Merge tag 'for-6.3/io_uring-2023-02-16' of git://git.kernel.dk/linux

Pull io_uring updates from Jens Axboe:

 - Cleanup series making the async prep and handling of
   REQ_F_FORCE_ASYNC easier to follow and verify (Dylan)

 - Enable specifying specific flags for OP_MSG_RING (Breno)

 - Enable use of KASAN with the internal request cache (Breno)

 - Split the opcode definition structs into a hot and cold part (Breno)

 - OP_MSG_RING fixes (Pavel, me)

 - Fix an issue with IOPOLL cancelation and PREEMPT_NONE (me)

 - Handle TIF_NOTIFY_RESUME for the io-wq threads that never return to
   userspace (me)

 - Add support for using io_uring_register() with a registered ring fd
   (Josh)

 - Improve handling of poll on the ring fd (Pavel)

 - Series improving the task_work handling (Pavel)

 - Misc cleanups, fixes, improvements (Dmitrii, Quanfa, Richard, Pavel,
   me)

* tag 'for-6.3/io_uring-2023-02-16' of git://git.kernel.dk/linux: (51 commits)
  io_uring: Support calling io_uring_register with a registered ring fd
  io_uring,audit: don't log IORING_OP_MADVISE
  io_uring: mark task TASK_RUNNING before handling resume/task work
  io_uring: always go async for unsupported open flags
  io_uring: always go async for unsupported fadvise flags
  io_uring: for requests that require async, force it
  io_uring: if a linked request has REQ_F_FORCE_ASYNC then run it async
  io_uring: add reschedule point to handle_tw_list()
  io_uring: add a conditional reschedule to the IOPOLL cancelation loop
  io_uring: return normal tw run linking optimisation
  io_uring: refactor tctx_task_work
  io_uring: refactor io_put_task helpers
  io_uring: refactor req allocation
  io_uring: improve io_get_sqe
  io_uring: kill outdated comment about overflow flush
  io_uring: use user visible tail in io_uring_poll()
  io_uring: pass in io_issue_def to io_assign_file()
  io_uring: Enable KASAN for request cache
  io_uring: handle TIF_NOTIFY_RESUME when checking for task_work
  io_uring/msg-ring: ensure flags passing works for task_work completions
  ...

include/linux/io_uring_types.h

@@ -195,21 +195,23 @@ struct io_alloc_cache {
 struct io_ring_ctx {
 	/* const or read-mostly hot data */
 	struct {
-		struct percpu_ref	refs;
-
-		struct io_rings		*rings;
 		unsigned int		flags;
-		enum task_work_notify_mode	notify_method;
-		unsigned int		compat: 1;
 		unsigned int		drain_next: 1;
 		unsigned int		restricted: 1;
 		unsigned int		off_timeout_used: 1;
 		unsigned int		drain_active: 1;
-		unsigned int		drain_disabled: 1;
 		unsigned int		has_evfd: 1;
-		unsigned int		syscall_iopoll: 1;
 		/* all CQEs should be posted only by the submitter task */
 		unsigned int		task_complete: 1;
+		unsigned int		syscall_iopoll: 1;
+		unsigned int		poll_activated: 1;
+		unsigned int		drain_disabled: 1;
+		unsigned int		compat: 1;
+
+		enum task_work_notify_mode	notify_method;
+		struct io_rings			*rings;
+		struct task_struct		*submitter_task;
+		struct percpu_ref		refs;
 	} ____cacheline_aligned_in_smp;
 
 	/* submission data */
@@ -293,6 +295,7 @@ struct io_ring_ctx {
 		spinlock_t		completion_lock;
 
 		bool			poll_multi_queue;
+		bool			cq_waiting;
 
 		/*
 		 * ->iopoll_list is protected by the ctx->uring_lock for
@@ -318,9 +321,8 @@ struct io_ring_ctx {
 	} ____cacheline_aligned_in_smp;
 
 	/* Keep this last, we don't need it for the fast path */
-
+	struct wait_queue_head		poll_wq;
 	struct io_restriction		restrictions;
-	struct task_struct		*submitter_task;
 
 	/* slow path rsrc auxilary data, used by update/register */
 	struct io_rsrc_node		*rsrc_backup_node;
@@ -357,6 +359,7 @@ struct io_ring_ctx {
 	u32				iowq_limits[2];
 	bool				iowq_limits_set;
 
+	struct callback_head		poll_wq_task_work;
 	struct list_head		defer_list;
 	unsigned			sq_thread_idle;
 	/* protected by ->completion_lock */

include/uapi/linux/io_uring.h

@@ -347,6 +347,8 @@ enum {
  *				applicable for IORING_MSG_DATA, obviously.
  */
 #define IORING_MSG_RING_CQE_SKIP	(1U << 0)
+/* Pass through the flags from sqe->file_index to cqe->flags */
+#define IORING_MSG_RING_FLAGS_PASS	(1U << 1)
 
 /*
  * IO completion data structure (Completion Queue Entry)
@@ -470,6 +472,7 @@ struct io_uring_params {
 #define IORING_FEAT_RSRC_TAGS		(1U << 10)
 #define IORING_FEAT_CQE_SKIP		(1U << 11)
 #define IORING_FEAT_LINKED_FILE		(1U << 12)
+#define IORING_FEAT_REG_REG_RING	(1U << 13)
 
 /*
  * io_uring_register(2) opcodes and arguments
@@ -517,7 +520,10 @@ enum {
 	IORING_REGISTER_FILE_ALLOC_RANGE	= 25,
 
 	/* this goes last */
-	IORING_REGISTER_LAST
+	IORING_REGISTER_LAST,
+
+	/* flag added to the opcode to use a registered ring fd */
+	IORING_REGISTER_USE_REGISTERED_RING	= 1U << 31
 };
 
 /* io-wq worker categories */

io_uring/advise.c

@@ -39,6 +39,7 @@ int io_madvise_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	ma->addr = READ_ONCE(sqe->addr);
 	ma->len = READ_ONCE(sqe->len);
 	ma->advice = READ_ONCE(sqe->fadvise_advice);
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 #else
 	return -EOPNOTSUPP;
@@ -51,8 +52,7 @@ int io_madvise(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_madvise *ma = io_kiocb_to_cmd(req, struct io_madvise);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = do_madvise(current->mm, ma->addr, ma->len, ma->advice);
 	io_req_set_res(req, ret, 0);
@@ -62,6 +62,18 @@ int io_madvise(struct io_kiocb *req, unsigned int issue_flags)
 #endif
 }
 
+static bool io_fadvise_force_async(struct io_fadvise *fa)
+{
+	switch (fa->advice) {
+	case POSIX_FADV_NORMAL:
+	case POSIX_FADV_RANDOM:
+	case POSIX_FADV_SEQUENTIAL:
+		return false;
+	default:
+		return true;
+	}
+}
+
 int io_fadvise_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	struct io_fadvise *fa = io_kiocb_to_cmd(req, struct io_fadvise);
@@ -72,6 +84,8 @@ int io_fadvise_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	fa->offset = READ_ONCE(sqe->off);
 	fa->len = READ_ONCE(sqe->len);
 	fa->advice = READ_ONCE(sqe->fadvise_advice);
+	if (io_fadvise_force_async(fa))
+		req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -80,16 +94,7 @@ int io_fadvise(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_fadvise *fa = io_kiocb_to_cmd(req, struct io_fadvise);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK) {
-		switch (fa->advice) {
-		case POSIX_FADV_NORMAL:
-		case POSIX_FADV_RANDOM:
-		case POSIX_FADV_SEQUENTIAL:
-			break;
-		default:
-			return -EAGAIN;
-		}
-	}
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK && io_fadvise_force_async(fa));
 
 	ret = vfs_fadvise(req->file, fa->offset, fa->len, fa->advice);
 	if (ret < 0)

io_uring/fs.c

@@ -74,6 +74,7 @@ int io_renameat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	}
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -82,8 +83,7 @@ int io_renameat(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_rename *ren = io_kiocb_to_cmd(req, struct io_rename);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = do_renameat2(ren->old_dfd, ren->oldpath, ren->new_dfd,
 				ren->newpath, ren->flags);
@@ -123,6 +123,7 @@ int io_unlinkat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 		return PTR_ERR(un->filename);
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -131,8 +132,7 @@ int io_unlinkat(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_unlink *un = io_kiocb_to_cmd(req, struct io_unlink);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	if (un->flags & AT_REMOVEDIR)
 		ret = do_rmdir(un->dfd, un->filename);
@@ -170,6 +170,7 @@ int io_mkdirat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 		return PTR_ERR(mkd->filename);
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -178,8 +179,7 @@ int io_mkdirat(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_mkdir *mkd = io_kiocb_to_cmd(req, struct io_mkdir);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = do_mkdirat(mkd->dfd, mkd->filename, mkd->mode);
 
@@ -220,6 +220,7 @@ int io_symlinkat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	}
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -228,8 +229,7 @@ int io_symlinkat(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_link *sl = io_kiocb_to_cmd(req, struct io_link);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = do_symlinkat(sl->oldpath, sl->new_dfd, sl->newpath);
 
@@ -265,6 +265,7 @@ int io_linkat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	}
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -273,8 +274,7 @@ int io_linkat(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_link *lnk = io_kiocb_to_cmd(req, struct io_link);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = do_linkat(lnk->old_dfd, lnk->oldpath, lnk->new_dfd,
 				lnk->newpath, lnk->flags);

io_uring/io_uring.c

@@ -151,7 +151,7 @@ static void io_move_task_work_from_local(struct io_ring_ctx *ctx);
 static void __io_submit_flush_completions(struct io_ring_ctx *ctx);
 static __cold void io_fallback_tw(struct io_uring_task *tctx);
 
-static struct kmem_cache *req_cachep;
+struct kmem_cache *req_cachep;
 
 struct sock *io_uring_get_socket(struct file *file)
 {
@@ -230,6 +230,7 @@ static inline void req_fail_link_node(struct io_kiocb *req, int res)
 static inline void io_req_add_to_cache(struct io_kiocb *req, struct io_ring_ctx *ctx)
 {
 	wq_stack_add_head(&req->comp_list, &ctx->submit_state.free_list);
+	kasan_poison_object_data(req_cachep, req);
 }
 
 static __cold void io_ring_ctx_ref_free(struct percpu_ref *ref)
@@ -245,17 +246,15 @@ static __cold void io_fallback_req_func(struct work_struct *work)
 						fallback_work.work);
 	struct llist_node *node = llist_del_all(&ctx->fallback_llist);
 	struct io_kiocb *req, *tmp;
-	bool locked = false;
+	bool locked = true;
 
-	percpu_ref_get(&ctx->refs);
+	mutex_lock(&ctx->uring_lock);
 	llist_for_each_entry_safe(req, tmp, node, io_task_work.node)
 		req->io_task_work.func(req, &locked);
-
-	if (locked) {
+	if (WARN_ON_ONCE(!locked))
+		return;
 	io_submit_flush_completions(ctx);
 	mutex_unlock(&ctx->uring_lock);
-	}
-	percpu_ref_put(&ctx->refs);
 }
 
 static int io_alloc_hash_table(struct io_hash_table *table, unsigned bits)
@@ -316,6 +315,7 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
 	xa_init_flags(&ctx->personalities, XA_FLAGS_ALLOC1);
 	mutex_init(&ctx->uring_lock);
 	init_waitqueue_head(&ctx->cq_wait);
+	init_waitqueue_head(&ctx->poll_wq);
 	spin_lock_init(&ctx->completion_lock);
 	spin_lock_init(&ctx->timeout_lock);
 	INIT_WQ_LIST(&ctx->iopoll_list);
@@ -407,7 +407,7 @@ static inline void io_arm_ltimeout(struct io_kiocb *req)
 
 static void io_prep_async_work(struct io_kiocb *req)
 {
-	const struct io_op_def *def = &io_op_defs[req->opcode];
+	const struct io_issue_def *def = &io_issue_defs[req->opcode];
 	struct io_ring_ctx *ctx = req->ctx;
 
 	if (!(req->flags & REQ_F_CREDS)) {
@@ -572,6 +572,8 @@ static void io_eventfd_flush_signal(struct io_ring_ctx *ctx)
 
 void __io_commit_cqring_flush(struct io_ring_ctx *ctx)
 {
+	if (ctx->poll_activated)
+		io_poll_wq_wake(ctx);
 	if (ctx->off_timeout_used)
 		io_flush_timeouts(ctx);
 	if (ctx->drain_active) {
@@ -618,6 +620,25 @@ static inline void __io_cq_unlock_post(struct io_ring_ctx *ctx)
 	io_cqring_wake(ctx);
 }
 
