public inbox for linux-block@vger.kernel.org
 help / color / mirror / Atom feed
From: Ming Lei <ming.lei@redhat.com>
To: Jens Axboe <axboe@kernel.dk>, linux-block@vger.kernel.org
Cc: Uday Shankar <ushankar@purestorage.com>,
	Caleb Sander Mateos <csander@purestorage.com>,
	Ming Lei <ming.lei@redhat.com>
Subject: [PATCH 2/2] ublk: run auto buf unregisgering in same io_ring_ctx with register
Date: Wed, 21 May 2025 10:55:00 +0800	[thread overview]
Message-ID: <20250521025502.71041-3-ming.lei@redhat.com> (raw)
In-Reply-To: <20250521025502.71041-1-ming.lei@redhat.com>

UBLK_F_AUTO_BUF_REG requires that the buffer registered automatically
is unregistered in same `io_ring_ctx`, so check it explicitly.

Meantime return the failure code if io_buffer_unregister_bvec() fails,
then ublk server can handle the failure in consistent way.

Also force to clear UBLK_IO_FLAG_AUTO_BUF_REG in ublk_io_release()
because ublk_io_release() may be triggered not from handling
UBLK_IO_COMMIT_AND_FETCH_REQ, and from releasing the `io_ring_ctx`
for registering the buffer.

Fixes: 99c1e4eb6a3f ("ublk: register buffer to local io_uring with provided buf index via UBLK_F_AUTO_BUF_REG")
Reported-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 drivers/block/ublk_drv.c      | 35 +++++++++++++++++++++++++++++++----
 include/uapi/linux/ublk_cmd.h |  3 ++-
 2 files changed, 33 insertions(+), 5 deletions(-)

diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index fcf568b89370..2af6422d6a89 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -84,6 +84,7 @@ struct ublk_rq_data {
 
 	/* for auto-unregister buffer in case of UBLK_F_AUTO_BUF_REG */
 	u16 buf_index;
+	unsigned long buf_ctx_id;
 };
 
 struct ublk_uring_cmd_pdu {
@@ -1192,6 +1193,11 @@ static void ublk_auto_buf_reg_fallback(struct request *req, struct ublk_io *io)
 	refcount_set(&data->ref, 1);
 }
 
+static unsigned long ublk_uring_cmd_ctx_id(struct io_uring_cmd *cmd)
+{
+	return (unsigned long)(cmd_to_io_kiocb(cmd)->ctx);
+}
+
 static bool ublk_auto_buf_reg(struct request *req, struct ublk_io *io,
 			      unsigned int issue_flags)
 {
@@ -1211,6 +1217,8 @@ static bool ublk_auto_buf_reg(struct request *req, struct ublk_io *io,
 	}
 	/* one extra reference is dropped by ublk_io_release */
 	refcount_set(&data->ref, 2);
+
+	data->buf_ctx_id = ublk_uring_cmd_ctx_id(io->cmd);
 	/* store buffer index in request payload */
 	data->buf_index = pdu->buf.index;
 	io->flags |= UBLK_IO_FLAG_AUTO_BUF_REG;
@@ -1994,6 +2002,21 @@ static void ublk_io_release(void *priv)
 {
 	struct request *rq = priv;
 	struct ublk_queue *ubq = rq->mq_hctx->driver_data;
+	struct ublk_io *io = &ubq->ios[rq->tag];
+
+	/*
+	 * In case of UBLK_F_AUTO_BUF_REG, the `io_uring_ctx` for registering
+	 * this buffer may be released, so we reach here not from handling
+	 * `UBLK_IO_COMMIT_AND_FETCH_REQ`.
+	 *
+	 * Force to clear UBLK_IO_FLAG_AUTO_BUF_REG, so that ublk server
+	 * still may complete this IO request by issuing uring_cmd from
+	 * another `io_uring_ctx` in case that the `io_ring_ctx` for
+	 * registering the buffer is gone
+	 */
+	if (ublk_support_auto_buf_reg(ubq) &&
+			(io->flags & UBLK_IO_FLAG_AUTO_BUF_REG))
+		io->flags &= ~UBLK_IO_FLAG_AUTO_BUF_REG;
 
 	ublk_put_req_ref(ubq, rq);
 }
@@ -2109,14 +2132,18 @@ static int ublk_commit_and_fetch(const struct ublk_queue *ubq,
 	}
 
 	if (ublk_support_auto_buf_reg(ubq)) {
+		struct ublk_rq_data *data = blk_mq_rq_to_pdu(req);
 		int ret;
 
 		if (io->flags & UBLK_IO_FLAG_AUTO_BUF_REG) {
-			struct ublk_rq_data *data = blk_mq_rq_to_pdu(req);
 
-			WARN_ON_ONCE(io_buffer_unregister_bvec(cmd,
-						data->buf_index,
-						issue_flags));
+			if (data->buf_ctx_id != ublk_uring_cmd_ctx_id(cmd))
+				return -EBADF;
+
+			ret = io_buffer_unregister_bvec(cmd, data->buf_index,
+							issue_flags);
+			if (ret)
+				return ret;
 			io->flags &= ~UBLK_IO_FLAG_AUTO_BUF_REG;
 		}
 
diff --git a/include/uapi/linux/ublk_cmd.h b/include/uapi/linux/ublk_cmd.h
index c4b9942697fc..3db604a3045e 100644
--- a/include/uapi/linux/ublk_cmd.h
+++ b/include/uapi/linux/ublk_cmd.h
@@ -226,7 +226,8 @@
  *
  * For using this feature:
  *
- * - ublk server has to create sparse buffer table
+ * - ublk server has to create sparse buffer table on the same `io_ring_ctx`
+ *   for issuing `UBLK_IO_FETCH_REQ` and `UBLK_IO_COMMIT_AND_FETCH_REQ`
  *
  * - ublk server passes auto buf register data via uring_cmd's sqe->addr,
  *   `struct ublk_auto_buf_reg` is populated from sqe->addr, please see
-- 
2.47.0


  parent reply	other threads:[~2025-05-21  2:55 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-21  2:54 [PATCH 0/2] ublk: two fixes on UBLK_F_AUTO_BUF_REG Ming Lei
2025-05-21  2:54 ` [PATCH 1/2] ublk: handle ublk_set_auto_buf_reg() failure correctly in ublk_fetch() Ming Lei
2025-05-21 14:53   ` Caleb Sander Mateos
2025-05-21  2:55 ` Ming Lei [this message]
2025-05-21 15:58   ` [PATCH 2/2] ublk: run auto buf unregisgering in same io_ring_ctx with register Caleb Sander Mateos
2025-05-22  0:42     ` Ming Lei
2025-05-22  1:05       ` Caleb Sander Mateos
2025-05-22  2:37         ` Ming Lei
2025-05-21 19:19   ` Caleb Sander Mateos
2025-05-21 13:11 ` [PATCH 0/2] ublk: two fixes on UBLK_F_AUTO_BUF_REG Jens Axboe
2025-05-21 20:31   ` Jens Axboe

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=20250521025502.71041-3-ming.lei@redhat.com \
    --to=ming.lei@redhat.com \
    --cc=axboe@kernel.dk \
    --cc=csander@purestorage.com \
    --cc=linux-block@vger.kernel.org \
    --cc=ushankar@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