[PATCH 4.14-rc fix-regression] nvme-rdma: Fix possible double free in reconnect flow

[sagi@grimberg.me (Sagi Grimberg)]

The fact that we free the async event buffer in
nvme_rdma_destroy_admin_queue can cause us to free it
more than once because this happens in every reconnect
attempt (we rely on the queue state flags DELETING
to avoid this for other resources).

A more complete fix is to not destroy the admin/io queues
unconditionally on every reconnect attempt, but its a bit
more extensive and will go in the next release.

Reported-by: Yi Zhang <yi.zhang at redhat.com>
Signed-off-by: Sagi Grimberg <sagi at grimberg.me>
---
Note that this might cause a minor conflict with nvme-4.15
patches, but it will be easy enough to resolve it.

 drivers/nvme/host/rdma.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index 92a03ff5fb4d..ec22da478ba3 100644
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -571,6 +571,11 @@ static void nvme_rdma_free_queue(struct nvme_rdma_queue *queue)
 	if (test_and_set_bit(NVME_RDMA_Q_DELETING, &queue->flags))
 		return;
 
+	if(nvme_rdma_queue_idx(queue) == 0)
+		nvme_rdma_free_qe(queue->device->dev,
+			&queue->ctrl->async_event_sqe,
+			sizeof(struct nvme_command), DMA_TO_DEVICE);
+
 	nvme_rdma_destroy_queue_ib(queue);
 	rdma_destroy_id(queue->cm_id);
 }
@@ -739,8 +744,6 @@ static struct blk_mq_tag_set *nvme_rdma_alloc_tagset(struct nvme_ctrl *nctrl,
 static void nvme_rdma_destroy_admin_queue(struct nvme_rdma_ctrl *ctrl,
 		bool remove)
 {
-	nvme_rdma_free_qe(ctrl->queues[0].device->dev, &ctrl->async_event_sqe,
-			sizeof(struct nvme_command), DMA_TO_DEVICE);
 	nvme_rdma_stop_queue(&ctrl->queues[0]);
 	if (remove) {
 		blk_cleanup_queue(ctrl->ctrl.admin_q);
-- 
2.7.4