From 59915143e89fb8dc7b5bd9dcaf628d8181fd54ac Mon Sep 17 00:00:00 2001
From: Jens Axboe <axboe@kernel.dk>
Date: Wed, 25 May 2022 08:57:27 -0600
Subject: [PATCH] io_uring: move timeout opcodes and handling into its own file

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/Makefile         |   2 +-
 io_uring/io_uring.c       | 667 +-------------------------------------
 io_uring/io_uring.h       |  13 +-
 io_uring/io_uring_types.h |  10 +
 io_uring/timeout.c        | 634 ++++++++++++++++++++++++++++++++++++
 io_uring/timeout.h        |  35 ++
 6 files changed, 701 insertions(+), 660 deletions(-)
 create mode 100644 io_uring/timeout.c
 create mode 100644 io_uring/timeout.h

diff --git a/io_uring/Makefile b/io_uring/Makefile
index d7cf992c841a..6ae4e45a15db 100644
--- a/io_uring/Makefile
+++ b/io_uring/Makefile
@@ -5,5 +5,5 @@
 obj-$(CONFIG_IO_URING)		+= io_uring.o xattr.o nop.o fs.o splice.o \
 					sync.o advise.o filetable.o \
 					openclose.o uring_cmd.o epoll.o \
-					statx.o net.o msg_ring.o
+					statx.o net.o msg_ring.o timeout.o
 obj-$(CONFIG_IO_WQ)		+= io-wq.o
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index eea5282b1ca2..3fc59a22d54e 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -105,6 +105,7 @@
 #include "statx.h"
 #include "net.h"
 #include "msg_ring.h"
+#include "timeout.h"
 
 #define IORING_MAX_ENTRIES	32768
 #define IORING_MAX_CQ_ENTRIES	(2 * IORING_MAX_ENTRIES)
@@ -288,14 +289,6 @@ struct io_poll_update {
 	bool				update_user_data;
 };
 
-struct io_timeout_data {
-	struct io_kiocb			*req;
-	struct hrtimer			timer;
-	struct timespec64		ts;
-	enum hrtimer_mode		mode;
-	u32				flags;
-};
-
 struct io_cancel {
 	struct file			*file;
 	u64				addr;
@@ -303,27 +296,6 @@ struct io_cancel {
 	s32				fd;
 };
 
-struct io_timeout {
-	struct file			*file;
-	u32				off;
-	u32				target_seq;
-	struct list_head		list;
-	/* head of the link, used by linked timeouts only */
-	struct io_kiocb			*head;
-	/* for linked completions */
-	struct io_kiocb			*prev;
-};
-
-struct io_timeout_rem {
-	struct file			*file;
-	u64				addr;
-
-	/* timeout update */
-	struct timespec64		ts;
-	u32				flags;
-	bool				ltimeout;
-};
-
 struct io_rw {
 	/* NOTE: kiocb has the file as the first member, so don't do it here */
 	struct kiocb			kiocb;
@@ -388,16 +360,6 @@ struct io_defer_entry {
 	u32			seq;
 };
 
-struct io_cancel_data {
-	struct io_ring_ctx *ctx;
-	union {
-		u64 data;
-		struct file *file;
-	};
-	u32 flags;
-	int seq;
-};
-
 struct io_op_def {
 	/* needs req->file assigned */
 	unsigned		needs_file : 1;
@@ -436,7 +398,6 @@ static const struct io_op_def io_op_defs[];
 #define IO_DISARM_MASK (REQ_F_ARM_LTIMEOUT | REQ_F_LINK_TIMEOUT | REQ_F_FAIL)
 #define IO_REQ_LINK_FLAGS (REQ_F_LINK | REQ_F_HARDLINK)
 
-static bool io_disarm_next(struct io_kiocb *req);
 static void io_uring_del_tctx_node(unsigned long index);
 static void io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
 					 struct task_struct *task,
@@ -444,7 +405,6 @@ static void io_uring_try_cancel_requests(struct io_ring_ctx *ctx,
 static void io_uring_cancel_generic(bool cancel_all, struct io_sq_data *sqd);
 
 static void io_dismantle_req(struct io_kiocb *req);
-static void io_queue_linked_timeout(struct io_kiocb *req);
 static int __io_register_rsrc_update(struct io_ring_ctx *ctx, unsigned type,
 				     struct io_uring_rsrc_update2 *up,
 				     unsigned nr_args);
@@ -456,9 +416,7 @@ static void io_req_task_queue(struct io_kiocb *req);
 static void __io_submit_flush_completions(struct io_ring_ctx *ctx);
 static int io_req_prep_async(struct io_kiocb *req);
 
-static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer);
 static void io_eventfd_signal(struct io_ring_ctx *ctx);
-static void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
 
 static struct kmem_cache *req_cachep;
 
@@ -609,9 +567,6 @@ static inline void io_tw_lock(struct io_ring_ctx *ctx, bool *locked)
 	}
 }
 
-#define io_for_each_link(pos, head) \
-	for (pos = (head); pos; pos = pos->link)
-
 static inline void io_submit_flush_completions(struct io_ring_ctx *ctx)
 {
 	if (!wq_list_empty(&ctx->submit_state.compl_reqs))
@@ -803,24 +758,6 @@ static void io_kbuf_recycle(struct io_kiocb *req, unsigned issue_flags)
 	io_ring_submit_unlock(ctx, issue_flags);
 }
 
-static bool io_match_task(struct io_kiocb *head, struct task_struct *task,
-			  bool cancel_all)
-	__must_hold(&req->ctx->timeout_lock)
-{
-	struct io_kiocb *req;
-
-	if (task && head->task != task)
-		return false;
-	if (cancel_all)
-		return true;
-
-	io_for_each_link(req, head) {
-		if (req->flags & REQ_F_INFLIGHT)
-			return true;
-	}
-	return false;
-}
-
 static bool io_match_linked(struct io_kiocb *head)
 {
 	struct io_kiocb *req;
@@ -877,13 +814,6 @@ static __cold void io_ring_ctx_ref_free(struct percpu_ref *ref)
 	complete(&ctx->ref_comp);
 }
 
-static inline bool io_is_timeout_noseq(struct io_kiocb *req)
-{
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-
-	return !timeout->off;
-}
-
 static __cold void io_fallback_req_func(struct work_struct *work)
 {
 	struct io_ring_ctx *ctx = container_of(work, struct io_ring_ctx,
@@ -1120,24 +1050,6 @@ static void io_queue_iowq(struct io_kiocb *req, bool *dont_use)
 		io_queue_linked_timeout(link);
 }
 
-static void io_kill_timeout(struct io_kiocb *req, int status)
-	__must_hold(&req->ctx->completion_lock)
-	__must_hold(&req->ctx->timeout_lock)
-{
-	struct io_timeout_data *io = req->async_data;
-
-	if (hrtimer_try_to_cancel(&io->timer) != -1) {
-		struct io_timeout *timeout = io_kiocb_to_cmd(req);
-
-		if (status)
-			req_set_fail(req);
-		atomic_set(&req->ctx->cq_timeouts,
-			atomic_read(&req->ctx->cq_timeouts) + 1);
-		list_del_init(&timeout->list);
-		io_req_tw_post_queue(req, status, 0);
-	}
-}
-
 static __cold void io_queue_deferred(struct io_ring_ctx *ctx)
 {
 	while (!list_empty(&ctx->defer_list)) {
@@ -1152,38 +1064,6 @@ static __cold void io_queue_deferred(struct io_ring_ctx *ctx)
 	}
 }
 
-static __cold void io_flush_timeouts(struct io_ring_ctx *ctx)
-	__must_hold(&ctx->completion_lock)
-{
-	u32 seq = ctx->cached_cq_tail - atomic_read(&ctx->cq_timeouts);
-	struct io_timeout *timeout, *tmp;
-
-	spin_lock_irq(&ctx->timeout_lock);
-	list_for_each_entry_safe(timeout, tmp, &ctx->timeout_list, list) {
-		struct io_kiocb *req = cmd_to_io_kiocb(timeout);
-		u32 events_needed, events_got;
-
-		if (io_is_timeout_noseq(req))
-			break;
-
-		/*
-		 * Since seq can easily wrap around over time, subtract
-		 * the last seq at which timeouts were flushed before comparing.
-		 * Assuming not more than 2^31-1 events have happened since,
-		 * these subtractions won't have wrapped, so we can check if
-		 * target is in [last_seq, current_seq] by comparing the two.
-		 */
-		events_needed = timeout->target_seq - ctx->cq_last_tm_flush;
-		events_got = seq - ctx->cq_last_tm_flush;
-		if (events_got < events_needed)
-			break;
-
-		io_kill_timeout(req, 0);
-	}
-	ctx->cq_last_tm_flush = seq;
-	spin_unlock_irq(&ctx->timeout_lock);
-}
-
 static void __io_commit_cqring_flush(struct io_ring_ctx *ctx)
 {
 	if (ctx->off_timeout_used || ctx->drain_active) {
@@ -1585,14 +1465,14 @@ static void __io_req_complete_put(struct io_kiocb *req)
 	}
 }
 
-static void __io_req_complete_post(struct io_kiocb *req)
+void __io_req_complete_post(struct io_kiocb *req)
 {
 	if (!(req->flags & REQ_F_CQE_SKIP))
 		__io_fill_cqe_req(req->ctx, req);
 	__io_req_complete_put(req);
 }
 
-static void io_req_complete_post(struct io_kiocb *req)
+void io_req_complete_post(struct io_kiocb *req)
 {
 	struct io_ring_ctx *ctx = req->ctx;
 
@@ -1717,7 +1597,7 @@ static inline void io_dismantle_req(struct io_kiocb *req)
 		io_put_file(req->file);
 }
 
-static __cold void io_free_req(struct io_kiocb *req)
+__cold void io_free_req(struct io_kiocb *req)
 {
 	struct io_ring_ctx *ctx = req->ctx;
 
@@ -1731,96 +1611,6 @@ static __cold void io_free_req(struct io_kiocb *req)
 	spin_unlock(&ctx->completion_lock);
 }
 
-static inline void io_remove_next_linked(struct io_kiocb *req)
-{
-	struct io_kiocb *nxt = req->link;
-
-	req->link = nxt->link;
-	nxt->link = NULL;
-}
-
-static struct io_kiocb *io_disarm_linked_timeout(struct io_kiocb *req)
-	__must_hold(&req->ctx->completion_lock)
-	__must_hold(&req->ctx->timeout_lock)
-{
-	struct io_kiocb *link = req->link;
-
-	if (link && link->opcode == IORING_OP_LINK_TIMEOUT) {
-		struct io_timeout_data *io = link->async_data;
-		struct io_timeout *timeout = io_kiocb_to_cmd(link);
-
-		io_remove_next_linked(req);
-		timeout->head = NULL;
-		if (hrtimer_try_to_cancel(&io->timer) != -1) {
-			list_del(&timeout->list);
-			return link;
-		}
-	}
-	return NULL;
-}
-
-static void io_fail_links(struct io_kiocb *req)
-	__must_hold(&req->ctx->completion_lock)
-{
-	struct io_kiocb *nxt, *link = req->link;
-	bool ignore_cqes = req->flags & REQ_F_SKIP_LINK_CQES;
-
-	req->link = NULL;
-	while (link) {
-		long res = -ECANCELED;
-
-		if (link->flags & REQ_F_FAIL)
-			res = link->cqe.res;
-
-		nxt = link->link;
-		link->link = NULL;
-
-		trace_io_uring_fail_link(req->ctx, req, req->cqe.user_data,
-					req->opcode, link);
-
-		if (ignore_cqes)
-			link->flags |= REQ_F_CQE_SKIP;
-		else
-			link->flags &= ~REQ_F_CQE_SKIP;
-		io_req_set_res(link, res, 0);
-		__io_req_complete_post(link);
-		link = nxt;
-	}
-}
-
-static bool io_disarm_next(struct io_kiocb *req)
-	__must_hold(&req->ctx->completion_lock)
-{
-	struct io_kiocb *link = NULL;
-	bool posted = false;
-
-	if (req->flags & REQ_F_ARM_LTIMEOUT) {
-		link = req->link;
-		req->flags &= ~REQ_F_ARM_LTIMEOUT;
-		if (link && link->opcode == IORING_OP_LINK_TIMEOUT) {
-			io_remove_next_linked(req);
-			io_req_tw_post_queue(link, -ECANCELED, 0);
-			posted = true;
-		}
-	} else if (req->flags & REQ_F_LINK_TIMEOUT) {
-		struct io_ring_ctx *ctx = req->ctx;
-
-		spin_lock_irq(&ctx->timeout_lock);
-		link = io_disarm_linked_timeout(req);
-		spin_unlock_irq(&ctx->timeout_lock);
-		if (link) {
-			posted = true;
-			io_req_tw_post_queue(link, -ECANCELED, 0);
-		}
-	}
-	if (unlikely((req->flags & REQ_F_FAIL) &&
-		     !(req->flags & REQ_F_HARDLINK))) {
-		posted |= (req->link != NULL);
-		io_fail_links(req);
-	}
-	return posted;
-}
-
 static void __io_req_find_next_prep(struct io_kiocb *req)
 {
 	struct io_ring_ctx *ctx = req->ctx;
@@ -2033,7 +1823,7 @@ static void io_req_tw_post(struct io_kiocb *req, bool *locked)
 	io_req_complete_post(req);
 }
 
-static void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags)
+void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags)
 {
 	io_req_set_res(req, res, cflags);
 	req->io_task_work.func = io_req_tw_post;
@@ -2057,7 +1847,7 @@ static void io_req_task_submit(struct io_kiocb *req, bool *locked)
 		io_req_complete_failed(req, -EFAULT);
 }
 
-static void io_req_task_queue_fail(struct io_kiocb *req, int ret)
+void io_req_task_queue_fail(struct io_kiocb *req, int ret)
 {
 	io_req_set_res(req, ret, 0);
 	req->io_task_work.func = io_req_task_cancel;
@@ -2076,7 +1866,7 @@ static void io_req_task_queue_reissue(struct io_kiocb *req)
 	io_req_task_work_add(req);
 }
 
-static void io_queue_next(struct io_kiocb *req)
+void io_queue_next(struct io_kiocb *req)
 {
 	struct io_kiocb *nxt = io_req_find_next(req);
 
@@ -2177,14 +1967,6 @@ static inline struct io_kiocb *io_put_req_find_next(struct io_kiocb *req)
 	return nxt;
 }
 
-static inline void io_put_req(struct io_kiocb *req)
-{
-	if (req_ref_put_and_test(req)) {
-		io_queue_next(req);
-		io_free_req(req);
-	}
-}
-
 static unsigned io_cqring_events(struct io_ring_ctx *ctx)
 {
 	/* See comment at the top of this file */
@@ -2451,7 +2233,7 @@ static bool __io_complete_rw_common(struct io_kiocb *req, long res)
 	return false;
 }
 
-static inline void io_req_task_complete(struct io_kiocb *req, bool *locked)
+inline void io_req_task_complete(struct io_kiocb *req, bool *locked)
 {
 	if (*locked) {
 		req->cqe.flags |= io_put_kbuf(req, 0);
@@ -4600,334 +4382,6 @@ static int io_poll_remove(struct io_kiocb *req, unsigned int issue_flags)
 	return IOU_OK;
 }
 
-static enum hrtimer_restart io_timeout_fn(struct hrtimer *timer)
-{
-	struct io_timeout_data *data = container_of(timer,
-						struct io_timeout_data, timer);
-	struct io_kiocb *req = data->req;
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-	struct io_ring_ctx *ctx = req->ctx;
-	unsigned long flags;
-
-	spin_lock_irqsave(&ctx->timeout_lock, flags);
-	list_del_init(&timeout->list);
-	atomic_set(&req->ctx->cq_timeouts,
-		atomic_read(&req->ctx->cq_timeouts) + 1);
-	spin_unlock_irqrestore(&ctx->timeout_lock, flags);
-
-	if (!(data->flags & IORING_TIMEOUT_ETIME_SUCCESS))
-		req_set_fail(req);
-
-	io_req_set_res(req, -ETIME, 0);
-	req->io_task_work.func = io_req_task_complete;
-	io_req_task_work_add(req);
-	return HRTIMER_NORESTART;
-}
-
-static struct io_kiocb *io_timeout_extract(struct io_ring_ctx *ctx,
-					   struct io_cancel_data *cd)
-	__must_hold(&ctx->timeout_lock)
-{
-	struct io_timeout *timeout;
-	struct io_timeout_data *io;
-	struct io_kiocb *req = NULL;
-
-	list_for_each_entry(timeout, &ctx->timeout_list, list) {
-		struct io_kiocb *tmp = cmd_to_io_kiocb(timeout);
-
-		if (!(cd->flags & IORING_ASYNC_CANCEL_ANY) &&
-		    cd->data != tmp->cqe.user_data)
-			continue;
-		if (cd->flags & (IORING_ASYNC_CANCEL_ALL|IORING_ASYNC_CANCEL_ANY)) {
-			if (cd->seq == tmp->work.cancel_seq)
-				continue;
-			tmp->work.cancel_seq = cd->seq;
-		}
-		req = tmp;
-		break;
-	}
-	if (!req)
-		return ERR_PTR(-ENOENT);
-
-	io = req->async_data;
-	if (hrtimer_try_to_cancel(&io->timer) == -1)
-		return ERR_PTR(-EALREADY);
-	timeout = io_kiocb_to_cmd(req);
-	list_del_init(&timeout->list);
-	return req;
-}
-
-static int io_timeout_cancel(struct io_ring_ctx *ctx, struct io_cancel_data *cd)
-	__must_hold(&ctx->completion_lock)
-{
-	struct io_kiocb *req;
-
-	spin_lock_irq(&ctx->timeout_lock);
-	req = io_timeout_extract(ctx, cd);
-	spin_unlock_irq(&ctx->timeout_lock);
-
-	if (IS_ERR(req))
-		return PTR_ERR(req);
-	io_req_task_queue_fail(req, -ECANCELED);
-	return 0;
-}
-
-static clockid_t io_timeout_get_clock(struct io_timeout_data *data)
-{
-	switch (data->flags & IORING_TIMEOUT_CLOCK_MASK) {
-	case IORING_TIMEOUT_BOOTTIME:
-		return CLOCK_BOOTTIME;
-	case IORING_TIMEOUT_REALTIME:
-		return CLOCK_REALTIME;
-	default:
-		/* can't happen, vetted at prep time */
-		WARN_ON_ONCE(1);
-		fallthrough;
-	case 0:
-		return CLOCK_MONOTONIC;
-	}
-}
-
-static int io_linked_timeout_update(struct io_ring_ctx *ctx, __u64 user_data,
-				    struct timespec64 *ts, enum hrtimer_mode mode)
-	__must_hold(&ctx->timeout_lock)
-{
-	struct io_timeout_data *io;
-	struct io_timeout *timeout;
-	struct io_kiocb *req = NULL;
-
-	list_for_each_entry(timeout, &ctx->ltimeout_list, list) {
-		struct io_kiocb *tmp = cmd_to_io_kiocb(timeout);
-
-		if (user_data == tmp->cqe.user_data) {
-			req = tmp;
-			break;
-		}
-	}
-	if (!req)
-		return -ENOENT;
-
-	io = req->async_data;
-	if (hrtimer_try_to_cancel(&io->timer) == -1)
-		return -EALREADY;
-	hrtimer_init(&io->timer, io_timeout_get_clock(io), mode);
-	io->timer.function = io_link_timeout_fn;
-	hrtimer_start(&io->timer, timespec64_to_ktime(*ts), mode);
-	return 0;
-}
-
-static int io_timeout_update(struct io_ring_ctx *ctx, __u64 user_data,
-			     struct timespec64 *ts, enum hrtimer_mode mode)
-	__must_hold(&ctx->timeout_lock)
-{
-	struct io_cancel_data cd = { .data = user_data, };
-	struct io_kiocb *req = io_timeout_extract(ctx, &cd);
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-	struct io_timeout_data *data;
-
-	if (IS_ERR(req))
-		return PTR_ERR(req);
-
-	timeout->off = 0; /* noseq */
-	data = req->async_data;
-	list_add_tail(&timeout->list, &ctx->timeout_list);
-	hrtimer_init(&data->timer, io_timeout_get_clock(data), mode);
-	data->timer.function = io_timeout_fn;
-	hrtimer_start(&data->timer, timespec64_to_ktime(*ts), mode);
-	return 0;
-}
-
-static int io_timeout_remove_prep(struct io_kiocb *req,
-				  const struct io_uring_sqe *sqe)
-{
-	struct io_timeout_rem *tr = io_kiocb_to_cmd(req);
-
-	if (unlikely(req->flags & (REQ_F_FIXED_FILE | REQ_F_BUFFER_SELECT)))
-		return -EINVAL;
-	if (sqe->buf_index || sqe->len || sqe->splice_fd_in)
-		return -EINVAL;
-
-	tr->ltimeout = false;
-	tr->addr = READ_ONCE(sqe->addr);
-	tr->flags = READ_ONCE(sqe->timeout_flags);
-	if (tr->flags & IORING_TIMEOUT_UPDATE_MASK) {
-		if (hweight32(tr->flags & IORING_TIMEOUT_CLOCK_MASK) > 1)
-			return -EINVAL;
-		if (tr->flags & IORING_LINK_TIMEOUT_UPDATE)
-			tr->ltimeout = true;
-		if (tr->flags & ~(IORING_TIMEOUT_UPDATE_MASK|IORING_TIMEOUT_ABS))
-			return -EINVAL;
-		if (get_timespec64(&tr->ts, u64_to_user_ptr(sqe->addr2)))
-			return -EFAULT;
-		if (tr->ts.tv_sec < 0 || tr->ts.tv_nsec < 0)
-			return -EINVAL;
-	} else if (tr->flags) {
-		/* timeout removal doesn't support flags */
-		return -EINVAL;
-	}
-
-	return 0;
-}
-
-static inline enum hrtimer_mode io_translate_timeout_mode(unsigned int flags)
-{
-	return (flags & IORING_TIMEOUT_ABS) ? HRTIMER_MODE_ABS
-					    : HRTIMER_MODE_REL;
-}
-
-/*
- * Remove or update an existing timeout command
- */
-static int io_timeout_remove(struct io_kiocb *req, unsigned int issue_flags)
-{
-	struct io_timeout_rem *tr = io_kiocb_to_cmd(req);
-	struct io_ring_ctx *ctx = req->ctx;
-	int ret;
-
-	if (!(tr->flags & IORING_TIMEOUT_UPDATE)) {
-		struct io_cancel_data cd = { .data = tr->addr, };
-
-		spin_lock(&ctx->completion_lock);
-		ret = io_timeout_cancel(ctx, &cd);
-		spin_unlock(&ctx->completion_lock);
-	} else {
-		enum hrtimer_mode mode = io_translate_timeout_mode(tr->flags);
-
-		spin_lock_irq(&ctx->timeout_lock);
-		if (tr->ltimeout)
-			ret = io_linked_timeout_update(ctx, tr->addr, &tr->ts, mode);
-		else
-			ret = io_timeout_update(ctx, tr->addr, &tr->ts, mode);
-		spin_unlock_irq(&ctx->timeout_lock);
-	}
-
-	if (ret < 0)
-		req_set_fail(req);
-	io_req_set_res(req, ret, 0);
-	return IOU_OK;
-}
-
-static int __io_timeout_prep(struct io_kiocb *req,
-			     const struct io_uring_sqe *sqe,
-			     bool is_timeout_link)
-{
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-	struct io_timeout_data *data;
-	unsigned flags;
-	u32 off = READ_ONCE(sqe->off);
-
-	if (sqe->buf_index || sqe->len != 1 || sqe->splice_fd_in)
-		return -EINVAL;
-	if (off && is_timeout_link)
-		return -EINVAL;
-	flags = READ_ONCE(sqe->timeout_flags);
-	if (flags & ~(IORING_TIMEOUT_ABS | IORING_TIMEOUT_CLOCK_MASK |
-		      IORING_TIMEOUT_ETIME_SUCCESS))
-		return -EINVAL;
-	/* more than one clock specified is invalid, obviously */
-	if (hweight32(flags & IORING_TIMEOUT_CLOCK_MASK) > 1)
-		return -EINVAL;
-
-	INIT_LIST_HEAD(&timeout->list);
-	timeout->off = off;
-	if (unlikely(off && !req->ctx->off_timeout_used))
-		req->ctx->off_timeout_used = true;
-
-	if (WARN_ON_ONCE(req_has_async_data(req)))
-		return -EFAULT;
-	if (io_alloc_async_data(req))
-		return -ENOMEM;
-
-	data = req->async_data;
-	data->req = req;
-	data->flags = flags;
-
-	if (get_timespec64(&data->ts, u64_to_user_ptr(sqe->addr)))
-		return -EFAULT;
-
-	if (data->ts.tv_sec < 0 || data->ts.tv_nsec < 0)
-		return -EINVAL;
-
-	INIT_LIST_HEAD(&timeout->list);
-	data->mode = io_translate_timeout_mode(flags);
-	hrtimer_init(&data->timer, io_timeout_get_clock(data), data->mode);
-
-	if (is_timeout_link) {
-		struct io_submit_link *link = &req->ctx->submit_state.link;
-
-		if (!link->head)
-			return -EINVAL;
-		if (link->last->opcode == IORING_OP_LINK_TIMEOUT)
-			return -EINVAL;
-		timeout->head = link->last;
-		link->last->flags |= REQ_F_ARM_LTIMEOUT;
-	}
-	return 0;
-}
-
-static int io_timeout_prep(struct io_kiocb *req,
-			   const struct io_uring_sqe *sqe)
-{
-	return __io_timeout_prep(req, sqe, false);
-}
-
-static int io_link_timeout_prep(struct io_kiocb *req,
-				const struct io_uring_sqe *sqe)
-{
-	return __io_timeout_prep(req, sqe, true);
-}
-
-static int io_timeout(struct io_kiocb *req, unsigned int issue_flags)
-{
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-	struct io_ring_ctx *ctx = req->ctx;
-	struct io_timeout_data *data = req->async_data;
-	struct list_head *entry;
-	u32 tail, off = timeout->off;
-
-	spin_lock_irq(&ctx->timeout_lock);
-
-	/*
-	 * sqe->off holds how many events that need to occur for this
-	 * timeout event to be satisfied. If it isn't set, then this is
-	 * a pure timeout request, sequence isn't used.
-	 */
-	if (io_is_timeout_noseq(req)) {
-		entry = ctx->timeout_list.prev;
-		goto add;
-	}
-
-	tail = ctx->cached_cq_tail - atomic_read(&ctx->cq_timeouts);
-	timeout->target_seq = tail + off;
-
-	/* Update the last seq here in case io_flush_timeouts() hasn't.
-	 * This is safe because ->completion_lock is held, and submissions
-	 * and completions are never mixed in the same ->completion_lock section.
-	 */
-	ctx->cq_last_tm_flush = tail;
-
-	/*
-	 * Insertion sort, ensuring the first entry in the list is always
-	 * the one we need first.
-	 */
-	list_for_each_prev(entry, &ctx->timeout_list) {
-		struct io_timeout *nextt = list_entry(entry, struct io_timeout, list);
-		struct io_kiocb *nxt = cmd_to_io_kiocb(nextt);
-
-		if (io_is_timeout_noseq(nxt))
-			continue;
-		/* nxt.seq is behind @tail, otherwise would've been completed */
-		if (off >= nextt->target_seq - tail)
-			break;
-	}
-add:
-	list_add(&timeout->list, entry);
-	data->timer.function = io_timeout_fn;
-	hrtimer_start(&data->timer, timespec64_to_ktime(data->ts), data->mode);
-	spin_unlock_irq(&ctx->timeout_lock);
-	return IOU_ISSUE_SKIP_COMPLETE;
-}
-
 static bool io_cancel_cb(struct io_wq_work *work, void *data)
 {
 	struct io_kiocb *req = container_of(work, struct io_kiocb, work);
@@ -4979,7 +4433,7 @@ static int io_async_cancel_one(struct io_uring_task *tctx,
 	return ret;
 }
 
