public inbox for [email protected]
 help / color / mirror / Atom feed
* [PATCH v1] fs/io_uring: Hoist ret2 == -EAGAIN check in tail of io_write
@ 2021-10-18  7:02 Noah Goldstein
  2021-10-18 12:17 ` Hao Xu
  2021-10-18 14:56 ` [PATCH v2] " Noah Goldstein
  0 siblings, 2 replies; 3+ messages in thread
From: Noah Goldstein @ 2021-10-18  7:02 UTC (permalink / raw)
  Cc: goldstein.w.n, axboe, asml.silence, io-uring, linux-kernel

This commit reorganizes the branches in the tail of io_write so that
the 'ret2 == -EAGAIN' check is not repeated and done first.

The previous version was duplicating the 'ret2 == -EAGAIN'. As well
'ret2 != -EAGAIN' gurantees the 'done:' path so it makes sense to
move that check to the front before the likely more expensive branches
which require memory derefences.

Signed-off-by: Noah Goldstein <[email protected]>
---
Generally I would want to rewrite this as:
```
if (ret2 != -EAGAIN
    || (req->flags & REQ_F_NOWAIT)
    || (!force_nonblock && !(req->ctx->flags & IORING_SETUP_IOPOLL)))
        kiocb_done(kiocb, ret2, issue_flags);
else {
    ...
```

But the style of the file seems to be to use gotos. If the above is
prefereable, let me know and I'll post a new version.    
 fs/io_uring.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index d1e672e7a2d1..932fc84d70d3 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -3648,12 +3648,15 @@ static int io_write(struct io_kiocb *req, unsigned int issue_flags)
 	 */
 	if (ret2 == -EOPNOTSUPP && (kiocb->ki_flags & IOCB_NOWAIT))
 		ret2 = -EAGAIN;
+
+	if (ret2 != -EAGAIN)
+		goto done;
 	/* no retry on NONBLOCK nor RWF_NOWAIT */
-	if (ret2 == -EAGAIN && (req->flags & REQ_F_NOWAIT))
+	if (req->flags & REQ_F_NOWAIT)
 		goto done;
-	if (!force_nonblock || ret2 != -EAGAIN) {
+	if (!force_nonblock) {
 		/* IOPOLL retry should happen for io-wq threads */
-		if (ret2 == -EAGAIN && (req->ctx->flags & IORING_SETUP_IOPOLL))
+		if (req->ctx->flags & IORING_SETUP_IOPOLL)
 			goto copy_iov;
 done:
 		kiocb_done(kiocb, ret2, issue_flags);
-- 
2.29.2


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH v1] fs/io_uring: Hoist ret2 == -EAGAIN check in tail of io_write
  2021-10-18  7:02 [PATCH v1] fs/io_uring: Hoist ret2 == -EAGAIN check in tail of io_write Noah Goldstein
@ 2021-10-18 12:17 ` Hao Xu
  2021-10-18 14:56 ` [PATCH v2] " Noah Goldstein
  1 sibling, 0 replies; 3+ messages in thread
From: Hao Xu @ 2021-10-18 12:17 UTC (permalink / raw)
  To: Noah Goldstein; +Cc: axboe, asml.silence, io-uring, linux-kernel

在 2021/10/18 下午3:02, Noah Goldstein 写道:
> This commit reorganizes the branches in the tail of io_write so that
> the 'ret2 == -EAGAIN' check is not repeated and done first.
> 
> The previous version was duplicating the 'ret2 == -EAGAIN'. As well
> 'ret2 != -EAGAIN' gurantees the 'done:' path so it makes sense to
> move that check to the front before the likely more expensive branches
> which require memory derefences.
> 
> Signed-off-by: Noah Goldstein <[email protected]>
> ---
> Generally I would want to rewrite this as:
> ```
> if (ret2 != -EAGAIN
>      || (req->flags & REQ_F_NOWAIT)
>      || (!force_nonblock && !(req->ctx->flags & IORING_SETUP_IOPOLL)))
>          kiocb_done(kiocb, ret2, issue_flags);
> else {
>      ...
> ```
To me, this one is clear enough and short, but I think better to:
if (ret2 != -EAGAIN || (req->flags & REQ_F_NOWAIT) ||
     (!force_nonblock && !(req->ctx->flags & IORING_SETUP_IOPOLL))

if the first line doesn't exceed the line limit.

