[PATCH OLK-6.6 0/2] Backport fs/pipe performance optimization patches
Backport some fs/pipe performance optimization patches from mainline. Max Kellermann (2): fs/pipe: remove unnecessary spinlock from pipe_write() fs/pipe: use spinlock in pipe_read() only if there is a watch_queue fs/pipe.c | 52 ++++++++++++++++++++++++++++++++-------------------- 1 file changed, 32 insertions(+), 20 deletions(-) -- 2.20.1
反馈： 您发送到kernel@openeuler.org的补丁/补丁集，已成功转换为PR！ PR链接地址： https://gitee.com/openeuler/kernel/pulls/18144 邮件列表地址：https://mailweb.openeuler.org/archives/list/kernel@openeuler.org/message/MIZ... FeedBack: The patch(es) which you have sent to kernel@openeuler.org mailing list has been converted to a pull request successfully! Pull request link: https://gitee.com/openeuler/kernel/pulls/18144 Mailing list address: https://mailweb.openeuler.org/archives/list/kernel@openeuler.org/message/MIZ...
From: Max Kellermann <max.kellermann@ionos.com> mainline inclusion from mainline-v6.7-rc1 commit dfaabf916b1ca83cfac856745db2fc9d57d9b13a category: performance bugzilla: https://gitee.com/src-openeuler/kernel/issues/ICZJGM CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i... -------------------------------- This reverts commit 8df441294dd3 ("pipe: Check for ring full inside of the spinlock in pipe_write()") which was obsoleted by commit c73be61cede ("pipe: Add general notification queue support") because now pipe_write() fails early with -EXDEV if there is a watch_queue. Without a watch_queue, no notifications can be posted to the pipe and mutex protection is enough, as can be seen in splice_pipe_to_pipe() which does not use the spinlock either. Signed-off-by: Max Kellermann <max.kellermann@ionos.com> Message-Id: <20230921075755.1378787-3-max.kellermann@ionos.com> Reviewed-by: David Howells <dhowells@redhat.com> Signed-off-by: Christian Brauner <brauner@kernel.org> Signed-off-by: Zheng Zengkai <zhengzengkai@huawei.com> --- fs/pipe.c | 9 --------- 1 file changed, 9 deletions(-) diff --git a/fs/pipe.c b/fs/pipe.c index cd7ad6093ce6..e89f293bd375 100644 --- a/fs/pipe.c +++ b/fs/pipe.c @@ -512,16 +512,7 @@ pipe_write(struct kiocb *iocb, struct iov_iter *from) * it, either the reader will consume it or it'll still * be there for the next write. */ - spin_lock_irq(&pipe->rd_wait.lock); - - head = pipe->head; - if (pipe_full(head, pipe->tail, pipe->max_usage)) { - spin_unlock_irq(&pipe->rd_wait.lock); - continue; - } - pipe->head = head + 1; - spin_unlock_irq(&pipe->rd_wait.lock); /* Insert it into the buffer array */ buf = &pipe->bufs[head & mask]; -- 2.20.1
From: Max Kellermann <max.kellermann@ionos.com> mainline inclusion from mainline-v6.7-rc1 commit 478dbf12176700f28d836dd03ae93a6888278230 category: performance bugzilla: https://gitee.com/src-openeuler/kernel/issues/ICZJGM CVE: NA Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i... -------------------------------- If there is no watch_queue, holding the pipe mutex is enough to prevent concurrent writes, and we can avoid the spinlock. O_NOTIFICATION_QUEUE is an exotic and rarely used feature, and of all the pipes that exist at any given time, only very few actually have a watch_queue, therefore it appears worthwile to optimize the common case. This patch does not optimize pipe_resize_ring() where the spinlocks could be avoided as well; that does not seem like a worthwile optimization because this function is not called often. Related commits: - commit 8df441294dd3 ("pipe: Check for ring full inside of the spinlock in pipe_write()") - commit b667b8673443 ("pipe: Advance tail pointer inside of wait spinlock in pipe_read()") - commit 189b0ddc2451 ("pipe: Fix missing lock in pipe_resize_ring()") Signed-off-by: Max Kellermann <max.kellermann@ionos.com> Message-Id: <20230921075755.1378787-4-max.kellermann@ionos.com> Reviewed-by: David Howells <dhowells@redhat.com> Signed-off-by: Christian Brauner <brauner@kernel.org> Signed-off-by: Zheng Zengkai <zhengzengkai@huawei.com> --- fs/pipe.c | 43 ++++++++++++++++++++++++++++++++----------- 1 file changed, 32 insertions(+), 11 deletions(-) diff --git a/fs/pipe.c b/fs/pipe.c index e89f293bd375..ad7db4e6280e 100644 --- a/fs/pipe.c +++ b/fs/pipe.c @@ -227,6 +227,36 @@ static inline bool pipe_readable(const struct pipe_inode_info *pipe) return !pipe_empty(head, tail) || !writers; } +static inline unsigned int pipe_update_tail(struct pipe_inode_info *pipe, + struct pipe_buffer *buf, + unsigned int tail) +{ + pipe_buf_release(pipe, buf); + + /* + * If the pipe has a watch_queue, we need additional protection + * by the spinlock because notifications get posted with only + * this spinlock, no mutex + */ + if (pipe_has_watch_queue(pipe)) { + spin_lock_irq(&pipe->rd_wait.lock); +#ifdef CONFIG_WATCH_QUEUE + if (buf->flags & PIPE_BUF_FLAG_LOSS) + pipe->note_loss = true; +#endif + pipe->tail = ++tail; + spin_unlock_irq(&pipe->rd_wait.lock); + return tail; + } + + /* + * Without a watch_queue, we can simply increment the tail + * without the spinlock - the mutex is enough. + */ + pipe->tail = ++tail; + return tail; +} + static ssize_t pipe_read(struct kiocb *iocb, struct iov_iter *to) { @@ -320,17 +350,8 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to) buf->len = 0; } - if (!buf->len) { - pipe_buf_release(pipe, buf); - spin_lock_irq(&pipe->rd_wait.lock); -#ifdef CONFIG_WATCH_QUEUE - if (buf->flags & PIPE_BUF_FLAG_LOSS) - pipe->note_loss = true; -#endif - tail++; - pipe->tail = tail; - spin_unlock_irq(&pipe->rd_wait.lock); - } + if (!buf->len) + tail = pipe_update_tail(pipe, buf, tail); total_len -= chars; if (!total_len) break; /* common path: read succeeded */ -- 2.20.1
participants (2)
-
patchwork bot -
Zheng Zengkai