From cddbaa8dee7e7d4eb678e6f04fec1aa43117c7f4 Mon Sep 17 00:00:00 2001 From: Ming Lei Date: Tue, 11 Jul 2023 17:40:40 +0800 Subject: [PATCH] nvme-tcp: fix potential unbalanced freeze & unfreeze commit 99dc264014d5aed66ee37ddf136a38b5a2b1b529 upstream. Move start_freeze into nvme_tcp_configure_io_queues(), and there is at least two benefits: 1) fix unbalanced freeze and unfreeze, since re-connection work may fail or be broken by removal 2) IO during error recovery can be failfast quickly because nvme fabrics unquiesces queues after teardown. One side-effect is that !mpath request may timeout during connecting because of queue topo change, but that looks not one big deal: 1) same problem exists with current code base 2) compared with !mpath, mpath use case is dominant Fixes: 2875b0aecabe ("nvme-tcp: fix controller reset hang during traffic") Cc: stable@vger.kernel.org Signed-off-by: Ming Lei Tested-by: Yi Zhang Reviewed-by: Sagi Grimberg Signed-off-by: Keith Busch Signed-off-by: Greg Kroah-Hartman --- drivers/nvme/host/tcp.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c index 8f17cbec5a0e4..f2fedd25915f9 100644 --- a/drivers/nvme/host/tcp.c +++ b/drivers/nvme/host/tcp.c @@ -1890,6 +1890,7 @@ static int nvme_tcp_configure_io_queues(struct nvme_ctrl *ctrl, bool new) goto out_cleanup_connect_q; if (!new) { + nvme_start_freeze(ctrl); nvme_start_queues(ctrl); if (!nvme_wait_freeze_timeout(ctrl, NVME_IO_TIMEOUT)) { /* @@ -1898,6 +1899,7 @@ static int nvme_tcp_configure_io_queues(struct nvme_ctrl *ctrl, bool new) * to be safe. */ ret = -ENODEV; + nvme_unfreeze(ctrl); goto out_wait_freeze_timed_out; } blk_mq_update_nr_hw_queues(ctrl->tagset, @@ -2002,7 +2004,6 @@ static void nvme_tcp_teardown_io_queues(struct nvme_ctrl *ctrl, if (ctrl->queue_count <= 1) return; nvme_stop_admin_queue(ctrl); - nvme_start_freeze(ctrl); nvme_stop_queues(ctrl); nvme_sync_io_queues(ctrl); nvme_tcp_stop_io_queues(ctrl);