+static inline void __io_cq_unlock_post_flush(struct io_ring_ctx *ctx)
+	__releases(ctx->completion_lock)
+{
+	io_commit_cqring(ctx);
+	__io_cq_unlock(ctx);
+	io_commit_cqring_flush(ctx);
+
+	/*
+	 * As ->task_complete implies that the ring is single tasked, cq_wait
+	 * may only be waited on by the current in io_cqring_wait(), but since
+	 * it will re-check the wakeup conditions once we return we can safely
+	 * skip waking it up.
+	 */
+	if (!(ctx->flags & IORING_SETUP_DEFER_TASKRUN)) {
+		smp_mb();
+		__io_cqring_wake(ctx);
+	}
+}
+
 void io_cq_unlock_post(struct io_ring_ctx *ctx)
 	__releases(ctx->completion_lock)
 {
@@ -645,7 +666,6 @@ static void io_cqring_overflow_kill(struct io_ring_ctx *ctx)
 	}
 }
 
-/* Returns true if there are no backlogged entries after the flush */
 static void __io_cqring_overflow_flush(struct io_ring_ctx *ctx)
 {
 	size_t cqe_size = sizeof(struct io_uring_cqe);
@@ -693,7 +713,8 @@ static void io_cqring_overflow_flush(struct io_ring_ctx *ctx)
 		io_cqring_do_overflow_flush(ctx);
 }
 
-void __io_put_task(struct task_struct *task, int nr)
+/* can be called by any task */
+static void io_put_task_remote(struct task_struct *task, int nr)
 {
 	struct io_uring_task *tctx = task->io_uring;
 
@@ -703,6 +724,21 @@ void __io_put_task(struct task_struct *task, int nr)
 	put_task_struct_many(task, nr);
 }
 
+/* used by a task to put its own references */
+static void io_put_task_local(struct task_struct *task, int nr)
+{
+	task->io_uring->cached_refs += nr;
+}
+
+/* must to be called somewhat shortly after putting a request */
+static inline void io_put_task(struct task_struct *task, int nr)
+{
+	if (likely(task == current))
+		io_put_task_local(task, nr);
+	else
+		io_put_task_remote(task, nr);
+}
+
 void io_task_refs_refill(struct io_uring_task *tctx)
 {
 	unsigned int refill = -tctx->cached_refs + IO_TCTX_REFS_CACHE_NR;
@@ -945,15 +981,15 @@ static void __io_req_complete_post(struct io_kiocb *req)
 				req->link = NULL;
 			}
 		}
+		io_put_kbuf_comp(req);
+		io_dismantle_req(req);
 		io_req_put_rsrc(req);
 		/*
 		 * Selected buffer deallocation in io_clean_op() assumes that
 		 * we don't hold ->completion_lock. Clean them here to avoid
 		 * deadlocks.
 		 */
-		io_put_kbuf_comp(req);
-		io_dismantle_req(req);
-		io_put_task(req->task, 1);
+		io_put_task_remote(req->task, 1);
 		wq_list_add_head(&req->comp_list, &ctx->locked_free_list);
 		ctx->locked_free_nr++;
 	}
@@ -980,7 +1016,7 @@ void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags)
 void io_req_defer_failed(struct io_kiocb *req, s32 res)
 	__must_hold(&ctx->uring_lock)
 {
-	const struct io_op_def *def = &io_op_defs[req->opcode];
+	const struct io_cold_def *def = &io_cold_defs[req->opcode];
 
 	lockdep_assert_held(&req->ctx->uring_lock);
 
@@ -1076,7 +1112,7 @@ __cold void io_free_req(struct io_kiocb *req)
 
 	io_req_put_rsrc(req);
 	io_dismantle_req(req);
-	io_put_task(req->task, 1);
+	io_put_task_remote(req->task, 1);
 
 	spin_lock(&ctx->completion_lock);
 	wq_list_add_head(&req->comp_list, &ctx->locked_free_list);
@@ -1130,7 +1166,7 @@ static unsigned int handle_tw_list(struct llist_node *node,
 {
 	unsigned int count = 0;
 
-	while (node != last) {
+	while (node && node != last) {
 		struct llist_node *next = node->next;
 		struct io_kiocb *req = container_of(node, struct io_kiocb,
 						    io_task_work.node);
@@ -1143,10 +1179,16 @@ static unsigned int handle_tw_list(struct llist_node *node,
 			/* if not contended, grab and improve batching */
 			*locked = mutex_trylock(&(*ctx)->uring_lock);
 			percpu_ref_get(&(*ctx)->refs);
-		}
+		} else if (!*locked)
+			*locked = mutex_trylock(&(*ctx)->uring_lock);
 		req->io_task_work.func(req, locked);
 		node = next;
 		count++;
+		if (unlikely(need_resched())) {
+			ctx_flush_and_put(*ctx, locked);
+			*ctx = NULL;
+			cond_resched();
+		}
 	}
 
 	return count;
@@ -1190,23 +1232,29 @@ void tctx_task_work(struct callback_head *cb)
 						  task_work);
 	struct llist_node fake = {};
 	struct llist_node *node;
-	unsigned int loops = 1;
-	unsigned int count;
+	unsigned int loops = 0;
+	unsigned int count = 0;
 
 	if (unlikely(current->flags & PF_EXITING)) {
 		io_fallback_tw(tctx);
 		return;
 	}
 
-	node = io_llist_xchg(&tctx->task_list, &fake);
-	count = handle_tw_list(node, &ctx, &uring_locked, NULL);
-	node = io_llist_cmpxchg(&tctx->task_list, &fake, NULL);
-	while (node != &fake) {
+	do {
 		loops++;
 		node = io_llist_xchg(&tctx->task_list, &fake);
 		count += handle_tw_list(node, &ctx, &uring_locked, &fake);
-		node = io_llist_cmpxchg(&tctx->task_list, &fake, NULL);
+
+		/* skip expensive cmpxchg if there are items in the list */
+		if (READ_ONCE(tctx->task_list.first) != &fake)
+			continue;
+		if (uring_locked && !wq_list_empty(&ctx->submit_state.compl_reqs)) {
+			io_submit_flush_completions(ctx);
+			if (READ_ONCE(tctx->task_list.first) != &fake)
+				continue;
 		}
+		node = io_llist_cmpxchg(&tctx->task_list, &fake, NULL);
+	} while (node != &fake);
 
 	ctx_flush_and_put(ctx, &uring_locked);
 
@@ -1241,7 +1289,7 @@ static void io_req_local_work_add(struct io_kiocb *req)
 		percpu_ref_put(&ctx->refs);
 		return;
 	}
-	/* need it for the following io_cqring_wake() */
+	/* needed for the following wake up */
 	smp_mb__after_atomic();
 
 	if (unlikely(atomic_read(&req->task->io_uring->in_idle))) {
@@ -1252,10 +1300,11 @@ static void io_req_local_work_add(struct io_kiocb *req)
 
 	if (ctx->flags & IORING_SETUP_TASKRUN_FLAG)
 		atomic_or(IORING_SQ_TASKRUN, &ctx->rings->sq_flags);
-
 	if (ctx->has_evfd)
 		io_eventfd_signal(ctx);
-	__io_cqring_wake(ctx);
+
+	if (READ_ONCE(ctx->cq_waiting))
+		wake_up_state(ctx->submitter_task, TASK_INTERRUPTIBLE);
 	percpu_ref_put(&ctx->refs);
 }
 
@@ -1296,21 +1345,19 @@ static void __cold io_move_task_work_from_local(struct io_ring_ctx *ctx)
 	}
 }
 
-int __io_run_local_work(struct io_ring_ctx *ctx, bool *locked)
+static int __io_run_local_work(struct io_ring_ctx *ctx, bool *locked)
 {
 	struct llist_node *node;
-	struct llist_node fake;
-	struct llist_node *current_final = NULL;
-	int ret;
-	unsigned int loops = 1;
+	unsigned int loops = 0;
+	int ret = 0;
 
-	if (unlikely(ctx->submitter_task != current))
+	if (WARN_ON_ONCE(ctx->submitter_task != current))
 		return -EEXIST;
-
-	node = io_llist_xchg(&ctx->work_llist, &fake);
-	ret = 0;
+	if (ctx->flags & IORING_SETUP_TASKRUN_FLAG)
+		atomic_andnot(IORING_SQ_TASKRUN, &ctx->rings->sq_flags);
 again:
-	while (node != current_final) {
+	node = io_llist_xchg(&ctx->work_llist, NULL);
+	while (node) {
 		struct llist_node *next = node->next;
 		struct io_kiocb *req = container_of(node, struct io_kiocb,
 						    io_task_work.node);
@@ -1319,26 +1366,20 @@ int __io_run_local_work(struct io_ring_ctx *ctx, bool *locked)
 		ret++;
 		node = next;
 	}
-
-	if (ctx->flags & IORING_SETUP_TASKRUN_FLAG)
-		atomic_andnot(IORING_SQ_TASKRUN, &ctx->rings->sq_flags);
-
-	node = io_llist_cmpxchg(&ctx->work_llist, &fake, NULL);
-	if (node != &fake) {
 	loops++;
-		current_final = &fake;
-		node = io_llist_xchg(&ctx->work_llist, &fake);
-		goto again;
-	}
 
-	if (*locked)
+	if (!llist_empty(&ctx->work_llist))
+		goto again;
+	if (*locked) {
 		io_submit_flush_completions(ctx);
+		if (!llist_empty(&ctx->work_llist))
+			goto again;
+	}
 	trace_io_uring_local_work_run(ctx, ret, loops);
 	return ret;
-
 }
 
-int io_run_local_work(struct io_ring_ctx *ctx)
+static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
 {
 	bool locked;
 	int ret;
@@ -1346,8 +1387,19 @@ int io_run_local_work(struct io_ring_ctx *ctx)
 	if (llist_empty(&ctx->work_llist))
 		return 0;
 
