Commit 0525af71 authored by Israel Rukshin's avatar Israel Rukshin Committed by Jens Axboe
Browse files

nvme-rdma: remove timeout for getting RDMA-CM established event 



In case many controllers start error recovery at the same time (i.e.,
when port is down and up), they may never succeed to reconnect again.
This is because the target can't handle all the connect requests at
three seconds (the arbitrary value set today). Even if some of the
connections are established, when a single queue fails to connect,
all the controller's queues are destroyed as well. So, on the
following reconnection attempts the number of connect requests may
remain the same. To fix this, remove the timeout and wait for RDMA-CM
event to abort/complete the connect request. RDMA-CM sends unreachable
event when a timeout of ~90 seconds is expired. This approach is used
at other RDMA-CM users like SRP and iSER at blocking mode. The commit
also renames NVME_RDMA_CONNECT_TIMEOUT_MS to NVME_RDMA_CM_TIMEOUT_MS.

Signed-off-by: default avatarIsrael Rukshin <israelr@nvidia.com>
Reviewed-by: default avatarMax Gurtovoy <mgurtovoy@nvidia.com>
Acked-by: default avatarSagi Grimberg <sagi@grimberg.me>
Signed-off-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
parent 7012eef5

drivers/nvme/host/rdma.c

+5 −8
Original line number Original line Diff line number Diff line
@@ -29,7 +29,7 @@ 
#include "fabrics.h"

#include "fabrics.h"









#define NVME_RDMA_CONNECT_TIMEOUT_MS	3000		/* 3 second */

#define NVME_RDMA_CM_TIMEOUT_MS		3000		/* 3 second */





#define NVME_RDMA_MAX_SEGMENTS		256

#define NVME_RDMA_MAX_SEGMENTS		256
@@ -248,12 +248,9 @@ static int nvme_rdma_wait_for_cm(struct nvme_rdma_queue *queue) 
{

{

	int ret;

	int ret;





	ret = wait_for_completion_interruptible_timeout(&queue->cm_done,

	ret = wait_for_completion_interruptible(&queue->cm_done);

			msecs_to_jiffies(NVME_RDMA_CONNECT_TIMEOUT_MS) + 1);

	if (ret)

	if (ret < 0)

		return ret;

		return ret;

	if (ret == 0)

		return -ETIMEDOUT;

	WARN_ON_ONCE(queue->cm_error > 0);

	WARN_ON_ONCE(queue->cm_error > 0);

	return queue->cm_error;

	return queue->cm_error;

}

}
@@ -612,7 +609,7 @@ static int nvme_rdma_alloc_queue(struct nvme_rdma_ctrl *ctrl, 
	queue->cm_error = -ETIMEDOUT;

	queue->cm_error = -ETIMEDOUT;

	ret = rdma_resolve_addr(queue->cm_id, src_addr,

	ret = rdma_resolve_addr(queue->cm_id, src_addr,

			(struct sockaddr *)&ctrl->addr,

			(struct sockaddr *)&ctrl->addr,

			NVME_RDMA_CONNECT_TIMEOUT_MS);

			NVME_RDMA_CM_TIMEOUT_MS);

	if (ret) {

	if (ret) {

		dev_info(ctrl->ctrl.device,

		dev_info(ctrl->ctrl.device,

			"rdma_resolve_addr failed (%d).\n", ret);

			"rdma_resolve_addr failed (%d).\n", ret);
@@ -1895,7 +1892,7 @@ static int nvme_rdma_addr_resolved(struct nvme_rdma_queue *queue) 




	if (ctrl->opts->tos >= 0)

	if (ctrl->opts->tos >= 0)

		rdma_set_service_type(queue->cm_id, ctrl->opts->tos);

		rdma_set_service_type(queue->cm_id, ctrl->opts->tos);

	ret = rdma_resolve_route(queue->cm_id, NVME_RDMA_CONNECT_TIMEOUT_MS);

	ret = rdma_resolve_route(queue->cm_id, NVME_RDMA_CM_TIMEOUT_MS);

	if (ret) {

	if (ret) {

		dev_err(ctrl->device, "rdma_resolve_route failed (%d).\n",

		dev_err(ctrl->device, "rdma_resolve_route failed (%d).\n",

			queue->cm_error);

			queue->cm_error);