[PATCH V3 0/2] io_uring: cancelable uring_cmd

[PATCH V3 0/2] io_uring: cancelable uring_cmd

[Ming Lei]

Hello,

Patch 1 retains top 8bits of  uring_cmd flags for kernel internal use.

Patch 2 implements cancelable uring_cmd.

git tree(with ublk change)

	https://github.com/ming1/linux/commits/uring_exit_and_ublk

V3:
	- code style change as suggested by Jens
	- add patch 1

V2:
	- use ->uring_cmd() with IO_URING_F_CANCEL for canceling command

Ming Lei (2):
  io_uring: retain top 8bits of uring_cmd flags for kernel internal use
  io_uring: cancelable uring_cmd

 include/linux/io_uring.h       | 19 +++++++++++++
 include/linux/io_uring_types.h |  6 ++++
 include/uapi/linux/io_uring.h  |  5 ++--
 io_uring/io_uring.c            | 34 +++++++++++++++++++++++
 io_uring/uring_cmd.c           | 51 +++++++++++++++++++++++++++++++++-
 5 files changed, 111 insertions(+), 4 deletions(-)

-- 
2.41.0

[PATCH V3 1/2] io_uring: retain top 8bits of uring_cmd flags for kernel internal use

[Ming Lei]

Retain top 8bits of uring_cmd flags for kernel internal use, so that we
can move IORING_URING_CMD_POLLED out of uapi header.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 include/linux/io_uring.h      | 3 +++
 include/uapi/linux/io_uring.h | 5 ++---
 io_uring/uring_cmd.c          | 2 +-
 3 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/include/linux/io_uring.h b/include/linux/io_uring.h
index 106cdc55ff3b..ae08d6f66e62 100644
--- a/include/linux/io_uring.h
+++ b/include/linux/io_uring.h
@@ -22,6 +22,9 @@ enum io_uring_cmd_flags {
 	IO_URING_F_IOPOLL		= (1 << 10),
 };
 
+/* only top 8 bits of sqe->uring_cmd_flags for kernel internal use */
+#define IORING_URING_CMD_POLLED		(1U << 31)
+
 struct io_uring_cmd {
 	struct file	*file;
 	const struct io_uring_sqe *sqe;
diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 8e61f8b7c2ce..4df2f11f264f 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -246,13 +246,12 @@ enum io_uring_op {
 };
 
 /*
- * sqe->uring_cmd_flags
+ * sqe->uring_cmd_flags		top 8bits aren't available for userspace
  * IORING_URING_CMD_FIXED	use registered buffer; pass this flag
  *				along with setting sqe->buf_index.
- * IORING_URING_CMD_POLLED	driver use only
  */
 #define IORING_URING_CMD_FIXED	(1U << 0)
-#define IORING_URING_CMD_POLLED	(1U << 31)
+#define IORING_URING_CMD_MASK	0x00ffffffU
 
 
 /*
diff --git a/io_uring/uring_cmd.c b/io_uring/uring_cmd.c
index 537795fddc87..52a455b67163 100644
--- a/io_uring/uring_cmd.c
+++ b/io_uring/uring_cmd.c
@@ -90,7 +90,7 @@ int io_uring_cmd_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	if (sqe->__pad1)
 		return -EINVAL;
 
-	ioucmd->flags = READ_ONCE(sqe->uring_cmd_flags);
+	ioucmd->flags = READ_ONCE(sqe->uring_cmd_flags) & IORING_URING_CMD_MASK;
 	if (ioucmd->flags & ~IORING_URING_CMD_FIXED)
 		return -EINVAL;
 
-- 
2.41.0

[PATCH V3 2/2] io_uring: cancelable uring_cmd

[Ming Lei]

uring_cmd may never complete, such as ublk, in which uring cmd isn't
completed until one new block request is coming from ublk block device.

Add cancelable uring_cmd to provide mechanism to driver for cancelling
pending commands in its own way.

Add API of io_uring_cmd_mark_cancelable() for driver to mark one command as
cancelable, then io_uring will cancel this command in
io_uring_cancel_generic(). ->uring_cmd() callback is reused for canceling
command in driver's way, then driver gets notified with the cancelling
from io_uring.

Add API of io_uring_cmd_get_task() to help driver cancel handler
deal with the canceling.

