[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[dm-devel] [PATCH 2/4] dm-userspace: use kernel threads for do_map_bio



No need to care about I/O reordering, so kernel threads can performs
do_map_bio (that might provide the better performance).

Signed-off-by: FUJITA Tomonori <fujita tomonori lab ntt co jp>
---
 drivers/md/dm-userspace-chardev.c |    9 ---------
 1 files changed, 0 insertions(+), 9 deletions(-)

diff --git a/drivers/md/dm-userspace-chardev.c b/drivers/md/dm-userspace-chardev.c
index 869e359..1853d0d 100644
--- a/drivers/md/dm-userspace-chardev.c
+++ b/drivers/md/dm-userspace-chardev.c
@@ -358,17 +358,8 @@ static void do_map_bio(struct dmu_device
 
 	memcpy(&req->response, msg, sizeof(req->response));
 
-#if 0
-	/* I think it would be nice to farm this out to a worker
-	 * thread, so that userspace does not have to do all the work,
-	 * but I wonder about the correctness of possibly reordering
-	 * requests to a single block
-	 */
 	INIT_WORK(&req->task, map_worker, req);
 	schedule_work(&req->task);
-#else
-	map_worker(req);
-#endif
 }
 
 static void do_map_done(struct dmu_device *dev, uint64_t id_of_op, int fail)
-- 
1.4.1


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]