-static int io_try_cancel(struct io_kiocb *req, struct io_cancel_data *cd)
+int io_try_cancel(struct io_kiocb *req, struct io_cancel_data *cd)
 {
 	struct io_ring_ctx *ctx = req->ctx;
 	int ret;
@@ -5462,84 +4916,6 @@ struct file *io_file_get_normal(struct io_kiocb *req, int fd)
 	return file;
 }
 
-static void io_req_task_link_timeout(struct io_kiocb *req, bool *locked)
-{
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-	struct io_kiocb *prev = timeout->prev;
-	int ret = -ENOENT;
-
-	if (prev) {
-		if (!(req->task->flags & PF_EXITING)) {
-			struct io_cancel_data cd = {
-				.ctx		= req->ctx,
-				.data		= prev->cqe.user_data,
-			};
-
-			ret = io_try_cancel(req, &cd);
-		}
-		io_req_set_res(req, ret ?: -ETIME, 0);
-		io_req_complete_post(req);
-		io_put_req(prev);
-	} else {
-		io_req_set_res(req, -ETIME, 0);
-		io_req_complete_post(req);
-	}
-}
-
-static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer)
-{
-	struct io_timeout_data *data = container_of(timer,
-						struct io_timeout_data, timer);
-	struct io_kiocb *prev, *req = data->req;
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-	struct io_ring_ctx *ctx = req->ctx;
-	unsigned long flags;
-
-	spin_lock_irqsave(&ctx->timeout_lock, flags);
-	prev = timeout->head;
-	timeout->head = NULL;
-
-	/*
-	 * We don't expect the list to be empty, that will only happen if we
-	 * race with the completion of the linked work.
-	 */
-	if (prev) {
-		io_remove_next_linked(prev);
-		if (!req_ref_inc_not_zero(prev))
-			prev = NULL;
-	}
-	list_del(&timeout->list);
-	timeout->prev = prev;
-	spin_unlock_irqrestore(&ctx->timeout_lock, flags);
-
-	req->io_task_work.func = io_req_task_link_timeout;
-	io_req_task_work_add(req);
-	return HRTIMER_NORESTART;
-}
-
-static void io_queue_linked_timeout(struct io_kiocb *req)
-{
-	struct io_timeout *timeout = io_kiocb_to_cmd(req);
-	struct io_ring_ctx *ctx = req->ctx;
-
-	spin_lock_irq(&ctx->timeout_lock);
-	/*
-	 * If the back reference is NULL, then our linked request finished
-	 * before we got a chance to setup the timer
-	 */
-	if (timeout->head) {
-		struct io_timeout_data *data = req->async_data;
-
-		data->timer.function = io_link_timeout_fn;
-		hrtimer_start(&data->timer, timespec64_to_ktime(data->ts),
-				data->mode);
-		list_add_tail(&timeout->list, &ctx->ltimeout_list);
-	}
-	spin_unlock_irq(&ctx->timeout_lock);
-	/* drop submission reference */
-	io_put_req(req);
-}
-
 static void io_queue_async(struct io_kiocb *req, int ret)
 	__must_hold(&req->ctx->uring_lock)
 {
@@ -8116,31 +7492,6 @@ static __cold void io_ring_exit_work(struct work_struct *work)
 	io_ring_ctx_free(ctx);
 }
 
-/* Returns true if we found and killed one or more timeouts */
-static __cold bool io_kill_timeouts(struct io_ring_ctx *ctx,
-				    struct task_struct *tsk, bool cancel_all)
-{
-	struct io_timeout *timeout, *tmp;
-	int canceled = 0;
-
-	spin_lock(&ctx->completion_lock);
-	spin_lock_irq(&ctx->timeout_lock);
-	list_for_each_entry_safe(timeout, tmp, &ctx->timeout_list, list) {
-		struct io_kiocb *req = cmd_to_io_kiocb(timeout);
-
-		if (io_match_task(req, tsk, cancel_all)) {
-			io_kill_timeout(req, -ECANCELED);
-			canceled++;
-		}
-	}
-	spin_unlock_irq(&ctx->timeout_lock);
-	io_commit_cqring(ctx);
-	spin_unlock(&ctx->completion_lock);
-	if (canceled != 0)
-		io_cqring_ev_posted(ctx);
-	return canceled != 0;
-}
-
 static __cold void io_ring_ctx_wait_and_kill(struct io_ring_ctx *ctx)
 {
 	unsigned long index;
diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h
index 4b46385720c5..e285e12ccbdb 100644
--- a/io_uring/io_uring.h
+++ b/io_uring/io_uring.h
@@ -65,7 +65,8 @@ static inline void io_commit_cqring(struct io_ring_ctx *ctx)
 }
 
 void __io_req_complete(struct io_kiocb *req, unsigned issue_flags);
-
+void io_req_complete_post(struct io_kiocb *req);
+void __io_req_complete_post(struct io_kiocb *req);
 bool io_fill_cqe_aux(struct io_ring_ctx *ctx, u64 user_data, s32 res,
 		     u32 cflags);
 void io_cqring_ev_posted(struct io_ring_ctx *ctx);
@@ -96,5 +97,15 @@ void io_rsrc_node_switch(struct io_ring_ctx *ctx,
 bool io_is_uring_fops(struct file *file);
 bool io_alloc_async_data(struct io_kiocb *req);
 void io_req_task_work_add(struct io_kiocb *req);
+void io_req_tw_post_queue(struct io_kiocb *req, s32 res, u32 cflags);
+void io_req_task_complete(struct io_kiocb *req, bool *locked);
+void io_req_task_queue_fail(struct io_kiocb *req, int ret);
+int io_try_cancel(struct io_kiocb *req, struct io_cancel_data *cd);
+
+void io_free_req(struct io_kiocb *req);
+void io_queue_next(struct io_kiocb *req);
+
+#define io_for_each_link(pos, head) \
+	for (pos = (head); pos; pos = pos->link)
 
 #endif
diff --git a/io_uring/io_uring_types.h b/io_uring/io_uring_types.h
index dba72113c59d..349524907b6b 100644
--- a/io_uring/io_uring_types.h
+++ b/io_uring/io_uring_types.h
@@ -488,4 +488,14 @@ struct io_kiocb {
 	struct io_wq_work		work;
 };
 
+struct io_cancel_data {
+	struct io_ring_ctx *ctx;
+	union {
+		u64 data;
+		struct file *file;
+	};
+	u32 flags;
+	int seq;
+};
+
 #endif
diff --git a/io_uring/timeout.c b/io_uring/timeout.c
new file mode 100644
index 000000000000..5e42bfcd683e
--- /dev/null
+++ b/io_uring/timeout.c
@@ -0,0 +1,634 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/file.h>
+#include <linux/io_uring.h>
+
+#include <trace/events/io_uring.h>
+
+#include <uapi/linux/io_uring.h>
+
+#include "io_uring_types.h"
+#include "io_uring.h"
+#include "refs.h"
+#include "timeout.h"
+
+struct io_timeout {
+	struct file			*file;
+	u32				off;
+	u32				target_seq;
+	struct list_head		list;
+	/* head of the link, used by linked timeouts only */
+	struct io_kiocb			*head;
+	/* for linked completions */
+	struct io_kiocb			*prev;
+};
+
+struct io_timeout_rem {
+	struct file			*file;
+	u64				addr;
+
+	/* timeout update */
+	struct timespec64		ts;
+	u32				flags;
+	bool				ltimeout;
+};
+
+static inline bool io_is_timeout_noseq(struct io_kiocb *req)
+{
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+
+	return !timeout->off;
+}
+
+static inline void io_put_req(struct io_kiocb *req)
+{
+	if (req_ref_put_and_test(req)) {
+		io_queue_next(req);
+		io_free_req(req);
+	}
+}
+
+static void io_kill_timeout(struct io_kiocb *req, int status)
+	__must_hold(&req->ctx->completion_lock)
+	__must_hold(&req->ctx->timeout_lock)
+{
+	struct io_timeout_data *io = req->async_data;
+
+	if (hrtimer_try_to_cancel(&io->timer) != -1) {
+		struct io_timeout *timeout = io_kiocb_to_cmd(req);
+
+		if (status)
+			req_set_fail(req);
+		atomic_set(&req->ctx->cq_timeouts,
+			atomic_read(&req->ctx->cq_timeouts) + 1);
+		list_del_init(&timeout->list);
+		io_req_tw_post_queue(req, status, 0);
+	}
+}
+
+__cold void io_flush_timeouts(struct io_ring_ctx *ctx)
+	__must_hold(&ctx->completion_lock)
+{
+	u32 seq = ctx->cached_cq_tail - atomic_read(&ctx->cq_timeouts);
+	struct io_timeout *timeout, *tmp;
+
+	spin_lock_irq(&ctx->timeout_lock);
+	list_for_each_entry_safe(timeout, tmp, &ctx->timeout_list, list) {
+		struct io_kiocb *req = cmd_to_io_kiocb(timeout);
+		u32 events_needed, events_got;
+
+		if (io_is_timeout_noseq(req))
+			break;
+
+		/*
+		 * Since seq can easily wrap around over time, subtract
+		 * the last seq at which timeouts were flushed before comparing.
+		 * Assuming not more than 2^31-1 events have happened since,
+		 * these subtractions won't have wrapped, so we can check if
+		 * target is in [last_seq, current_seq] by comparing the two.
+		 */
+		events_needed = timeout->target_seq - ctx->cq_last_tm_flush;
+		events_got = seq - ctx->cq_last_tm_flush;
+		if (events_got < events_needed)
+			break;
+
+		io_kill_timeout(req, 0);
+	}
+	ctx->cq_last_tm_flush = seq;
+	spin_unlock_irq(&ctx->timeout_lock);
+}
+
+static void io_fail_links(struct io_kiocb *req)
+	__must_hold(&req->ctx->completion_lock)
+{
+	struct io_kiocb *nxt, *link = req->link;
+	bool ignore_cqes = req->flags & REQ_F_SKIP_LINK_CQES;
+
+	req->link = NULL;
+	while (link) {
+		long res = -ECANCELED;
+
+		if (link->flags & REQ_F_FAIL)
+			res = link->cqe.res;
+
+		nxt = link->link;
+		link->link = NULL;
+
+		trace_io_uring_fail_link(req->ctx, req, req->cqe.user_data,
+					req->opcode, link);
+
+		if (ignore_cqes)
+			link->flags |= REQ_F_CQE_SKIP;
+		else
+			link->flags &= ~REQ_F_CQE_SKIP;
+		io_req_set_res(link, res, 0);
+		__io_req_complete_post(link);
+		link = nxt;
+	}
+}
+
+static inline void io_remove_next_linked(struct io_kiocb *req)
+{
+	struct io_kiocb *nxt = req->link;
+
+	req->link = nxt->link;
+	nxt->link = NULL;
+}
+
+bool io_disarm_next(struct io_kiocb *req)
+	__must_hold(&req->ctx->completion_lock)
+{
+	struct io_kiocb *link = NULL;
+	bool posted = false;
+
+	if (req->flags & REQ_F_ARM_LTIMEOUT) {
+		link = req->link;
+		req->flags &= ~REQ_F_ARM_LTIMEOUT;
+		if (link && link->opcode == IORING_OP_LINK_TIMEOUT) {
+			io_remove_next_linked(req);
+			io_req_tw_post_queue(link, -ECANCELED, 0);
+			posted = true;
+		}
+	} else if (req->flags & REQ_F_LINK_TIMEOUT) {
+		struct io_ring_ctx *ctx = req->ctx;
+
+		spin_lock_irq(&ctx->timeout_lock);
+		link = io_disarm_linked_timeout(req);
+		spin_unlock_irq(&ctx->timeout_lock);
+		if (link) {
+			posted = true;
+			io_req_tw_post_queue(link, -ECANCELED, 0);
+		}
+	}
+	if (unlikely((req->flags & REQ_F_FAIL) &&
+		     !(req->flags & REQ_F_HARDLINK))) {
+		posted |= (req->link != NULL);
+		io_fail_links(req);
+	}
+	return posted;
+}
+
+struct io_kiocb *__io_disarm_linked_timeout(struct io_kiocb *req,
+					    struct io_kiocb *link)
+	__must_hold(&req->ctx->completion_lock)
+	__must_hold(&req->ctx->timeout_lock)
+{
+	struct io_timeout_data *io = link->async_data;
+	struct io_timeout *timeout = io_kiocb_to_cmd(link);
+
+	io_remove_next_linked(req);
+	timeout->head = NULL;
+	if (hrtimer_try_to_cancel(&io->timer) != -1) {
+		list_del(&timeout->list);
+		return link;
+	}
+
+	return NULL;
+}
+
+static enum hrtimer_restart io_timeout_fn(struct hrtimer *timer)
+{
+	struct io_timeout_data *data = container_of(timer,
+						struct io_timeout_data, timer);
+	struct io_kiocb *req = data->req;
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+	struct io_ring_ctx *ctx = req->ctx;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ctx->timeout_lock, flags);
+	list_del_init(&timeout->list);
+	atomic_set(&req->ctx->cq_timeouts,
+		atomic_read(&req->ctx->cq_timeouts) + 1);
+	spin_unlock_irqrestore(&ctx->timeout_lock, flags);
+
+	if (!(data->flags & IORING_TIMEOUT_ETIME_SUCCESS))
+		req_set_fail(req);
+
+	io_req_set_res(req, -ETIME, 0);
+	req->io_task_work.func = io_req_task_complete;
+	io_req_task_work_add(req);
+	return HRTIMER_NORESTART;
+}
+
+static struct io_kiocb *io_timeout_extract(struct io_ring_ctx *ctx,
+					   struct io_cancel_data *cd)
+	__must_hold(&ctx->timeout_lock)
+{
+	struct io_timeout *timeout;
+	struct io_timeout_data *io;
+	struct io_kiocb *req = NULL;
+
+	list_for_each_entry(timeout, &ctx->timeout_list, list) {
+		struct io_kiocb *tmp = cmd_to_io_kiocb(timeout);
+
+		if (!(cd->flags & IORING_ASYNC_CANCEL_ANY) &&
+		    cd->data != tmp->cqe.user_data)
+			continue;
+		if (cd->flags & (IORING_ASYNC_CANCEL_ALL|IORING_ASYNC_CANCEL_ANY)) {
+			if (cd->seq == tmp->work.cancel_seq)
+				continue;
+			tmp->work.cancel_seq = cd->seq;
