public inbox for [email protected]
 help / color / mirror / Atom feed
From: Jens Axboe <[email protected]>
To: [email protected]
Cc: [email protected], [email protected],
	[email protected], [email protected],
	Jens Axboe <[email protected]>
Subject: [PATCH 15/15] io_uring: support true async buffered reads, if file provides it
Date: Thu, 18 Jun 2020 08:43:55 -0600	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

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 <[email protected]>
---
 fs/io_uring.c | 145 +++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 137 insertions(+), 8 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 40413fb9d07b..94282be1c413 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -78,6 +78,7 @@
 #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>
@@ -503,6 +504,8 @@ 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 {
@@ -2750,6 +2753,126 @@ static int io_read_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	return 0;
 }
 
+static void __io_async_buf_error(struct io_kiocb *req, int error)
+{
+	struct io_ring_ctx *ctx = req->ctx;
+
+	spin_lock_irq(&ctx->completion_lock);
+	io_cqring_fill_event(req, error);
+	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_cancel(struct callback_head *cb)
+{
+	struct io_async_rw *rw;
+	struct io_kiocb *req;
+
+	rw = container_of(cb, struct io_async_rw, task_work);
+	req = rw->wpq.wait.private;
+	__io_async_buf_error(req, -ECANCELED);
+}
+
+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);
+	if (!io_sq_thread_acquire_mm(ctx, req)) {
+		mutex_lock(&ctx->uring_lock);
+		__io_queue_sqe(req, NULL);
+		mutex_unlock(&ctx->uring_lock);
+	} else {
+		__io_async_buf_error(req, -EFAULT);
+	}
+}
+
+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) {
+		io_get_req_task(req);
+		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;
@@ -2784,10 +2907,7 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 		unsigned long nr_segs = iter.nr_segs;
 		ssize_t ret2 = 0;
 
-		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 && ret2 != -EIO)) {
@@ -2799,17 +2919,26 @@ static int io_read(struct io_kiocb *req, bool force_nonblock)
 			ret = io_setup_async_rw(req, io_size, iovec,
 						inline_vecs, &iter);
 			if (ret)
-				goto out_free;
+				goto out;
 			/* any defer here is final, must blocking retry */
 			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 == -EIOCBQUEUED) {
+					goto out;
+				} else if (ret2 != -EAGAIN) {
+					kiocb_done(kiocb, ret2);
+					goto out;
+				}
+			}
+			kiocb->ki_flags &= ~IOCB_WAITQ;
 			return -EAGAIN;
 		}
 	}
-out_free:
-	kfree(iovec);
-	req->flags &= ~REQ_F_NEED_CLEANUP;
+out:
 	return ret;
 }
 
-- 
2.27.0


  parent reply	other threads:[~2020-06-18 14:45 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-06-18 14:43 Jens Axboe
2020-06-18 14:43 ` [PATCH 01/15] block: provide plug based way of signaling forced no-wait semantics Jens Axboe
2020-06-18 14:43 ` [PATCH 02/15] io_uring: always plug for any number of IOs Jens Axboe
2020-06-18 14:43 ` [PATCH 03/15] io_uring: catch -EIO from buffered issue request failure Jens Axboe
2020-06-18 14:43 ` [PATCH 04/15] io_uring: re-issue block requests that failed because of resources Jens Axboe
2020-06-19 14:12   ` Pavel Begunkov
2020-06-19 14:22     ` Jens Axboe
2020-06-19 14:30       ` Pavel Begunkov
2020-06-19 14:36         ` Jens Axboe
2020-06-18 14:43 ` [PATCH 05/15] mm: allow read-ahead with IOCB_NOWAIT set Jens Axboe
2020-06-24  1:02   ` Dave Chinner
2020-06-24  1:46     ` Matthew Wilcox
2020-06-24 15:00       ` Jens Axboe
2020-06-24 15:35         ` Jens Axboe
2020-06-24 16:41           ` Matthew Wilcox
2020-06-24 16:44             ` Jens Axboe
2020-07-07 11:38               ` Andreas Grünbacher
2020-07-07 14:31                 ` Jens Axboe
2020-08-10 22:56               ` Dave Chinner
2020-08-10 23:03                 ` Jens Axboe
2020-06-24  4:38   ` Dave Chinner
2020-06-24 15:01     ` Jens Axboe
2020-06-18 14:43 ` [PATCH 06/15] mm: abstract out wake_page_match() from wake_page_function() Jens Axboe
2020-06-18 14:43 ` [PATCH 07/15] mm: add support for async page locking Jens Axboe
2020-07-07 11:32   ` Andreas Grünbacher
2020-07-07 14:32     ` Jens Axboe
2020-06-18 14:43 ` [PATCH 08/15] mm: support async buffered reads in generic_file_buffered_read() Jens Axboe
2020-06-18 14:43 ` [PATCH 09/15] fs: add FMODE_BUF_RASYNC Jens Axboe
2020-06-18 14:43 ` [PATCH 10/15] block: flag block devices as supporting IOCB_WAITQ Jens Axboe
2020-06-18 14:43 ` [PATCH 11/15] xfs: flag files as supporting buffered async reads Jens Axboe
2020-06-18 14:43 ` [PATCH 12/15] btrfs: " Jens Axboe
2020-06-19 11:11   ` David Sterba
2020-06-18 14:43 ` [PATCH 13/15] ext4: flag " Jens Axboe
2020-06-18 14:43 ` [PATCH 14/15] mm: add kiocb_wait_page_queue_init() helper Jens Axboe
2020-06-18 14:43 ` Jens Axboe [this message]
2020-06-23 12:39   ` [PATCH 15/15] io_uring: support true async buffered reads, if file provides it Pavel Begunkov
2020-06-23 14:38     ` Jens Axboe
2020-06-18 14:45 ` [PATCHSET v7 0/12] Add support for async buffered reads Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    [email protected] \
    [email protected] \
    --cc="[PATCHSET v7 0/15]"@vger.kernel.org \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox