Re: [pipe] 3a34b13a88: hackbench.throughput -12.6% regression

[Linus Torvalds]

On Mon, Aug 2, 2021 at 10:06 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> Sandeep, does something like the attached patch (written to be on top
> of the existing one) work for you?

.. and by "attached patch" I obviously meant the patch that wasn't attached.

It's here.

                 Linus
 fs/pipe.c                 | 15 +++++++++------
 include/linux/pipe_fs_i.h |  2 ++
 2 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/fs/pipe.c b/fs/pipe.c
index 9ef4231cce61..e1704c14a918 100644
--- a/fs/pipe.c
+++ b/fs/pipe.c
@@ -429,9 +429,6 @@ pipe_write(struct kiocb *iocb, struct iov_iter *from)
 #endif
 
 	/*
-	 * Epoll nonsensically wants a wakeup whether the pipe
-	 * was already empty or not.
-	 *
 	 * If it wasn't empty we try to merge new data into
 	 * the last buffer.
 	 *
@@ -440,9 +437,9 @@ pipe_write(struct kiocb *iocb, struct iov_iter *from)
 	 * spanning multiple pages.
 	 */
 	head = pipe->head;
-	was_empty = true;
+	was_empty = pipe_empty(head, pipe->tail);
 	chars = total_len & (PAGE_SIZE-1);
-	if (chars && !pipe_empty(head, pipe->tail)) {
+	if (chars && !was_empty) {
 		unsigned int mask = pipe->ring_size - 1;
 		struct pipe_buffer *buf = &pipe->bufs[(head - 1) & mask];
 		int offset = buf->offset + buf->len;
@@ -575,8 +572,11 @@ pipe_write(struct kiocb *iocb, struct iov_iter *from)
 	 * This is particularly important for small writes, because of
 	 * how (for example) the GNU make jobserver uses small writes to
 	 * wake up pending jobs
+	 *
+	 * Epoll nonsensically wants a wakeup whether the pipe
+	 * was already empty or not.
 	 */
-	if (was_empty) {
+	if (was_empty || pipe->poll_usage) {
 		wake_up_interruptible_sync_poll(&pipe->rd_wait, EPOLLIN | EPOLLRDNORM);
 		kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
 	}
@@ -639,6 +639,9 @@ pipe_poll(struct file *filp, poll_table *wait)
 	struct pipe_inode_info *pipe = filp->private_data;
 	unsigned int head, tail;
 
+	/* Epoll has some historical nasty semantics, this enables them */
+	pipe->poll_usage = 1;
+
 	/*
 	 * Reading pipe state only -- no need for acquiring the semaphore.
 	 *
diff --git a/include/linux/pipe_fs_i.h b/include/linux/pipe_fs_i.h
index 5d2705f1d01c..fc5642431b92 100644
--- a/include/linux/pipe_fs_i.h
+++ b/include/linux/pipe_fs_i.h
@@ -48,6 +48,7 @@ struct pipe_buffer {
  *	@files: number of struct file referring this pipe (protected by ->i_lock)
  *	@r_counter: reader counter
  *	@w_counter: writer counter
+ *	@poll_usage: is this pipe used for epoll, which has crazy wakeups?
  *	@fasync_readers: reader side fasync
  *	@fasync_writers: writer side fasync
  *	@bufs: the circular array of pipe buffers
@@ -70,6 +71,7 @@ struct pipe_inode_info {
 	unsigned int files;
 	unsigned int r_counter;
 	unsigned int w_counter;
+	unsigned int poll_usage;
 	struct page *tmp_page;
 	struct fasync_struct *fasync_readers;
 	struct fasync_struct *fasync_writers;