+		}
+		req = tmp;
+		break;
+	}
+	if (!req)
+		return ERR_PTR(-ENOENT);
+
+	io = req->async_data;
+	if (hrtimer_try_to_cancel(&io->timer) == -1)
+		return ERR_PTR(-EALREADY);
+	timeout = io_kiocb_to_cmd(req);
+	list_del_init(&timeout->list);
+	return req;
+}
+
+int io_timeout_cancel(struct io_ring_ctx *ctx, struct io_cancel_data *cd)
+	__must_hold(&ctx->completion_lock)
+{
+	struct io_kiocb *req;
+
+	spin_lock_irq(&ctx->timeout_lock);
+	req = io_timeout_extract(ctx, cd);
+	spin_unlock_irq(&ctx->timeout_lock);
+
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+	io_req_task_queue_fail(req, -ECANCELED);
+	return 0;
+}
+
+static void io_req_task_link_timeout(struct io_kiocb *req, bool *locked)
+{
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+	struct io_kiocb *prev = timeout->prev;
+	int ret = -ENOENT;
+
+	if (prev) {
+		if (!(req->task->flags & PF_EXITING)) {
+			struct io_cancel_data cd = {
+				.ctx		= req->ctx,
+				.data		= prev->cqe.user_data,
+			};
+
+			ret = io_try_cancel(req, &cd);
+		}
+		io_req_set_res(req, ret ?: -ETIME, 0);
+		io_req_complete_post(req);
+		io_put_req(prev);
+	} else {
+		io_req_set_res(req, -ETIME, 0);
+		io_req_complete_post(req);
+	}
+}
+
+static enum hrtimer_restart io_link_timeout_fn(struct hrtimer *timer)
+{
+	struct io_timeout_data *data = container_of(timer,
+						struct io_timeout_data, timer);
+	struct io_kiocb *prev, *req = data->req;
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+	struct io_ring_ctx *ctx = req->ctx;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ctx->timeout_lock, flags);
+	prev = timeout->head;
+	timeout->head = NULL;
+
+	/*
+	 * We don't expect the list to be empty, that will only happen if we
+	 * race with the completion of the linked work.
+	 */
+	if (prev) {
+		io_remove_next_linked(prev);
+		if (!req_ref_inc_not_zero(prev))
+			prev = NULL;
+	}
+	list_del(&timeout->list);
+	timeout->prev = prev;
+	spin_unlock_irqrestore(&ctx->timeout_lock, flags);
+
+	req->io_task_work.func = io_req_task_link_timeout;
+	io_req_task_work_add(req);
+	return HRTIMER_NORESTART;
+}
+
+static clockid_t io_timeout_get_clock(struct io_timeout_data *data)
+{
+	switch (data->flags & IORING_TIMEOUT_CLOCK_MASK) {
+	case IORING_TIMEOUT_BOOTTIME:
+		return CLOCK_BOOTTIME;
+	case IORING_TIMEOUT_REALTIME:
+		return CLOCK_REALTIME;
+	default:
+		/* can't happen, vetted at prep time */
+		WARN_ON_ONCE(1);
+		fallthrough;
+	case 0:
+		return CLOCK_MONOTONIC;
+	}
+}
+
+static int io_linked_timeout_update(struct io_ring_ctx *ctx, __u64 user_data,
+				    struct timespec64 *ts, enum hrtimer_mode mode)
+	__must_hold(&ctx->timeout_lock)
+{
+	struct io_timeout_data *io;
+	struct io_timeout *timeout;
+	struct io_kiocb *req = NULL;
+
+	list_for_each_entry(timeout, &ctx->ltimeout_list, list) {
+		struct io_kiocb *tmp = cmd_to_io_kiocb(timeout);
+
+		if (user_data == tmp->cqe.user_data) {
+			req = tmp;
+			break;
+		}
+	}
+	if (!req)
+		return -ENOENT;
+
+	io = req->async_data;
+	if (hrtimer_try_to_cancel(&io->timer) == -1)
+		return -EALREADY;
+	hrtimer_init(&io->timer, io_timeout_get_clock(io), mode);
+	io->timer.function = io_link_timeout_fn;
+	hrtimer_start(&io->timer, timespec64_to_ktime(*ts), mode);
+	return 0;
+}
+
+static int io_timeout_update(struct io_ring_ctx *ctx, __u64 user_data,
+			     struct timespec64 *ts, enum hrtimer_mode mode)
+	__must_hold(&ctx->timeout_lock)
+{
+	struct io_cancel_data cd = { .data = user_data, };
+	struct io_kiocb *req = io_timeout_extract(ctx, &cd);
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+	struct io_timeout_data *data;
+
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+
+	timeout->off = 0; /* noseq */
+	data = req->async_data;
+	list_add_tail(&timeout->list, &ctx->timeout_list);
+	hrtimer_init(&data->timer, io_timeout_get_clock(data), mode);
+	data->timer.function = io_timeout_fn;
+	hrtimer_start(&data->timer, timespec64_to_ktime(*ts), mode);
+	return 0;
+}
+
+int io_timeout_remove_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+	struct io_timeout_rem *tr = io_kiocb_to_cmd(req);
+
+	if (unlikely(req->flags & (REQ_F_FIXED_FILE | REQ_F_BUFFER_SELECT)))
+		return -EINVAL;
+	if (sqe->buf_index || sqe->len || sqe->splice_fd_in)
+		return -EINVAL;
+
+	tr->ltimeout = false;
+	tr->addr = READ_ONCE(sqe->addr);
+	tr->flags = READ_ONCE(sqe->timeout_flags);
+	if (tr->flags & IORING_TIMEOUT_UPDATE_MASK) {
+		if (hweight32(tr->flags & IORING_TIMEOUT_CLOCK_MASK) > 1)
+			return -EINVAL;
+		if (tr->flags & IORING_LINK_TIMEOUT_UPDATE)
+			tr->ltimeout = true;
+		if (tr->flags & ~(IORING_TIMEOUT_UPDATE_MASK|IORING_TIMEOUT_ABS))
+			return -EINVAL;
+		if (get_timespec64(&tr->ts, u64_to_user_ptr(sqe->addr2)))
+			return -EFAULT;
+		if (tr->ts.tv_sec < 0 || tr->ts.tv_nsec < 0)
+			return -EINVAL;
+	} else if (tr->flags) {
+		/* timeout removal doesn't support flags */
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static inline enum hrtimer_mode io_translate_timeout_mode(unsigned int flags)
+{
+	return (flags & IORING_TIMEOUT_ABS) ? HRTIMER_MODE_ABS
+					    : HRTIMER_MODE_REL;
+}
+
+/*
+ * Remove or update an existing timeout command
+ */
+int io_timeout_remove(struct io_kiocb *req, unsigned int issue_flags)
+{
+	struct io_timeout_rem *tr = io_kiocb_to_cmd(req);
+	struct io_ring_ctx *ctx = req->ctx;
+	int ret;
+
+	if (!(tr->flags & IORING_TIMEOUT_UPDATE)) {
+		struct io_cancel_data cd = { .data = tr->addr, };
+
+		spin_lock(&ctx->completion_lock);
+		ret = io_timeout_cancel(ctx, &cd);
+		spin_unlock(&ctx->completion_lock);
+	} else {
+		enum hrtimer_mode mode = io_translate_timeout_mode(tr->flags);
+
+		spin_lock_irq(&ctx->timeout_lock);
+		if (tr->ltimeout)
+			ret = io_linked_timeout_update(ctx, tr->addr, &tr->ts, mode);
+		else
+			ret = io_timeout_update(ctx, tr->addr, &tr->ts, mode);
+		spin_unlock_irq(&ctx->timeout_lock);
+	}
+
+	if (ret < 0)
+		req_set_fail(req);
+	io_req_set_res(req, ret, 0);
+	return IOU_OK;
+}
+
+static int __io_timeout_prep(struct io_kiocb *req,
+			     const struct io_uring_sqe *sqe,
+			     bool is_timeout_link)
+{
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+	struct io_timeout_data *data;
+	unsigned flags;
+	u32 off = READ_ONCE(sqe->off);
+
+	if (sqe->buf_index || sqe->len != 1 || sqe->splice_fd_in)
+		return -EINVAL;