-	__set_current_state(TASK_RUNNING);
-	locked = mutex_trylock(&ctx->uring_lock);
+	locked = true;
+	ret = __io_run_local_work(ctx, &locked);
+	/* shouldn't happen! */
+	if (WARN_ON_ONCE(!locked))
+		mutex_lock(&ctx->uring_lock);
+	return ret;
+}
+
+static int io_run_local_work(struct io_ring_ctx *ctx)
+{
+	bool locked = mutex_trylock(&ctx->uring_lock);
+	int ret;
+
 	ret = __io_run_local_work(ctx, &locked);
 	if (locked)
 		mutex_unlock(&ctx->uring_lock);
@@ -1365,10 +1417,12 @@ void io_req_task_submit(struct io_kiocb *req, bool *locked)
 {
 	io_tw_lock(req->ctx, locked);
 	/* req->task == current here, checking PF_EXITING is safe */
-	if (likely(!(req->task->flags & PF_EXITING)))
-		io_queue_sqe(req);
-	else
+	if (unlikely(req->task->flags & PF_EXITING))
 		io_req_defer_failed(req, -EFAULT);
+	else if (req->flags & REQ_F_FORCE_ASYNC)
+		io_queue_iowq(req, locked);
+	else
+		io_queue_sqe(req);
 }
 
 void io_req_task_queue_fail(struct io_kiocb *req, int ret)
@@ -1467,7 +1521,7 @@ static void __io_submit_flush_completions(struct io_ring_ctx *ctx)
 			}
 		}
 	}
-	__io_cq_unlock_post(ctx);
+	__io_cq_unlock_post_flush(ctx);
 
 	if (!wq_list_empty(&ctx->submit_state.compl_reqs)) {
 		io_free_batch_list(ctx, state->compl_reqs.first);
@@ -1708,8 +1762,8 @@ unsigned int io_file_get_flags(struct file *file)
 
 bool 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);
+	WARN_ON_ONCE(!io_cold_defs[req->opcode].async_size);
+	req->async_data = kmalloc(io_cold_defs[req->opcode].async_size, GFP_KERNEL);
 	if (req->async_data) {
 		req->flags |= REQ_F_ASYNC_DATA;
 		return false;
@@ -1719,20 +1773,21 @@ bool io_alloc_async_data(struct io_kiocb *req)
 
 int io_req_prep_async(struct io_kiocb *req)
 {
-	const struct io_op_def *def = &io_op_defs[req->opcode];
+	const struct io_cold_def *cdef = &io_cold_defs[req->opcode];
+	const struct io_issue_def *def = &io_issue_defs[req->opcode];
 
 	/* assign early for deferred execution for non-fixed file */
 	if (def->needs_file && !(req->flags & REQ_F_FIXED_FILE))
 		req->file = io_file_get_normal(req, req->cqe.fd);
-	if (!def->prep_async)
+	if (!cdef->prep_async)
 		return 0;
 	if (WARN_ON_ONCE(req_has_async_data(req)))
 		return -EFAULT;
-	if (!io_op_defs[req->opcode].manual_alloc) {
+	if (!def->manual_alloc) {
 		if (io_alloc_async_data(req))
 			return -EAGAIN;
 	}
-	return def->prep_async(req);
+	return cdef->prep_async(req);
 }
 
 static u32 io_get_sequence(struct io_kiocb *req)
@@ -1796,7 +1851,7 @@ static void io_clean_op(struct io_kiocb *req)
 	}
 
 	if (req->flags & REQ_F_NEED_CLEANUP) {
-		const struct io_op_def *def = &io_op_defs[req->opcode];
+		const struct io_cold_def *def = &io_cold_defs[req->opcode];
 
 		if (def->cleanup)
 			def->cleanup(req);
@@ -1820,9 +1875,10 @@ static void io_clean_op(struct io_kiocb *req)
 	req->flags &= ~IO_REQ_CLEAN_FLAGS;
 }
 
-static bool io_assign_file(struct io_kiocb *req, unsigned int issue_flags)
+static bool io_assign_file(struct io_kiocb *req, const struct io_issue_def *def,
+			   unsigned int issue_flags)
 {
-	if (req->file || !io_op_defs[req->opcode].needs_file)
+	if (req->file || !def->needs_file)
 		return true;
 
 	if (req->flags & REQ_F_FIXED_FILE)
@@ -1835,11 +1891,11 @@ static bool io_assign_file(struct io_kiocb *req, unsigned int issue_flags)
 
 static int io_issue_sqe(struct io_kiocb *req, unsigned int issue_flags)
 {
-	const struct io_op_def *def = &io_op_defs[req->opcode];
+	const struct io_issue_def *def = &io_issue_defs[req->opcode];
 	const struct cred *creds = NULL;
 	int ret;
 
-	if (unlikely(!io_assign_file(req, issue_flags)))
+	if (unlikely(!io_assign_file(req, def, issue_flags)))
 		return -EBADF;
 
 	if (unlikely((req->flags & REQ_F_CREDS) && req->creds != current_cred()))
@@ -1889,7 +1945,7 @@ struct io_wq_work *io_wq_free_work(struct io_wq_work *work)
 void io_wq_submit_work(struct io_wq_work *work)
 {
 	struct io_kiocb *req = container_of(work, struct io_kiocb, work);
-	const struct io_op_def *def = &io_op_defs[req->opcode];
+	const struct io_issue_def *def = &io_issue_defs[req->opcode];
 	unsigned int issue_flags = IO_URING_F_UNLOCKED | IO_URING_F_IOWQ;
 	bool needs_poll = false;
 	int ret = 0, err = -ECANCELED;
@@ -1908,7 +1964,7 @@ void io_wq_submit_work(struct io_wq_work *work)
 		io_req_task_queue_fail(req, err);
 		return;
 	}
-	if (!io_assign_file(req, issue_flags)) {
+	if (!io_assign_file(req, def, issue_flags)) {
 		err = -EBADF;
 		work->flags |= IO_WQ_WORK_CANCEL;
 		goto fail;
@@ -2104,7 +2160,7 @@ static int io_init_req(struct io_ring_ctx *ctx, struct io_kiocb *req,
 		       const struct io_uring_sqe *sqe)
 	__must_hold(&ctx->uring_lock)
 {
-	const struct io_op_def *def;
+	const struct io_issue_def *def;
 	unsigned int sqe_flags;
 	int personality;
 	u8 opcode;
@@ -2122,7 +2178,7 @@ static int io_init_req(struct io_ring_ctx *ctx, struct io_kiocb *req,
 		req->opcode = 0;
 		return -EINVAL;
 	}
-	def = &io_op_defs[opcode];
+	def = &io_issue_defs[opcode];
 	if (unlikely(sqe_flags & ~SQE_COMMON_FLAGS)) {
 		/* enforce forwards compatibility on users */
 		if (sqe_flags & ~SQE_VALID_FLAGS)
@@ -2333,7 +2389,7 @@ static void io_commit_sqring(struct io_ring_ctx *ctx)
  * used, it's important that those reads are done through READ_ONCE() to
  * prevent a re-load down the line.
  */
-static const struct io_uring_sqe *io_get_sqe(struct io_ring_ctx *ctx)
+static bool io_get_sqe(struct io_ring_ctx *ctx, const struct io_uring_sqe **sqe)
 {
 	unsigned head, mask = ctx->sq_entries - 1;
 	unsigned sq_idx = ctx->cached_sq_head++ & mask;
@@ -2351,14 +2407,15 @@ static const struct io_uring_sqe *io_get_sqe(struct io_ring_ctx *ctx)
 		/* double index for 128-byte SQEs, twice as long */
 		if (ctx->flags & IORING_SETUP_SQE128)
 			head <<= 1;
-		return &ctx->sq_sqes[head];
+		*sqe = &ctx->sq_sqes[head];
+		return true;
 	}
 
 	/* drop invalid entries */
 	ctx->cq_extra--;
 	WRITE_ONCE(ctx->rings->sq_dropped,
 		   READ_ONCE(ctx->rings->sq_dropped) + 1);
-	return NULL;
+	return false;
 }
 
 int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr)
@@ -2379,11 +2436,9 @@ int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr)
 		const struct io_uring_sqe *sqe;
 		struct io_kiocb *req;
 
-		if (unlikely(!io_alloc_req_refill(ctx)))
+		if (unlikely(!io_alloc_req(ctx, &req)))
 			break;
-		req = io_alloc_req(ctx);
-		sqe = io_get_sqe(ctx);
-		if (unlikely(!sqe)) {
+		if (unlikely(!io_get_sqe(ctx, &sqe))) {
 			io_req_add_to_cache(req, ctx);
 			break;
 		}
@@ -2418,13 +2473,13 @@ struct io_wait_queue {
 	struct io_ring_ctx *ctx;
 	unsigned cq_tail;
 	unsigned nr_timeouts;
+	ktime_t timeout;
 };
 
 static inline bool io_has_work(struct io_ring_ctx *ctx)
 {
 	return test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq) ||
-	       ((ctx->flags & IORING_SETUP_DEFER_TASKRUN) &&
-		!llist_empty(&ctx->work_llist));
+	       !llist_empty(&ctx->work_llist);
 }
 
 static inline bool io_should_wake(struct io_wait_queue *iowq)
@@ -2443,22 +2498,25 @@ static inline bool io_should_wake(struct io_wait_queue *iowq)
 static int io_wake_function(struct wait_queue_entry *curr, unsigned int mode,
 			    int wake_flags, void *key)
 {
-	struct io_wait_queue *iowq = container_of(curr, struct io_wait_queue,
-							wq);
-	struct io_ring_ctx *ctx = iowq->ctx;
+	struct io_wait_queue *iowq = container_of(curr, struct io_wait_queue, wq);
 
 	/*
 	 * Cannot safely flush overflowed CQEs from here, ensure we wake up
 	 * the task, and the next invocation will do it.
 	 */
-	if (io_should_wake(iowq) || io_has_work(ctx))
+	if (io_should_wake(iowq) || io_has_work(iowq->ctx))
 		return autoremove_wake_function(curr, mode, wake_flags, key);
 	return -1;
 }
 
 int io_run_task_work_sig(struct io_ring_ctx *ctx)
 {
-	if (io_run_task_work_ctx(ctx) > 0)
+	if (!llist_empty(&ctx->work_llist)) {
+		__set_current_state(TASK_RUNNING);
+		if (io_run_local_work(ctx) > 0)
+			return 1;
+	}
+	if (io_run_task_work() > 0)
 		return 1;
 	if (task_sigpending(current))
 		return -EINTR;
@@ -2467,35 +2525,23 @@ int io_run_task_work_sig(struct io_ring_ctx *ctx)
 
 /* when returns >0, the caller should retry */
 static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
-					  struct io_wait_queue *iowq,
-					  ktime_t *timeout)
+					  struct io_wait_queue *iowq)
 {
-	int ret;
-	unsigned long check_cq;
-
-	/* make sure we run task_work before checking for signals */
-	ret = io_run_task_work_sig(ctx);
-	if (ret || io_should_wake(iowq))
-		return ret;
-
-	check_cq = READ_ONCE(ctx->check_cq);
-	if (unlikely(check_cq)) {
-		/* let the caller flush overflows, retry */
-		if (check_cq & BIT(IO_CHECK_CQ_OVERFLOW_BIT))
+	if (unlikely(READ_ONCE(ctx->check_cq)))
 		return 1;
-		if (check_cq & BIT(IO_CHECK_CQ_DROPPED_BIT))
-			return -EBADR;
-	}
-	if (!schedule_hrtimeout(timeout, HRTIMER_MODE_ABS))
+	if (unlikely(!llist_empty(&ctx->work_llist)))
+		return 1;
+	if (unlikely(test_thread_flag(TIF_NOTIFY_SIGNAL)))
+		return 1;
+	if (unlikely(task_sigpending(current)))
+		return -EINTR;
+	if (unlikely(io_should_wake(iowq)))
+		return 0;
+	if (iowq->timeout == KTIME_MAX)
+		schedule();
+	else if (!schedule_hrtimeout(&iowq->timeout, HRTIMER_MODE_ABS))
 		return -ETIME;
-
-	/*
-	 * Run task_work after scheduling. If we got woken because of
-	 * task_work being processed, run it now rather than let the caller
-	 * do another wait loop.
-	 */
-	ret = io_run_task_work_sig(ctx);
-	return ret < 0 ? ret : 1;
+	return 0;
 }
 
