Commit de3eb36a authored Feb 20, 2019 by Changpeng Liu Committed by Jim Harris Feb 22, 2019

bdev/nvme: exit controller removal callback if the destruction was started



For real PCIe drives, if we removed one drive, existing hotplug
monitor will trigger the remove callback twice, there is one
workaround for vfio-attached device hot remove detection which
will also trigger the hot removal callback.  For now we add
the check in the bdev_nvme layer so that coredump will not happen.

Fix issue #606.

Change-Id: I0605fbdf391fed20c4aa9a2d54b4f059f29dc483
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/445642


Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>

parent 1e5b58a1

lib/bdev/nvme/bdev_nvme.c

+5 −1

Original line number	Diff line number	Diff line
		@@ -1099,6 +1099,11 @@ remove_cb(void cb_ctx, struct spdk_nvme_ctrlr ctrlr)
		pthread_mutex_lock(&g_bdev_nvme_mutex);
		TAILQ_FOREACH(nvme_ctrlr, &g_nvme_ctrlrs, tailq) {
		if (nvme_ctrlr->ctrlr == ctrlr) {
		/* The controller's destruction was already started */
		if (nvme_ctrlr->destruct) {
		pthread_mutex_unlock(&g_bdev_nvme_mutex);
		return;
		}
		pthread_mutex_unlock(&g_bdev_nvme_mutex);
		for (i = 0; i < nvme_ctrlr->num_ns; i++) {
		uint32_t nsid = i + 1;
		@@ -1111,7 +1116,6 @@ remove_cb(void cb_ctx, struct spdk_nvme_ctrlr ctrlr)
		}

		pthread_mutex_lock(&g_bdev_nvme_mutex);
		assert(!nvme_ctrlr->destruct);
		nvme_ctrlr->destruct = true;
		if (nvme_ctrlr->ref == 0) {
		pthread_mutex_unlock(&g_bdev_nvme_mutex);