From: Bart Van Assche <bvanassche@acm.org>
To: Jens Axboe <axboe@kernel.dk>
Cc: linux-block@vger.kernel.org, linux-scsi@vger.kernel.org,
Christoph Hellwig <hch@lst.de>,
Damien Le Moal <dlemoal@kernel.org>,
Jaegeuk Kim <jaegeuk@kernel.org>,
Bart Van Assche <bvanassche@acm.org>,
"Bao D. Nguyen" <quic_nguyenb@quicinc.com>,
Can Guo <quic_cang@quicinc.com>,
"Martin K. Petersen" <martin.petersen@oracle.com>,
Avri Altman <avri.altman@wdc.com>
Subject: [PATCH v16 26/26] scsi: ufs: Inform the block layer about write ordering
Date: Mon, 18 Nov 2024 16:28:15 -0800 [thread overview]
Message-ID: <20241119002815.600608-27-bvanassche@acm.org> (raw)
In-Reply-To: <20241119002815.600608-1-bvanassche@acm.org>
From the UFSHCI 4.0 specification, about the legacy (single queue) mode:
"The host controller always process transfer requests in-order according
to the order submitted to the list. In case of multiple commands with
single doorbell register ringing (batch mode), The dispatch order for
these transfer requests by host controller will base on their index in
the List. A transfer request with lower index value will be executed
before a transfer request with higher index value."
From the UFSHCI 4.0 specification, about the MCQ mode:
"Command Submission
1. Host SW writes an Entry to SQ
2. Host SW updates SQ doorbell tail pointer
Command Processing
3. After fetching the Entry, Host Controller updates SQ doorbell head
pointer
4. Host controller sends COMMAND UPIU to UFS device"
In other words, for both legacy and MCQ mode, UFS controllers are
required to forward commands to the UFS device in the order these
commands have been received from the host.
Notes:
- For legacy mode this is only correct if the host submits one
command at a time. The UFS driver does this.
- Also in legacy mode, the command order is not preserved if
auto-hibernation is enabled in the UFS controller.
This patch improves performance as follows on a test setup with UFSHCI
3.0 controller:
- With the mq-deadline scheduler: 2.5x more IOPS for small writes.
- When not using an I/O scheduler compared to using mq-deadline with
zone locking: 4x more IOPS for small writes.
Cc: Bao D. Nguyen <quic_nguyenb@quicinc.com>
Cc: Can Guo <quic_cang@quicinc.com>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Avri Altman <avri.altman@wdc.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
---
drivers/ufs/core/ufshcd.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index abbe7135a977..a6dec3b7e3fd 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -5241,6 +5241,13 @@ static int ufshcd_device_configure(struct scsi_device *sdev,
struct ufs_hba *hba = shost_priv(sdev->host);
struct request_queue *q = sdev->request_queue;
+ /*
+ * With auto-hibernation disabled, the write order is preserved per
+ * MCQ. Auto-hibernation may cause write reordering that results in
+ * unaligned write errors. The SCSI core will retry the failed writes.
+ */
+ lim->driver_preserves_write_order = true;
+
lim->dma_pad_mask = PRDT_DATA_BYTE_COUNT_PAD - 1;
/*
next prev parent reply other threads:[~2024-11-19 0:29 UTC|newest]
Thread overview: 73+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-11-19 0:27 [PATCH v16 00/26] Improve write performance for zoned UFS devices Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 01/26] blk-zoned: Fix a reference count leak Bart Van Assche
2024-11-19 2:23 ` Damien Le Moal
2024-11-19 20:21 ` Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 02/26] blk-zoned: Split disk_zone_wplugs_work() Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 03/26] blk-zoned: Split queue_zone_wplugs_show() Bart Van Assche
2024-11-19 2:25 ` Damien Le Moal
2024-11-19 0:27 ` [PATCH v16 04/26] blk-zoned: Only handle errors after pending zoned writes have completed Bart Van Assche
2024-11-19 2:50 ` Damien Le Moal
2024-11-19 20:51 ` Bart Van Assche
2024-11-21 3:23 ` Damien Le Moal
2024-11-21 17:43 ` Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 05/26] blk-zoned: Fix a deadlock triggered by unaligned writes Bart Van Assche
2024-11-19 2:57 ` Damien Le Moal
2024-11-19 21:04 ` Bart Van Assche
2024-11-21 3:32 ` Damien Le Moal
2024-11-21 17:51 ` Bart Van Assche
2024-11-25 4:00 ` Damien Le Moal
2024-11-25 4:19 ` Damien Le Moal
2025-01-09 19:11 ` Bart Van Assche
2025-01-10 5:07 ` Damien Le Moal
2025-01-10 18:17 ` Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 06/26] blk-zoned: Fix requeuing of zoned writes Bart Van Assche
2024-11-19 3:00 ` Damien Le Moal
2024-11-19 21:06 ` Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 07/26] block: Support block drivers that preserve the order of write requests Bart Van Assche
2024-11-19 7:37 ` Damien Le Moal
2024-11-19 21:08 ` Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 08/26] dm-linear: Report to the block layer that the write order is preserved Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 09/26] mq-deadline: Remove a local variable Bart Van Assche
2024-11-19 7:38 ` Damien Le Moal
2024-11-19 21:11 ` Bart Van Assche
2024-11-19 0:27 ` [PATCH v16 10/26] blk-mq: Clean up blk_mq_requeue_work() Bart Van Assche
2024-11-19 7:39 ` Damien Le Moal
2024-11-19 0:28 ` [PATCH v16 11/26] block: Optimize blk_mq_submit_bio() for the cache hit scenario Bart Van Assche
2024-11-19 7:40 ` Damien Le Moal
2024-11-19 0:28 ` [PATCH v16 12/26] block: Rework request allocation in blk_mq_submit_bio() Bart Van Assche
2024-11-19 7:44 ` Damien Le Moal
2024-11-19 0:28 ` [PATCH v16 13/26] block: Support allocating from a specific software queue Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 14/26] blk-mq: Restore the zoned write order when requeuing Bart Van Assche
2024-11-19 7:52 ` Damien Le Moal
2024-11-19 21:16 ` Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 15/26] blk-zoned: Document the locking order Bart Van Assche
2024-11-19 7:52 ` Damien Le Moal
2024-11-19 0:28 ` [PATCH v16 16/26] blk-zoned: Document locking assumptions Bart Van Assche
2024-11-19 7:53 ` Damien Le Moal
2024-11-19 21:18 ` Bart Van Assche
2024-11-21 3:34 ` Damien Le Moal
2024-11-19 0:28 ` [PATCH v16 17/26] blk-zoned: Uninline functions that are not in the hot path Bart Van Assche
2024-11-19 7:55 ` Damien Le Moal
2024-11-19 21:20 ` Bart Van Assche
2024-11-21 3:36 ` Damien Le Moal
2024-11-19 0:28 ` [PATCH v16 18/26] blk-zoned: Make disk_should_remove_zone_wplug() more robust Bart Van Assche
2024-11-19 7:58 ` Damien Le Moal
2024-11-19 0:28 ` [PATCH v16 19/26] blk-zoned: Add an argument to blk_zone_plug_bio() Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 20/26] blk-zoned: Support pipelining of zoned writes Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 21/26] scsi: core: Retry unaligned " Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 22/26] scsi: sd: Increase retry count for " Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 23/26] scsi: scsi_debug: Add the preserves_write_order module parameter Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 24/26] scsi: scsi_debug: Support injecting unaligned write errors Bart Van Assche
2024-11-19 0:28 ` [PATCH v16 25/26] scsi: scsi_debug: Skip host/bus reset settle delay Bart Van Assche
2024-11-19 0:28 ` Bart Van Assche [this message]
[not found] ` <37f95f44-ab1d-20db-e0c7-94946cb9d4eb@quicinc.com>
2024-11-22 18:20 ` [PATCH v16 26/26] scsi: ufs: Inform the block layer about write ordering Bart Van Assche
2024-11-23 0:34 ` Can Guo
2024-11-19 8:01 ` [PATCH v16 00/26] Improve write performance for zoned UFS devices Damien Le Moal
2024-11-19 19:08 ` Bart Van Assche
2024-11-21 3:20 ` Damien Le Moal
2024-11-21 18:00 ` Bart Van Assche
2024-11-25 3:59 ` Damien Le Moal
2025-01-09 19:02 ` Bart Van Assche
2025-01-10 5:10 ` Damien Le Moal
2024-11-19 12:25 ` Christoph Hellwig
2024-11-19 18:52 ` Bart Van Assche
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=20241119002815.600608-27-bvanassche@acm.org \
--to=bvanassche@acm.org \
--cc=avri.altman@wdc.com \
--cc=axboe@kernel.dk \
--cc=dlemoal@kernel.org \
--cc=hch@lst.de \
--cc=jaegeuk@kernel.org \
--cc=linux-block@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=martin.petersen@oracle.com \
--cc=quic_cang@quicinc.com \
--cc=quic_nguyenb@quicinc.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