From: "Stefan Bühler" <source@stbuehler.de>
To: Jens Axboe <axboe@kernel.dk>,
linux-block@vger.kernel.org, linux-fsdevel@vger.kernel.org
Subject: [PATCH 1/2] io_uring: restructure io_{read,write} control flow
Date: Fri, 3 May 2019 11:47:14 +0200 [thread overview]
Message-ID: <20190503094715.2381-1-source@stbuehler.de> (raw)
In-Reply-To: <37071226-375a-07a6-d3d3-21323145de71@kernel.dk>
Call io_async_list_note at the end if -EAGAIN is going to be returned;
we need iov_count for that, which we have (almost) at the same time as
we need to free iovec.
Instead of using a second return value reset the normal one after
passing it to io_rw_done.
Unless rw_verify_area returns -EAGAIN this shouldn't result in different
behavior.
This change should make it easier to punt a request to the workers by
returning -EAGAIN and still calling io_async_list_note if needed.
Signed-off-by: Stefan Bühler <source@stbuehler.de>
---
fs/io_uring.c | 89 ++++++++++++++++++++++-----------------------------
1 file changed, 39 insertions(+), 50 deletions(-)
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 84efb8956734..52e435a72b6f 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -1062,26 +1062,24 @@ static int io_read(struct io_kiocb *req, const struct sqe_submit *s,
ret = io_import_iovec(req->ctx, READ, s, &iovec, &iter);
if (ret)
return ret;
-
iov_count = iov_iter_count(&iter);
+
ret = rw_verify_area(READ, file, &kiocb->ki_pos, iov_count);
- if (!ret) {
- ssize_t ret2;
+ if (ret)
+ goto out_free;
- /* Catch -EAGAIN return for forced non-blocking submission */
- ret2 = call_read_iter(file, kiocb, &iter);
- if (!force_nonblock || ret2 != -EAGAIN) {
- io_rw_done(kiocb, ret2);
- } else {
- /*
- * If ->needs_lock is true, we're already in async
- * context.
- */
- if (!s->needs_lock)
- io_async_list_note(READ, req, iov_count);
- ret = -EAGAIN;
- }
+ /* Passthrough -EAGAIN return for forced non-blocking submission */
+ ret = call_read_iter(file, kiocb, &iter);
+ if (!(force_nonblock && ret == -EAGAIN)) {
+ io_rw_done(kiocb, ret);
+ ret = 0;
}
+
+out_free:
+ /* If ->needs_lock is true, we're already in async context. */
+ if (ret == -EAGAIN && !s->needs_lock)
+ io_async_list_note(READ, req, iov_count);
+
kfree(iovec);
return ret;
}
@@ -1109,50 +1107,41 @@ static int io_write(struct io_kiocb *req, const struct sqe_submit *s,
ret = io_import_iovec(req->ctx, WRITE, s, &iovec, &iter);
if (ret)
return ret;
-
iov_count = iov_iter_count(&iter);
ret = -EAGAIN;
- if (force_nonblock && !(kiocb->ki_flags & IOCB_DIRECT)) {
- /* If ->needs_lock is true, we're already in async context. */
- if (!s->needs_lock)
- io_async_list_note(WRITE, req, iov_count);
+ if (force_nonblock && !(kiocb->ki_flags & IOCB_DIRECT))
goto out_free;
- }
ret = rw_verify_area(WRITE, file, &kiocb->ki_pos, iov_count);
- if (!ret) {
- ssize_t ret2;
+ if (ret)
+ goto out_free;
- /*
- * Open-code file_start_write here to grab freeze protection,
- * which will be released by another thread in
- * io_complete_rw(). Fool lockdep by telling it the lock got
- * released so that it doesn't complain about the held lock when
- * we return to userspace.
- */
- if (S_ISREG(file_inode(file)->i_mode)) {
- __sb_start_write(file_inode(file)->i_sb,
- SB_FREEZE_WRITE, true);
- __sb_writers_release(file_inode(file)->i_sb,
- SB_FREEZE_WRITE);
- }
- kiocb->ki_flags |= IOCB_WRITE;
+ /*
+ * Open-code file_start_write here to grab freeze protection,
+ * which will be released by another thread in
+ * io_complete_rw(). Fool lockdep by telling it the lock got
+ * released so that it doesn't complain about the held lock when
+ * we return to userspace.
