[PATCH for-next 00/13] simple 5.13 cleanups

[PATCH for-next 00/13] simple 5.13 cleanups

[Pavel Begunkov]

1-8 are just random stuff
9-13 are leftovers after prep_async patches, further cleaning it up

Pavel Begunkov (13):
  io_uring: introduce a new helper for ctx quiesce
  io_uring: avoid taking ctx refs for task-cancel
  io_uring: reuse io_req_task_queue_fail()
  io_uring: further deduplicate file slot selection
  io_uring: add a helper failing not issued requests
  io_uring: refactor provide/remove buffer locking
  io_uring: don't restirct issue_flags for io_openat
  io_uring: use better types for cflags
  io_uring: refactor out send/recv async setup
  io_uring: untie alloc_async_data and needs_async_data
  io_uring: rethink def->needs_async_data
  io_uring: merge defer_prep() and prep_async()
  io_uring: simplify io_resubmit_prep()

 fs/io_uring.c | 262 +++++++++++++++++++-------------------------------
 1 file changed, 97 insertions(+), 165 deletions(-)

-- 
2.24.0

[PATCH 01/13] io_uring: introduce a new helper for ctx quiesce

[Pavel Begunkov]

Extract a helper io_uring_quiesce_ctx(), which takes care of ctx
quiesce, it's complex enough.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 52 ++++++++++++++++++++++++++++-----------------------
 1 file changed, 29 insertions(+), 23 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 3f764f2f2982..6e781c064e74 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -9683,6 +9683,33 @@ static bool io_register_op_must_quiesce(int op)
 	}
 }
 
+static int io_uring_quiesce_ctx(struct io_ring_ctx *ctx)
+{
+	int ret;
+
+	percpu_ref_kill(&ctx->refs);
+	/*
+	 * Drop uring mutex before waiting for references to exit. If
+	 * another thread is currently inside io_uring_enter() it might
+	 * need to grab the uring_lock to make progress. If we hold it
+	 * here across the drain wait, then we can deadlock. It's safe
+	 * to drop the mutex here, since no new references will come in
+	 * after we've killed the percpu ref.
+	 */
+	mutex_unlock(&ctx->uring_lock);
+	do {
+		ret = wait_for_completion_interruptible(&ctx->ref_comp);
+		if (!ret)
+			break;
+		ret = io_run_task_work_sig();
+	} while (ret >= 0);
+	mutex_lock(&ctx->uring_lock);
+
+	if (ret && io_refs_resurrect(&ctx->refs, &ctx->ref_comp))
+		return ret;
+	return 0;
+}
+
 static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 			       void __user *arg, unsigned nr_args)
 	__releases(ctx->uring_lock)
@@ -9699,29 +9726,8 @@ static int __io_uring_register(struct io_ring_ctx *ctx, unsigned opcode,
 		return -ENXIO;
 
 	if (io_register_op_must_quiesce(opcode)) {
-		percpu_ref_kill(&ctx->refs);
-
-		/*
-		 * Drop uring mutex before waiting for references to exit. If
-		 * another thread is currently inside io_uring_enter() it might
-		 * need to grab the uring_lock to make progress. If we hold it
-		 * here across the drain wait, then we can deadlock. It's safe
-		 * to drop the mutex here, since no new references will come in
-		 * after we've killed the percpu ref.
-		 */
-		mutex_unlock(&ctx->uring_lock);
-		do {
-			ret = wait_for_completion_interruptible(&ctx->ref_comp);
-			if (!ret)
-				break;
-			ret = io_run_task_work_sig();
-			if (ret < 0)
-				break;
-		} while (1);
-
-		mutex_lock(&ctx->uring_lock);
-
-		if (ret && io_refs_resurrect(&ctx->refs, &ctx->ref_comp))
+		ret = io_uring_quiesce_ctx(ctx);
+		if (ret)
 			return ret;
 	}
 
-- 
2.24.0

[PATCH 02/13] io_uring: avoid taking ctx refs for task-cancel

[Pavel Begunkov]

Don't bother to take a ctx->refs for io_req_task_cancel() because it
take uring_lock before putting a request, and the context is promised to
stay alive until unlock happens.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 6e781c064e74..2d66d0afc6c0 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2010,10 +2010,10 @@ static void io_req_task_cancel(struct callback_head *cb)
 	struct io_kiocb *req = container_of(cb, struct io_kiocb, task_work);
 	struct io_ring_ctx *ctx = req->ctx;
 
+	/* ctx is guaranteed to stay alive while we hold uring_lock */
 	mutex_lock(&ctx->uring_lock);
 	__io_req_task_cancel(req, req->result);
 	mutex_unlock(&ctx->uring_lock);
-	percpu_ref_put(&ctx->refs);
 }
 
 static void __io_req_task_submit(struct io_kiocb *req)
@@ -2044,14 +2044,12 @@ static void io_req_task_queue(struct io_kiocb *req)
 	ret = io_req_task_work_add(req);
 	if (unlikely(ret)) {
 		req->result = -ECANCELED;
-		percpu_ref_get(&req->ctx->refs);
 		io_req_task_work_add_fallback(req, io_req_task_cancel);
 	}
 }
 
 static void io_req_task_queue_fail(struct io_kiocb *req, int ret)
 {
-	percpu_ref_get(&req->ctx->refs);
 	req->result = ret;
 	req->task_work.func = io_req_task_cancel;
 
-- 
2.24.0

[PATCH 03/13] io_uring: reuse io_req_task_queue_fail()

[Pavel Begunkov]

Use io_req_task_queue_fail() on the fail path of io_req_task_queue().
It's unlikely to happen, so don't care about additional overhead, but
allows to keep all the req->result invariant in a single function.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 18 +++++++-----------
 1 file changed, 7 insertions(+), 11 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 2d66d0afc6c0..aa2ad863af96 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2036,25 +2036,21 @@ static void io_req_task_submit(struct callback_head *cb)
 	__io_req_task_submit(req);
 }
 
-static void io_req_task_queue(struct io_kiocb *req)
+static void io_req_task_queue_fail(struct io_kiocb *req, int ret)
 {
-	int ret;
+	req->result = ret;
+	req->task_work.func = io_req_task_cancel;
 
-	req->task_work.func = io_req_task_submit;
-	ret = io_req_task_work_add(req);
-	if (unlikely(ret)) {
-		req->result = -ECANCELED;
+	if (unlikely(io_req_task_work_add(req)))
 		io_req_task_work_add_fallback(req, io_req_task_cancel);
-	}
 }
 
-static void io_req_task_queue_fail(struct io_kiocb *req, int ret)
+static void io_req_task_queue(struct io_kiocb *req)
 {
-	req->result = ret;
-	req->task_work.func = io_req_task_cancel;
+	req->task_work.func = io_req_task_submit;
 
 	if (unlikely(io_req_task_work_add(req)))
-		io_req_task_work_add_fallback(req, io_req_task_cancel);
+		io_req_task_queue_fail(req, -ECANCELED);
 }
 
 static inline void io_queue_next(struct io_kiocb *req)
-- 
2.24.0

[PATCH 04/13] io_uring: further deduplicate file slot selection

[Pavel Begunkov]

io_fixed_file_slot() and io_file_from_index() behave pretty similarly,
DRY and call one from another.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 23 ++++++++++-------------
 1 file changed, 10 insertions(+), 13 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index aa2ad863af96..0f85506448ac 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -6107,13 +6107,19 @@ static void io_wq_submit_work(struct io_wq_work *work)
 	}
 }
 
-static inline struct file *io_file_from_index(struct io_ring_ctx *ctx,
-					      int index)
+static inline struct file **io_fixed_file_slot(struct fixed_rsrc_data *file_data,
+					       unsigned i)
 {
 	struct fixed_rsrc_table *table;
 
-	table = &ctx->file_data->table[index >> IORING_FILE_TABLE_SHIFT];
-	return table->files[index & IORING_FILE_TABLE_MASK];
+	table = &file_data->table[i >> IORING_FILE_TABLE_SHIFT];
+	return &table->files[i & IORING_FILE_TABLE_MASK];
+}
+
+static inline struct file *io_file_from_index(struct io_ring_ctx *ctx,
+					      int index)
+{
+	return *io_fixed_file_slot(ctx->file_data, index);
 }
 
 static struct file *io_file_get(struct io_submit_state *state,
@@ -7422,15 +7428,6 @@ static void io_rsrc_put_work(struct work_struct *work)
 	}
 }
 
-static struct file **io_fixed_file_slot(struct fixed_rsrc_data *file_data,
-					unsigned i)
-{
-	struct fixed_rsrc_table *table;
-
-	table = &file_data->table[i >> IORING_FILE_TABLE_SHIFT];
-	return &table->files[i & IORING_FILE_TABLE_MASK];
-}
-
 static void io_rsrc_node_ref_zero(struct percpu_ref *ref)
 {
 	struct fixed_rsrc_ref_node *ref_node;
-- 
2.24.0

[PATCH 05/13] io_uring: add a helper failing not issued requests

[Pavel Begunkov]

Add a simple helper doing CQE posting, marking request for link-failure,
and putting both submission and completion references.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 29 ++++++++++++++---------------
 1 file changed, 14 insertions(+), 15 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 0f85506448ac..692fe7399c94 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1592,8 +1592,8 @@ static void io_cqring_fill_event(struct io_kiocb *req, long res)
 	__io_cqring_fill_event(req, res, 0);
 }
 
-static inline void io_req_complete_post(struct io_kiocb *req, long res,
-					unsigned int cflags)
+static void io_req_complete_post(struct io_kiocb *req, long res,
+				 unsigned int cflags)
 {
 	struct io_ring_ctx *ctx = req->ctx;
 	unsigned long flags;
@@ -1646,6 +1646,13 @@ static inline void io_req_complete(struct io_kiocb *req, long res)
 	__io_req_complete(req, 0, res, 0);
 }
 
+static void io_req_complete_failed(struct io_kiocb *req, long res)
+{
+	req_set_fail_links(req);
+	io_put_req(req);
+	io_req_complete_post(req, res, 0);
+}
+
 static bool io_flush_cached_reqs(struct io_ring_ctx *ctx)
 {
 	struct io_submit_state *state = &ctx->submit_state;
@@ -6258,9 +6265,7 @@ static void __io_queue_sqe(struct io_kiocb *req)
 			io_put_req(req);
 		}
 	} else {
-		req_set_fail_links(req);
-		io_put_req(req);
-		io_req_complete(req, ret);
+		io_req_complete_failed(req, ret);
 	}
 	if (linked_timeout)
 		io_queue_linked_timeout(linked_timeout);
@@ -6274,9 +6279,7 @@ static void io_queue_sqe(struct io_kiocb *req)
 	if (ret) {
 		if (ret != -EIOCBQUEUED) {
 fail_req:
-			req_set_fail_links(req);
-			io_put_req(req);
-			io_req_complete(req, ret);
+			io_req_complete_failed(req, ret);
 		}
 	} else if (req->flags & REQ_F_FORCE_ASYNC) {
 		ret = io_req_defer_prep(req);
@@ -6393,13 +6396,11 @@ static int io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
 	ret = io_init_req(ctx, req, sqe);
 	if (unlikely(ret)) {
 fail_req:
-		io_put_req(req);
-		io_req_complete(req, ret);
+		io_req_complete_failed(req, ret);
 		if (link->head) {
 			/* fail even hard links since we don't submit */
 			link->head->flags |= REQ_F_FAIL_LINK;
-			io_put_req(link->head);
-			io_req_complete(link->head, -ECANCELED);
+			io_req_complete_failed(link->head, -ECANCELED);
 			link->head = NULL;
 		}
 		return ret;
@@ -8591,9 +8592,7 @@ static void io_cancel_defer_files(struct io_ring_ctx *ctx,
 	while (!list_empty(&list)) {
 		de = list_first_entry(&list, struct io_defer_entry, list);
 		list_del_init(&de->list);
-		req_set_fail_links(de->req);
-		io_put_req(de->req);
-		io_req_complete(de->req, -ECANCELED);
+		io_req_complete_failed(de->req, -ECANCELED);
 		kfree(de);
 	}
 }
-- 
2.24.0

[PATCH 06/13] io_uring: refactor provide/remove buffer locking

[Pavel Begunkov]

Always complete request holding the mutex instead of doing that strange
dancing with conditional ordering.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 23 ++++++-----------------
 1 file changed, 6 insertions(+), 17 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 692fe7399c94..39aa7eef39c2 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -3937,14 +3937,9 @@ static int io_remove_buffers(struct io_kiocb *req, unsigned int issue_flags)
 	if (ret < 0)
 		req_set_fail_links(req);
 
-	/* need to hold the lock to complete IOPOLL requests */
-	if (ctx->flags & IORING_SETUP_IOPOLL) {
-		__io_req_complete(req, issue_flags, ret, 0);
-		io_ring_submit_unlock(ctx, !force_nonblock);
-	} else {
-		io_ring_submit_unlock(ctx, !force_nonblock);
-		__io_req_complete(req, issue_flags, ret, 0);
-	}
+	/* complete before unlock, IOPOLL may need the lock */
+	__io_req_complete(req, issue_flags, ret, 0);
+	io_ring_submit_unlock(ctx, !force_nonblock);
 	return 0;
 }
 
@@ -4031,15 +4026,9 @@ static int io_provide_buffers(struct io_kiocb *req, unsigned int issue_flags)
 out:
 	if (ret < 0)
 		req_set_fail_links(req);
-
-	/* need to hold the lock to complete IOPOLL requests */
-	if (ctx->flags & IORING_SETUP_IOPOLL) {
-		__io_req_complete(req, issue_flags, ret, 0);
-		io_ring_submit_unlock(ctx, !force_nonblock);
-	} else {
-		io_ring_submit_unlock(ctx, !force_nonblock);
-		__io_req_complete(req, issue_flags, ret, 0);
-	}
+	/* complete before unlock, IOPOLL may need the lock */
+	__io_req_complete(req, issue_flags, ret, 0);
+	io_ring_submit_unlock(ctx, !force_nonblock);
 	return 0;
 }
 
-- 
2.24.0

[PATCH 07/13] io_uring: don't restirct issue_flags for io_openat

[Pavel Begunkov]

45d189c606292 ("io_uring: replace force_nonblock with flags") did
something strange for io_openat() slicing all issue_flags but
IO_URING_F_NONBLOCK. Not a bug for now, but better to just forward the
flags.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 39aa7eef39c2..d2cbee73fd4a 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -3870,7 +3870,7 @@ static int io_openat2(struct io_kiocb *req, unsigned int issue_flags)
 
 static int io_openat(struct io_kiocb *req, unsigned int issue_flags)
 {
-	return io_openat2(req, issue_flags & IO_URING_F_NONBLOCK);
+	return io_openat2(req, issue_flags);
 }
 
 static int io_remove_buffers_prep(struct io_kiocb *req,
-- 
2.24.0

[PATCH 08/13] io_uring: use better types for cflags

[Pavel Begunkov]

__io_cqring_fill_event() takes cflags as long to squeeze it into u32 in
an CQE, awhile all users pass int or unsigned. Replace it with unsigned
int and store it as u32 in struct io_completion to match CQE.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index d2cbee73fd4a..b4d69f9db5fb 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -643,7 +643,7 @@ struct io_unlink {
 struct io_completion {
 	struct file			*file;
 	struct list_head		list;
-	int				cflags;
+	u32				cflags;
 };
 
 struct io_async_connect {
@@ -1547,7 +1547,8 @@ static void io_cqring_overflow_flush(struct io_ring_ctx *ctx, bool force,
 	}
 }
 
-static void __io_cqring_fill_event(struct io_kiocb *req, long res, long cflags)
+static void __io_cqring_fill_event(struct io_kiocb *req, long res,
+				   unsigned int cflags)
 {
 	struct io_ring_ctx *ctx = req->ctx;
 	struct io_uring_cqe *cqe;
-- 
2.24.0

[PATCH 09/13] io_uring: refactor out send/recv async setup

[Pavel Begunkov]

IORING_OP_[SEND,RECV] don't need async setup neither will get into
io_req_prep_async(). Remove them from io_req_prep_async() and remove
needs_async_data checks from the related setup functions.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index b4d69f9db5fb..b3585226bfb5 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -4325,8 +4325,6 @@ static int io_sendmsg_prep_async(struct io_kiocb *req)
 {
 	int ret;
 
-	if (!io_op_defs[req->opcode].needs_async_data)
-		return 0;
 	ret = io_sendmsg_copy_hdr(req, req->async_data);
 	if (!ret)
 		req->flags |= REQ_F_NEED_CLEANUP;
@@ -4546,8 +4544,6 @@ static int io_recvmsg_prep_async(struct io_kiocb *req)
 {
 	int ret;
 
-	if (!io_op_defs[req->opcode].needs_async_data)
-		return 0;
 	ret = io_recvmsg_copy_hdr(req, req->async_data);
 	if (!ret)
 		req->flags |= REQ_F_NEED_CLEANUP;
@@ -5802,10 +5798,8 @@ static int io_req_prep_async(struct io_kiocb *req)
 	case IORING_OP_WRITE:
 		return io_rw_prep_async(req, WRITE);
 	case IORING_OP_SENDMSG:
-	case IORING_OP_SEND:
 		return io_sendmsg_prep_async(req);
 	case IORING_OP_RECVMSG:
-	case IORING_OP_RECV:
 		return io_recvmsg_prep_async(req);
 	case IORING_OP_CONNECT:
 		return io_connect_prep_async(req);
-- 
2.24.0

[PATCH 10/13] io_uring: untie alloc_async_data and needs_async_data

[Pavel Begunkov]

All opcode handlers pretty well know whether they need async data or
not, and can skip testing for needs_async_data. The exception is rw
the generic path, but those test the flag by hand anyway. So, check the
flag and make io_alloc_async_data() allocating unconditionally.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 14 +++-----------
 1 file changed, 3 insertions(+), 11 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index b3585226bfb5..8f25371ae904 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -3105,21 +3105,13 @@ static void io_req_map_rw(struct io_kiocb *req, const struct iovec *iovec,
 	}
 }
 
-static inline int __io_alloc_async_data(struct io_kiocb *req)
+static inline int io_alloc_async_data(struct io_kiocb *req)
 {
 	WARN_ON_ONCE(!io_op_defs[req->opcode].async_size);
 	req->async_data = kmalloc(io_op_defs[req->opcode].async_size, GFP_KERNEL);
 	return req->async_data == NULL;
 }
 
-static int io_alloc_async_data(struct io_kiocb *req)
-{
-	if (!io_op_defs[req->opcode].needs_async_data)
-		return 0;
-
-	return  __io_alloc_async_data(req);
-}
-
 static int io_setup_async_rw(struct io_kiocb *req, const struct iovec *iovec,
 			     const struct iovec *fast_iov,
 			     struct iov_iter *iter, bool force)
@@ -3127,7 +3119,7 @@ static int io_setup_async_rw(struct io_kiocb *req, const struct iovec *iovec,
 	if (!force && !io_op_defs[req->opcode].needs_async_data)
 		return 0;
 	if (!req->async_data) {
-		if (__io_alloc_async_data(req)) {
+		if (io_alloc_async_data(req)) {
 			kfree(iovec);
 			return -ENOMEM;
 		}
@@ -5814,7 +5806,7 @@ static int io_req_defer_prep(struct io_kiocb *req)
 	/* some opcodes init it during the inital prep */
 	if (req->async_data)
 		return 0;
-	if (__io_alloc_async_data(req))
+	if (io_alloc_async_data(req))
 		return -EAGAIN;
 	return io_req_prep_async(req);
 }
-- 
2.24.0

[PATCH 11/13] io_uring: rethink def->needs_async_data

[Pavel Begunkov]

needs_async_data controls allocation of async_data, and used in two
cases. 1) when async setup requires it (by io_req_prep_async() or
handler themselves), and 2) when op always needs additional space to
operate, like timeouts do.

Opcode preps already don't bother about the second case and do
allocation unconditionally, restrict needs_async_data to the first case
only and rename it into needs_async_setup.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 29 +++++++++++------------------
 1 file changed, 11 insertions(+), 18 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 8f25371ae904..5647bc73969b 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -833,8 +833,8 @@ struct io_op_def {
 	unsigned		pollout : 1;
 	/* op supports buffer selection */
 	unsigned		buffer_select : 1;
-	/* must always have async data allocated */
-	unsigned		needs_async_data : 1;
+	/* do prep async if is going to be punted */
+	unsigned		needs_async_setup : 1;
 	/* should block plug */
 	unsigned		plug : 1;
 	/* size of async data needed, if any */
@@ -848,7 +848,7 @@ static const struct io_op_def io_op_defs[] = {
 		.unbound_nonreg_file	= 1,
 		.pollin			= 1,
 		.buffer_select		= 1,
-		.needs_async_data	= 1,
+		.needs_async_setup	= 1,
 		.plug			= 1,
 		.async_size		= sizeof(struct io_async_rw),
 	},
@@ -857,7 +857,7 @@ static const struct io_op_def io_op_defs[] = {
 		.hash_reg_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.pollout		= 1,
-		.needs_async_data	= 1,
+		.needs_async_setup	= 1,
 		.plug			= 1,
 		.async_size		= sizeof(struct io_async_rw),
 	},
@@ -891,7 +891,7 @@ static const struct io_op_def io_op_defs[] = {
 		.needs_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.pollout		= 1,
-		.needs_async_data	= 1,
+		.needs_async_setup	= 1,
 		.async_size		= sizeof(struct io_async_msghdr),
 	},
 	[IORING_OP_RECVMSG] = {
@@ -899,11 +899,10 @@ static const struct io_op_def io_op_defs[] = {
 		.unbound_nonreg_file	= 1,
 		.pollin			= 1,
 		.buffer_select		= 1,
-		.needs_async_data	= 1,
+		.needs_async_setup	= 1,
 		.async_size		= sizeof(struct io_async_msghdr),
 	},
 	[IORING_OP_TIMEOUT] = {
-		.needs_async_data	= 1,
 		.async_size		= sizeof(struct io_timeout_data),
 	},
 	[IORING_OP_TIMEOUT_REMOVE] = {
@@ -916,14 +915,13 @@ static const struct io_op_def io_op_defs[] = {
 	},
 	[IORING_OP_ASYNC_CANCEL] = {},
 	[IORING_OP_LINK_TIMEOUT] = {
-		.needs_async_data	= 1,
 		.async_size		= sizeof(struct io_timeout_data),
 	},
 	[IORING_OP_CONNECT] = {
 		.needs_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.pollout		= 1,
-		.needs_async_data	= 1,
+		.needs_async_setup	= 1,
 		.async_size		= sizeof(struct io_async_connect),
 	},
 	[IORING_OP_FALLOCATE] = {
@@ -3116,7 +3114,7 @@ static int io_setup_async_rw(struct io_kiocb *req, const struct iovec *iovec,
 			     const struct iovec *fast_iov,
 			     struct iov_iter *iter, bool force)
 {
-	if (!force && !io_op_defs[req->opcode].needs_async_data)
+	if (!force && !io_op_defs[req->opcode].needs_async_setup)
 		return 0;
 	if (!req->async_data) {
 		if (io_alloc_async_data(req)) {
@@ -5782,12 +5780,8 @@ static int io_req_prep_async(struct io_kiocb *req)
 {
 	switch (req->opcode) {
 	case IORING_OP_READV:
-	case IORING_OP_READ_FIXED:
-	case IORING_OP_READ:
 		return io_rw_prep_async(req, READ);
 	case IORING_OP_WRITEV:
-	case IORING_OP_WRITE_FIXED:
-	case IORING_OP_WRITE:
 		return io_rw_prep_async(req, WRITE);
 	case IORING_OP_SENDMSG:
 		return io_sendmsg_prep_async(req);
@@ -5801,11 +5795,10 @@ static int io_req_prep_async(struct io_kiocb *req)
 
 static int io_req_defer_prep(struct io_kiocb *req)
 {
-	if (!io_op_defs[req->opcode].needs_async_data)
-		return 0;
-	/* some opcodes init it during the inital prep */
-	if (req->async_data)
+	if (!io_op_defs[req->opcode].needs_async_setup)
 		return 0;
+	if (WARN_ON_ONCE(req->async_data))
+		return -EFAULT;
 	if (io_alloc_async_data(req))
 		return -EAGAIN;
 	return io_req_prep_async(req);
-- 
2.24.0

[PATCH 12/13] io_uring: merge defer_prep() and prep_async()

[Pavel Begunkov]

Merge two function and do renaming in favour of the second one, it
relays the meaning better.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 28 +++++++++++++---------------
 1 file changed, 13 insertions(+), 15 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 5647bc73969b..34cd1b9545c1 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -5778,6 +5778,13 @@ static int io_req_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 
 static int io_req_prep_async(struct io_kiocb *req)
 {
+	if (!io_op_defs[req->opcode].needs_async_setup)
+		return 0;
+	if (WARN_ON_ONCE(req->async_data))
+		return -EFAULT;
+	if (io_alloc_async_data(req))
+		return -EAGAIN;
+
 	switch (req->opcode) {
 	case IORING_OP_READV:
 		return io_rw_prep_async(req, READ);
@@ -5790,18 +5797,9 @@ static int io_req_prep_async(struct io_kiocb *req)
 	case IORING_OP_CONNECT:
 		return io_connect_prep_async(req);
 	}
-	return 0;
-}
-
-static int io_req_defer_prep(struct io_kiocb *req)
-{
-	if (!io_op_defs[req->opcode].needs_async_setup)
-		return 0;
-	if (WARN_ON_ONCE(req->async_data))
-		return -EFAULT;
-	if (io_alloc_async_data(req))
-		return -EAGAIN;
-	return io_req_prep_async(req);
+	printk_once(KERN_WARNING "io_uring: prep_async() bad opcode %d\n",
+		    req->opcode);
+	return -EFAULT;
 }
 
 static u32 io_get_sequence(struct io_kiocb *req)
@@ -5834,7 +5832,7 @@ static int io_req_defer(struct io_kiocb *req)
 	if (!req_need_defer(req, seq) && list_empty_careful(&ctx->defer_list))
 		return 0;
 
-	ret = io_req_defer_prep(req);
+	ret = io_req_prep_async(req);
 	if (ret)
 		return ret;
 	io_prep_async_link(req);
@@ -6251,7 +6249,7 @@ static void io_queue_sqe(struct io_kiocb *req)
 			io_req_complete_failed(req, ret);
 		}
 	} else if (req->flags & REQ_F_FORCE_ASYNC) {
-		ret = io_req_defer_prep(req);
+		ret = io_req_prep_async(req);
 		if (unlikely(ret))
 			goto fail_req;
 		io_queue_async_work(req);
@@ -6403,7 +6401,7 @@ static int io_submit_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req,
 			head->flags |= REQ_F_IO_DRAIN;
 			ctx->drain_next = 1;
 		}
-		ret = io_req_defer_prep(req);
+		ret = io_req_prep_async(req);
 		if (unlikely(ret))
 			goto fail_req;
 		trace_io_uring_link(ctx, req, head);
-- 
2.24.0

[PATCH 13/13] io_uring: simplify io_resubmit_prep()

[Pavel Begunkov]

If not for async_data NULL check, io_resubmit_prep() is already an rw
specific version of io_req_prep_async(), but slower because 1) it always
goes through io_import_iovec() even if following io_setup_async_rw() the
result 2) instead of initialising iovec/iter in-place it does it
on-stack and then copies with io_setup_async_rw().

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
---
 fs/io_uring.c | 37 +++----------------------------------
 1 file changed, 3 insertions(+), 34 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 34cd1b9545c1..97002de5e0d2 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1014,14 +1014,10 @@ static struct file *io_file_get(struct io_submit_state *state,
 static void __io_queue_sqe(struct io_kiocb *req);
 static void io_rsrc_put_work(struct work_struct *work);
 
-static int io_import_iovec(int rw, struct io_kiocb *req, struct iovec **iovec,
-			   struct iov_iter *iter, bool needs_lock);
-static int io_setup_async_rw(struct io_kiocb *req, const struct iovec *iovec,
-			     const struct iovec *fast_iov,
-			     struct iov_iter *iter, bool force);
 static void io_req_task_queue(struct io_kiocb *req);
 static void io_submit_flush_completions(struct io_comp_state *cs,
 					struct io_ring_ctx *ctx);
+static int io_req_prep_async(struct io_kiocb *req);
 
 static struct kmem_cache *req_cachep;
 
@@ -2454,35 +2450,8 @@ static void kiocb_end_write(struct io_kiocb *req)
 #ifdef CONFIG_BLOCK
 static bool io_resubmit_prep(struct io_kiocb *req)
 {
-	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
-	int rw, ret;
-	struct iov_iter iter;
-
-	/* already prepared */
-	if (req->async_data)
-		return true;
-
-	switch (req->opcode) {
-	case IORING_OP_READV:
-	case IORING_OP_READ_FIXED:
-	case IORING_OP_READ:
-		rw = READ;
-		break;
-	case IORING_OP_WRITEV:
-	case IORING_OP_WRITE_FIXED:
-	case IORING_OP_WRITE:
-		rw = WRITE;
-		break;
-	default:
-		printk_once(KERN_WARNING "io_uring: bad opcode in resubmit %d\n",
-				req->opcode);
-		return false;
-	}
-
-	ret = io_import_iovec(rw, req, &iovec, &iter, false);
-	if (ret < 0)
-		return false;
-	return !io_setup_async_rw(req, iovec, inline_vecs, &iter, false);
+	/* either already prepared or successfully done */
+	return req->async_data || !io_req_prep_async(req);
 }
 #endif
 
-- 
2.24.0