Skip to contentSkip to conversation

null_blk: Always split BIOs to respect queue limits

PATCH In review Damien Le Moal 4 participants 5 messages Jan 26, 2024 Block layer

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 null_blk: Always split BIOs to respect queue limits

The function null_submit_bio() used for null_blk devices configured with a BIO-based queue never splits BIOs according to the queue limits set with the various module and configfs parameters that the user can specify.

Add a call to bio_split_to_limits() to correctly handle large BIOs that need splitting. Doing so also fixes issues with zoned devices as a large BIO may cross over a zone boundary, which breaks null_blk zone emulation.

While at it, remove all the local variable that are not necessary.

---
 drivers/block/null_blk/main.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)
drivers/block/null_blk/main.c +9 -5
index 36755f263e8e..514c2592046a 100644
@@ -1528,12 +1528,16 @@ static struct nullb_queue *nullb_to_queue(struct nullb *nullb)
 
 static void null_submit_bio(struct bio *bio)
 {
-	sector_t sector = bio->bi_iter.bi_sector;
-	sector_t nr_sectors = bio_sectors(bio);
-	struct nullb *nullb = bio->bi_bdev->bd_disk->private_data;
-	struct nullb_queue *nq = nullb_to_queue(nullb);
+	struct nullb_queue *nq =
+		nullb_to_queue(bio->bi_bdev->bd_disk->private_data);
+
+	/* Respect the queue limits */
+	bio = bio_split_to_limits(bio);
+	if (!bio)
+		return;
 
-	null_handle_cmd(alloc_cmd(nq, bio), sector, nr_sectors, bio_op(bio));
+	null_handle_cmd(alloc_cmd(nq, bio), bio->bi_iter.bi_sector,
+			bio_sectors(bio), bio_op(bio));
 }
 
 #ifdef CONFIG_BLK_DEV_NULL_BLK_FAULT_INJECTION
Signature · 1 line
2.43.0
#2 lore raw
In reply to Hannes Reinecke
Quoted 6 lines · Hannes Reinecke
Quoted 5 lines · Damien Le Moal

The function null_submit_bio() used for null_blk devices configured with a BIO-based queue never splits BIOs according to the queue limits set with the various module and configfs parameters that the user can specify.

Quoted 7 lines
Quoted 5 lines

Add a call to bio_split_to_limits() to correctly handle large BIOs that need splitting. Doing so also fixes issues with zoned devices as a large BIO may cross over a zone boundary, which breaks null_blk zone emulation.

That feels so wrong. Why would we need to apply queue limits to a bio? (Yes, I know why. We still shouldn't be doing it.)

Splitting is at least needed for zoned devices. Otherwise, everything breaks with the zone emulation.

Quoted 1 line

Maybe indeed time to kill the bio-based path.

I have nothing against that :)

Quoted 8 lines

But until that happens:

Cheers,

Hannes

Signature · 2 lines
Damien Le Moal
Western Digital Research
Reviewed-by on v1 1/1
#1 lore raw
In reply to Damien Le Moal
Quoted 10 lines · Damien Le Moal

The function null_submit_bio() used for null_blk devices configured with a BIO-based queue never splits BIOs according to the queue limits set with the various module and configfs parameters that the user can specify.

Add a call to bio_split_to_limits() to correctly handle large BIOs that need splitting. Doing so also fixes issues with zoned devices as a large BIO may cross over a zone boundary, which breaks null_blk zone emulation.

That feels so wrong. Why would we need to apply queue limits to a bio? (Yes, I know why. We still shouldn't be doing it.)

Maybe indeed time to kill the bio-based path.

But until that happens:

Cheers,

Hannes

Signature · 5 lines
Dr. Hannes Reinecke                Kernel Storage Architect
[email protected]                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Ivo Totev, Andrew McDonald,
Werner Knoblich
Reviewed-by
#3 lore raw
In reply to Damien Le Moal
Quoted 12 lines · Damien Le Moal
Quoted 11 lines · Hannes Reinecke
Quoted 10 lines · Damien Le Moal

The function null_submit_bio() used for null_blk devices configured with a BIO-based queue never splits BIOs according to the queue limits set with the various module and configfs parameters that the user can specify.

Add a call to bio_split_to_limits() to correctly handle large BIOs that need splitting. Doing so also fixes issues with zoned devices as a large BIO may cross over a zone boundary, which breaks null_blk zone emulation.

Quoted 14 lines
Quoted 2 lines

That feels so wrong. Why would we need to apply queue limits to a bio? (Yes, I know why. We still shouldn't be doing it.)

Splitting is at least needed for zoned devices. Otherwise, everything breaks with the zone emulation.

Quoted 1 line

Maybe indeed time to kill the bio-based path.

I have nothing against that :)

Quoted 7 lines

But until that happens:

Cheers,

Hannes

If we are going to kill it we really don't need this patch, irrespective of that :-

-ck

Reviewed-by
#4 lore raw
In reply to Hannes Reinecke
Quoted 6 lines · Hannes Reinecke
Quoted 4 lines

BIOs that need splitting. Doing so also fixes issues with zoned devices as a large BIO may cross over a zone boundary, which breaks null_blk zone emulation.

That feels so wrong. Why would we need to apply queue limits to a bio? (Yes, I know why. We still shouldn't be doing it.)

Because a driver that has limits should enforce them. Your hardware doesn't suddenly use limits because you're using a bio based driver, and null_blk shouldn't suddenly ignore the configured limits just because you're using it in bio mode.

The patch looks good:

But yeah, I'd rather kill the bio mode. Jens, are you attached to the bio mode? Otherwise I'll cook up a patch over the weekend.