 /*
@@ -2508,23 +2554,17 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 {
 	struct io_wait_queue iowq;
 	struct io_rings *rings = ctx->rings;
-	ktime_t timeout = KTIME_MAX;
 	int ret;
 
 	if (!io_allowed_run_tw(ctx))
 		return -EEXIST;
-
-	do {
-		/* always run at least 1 task work to process local work */
-		ret = io_run_task_work_ctx(ctx);
-		if (ret < 0)
-			return ret;
+	if (!llist_empty(&ctx->work_llist))
+		io_run_local_work(ctx);
+	io_run_task_work();
 	io_cqring_overflow_flush(ctx);
-
 	/* if user messes with these they will just get an early return */
 	if (__io_cqring_events_user(ctx) >= min_events)
 		return 0;
-	} while (ret > 0);
 
 	if (sig) {
 #ifdef CONFIG_COMPAT
@@ -2539,35 +2579,68 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
 			return ret;
 	}
 
-	if (uts) {
-		struct timespec64 ts;
-
-		if (get_timespec64(&ts, uts))
-			return -EFAULT;
-		timeout = ktime_add_ns(timespec64_to_ktime(ts), ktime_get_ns());
-	}
-
 	init_waitqueue_func_entry(&iowq.wq, io_wake_function);
 	iowq.wq.private = current;
 	INIT_LIST_HEAD(&iowq.wq.entry);
 	iowq.ctx = ctx;
 	iowq.nr_timeouts = atomic_read(&ctx->cq_timeouts);
 	iowq.cq_tail = READ_ONCE(ctx->rings->cq.head) + min_events;
+	iowq.timeout = KTIME_MAX;
+
+	if (uts) {
+		struct timespec64 ts;
+
+		if (get_timespec64(&ts, uts))
+			return -EFAULT;
+		iowq.timeout = ktime_add_ns(timespec64_to_ktime(ts), ktime_get_ns());
+	}
 
 	trace_io_uring_cqring_wait(ctx, min_events);
 	do {
-		if (test_bit(IO_CHECK_CQ_OVERFLOW_BIT, &ctx->check_cq)) {
-			finish_wait(&ctx->cq_wait, &iowq.wq);
-			io_cqring_do_overflow_flush(ctx);
-		}
+		unsigned long check_cq;
+
+		if (ctx->flags & IORING_SETUP_DEFER_TASKRUN) {
+			WRITE_ONCE(ctx->cq_waiting, 1);
+			set_current_state(TASK_INTERRUPTIBLE);
+		} else {
 			prepare_to_wait_exclusive(&ctx->cq_wait, &iowq.wq,
 							TASK_INTERRUPTIBLE);
-		ret = io_cqring_wait_schedule(ctx, &iowq, &timeout);
-		if (__io_cqring_events_user(ctx) >= min_events)
+		}
+
+		ret = io_cqring_wait_schedule(ctx, &iowq);
+		__set_current_state(TASK_RUNNING);
+		WRITE_ONCE(ctx->cq_waiting, 0);
+
+		if (ret < 0)
 			break;
+		/*
+		 * Run task_work after scheduling and before io_should_wake().
+		 * If we got woken because of task_work being processed, run it
+		 * now rather than let the caller do another wait loop.
+		 */
+		io_run_task_work();
+		if (!llist_empty(&ctx->work_llist))
+			io_run_local_work(ctx);
+
+		check_cq = READ_ONCE(ctx->check_cq);
+		if (unlikely(check_cq)) {
+			/* let the caller flush overflows, retry */
+			if (check_cq & BIT(IO_CHECK_CQ_OVERFLOW_BIT))
+				io_cqring_do_overflow_flush(ctx);
+			if (check_cq & BIT(IO_CHECK_CQ_DROPPED_BIT)) {
+				ret = -EBADR;
+				break;
+			}
+		}
+
+		if (io_should_wake(&iowq)) {
+			ret = 0;
+			break;
+		}
 		cond_resched();
-	} while (ret > 0);
+	} while (1);
 
+	if (!(ctx->flags & IORING_SETUP_DEFER_TASKRUN))
 		finish_wait(&ctx->cq_wait, &iowq.wq);
 	restore_saved_sigmask_unless(ret == -EINTR);
 
@@ -2683,14 +2756,14 @@ static int io_eventfd_unregister(struct io_ring_ctx *ctx)
 
 static void io_req_caches_free(struct io_ring_ctx *ctx)
 {
+	struct io_kiocb *req;
 	int nr = 0;
 
 	mutex_lock(&ctx->uring_lock);
 	io_flush_cached_locked_reqs(ctx, &ctx->submit_state);
 
 	while (!io_req_cache_empty(ctx)) {
-		struct io_kiocb *req = io_alloc_req(ctx);
-
+		req = io_extract_req(ctx);
 		kmem_cache_free(req_cachep, req);
 		nr++;
 	}
@@ -2762,12 +2835,54 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)
 	kfree(ctx);
 }
 
+static __cold void io_activate_pollwq_cb(struct callback_head *cb)
+{
+	struct io_ring_ctx *ctx = container_of(cb, struct io_ring_ctx,
+					       poll_wq_task_work);
+
+	mutex_lock(&ctx->uring_lock);
+	ctx->poll_activated = true;
+	mutex_unlock(&ctx->uring_lock);
+
+	/*
+	 * Wake ups for some events between start of polling and activation
+	 * might've been lost due to loose synchronisation.
+	 */
+	wake_up_all(&ctx->poll_wq);
+	percpu_ref_put(&ctx->refs);
+}
+
+static __cold void io_activate_pollwq(struct io_ring_ctx *ctx)
+{
+	spin_lock(&ctx->completion_lock);
+	/* already activated or in progress */
+	if (ctx->poll_activated || ctx->poll_wq_task_work.func)
+		goto out;
+	if (WARN_ON_ONCE(!ctx->task_complete))
+		goto out;
+	if (!ctx->submitter_task)
+		goto out;
+	/*
+	 * with ->submitter_task only the submitter task completes requests, we
+	 * only need to sync with it, which is done by injecting a tw
+	 */
+	init_task_work(&ctx->poll_wq_task_work, io_activate_pollwq_cb);
+	percpu_ref_get(&ctx->refs);
+	if (task_work_add(ctx->submitter_task, &ctx->poll_wq_task_work, TWA_SIGNAL))
+		percpu_ref_put(&ctx->refs);
+out:
+	spin_unlock(&ctx->completion_lock);
+}
+
 static __poll_t io_uring_poll(struct file *file, poll_table *wait)
 {
 	struct io_ring_ctx *ctx = file->private_data;
 	__poll_t mask = 0;
 
-	poll_wait(file, &ctx->cq_wait, wait);
+	if (unlikely(!ctx->poll_activated))
+		io_activate_pollwq(ctx);
+
+	poll_wait(file, &ctx->poll_wq, wait);
 	/*
 	 * synchronizes with barrier from wq_has_sleeper call in
 	 * io_commit_cqring
@@ -2790,7 +2905,7 @@ static __poll_t io_uring_poll(struct file *file, poll_table *wait)
 	 * pushes them to do the flush.
 	 */
 
-	if (io_cqring_events(ctx) || io_has_work(ctx))
+	if (__io_cqring_events_user(ctx) || io_has_work(ctx))
 		mask |= EPOLLIN | EPOLLRDNORM;
 
 	return mask;
@@ -3053,10 +3168,12 @@ static __cold bool io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
 		while (!wq_list_empty(&ctx->iopoll_list)) {
 			io_iopoll_try_reap_events(ctx);
 			ret = true;
+			cond_resched();
 		}
 	}
 
-	if (ctx->flags & IORING_SETUP_DEFER_TASKRUN)
+	if ((ctx->flags & IORING_SETUP_DEFER_TASKRUN) &&
+	    io_allowed_defer_tw_run(ctx))
 		ret |= io_run_local_work(ctx) > 0;
 	ret |= io_cancel_defer_files(ctx, task, cancel_all);
 	mutex_lock(&ctx->uring_lock);
@@ -3328,11 +3445,9 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
 		}
 		if (flags & IORING_ENTER_SQ_WAKEUP)
 			wake_up(&ctx->sq_data->wait);
-		if (flags & IORING_ENTER_SQ_WAIT) {
-			ret = io_sqpoll_wait_sq(ctx);
-			if (ret)
-				goto out;
-		}
+		if (flags & IORING_ENTER_SQ_WAIT)
+			io_sqpoll_wait_sq(ctx);
+
 		ret = to_submit;
 	} else if (to_submit) {
 		ret = io_uring_add_tctx_node(ctx);
@@ -3572,6 +3687,13 @@ static __cold int io_uring_create(unsigned entries, struct io_uring_params *p,
 	    !(ctx->flags & IORING_SETUP_SQPOLL))
 		ctx->task_complete = true;
 
+	/*
+	 * lazy poll_wq activation relies on ->task_complete for synchronisation
+	 * purposes, see io_activate_pollwq()
+	 */
+	if (!ctx->task_complete)
+		ctx->poll_activated = true;
+
 	/*
 	 * When SETUP_IOPOLL and SETUP_SQPOLL are both enabled, user
 	 * space applications don't need to do io completion events
@@ -3663,7 +3785,7 @@ static __cold int io_uring_create(unsigned entries, struct io_uring_params *p,
 			IORING_FEAT_POLL_32BITS | IORING_FEAT_SQPOLL_NONFIXED |
 			IORING_FEAT_EXT_ARG | IORING_FEAT_NATIVE_WORKERS |
 			IORING_FEAT_RSRC_TAGS | IORING_FEAT_CQE_SKIP |
-			IORING_FEAT_LINKED_FILE;
+			IORING_FEAT_LINKED_FILE | IORING_FEAT_REG_REG_RING;
 
 	if (copy_to_user(params, p, sizeof(*p))) {
 		ret = -EFAULT;
@@ -3760,7 +3882,7 @@ static __cold int io_probe(struct io_ring_ctx *ctx, void __user *arg,
 
 	for (i = 0; i < nr_args; i++) {
 		p->ops[i].op = i;
-		if (!io_op_defs[i].not_supported)
+		if (!io_issue_defs[i].not_supported)
 			p->ops[i].flags = IO_URING_OP_SUPPORTED;
 	}
 	p->ops_len = i;
@@ -3865,8 +3987,15 @@ static int io_register_enable_rings(struct io_ring_ctx *ctx)
 	if (!(ctx->flags & IORING_SETUP_R_DISABLED))
 		return -EBADFD;
 
-	if (ctx->flags & IORING_SETUP_SINGLE_ISSUER && !ctx->submitter_task)
+	if (ctx->flags & IORING_SETUP_SINGLE_ISSUER && !ctx->submitter_task) {
 		WRITE_ONCE(ctx->submitter_task, get_task_struct(current));
+		/*
+		 * Lazy activation attempts would fail if it was polled before
+		 * submitter_task is set.
+		 */
+		if (wq_has_sleeper(&ctx->poll_wq))
+			io_activate_pollwq(ctx);
+	}
 
 	if (ctx->restrictions.registered)
 		ctx->restricted = 1;
@@ -4177,17 +4306,36 @@ SYSCALL_DEFINE4(io_uring_register, unsigned int, fd, unsigned int, opcode,
 	struct io_ring_ctx *ctx;
 	long ret = -EBADF;
 	struct fd f;
+	bool use_registered_ring;
+
+	use_registered_ring = !!(opcode & IORING_REGISTER_USE_REGISTERED_RING);
+	opcode &= ~IORING_REGISTER_USE_REGISTERED_RING;
 
 	if (opcode >= IORING_REGISTER_LAST)
 		return -EINVAL;
 
