Commit b4692083 authored Dec 04, 2018 by Ziye Yang Committed by Ben Walker Dec 05, 2018

nvme: Fix the race condition in nvme_ctrlr_get_cc



When the applications call spdk_nvme_ctrlr_alloc_io_qpair,
there will be cmd to the admin qpairs in nvme_ctrlr_get_cc,
so there is contention. We should use the lock to protect
nvme_ctrl_get_cc.  Otherwise, the multiple threads will have
contention on the admin qpair, thus there will be coredump issue.

We get the bug when testing NVMe-oF TCP transport, and this
patch can address this issue.

Change-Id: I7247f98cdf890c2eafaf8fb94580ecd714010bd5
Signed-off-by: Ziye Yang <optimistyzy@gmail.com>
Reviewed-on: https://review.gerrithub.io/435577


Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>

parent d6c0c192

lib/nvme/nvme_ctrlr.c

+4 −2

Original line number	Diff line number	Diff line
		@@ -248,13 +248,16 @@ spdk_nvme_ctrlr_alloc_io_qpair(struct spdk_nvme_ctrlr *ctrlr,
		memcpy(&opts, user_opts, spdk_min(sizeof(opts), opts_size));
		}

		nvme_robust_mutex_lock(&ctrlr->ctrlr_lock);
		if (nvme_ctrlr_get_cc(ctrlr, &cc)) {
		SPDK_ERRLOG("get_cc failed\n");
		nvme_robust_mutex_unlock(&ctrlr->ctrlr_lock);
		return NULL;
		}

		/* Only the low 2 bits (values 0, 1, 2, 3) of QPRIO are valid. */
		if ((opts.qprio & 3) != opts.qprio) {
		nvme_robust_mutex_unlock(&ctrlr->ctrlr_lock);
		return NULL;
		}

		@@ -264,11 +267,10 @@ spdk_nvme_ctrlr_alloc_io_qpair(struct spdk_nvme_ctrlr *ctrlr,
		*/
		if ((cc.bits.ams == SPDK_NVME_CC_AMS_RR) && (opts.qprio != SPDK_NVME_QPRIO_URGENT)) {
		SPDK_ERRLOG("invalid queue priority for default round robin arbitration method\n");
		nvme_robust_mutex_unlock(&ctrlr->ctrlr_lock);
		return NULL;
		}

		nvme_robust_mutex_lock(&ctrlr->ctrlr_lock);

		/*
		* Get the first available I/O queue ID.
		*/