[PATCH 1/1] block: return BLK_STS_NOTSUPP if operation is not supported

[PATCH 1/1] block: return BLK_STS_NOTSUPP if operation is not supported

[Ritika Srivastava]

If WRITE_ZERO/WRITE_SAME operation is not supported by the storage,
blk_cloned_rq_check_limits() will return -EIO which will cause
device-mapper to fail the paths.

Instead, if the queue limit is set to 0, return BLK_STS_NOTSUPP.
BLK_STS_NOTSUPP will be ignored by device-mapper and will not fail the
paths.

Suggested-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Ritika Srivastava <ritika.srivastava@oracle.com>
---
 block/blk-core.c | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 9bfaee0..173bb04 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1223,10 +1223,13 @@ blk_qc_t submit_bio(struct bio *bio)
 static int blk_cloned_rq_check_limits(struct request_queue *q,
 				      struct request *rq)
 {
-	if (blk_rq_sectors(rq) > blk_queue_get_max_sectors(q, req_op(rq))) {
+	unsigned int queue_max_sector = blk_queue_get_max_sectors(q, req_op(rq));
+
+	if (blk_rq_sectors(rq) > queue_max_sector) {
 		printk(KERN_ERR "%s: over max size limit. (%u > %u)\n",
-			__func__, blk_rq_sectors(rq),
-			blk_queue_get_max_sectors(q, req_op(rq)));
+			__func__, blk_rq_sectors(rq), queue_max_sector);
+		if (queue_max_sector == 0)
+			return -EOPNOTSUPP;
 		return -EIO;
 	}
 
@@ -1253,7 +1256,11 @@ static int blk_cloned_rq_check_limits(struct request_queue *q,
  */
 blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *rq)
 {
-	if (blk_cloned_rq_check_limits(q, rq))
+	int cloned_limit_check = blk_cloned_rq_check_limits(q, rq);
+
+	if (cloned_limit_check == -EOPNOTSUPP)
+		return BLK_STS_NOTSUPP;
+	else if (cloned_limit_check)
 		return BLK_STS_IOERR;
 
 	if (rq->rq_disk &&
-- 
1.8.3.1

Re: [PATCH 1/1] block: return BLK_STS_NOTSUPP if operation is not supported

[Christoph Hellwig]

On Fri, Jul 24, 2020 at 09:33:22AM -0700, Ritika Srivastava wrote:
> If WRITE_ZERO/WRITE_SAME operation is not supported by the storage,
> blk_cloned_rq_check_limits() will return -EIO which will cause
> device-mapper to fail the paths.
> 
> Instead, if the queue limit is set to 0, return BLK_STS_NOTSUPP.
> BLK_STS_NOTSUPP will be ignored by device-mapper and will not fail the
> paths.

How do we end up calling blk_cloned_rq_check_limits when the operations
aren't actually supported?  The upper layers should make sure this never
happens, so you're just fixing symptoms here and not the root cause.

Re: [PATCH 1/1] block: return BLK_STS_NOTSUPP if operation is not supported

[Ritika Srivastava]

Hi Christoph,

Thank you for the review.

> On Jul 26, 2020, at 8:10 AM, Christoph Hellwig <hch@infradead.org> wrote:
> 
> On Fri, Jul 24, 2020 at 09:33:22AM -0700, Ritika Srivastava wrote:
>> If WRITE_ZERO/WRITE_SAME operation is not supported by the storage,
>> blk_cloned_rq_check_limits() will return -EIO which will cause
>> device-mapper to fail the paths.
>> 
>> Instead, if the queue limit is set to 0, return BLK_STS_NOTSUPP.
>> BLK_STS_NOTSUPP will be ignored by device-mapper and will not fail the
>> paths.
> 
> How do we end up calling blk_cloned_rq_check_limits when the operations
> aren't actually supported?  The upper layers should make sure this never
> happens, so you're just fixing symptoms here and not the root cause.
The device advertises it’s write zero value as non zero value:
cat /sys/block/sdh/queue/write_zeroes_max_bytes
33553920

Hence block layer issues write zeroes in blkdev_issue_zeroout()

In response, the storage returns the following SCSI error
Sense Key : Illegal Request [current]
Add. Sense: Invalid command operation code

Once this error is received, the write zero capability of the device is disabled and write_zeroes_max_bytes is set to 0.
DM device queue limits are also set to 0 and device-mapper fails the path.
To avoid this, if BLK_STS_NOTSUPP could be returned, then device-mapper would not fail the paths.

Once the write zero capability has been disabled, blk-lib issues zeroes via __blkdev_issue_zero_pages().

Please let me know if I missed something.

Re: [PATCH 1/1] block: return BLK_STS_NOTSUPP if operation is not supported

[Christoph Hellwig]

On Mon, Jul 27, 2020 at 12:11:19PM -0700, Ritika Srivastava wrote:
> Hence block layer issues write zeroes in blkdev_issue_zeroout()
> 
> In response, the storage returns the following SCSI error
> Sense Key : Illegal Request [current]
> Add. Sense: Invalid command operation code
> 
> Once this error is received, the write zero capability of the device is disabled and write_zeroes_max_bytes is set to 0.
> DM device queue limits are also set to 0 and device-mapper fails the path.
> To avoid this, if BLK_STS_NOTSUPP could be returned, then device-mapper would not fail the paths.
> 
> Once the write zero capability has been disabled, blk-lib issues zeroes via __blkdev_issue_zero_pages().
> 
> Please let me know if I missed something.

Oh, the stupid SCSI runtime detection case again.  Can you please
document this in a comment?

Also please switch blk_cloned_rq_check_limits to just return the
blk_status_t directly.

Re: [PATCH 1/1] block: return BLK_STS_NOTSUPP if operation is not supported

[Ritika Srivastava]

> On Jul 28, 2020, at 12:28 AM, Christoph Hellwig <hch@infradead.org> wrote:
> 
> On Mon, Jul 27, 2020 at 12:11:19PM -0700, Ritika Srivastava wrote:
>> Hence block layer issues write zeroes in blkdev_issue_zeroout()
>> 
>> In response, the storage returns the following SCSI error
>> Sense Key : Illegal Request [current]
>> Add. Sense: Invalid command operation code
>> 
>> Once this error is received, the write zero capability of the device is disabled and write_zeroes_max_bytes is set to 0.
>> DM device queue limits are also set to 0 and device-mapper fails the path.
>> To avoid this, if BLK_STS_NOTSUPP could be returned, then device-mapper would not fail the paths.
>> 
>> Once the write zero capability has been disabled, blk-lib issues zeroes via __blkdev_issue_zero_pages().
>> 
>> Please let me know if I missed something.
> 
> Oh, the stupid SCSI runtime detection case again.  Can you please
> document this in a comment?
> 
> Also please switch blk_cloned_rq_check_limits to just return the
> blk_status_t directly.

Sure, I will make these changes and send updated version.

Thanks,
Ritika