nvme-pci: Fix a race in controller removal

User space programs like udevd may try to read to partitions at the
same time the driver detects a namespace is unusable, and may deadlock
if revalidate_disk() is called while such a process is waiting to
enter the frozen queue. On detecting a dead namespace, move the disk
revalidate after unblocking dispatchers that may be holding bd_butex.

changelog Suggested-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Balbir Singh <sblbir@amzn.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 108f60b46804a440c6acf07e336d763c4cad29f3..0c385b1994fe0e36d0351978fb36c19942129fe9 100644 (file)
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -102,10 +102,13 @@ static void nvme_set_queue_dying(struct nvme_ns *ns)
         */
        if (!ns->disk || test_and_set_bit(NVME_NS_DEAD, &ns->flags))
                return;
-       revalidate_disk(ns->disk);
        blk_set_queue_dying(ns->queue);
        /* Forcibly unquiesce queues to avoid blocking dispatch */
        blk_mq_unquiesce_queue(ns->queue);
+       /*
+        * Revalidate after unblocking dispatchers that may be holding bd_butex
+        */
+       revalidate_disk(ns->disk);
 }
 
 static void nvme_queue_scan(struct nvme_ctrl *ctrl)