[PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Jens Axboe]

If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
the buffered read to an io-wq worker. Instead we can rely on page
unlocking callbacks to support retry based async IO. This is a lot more
efficient than doing async thread offload.

The retry is done similarly to how we handle poll based retry. From
the unlock callback, we simply queue the retry to a task_work based
handler.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/io_uring.c | 99 +++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 99 insertions(+)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index e95481c552ff..dd532d2634c2 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -498,6 +498,8 @@ struct io_async_rw {
 	struct iovec			*iov;
 	ssize_t				nr_segs;
 	ssize_t				size;
+	struct wait_page_queue		wpq;
+	struct callback_head		task_work;
 };
 
 struct io_async_ctx {
@@ -2568,6 +2570,99 @@ static int io_read_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	return 0;
 }
 
+static void io_async_buf_cancel(struct callback_head *cb)
+{
+	struct io_async_rw *rw;
+	struct io_ring_ctx *ctx;
+	struct io_kiocb *req;
+
+	rw = container_of(cb, struct io_async_rw, task_work);
+	req = rw->wpq.wait.private;
+	ctx = req->ctx;
+
+	spin_lock_irq(&ctx->completion_lock);
+	io_cqring_fill_event(req, -ECANCELED);
+	io_commit_cqring(ctx);
+	spin_unlock_irq(&ctx->completion_lock);
+
+	io_cqring_ev_posted(ctx);
+	req_set_fail_links(req);
+	io_double_put_req(req);
+}
+
+static void io_async_buf_retry(struct callback_head *cb)
+{
+	struct io_async_rw *rw;
+	struct io_ring_ctx *ctx;
+	struct io_kiocb *req;
+
+	rw = container_of(cb, struct io_async_rw, task_work);
+	req = rw->wpq.wait.private;
+	ctx = req->ctx;
+
+	__set_current_state(TASK_RUNNING);
+	mutex_lock(&ctx->uring_lock);
+	__io_queue_sqe(req, NULL);
+	mutex_unlock(&ctx->uring_lock);
+}
+
+static int io_async_buf_func(struct wait_queue_entry *wait, unsigned mode,
+			     int sync, void *arg)
+{
+	struct wait_page_queue *wpq;
+	struct io_kiocb *req = wait->private;
+	struct io_async_rw *rw = &req->io->rw;
+	struct wait_page_key *key = arg;
+	struct task_struct *tsk;
+	int ret;
+
+	wpq = container_of(wait, struct wait_page_queue, wait);
+
+	ret = wake_page_match(wpq, key);
+	if (ret != 1)
+		return ret;
+
+	list_del_init(&wait->entry);
+
+	init_task_work(&rw->task_work, io_async_buf_retry);
+	/* submit ref gets dropped, acquire a new one */
+	refcount_inc(&req->refs);
+	tsk = req->task;
+	ret = task_work_add(tsk, &rw->task_work, true);
+	if (unlikely(ret)) {
+		/* queue just for cancelation */
+		init_task_work(&rw->task_work, io_async_buf_cancel);
+		tsk = io_wq_get_task(req->ctx->io_wq);
+		task_work_add(tsk, &rw->task_work, true);
+	}
+	wake_up_process(tsk);
+	return 1;
+}
+
+static bool io_rw_should_retry(struct io_kiocb *req)
+{
+	struct kiocb *kiocb = &req->rw.kiocb;
+	int ret;
+
+	/* already tried, or we're doing O_DIRECT */
+	if (kiocb->ki_flags & (IOCB_DIRECT | IOCB_WAITQ))
+		return false;
+	/*
+	 * just use poll if we can, and don't attempt if the fs doesn't
+	 * support callback based unlocks
+	 */
+	if (file_can_poll(req->file) || !(req->file->f_mode & FMODE_BUF_RASYNC))
+		return false;
+
+	ret = kiocb_wait_page_queue_init(kiocb, &req->io->rw.wpq,
+						io_async_buf_func, req);
+	if (ret)
+		return false;
+	get_task_struct(current);
+	req->task = current;
+	return true;
+}
+
 static int io_read(struct io_kiocb *req, bool force_nonblock)
 {
 	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
@@ -2601,6 +2696,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 	if (!ret) {
 		ssize_t ret2;
 
+retry:
 		if (req->file->f_op->read_iter)
 			ret2 = call_read_iter(req->file, kiocb, &iter);
 		else
@@ -2619,6 +2715,9 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 			if (!(req->flags & REQ_F_NOWAIT) &&
 			    !file_can_poll(req->file))
 				req->flags |= REQ_F_MUST_PUNT;
+			if (io_rw_should_retry(req))
+				goto retry;
+			kiocb->ki_flags &= ~IOCB_WAITQ;
 			return -EAGAIN;
 		}
 	}
-- 
2.26.2

[PATCHSET v4 0/12] Add support for async buffered reads

[Jens Axboe]

We technically support this already through io_uring, but it's
implemented with a thread backend to support cases where we would
block. This isn't ideal.

After a few prep patches, the core of this patchset is adding support
for async callbacks on page unlock. With this primitive, we can simply
retry the IO operation. With io_uring, this works a lot like poll based
retry for files that support it. If a page is currently locked and
needed, -EIOCBQUEUED is returned with a callback armed. The callers
callback is responsible for restarting the operation.

With this callback primitive, we can add support for
generic_file_buffered_read(), which is what most file systems end up
using for buffered reads. XFS/ext4/btrfs/bdev is wired up, but probably
trivial to add more.

The file flags support for this by setting FMODE_BUF_RASYNC, similar
to what we do for FMODE_NOWAIT. Open to suggestions here if this is
the preferred method or not.

In terms of results, I wrote a small test app that randomly reads 4G
of data in 4K chunks from a file hosted by ext4. The app uses a queue
depth of 32. If you want to test yourself, you can just use buffered=1
with ioengine=io_uring with fio. No application changes are needed to
use the more optimized buffered async read.

preadv for comparison:
	real    1m13.821s
	user    0m0.558s
	sys     0m11.125s
	CPU	~13%

Mainline:
	real    0m12.054s
	user    0m0.111s
	sys     0m5.659s
	CPU	~32% + ~50% == ~82%

This patchset:
	real    0m9.283s
	user    0m0.147s
	sys     0m4.619s
	CPU	~52%

The CPU numbers are just a rough estimate. For the mainline io_uring
run, this includes the app itself and all the threads doing IO on its
behalf (32% for the app, ~1.6% per worker and 32 of them). Context
switch rate is much smaller with the patchset, since we only have the
one task performing IO.

Also ran a simple fio based test case, varying the queue depth from 1
to 16, doubling every time:

[buf-test]
filename=/data/file
direct=0
ioengine=io_uring
norandommap
rw=randread
bs=4k
iodepth=${QD}
randseed=89
runtime=10s

QD/Test		Patchset IOPS		Mainline IOPS
1		9046			8294
2		19.8k			18.9k
4		39.2k			28.5k
8		64.4k			31.4k
16		65.7k			37.8k

Outside of my usual environment, so this is just running on a virtualized
NVMe device in qemu, using ext4 as the file system. NVMe isn't very
efficient virtualized, so we run out of steam at ~65K which is why we
flatline on the patched side (nvme_submit_cmd() eats ~75% of the test app
CPU). Before that happens, it's a linear increase. Not shown is context
switch rate, which is massively lower with the new code. The old thread
offload adds a blocking thread per pending IO, so context rate quickly
goes through the roof.

The goal here is efficiency. Async thread offload adds latency, and
it also adds noticable overhead on items such as adding pages to the
page cache. By allowing proper async buffered read support, we don't
have X threads hammering on the same inode page cache, we have just
the single app actually doing IO.

Been beating on this and it's solid for me, and I'm now pretty happy
with how it all turned out. Not aware of any missing bits/pieces or
code cleanups that need doing.

Series can also be found here:

https://git.kernel.dk/cgit/linux-block/log/?h=async-buffered.4

or pull from:

git://git.kernel.dk/linux-block async-buffered.4

 fs/block_dev.c            |   2 +-
 fs/btrfs/file.c           |   2 +-
 fs/ext4/file.c            |   2 +-
 fs/io_uring.c             | 114 ++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_file.c         |   2 +-
 include/linux/blk_types.h |   3 +-
 include/linux/fs.h        |  10 +++-
 include/linux/pagemap.h   |  67 ++++++++++++++++++++++
 mm/filemap.c              | 111 ++++++++++++++++++++++++-------------
 9 files changed, 267 insertions(+), 46 deletions(-)

Changes since v3:
- io_uring: don't retry if REQ_F_NOWAIT is set
- io_uring: alloc req->io if the request type didn't already
- Add iocb->ki_waitq instead of (ab)using iocb->private
Changes since v2:
- Get rid of unnecessary wait_page_async struct, just use wait_page_async
- Add another prep handler, adding wake_page_match()
- Use wake_page_match() in both callers
Changes since v1:
- Fix an issue with inline page locking
- Fix a potential race with __wait_on_page_locked_async()
- Fix a hang related to not setting page_match, thus missing a wakeup

-- 
Jens Axboe

[PATCH 01/12] block: read-ahead submission should imply no-wait as well

[Jens Axboe]

As read-ahead is opportunistic, don't block for request allocation.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/blk_types.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index ccb895f911b1..c296463c15eb 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -374,7 +374,8 @@ enum req_flag_bits {
 #define REQ_INTEGRITY		(1ULL << __REQ_INTEGRITY)
 #define REQ_FUA			(1ULL << __REQ_FUA)
 #define REQ_PREFLUSH		(1ULL << __REQ_PREFLUSH)
-#define REQ_RAHEAD		(1ULL << __REQ_RAHEAD)
+#define REQ_RAHEAD		\
+	((1ULL << __REQ_RAHEAD) | (1ULL << __REQ_NOWAIT))
 #define REQ_BACKGROUND		(1ULL << __REQ_BACKGROUND)
 #define REQ_NOWAIT		(1ULL << __REQ_NOWAIT)
 #define REQ_CGROUP_PUNT		(1ULL << __REQ_CGROUP_PUNT)
-- 
2.26.2

[PATCH 02/12] mm: allow read-ahead with IOCB_NOWAIT set

[Jens Axboe]

The read-ahead shouldn't block, so allow it to be done even if
IOCB_NOWAIT is set in the kiocb.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 mm/filemap.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 23a051a7ef0f..80747f1377d5 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2031,8 +2031,6 @@ static ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 		page = find_get_page(mapping, index);
 		if (!page) {
-			if (iocb->ki_flags & IOCB_NOWAIT)
-				goto would_block;
 			page_cache_sync_readahead(mapping,
 					ra, filp,
 					index, last_index - index);
-- 
2.26.2

[PATCH 03/12] mm: abstract out wake_page_match() from wake_page_function()

[Jens Axboe]

No functional changes in this patch, just in preparation for allowing
more callers.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/pagemap.h | 37 +++++++++++++++++++++++++++++++++++++
 mm/filemap.c            | 35 ++++-------------------------------
 2 files changed, 41 insertions(+), 31 deletions(-)

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index a8f7bd8ea1c6..53d980f2208d 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -456,6 +456,43 @@ static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
 	return pgoff;
 }
 
+/* This has the same layout as wait_bit_key - see fs/cachefiles/rdwr.c */
+struct wait_page_key {
+	struct page *page;
+	int bit_nr;
+	int page_match;
+};
+
+struct wait_page_queue {
+	struct page *page;
+	int bit_nr;
+	wait_queue_entry_t wait;
+};
+
+static inline int wake_page_match(struct wait_page_queue *wait_page,
+				  struct wait_page_key *key)
+{
+	if (wait_page->page != key->page)
+	       return 0;
+	key->page_match = 1;
+
+	if (wait_page->bit_nr != key->bit_nr)
+		return 0;
+
+	/*
+	 * Stop walking if it's locked.
+	 * Is this safe if put_and_wait_on_page_locked() is in use?
+	 * Yes: the waker must hold a reference to this page, and if PG_locked
+	 * has now already been set by another task, that task must also hold
+	 * a reference to the *same usage* of this page; so there is no need
+	 * to walk on to wake even the put_and_wait_on_page_locked() callers.
+	 */
+	if (test_bit(key->bit_nr, &key->page->flags))
+		return -1;
+
+	return 1;
+}
+
 extern void __lock_page(struct page *page);
 extern int __lock_page_killable(struct page *page);
 extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm,
diff --git a/mm/filemap.c b/mm/filemap.c
index 80747f1377d5..e891b5bee8fd 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -990,43 +990,16 @@ void __init pagecache_init(void)
 	page_writeback_init();
 }
 
-/* This has the same layout as wait_bit_key - see fs/cachefiles/rdwr.c */
-struct wait_page_key {
-	struct page *page;
-	int bit_nr;
-	int page_match;
-};
-
-struct wait_page_queue {
-	struct page *page;
-	int bit_nr;
-	wait_queue_entry_t wait;
-};
-
 static int wake_page_function(wait_queue_entry_t *wait, unsigned mode, int sync, void *arg)
 {
 	struct wait_page_key *key = arg;
 	struct wait_page_queue *wait_page
 		= container_of(wait, struct wait_page_queue, wait);
+	int ret;
 
-	if (wait_page->page != key->page)
-	       return 0;
-	key->page_match = 1;
-
-	if (wait_page->bit_nr != key->bit_nr)
-		return 0;
-
-	/*
-	 * Stop walking if it's locked.
-	 * Is this safe if put_and_wait_on_page_locked() is in use?
-	 * Yes: the waker must hold a reference to this page, and if PG_locked
-	 * has now already been set by another task, that task must also hold
-	 * a reference to the *same usage* of this page; so there is no need
-	 * to walk on to wake even the put_and_wait_on_page_locked() callers.
-	 */
-	if (test_bit(key->bit_nr, &key->page->flags))
-		return -1;
-
+	ret = wake_page_match(wait_page, key);
+	if (ret != 1)
+		return ret;
 	return autoremove_wake_function(wait, mode, sync, key);
 }
 
-- 
2.26.2

[PATCH 04/12] mm: add support for async page locking

[Jens Axboe]

Normally waiting for a page to become unlocked, or locking the page,
requires waiting for IO to complete. Add support for lock_page_async()
and wait_on_page_locked_async(), which are callback based instead. This
allows a caller to get notified when a page becomes unlocked, rather
than wait for it.

We use the iocb->private field to pass in this necessary data for this
to happen. struct wait_page_key is made public, and we define struct
wait_page_async as the interface between the caller and the core.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/fs.h      |  7 ++++++-
 include/linux/pagemap.h |  9 +++++++++
 mm/filemap.c            | 41 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 56 insertions(+), 1 deletion(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index 7e84d823c6a8..5a5434ff7543 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -314,6 +314,8 @@ enum rw_hint {
 #define IOCB_SYNC		(1 << 5)
 #define IOCB_WRITE		(1 << 6)
 #define IOCB_NOWAIT		(1 << 7)
+/* iocb->ki_waitq is valid */
+#define IOCB_WAITQ		(1 << 8)
 
 struct kiocb {
 	struct file		*ki_filp;
@@ -327,7 +329,10 @@ struct kiocb {
 	int			ki_flags;
 	u16			ki_hint;
 	u16			ki_ioprio; /* See linux/ioprio.h */
-	unsigned int		ki_cookie; /* for ->iopoll */
+	union {
+		unsigned int		ki_cookie; /* for ->iopoll */
+		struct wait_page_queue	*ki_waitq; /* for async buffered IO */
+	};
 
 	randomized_struct_fields_end
 };
diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 53d980f2208d..d3e63c9c61ae 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -495,6 +495,7 @@ static inline int wake_page_match(struct wait_page_queue *wait_page,
 
 extern void __lock_page(struct page *page);
 extern int __lock_page_killable(struct page *page);
+extern int __lock_page_async(struct page *page, struct wait_page_queue *wait);
 extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm,
 				unsigned int flags);
 extern void unlock_page(struct page *page);
@@ -531,6 +532,14 @@ static inline int lock_page_killable(struct page *page)
 	return 0;
 }
 
+static inline int lock_page_async(struct page *page,
+				  struct wait_page_queue *wait)
+{
+	if (!trylock_page(page))
+		return __lock_page_async(page, wait);
+	return 0;
+}
+
 /*
  * lock_page_or_retry - Lock the page, unless this would block and the
  * caller indicated that it can handle a retry.
diff --git a/mm/filemap.c b/mm/filemap.c
index e891b5bee8fd..c746541b1d49 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1183,6 +1183,42 @@ int wait_on_page_bit_killable(struct page *page, int bit_nr)
 }
 EXPORT_SYMBOL(wait_on_page_bit_killable);
 
+static int __wait_on_page_locked_async(struct page *page,
+				       struct wait_page_queue *wait, bool set)
+{
+	struct wait_queue_head *q = page_waitqueue(page);
+	int ret = 0;
+
+	wait->page = page;
+	wait->bit_nr = PG_locked;
+
+	spin_lock_irq(&q->lock);
+	if (set)
+		ret = !trylock_page(page);
+	else
+		ret = PageLocked(page);
+	if (ret) {
+		__add_wait_queue_entry_tail(q, &wait->wait);
+		SetPageWaiters(page);
+		if (set)
+			ret = !trylock_page(page);
+		else
+			ret = PageLocked(page);
+		/*
+		 * If we were succesful now, we know we're still on the
+		 * waitqueue as we're still under the lock. This means it's
+		 * safe to remove and return success, we know the callback
+		 * isn't going to trigger.
+		 */
+		if (!ret)
+			__remove_wait_queue(q, &wait->wait);
+		else
+			ret = -EIOCBQUEUED;
+	}
+	spin_unlock_irq(&q->lock);
+	return ret;
+}
+
 /**
  * put_and_wait_on_page_locked - Drop a reference and wait for it to be unlocked
  * @page: The page to wait for.
@@ -1345,6 +1381,11 @@ int __lock_page_killable(struct page *__page)
 }
 EXPORT_SYMBOL_GPL(__lock_page_killable);
 
+int __lock_page_async(struct page *page, struct wait_page_queue *wait)
+{
+	return __wait_on_page_locked_async(page, wait, true);
+}
+
 /*
  * Return values:
  * 1 - page is locked; mmap_sem is still held.
-- 
2.26.2

[PATCH 05/12] mm: support async buffered reads in generic_file_buffered_read()

[Jens Axboe]

Use the async page locking infrastructure, if IOCB_WAITQ is set in the
passed in iocb. The caller must expect an -EIOCBQUEUED return value,
which means that IO is started but not done yet. This is similar to how
O_DIRECT signals the same operation. Once the callback is received by
the caller for IO completion, the caller must retry the operation.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 mm/filemap.c | 33 ++++++++++++++++++++++++++-------
 1 file changed, 26 insertions(+), 7 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index c746541b1d49..18022de7dc33 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1219,6 +1219,14 @@ static int __wait_on_page_locked_async(struct page *page,
 	return ret;
 }
 
+static int wait_on_page_locked_async(struct page *page,
+				     struct wait_page_queue *wait)
+{
+	if (!PageLocked(page))
+		return 0;
+	return __wait_on_page_locked_async(compound_head(page), wait, false);
+}
+
 /**
  * put_and_wait_on_page_locked - Drop a reference and wait for it to be unlocked
  * @page: The page to wait for.
@@ -2058,17 +2066,25 @@ static ssize_t generic_file_buffered_read(struct kiocb *iocb,
 					index, last_index - index);
 		}
 		if (!PageUptodate(page)) {
-			if (iocb->ki_flags & IOCB_NOWAIT) {
-				put_page(page);
-				goto would_block;
-			}
-
 			/*
 			 * See comment in do_read_cache_page on why
 			 * wait_on_page_locked is used to avoid unnecessarily
 			 * serialisations and why it's safe.
 			 */
-			error = wait_on_page_locked_killable(page);
+			if (iocb->ki_flags & IOCB_WAITQ) {
+				if (written) {
+					put_page(page);
+					goto out;
+				}
+				error = wait_on_page_locked_async(page,
+								iocb->ki_waitq);
+			} else {
+				if (iocb->ki_flags & IOCB_NOWAIT) {
+					put_page(page);
+					goto would_block;
+				}
+				error = wait_on_page_locked_killable(page);
+			}
 			if (unlikely(error))
 				goto readpage_error;
 			if (PageUptodate(page))
@@ -2156,7 +2172,10 @@ static ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 page_not_up_to_date:
 		/* Get exclusive access to the page ... */
-		error = lock_page_killable(page);
+		if (iocb->ki_flags & IOCB_WAITQ)
+			error = lock_page_async(page, iocb->ki_waitq);
+		else
+			error = lock_page_killable(page);
 		if (unlikely(error))
 			goto readpage_error;
 
-- 
2.26.2

[PATCH 06/12] fs: add FMODE_BUF_RASYNC

[Jens Axboe]

If set, this indicates that the file system supports IOCB_WAITQ for
buffered reads.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/fs.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index 5a5434ff7543..f7b1eb765c6e 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -175,6 +175,9 @@ typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
 /* File does not contribute to nr_files count */
 #define FMODE_NOACCOUNT		((__force fmode_t)0x20000000)
 
+/* File supports async buffered reads */
+#define FMODE_BUF_RASYNC	((__force fmode_t)0x40000000)
+
 /*
  * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
  * that indicates that they should check the contents of the iovec are
-- 
2.26.2

[PATCH 07/12] ext4: flag as supporting buffered async reads

[Jens Axboe]

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/ext4/file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 0d624250a62b..9f7d9bf427b4 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -826,7 +826,7 @@ static int ext4_file_open(struct inode * inode, struct file * filp)
 			return ret;
 	}
 
-	filp->f_mode |= FMODE_NOWAIT;
+	filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC;
 	return dquot_file_open(inode, filp);
 }
 
-- 
2.26.2

[PATCH 08/12] block: flag block devices as supporting IOCB_WAITQ

[Jens Axboe]

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/block_dev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index 86e2a7134513..ec8dccc81b65 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -1851,7 +1851,7 @@ static int blkdev_open(struct inode * inode, struct file * filp)
 	 */
 	filp->f_flags |= O_LARGEFILE;
 
-	filp->f_mode |= FMODE_NOWAIT;
+	filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC;
 
 	if (filp->f_flags & O_NDELAY)
 		filp->f_mode |= FMODE_NDELAY;
-- 
2.26.2

[PATCH 09/12] xfs: flag files as supporting buffered async reads

[Jens Axboe]

XFS uses generic_file_read_iter(), which already supports this.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/xfs/xfs_file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 4b8bdecc3863..97f44fbf17f2 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1080,7 +1080,7 @@ xfs_file_open(
 		return -EFBIG;
 	if (XFS_FORCED_SHUTDOWN(XFS_M(inode->i_sb)))
 		return -EIO;
-	file->f_mode |= FMODE_NOWAIT;
+	file->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC;
 	return 0;
 }
 
-- 
2.26.2

[PATCH 10/12] btrfs: flag files as supporting buffered async reads

[Jens Axboe]

btrfs uses generic_file_read_iter(), which already supports this.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/btrfs/file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 719e68ab552c..c933b6a1b4a8 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -3480,7 +3480,7 @@ static loff_t btrfs_file_llseek(struct file *file, loff_t offset, int whence)
 
 static int btrfs_file_open(struct inode *inode, struct file *filp)
 {
-	filp->f_mode |= FMODE_NOWAIT;
+	filp->f_mode |= FMODE_NOWAIT | FMODE_BUF_RASYNC;
 	return generic_file_open(inode, filp);
 }
 
-- 
2.26.2

[PATCH 11/12] mm: add kiocb_wait_page_queue_init() helper

[Jens Axboe]

Checks if the file supports it, and initializes the values that we need.
Caller passes in 'data' pointer, if any, and the callback function to
be used.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/linux/pagemap.h | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index d3e63c9c61ae..8b65420410ee 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -493,6 +493,27 @@ static inline int wake_page_match(struct wait_page_queue *wait_page,
 	return 1;
 }
 
+static inline int kiocb_wait_page_queue_init(struct kiocb *kiocb,
+					     struct wait_page_queue *wait,
+					     wait_queue_func_t func,
+					     void *data)
+{
+	/* Can't support async wakeup with polled IO */
+	if (kiocb->ki_flags & IOCB_HIPRI)
+		return -EINVAL;
+	if (kiocb->ki_filp->f_mode & FMODE_BUF_RASYNC) {
+		wait->wait.func = func;
+		wait->wait.private = data;
+		wait->wait.flags = 0;
+		INIT_LIST_HEAD(&wait->wait.entry);
+		kiocb->ki_flags |= IOCB_WAITQ;
+		kiocb->ki_waitq = wait;
+		return 0;
+	}
+
+	return -EOPNOTSUPP;
+}
+
 extern void __lock_page(struct page *page);
 extern int __lock_page_killable(struct page *page);
 extern int __lock_page_async(struct page *page, struct wait_page_queue *wait);
-- 
2.26.2

[PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Jens Axboe]

If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
the buffered read to an io-wq worker. Instead we can rely on page
unlocking callbacks to support retry based async IO. This is a lot more
efficient than doing async thread offload.

The retry is done similarly to how we handle poll based retry. From
the unlock callback, we simply queue the retry to a task_work based
handler.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/io_uring.c | 112 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 112 insertions(+)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index e95481c552ff..23073857239c 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -498,6 +498,8 @@ struct io_async_rw {
 	struct iovec			*iov;
 	ssize_t				nr_segs;
 	ssize_t				size;
+	struct wait_page_queue		wpq;
+	struct callback_head		task_work;
 };
 
 struct io_async_ctx {
@@ -2568,6 +2570,112 @@ static int io_read_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	return 0;
 }
 
+static void io_async_buf_cancel(struct callback_head *cb)
+{
+	struct io_async_rw *rw;
+	struct io_ring_ctx *ctx;
+	struct io_kiocb *req;
+
+	rw = container_of(cb, struct io_async_rw, task_work);
+	req = rw->wpq.wait.private;
+	ctx = req->ctx;
+
+	spin_lock_irq(&ctx->completion_lock);
+	io_cqring_fill_event(req, -ECANCELED);
+	io_commit_cqring(ctx);
+	spin_unlock_irq(&ctx->completion_lock);
+
+	io_cqring_ev_posted(ctx);
+	req_set_fail_links(req);
+	io_double_put_req(req);
+}
+
+static void io_async_buf_retry(struct callback_head *cb)
+{
+	struct io_async_rw *rw;
+	struct io_ring_ctx *ctx;
+	struct io_kiocb *req;
+
+	rw = container_of(cb, struct io_async_rw, task_work);
+	req = rw->wpq.wait.private;
+	ctx = req->ctx;
+
+	__set_current_state(TASK_RUNNING);
+	mutex_lock(&ctx->uring_lock);
+	__io_queue_sqe(req, NULL);
+	mutex_unlock(&ctx->uring_lock);
+}
+
+static int io_async_buf_func(struct wait_queue_entry *wait, unsigned mode,
+			     int sync, void *arg)
+{
+	struct wait_page_queue *wpq;
+	struct io_kiocb *req = wait->private;
+	struct io_async_rw *rw = &req->io->rw;
+	struct wait_page_key *key = arg;
+	struct task_struct *tsk;
+	int ret;
+
+	wpq = container_of(wait, struct wait_page_queue, wait);
+
+	ret = wake_page_match(wpq, key);
+	if (ret != 1)
+		return ret;
+
+	list_del_init(&wait->entry);
+
+	init_task_work(&rw->task_work, io_async_buf_retry);
+	/* submit ref gets dropped, acquire a new one */
+	refcount_inc(&req->refs);
+	tsk = req->task;
+	ret = task_work_add(tsk, &rw->task_work, true);
+	if (unlikely(ret)) {
+		/* queue just for cancelation */
+		init_task_work(&rw->task_work, io_async_buf_cancel);
+		tsk = io_wq_get_task(req->ctx->io_wq);
+		task_work_add(tsk, &rw->task_work, true);
+	}
+	wake_up_process(tsk);
+	return 1;
+}
+
+static bool io_rw_should_retry(struct io_kiocb *req)
+{
+	struct kiocb *kiocb = &req->rw.kiocb;
+	int ret;
+
+	/* never retry for NOWAIT, we just complete with -EAGAIN */
+	if (req->flags & REQ_F_NOWAIT)
+		return false;
+
+	/* already tried, or we're doing O_DIRECT */
+	if (kiocb->ki_flags & (IOCB_DIRECT | IOCB_WAITQ))
+		return false;
+	/*
+	 * just use poll if we can, and don't attempt if the fs doesn't
+	 * support callback based unlocks
+	 */
+	if (file_can_poll(req->file) || !(req->file->f_mode & FMODE_BUF_RASYNC))
+		return false;
+
+	/*
+	 * If request type doesn't require req->io to defer in general,
+	 * we need to allocate it here
+	 */
+	if (!req->io && __io_alloc_async_ctx(req))
+		return false;
+
+	ret = kiocb_wait_page_queue_init(kiocb, &req->io->rw.wpq,
+						io_async_buf_func, req);
+	if (!ret) {
+		get_task_struct(current);
+		req->task = current;
+		return true;
+	}
+
+	return false;
+}
+
 static int io_read(struct io_kiocb *req, bool force_nonblock)
 {
 	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
@@ -2601,6 +2709,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 	if (!ret) {
 		ssize_t ret2;
 
