Re: [PATCH v8 5/9] scsi: core: Retry unaligned zoned writes

On 8/15/23 02:57, Bart Van Assche wrote:

On 8/14/23 05:36, Damien Le Moal wrote:

quoted

On 8/12/23 06:35, Bart Van Assche wrote:

quoted

+	if (!rq->q->limits.use_zone_write_lock && blk_rq_is_seq_zoned_write(rq))

This condition could be written as a little inline helper
blk_req_need_zone_write_lock(), which could be used in mq-dealine patch 2.

Hi Damien,

Since q->limits.use_zone_write_lock is being introduced, how about
modifying blk_req_needs_zone_write_lock() such that it tests that new member
variable instead of checking rq->q->disk->seq_zones_wlock? That would allow
me to leave out one change from block/mq-deadline.c. I do not have a strong
opinion about whether the name blk_req_needs_zone_write_lock() should be
retained or whether that function should be renamed into
blk_req_use_zone_write_lock().

Something like this ?

diff --git a/block/blk-zoned.c b/block/blk-zoned.c
index 619ee41a51cc..a3980a71c0ac 100644
--- a/block/blk-zoned.c
+++ b/block/blk-zoned.c

@@ -57,7 +57,12 @@ EXPORT_SYMBOL_GPL(blk_zone_cond_str);
  */
 bool blk_req_needs_zone_write_lock(struct request *rq)
 {
-       if (!rq->q->disk->seq_zones_wlock)
+       struct request_queue *q = rq->q;
+
+       if (!q->limits.use_zone_write_lock)
+               return false;
+
+       if (!q->disk->seq_zones_wlock)
                return false;

        return blk_rq_is_seq_zoned_write(rq);

I think that is fine and avoids adding yet another helper.
I am OK with this.

-- 
Damien Le Moal
Western Digital Research