On Fri Mar 7, 2025 at 11:10 AM CET, zhang.guanghui@xxxxxxxx wrote: > > Hi > > After testing this patch, sending request failure occurred, unfortunately, the issue still persists. Maybe I am completely wrong but I am still quite convinced that the problem here is that, despite the nvme_tcp_try_send() failure, the target sends a response capsule for the command, leading to a double-completion in the host. Sagi, what about taking this patch: https://lore.kernel.org/linux-nvme/20250306160322.1370300-2-mlombard@xxxxxxxxxx/T/#u and do a step further by not completing the request, leaving the error recovery handler the task of cleaning everything up? diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c index 327f3f2f5399..72c1d7948386 100644 --- a/drivers/nvme/host/tcp.c +++ b/drivers/nvme/host/tcp.c @@ -1328,8 +1328,8 @@ static int nvme_tcp_try_send(struct nvme_tcp_queue *queue) } else if (ret < 0) { dev_err(queue->ctrl->ctrl.device, "failed to send request %d\n", ret); - nvme_tcp_fail_request(queue->request); nvme_tcp_done_send_req(queue); + nvme_tcp_error_recovery(&queue->ctrl->ctrl); } out: memalloc_noreclaim_restore(noreclaim_flag); Maurizio