Cc: Gabriel Krisman Bertazi <krisman@suse.de>
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 include/linux/io_uring.h       | 16 +++++++++++
 include/linux/io_uring_types.h |  6 +++++
 io_uring/io_uring.c            | 34 +++++++++++++++++++++++
 io_uring/uring_cmd.c           | 49 ++++++++++++++++++++++++++++++++++
 4 files changed, 105 insertions(+)

diff --git a/include/linux/io_uring.h b/include/linux/io_uring.h
index ae08d6f66e62..a0307289bdc7 100644
--- a/include/linux/io_uring.h
+++ b/include/linux/io_uring.h
@@ -20,9 +20,13 @@ enum io_uring_cmd_flags {
 	IO_URING_F_SQE128		= (1 << 8),
 	IO_URING_F_CQE32		= (1 << 9),
 	IO_URING_F_IOPOLL		= (1 << 10),
+
+	/* set when uring wants to cancel one issued command */
+	IO_URING_F_CANCEL		= (1 << 11),
 };
 
 /* only top 8 bits of sqe->uring_cmd_flags for kernel internal use */
+#define IORING_URING_CMD_CANCELABLE	(1U << 30)
 #define IORING_URING_CMD_POLLED		(1U << 31)
 
 struct io_uring_cmd {
@@ -85,6 +89,9 @@ static inline void io_uring_free(struct task_struct *tsk)
 		__io_uring_free(tsk);
 }
 int io_uring_cmd_sock(struct io_uring_cmd *cmd, unsigned int issue_flags);
+int io_uring_cmd_mark_cancelable(struct io_uring_cmd *cmd,
+		unsigned int issue_flags);
+struct task_struct *io_uring_cmd_get_task(struct io_uring_cmd *cmd);
 #else
 static inline int io_uring_cmd_import_fixed(u64 ubuf, unsigned long len, int rw,
 			      struct iov_iter *iter, void *ioucmd)
@@ -125,6 +132,15 @@ static inline int io_uring_cmd_sock(struct io_uring_cmd *cmd,
 {
 	return -EOPNOTSUPP;
 }
+static inline int io_uring_cmd_mark_cancelable(struct io_uring_cmd *cmd,
+		unsigned int issue_flags)
+{
+	return -EOPNOTSUPP;
+}
+static inline struct task_struct *io_uring_cmd_get_task(struct io_uring_cmd *cmd)
+{
+	return NULL;
+}
 #endif
 
 #endif
diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 13d19b9be9f4..1571db76bec1 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -265,6 +265,12 @@ struct io_ring_ctx {
 		 */
 		struct io_wq_work_list	iopoll_list;
 		bool			poll_multi_queue;
+
+		/*
+		 * Any cancelable uring_cmd is added to this list in
+		 * ->uring_cmd() by io_uring_cmd_insert_cancelable()
+		 */
+		struct hlist_head	cancelable_uring_cmd;
 	} ____cacheline_aligned_in_smp;
 
 	struct {
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 783ed0fff71b..1e3de74c2ba3 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -3256,6 +3256,39 @@ static __cold bool io_uring_try_cancel_iowq(struct io_ring_ctx *ctx)
 	return ret;
 }
 
+static bool io_uring_try_cancel_uring_cmd(struct io_ring_ctx *ctx,
+		struct task_struct *task, bool cancel_all)
+	__acquires(ctx->uring_lock)
+{
+	struct hlist_node *tmp;
+	struct io_kiocb *req;
+	bool ret = false;
+
+	hlist_for_each_entry_safe(req, tmp, &ctx->cancelable_uring_cmd,
+			hash_node) {
+		struct io_uring_cmd *cmd = io_kiocb_to_cmd(req,
+				struct io_uring_cmd);
+		struct file *file = req->file;
+
+		if (WARN_ON_ONCE(!file->f_op->uring_cmd))
+			continue;
+
+		if (!cancel_all && req->task != task)
+			continue;
+
+		if (cmd->flags & IORING_URING_CMD_CANCELABLE) {
+			/* ->sqe isn't available if no async data */
+			if (!req_has_async_data(req))
+				cmd->sqe = NULL;
+			file->f_op->uring_cmd(cmd, IO_URING_F_CANCEL);
+			ret = true;
+		}
+	}
+	io_submit_flush_completions(ctx);
+
+	return ret;
+}
+
 static __cold bool io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
 						struct task_struct *task,
 						bool cancel_all)
@@ -3303,6 +3336,7 @@ static __cold bool io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
 	ret |= io_cancel_defer_files(ctx, task, cancel_all);
 	mutex_lock(&ctx->uring_lock);
 	ret |= io_poll_remove_all(ctx, task, cancel_all);
+	ret |= io_uring_try_cancel_uring_cmd(ctx, task, cancel_all);
 	mutex_unlock(&ctx->uring_lock);
 	ret |= io_kill_timeouts(ctx, task, cancel_all);
 	if (task)
diff --git a/io_uring/uring_cmd.c b/io_uring/uring_cmd.c
index 52a455b67163..5832a316359c 100644
--- a/io_uring/uring_cmd.c
+++ b/io_uring/uring_cmd.c
@@ -13,6 +13,53 @@
 #include "rsrc.h"
 #include "uring_cmd.h"
 