+retry:
 		if (req->file->f_op->read_iter)
 			ret2 = call_read_iter(req->file, kiocb, &iter);
 		else
@@ -2619,6 +2728,9 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 			if (!(req->flags & REQ_F_NOWAIT) &&
 			    !file_can_poll(req->file))
 				req->flags |= REQ_F_MUST_PUNT;
+			if (io_rw_should_retry(req))
+				goto retry;
+			kiocb->ki_flags &= ~IOCB_WAITQ;
 			return -EAGAIN;
 		}
 	}
-- 
2.26.2

Re: [PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Pavel Begunkov]

On 23/05/2020 21:57, Jens Axboe wrote:
> If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
> the buffered read to an io-wq worker. Instead we can rely on page
> unlocking callbacks to support retry based async IO. This is a lot more
> efficient than doing async thread offload.
> 
> The retry is done similarly to how we handle poll based retry. From
> the unlock callback, we simply queue the retry to a task_work based
> handler.
> 
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> ---
>  fs/io_uring.c | 99 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 99 insertions(+)
> 
...
> +
> +	init_task_work(&rw->task_work, io_async_buf_retry);
> +	/* submit ref gets dropped, acquire a new one */
> +	refcount_inc(&req->refs);
> +	tsk = req->task;
> +	ret = task_work_add(tsk, &rw->task_work, true);
> +	if (unlikely(ret)) {
> +		/* queue just for cancelation */
> +		init_task_work(&rw->task_work, io_async_buf_cancel);
> +		tsk = io_wq_get_task(req->ctx->io_wq);

IIRC, task will be put somewhere around io_free_req(). Then shouldn't here be
some juggling with reassigning req->task with task_{get,put}()?

> +		task_work_add(tsk, &rw->task_work, true);
> +	}
> +	wake_up_process(tsk);
> +	return 1;
> +}
...
>  static int io_read(struct io_kiocb *req, bool force_nonblock)
>  {
>  	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
> @@ -2601,6 +2696,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>  	if (!ret) {
>  		ssize_t ret2;
>  
> +retry:
>  		if (req->file->f_op->read_iter)
>  			ret2 = call_read_iter(req->file, kiocb, &iter);
>  		else
> @@ -2619,6 +2715,9 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>  			if (!(req->flags & REQ_F_NOWAIT) &&
>  			    !file_can_poll(req->file))
>  				req->flags |= REQ_F_MUST_PUNT;
> +			if (io_rw_should_retry(req))

It looks like a state machine with IOCB_WAITQ and gotos. Wouldn't it be cleaner
to call call_read_iter()/loop_rw_iter() here directly instead of "goto retry" ?

BTW, can this async stuff return -EAGAIN ?

> +				goto retry;
> +			kiocb->ki_flags &= ~IOCB_WAITQ;
>  			return -EAGAIN;
>  		}
>  	}
> 

-- 
Pavel Begunkov

Re: [PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Jens Axboe]

On 5/25/20 1:29 AM, Pavel Begunkov wrote:
> On 23/05/2020 21:57, Jens Axboe wrote:
>> If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
>> the buffered read to an io-wq worker. Instead we can rely on page
>> unlocking callbacks to support retry based async IO. This is a lot more
>> efficient than doing async thread offload.
>>
>> The retry is done similarly to how we handle poll based retry. From
>> the unlock callback, we simply queue the retry to a task_work based
>> handler.
>>
>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
>> ---
>>  fs/io_uring.c | 99 +++++++++++++++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 99 insertions(+)
>>
> ...
>> +
>> +	init_task_work(&rw->task_work, io_async_buf_retry);
>> +	/* submit ref gets dropped, acquire a new one */
>> +	refcount_inc(&req->refs);
>> +	tsk = req->task;
>> +	ret = task_work_add(tsk, &rw->task_work, true);
>> +	if (unlikely(ret)) {
>> +		/* queue just for cancelation */
>> +		init_task_work(&rw->task_work, io_async_buf_cancel);
>> +		tsk = io_wq_get_task(req->ctx->io_wq);
> 
> IIRC, task will be put somewhere around io_free_req(). Then shouldn't here be
> some juggling with reassigning req->task with task_{get,put}()?

Not sure I follow? Yes, we'll put this task again when the request
is freed, but not sure what you mean with juggling?

>> +		task_work_add(tsk, &rw->task_work, true);
>> +	}
>> +	wake_up_process(tsk);
>> +	return 1;
>> +}
> ...
>>  static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  {
>>  	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
>> @@ -2601,6 +2696,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  	if (!ret) {
>>  		ssize_t ret2;
>>  
>> +retry:
>>  		if (req->file->f_op->read_iter)
>>  			ret2 = call_read_iter(req->file, kiocb, &iter);
>>  		else
>> @@ -2619,6 +2715,9 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  			if (!(req->flags & REQ_F_NOWAIT) &&
>>  			    !file_can_poll(req->file))
>>  				req->flags |= REQ_F_MUST_PUNT;
>> +			if (io_rw_should_retry(req))
> 
> It looks like a state machine with IOCB_WAITQ and gotos. Wouldn't it be cleaner
> to call call_read_iter()/loop_rw_iter() here directly instead of "goto retry" ?

We could, probably making that part a separate helper then. How about the
below incremental?

> BTW, can this async stuff return -EAGAIN ?

Probably? Prefer not to make any definitive calls on that being possible or
not, as it's sure to disappoint. If it does and IOCB_WAITQ is already set,
then we'll punt to a thread like before.


diff --git a/fs/io_uring.c b/fs/io_uring.c
index a5a4d9602915..669dccd81207 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -2677,6 +2677,13 @@ static bool io_rw_should_retry(struct io_kiocb *req)
 	return false;
 }
 
+static int __io_read(struct io_kiocb *req, struct iov_iter *iter)
+{
+	if (req->file->f_op->read_iter)
+		return call_read_iter(req->file, &req->rw.kiocb, iter);
+	return loop_rw_iter(READ, req->file, &req->rw.kiocb, iter);
+}
+
 static int io_read(struct io_kiocb *req, bool force_nonblock)
 {
 	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
@@ -2710,11 +2717,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 	if (!ret) {
 		ssize_t ret2;
 
-retry:
-		if (req->file->f_op->read_iter)
-			ret2 = call_read_iter(req->file, kiocb, &iter);
-		else
-			ret2 = loop_rw_iter(READ, req->file, kiocb, &iter);
+		ret2 = __io_read(req, &iter);
 
 		/* Catch -EAGAIN return for forced non-blocking submission */
 		if (!force_nonblock || ret2 != -EAGAIN) {
@@ -2729,8 +2732,11 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 			if (!(req->flags & REQ_F_NOWAIT) &&
 			    !file_can_poll(req->file))
 				req->flags |= REQ_F_MUST_PUNT;
-			if (io_rw_should_retry(req))
-				goto retry;
+			if (io_rw_should_retry(req)) {
+				ret2 = __io_read(req, &iter);
+				if (ret2 != -EAGAIN)
+					goto out_free;
+			}
 			kiocb->ki_flags &= ~IOCB_WAITQ;
 			return -EAGAIN;
 		}

-- 
Jens Axboe

Re: [PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Pavel Begunkov]

On 23/05/2020 21:57, Jens Axboe wrote:
> If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
> the buffered read to an io-wq worker. Instead we can rely on page
> unlocking callbacks to support retry based async IO. This is a lot more
> efficient than doing async thread offload.
> 
> The retry is done similarly to how we handle poll based retry. From
> the unlock callback, we simply queue the retry to a task_work based
> handler.
> 
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> ---
>  fs/io_uring.c | 99 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 99 insertions(+)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index e95481c552ff..dd532d2634c2 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -498,6 +498,8 @@ struct io_async_rw {
>  	struct iovec			*iov;
>  	ssize_t				nr_segs;
>  	ssize_t				size;
> +	struct wait_page_queue		wpq;
> +	struct callback_head		task_work;
>  };
>  
>  struct io_async_ctx {
> @@ -2568,6 +2570,99 @@ static int io_read_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe,
>  	return 0;
>  }
>  
> +static void io_async_buf_cancel(struct callback_head *cb)
> +{
> +	struct io_async_rw *rw;
> +	struct io_ring_ctx *ctx;
> +	struct io_kiocb *req;
> +
> +	rw = container_of(cb, struct io_async_rw, task_work);
> +	req = rw->wpq.wait.private;
> +	ctx = req->ctx;
> +
> +	spin_lock_irq(&ctx->completion_lock);
> +	io_cqring_fill_event(req, -ECANCELED);

It seems like it should go through kiocb_done()/io_complete_rw_common().
My concern is missing io_put_kbuf().

> +	io_commit_cqring(ctx);
> +	spin_unlock_irq(&ctx->completion_lock);
> +
> +	io_cqring_ev_posted(ctx);
> +	req_set_fail_links(req);
> +	io_double_put_req(req);
> +}


-- 
Pavel Begunkov

Re: [PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Pavel Begunkov]

On 25/05/2020 22:59, Jens Axboe wrote:
> On 5/25/20 1:29 AM, Pavel Begunkov wrote:
>> On 23/05/2020 21:57, Jens Axboe wrote:
>>> If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
>>> the buffered read to an io-wq worker. Instead we can rely on page
>>> unlocking callbacks to support retry based async IO. This is a lot more
>>> efficient than doing async thread offload.
>>>
>>> The retry is done similarly to how we handle poll based retry. From
>>> the unlock callback, we simply queue the retry to a task_work based
>>> handler.
>>>
>>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
>>> ---
>>>  fs/io_uring.c | 99 +++++++++++++++++++++++++++++++++++++++++++++++++++
>>>  1 file changed, 99 insertions(+)
>>>
>> ...
>>> +
>>> +	init_task_work(&rw->task_work, io_async_buf_retry);
>>> +	/* submit ref gets dropped, acquire a new one */
>>> +	refcount_inc(&req->refs);
>>> +	tsk = req->task;
>>> +	ret = task_work_add(tsk, &rw->task_work, true);
>>> +	if (unlikely(ret)) {
>>> +		/* queue just for cancelation */
>>> +		init_task_work(&rw->task_work, io_async_buf_cancel);
>>> +		tsk = io_wq_get_task(req->ctx->io_wq);
>>
>> IIRC, task will be put somewhere around io_free_req(). Then shouldn't here be
>> some juggling with reassigning req->task with task_{get,put}()?
> 
> Not sure I follow? Yes, we'll put this task again when the request
> is freed, but not sure what you mean with juggling?

I meant something like:

...
/* queue just for cancelation */
init_task_work(&rw->task_work, io_async_buf_cancel);
+ put_task_struct(req->task);
+ req->task = get_task_struct(io_wq_task);


but, thinking twice, if I got the whole idea right, it should be ok as is --
io-wq won't go away before the request anyway, and leaving req->task pinned down
for a bit is not a problem.

>>> +		task_work_add(tsk, &rw->task_work, true);
>>> +	}
>>> +	wake_up_process(tsk);
>>> +	return 1;
>>> +}
>> ...
>>>  static int io_read(struct io_kiocb *req, bool force_nonblock)
>>>  {
>>>  	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
>>> @@ -2601,6 +2696,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>>  	if (!ret) {
>>>  		ssize_t ret2;
>>>  
>>> +retry:
>>>  		if (req->file->f_op->read_iter)
>>>  			ret2 = call_read_iter(req->file, kiocb, &iter);
>>>  		else
>>> @@ -2619,6 +2715,9 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>>  			if (!(req->flags & REQ_F_NOWAIT) &&
>>>  			    !file_can_poll(req->file))
>>>  				req->flags |= REQ_F_MUST_PUNT;
>>> +			if (io_rw_should_retry(req))
>>
>> It looks like a state machine with IOCB_WAITQ and gotos. Wouldn't it be cleaner
>> to call call_read_iter()/loop_rw_iter() here directly instead of "goto retry" ?
> 
> We could, probably making that part a separate helper then. How about the
> below incremental?

IMHO, it was easy to get lost with such implicit state switching.
Looks better now! See a small comment below.

> 
>> BTW, can this async stuff return -EAGAIN ?
> 
> Probably? Prefer not to make any definitive calls on that being possible or
> not, as it's sure to disappoint. If it does and IOCB_WAITQ is already set,
> then we'll punt to a thread like before.

Sounds reasonable

> 
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index a5a4d9602915..669dccd81207 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -2677,6 +2677,13 @@ static bool io_rw_should_retry(struct io_kiocb *req)
>  	return false;
>  }
>  
> +static int __io_read(struct io_kiocb *req, struct iov_iter *iter)
> +{
> +	if (req->file->f_op->read_iter)
> +		return call_read_iter(req->file, &req->rw.kiocb, iter);
> +	return loop_rw_iter(READ, req->file, &req->rw.kiocb, iter);
> +}
> +
>  static int io_read(struct io_kiocb *req, bool force_nonblock)
>  {
>  	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
> @@ -2710,11 +2717,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>  	if (!ret) {
>  		ssize_t ret2;
>  
> -retry:
> -		if (req->file->f_op->read_iter)
> -			ret2 = call_read_iter(req->file, kiocb, &iter);
> -		else
> -			ret2 = loop_rw_iter(READ, req->file, kiocb, &iter);
> +		ret2 = __io_read(req, &iter);
>  
>  		/* Catch -EAGAIN return for forced non-blocking submission */
>  		if (!force_nonblock || ret2 != -EAGAIN) {
> @@ -2729,8 +2732,11 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>  			if (!(req->flags & REQ_F_NOWAIT) &&
>  			    !file_can_poll(req->file))
>  				req->flags |= REQ_F_MUST_PUNT;
> -			if (io_rw_should_retry(req))
> -				goto retry;
> +			if (io_rw_should_retry(req)) {
> +				ret2 = __io_read(req, &iter);
> +				if (ret2 != -EAGAIN)
> +					goto out_free;

"goto out_free" returns ret=0, so someone should add a cqe

if (ret2 != -EAGAIN) {
	kiocb_done(kiocb, ret2);
	goto free_out;
}


> +			}
>  			kiocb->ki_flags &= ~IOCB_WAITQ;
>  			return -EAGAIN;
>  		}
> 

-- 
Pavel Begunkov

Re: [PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Jens Axboe]

On 5/26/20 1:38 AM, Pavel Begunkov wrote:
> On 23/05/2020 21:57, Jens Axboe wrote:
>> If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
>> the buffered read to an io-wq worker. Instead we can rely on page
>> unlocking callbacks to support retry based async IO. This is a lot more
>> efficient than doing async thread offload.
>>
>> The retry is done similarly to how we handle poll based retry. From
>> the unlock callback, we simply queue the retry to a task_work based
>> handler.
>>
>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
>> ---
>>  fs/io_uring.c | 99 +++++++++++++++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 99 insertions(+)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index e95481c552ff..dd532d2634c2 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -498,6 +498,8 @@ struct io_async_rw {
>>  	struct iovec			*iov;
>>  	ssize_t				nr_segs;
>>  	ssize_t				size;
>> +	struct wait_page_queue		wpq;
>> +	struct callback_head		task_work;
>>  };
>>  
>>  struct io_async_ctx {
>> @@ -2568,6 +2570,99 @@ static int io_read_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe,
>>  	return 0;
>>  }
>>  
>> +static void io_async_buf_cancel(struct callback_head *cb)
>> +{
>> +	struct io_async_rw *rw;
>> +	struct io_ring_ctx *ctx;
>> +	struct io_kiocb *req;
>> +
>> +	rw = container_of(cb, struct io_async_rw, task_work);
>> +	req = rw->wpq.wait.private;
>> +	ctx = req->ctx;
>> +
>> +	spin_lock_irq(&ctx->completion_lock);
>> +	io_cqring_fill_event(req, -ECANCELED);
> 
> It seems like it should go through kiocb_done()/io_complete_rw_common().
> My concern is missing io_put_kbuf().

Yeah, I noticed that too after sending it out. If you look at the
current one that I updated yesterday, it does add that (and also
renames the iter read helper):

https://git.kernel.dk/cgit/linux-block/commit/?h=async-buffered.5&id=6f4e3a4066d0db3e3478e58cc250afb16d8d4d91

-- 
Jens Axboe

Re: [PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Jens Axboe]

On 5/26/20 1:44 AM, Pavel Begunkov wrote:
> On 25/05/2020 22:59, Jens Axboe wrote:
>> On 5/25/20 1:29 AM, Pavel Begunkov wrote:
>>> On 23/05/2020 21:57, Jens Axboe wrote:
>>>> If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
>>>> the buffered read to an io-wq worker. Instead we can rely on page
>>>> unlocking callbacks to support retry based async IO. This is a lot more
>>>> efficient than doing async thread offload.
>>>>
>>>> The retry is done similarly to how we handle poll based retry. From
>>>> the unlock callback, we simply queue the retry to a task_work based
>>>> handler.
>>>>
>>>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
>>>> ---
>>>>  fs/io_uring.c | 99 +++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>  1 file changed, 99 insertions(+)
>>>>
>>> ...
>>>> +
>>>> +	init_task_work(&rw->task_work, io_async_buf_retry);
>>>> +	/* submit ref gets dropped, acquire a new one */
>>>> +	refcount_inc(&req->refs);
>>>> +	tsk = req->task;
>>>> +	ret = task_work_add(tsk, &rw->task_work, true);
>>>> +	if (unlikely(ret)) {
>>>> +		/* queue just for cancelation */
>>>> +		init_task_work(&rw->task_work, io_async_buf_cancel);
>>>> +		tsk = io_wq_get_task(req->ctx->io_wq);
>>>
>>> IIRC, task will be put somewhere around io_free_req(). Then shouldn't here be
>>> some juggling with reassigning req->task with task_{get,put}()?
>>
>> Not sure I follow? Yes, we'll put this task again when the request
>> is freed, but not sure what you mean with juggling?
> 
> I meant something like:
> 
> ...
> /* queue just for cancelation */
> init_task_work(&rw->task_work, io_async_buf_cancel);
> + put_task_struct(req->task);
> + req->task = get_task_struct(io_wq_task);
> 
> 
> but, thinking twice, if I got the whole idea right, it should be ok as
> is -- io-wq won't go away before the request anyway, and leaving
> req->task pinned down for a bit is not a problem.

OK good, then I thin kwe agree it's fine.

>>>> +		task_work_add(tsk, &rw->task_work, true);
>>>> +	}
>>>> +	wake_up_process(tsk);
>>>> +	return 1;
>>>> +}
>>> ...
>>>>  static int io_read(struct io_kiocb *req, bool force_nonblock)
>>>>  {
>>>>  	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
>>>> @@ -2601,6 +2696,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>>>  	if (!ret) {
>>>>  		ssize_t ret2;
>>>>  
>>>> +retry:
>>>>  		if (req->file->f_op->read_iter)
>>>>  			ret2 = call_read_iter(req->file, kiocb, &iter);
>>>>  		else
>>>> @@ -2619,6 +2715,9 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>>>  			if (!(req->flags & REQ_F_NOWAIT) &&
>>>>  			    !file_can_poll(req->file))
>>>>  				req->flags |= REQ_F_MUST_PUNT;
>>>> +			if (io_rw_should_retry(req))
>>>
>>> It looks like a state machine with IOCB_WAITQ and gotos. Wouldn't it be cleaner
>>> to call call_read_iter()/loop_rw_iter() here directly instead of "goto retry" ?
>>
>> We could, probably making that part a separate helper then. How about the
>> below incremental?
> 
> IMHO, it was easy to get lost with such implicit state switching.
> Looks better now! See a small comment below.

Agree, that is cleaner.

