Linux-Block Archive mirror
 help / color / mirror / Atom feed
From: Uday Shankar <ushankar@purestorage.com>
To: Ming Lei <ming.lei@redhat.com>, Jens Axboe <axboe@kernel.dk>
Cc: linux-block@vger.kernel.org,
	Uday Shankar <ushankar@purestorage.com>,
	Riley Thomasson <riley@purestorage.com>
Subject: [PATCH] ublk: remove segment count and size limits
Date: Mon, 29 Apr 2024 18:53:31 -0600	[thread overview]
Message-ID: <20240430005330.2786014-1-ushankar@purestorage.com> (raw)

ublk_drv currently creates block devices with the default max_segments
and max_segment_size limits of BLK_MAX_SEGMENTS (128) and
BLK_MAX_SEGMENT_SIZE (65536) respectively. These defaults can
artificially constrain the I/O size seen by the ublk server - for
example, suppose that the ublk server has configured itself to accept
I/Os up to 1M and the application is also issuing 1M sized I/Os. If the
I/O buffer used by the application is backed by 4K pages, the buffer
could consist of up to 1M / 4K = 256 physically discontiguous segments
(even if the buffer is virtually contiguous). As such, the I/O could
exceed the default max_segments limit and get split. This can cause
unnecessary performance issues if the ublk server is optimized to handle
1M I/Os. The block layer's segment count/size limits exist to model
hardware constraints which don't exist in ublk_drv's case, so just
remove those limits for the block devices created by ublk_drv.

Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Reviewed-by: Riley Thomasson <riley@purestorage.com>
---
 drivers/block/ublk_drv.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index bea3d5cf8a83..835b0cc7c032 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -2209,6 +2209,9 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd)
 		lim.max_zone_append_sectors = p->max_zone_append_sectors;
 	}
 
+	lim.max_segments = USHRT_MAX;
+	lim.max_segment_size = UINT_MAX;
+
 	if (wait_for_completion_interruptible(&ub->completion) != 0)
 		return -EINTR;
 
-- 
2.34.1


             reply	other threads:[~2024-04-30  0:54 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-04-30  0:53 Uday Shankar [this message]
2024-04-30  9:15 ` [PATCH] ublk: remove segment count and size limits Ming Lei
2024-04-30 12:59 ` Christoph Hellwig

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240430005330.2786014-1-ushankar@purestorage.com \
    --to=ushankar@purestorage.com \
    --cc=axboe@kernel.dk \
    --cc=linux-block@vger.kernel.org \
    --cc=ming.lei@redhat.com \
    --cc=riley@purestorage.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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 read-only IMAP folder(s) and NNTP newsgroup(s).