[PATCH 1/2] block: add a helper to cancel atomic queue limit updates

[PATCH 1/2] block: add a helper to cancel atomic queue limit updates

[Christoph Hellwig]

Drivers might have to perform complex actions to determine queue limits,
and those might fail.  Add a helper to cancel a queue limit update
that can be called in those cases.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/blkdev.h | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index f9b87c39cab047..39fedc8ef9c41f 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -892,6 +892,19 @@ int queue_limits_commit_update(struct request_queue *q,
 		struct queue_limits *lim);
 int queue_limits_set(struct request_queue *q, struct queue_limits *lim);
 
+/**
+ * queue_limits_cancel_update - cancel an atomic update of queue limits
+ * @q:		queue to update
+ *
+ * This functions cancels an atomic update of the queue limits started by
+ * queue_limits_start_update() and should be used when an error occurs after
+ * starting update.
+ */
+static inline void queue_limits_cancel_update(struct request_queue *q)
+{
+	mutex_lock(&q->limits_lock);
+}
+
 /*
  * Access functions for manipulating queue properties
  */
-- 
2.39.2

[PATCH 2/2] nvme: cancel the queue limit update when nvme_update_zone_info fails

[Christoph Hellwig]

Starting an atomic queue limits update takes a mutex and thus needs
to be finished, or with the newly added helper, canceled to not leak
the lock critical section.

Fixes: 9b130d681443 ("nvme: use the atomic queue limits update API")
Reported-by: Kanchan Joshi <joshi.k@samsung.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/nvme/host/core.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 943d72bdd794ca..f8a9565bee41d2 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -2115,6 +2115,7 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 	    ns->head->ids.csi == NVME_CSI_ZNS) {
 		ret = nvme_update_zone_info(ns, lbaf, &lim);
 		if (ret) {
+			queue_limits_cancel_update(ns->disk->queue);
 			blk_mq_unfreeze_queue(ns->disk->queue);
 			goto out;
 		}
-- 
2.39.2

Re: [PATCH 2/2] nvme: cancel the queue limit update when nvme_update_zone_info fails

[Keith Busch]

On Wed, Mar 27, 2024 at 06:21:45PM +0100, Christoph Hellwig wrote:
> @@ -2115,6 +2115,7 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
>  	    ns->head->ids.csi == NVME_CSI_ZNS) {
>  		ret = nvme_update_zone_info(ns, lbaf, &lim);
>  		if (ret) {
> +			queue_limits_cancel_update(ns->disk->queue);

Could you instead move nvme_update_zone_info() outside the
queue_limits_start_update()? That way we wouldn't need to "cancel" the
update. You'd just need to save a copy of "mor" and "mar" in the ns
instead of writing these directly to the queue_limits.

Re: [PATCH 1/2] block: add a helper to cancel atomic queue limit updates

[Kanchan Joshi]

On 3/27/2024 10:51 PM, Christoph Hellwig wrote:
> Drivers might have to perform complex actions to determine queue limits,
> and those might fail.  Add a helper to cancel a queue limit update
> that can be called in those cases.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>   include/linux/blkdev.h | 13 +++++++++++++
>   1 file changed, 13 insertions(+)
> 
> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> index f9b87c39cab047..39fedc8ef9c41f 100644
> --- a/include/linux/blkdev.h
> +++ b/include/linux/blkdev.h
> @@ -892,6 +892,19 @@ int queue_limits_commit_update(struct request_queue *q,
>   		struct queue_limits *lim);
>   int queue_limits_set(struct request_queue *q, struct queue_limits *lim);
>   
> +/**
> + * queue_limits_cancel_update - cancel an atomic update of queue limits
> + * @q:		queue to update
> + *
> + * This functions cancels an atomic update of the queue limits started by
> + * queue_limits_start_update() and should be used when an error occurs after
> + * starting update.
> + */
> +static inline void queue_limits_cancel_update(struct request_queue *q)
> +{
> +	mutex_lock(&q->limits_lock);

mutex_unlock is needed here.