+static void io_uring_cmd_del_cancelable(struct io_uring_cmd *cmd,
+		unsigned int issue_flags)
+{
+	struct io_kiocb *req = cmd_to_io_kiocb(cmd);
+	struct io_ring_ctx *ctx = req->ctx;
+
+	if (!(cmd->flags & IORING_URING_CMD_CANCELABLE))
+		return;
+
+	cmd->flags &= ~IORING_URING_CMD_CANCELABLE;
+	io_ring_submit_lock(ctx, issue_flags);
+	hlist_del(&req->hash_node);
+	io_ring_submit_unlock(ctx, issue_flags);
+}
+
+/*
+ * Mark this command as concelable, then io_uring_try_cancel_uring_cmd()
+ * will try to cancel this issued command by sending ->uring_cmd() with
+ * issue_flags of IO_URING_F_CANCEL.
+ *
+ * The command is guaranteed to not be done when calling ->uring_cmd()
+ * with IO_URING_F_CANCEL, but it is driver's responsibility to deal
+ * with race between io_uring canceling and normal completion.
+ */
+int io_uring_cmd_mark_cancelable(struct io_uring_cmd *cmd,
+		unsigned int issue_flags)
+{
+	struct io_kiocb *req = cmd_to_io_kiocb(cmd);
+	struct io_ring_ctx *ctx = req->ctx;
+
+	if (!(cmd->flags & IORING_URING_CMD_CANCELABLE)) {
+		cmd->flags |= IORING_URING_CMD_CANCELABLE;
+		io_ring_submit_lock(ctx, issue_flags);
+		hlist_add_head(&req->hash_node, &ctx->cancelable_uring_cmd);
+		io_ring_submit_unlock(ctx, issue_flags);
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(io_uring_cmd_mark_cancelable);
+
+struct task_struct *io_uring_cmd_get_task(struct io_uring_cmd *cmd)
+{
+	return cmd_to_io_kiocb(cmd)->task;
+}
+EXPORT_SYMBOL_GPL(io_uring_cmd_get_task);
+
 static void io_uring_cmd_work(struct io_kiocb *req, struct io_tw_state *ts)
 {
 	struct io_uring_cmd *ioucmd = io_kiocb_to_cmd(req, struct io_uring_cmd);
@@ -56,6 +103,8 @@ void io_uring_cmd_done(struct io_uring_cmd *ioucmd, ssize_t ret, ssize_t res2,
 {
 	struct io_kiocb *req = cmd_to_io_kiocb(ioucmd);
 
+	io_uring_cmd_del_cancelable(ioucmd, issue_flags);
+
 	if (ret < 0)
 		req_set_fail(req);
 
-- 
2.41.0

Re: [PATCH V3 2/2] io_uring: cancelable uring_cmd

[Jens Axboe]

On 9/22/23 10:09 AM, Ming Lei wrote:
> diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
> index 783ed0fff71b..1e3de74c2ba3 100644
> --- a/io_uring/io_uring.c
> +++ b/io_uring/io_uring.c
> @@ -3256,6 +3256,39 @@ static __cold bool io_uring_try_cancel_iowq(struct io_ring_ctx *ctx)
>  	return ret;
>  }
>  
> +static bool io_uring_try_cancel_uring_cmd(struct io_ring_ctx *ctx,
> +		struct task_struct *task, bool cancel_all)
> +	__acquires(ctx->uring_lock)
> +{

Minor nit - I don't think the static checker will be happy with this, as
we're now called with it held already. Might not be a bad idea to add a:

lockdep_assert_held(&ctx->uring_lock);

at the start of the body, both for safety but also to document this
requirement.

-- 
Jens Axboe

Re: [PATCH V3 1/2] io_uring: retain top 8bits of uring_cmd flags for kernel internal use

[Jens Axboe]

On 9/22/23 10:09 AM, Ming Lei wrote:
> diff --git a/io_uring/uring_cmd.c b/io_uring/uring_cmd.c
> index 537795fddc87..52a455b67163 100644
> --- a/io_uring/uring_cmd.c
> +++ b/io_uring/uring_cmd.c
> @@ -90,7 +90,7 @@ int io_uring_cmd_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>  	if (sqe->__pad1)
>  		return -EINVAL;
>  
> -	ioucmd->flags = READ_ONCE(sqe->uring_cmd_flags);
> +	ioucmd->flags = READ_ONCE(sqe->uring_cmd_flags) & IORING_URING_CMD_MASK;
>  	if (ioucmd->flags & ~IORING_URING_CMD_FIXED)
>  		return -EINVAL;
>  

Do we want to mask it here? If any of the upper bits are set at prep
time, that should be an EINVAL condition.

-- 
Jens Axboe