Reviewed-by: Hao Xu <[email protected]>
> 
> But the style of the file seems to be to use gotos. If the above is
> prefereable, let me know and I'll post a new version.
>   fs/io_uring.c | 9 ++++++---
>   1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index d1e672e7a2d1..932fc84d70d3 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -3648,12 +3648,15 @@ static int io_write(struct io_kiocb *req, unsigned int issue_flags)
>   	 */
>   	if (ret2 == -EOPNOTSUPP && (kiocb->ki_flags & IOCB_NOWAIT))
>   		ret2 = -EAGAIN;
> +
> +	if (ret2 != -EAGAIN)
> +		goto done;
>   	/* no retry on NONBLOCK nor RWF_NOWAIT */
> -	if (ret2 == -EAGAIN && (req->flags & REQ_F_NOWAIT))
> +	if (req->flags & REQ_F_NOWAIT)
>   		goto done;
> -	if (!force_nonblock || ret2 != -EAGAIN) {
> +	if (!force_nonblock) {
>   		/* IOPOLL retry should happen for io-wq threads */
> -		if (ret2 == -EAGAIN && (req->ctx->flags & IORING_SETUP_IOPOLL))
> +		if (req->ctx->flags & IORING_SETUP_IOPOLL)
>   			goto copy_iov;
>   done:
>   		kiocb_done(kiocb, ret2, issue_flags);
> 


^ permalink raw reply	[flat|nested] 3+ messages in thread

* [PATCH v2] fs/io_uring: Hoist ret2 == -EAGAIN check in tail of io_write
  2021-10-18  7:02 [PATCH v1] fs/io_uring: Hoist ret2 == -EAGAIN check in tail of io_write Noah Goldstein
  2021-10-18 12:17 ` Hao Xu
@ 2021-10-18 14:56 ` Noah Goldstein
  1 sibling, 0 replies; 3+ messages in thread
From: Noah Goldstein @ 2021-10-18 14:56 UTC (permalink / raw)
  Cc: goldstein.w.n, axboe, asml.silence, io-uring, linux-kernel

This commit reorganizes the branches in the tail of io_write so that
the 'ret2 == -EAGAIN' check is not repeated and done first.

The previous version was duplicating the 'ret2 == -EAGAIN'. As well
'ret2 != -EAGAIN' gurantees the 'done:' path so it makes sense to
move that check to the front before the likely more expensive branches
which require memory derefences.

Signed-off-by: Noah Goldstein <[email protected]>
---
 fs/io_uring.c | 16 +++++++---------
 1 file changed, 7 insertions(+), 9 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index d1e672e7a2d1..18293407e8bc 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -3648,21 +3648,19 @@ static int io_write(struct io_kiocb *req, unsigned int issue_flags)
 	 */
 	if (ret2 == -EOPNOTSUPP && (kiocb->ki_flags & IOCB_NOWAIT))
 		ret2 = -EAGAIN;
-	/* no retry on NONBLOCK nor RWF_NOWAIT */
-	if (ret2 == -EAGAIN && (req->flags & REQ_F_NOWAIT))
-		goto done;
-	if (!force_nonblock || ret2 != -EAGAIN) {
-		/* IOPOLL retry should happen for io-wq threads */
-		if (ret2 == -EAGAIN && (req->ctx->flags & IORING_SETUP_IOPOLL))
-			goto copy_iov;
-done:
+
+	if (ret2 != -EAGAIN ||
+	    /* no retry on NONBLOCK nor RWF_NOWAIT */
+	    (req->flags & REQ_F_NOWAIT)
+	    /* IOPOLL retry should happen for io-wq threads */
+	    || (!force_nonblock && !(req->ctx->flags & IORING_SETUP_IOPOLL))) {
 		kiocb_done(kiocb, ret2, issue_flags);
 	} else {
 copy_iov:
 		iov_iter_restore(&s->iter, &s->iter_state);
 		ret = io_setup_async_rw(req, iovec, s, false);
 		return ret ?: -EAGAIN;
-	}
+}
 out_free:
 	/* it's reportedly faster than delegating the null check to kfree() */
 	if (iovec)
-- 
2.29.2


^ permalink raw reply related	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2021-10-18 14:56 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2021-10-18  7:02 [PATCH v1] fs/io_uring: Hoist ret2 == -EAGAIN check in tail of io_write Noah Goldstein
2021-10-18 12:17 ` Hao Xu
2021-10-18 14:56 ` [PATCH v2] " Noah Goldstein

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox