From: K Prateek Nayak <kprateek.nayak@amd.com>
To: Mateusz Guzik <mjguzik@gmail.com>
Cc: "Sapkal, Swapnil" <swapnil.sapkal@amd.com>,
Oleg Nesterov <oleg@redhat.com>,
Manfred Spraul <manfred@colorfullife.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
Christian Brauner <brauner@kernel.org>,
David Howells <dhowells@redhat.com>,
WangYuli <wangyuli@uniontech.com>,
<linux-fsdevel@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
"Shenoy, Gautham Ranjal" <gautham.shenoy@amd.com>,
<Neeraj.Upadhyay@amd.com>, <Ananth.narayan@amd.com>
Subject: Re: [PATCH] pipe_read: don't wake up the writer if the pipe is still full
Date: Tue, 4 Mar 2025 00:02:59 +0530 [thread overview]
Message-ID: <da2b1976-b392-4b3e-973b-3029a1ca1d72@amd.com> (raw)
In-Reply-To: <3jnnhipk2at3f7r23qb7fvznqg6dqw4rfrhajc7h6j2nu7twi2@wc3g5sdlfewt>
Hello Mateusz,
On 3/3/2025 11:24 PM, Mateusz Guzik wrote:
> Can you guys try out the patch below?
>
> It changes things up so that there is no need to read 2 different vars.
>
> It is not the final version and I don't claim to be able to fully
> justify the thing at the moment either, but I would like to know if it
> fixes the problem.
Happy to help! We've queued the below patch for an overnight run, will
report back once it is done.
Full disclaimer: We're testing on top of commit aaec5a95d596
("pipe_read: don't wake up the writer if the pipe is still full") where the
issue is more reproducible. I've replaced the VFS_BUG_ON() with a plain
BUG_ON() based on [1] since v6.14-rc1 did not include the CONFIG_DEBUG_VFS
bits. Hope that is alright.
[1] https://lore.kernel.org/lkml/20250209185523.745956-2-mjguzik@gmail.com/
/off to get some shut eyes/
--
Thanks and Regards,
Prateek
>
> If you don't have time that's fine, this is a quick jab. While I can't
> reproduce the bug myself even after inserting a delay by hand with
> msleep between the loads, I verified it does not outright break either.
> :P
>
> diff --git a/fs/pipe.c b/fs/pipe.c
> index 19a7948ab234..e61ad589fc2c 100644
> --- a/fs/pipe.c
> +++ b/fs/pipe.c
> @@ -210,11 +210,21 @@ static const struct pipe_buf_operations anon_pipe_buf_ops = {
> /* Done while waiting without holding the pipe lock - thus the READ_ONCE() */
> static inline bool pipe_readable(const struct pipe_inode_info *pipe)
> {
> - unsigned int head = READ_ONCE(pipe->head);
> - unsigned int tail = READ_ONCE(pipe->tail);
> - unsigned int writers = READ_ONCE(pipe->writers);
> + return !READ_ONCE(pipe->isempty) || !READ_ONCE(pipe->writers);
> +}
> +
> +static inline void pipe_recalc_state(struct pipe_inode_info *pipe)
> +{
> + pipe->isempty = pipe_empty(pipe->head, pipe->tail);
> + pipe->isfull = pipe_full(pipe->head, pipe->tail, pipe->max_usage);
> + VFS_BUG_ON(pipe->isempty && pipe->isfull);
> +}
>
> - return !pipe_empty(head, tail) || !writers;
> +static inline void pipe_update_head(struct pipe_inode_info *pipe,
> + unsigned int head)
> +{
> + pipe->head = ++head;
> + pipe_recalc_state(pipe);
> }
>
> static inline unsigned int pipe_update_tail(struct pipe_inode_info *pipe,
> @@ -244,6 +254,7 @@ static inline unsigned int pipe_update_tail(struct pipe_inode_info *pipe,
> * without the spinlock - the mutex is enough.
> */
> pipe->tail = ++tail;
> + pipe_recalc_state(pipe);
> return tail;
> }
>
> @@ -403,12 +414,7 @@ static inline int is_packetized(struct file *file)
> /* Done while waiting without holding the pipe lock - thus the READ_ONCE() */
> static inline bool pipe_writable(const struct pipe_inode_info *pipe)
> {
> - unsigned int head = READ_ONCE(pipe->head);
> - unsigned int tail = READ_ONCE(pipe->tail);
> - unsigned int max_usage = READ_ONCE(pipe->max_usage);
> -
> - return !pipe_full(head, tail, max_usage) ||
> - !READ_ONCE(pipe->readers);
> + return !READ_ONCE(pipe->isfull) || !READ_ONCE(pipe->readers);
> }
>
> static ssize_t
> @@ -512,7 +518,7 @@ anon_pipe_write(struct kiocb *iocb, struct iov_iter *from)
> break;
> }
>
> - pipe->head = head + 1;
> + pipe_update_head(pipe, head);
> pipe->tmp_page = NULL;
> /* Insert it into the buffer array */
> buf = &pipe->bufs[head & mask];
> @@ -529,10 +535,9 @@ anon_pipe_write(struct kiocb *iocb, struct iov_iter *from)
>
> if (!iov_iter_count(from))
> break;
> - }
>
> - if (!pipe_full(head, pipe->tail, pipe->max_usage))
> continue;
> + }
>
> /* Wait for buffer space to become available. */
> if ((filp->f_flags & O_NONBLOCK) ||
> diff --git a/include/linux/pipe_fs_i.h b/include/linux/pipe_fs_i.h
> index 8ff23bf5a819..d4b7539399b5 100644
> --- a/include/linux/pipe_fs_i.h
> +++ b/include/linux/pipe_fs_i.h
> @@ -69,6 +69,8 @@ struct pipe_inode_info {
> unsigned int r_counter;
> unsigned int w_counter;
> bool poll_usage;
> + bool isempty;
> + bool isfull;
> #ifdef CONFIG_WATCH_QUEUE
> bool note_loss;
> #endif
next prev parent reply other threads:[~2025-03-03 18:33 UTC|newest]
Thread overview: 109+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-01-02 14:07 Oleg Nesterov
2025-01-02 16:20 ` WangYuli
2025-01-02 16:46 ` Oleg Nesterov
2025-01-04 8:42 ` Christian Brauner
2025-01-31 9:49 ` K Prateek Nayak
2025-01-31 13:23 ` Oleg Nesterov
2025-01-31 20:06 ` Linus Torvalds
2025-02-02 17:01 ` Oleg Nesterov
2025-02-02 18:39 ` Linus Torvalds
2025-02-02 19:32 ` Oleg Nesterov
2025-02-04 11:17 ` Christian Brauner
2025-02-03 9:05 ` K Prateek Nayak
2025-02-04 13:49 ` Oleg Nesterov
2025-02-24 9:26 ` Sapkal, Swapnil
2025-02-24 14:24 ` Oleg Nesterov
2025-02-24 18:36 ` Linus Torvalds
2025-02-25 14:26 ` Oleg Nesterov
2025-02-25 11:57 ` Oleg Nesterov
2025-02-26 5:55 ` Sapkal, Swapnil
2025-02-26 11:38 ` Oleg Nesterov
2025-02-26 17:56 ` Sapkal, Swapnil
2025-02-26 18:12 ` Oleg Nesterov
2025-03-03 13:00 ` Alexey Gladkov
2025-03-03 15:46 ` K Prateek Nayak
2025-03-03 17:18 ` Alexey Gladkov
2025-02-26 13:18 ` Mateusz Guzik
2025-02-26 13:21 ` Mateusz Guzik
2025-02-26 17:16 ` Oleg Nesterov
2025-02-27 16:18 ` Sapkal, Swapnil
2025-02-27 16:34 ` Mateusz Guzik
2025-02-27 21:12 ` Oleg Nesterov
2025-02-28 5:58 ` Sapkal, Swapnil
2025-02-28 14:30 ` Oleg Nesterov
2025-02-28 16:33 ` Oleg Nesterov
2025-03-03 9:46 ` Sapkal, Swapnil
2025-03-03 14:37 ` Mateusz Guzik
2025-03-03 14:51 ` Mateusz Guzik
2025-03-03 15:31 ` K Prateek Nayak
2025-03-03 17:54 ` Mateusz Guzik
2025-03-03 18:11 ` Linus Torvalds
2025-03-03 18:33 ` Mateusz Guzik
2025-03-03 18:55 ` Linus Torvalds
2025-03-03 19:06 ` Mateusz Guzik
2025-03-03 20:27 ` Oleg Nesterov
2025-03-03 20:46 ` Linus Torvalds
2025-03-04 5:31 ` K Prateek Nayak
2025-03-04 6:32 ` Linus Torvalds
2025-03-04 12:54 ` Oleg Nesterov
2025-03-04 13:25 ` Oleg Nesterov
2025-03-04 18:28 ` Linus Torvalds
2025-03-04 22:11 ` Oleg Nesterov
2025-03-05 4:40 ` K Prateek Nayak
2025-03-05 4:52 ` Linus Torvalds
2025-03-04 13:51 ` [PATCH] fs/pipe: Read pipe->{head,tail} atomically outside pipe->mutex K Prateek Nayak
2025-03-04 18:36 ` Alexey Gladkov
2025-03-04 19:03 ` Linus Torvalds
2025-03-05 15:31 ` [PATCH] pipe_read: don't wake up the writer if the pipe is still full Rasmus Villemoes
2025-03-05 16:50 ` Linus Torvalds
2025-03-06 9:48 ` Rasmus Villemoes
2025-03-06 14:42 ` Rasmus Villemoes
2025-03-05 16:40 ` Linus Torvalds
2025-03-06 8:35 ` Rasmus Villemoes
2025-03-06 17:59 ` Linus Torvalds
2025-03-06 9:28 ` Rasmus Villemoes
2025-03-06 11:39 ` [RFC PATCH 0/3] pipe: Convert pipe->{head,tail} to unsigned short K Prateek Nayak
2025-03-06 11:39 ` [RFC PATCH 1/3] fs/pipe: Limit the slots in pipe_resize_ring() K Prateek Nayak
2025-03-06 12:28 ` Oleg Nesterov
2025-03-06 15:26 ` K Prateek Nayak
2025-03-06 11:39 ` [RFC PATCH 2/3] fs/splice: Atomically read pipe->{head,tail} in opipe_prep() K Prateek Nayak
2025-03-06 11:39 ` [RFC PATCH 3/3] treewide: pipe: Convert all references to pipe->{head,tail,max_usage,ring_size} to unsigned short K Prateek Nayak
2025-03-06 12:32 ` Oleg Nesterov
2025-03-06 12:41 ` Oleg Nesterov
2025-03-06 15:33 ` K Prateek Nayak
2025-03-06 18:04 ` Linus Torvalds
2025-03-06 14:27 ` Rasmus Villemoes
2025-03-03 18:32 ` K Prateek Nayak [this message]
2025-03-04 5:22 ` [PATCH] pipe_read: don't wake up the writer if the pipe is still full K Prateek Nayak
2025-03-03 16:49 ` Oleg Nesterov
2025-03-04 5:06 ` Hillf Danton
2025-03-04 5:35 ` K Prateek Nayak
2025-03-04 10:29 ` Hillf Danton
2025-03-04 12:34 ` Oleg Nesterov
2025-03-04 23:35 ` Hillf Danton
2025-03-04 23:49 ` Oleg Nesterov
2025-03-05 4:56 ` Hillf Danton
2025-03-05 11:44 ` Oleg Nesterov
2025-03-05 22:46 ` Hillf Danton
2025-03-06 9:30 ` Oleg Nesterov
2025-03-07 6:08 ` Hillf Danton
2025-03-07 6:24 ` K Prateek Nayak
2025-03-07 10:46 ` Hillf Danton
2025-03-07 11:29 ` Oleg Nesterov
2025-03-07 12:34 ` Oleg Nesterov
2025-03-07 23:56 ` Hillf Danton
2025-03-09 14:01 ` K Prateek Nayak
2025-03-09 17:02 ` Oleg Nesterov
2025-03-10 10:49 ` Hillf Danton
2025-03-10 11:09 ` Oleg Nesterov
2025-03-10 11:37 ` Hillf Danton
2025-03-10 12:43 ` Oleg Nesterov
2025-03-10 23:33 ` Hillf Danton
2025-03-11 0:26 ` Linus Torvalds
2025-03-11 6:54 ` Oleg Nesterov
[not found] ` <20250311112922.3342-1-hdanton@sina.com>
2025-03-11 11:53 ` Oleg Nesterov
2025-03-07 11:26 ` Oleg Nesterov
2025-02-27 12:50 ` Oleg Nesterov
2025-02-27 13:52 ` Oleg Nesterov
2025-02-27 15:59 ` Mateusz Guzik
2025-02-27 16:28 ` Oleg Nesterov
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=da2b1976-b392-4b3e-973b-3029a1ca1d72@amd.com \
--to=kprateek.nayak@amd.com \
--cc=Ananth.narayan@amd.com \
--cc=Neeraj.Upadhyay@amd.com \
--cc=brauner@kernel.org \
--cc=dhowells@redhat.com \
--cc=gautham.shenoy@amd.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=manfred@colorfullife.com \
--cc=mjguzik@gmail.com \
--cc=oleg@redhat.com \
--cc=swapnil.sapkal@amd.com \
--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
all inboxes | Powered by JetHome®