On Thu, Feb 03 2011 at  8:24am -0500,
Jens Axboe [off-list ref] wrote:

On 2011-02-02 23:55, Mike Snitzer wrote:

quoted

Set REQ_SORTED, in the @rw_flags passed to the request allocator, for
any request that may be put on IO scheduler.  Skip elevator data
initialization during request allocation if REQ_SORTED is not set.

REQ_SORTED is not set for flush requests because they are never put on
the IO scheduler.

That looks very wrong. REQ_SORTED gets set _when_ the request is sorted
into the IO scheduler.

Yes, considerable oversight on my part.

This is gross misuse, a bad hack.

As my recently posted v4 stated, even without the bug it is still a hack.

My initial version didn't get so cute with overloading flags, etc, e.g:

diff --git a/block/blk-core.c b/block/blk-core.c
index 72dd23b..e098598 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c

@@ -764,7 +764,7 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
 	struct request_list *rl = &q->rq;
 	struct io_context *ioc = NULL;
 	const bool is_sync = rw_is_sync(rw_flags) != 0;
-	int may_queue, priv;
+	int may_queue, priv = 0;
 
 	may_queue = elv_may_queue(q, rw_flags);
 	if (may_queue == ELV_MQUEUE_NO)

@@ -808,9 +808,14 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
 	rl->count[is_sync]++;
 	rl->starved[is_sync] = 0;
 
-	priv = !test_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
-	if (priv)
-		rl->elvpriv++;
+	/*
+	 * Skip elevator data initialization for flush requests.
+	 */
+	if (!(bio && (bio->bi_rw & (REQ_FLUSH | REQ_FUA)))) {
+		priv = !test_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
+		if (priv)
+			rl->elvpriv++;
+	}
 
 	if (blk_queue_io_stat(q))
 		rw_flags |= REQ_IO_STAT;

I ran with Tejun's suggestion of overloading @rw_flags when it became
clear he wasn't happy with the above (because it spread FLUSH/FUA
special casing too thin).

Regardless, other ideas for how to convey this info to get_request()
would be appreciated.

Mike