On 11/26/19 5:09 PM, Jaegeuk Kim wrote:
+ if (drop_request)
+ blk_set_queue_dying(lo->lo_queue);
/* I/O need to be drained during transfer transition */
blk_mq_freeze_queue(lo->lo_queue);
Since blk_set_queue_dying() calls blk_freeze_queue_start(), I think the
above code will increase q->mq_freeze_depth by one or by two depending
on which path is taken. How about changing the above code into the
following:
if (drop_request) {
blk_set_queue_dying(lo->lo_queue);
blk_mq_freeze_queue_wait(lo->lo_queue);
} else {
blk_mq_freeze_queue(lo->lo_queue);
}
Otherwise this patch looks good to me.
Thanks,
Bart.