EBUSY)
> [  350.960452] device-mapper: ioctl: error adding target to table
> [  351.074392] device-mapper: table: 253:1: multipath: error getting
> device (-EBUSY)
> [  351.074405] device-mapper: ioctl: error adding target to table
> [  355.638030] nvme nvme0: Removing ctrl: NQN "nvme-test"
> [  355.825297] nvmet_rdma:nvmet_rdma_cm_handler: nvmet_rdma:
> disconnected (10): status 0 id 000000000c220555
> [  355.825306] nvmet_rdma:__nvmet_rdma_queue_disconnect: nvmet_rdma:
> cm_id= 000000000c220555 queue->state= 1
> [  355.825316] nvmet_rdma:nvmet_rdma_free_queue: nvmet_rdma: freeing queue 1
> [  355.825360] nvmet_rdma:nvmet_rdma_cm_handler: nvmet_rdma:
> disconnected (10): status 0 id 00000000e169966e


Is is possible to bisect this Yi ?

-ck



More information about the Linux-nvme mailing list