+ */
+ if (S_ISREG(file_inode(file)->i_mode)) {
+ __sb_start_write(file_inode(file)->i_sb, SB_FREEZE_WRITE, true);
+ __sb_writers_release(file_inode(file)->i_sb, SB_FREEZE_WRITE);
+ }
+ kiocb->ki_flags |= IOCB_WRITE;
- ret2 = call_write_iter(file, kiocb, &iter);
- if (!force_nonblock || ret2 != -EAGAIN) {
- io_rw_done(kiocb, ret2);
- } else {
- /*
- * If ->needs_lock is true, we're already in async
- * context.
- */
- if (!s->needs_lock)
- io_async_list_note(WRITE, req, iov_count);
- ret = -EAGAIN;
- }
+ /* Passthrough -EAGAIN return for forced non-blocking submission */
+ ret = call_write_iter(file, kiocb, &iter);
+ if (!(force_nonblock && ret == -EAGAIN)) {
+ io_rw_done(kiocb, ret);
+ ret = 0;
}
+
out_free:
+ /* If ->needs_lock is true, we're already in async context. */
+ if (ret == -EAGAIN && !s->needs_lock)
+ io_async_list_note(WRITE, req, iov_count);
+
kfree(iovec);
return ret;
}
--
2.20.1
next prev parent reply other threads:[~2019-05-03 9:47 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-04-23 19:06 io_uring: not good enough for release Stefan Bühler
2019-04-23 20:31 ` Jens Axboe
2019-04-23 22:07 ` Jens Axboe
2019-04-24 16:09 ` Jens Axboe
2019-04-27 16:05 ` io_uring: RWF_NOWAIT support Stefan Bühler
2019-04-27 18:34 ` [PATCH v1 1/1] [io_uring] fix handling SQEs requesting NOWAIT Stefan Bühler
2019-04-30 15:40 ` Jens Axboe
2019-04-27 15:50 ` io_uring: submission error handling Stefan Bühler
2019-04-30 16:02 ` Jens Axboe
2019-04-30 16:15 ` Jens Axboe
2019-04-30 18:15 ` Stefan Bühler
2019-04-30 18:42 ` Jens Axboe
2019-05-01 11:49 ` [PATCH v1 1/1] [io_uring] don't stall on submission errors Stefan Bühler
2019-05-01 12:43 ` Jens Axboe
2019-04-27 21:07 ` io_uring: closing / release Stefan Bühler
2019-05-11 16:26 ` Stefan Bühler
2019-04-28 15:54 ` io_uring: O_NONBLOCK/IOCB_NOWAIT/RWF_NOWAIT mess Stefan Bühler
2019-05-11 16:34 ` Stefan Bühler
2019-05-11 16:57 ` [PATCH 1/5] fs: RWF flags override default IOCB flags from file flags Stefan Bühler
2019-05-11 16:57 ` [PATCH 2/5] tcp: handle SPLICE_F_NONBLOCK in tcp_splice_read Stefan Bühler
2019-05-11 16:57 ` [PATCH 3/5] pipe: use IOCB_NOWAIT instead of O_NONBLOCK Stefan Bühler
2019-05-11 16:57 ` [PATCH 4/5] socket: " Stefan Bühler
2019-05-11 16:57 ` [PATCH 5/5] io_uring: use FMODE_NOWAIT to detect files supporting IOCB_NOWAIT Stefan Bühler
2019-05-03 9:47 ` Stefan Bühler [this message]
2019-05-03 9:47 ` [PATCH 2/2] io_uring: punt to workers if file doesn't support async Stefan Bühler
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 \
--in-reply-to=20190503094715.2381-1-source@stbuehler.de \
--to=source@stbuehler.de \
--cc=axboe@kernel.dk \
--cc=linux-block@vger.kernel.org \
--cc=linux-fsdevel@vger.kernel.org \
/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;
as well as URLs for NNTP newsgroup(s).