d/numam-spdk
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

nvmf/rdma: On getting a wc error, force the qpair into the error state

Browse Source 
This initiates an error recovery instead of a disconnect. The
error recovery may result in a disconnect if the qpair is not
recoverable. This also resolves an issue where the disconnect
may immediately release the resources associated with the rqpair,
but upcoming wc entries may still reference it.

Change-Id: I9d9e212a83129412e049c91c02725699ce2cac11
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/425010
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This commit is contained in:
Ben Walker 2018-09-07 11:16:32 -07:00 committed by Jim Harris
parent 756bf3be20
commit e06896b94c
1 changed files with 5 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
lib/nvmf/rdma.c
View File

@ -2588,7 +2588,7 @@ spdk_nvmf_rdma_poller_poll(struct spdk_nvmf_rdma_transport *rtransport,
rdma_req = get_rdma_req_from_wc(&wc[i]);
rqpair = SPDK_CONTAINEROF(rdma_req->req.qpair, struct spdk_nvmf_rdma_qpair, qpair);
/* We're going to kill the connection, so force the request into
/* We're going to attempt an error recovery, so force the request into
* the completed state. */
spdk_nvmf_rdma_request_set_state(rdma_req, RDMA_REQUEST_STATE_COMPLETED);
spdk_nvmf_rdma_request_process(rtransport, rdma_req);
@ -2598,17 +2598,16 @@ spdk_nvmf_rdma_poller_poll(struct spdk_nvmf_rdma_transport *rtransport,
rqpair = rdma_recv->qpair;
/* Dump this into the incoming queue. This gets cleaned up when
* the queue pair disconnects. */
* the queue pair disconnects or recovers. */
TAILQ_INSERT_TAIL(&rqpair->incoming_queue, rdma_recv, link);
break;
default:
SPDK_ERRLOG("Received an unknown opcode on the CQ: %d\n", wc[i].opcode);
continue;
}
/* Begin disconnecting the qpair. This is ok to call multiple times if lots of
* errors occur on the same qpair in the same ibv_poll_cq batch. */
spdk_nvmf_qpair_disconnect(&rqpair->qpair, NULL, NULL);
/* Set the qpair to the error state. This will initiate a recovery. */
spdk_nvmf_rdma_set_ibv_state(rqpair, IBV_QPS_ERR);
continue;
}