| Message ID | 20180806083058.14724-5-hch@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/4] timerfd: add support for keyed wakeups | expand |
On Mon, 6 Aug 2018 10:30:58 +0200 Christoph Hellwig <hch@lst.de> wrote: > If we get a keyed wakeup for a aio poll waitqueue and wake can acquire the > ctx_lock without spinning we can just complete the iocb straight from the > wakeup callback to avoid a context switch. Why do we try to avoid spinning on the lock? > --- a/fs/aio.c > +++ b/fs/aio.c > @@ -1672,13 +1672,26 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync, > void *key) > { > struct poll_iocb *req = container_of(wait, struct poll_iocb, wait); > + struct aio_kiocb *iocb = container_of(req, struct aio_kiocb, poll); > __poll_t mask = key_to_poll(key); > > req->woken = true; > > /* for instances that support it check for an event match first: */ > - if (mask && !(mask & req->events)) > - return 0; > + if (mask) { > + if (!(mask & req->events)) > + return 0; > + > + /* try to complete the iocb inline if we can: */ ie, this comment explains 'what" but not "why". (There's a typo in Subject:, btw) > + if (spin_trylock(&iocb->ki_ctx->ctx_lock)) { > + list_del(&iocb->ki_list); > + spin_unlock(&iocb->ki_ctx->ctx_lock); > + > + list_del_init(&req->wait.entry); > + aio_poll_complete(iocb, mask); > + return 1; > + } > + } > > list_del_init(&req->wait.entry); > schedule_work(&req->work);
On Mon, Aug 06, 2018 at 03:27:05PM -0700, Andrew Morton wrote: > On Mon, 6 Aug 2018 10:30:58 +0200 Christoph Hellwig <hch@lst.de> wrote: > > > If we get a keyed wakeup for a aio poll waitqueue and wake can acquire the > > ctx_lock without spinning we can just complete the iocb straight from the > > wakeup callback to avoid a context switch. > > Why do we try to avoid spinning on the lock? Because we are called with the lock on the waitqueue called, which nests inside it. > > + /* try to complete the iocb inline if we can: */ > > ie, this comment explains 'what" but not "why". > > (There's a typo in Subject:, btw) Because it is faster obviously. I can update the comment.
On Tue, 7 Aug 2018 09:25:55 +0200 Christoph Hellwig <hch@lst.de> wrote: > On Mon, Aug 06, 2018 at 03:27:05PM -0700, Andrew Morton wrote: > > On Mon, 6 Aug 2018 10:30:58 +0200 Christoph Hellwig <hch@lst.de> wrote: > > > > > If we get a keyed wakeup for a aio poll waitqueue and wake can acquire the > > > ctx_lock without spinning we can just complete the iocb straight from the > > > wakeup callback to avoid a context switch. > > > > Why do we try to avoid spinning on the lock? > > Because we are called with the lock on the waitqueue called, which > nests inside it. Ah. > > > + /* try to complete the iocb inline if we can: */ > > > > ie, this comment explains 'what" but not "why". > > > > (There's a typo in Subject:, btw) > > Because it is faster obviously. I can update the comment. I meant the comment could explain why it's a trylock instead of a spin_lock().
On Tue, Aug 07, 2018 at 09:04:41AM -0700, Andrew Morton wrote: > > Because it is faster obviously. I can update the comment. > > I meant the comment could explain why it's a trylock instead of a > spin_lock(). We could something like this the patch below. Al, do you want me to resend or can you just fold it in? diff --git a/fs/aio.c b/fs/aio.c index 5943098a87c6..84df2c2bf80b 100644 --- a/fs/aio.c +++ b/fs/aio.c @@ -1684,7 +1684,8 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync, /* * Try to complete the iocb inline if we can to avoid a costly - * context switch. + * context switch. As the waitqueue lock nests inside the ctx + * lock we can only do that if we can get it without waiting. */ if (spin_trylock(&iocb->ki_ctx->ctx_lock)) { list_del(&iocb->ki_list);
diff --git a/fs/aio.c b/fs/aio.c index 2fd19521d8a8..29f2b5b57d32 100644 --- a/fs/aio.c +++ b/fs/aio.c @@ -1672,13 +1672,26 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync, void *key) { struct poll_iocb *req = container_of(wait, struct poll_iocb, wait); + struct aio_kiocb *iocb = container_of(req, struct aio_kiocb, poll); __poll_t mask = key_to_poll(key); req->woken = true; /* for instances that support it check for an event match first: */ - if (mask && !(mask & req->events)) - return 0; + if (mask) { + if (!(mask & req->events)) + return 0; + + /* try to complete the iocb inline if we can: */ + if (spin_trylock(&iocb->ki_ctx->ctx_lock)) { + list_del(&iocb->ki_list); + spin_unlock(&iocb->ki_ctx->ctx_lock); + + list_del_init(&req->wait.entry); + aio_poll_complete(iocb, mask); + return 1; + } + } list_del_init(&req->wait.entry); schedule_work(&req->work);