2022-10-06 13:54:22

by Christian Loehle

Subject: [PATCH 2/2] mmc: queue: Flush recovery work on cleanup

To prevent any recovery work running after the queue cleanup flush it.
Any recovery running post-cleanup dereferenced mq->card as NULL
and was not meaningful to begin with.

Cc: [email protected]

Signed-off-by: Christian Loehle <[email protected]>
---
drivers/mmc/core/queue.c | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
index fefaa901b50f..a1b985d3dfda 100644
--- a/drivers/mmc/core/queue.c
+++ b/drivers/mmc/core/queue.c
@@ -493,6 +493,13 @@ void mmc_cleanup_queue(struct mmc_queue *mq)
if (blk_queue_quiesced(q))
blk_mq_unquiesce_queue(q);

+ /*
+ * If the recovery completes the last (and only remaining) request in
+ * the queue, and the card has been removed, we could end up here with
+ * the recovery not quite finished yet, so flush it.
+ */
+ flush_work(&mq->recovery_work);
+
blk_mq_free_tag_set(&mq->tag_set);

/*
--
2.37.3

Hyperstone GmbH | Reichenaustr. 39a | 78467 Konstanz
Managing Director: Dr. Jan Peter Berns.
Commercial register of local courts: Freiburg HRB381782

2022-10-07 13:20:34

by Ulf Hansson

Subject: Re: [PATCH 2/2] mmc: queue: Flush recovery work on cleanup

On Thu, 6 Oct 2022 at 15:39, Christian Löhle <[email protected]> wrote:
>
> To prevent any recovery work running after the queue cleanup flush it.
> Any recovery running post-cleanup dereferenced mq->card as NULL
> and was not meaningful to begin with.
>
> Cc: [email protected]
>
> Signed-off-by: Christian Loehle <[email protected]>
> ---
> drivers/mmc/core/queue.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c
> index fefaa901b50f..a1b985d3dfda 100644
> --- a/drivers/mmc/core/queue.c
> +++ b/drivers/mmc/core/queue.c
> @@ -493,6 +493,13 @@ void mmc_cleanup_queue(struct mmc_queue *mq)
> if (blk_queue_quiesced(q))
> blk_mq_unquiesce_queue(q);
>
> + /*
> + * If the recovery completes the last (and only remaining) request in
> + * the queue, and the card has been removed, we could end up here with
> + * the recovery not quite finished yet, so flush it.
> + */
> + flush_work(&mq->recovery_work);
> +

Not sure if it really matters in this case, but isn't
cancel_work_sync() the more proper thing to call instead?

> blk_mq_free_tag_set(&mq->tag_set);
>
> /*

Kind regards
Uffe