On 6/25/25 4:38 PM, Damien Le Moal wrote:
On 6/26/25 00:48, Bart Van Assche wrote:
On 6/25/25 2:33 AM, Damien Le Moal wrote:
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 4806b867e37d..0c61492724d2 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -3169,8 +3169,10 @@ void blk_mq_submit_bio(struct bio *bio)
if (blk_mq_attempt_bio_merge(q, bio, nr_segs))
goto queue_exit;
- if (blk_queue_is_zoned(q) && blk_zone_plug_bio(bio, nr_segs))
- goto queue_exit;
+ if (bio_needs_zone_write_plugging(bio)) {
+ if (blk_zone_plug_bio(bio, nr_segs))
+ goto queue_exit;
+ }
Why nested if-statements instead of keeping "&&"? I prefer "&&".
I did this because bio_needs_zone_write_plugging() is inline and
blk_zone_plug_bio() is not, so this ensures that we do not have the function
call for nothing. Though I may be overthinking this since normally, the
generated assembler will not test the second part of a && condition if the first
part is false already.
From a (draft version of) the C standard, section "6.5.14 Logical AND
operator": "If the first operand compares equal to 0, the second operand
is not evaluated." I think that text was already present in the
K&R C book.
See also https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3220.pdf.
Thanks,
Bart.