+	if (off && is_timeout_link)
+		return -EINVAL;
+	flags = READ_ONCE(sqe->timeout_flags);
+	if (flags & ~(IORING_TIMEOUT_ABS | IORING_TIMEOUT_CLOCK_MASK |
+		      IORING_TIMEOUT_ETIME_SUCCESS))
+		return -EINVAL;
+	/* more than one clock specified is invalid, obviously */
+	if (hweight32(flags & IORING_TIMEOUT_CLOCK_MASK) > 1)
+		return -EINVAL;
+
+	INIT_LIST_HEAD(&timeout->list);
+	timeout->off = off;
+	if (unlikely(off && !req->ctx->off_timeout_used))
+		req->ctx->off_timeout_used = true;
+
+	if (WARN_ON_ONCE(req_has_async_data(req)))
+		return -EFAULT;
+	if (io_alloc_async_data(req))
+		return -ENOMEM;
+
+	data = req->async_data;
+	data->req = req;
+	data->flags = flags;
+
+	if (get_timespec64(&data->ts, u64_to_user_ptr(sqe->addr)))
+		return -EFAULT;
+
+	if (data->ts.tv_sec < 0 || data->ts.tv_nsec < 0)
+		return -EINVAL;
+
+	INIT_LIST_HEAD(&timeout->list);
+	data->mode = io_translate_timeout_mode(flags);
+	hrtimer_init(&data->timer, io_timeout_get_clock(data), data->mode);
+
+	if (is_timeout_link) {
+		struct io_submit_link *link = &req->ctx->submit_state.link;
+
+		if (!link->head)
+			return -EINVAL;
+		if (link->last->opcode == IORING_OP_LINK_TIMEOUT)
+			return -EINVAL;
+		timeout->head = link->last;
+		link->last->flags |= REQ_F_ARM_LTIMEOUT;
+	}
+	return 0;
+}
+
+int io_timeout_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+	return __io_timeout_prep(req, sqe, false);
+}
+
+int io_link_timeout_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+	return __io_timeout_prep(req, sqe, true);
+}
+
+int io_timeout(struct io_kiocb *req, unsigned int issue_flags)
+{
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+	struct io_ring_ctx *ctx = req->ctx;
+	struct io_timeout_data *data = req->async_data;
+	struct list_head *entry;
+	u32 tail, off = timeout->off;
+
+	spin_lock_irq(&ctx->timeout_lock);
+
+	/*
+	 * sqe->off holds how many events that need to occur for this
+	 * timeout event to be satisfied. If it isn't set, then this is
+	 * a pure timeout request, sequence isn't used.
+	 */
+	if (io_is_timeout_noseq(req)) {
+		entry = ctx->timeout_list.prev;
+		goto add;
+	}
+
+	tail = ctx->cached_cq_tail - atomic_read(&ctx->cq_timeouts);
+	timeout->target_seq = tail + off;
+
+	/* Update the last seq here in case io_flush_timeouts() hasn't.
+	 * This is safe because ->completion_lock is held, and submissions
+	 * and completions are never mixed in the same ->completion_lock section.
+	 */
+	ctx->cq_last_tm_flush = tail;
+
+	/*
+	 * Insertion sort, ensuring the first entry in the list is always
+	 * the one we need first.
+	 */
+	list_for_each_prev(entry, &ctx->timeout_list) {
+		struct io_timeout *nextt = list_entry(entry, struct io_timeout, list);
+		struct io_kiocb *nxt = cmd_to_io_kiocb(nextt);
+
+		if (io_is_timeout_noseq(nxt))
+			continue;
+		/* nxt.seq is behind @tail, otherwise would've been completed */
+		if (off >= nextt->target_seq - tail)
+			break;
+	}
+add:
+	list_add(&timeout->list, entry);
+	data->timer.function = io_timeout_fn;
+	hrtimer_start(&data->timer, timespec64_to_ktime(data->ts), data->mode);
+	spin_unlock_irq(&ctx->timeout_lock);
+	return IOU_ISSUE_SKIP_COMPLETE;
+}
+
+void io_queue_linked_timeout(struct io_kiocb *req)
+{
+	struct io_timeout *timeout = io_kiocb_to_cmd(req);
+	struct io_ring_ctx *ctx = req->ctx;
+
+	spin_lock_irq(&ctx->timeout_lock);
+	/*
+	 * If the back reference is NULL, then our linked request finished
+	 * before we got a chance to setup the timer
+	 */
+	if (timeout->head) {
+		struct io_timeout_data *data = req->async_data;
+
+		data->timer.function = io_link_timeout_fn;
+		hrtimer_start(&data->timer, timespec64_to_ktime(data->ts),
+				data->mode);
+		list_add_tail(&timeout->list, &ctx->ltimeout_list);
+	}
+	spin_unlock_irq(&ctx->timeout_lock);
+	/* drop submission reference */
+	io_put_req(req);
+}
+
+static bool io_match_task(struct io_kiocb *head, struct task_struct *task,
+			  bool cancel_all)
+	__must_hold(&req->ctx->timeout_lock)
+{
+	struct io_kiocb *req;
+
+	if (task && head->task != task)
+		return false;
+	if (cancel_all)
+		return true;
+
+	io_for_each_link(req, head) {
+		if (req->flags & REQ_F_INFLIGHT)
+			return true;
+	}
+	return false;
+}
+
+/* Returns true if we found and killed one or more timeouts */
+__cold bool io_kill_timeouts(struct io_ring_ctx *ctx, struct task_struct *tsk,
+			     bool cancel_all)
+{
+	struct io_timeout *timeout, *tmp;
+	int canceled = 0;
+
+	spin_lock(&ctx->completion_lock);
+	spin_lock_irq(&ctx->timeout_lock);
+	list_for_each_entry_safe(timeout, tmp, &ctx->timeout_list, list) {
+		struct io_kiocb *req = cmd_to_io_kiocb(timeout);
+
+		if (io_match_task(req, tsk, cancel_all)) {
+			io_kill_timeout(req, -ECANCELED);
+			canceled++;
+		}
+	}
+	spin_unlock_irq(&ctx->timeout_lock);
+	io_commit_cqring(ctx);
+	spin_unlock(&ctx->completion_lock);
+	if (canceled != 0)
+		io_cqring_ev_posted(ctx);
+	return canceled != 0;
+}
diff --git a/io_uring/timeout.h b/io_uring/timeout.h
new file mode 100644
index 000000000000..dd7cfb0d9366
--- /dev/null
+++ b/io_uring/timeout.h
@@ -0,0 +1,35 @@
+// SPDX-License-Identifier: GPL-2.0
+
+struct io_timeout_data {
+	struct io_kiocb			*req;
+	struct hrtimer			timer;
+	struct timespec64		ts;
+	enum hrtimer_mode		mode;
+	u32				flags;
+};
+
+struct io_kiocb *__io_disarm_linked_timeout(struct io_kiocb *req,
+					    struct io_kiocb *link);
+
+static inline struct io_kiocb *io_disarm_linked_timeout(struct io_kiocb *req)
+{
+	struct io_kiocb *link = req->link;
+
+	if (link && link->opcode == IORING_OP_LINK_TIMEOUT)
+		return __io_disarm_linked_timeout(req, link);
+
+	return NULL;
+}
+
+__cold void io_flush_timeouts(struct io_ring_ctx *ctx);
+int io_timeout_cancel(struct io_ring_ctx *ctx, struct io_cancel_data *cd);
+__cold bool io_kill_timeouts(struct io_ring_ctx *ctx, struct task_struct *tsk,
+			     bool cancel_all);
+void io_queue_linked_timeout(struct io_kiocb *req);
+bool io_disarm_next(struct io_kiocb *req);
+
+int io_timeout_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
+int io_link_timeout_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
+int io_timeout(struct io_kiocb *req, unsigned int issue_flags);
+int io_timeout_remove_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
+int io_timeout_remove(struct io_kiocb *req, unsigned int issue_flags);