+	if (use_registered_ring) {
+		/*
+		 * Ring fd has been registered via IORING_REGISTER_RING_FDS, we
+		 * need only dereference our task private array to find it.
+		 */
+		struct io_uring_task *tctx = current->io_uring;
+
+		if (unlikely(!tctx || fd >= IO_RINGFD_REG_MAX))
+			return -EINVAL;
+		fd = array_index_nospec(fd, IO_RINGFD_REG_MAX);
+		f.file = tctx->registered_rings[fd];
+		f.flags = 0;
+		if (unlikely(!f.file))
+			return -EBADF;
+	} else {
 		f = fdget(fd);
-	if (!f.file)
+		if (unlikely(!f.file))
 			return -EBADF;
-
 		ret = -EOPNOTSUPP;
 		if (!io_is_uring_fops(f.file))
 			goto out_fput;
+	}
 
 	ctx = f.file->private_data;
 

io_uring/io_uring.h

@@ -3,6 +3,8 @@
 
 #include <linux/errno.h>
 #include <linux/lockdep.h>
+#include <linux/resume_user_mode.h>
+#include <linux/kasan.h>
 #include <linux/io_uring_types.h>
 #include <uapi/linux/eventpoll.h>
 #include "io-wq.h"
@@ -28,8 +30,6 @@ enum {
 struct io_uring_cqe *__io_get_cqe(struct io_ring_ctx *ctx, bool overflow);
 bool io_req_cqe_overflow(struct io_kiocb *req);
 int io_run_task_work_sig(struct io_ring_ctx *ctx);
-int __io_run_local_work(struct io_ring_ctx *ctx, bool *locked);
-int io_run_local_work(struct io_ring_ctx *ctx);
 void io_req_defer_failed(struct io_kiocb *req, s32 res);
 void io_req_complete_post(struct io_kiocb *req, unsigned issue_flags);
 bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags);
@@ -72,7 +72,6 @@ void io_wq_submit_work(struct io_wq_work *work);
 
 void io_free_req(struct io_kiocb *req);
 void io_queue_next(struct io_kiocb *req);
-void __io_put_task(struct task_struct *task, int nr);
 void io_task_refs_refill(struct io_uring_task *tctx);
 bool __io_alloc_req_refill(struct io_ring_ctx *ctx);
 
@@ -222,6 +221,13 @@ static inline void io_commit_cqring(struct io_ring_ctx *ctx)
 	smp_store_release(&ctx->rings->cq.tail, ctx->cached_cq_tail);
 }
 
+static inline void io_poll_wq_wake(struct io_ring_ctx *ctx)
+{
+	if (wq_has_sleeper(&ctx->poll_wq))
+		__wake_up(&ctx->poll_wq, TASK_NORMAL, 0,
+				poll_to_key(EPOLL_URING_WAKE | EPOLLIN));
+}
+
 /* requires smb_mb() prior, see wq_has_sleeper() */
 static inline void __io_cqring_wake(struct io_ring_ctx *ctx)
 {
@@ -270,6 +276,15 @@ static inline int io_run_task_work(void)
 	 */
 	if (test_thread_flag(TIF_NOTIFY_SIGNAL))
 		clear_notify_signal();
+	/*
+	 * PF_IO_WORKER never returns to userspace, so check here if we have
+	 * notify work that needs processing.
+	 */
+	if (current->flags & PF_IO_WORKER &&
+	    test_thread_flag(TIF_NOTIFY_RESUME)) {
+		__set_current_state(TASK_RUNNING);
+		resume_user_mode_work(NULL);
+	}
 	if (task_work_pending(current)) {
 		__set_current_state(TASK_RUNNING);
 		task_work_run();
@@ -284,42 +299,6 @@ static inline bool io_task_work_pending(struct io_ring_ctx *ctx)
 	return task_work_pending(current) || !wq_list_empty(&ctx->work_llist);
 }
 
-static inline int io_run_task_work_ctx(struct io_ring_ctx *ctx)
-{
-	int ret = 0;
-	int ret2;
-
-	if (ctx->flags & IORING_SETUP_DEFER_TASKRUN)
-		ret = io_run_local_work(ctx);
-
-	/* want to run this after in case more is added */
-	ret2 = io_run_task_work();
-
-	/* Try propagate error in favour of if tasks were run,
-	 * but still make sure to run them if requested
-	 */
-	if (ret >= 0)
-		ret += ret2;
-
-	return ret;
-}
-
-static inline int io_run_local_work_locked(struct io_ring_ctx *ctx)
-{
-	bool locked;
-	int ret;
-
-	if (llist_empty(&ctx->work_llist))
-		return 0;
-
-	locked = true;
-	ret = __io_run_local_work(ctx, &locked);
-	/* shouldn't happen! */
-	if (WARN_ON_ONCE(!locked))
-		mutex_lock(&ctx->uring_lock);
-	return ret;
-}
-
 static inline void io_tw_lock(struct io_ring_ctx *ctx, bool *locked)
 {
 	if (!*locked) {
@@ -345,19 +324,11 @@ static inline void io_req_complete_defer(struct io_kiocb *req)
 
 static inline void io_commit_cqring_flush(struct io_ring_ctx *ctx)
 {
-	if (unlikely(ctx->off_timeout_used || ctx->drain_active || ctx->has_evfd))
+	if (unlikely(ctx->off_timeout_used || ctx->drain_active ||
+		     ctx->has_evfd || ctx->poll_activated))
 		__io_commit_cqring_flush(ctx);
 }
 
-/* must to be called somewhat shortly after putting a request */
-static inline void io_put_task(struct task_struct *task, int nr)
-{
-	if (likely(task == current))
-		task->io_uring->cached_refs += nr;
-	else
-		__io_put_task(task, nr);
-}
-
 static inline void io_get_task_refs(int nr)
 {
 	struct io_uring_task *tctx = current->io_uring;
@@ -372,19 +343,31 @@ static inline bool io_req_cache_empty(struct io_ring_ctx *ctx)
 	return !ctx->submit_state.free_list.next;
 }
 
-static inline bool io_alloc_req_refill(struct io_ring_ctx *ctx)
+extern struct kmem_cache *req_cachep;
+
+static inline struct io_kiocb *io_extract_req(struct io_ring_ctx *ctx)
 {
-	if (unlikely(io_req_cache_empty(ctx)))
-		return __io_alloc_req_refill(ctx);
-	return true;
+	struct io_kiocb *req;
+
+	req = container_of(ctx->submit_state.free_list.next, struct io_kiocb, comp_list);
+	kasan_unpoison_object_data(req_cachep, req);
+	wq_stack_extract(&ctx->submit_state.free_list);
+	return req;
 }
 
-static inline struct io_kiocb *io_alloc_req(struct io_ring_ctx *ctx)
+static inline bool io_alloc_req(struct io_ring_ctx *ctx, struct io_kiocb **req)
 {
-	struct io_wq_work_node *node;
+	if (unlikely(io_req_cache_empty(ctx))) {
+		if (!__io_alloc_req_refill(ctx))
+			return false;
+	}
+	*req = io_extract_req(ctx);
+	return true;
+}
 
-	node = wq_stack_extract(&ctx->submit_state.free_list);
-	return container_of(node, struct io_kiocb, comp_list);
+static inline bool io_allowed_defer_tw_run(struct io_ring_ctx *ctx)
+{
+	return likely(ctx->submitter_task == current);
 }
 
 static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)

io_uring/msg_ring.c

@@ -13,6 +13,11 @@
 #include "filetable.h"
 #include "msg_ring.h"
 
+
+/* All valid masks for MSG_RING */
+#define IORING_MSG_RING_MASK		(IORING_MSG_RING_CQE_SKIP | \
+					IORING_MSG_RING_FLAGS_PASS)
+
 struct io_msg {
 	struct file			*file;
 	struct file			*src_file;
@@ -21,7 +26,10 @@ struct io_msg {
 	u32 len;
 	u32 cmd;
 	u32 src_fd;
+	union {
 		u32 dst_fd;
+		u32 cqe_flags;
+	};
 	u32 flags;
 };
 
@@ -91,6 +99,11 @@ static void io_msg_tw_complete(struct callback_head *head)
 	if (current->flags & PF_EXITING) {
 		ret = -EOWNERDEAD;
 	} else {
+		u32 flags = 0;
+
+		if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
+			flags = msg->cqe_flags;
+
 		/*
 		 * If the target ring is using IOPOLL mode, then we need to be
 		 * holding the uring_lock for posting completions. Other ring
@@ -99,7 +112,7 @@ static void io_msg_tw_complete(struct callback_head *head)
 		 */
 		if (target_ctx->flags & IORING_SETUP_IOPOLL)
 			mutex_lock(&target_ctx->uring_lock);
-		if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
+		if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
 			ret = -EOVERFLOW;
 		if (target_ctx->flags & IORING_SETUP_IOPOLL)
 			mutex_unlock(&target_ctx->uring_lock);
@@ -114,9 +127,12 @@ static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags)
 {
 	struct io_ring_ctx *target_ctx = req->file->private_data;
 	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
+	u32 flags = 0;
 	int ret;
 
-	if (msg->src_fd || msg->dst_fd || msg->flags)
+	if (msg->src_fd || msg->flags & ~IORING_MSG_RING_FLAGS_PASS)
+		return -EINVAL;
+	if (!(msg->flags & IORING_MSG_RING_FLAGS_PASS) && msg->dst_fd)
 		return -EINVAL;
 	if (target_ctx->flags & IORING_SETUP_R_DISABLED)
 		return -EBADFD;
@@ -124,15 +140,18 @@ static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags)
 	if (io_msg_need_remote(target_ctx))
 		return io_msg_exec_remote(req, io_msg_tw_complete);
 
+	if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
+		flags = msg->cqe_flags;
+
 	ret = -EOVERFLOW;
 	if (target_ctx->flags & IORING_SETUP_IOPOLL) {
 		if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
 			return -EAGAIN;
-		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
+		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
 			ret = 0;
 		io_double_unlock_ctx(target_ctx);
 	} else {
-		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
+		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
 			ret = 0;
 	}
 	return ret;
@@ -241,7 +260,7 @@ int io_msg_ring_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	msg->src_fd = READ_ONCE(sqe->addr3);
 	msg->dst_fd = READ_ONCE(sqe->file_index);
 	msg->flags = READ_ONCE(sqe->msg_ring_flags);
-	if (msg->flags & ~IORING_MSG_RING_CQE_SKIP)
+	if (msg->flags & ~IORING_MSG_RING_MASK)
 		return -EINVAL;
 
 	return 0;

io_uring/net.c

@@ -90,6 +90,7 @@ int io_shutdown_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 		return -EINVAL;
 
 	shutdown->how = READ_ONCE(sqe->len);
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -99,8 +100,7 @@ int io_shutdown(struct io_kiocb *req, unsigned int issue_flags)
 	struct socket *sock;
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	sock = sock_from_file(req->file);
 	if (unlikely(!sock))

io_uring/notif.c

@@ -68,9 +68,8 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
 	struct io_kiocb *notif;
 	struct io_notif_data *nd;
 
-	if (unlikely(!io_alloc_req_refill(ctx)))
+	if (unlikely(!io_alloc_req(ctx, &notif)))
 		return NULL;
-	notif = io_alloc_req(ctx);
 	notif->opcode = IORING_OP_NOP;
 	notif->flags = 0;
 	notif->file = NULL;

io_uring/opdef.c

@@ -46,11 +46,10 @@ static __maybe_unused int io_eopnotsupp_prep(struct io_kiocb *kiocb,
 	return -EOPNOTSUPP;
 }
 
