* [PATCH] loop: change fsync to fdatasync when update dio [not found] <CGME20230126051713epcas1p10a9005ad21887893a486100cbbd376e5@epcas1p1.samsung.com> @ 2023-01-26 5:16 ` Huijin Park 2023-01-26 5:31 ` Christoph Hellwig 0 siblings, 1 reply; 3+ messages in thread From: Huijin Park @ 2023-01-26 5:16 UTC (permalink / raw) To: Jens Axboe, Christoph Hellwig, Chaitanya Kulkarni Cc: Ming Lei, linux-block, linux-kernel, huijin.park, bbanghj.park In general, fsync has a larger overhead than fdatasync. And since the dio option is for data, it seems like fdatasync is enough. So this patch changes it to fdatasync which has little load relatively. Signed-off-by: Huijin Park <huijin.park@samsung.com> diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 1518a6423279..6c7ce8be8c0c 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -203,7 +203,7 @@ static void __loop_update_dio(struct loop_device *lo, bool dio) return; /* flush dirty pages before changing direct IO */ - vfs_fsync(file, 0); + vfs_fsync(file, 1); /* * The flag of LO_FLAGS_DIRECT_IO is handled similarly with -- 2.17.1 ^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] loop: change fsync to fdatasync when update dio 2023-01-26 5:16 ` [PATCH] loop: change fsync to fdatasync when update dio Huijin Park @ 2023-01-26 5:31 ` Christoph Hellwig 2023-02-02 18:12 ` Huijin Park 0 siblings, 1 reply; 3+ messages in thread From: Christoph Hellwig @ 2023-01-26 5:31 UTC (permalink / raw) To: Huijin Park Cc: Jens Axboe, Christoph Hellwig, Chaitanya Kulkarni, linux-block, linux-kernel, bbanghj.park On Thu, Jan 26, 2023 at 02:16:57PM +0900, Huijin Park wrote: > In general, fsync has a larger overhead than fdatasync. And since the > dio option is for data, it seems like fdatasync is enough. > So this patch changes it to fdatasync which has little load relatively. The only difference is that fsync also syncs the timestamps. So this change looks correct, but also a bit useless given that buffered to direct I/O or back changes aren't exactly a fast path. ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] loop: change fsync to fdatasync when update dio 2023-01-26 5:31 ` Christoph Hellwig @ 2023-02-02 18:12 ` Huijin Park 0 siblings, 0 replies; 3+ messages in thread From: Huijin Park @ 2023-02-02 18:12 UTC (permalink / raw) To: Christoph Hellwig Cc: Huijin Park, Jens Axboe, Chaitanya Kulkarni, linux-block, linux-kernel On Thu, Jan 26, 2023 at 2:31 PM Christoph Hellwig <hch@lst.de> wrote: > > On Thu, Jan 26, 2023 at 02:16:57PM +0900, Huijin Park wrote: > > In general, fsync has a larger overhead than fdatasync. And since the > > dio option is for data, it seems like fdatasync is enough. > > So this patch changes it to fdatasync which has little load relatively. > > The only difference is that fsync also syncs the timestamps. So this > change looks correct, but also a bit useless given that buffered to > direct I/O or back changes aren't exactly a fast path. Although the difference will be minimal, why I suggested it is because it can reduce unnecessary metadata i/o (helpful on slow i/o devices), and using fdatasync looked correct like your opinion. In some environment cases, loop setup for mount is required when application is initialized and this change will help. ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2023-02-02 18:12 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <CGME20230126051713epcas1p10a9005ad21887893a486100cbbd376e5@epcas1p1.samsung.com>
2023-01-26 5:16 ` [PATCH] loop: change fsync to fdatasync when update dio Huijin Park
2023-01-26 5:31 ` Christoph Hellwig
2023-02-02 18:12 ` Huijin Park
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).