>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index a5a4d9602915..669dccd81207 100644
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -2677,6 +2677,13 @@ static bool io_rw_should_retry(struct io_kiocb *req)
>>  	return false;
>>  }
>>  
>> +static int __io_read(struct io_kiocb *req, struct iov_iter *iter)
>> +{
>> +	if (req->file->f_op->read_iter)
>> +		return call_read_iter(req->file, &req->rw.kiocb, iter);
>> +	return loop_rw_iter(READ, req->file, &req->rw.kiocb, iter);
>> +}
>> +
>>  static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  {
>>  	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
>> @@ -2710,11 +2717,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  	if (!ret) {
>>  		ssize_t ret2;
>>  
>> -retry:
>> -		if (req->file->f_op->read_iter)
>> -			ret2 = call_read_iter(req->file, kiocb, &iter);
>> -		else
>> -			ret2 = loop_rw_iter(READ, req->file, kiocb, &iter);
>> +		ret2 = __io_read(req, &iter);
>>  
>>  		/* Catch -EAGAIN return for forced non-blocking submission */
>>  		if (!force_nonblock || ret2 != -EAGAIN) {
>> @@ -2729,8 +2732,11 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
>>  			if (!(req->flags & REQ_F_NOWAIT) &&
>>  			    !file_can_poll(req->file))
>>  				req->flags |= REQ_F_MUST_PUNT;
>> -			if (io_rw_should_retry(req))
>> -				goto retry;
>> +			if (io_rw_should_retry(req)) {
>> +				ret2 = __io_read(req, &iter);
>> +				if (ret2 != -EAGAIN)
>> +					goto out_free;
> 
> "goto out_free" returns ret=0, so someone should add a cqe
> 
> if (ret2 != -EAGAIN) {
> 	kiocb_done(kiocb, ret2);
> 	goto free_out;
> }

Fixed up in the current one.

-- 
Jens Axboe

[PATCH 12/12] io_uring: support true async buffered reads, if file provides it

[Jens Axboe]

If the file is flagged with FMODE_BUF_RASYNC, then we don't have to punt
the buffered read to an io-wq worker. Instead we can rely on page
unlocking callbacks to support retry based async IO. This is a lot more
efficient than doing async thread offload.

The retry is done similarly to how we handle poll based retry. From
the unlock callback, we simply queue the retry to a task_work based
handler.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/io_uring.c | 130 ++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 126 insertions(+), 4 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index e6865afa8467..95df63b0b2ce 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -79,6 +79,7 @@
 #include <linux/fs_struct.h>
 #include <linux/splice.h>
 #include <linux/task_work.h>
+#include <linux/pagemap.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/io_uring.h>
@@ -498,6 +499,8 @@ struct io_async_rw {
 	struct iovec			*iov;
 	ssize_t				nr_segs;
 	ssize_t				size;
+	struct wait_page_queue		wpq;
+	struct callback_head		task_work;
 };
 
 struct io_async_ctx {
@@ -2568,6 +2571,119 @@ static int io_read_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	return 0;
 }
 
+static void io_async_buf_cancel(struct callback_head *cb)
+{
+	struct io_async_rw *rw;
+	struct io_ring_ctx *ctx;
+	struct io_kiocb *req;
+
+	rw = container_of(cb, struct io_async_rw, task_work);
+	req = rw->wpq.wait.private;
+	ctx = req->ctx;
+
+	spin_lock_irq(&ctx->completion_lock);
+	io_cqring_fill_event(req, -ECANCELED);
+	io_commit_cqring(ctx);
+	spin_unlock_irq(&ctx->completion_lock);
+
+	io_cqring_ev_posted(ctx);
+	req_set_fail_links(req);
+	io_double_put_req(req);
+}
+
+static void io_async_buf_retry(struct callback_head *cb)
+{
+	struct io_async_rw *rw;
+	struct io_ring_ctx *ctx;
+	struct io_kiocb *req;
+
+	rw = container_of(cb, struct io_async_rw, task_work);
+	req = rw->wpq.wait.private;
+	ctx = req->ctx;
+
+	__set_current_state(TASK_RUNNING);
+	mutex_lock(&ctx->uring_lock);
+	__io_queue_sqe(req, NULL);
+	mutex_unlock(&ctx->uring_lock);
+}
+
+static int io_async_buf_func(struct wait_queue_entry *wait, unsigned mode,
+			     int sync, void *arg)
+{
+	struct wait_page_queue *wpq;
+	struct io_kiocb *req = wait->private;
+	struct io_async_rw *rw = &req->io->rw;
+	struct wait_page_key *key = arg;
+	struct task_struct *tsk;
+	int ret;
+
+	wpq = container_of(wait, struct wait_page_queue, wait);
+
+	ret = wake_page_match(wpq, key);
+	if (ret != 1)
+		return ret;
+
+	list_del_init(&wait->entry);
+
+	init_task_work(&rw->task_work, io_async_buf_retry);
+	/* submit ref gets dropped, acquire a new one */
+	refcount_inc(&req->refs);
+	tsk = req->task;
+	ret = task_work_add(tsk, &rw->task_work, true);
+	if (unlikely(ret)) {
+		/* queue just for cancelation */
+		init_task_work(&rw->task_work, io_async_buf_cancel);
+		tsk = io_wq_get_task(req->ctx->io_wq);
+		task_work_add(tsk, &rw->task_work, true);
+	}
+	wake_up_process(tsk);
+	return 1;
+}
+
+static bool io_rw_should_retry(struct io_kiocb *req)
+{
+	struct kiocb *kiocb = &req->rw.kiocb;
+	int ret;
+
+	/* never retry for NOWAIT, we just complete with -EAGAIN */
+	if (req->flags & REQ_F_NOWAIT)
+		return false;
+
+	/* already tried, or we're doing O_DIRECT */
+	if (kiocb->ki_flags & (IOCB_DIRECT | IOCB_WAITQ))
+		return false;
+	/*
+	 * just use poll if we can, and don't attempt if the fs doesn't
+	 * support callback based unlocks
+	 */
+	if (file_can_poll(req->file) || !(req->file->f_mode & FMODE_BUF_RASYNC))
+		return false;
+
+	/*
+	 * If request type doesn't require req->io to defer in general,
+	 * we need to allocate it here
+	 */
+	if (!req->io && __io_alloc_async_ctx(req))
+		return false;
+
+	ret = kiocb_wait_page_queue_init(kiocb, &req->io->rw.wpq,
+						io_async_buf_func, req);
+	if (!ret) {
+		get_task_struct(current);
+		req->task = current;
+		return true;
+	}
+
+	return false;
+}
+
+static int io_iter_do_read(struct io_kiocb *req, struct iov_iter *iter)
+{
+	if (req->file->f_op->read_iter)
+		return call_read_iter(req->file, &req->rw.kiocb, iter);
+	return loop_rw_iter(READ, req->file, &req->rw.kiocb, iter);
+}
+
 static int io_read(struct io_kiocb *req, bool force_nonblock)
 {
 	struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
@@ -2601,10 +2717,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 	if (!ret) {
 		ssize_t ret2;
 
-		if (req->file->f_op->read_iter)
-			ret2 = call_read_iter(req->file, kiocb, &iter);
-		else
-			ret2 = loop_rw_iter(READ, req->file, kiocb, &iter);
+		ret2 = io_iter_do_read(req, &iter);
 
 		/* Catch -EAGAIN return for forced non-blocking submission */
 		if (!force_nonblock || ret2 != -EAGAIN) {
@@ -2619,6 +2732,15 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 			if (!(req->flags & REQ_F_NOWAIT) &&
 			    !file_can_poll(req->file))
 				req->flags |= REQ_F_MUST_PUNT;
+			/* if we can retry, do so with the callbacks armed */
+			if (io_rw_should_retry(req)) {
+				ret2 = io_iter_do_read(req, &iter);
+				if (ret2 != -EAGAIN) {
+					kiocb_done(kiocb, ret2);
+					goto out_free;
+				}
+			}
+			kiocb->ki_flags &= ~IOCB_WAITQ;
 			return -EAGAIN;
 		}
 	}
-- 
2.26.2