-const struct io_op_def io_op_defs[] = {
+const struct io_issue_def io_issue_defs[] = {
 	[IORING_OP_NOP] = {
 		.audit_skip		= 1,
 		.iopoll			= 1,
-		.name			= "NOP",
 		.prep			= io_nop_prep,
 		.issue			= io_nop,
 	},
@@ -64,13 +63,8 @@ const struct io_op_def io_op_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
-		.async_size		= sizeof(struct io_async_rw),
-		.name			= "READV",
 		.prep			= io_prep_rw,
 		.issue			= io_read,
-		.prep_async		= io_readv_prep_async,
-		.cleanup		= io_readv_writev_cleanup,
-		.fail			= io_rw_fail,
 	},
 	[IORING_OP_WRITEV] = {
 		.needs_file		= 1,
@@ -82,18 +76,12 @@ const struct io_op_def io_op_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
-		.async_size		= sizeof(struct io_async_rw),
-		.name			= "WRITEV",
 		.prep			= io_prep_rw,
 		.issue			= io_write,
-		.prep_async		= io_writev_prep_async,
-		.cleanup		= io_readv_writev_cleanup,
-		.fail			= io_rw_fail,
 	},
 	[IORING_OP_FSYNC] = {
 		.needs_file		= 1,
 		.audit_skip		= 1,
-		.name			= "FSYNC",
 		.prep			= io_fsync_prep,
 		.issue			= io_fsync,
 	},
@@ -106,11 +94,8 @@ const struct io_op_def io_op_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
-		.async_size		= sizeof(struct io_async_rw),
-		.name			= "READ_FIXED",
 		.prep			= io_prep_rw,
 		.issue			= io_read,
-		.fail			= io_rw_fail,
 	},
 	[IORING_OP_WRITE_FIXED] = {
 		.needs_file		= 1,
@@ -122,30 +107,24 @@ const struct io_op_def io_op_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
-		.async_size		= sizeof(struct io_async_rw),
-		.name			= "WRITE_FIXED",
 		.prep			= io_prep_rw,
 		.issue			= io_write,
-		.fail			= io_rw_fail,
 	},
 	[IORING_OP_POLL_ADD] = {
 		.needs_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.audit_skip		= 1,
-		.name			= "POLL_ADD",
 		.prep			= io_poll_add_prep,
 		.issue			= io_poll_add,
 	},
 	[IORING_OP_POLL_REMOVE] = {
 		.audit_skip		= 1,
-		.name			= "POLL_REMOVE",
 		.prep			= io_poll_remove_prep,
 		.issue			= io_poll_remove,
 	},
 	[IORING_OP_SYNC_FILE_RANGE] = {
 		.needs_file		= 1,
 		.audit_skip		= 1,
-		.name			= "SYNC_FILE_RANGE",
 		.prep			= io_sfr_prep,
 		.issue			= io_sync_file_range,
 	},
@@ -155,14 +134,9 @@ const struct io_op_def io_op_defs[] = {
 		.pollout		= 1,
 		.ioprio			= 1,
 		.manual_alloc		= 1,
-		.name			= "SENDMSG",
 #if defined(CONFIG_NET)
-		.async_size		= sizeof(struct io_async_msghdr),
 		.prep			= io_sendmsg_prep,
 		.issue			= io_sendmsg,
-		.prep_async		= io_sendmsg_prep_async,
-		.cleanup		= io_sendmsg_recvmsg_cleanup,
-		.fail			= io_sendrecv_fail,
 #else
 		.prep			= io_eopnotsupp_prep,
 #endif
@@ -174,29 +148,21 @@ const struct io_op_def io_op_defs[] = {
 		.buffer_select		= 1,
 		.ioprio			= 1,
 		.manual_alloc		= 1,
-		.name			= "RECVMSG",
 #if defined(CONFIG_NET)
-		.async_size		= sizeof(struct io_async_msghdr),
 		.prep			= io_recvmsg_prep,
 		.issue			= io_recvmsg,
-		.prep_async		= io_recvmsg_prep_async,
-		.cleanup		= io_sendmsg_recvmsg_cleanup,
-		.fail			= io_sendrecv_fail,
 #else
 		.prep			= io_eopnotsupp_prep,
 #endif
 	},
 	[IORING_OP_TIMEOUT] = {
 		.audit_skip		= 1,
-		.async_size		= sizeof(struct io_timeout_data),
-		.name			= "TIMEOUT",
 		.prep			= io_timeout_prep,
 		.issue			= io_timeout,
 	},
 	[IORING_OP_TIMEOUT_REMOVE] = {
 		/* used by timeout updates' prep() */
 		.audit_skip		= 1,
-		.name			= "TIMEOUT_REMOVE",
 		.prep			= io_timeout_remove_prep,
 		.issue			= io_timeout_remove,
 	},
@@ -206,7 +172,6 @@ const struct io_op_def io_op_defs[] = {
 		.pollin			= 1,
 		.poll_exclusive		= 1,
 		.ioprio			= 1,	/* used for flags */
-		.name			= "ACCEPT",
 #if defined(CONFIG_NET)
 		.prep			= io_accept_prep,
 		.issue			= io_accept,
@@ -216,14 +181,11 @@ const struct io_op_def io_op_defs[] = {
 	},
 	[IORING_OP_ASYNC_CANCEL] = {
 		.audit_skip		= 1,
-		.name			= "ASYNC_CANCEL",
 		.prep			= io_async_cancel_prep,
 		.issue			= io_async_cancel,
 	},
 	[IORING_OP_LINK_TIMEOUT] = {
 		.audit_skip		= 1,
-		.async_size		= sizeof(struct io_timeout_data),
-		.name			= "LINK_TIMEOUT",
 		.prep			= io_link_timeout_prep,
 		.issue			= io_no_issue,
 	},
@@ -231,46 +193,36 @@ const struct io_op_def io_op_defs[] = {
 		.needs_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.pollout		= 1,
-		.name			= "CONNECT",
 #if defined(CONFIG_NET)
-		.async_size		= sizeof(struct io_async_connect),
 		.prep			= io_connect_prep,
 		.issue			= io_connect,
-		.prep_async		= io_connect_prep_async,
 #else
 		.prep			= io_eopnotsupp_prep,
 #endif
 	},
 	[IORING_OP_FALLOCATE] = {
 		.needs_file		= 1,
-		.name			= "FALLOCATE",
 		.prep			= io_fallocate_prep,
 		.issue			= io_fallocate,
 	},
 	[IORING_OP_OPENAT] = {
-		.name			= "OPENAT",
 		.prep			= io_openat_prep,
 		.issue			= io_openat,
-		.cleanup		= io_open_cleanup,
 	},
 	[IORING_OP_CLOSE] = {
-		.name			= "CLOSE",
 		.prep			= io_close_prep,
 		.issue			= io_close,
 	},
 	[IORING_OP_FILES_UPDATE] = {
 		.audit_skip		= 1,
 		.iopoll			= 1,
-		.name			= "FILES_UPDATE",
 		.prep			= io_files_update_prep,
 		.issue			= io_files_update,
 	},
 	[IORING_OP_STATX] = {
 		.audit_skip		= 1,
-		.name			= "STATX",
 		.prep			= io_statx_prep,
 		.issue			= io_statx,
-		.cleanup		= io_statx_cleanup,
 	},
 	[IORING_OP_READ] = {
 		.needs_file		= 1,
@@ -282,11 +234,8 @@ const struct io_op_def io_op_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
-		.async_size		= sizeof(struct io_async_rw),
-		.name			= "READ",
 		.prep			= io_prep_rw,
 		.issue			= io_read,
-		.fail			= io_rw_fail,
 	},
 	[IORING_OP_WRITE] = {
 		.needs_file		= 1,
@@ -298,21 +247,17 @@ const struct io_op_def io_op_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
-		.async_size		= sizeof(struct io_async_rw),
-		.name			= "WRITE",
 		.prep			= io_prep_rw,
 		.issue			= io_write,
-		.fail			= io_rw_fail,
 	},
 	[IORING_OP_FADVISE] = {
 		.needs_file		= 1,
 		.audit_skip		= 1,
-		.name			= "FADVISE",
 		.prep			= io_fadvise_prep,
 		.issue			= io_fadvise,
 	},
 	[IORING_OP_MADVISE] = {
-		.name			= "MADVISE",
+		.audit_skip		= 1,
 		.prep			= io_madvise_prep,
 		.issue			= io_madvise,
 	},
@@ -323,13 +268,9 @@ const struct io_op_def io_op_defs[] = {
 		.audit_skip		= 1,
 		.ioprio			= 1,
 		.manual_alloc		= 1,
-		.name			= "SEND",
 #if defined(CONFIG_NET)
-		.async_size		= sizeof(struct io_async_msghdr),
 		.prep			= io_sendmsg_prep,
 		.issue			= io_send,
-		.fail			= io_sendrecv_fail,
-		.prep_async		= io_send_prep_async,
 #else
 		.prep			= io_eopnotsupp_prep,
 #endif
@@ -341,25 +282,20 @@ const struct io_op_def io_op_defs[] = {
 		.buffer_select		= 1,
 		.audit_skip		= 1,
 		.ioprio			= 1,
-		.name			= "RECV",
 #if defined(CONFIG_NET)
 		.prep			= io_recvmsg_prep,
 		.issue			= io_recv,
-		.fail			= io_sendrecv_fail,
 #else
 		.prep			= io_eopnotsupp_prep,
 #endif
 	},
 	[IORING_OP_OPENAT2] = {
-		.name			= "OPENAT2",
 		.prep			= io_openat2_prep,
 		.issue			= io_openat2,
-		.cleanup		= io_open_cleanup,
 	},
 	[IORING_OP_EPOLL_CTL] = {
 		.unbound_nonreg_file	= 1,
 		.audit_skip		= 1,
-		.name			= "EPOLL",
 #if defined(CONFIG_EPOLL)
 		.prep			= io_epoll_ctl_prep,
 		.issue			= io_epoll_ctl,
@@ -372,21 +308,18 @@ const struct io_op_def io_op_defs[] = {
 		.hash_reg_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.audit_skip		= 1,
-		.name			= "SPLICE",
 		.prep			= io_splice_prep,
 		.issue			= io_splice,
 	},
 	[IORING_OP_PROVIDE_BUFFERS] = {
 		.audit_skip		= 1,
 		.iopoll			= 1,
-		.name			= "PROVIDE_BUFFERS",
 		.prep			= io_provide_buffers_prep,
 		.issue			= io_provide_buffers,
 	},
 	[IORING_OP_REMOVE_BUFFERS] = {
 		.audit_skip		= 1,
 		.iopoll			= 1,
-		.name			= "REMOVE_BUFFERS",
 		.prep			= io_remove_buffers_prep,
 		.issue			= io_remove_buffers,
 	},
@@ -395,13 +328,11 @@ const struct io_op_def io_op_defs[] = {
 		.hash_reg_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.audit_skip		= 1,
-		.name			= "TEE",
 		.prep			= io_tee_prep,
 		.issue			= io_tee,
 	},
 	[IORING_OP_SHUTDOWN] = {
 		.needs_file		= 1,
-		.name			= "SHUTDOWN",
 #if defined(CONFIG_NET)
 		.prep			= io_shutdown_prep,
 		.issue			= io_shutdown,
@@ -410,72 +341,51 @@ const struct io_op_def io_op_defs[] = {
 #endif
 	},
 	[IORING_OP_RENAMEAT] = {
-		.name			= "RENAMEAT",
 		.prep			= io_renameat_prep,
 		.issue			= io_renameat,
