From: Oleg Nesterov <oleg@redhat.com>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: WangYuli <wangyuli@uniontech.com>,
linux-fsdevel <linux-fsdevel@vger.kernel.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Christian Brauner <brauner@kernel.org>
Subject: Re: [RESEND PATCH] fs/pipe: Introduce a check to skip sleeping processes during pipe read/write
Date: Fri, 27 Dec 2024 17:43:48 +0100 [thread overview]
Message-ID: <20241227164348.GB15300@redhat.com> (raw)
In-Reply-To: <20241227155428.GA15300@redhat.com>
On 12/27, Oleg Nesterov wrote:
>
> Consider
>
> int main(void)
> {
> int fd[2], cnt;
> char c;
>
> pipe(fd);
>
> if (!fork()) {
> // wait until the parent blocks in pipe_write() ->
> // wait_event_interruptible_exclusive(pipe->wr_wait, pipe_writable(pipe));
> sleep(1);
>
> for (cnt = 0; cnt < 4096; ++cnt)
> read(fd[0], &c, 1);
> return 0;
> }
>
> // parent
> for (;;)
> write(fd[1], &c, 1);
> }
>
> In this case the child will wakeup the parent 4095 times for no reason,
> pipe_writable() == !pipe_pull() will still be true until the last
> read(fd[0], &c, 1) does
>
> if (!buf->len)
> tail = pipe_update_tail(pipe, buf, tail);
>
> and after that the parent can write the next char.
perhaps something like below makes sense in this particular case.
Incomplete and ugly, just for illustration.
Oleg.
diff --git a/fs/pipe.c b/fs/pipe.c
index 12b22c2723b7..b8eef9e75639 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -253,7 +253,7 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to)
size_t total_len = iov_iter_count(to);
struct file *filp = iocb->ki_filp;
struct pipe_inode_info *pipe = filp->private_data;
- bool was_full, wake_next_reader = false;
+ bool was_full, xxx, wake_next_reader = false;
ssize_t ret;
/* Null read succeeds. */
@@ -277,6 +277,7 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to)
unsigned int head = smp_load_acquire(&pipe->head);
unsigned int tail = pipe->tail;
unsigned int mask = pipe->ring_size - 1;
+ xxx = false;
#ifdef CONFIG_WATCH_QUEUE
if (pipe->note_loss) {
@@ -340,8 +341,10 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to)
buf->len = 0;
}
- if (!buf->len)
+ if (!buf->len) {
tail = pipe_update_tail(pipe, buf, tail);
+ xxx = true;
+ }
total_len -= chars;
if (!total_len)
break; /* common path: read succeeded */
@@ -398,7 +401,7 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to)
wake_next_reader = false;
mutex_unlock(&pipe->mutex);
- if (was_full)
+ if (was_full && xxx)
wake_up_interruptible_sync_poll(&pipe->wr_wait, EPOLLOUT | EPOLLWRNORM);
if (wake_next_reader)
wake_up_interruptible_sync_poll(&pipe->rd_wait, EPOLLIN | EPOLLRDNORM);
next prev parent reply other threads:[~2024-12-27 16:44 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-12-25 9:42 [RESEND PATCH] fs/pipe: Introduce a check to skip sleeping processes during pipe read/write WangYuli
2024-12-25 13:30 ` Andy Shevchenko
2024-12-25 13:53 ` Kent Overstreet
2024-12-25 16:04 ` Mateusz Guzik
2024-12-25 16:32 ` Kent Overstreet
2024-12-25 17:22 ` Mateusz Guzik
2024-12-25 17:41 ` Kent Overstreet
2024-12-25 15:42 ` WangYuli
2024-12-25 16:00 ` Willy Tarreau
2024-12-25 16:32 ` WangYuli
2024-12-25 16:56 ` Willy Tarreau
2024-12-26 16:00 ` Oleg Nesterov
2024-12-26 19:02 ` Linus Torvalds
2024-12-26 20:11 ` Oleg Nesterov
2024-12-26 20:29 ` Linus Torvalds
2024-12-26 20:57 ` Oleg Nesterov
2024-12-27 15:54 ` Oleg Nesterov
2024-12-27 16:43 ` Oleg Nesterov [this message]
2024-12-27 18:39 ` Manfred Spraul
2024-12-28 14:32 ` Oleg Nesterov
2024-12-28 15:22 ` Oleg Nesterov
2024-12-28 16:32 ` Oleg Nesterov
2024-12-28 18:53 ` Manfred Spraul
2024-12-29 11:54 ` Oleg Nesterov
2024-12-28 16:45 ` Manfred Spraul
2024-12-29 11:57 ` Oleg Nesterov
2024-12-29 12:41 ` Manfred Spraul
2024-12-29 13:05 ` Oleg Nesterov
2024-12-29 13:13 ` Oleg Nesterov
2024-12-29 19:54 ` Manfred Spraul
2024-12-30 15:38 ` Oleg Nesterov
2024-12-31 11:14 ` Manfred Spraul
2024-12-31 19:38 ` Linus Torvalds
2024-12-31 20:24 ` Oleg Nesterov
2024-12-31 22:31 ` Linus Torvalds
2025-01-02 13:57 ` Oleg Nesterov
2025-01-04 21:15 ` RFC: Checkpatch: Introduce list of functions that need memory barriers Manfred Spraul
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=20241227164348.GB15300@redhat.com \
--to=oleg@redhat.com \
--cc=brauner@kernel.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=torvalds@linux-foundation.org \
--cc=wangyuli@uniontech.com \
/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