On 7/8/20 3:27 AM, John Garry wrote:

On 03/07/2020 14:01, Hannes Reinecke wrote:

+linux-block

I figure that linux-block should be cc'ed here

quoted

Some drivers require to allocate requests for internal command
submission. These request will never be passed through the block
layer, but nevertheless require a valid tag to avoid them clashing
with normal I/O commands.
This patch adds a new request flag REQ_INTERNAL to mark such
requests and a terminates any such commands in blk_execute_rq_nowait()
with a WARN_ON_ONCE to signal such an invalid usage.

Signed-off-by: Hannes Reinecke <hare@suse.de>
---
  block/blk-exec.c          | 5 +++++
  include/linux/blk_types.h | 2 ++
  include/linux/blkdev.h    | 5 +++++
  3 files changed, 12 insertions(+)

diff --git a/block/blk-exec.c b/block/blk-exec.c
index 85324d53d072..6869877e0d21 100644
--- a/block/blk-exec.c
+++ b/block/blk-exec.c

@@ -55,6 +55,11 @@ void blk_execute_rq_nowait(struct request_queue *q, struct gendisk *bd_disk,
  	rq->rq_disk = bd_disk;
  	rq->end_io = done;
  
+	if (WARN_ON_ONCE(blk_rq_is_internal(rq))) {
+		blk_mq_end_request(rq, BLK_STS_NOTSUPP);
+		return;
+	}
+
  	blk_account_io_start(rq);

The whole concept seems very odd, and then there's this seemingly
randomly placed check and error condition. As I haven't seen the
actual use case for this, hard to make suggestions though.

-- 
Jens Axboe