-		.cleanup		= io_renameat_cleanup,
 	},
 	[IORING_OP_UNLINKAT] = {
-		.name			= "UNLINKAT",
 		.prep			= io_unlinkat_prep,
 		.issue			= io_unlinkat,
-		.cleanup		= io_unlinkat_cleanup,
 	},
 	[IORING_OP_MKDIRAT] = {
-		.name			= "MKDIRAT",
 		.prep			= io_mkdirat_prep,
 		.issue			= io_mkdirat,
-		.cleanup		= io_mkdirat_cleanup,
 	},
 	[IORING_OP_SYMLINKAT] = {
-		.name			= "SYMLINKAT",
 		.prep			= io_symlinkat_prep,
 		.issue			= io_symlinkat,
-		.cleanup		= io_link_cleanup,
 	},
 	[IORING_OP_LINKAT] = {
-		.name			= "LINKAT",
 		.prep			= io_linkat_prep,
 		.issue			= io_linkat,
-		.cleanup		= io_link_cleanup,
 	},
 	[IORING_OP_MSG_RING] = {
 		.needs_file		= 1,
 		.iopoll			= 1,
-		.name			= "MSG_RING",
 		.prep			= io_msg_ring_prep,
 		.issue			= io_msg_ring,
-		.cleanup		= io_msg_ring_cleanup,
 	},
 	[IORING_OP_FSETXATTR] = {
 		.needs_file = 1,
-		.name			= "FSETXATTR",
 		.prep			= io_fsetxattr_prep,
 		.issue			= io_fsetxattr,
-		.cleanup		= io_xattr_cleanup,
 	},
 	[IORING_OP_SETXATTR] = {
-		.name			= "SETXATTR",
 		.prep			= io_setxattr_prep,
 		.issue			= io_setxattr,
-		.cleanup		= io_xattr_cleanup,
 	},
 	[IORING_OP_FGETXATTR] = {
 		.needs_file = 1,
-		.name			= "FGETXATTR",
 		.prep			= io_fgetxattr_prep,
 		.issue			= io_fgetxattr,
-		.cleanup		= io_xattr_cleanup,
 	},
 	[IORING_OP_GETXATTR] = {
-		.name			= "GETXATTR",
 		.prep			= io_getxattr_prep,
 		.issue			= io_getxattr,
-		.cleanup		= io_xattr_cleanup,
 	},
 	[IORING_OP_SOCKET] = {
 		.audit_skip		= 1,
-		.name			= "SOCKET",
 #if defined(CONFIG_NET)
 		.prep			= io_socket_prep,
 		.issue			= io_socket,
@@ -486,16 +396,12 @@ const struct io_op_def io_op_defs[] = {
 	[IORING_OP_URING_CMD] = {
 		.needs_file		= 1,
 		.plug			= 1,
-		.name			= "URING_CMD",
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
-		.async_size		= uring_cmd_pdu_size(1),
 		.prep			= io_uring_cmd_prep,
 		.issue			= io_uring_cmd,
-		.prep_async		= io_uring_cmd_prep_async,
 	},
 	[IORING_OP_SEND_ZC] = {
-		.name			= "SEND_ZC",
 		.needs_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.pollout		= 1,
@@ -503,32 +409,243 @@ const struct io_op_def io_op_defs[] = {
 		.ioprio			= 1,
 		.manual_alloc		= 1,
 #if defined(CONFIG_NET)
-		.async_size		= sizeof(struct io_async_msghdr),
 		.prep			= io_send_zc_prep,
 		.issue			= io_send_zc,
-		.prep_async		= io_send_prep_async,
-		.cleanup		= io_send_zc_cleanup,
-		.fail			= io_sendrecv_fail,
 #else
 		.prep			= io_eopnotsupp_prep,
 #endif
 	},
 	[IORING_OP_SENDMSG_ZC] = {
-		.name			= "SENDMSG_ZC",
 		.needs_file		= 1,
 		.unbound_nonreg_file	= 1,
 		.pollout		= 1,
 		.ioprio			= 1,
 		.manual_alloc		= 1,
 #if defined(CONFIG_NET)
-		.async_size		= sizeof(struct io_async_msghdr),
 		.prep			= io_send_zc_prep,
 		.issue			= io_sendmsg_zc,
+#else
+		.prep			= io_eopnotsupp_prep,
+#endif
+	},
+};
+
+
+const struct io_cold_def io_cold_defs[] = {
+	[IORING_OP_NOP] = {
+		.name			= "NOP",
+	},
+	[IORING_OP_READV] = {
+		.async_size		= sizeof(struct io_async_rw),
+		.name			= "READV",
+		.prep_async		= io_readv_prep_async,
+		.cleanup		= io_readv_writev_cleanup,
+		.fail			= io_rw_fail,
+	},
+	[IORING_OP_WRITEV] = {
+		.async_size		= sizeof(struct io_async_rw),
+		.name			= "WRITEV",
+		.prep_async		= io_writev_prep_async,
+		.cleanup		= io_readv_writev_cleanup,
+		.fail			= io_rw_fail,
+	},
+	[IORING_OP_FSYNC] = {
+		.name			= "FSYNC",
+	},
+	[IORING_OP_READ_FIXED] = {
+		.async_size		= sizeof(struct io_async_rw),
+		.name			= "READ_FIXED",
+		.fail			= io_rw_fail,
+	},
+	[IORING_OP_WRITE_FIXED] = {
+		.async_size		= sizeof(struct io_async_rw),
+		.name			= "WRITE_FIXED",
+		.fail			= io_rw_fail,
+	},
+	[IORING_OP_POLL_ADD] = {
+		.name			= "POLL_ADD",
+	},
+	[IORING_OP_POLL_REMOVE] = {
+		.name			= "POLL_REMOVE",
+	},
+	[IORING_OP_SYNC_FILE_RANGE] = {
+		.name			= "SYNC_FILE_RANGE",
+	},
+	[IORING_OP_SENDMSG] = {
+		.name			= "SENDMSG",
+#if defined(CONFIG_NET)
+		.async_size		= sizeof(struct io_async_msghdr),
+		.prep_async		= io_sendmsg_prep_async,
+		.cleanup		= io_sendmsg_recvmsg_cleanup,
+		.fail			= io_sendrecv_fail,
+#endif
+	},
+	[IORING_OP_RECVMSG] = {
+		.name			= "RECVMSG",
+#if defined(CONFIG_NET)
+		.async_size		= sizeof(struct io_async_msghdr),
+		.prep_async		= io_recvmsg_prep_async,
+		.cleanup		= io_sendmsg_recvmsg_cleanup,
+		.fail			= io_sendrecv_fail,
+#endif
+	},
+	[IORING_OP_TIMEOUT] = {
+		.async_size		= sizeof(struct io_timeout_data),
+		.name			= "TIMEOUT",
+	},
+	[IORING_OP_TIMEOUT_REMOVE] = {
+		.name			= "TIMEOUT_REMOVE",
+	},
+	[IORING_OP_ACCEPT] = {
+		.name			= "ACCEPT",
+	},
+	[IORING_OP_ASYNC_CANCEL] = {
+		.name			= "ASYNC_CANCEL",
+	},
+	[IORING_OP_LINK_TIMEOUT] = {
+		.async_size		= sizeof(struct io_timeout_data),
+		.name			= "LINK_TIMEOUT",
+	},
+	[IORING_OP_CONNECT] = {
+		.name			= "CONNECT",
+#if defined(CONFIG_NET)
+		.async_size		= sizeof(struct io_async_connect),
+		.prep_async		= io_connect_prep_async,
+#endif
+	},
+	[IORING_OP_FALLOCATE] = {
+		.name			= "FALLOCATE",
+	},
+	[IORING_OP_OPENAT] = {
+		.name			= "OPENAT",
+		.cleanup		= io_open_cleanup,
+	},
+	[IORING_OP_CLOSE] = {
+		.name			= "CLOSE",
+	},
+	[IORING_OP_FILES_UPDATE] = {
+		.name			= "FILES_UPDATE",
+	},
+	[IORING_OP_STATX] = {
+		.name			= "STATX",
+		.cleanup		= io_statx_cleanup,
+	},
+	[IORING_OP_READ] = {
+		.async_size		= sizeof(struct io_async_rw),
+		.name			= "READ",
+		.fail			= io_rw_fail,
+	},
+	[IORING_OP_WRITE] = {
+		.async_size		= sizeof(struct io_async_rw),
+		.name			= "WRITE",
+		.fail			= io_rw_fail,
+	},
+	[IORING_OP_FADVISE] = {
+		.name			= "FADVISE",
+	},
+	[IORING_OP_MADVISE] = {
+		.name			= "MADVISE",
+	},
+	[IORING_OP_SEND] = {
+		.name			= "SEND",
+#if defined(CONFIG_NET)
+		.async_size		= sizeof(struct io_async_msghdr),
+		.fail			= io_sendrecv_fail,
+		.prep_async		= io_send_prep_async,
+#endif
+	},
+	[IORING_OP_RECV] = {
+		.name			= "RECV",
+#if defined(CONFIG_NET)
+		.fail			= io_sendrecv_fail,
+#endif
+	},
+	[IORING_OP_OPENAT2] = {
+		.name			= "OPENAT2",
+		.cleanup		= io_open_cleanup,
+	},
+	[IORING_OP_EPOLL_CTL] = {
+		.name			= "EPOLL",
+	},
+	[IORING_OP_SPLICE] = {
+		.name			= "SPLICE",
+	},
+	[IORING_OP_PROVIDE_BUFFERS] = {
+		.name			= "PROVIDE_BUFFERS",
+	},
+	[IORING_OP_REMOVE_BUFFERS] = {
+		.name			= "REMOVE_BUFFERS",
+	},
+	[IORING_OP_TEE] = {
+		.name			= "TEE",
+	},
+	[IORING_OP_SHUTDOWN] = {
+		.name			= "SHUTDOWN",
+	},
+	[IORING_OP_RENAMEAT] = {
+		.name			= "RENAMEAT",
+		.cleanup		= io_renameat_cleanup,
+	},
+	[IORING_OP_UNLINKAT] = {
+		.name			= "UNLINKAT",
+		.cleanup		= io_unlinkat_cleanup,
+	},
+	[IORING_OP_MKDIRAT] = {
+		.name			= "MKDIRAT",
+		.cleanup		= io_mkdirat_cleanup,
+	},
+	[IORING_OP_SYMLINKAT] = {
+		.name			= "SYMLINKAT",
+		.cleanup		= io_link_cleanup,
+	},
+	[IORING_OP_LINKAT] = {
+		.name			= "LINKAT",
+		.cleanup		= io_link_cleanup,
+	},
+	[IORING_OP_MSG_RING] = {
+		.name			= "MSG_RING",
+		.cleanup		= io_msg_ring_cleanup,
+	},
+	[IORING_OP_FSETXATTR] = {
+		.name			= "FSETXATTR",
+		.cleanup		= io_xattr_cleanup,
+	},
+	[IORING_OP_SETXATTR] = {
+		.name			= "SETXATTR",
+		.cleanup		= io_xattr_cleanup,
+	},
+	[IORING_OP_FGETXATTR] = {
+		.name			= "FGETXATTR",
+		.cleanup		= io_xattr_cleanup,
+	},
+	[IORING_OP_GETXATTR] = {
+		.name			= "GETXATTR",
+		.cleanup		= io_xattr_cleanup,
+	},
+	[IORING_OP_SOCKET] = {
+		.name			= "SOCKET",
+	},
+	[IORING_OP_URING_CMD] = {
+		.name			= "URING_CMD",
+		.async_size		= uring_cmd_pdu_size(1),
+		.prep_async		= io_uring_cmd_prep_async,
+	},
+	[IORING_OP_SEND_ZC] = {
+		.name			= "SEND_ZC",
+#if defined(CONFIG_NET)
+		.async_size		= sizeof(struct io_async_msghdr),
+		.prep_async		= io_send_prep_async,
+		.cleanup		= io_send_zc_cleanup,
+		.fail			= io_sendrecv_fail,
+#endif
+	},
+	[IORING_OP_SENDMSG_ZC] = {
+		.name			= "SENDMSG_ZC",
+#if defined(CONFIG_NET)
+		.async_size		= sizeof(struct io_async_msghdr),
 		.prep_async		= io_sendmsg_prep_async,
 		.cleanup		= io_send_zc_cleanup,
 		.fail			= io_sendrecv_fail,
-#else
-		.prep			= io_eopnotsupp_prep,
 #endif
 	},
 };
@@ -536,7 +653,7 @@ const struct io_op_def io_op_defs[] = {
 const char *io_uring_get_opcode(u8 opcode)
 {
 	if (opcode < IORING_OP_LAST)
-		return io_op_defs[opcode].name;
+		return io_cold_defs[opcode].name;
 	return "INVALID";
 }
 
@@ -544,12 +661,13 @@ void __init io_uring_optable_init(void)
 {
 	int i;
 
-	BUILD_BUG_ON(ARRAY_SIZE(io_op_defs) != IORING_OP_LAST);
+	BUILD_BUG_ON(ARRAY_SIZE(io_cold_defs) != IORING_OP_LAST);
+	BUILD_BUG_ON(ARRAY_SIZE(io_issue_defs) != IORING_OP_LAST);
 
-	for (i = 0; i < ARRAY_SIZE(io_op_defs); i++) {
-		BUG_ON(!io_op_defs[i].prep);
-		if (io_op_defs[i].prep != io_eopnotsupp_prep)
-			BUG_ON(!io_op_defs[i].issue);
-		WARN_ON_ONCE(!io_op_defs[i].name);
+	for (i = 0; i < ARRAY_SIZE(io_issue_defs); i++) {
+		BUG_ON(!io_issue_defs[i].prep);
+		if (io_issue_defs[i].prep != io_eopnotsupp_prep)
+			BUG_ON(!io_issue_defs[i].issue);
+		WARN_ON_ONCE(!io_cold_defs[i].name);
 	}
 }

io_uring/opdef.h

@@ -2,7 +2,7 @@
 #ifndef IOU_OP_DEF_H
 #define IOU_OP_DEF_H
 
-struct io_op_def {
+struct io_issue_def {
 	/* needs req->file assigned */
 	unsigned		needs_file : 1;
 	/* should block plug */
@@ -29,19 +29,24 @@ struct io_op_def {
 	unsigned		iopoll_queue : 1;
 	/* opcode specific path will handle ->async_data allocation if needed */
 	unsigned		manual_alloc : 1;
+
+	int (*issue)(struct io_kiocb *, unsigned int);
+	int (*prep)(struct io_kiocb *, const struct io_uring_sqe *);
+};
+
+struct io_cold_def {
 	/* size of async data needed, if any */
 	unsigned short		async_size;
 
 	const char		*name;
 
-	int (*prep)(struct io_kiocb *, const struct io_uring_sqe *);
-	int (*issue)(struct io_kiocb *, unsigned int);
 	int (*prep_async)(struct io_kiocb *);
 	void (*cleanup)(struct io_kiocb *);
 	void (*fail)(struct io_kiocb *);
 };
 
-extern const struct io_op_def io_op_defs[];
+extern const struct io_issue_def io_issue_defs[];
+extern const struct io_cold_def io_cold_defs[];
 
 void io_uring_optable_init(void);
 #endif

io_uring/openclose.c

@@ -31,6 +31,15 @@ struct io_close {
 	u32				file_slot;
 };
 
+static bool io_openat_force_async(struct io_open *open)
+{
+	/*
+	 * Don't bother trying for O_TRUNC, O_CREAT, or O_TMPFILE open,
+	 * it'll always -EAGAIN
+	 */
+	return open->how.flags & (O_TRUNC | O_CREAT | O_TMPFILE);
+}
+
 static int __io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 {
 	struct io_open *open = io_kiocb_to_cmd(req, struct io_open);
@@ -61,6 +70,8 @@ static int __io_openat_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe
 
 	open->nofile = rlimit(RLIMIT_NOFILE);
 	req->flags |= REQ_F_NEED_CLEANUP;
+	if (io_openat_force_async(open))
+		req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -108,12 +119,7 @@ int io_openat2(struct io_kiocb *req, unsigned int issue_flags)
 	nonblock_set = op.open_flag & O_NONBLOCK;
 	resolve_nonblock = open->how.resolve & RESOLVE_CACHED;
 	if (issue_flags & IO_URING_F_NONBLOCK) {
-		/*
-		 * Don't bother trying for O_TRUNC, O_CREAT, or O_TMPFILE open,
-		 * it'll always -EAGAIN
-		 */
-		if (open->how.flags & (O_TRUNC | O_CREAT | O_TMPFILE))
-			return -EAGAIN;
+		WARN_ON_ONCE(io_openat_force_async(open));
 		op.lookup_flags |= LOOKUP_CACHED;
 		op.open_flag |= O_NONBLOCK;
 	}

io_uring/poll.c

@@ -678,7 +678,7 @@ static struct async_poll *io_req_alloc_apoll(struct io_kiocb *req,
 
 int io_arm_poll_handler(struct io_kiocb *req, unsigned issue_flags)
 {
-	const struct io_op_def *def = &io_op_defs[req->opcode];
+	const struct io_issue_def *def = &io_issue_defs[req->opcode];
 	struct async_poll *apoll;
 	struct io_poll_table ipt;
 	__poll_t mask = POLLPRI | POLLERR | EPOLLET;

io_uring/rw.c

@@ -410,7 +410,7 @@ static inline int io_import_iovec(int rw, struct io_kiocb *req,
 				  unsigned int issue_flags)
 {
 	*iovec = __io_import_iovec(rw, req, s, issue_flags);
-	if (unlikely(IS_ERR(*iovec)))
+	if (IS_ERR(*iovec))
 		return PTR_ERR(*iovec);
 
 	iov_iter_save_state(&s->iter, &s->iter_state);
@@ -516,7 +516,7 @@ static void io_req_map_rw(struct io_kiocb *req, const struct iovec *iovec,
 static int io_setup_async_rw(struct io_kiocb *req, const struct iovec *iovec,
 			     struct io_rw_state *s, bool force)
 {
-	if (!force && !io_op_defs[req->opcode].prep_async)
+	if (!force && !io_cold_defs[req->opcode].prep_async)
 		return 0;
 	if (!req_has_async_data(req)) {
 		struct io_async_rw *iorw;

io_uring/splice.c

@@ -34,6 +34,7 @@ static int __io_splice_prep(struct io_kiocb *req,
 	if (unlikely(sp->flags & ~valid_flags))
 		return -EINVAL;
 	sp->splice_fd_in = READ_ONCE(sqe->splice_fd_in);
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -52,8 +53,7 @@ int io_tee(struct io_kiocb *req, unsigned int issue_flags)
 	struct file *in;
 	long ret = 0;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	if (sp->flags & SPLICE_F_FD_IN_FIXED)
 		in = io_file_get_fixed(req, sp->splice_fd_in, issue_flags);
@@ -94,8 +94,7 @@ int io_splice(struct io_kiocb *req, unsigned int issue_flags)
 	struct file *in;
 	long ret = 0;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	if (sp->flags & SPLICE_F_FD_IN_FIXED)
 		in = io_file_get_fixed(req, sp->splice_fd_in, issue_flags);

io_uring/sqpoll.c

@@ -312,7 +312,7 @@ static int io_sq_thread(void *data)
 	do_exit(0);
 }
 
-int io_sqpoll_wait_sq(struct io_ring_ctx *ctx)
+void io_sqpoll_wait_sq(struct io_ring_ctx *ctx)
 {
 	DEFINE_WAIT(wait);
 
@@ -327,7 +327,6 @@ int io_sqpoll_wait_sq(struct io_ring_ctx *ctx)
 	} while (!signal_pending(current));
 
 	finish_wait(&ctx->sqo_sq_wait, &wait);
-	return 0;
 }
 
 __cold int io_sq_offload_create(struct io_ring_ctx *ctx,

io_uring/sqpoll.h

@@ -26,4 +26,4 @@ void io_sq_thread_stop(struct io_sq_data *sqd);
 void io_sq_thread_park(struct io_sq_data *sqd);
 void io_sq_thread_unpark(struct io_sq_data *sqd);
 void io_put_sq_data(struct io_sq_data *sqd);
-int io_sqpoll_wait_sq(struct io_ring_ctx *ctx);
+void io_sqpoll_wait_sq(struct io_ring_ctx *ctx);

io_uring/statx.c

@@ -48,6 +48,7 @@ int io_statx_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	}
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -56,8 +57,7 @@ int io_statx(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_statx *sx = io_kiocb_to_cmd(req, struct io_statx);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = do_statx(sx->dfd, sx->filename, sx->flags, sx->mask, sx->buffer);
 	io_req_set_res(req, ret, 0);

io_uring/sync.c

@@ -32,6 +32,8 @@ int io_sfr_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	sync->off = READ_ONCE(sqe->off);
 	sync->len = READ_ONCE(sqe->len);
 	sync->flags = READ_ONCE(sqe->sync_range_flags);
+	req->flags |= REQ_F_FORCE_ASYNC;
+
 	return 0;
 }
 
@@ -41,8 +43,7 @@ int io_sync_file_range(struct io_kiocb *req, unsigned int issue_flags)
 	int ret;
 
 	/* sync_file_range always requires a blocking context */
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = sync_file_range(req->file, sync->off, sync->len, sync->flags);
 	io_req_set_res(req, ret, 0);
@@ -62,6 +63,7 @@ int io_fsync_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 
 	sync->off = READ_ONCE(sqe->off);
 	sync->len = READ_ONCE(sqe->len);
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -72,8 +74,7 @@ int io_fsync(struct io_kiocb *req, unsigned int issue_flags)
 	int ret;
 
 	/* fsync always requires a blocking context */
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = vfs_fsync_range(req->file, sync->off, end > 0 ? end : LLONG_MAX,
 				sync->flags & IORING_FSYNC_DATASYNC);
@@ -91,6 +92,7 @@ int io_fallocate_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	sync->off = READ_ONCE(sqe->off);
 	sync->len = READ_ONCE(sqe->addr);
 	sync->mode = READ_ONCE(sqe->len);
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -100,8 +102,8 @@ int io_fallocate(struct io_kiocb *req, unsigned int issue_flags)
 	int ret;
 
 	/* fallocate always requiring blocking context */
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
+
 	ret = vfs_fallocate(req->file, sync->mode, sync->off, sync->len);
 	if (ret >= 0)
 		fsnotify_modify(req->file);

io_uring/xattr.c

@@ -75,6 +75,7 @@ static int __io_getxattr_prep(struct io_kiocb *req,
 	}
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -109,8 +110,7 @@ int io_fgetxattr(struct io_kiocb *req, unsigned int issue_flags)
 	struct io_xattr *ix = io_kiocb_to_cmd(req, struct io_xattr);
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = do_getxattr(mnt_idmap(req->file->f_path.mnt),
 			req->file->f_path.dentry,
@@ -127,8 +127,7 @@ int io_getxattr(struct io_kiocb *req, unsigned int issue_flags)
 	struct path path;
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 retry:
 	ret = filename_lookup(AT_FDCWD, ix->filename, lookup_flags, &path, NULL);
@@ -174,6 +173,7 @@ static int __io_setxattr_prep(struct io_kiocb *req,
 	}
 
 	req->flags |= REQ_F_NEED_CLEANUP;
+	req->flags |= REQ_F_FORCE_ASYNC;
 	return 0;
 }
 
@@ -222,8 +222,7 @@ int io_fsetxattr(struct io_kiocb *req, unsigned int issue_flags)
 {
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 	ret = __io_setxattr(req, issue_flags, &req->file->f_path);
 	io_xattr_finish(req, ret);
@@ -237,8 +236,7 @@ int io_setxattr(struct io_kiocb *req, unsigned int issue_flags)
 	struct path path;
 	int ret;
 
-	if (issue_flags & IO_URING_F_NONBLOCK)
-		return -EAGAIN;
+	WARN_ON_ONCE(issue_flags & IO_URING_F_NONBLOCK);
 
 retry:
 	ret = filename_lookup(AT_FDCWD, ix->filename, lookup_flags, &path, NULL);