| Message ID | 20211122041135.2450220-1-yangerkun@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | pipe: remove needless spin_lock in pipe_read/pipe_write | expand |
ping On 2021/11/22 12:11, yangerkun wrote: > Once enable CONFIG_WATCH_QUEUE, we should protect pipe with > pipe->rd_wait.lock since post_one_notification may write pipe from > contexts where pipe->mutex cannot be token. But nowdays we will try > take it for anycase, it seems needless. Besides, pipe_write will break > once it's pipe with O_NOTIFICATION_PIPE, pipe->rd_wait.lock seems no > need at all. We make some change base on that, and it can help improve > performance for some case like pipe_pst1 in libMicro. > > ARMv7 for our scene, before this patch: > 5483 nsecs/call > After this patch: > 4854 nsecs/call > > Signed-off-by: yangerkun <yangerkun@huawei.com> > --- > fs/pipe.c | 15 +++++++-------- > 1 file changed, 7 insertions(+), 8 deletions(-) > > diff --git a/fs/pipe.c b/fs/pipe.c > index 6d4342bad9f1..e8ced0c50824 100644 > --- a/fs/pipe.c > +++ b/fs/pipe.c > @@ -320,14 +320,18 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to) > > if (!buf->len) { > pipe_buf_release(pipe, buf); > - spin_lock_irq(&pipe->rd_wait.lock); > #ifdef CONFIG_WATCH_QUEUE > + if (pipe->watch_queue) > + spin_lock_irq(&pipe->rd_wait.lock); > if (buf->flags & PIPE_BUF_FLAG_LOSS) > pipe->note_loss = true; > #endif > tail++; > pipe->tail = tail; > - spin_unlock_irq(&pipe->rd_wait.lock); > +#ifdef CONFIG_WATCH_QUEUE > + if (pipe->watch_queue) > + spin_unlock_irq(&pipe->rd_wait.lock); > +#endif > } > total_len -= chars; > if (!total_len) > @@ -504,16 +508,11 @@ 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); > + if (pipe_full(head, pipe->tail, pipe->max_usage)) > continue; > - } > > pipe->head = head + 1; > - spin_unlock_irq(&pipe->rd_wait.lock); > > /* Insert it into the buffer array */ > buf = &pipe->bufs[head & mask]; >
ping... On 2021/11/29 14:22, yangerkun wrote: > ping > > On 2021/11/22 12:11, yangerkun wrote: >> Once enable CONFIG_WATCH_QUEUE, we should protect pipe with >> pipe->rd_wait.lock since post_one_notification may write pipe from >> contexts where pipe->mutex cannot be token. But nowdays we will try >> take it for anycase, it seems needless. Besides, pipe_write will break >> once it's pipe with O_NOTIFICATION_PIPE, pipe->rd_wait.lock seems no >> need at all. We make some change base on that, and it can help improve >> performance for some case like pipe_pst1 in libMicro. >> >> ARMv7 for our scene, before this patch: >> 5483 nsecs/call >> After this patch: >> 4854 nsecs/call >> >> Signed-off-by: yangerkun <yangerkun@huawei.com> >> --- >> fs/pipe.c | 15 +++++++-------- >> 1 file changed, 7 insertions(+), 8 deletions(-) >> >> diff --git a/fs/pipe.c b/fs/pipe.c >> index 6d4342bad9f1..e8ced0c50824 100644 >> --- a/fs/pipe.c >> +++ b/fs/pipe.c >> @@ -320,14 +320,18 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to) >> if (!buf->len) { >> pipe_buf_release(pipe, buf); >> - spin_lock_irq(&pipe->rd_wait.lock); >> #ifdef CONFIG_WATCH_QUEUE >> + if (pipe->watch_queue) >> + spin_lock_irq(&pipe->rd_wait.lock); >> if (buf->flags & PIPE_BUF_FLAG_LOSS) >> pipe->note_loss = true; >> #endif >> tail++; >> pipe->tail = tail; >> - spin_unlock_irq(&pipe->rd_wait.lock); >> +#ifdef CONFIG_WATCH_QUEUE >> + if (pipe->watch_queue) >> + spin_unlock_irq(&pipe->rd_wait.lock); >> +#endif >> } >> total_len -= chars; >> if (!total_len) >> @@ -504,16 +508,11 @@ 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); >> + if (pipe_full(head, pipe->tail, pipe->max_usage)) >> continue; >> - } >> pipe->head = head + 1; >> - spin_unlock_irq(&pipe->rd_wait.lock); >> /* Insert it into the buffer array */ >> buf = &pipe->bufs[head & mask]; >> > .
diff --git a/fs/pipe.c b/fs/pipe.c index 6d4342bad9f1..e8ced0c50824 100644 --- a/fs/pipe.c +++ b/fs/pipe.c @@ -320,14 +320,18 @@ pipe_read(struct kiocb *iocb, struct iov_iter *to) if (!buf->len) { pipe_buf_release(pipe, buf); - spin_lock_irq(&pipe->rd_wait.lock); #ifdef CONFIG_WATCH_QUEUE + if (pipe->watch_queue) + spin_lock_irq(&pipe->rd_wait.lock); if (buf->flags & PIPE_BUF_FLAG_LOSS) pipe->note_loss = true; #endif tail++; pipe->tail = tail; - spin_unlock_irq(&pipe->rd_wait.lock); +#ifdef CONFIG_WATCH_QUEUE + if (pipe->watch_queue) + spin_unlock_irq(&pipe->rd_wait.lock); +#endif } total_len -= chars; if (!total_len) @@ -504,16 +508,11 @@ 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); + if (pipe_full(head, pipe->tail, pipe->max_usage)) continue; - } pipe->head = head + 1; - spin_unlock_irq(&pipe->rd_wait.lock); /* Insert it into the buffer array */ buf = &pipe->bufs[head & mask];
Once enable CONFIG_WATCH_QUEUE, we should protect pipe with pipe->rd_wait.lock since post_one_notification may write pipe from contexts where pipe->mutex cannot be token. But nowdays we will try take it for anycase, it seems needless. Besides, pipe_write will break once it's pipe with O_NOTIFICATION_PIPE, pipe->rd_wait.lock seems no need at all. We make some change base on that, and it can help improve performance for some case like pipe_pst1 in libMicro. ARMv7 for our scene, before this patch: 5483 nsecs/call After this patch: 4854 nsecs/call Signed-off-by: yangerkun <yangerkun@huawei.com> --- fs/pipe.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-)