Commit 37b5e521 authored by Jason Baron's avatar Jason Baron Committed by Linus Torvalds
Browse files

epoll: remove ep_call_nested() from ep_eventpoll_poll() 

The use of ep_call_nested() in ep_eventpoll_poll(), which is the .poll
routine for an epoll fd, is used to prevent excessively deep epoll
nesting, and to prevent circular paths.

However, we are already preventing these conditions during
EPOLL_CTL_ADD.  In terms of too deep epoll chains, we do in fact allow
deep nesting of the epoll fds themselves (deeper than EP_MAX_NESTS),
however we don't allow more than EP_MAX_NESTS when an epoll file
descriptor is actually connected to a wakeup source.  Thus, we do not
require the use of ep_call_nested(), since ep_eventpoll_poll(), which is
called via ep_scan_ready_list() only continues nesting if there are
events available.

Since ep_call_nested() is implemented using a global lock, applications
that make use of nested epoll can see large performance improvements
with this change.

Davidlohr said:

: Improvements are quite obscene actually, such as for the following
: epoll_wait() benchmark with 2 level nesting on a 80 core IvyBridge:
:
: ncpus  vanilla     dirty     delta
: 1      2447092     3028315   +23.75%
: 4      231265      2986954   +1191.57%
: 8      121631      2898796   +2283.27%
: 16     59749       2902056   +4757.07%
: 32     26837	     2326314   +8568.30%
: 64     12926       1341281   +10276.61%
:
: (http://linux-scalability.org/epoll/epoll-test.c)

Link: http://lkml.kernel.org/r/1509430214-5599-1-git-send-email-jbaron@akamai.com


Signed-off-by: default avatarJason Baron <jbaron@akamai.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Salman Qazi <sqazi@google.com>
Cc: Hou Tao <houtao1@huawei.com>
Signed-off-by: default avatarAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
parent 57a173bd

fs/eventpoll.c

+35 −45
Original line number Diff line number Diff line
@@ -276,9 +276,6 @@ static DEFINE_MUTEX(epmutex); 
/* Used to check for epoll file descriptor inclusion loops */

static struct nested_calls poll_loop_ncalls;



/* Used to call file's f_op->poll() under the nested calls boundaries */

static struct nested_calls poll_readywalk_ncalls;



/* Slab cache used to allocate "struct epitem" */

static struct kmem_cache *epi_cache __read_mostly;
@@ -867,11 +864,33 @@ static int ep_eventpoll_release(struct inode *inode, struct file *file) 
	return 0;

}



static inline unsigned int ep_item_poll(struct epitem *epi, poll_table *pt)

static int ep_read_events_proc(struct eventpoll *ep, struct list_head *head,

			       void *priv);

static void ep_ptable_queue_proc(struct file *file, wait_queue_head_t *whead,

				 poll_table *pt);



/*

 * Differs from ep_eventpoll_poll() in that internal callers already have

 * the ep->mtx so we need to start from depth=1, such that mutex_lock_nested()

 * is correctly annotated.

 */

static unsigned int ep_item_poll(struct epitem *epi, poll_table *pt, int depth)

{

	struct eventpoll *ep;

	bool locked;



	pt->_key = epi->event.events;

	if (!is_file_epoll(epi->ffd.file))

		return epi->ffd.file->f_op->poll(epi->ffd.file, pt) &

		       epi->event.events;



	return epi->ffd.file->f_op->poll(epi->ffd.file, pt) & epi->event.events;

	ep = epi->ffd.file->private_data;

	poll_wait(epi->ffd.file, &ep->poll_wait, pt);

	locked = pt && (pt->_qproc == ep_ptable_queue_proc);



	return ep_scan_ready_list(epi->ffd.file->private_data,

				  ep_read_events_proc, &depth, depth,

				  locked) & epi->event.events;

}



static int ep_read_events_proc(struct eventpoll *ep, struct list_head *head,
@@ -879,13 +898,15 @@ static int ep_read_events_proc(struct eventpoll *ep, struct list_head *head, 
{

	struct epitem *epi, *tmp;

	poll_table pt;

	int depth = *(int *)priv;



	init_poll_funcptr(&pt, NULL);

	depth++;



	list_for_each_entry_safe(epi, tmp, head, rdllink) {

		if (ep_item_poll(epi, &pt))

		if (ep_item_poll(epi, &pt, depth)) {

			return POLLIN | POLLRDNORM;

		else {

		} else {

			/*

			 * Item has been dropped into the ready list by the poll

			 * callback, but it's not actually ready, as far as
@@ -899,48 +920,20 @@ static int ep_read_events_proc(struct eventpoll *ep, struct list_head *head, 
	return 0;

}



static void ep_ptable_queue_proc(struct file *file, wait_queue_head_t *whead,

				 poll_table *pt);



struct readyevents_arg {

	struct eventpoll *ep;

	bool locked;

};



static int ep_poll_readyevents_proc(void *priv, void *cookie, int call_nests)

{

	struct readyevents_arg *arg = priv;



	return ep_scan_ready_list(arg->ep, ep_read_events_proc, NULL,

				  call_nests + 1, arg->locked);

}



static unsigned int ep_eventpoll_poll(struct file *file, poll_table *wait)

{

	int pollflags;

	struct eventpoll *ep = file->private_data;

	struct readyevents_arg arg;



	/*

	 * During ep_insert() we already hold the ep->mtx for the tfile.

	 * Prevent re-aquisition.

	 */

	arg.locked = wait && (wait->_qproc == ep_ptable_queue_proc);

	arg.ep = ep;

	int depth = 0;



	/* Insert inside our poll wait queue */

	poll_wait(file, &ep->poll_wait, wait);



	/*

	 * Proceed to find out if wanted events are really available inside

	 * the ready list. This need to be done under ep_call_nested()

	 * supervision, since the call to f_op->poll() done on listed files

	 * could re-enter here.

	 * the ready list.

	 */

	pollflags = ep_call_nested(&poll_readywalk_ncalls, EP_MAX_NESTS,

				   ep_poll_readyevents_proc, &arg, ep, current);



	return pollflags != -1 ? pollflags : 0;

	return ep_scan_ready_list(ep, ep_read_events_proc,

				  &depth, depth, false);

}



#ifdef CONFIG_PROC_FS
@@ -1459,7 +1452,7 @@ static int ep_insert(struct eventpoll *ep, struct epoll_event *event, 
	 * this operation completes, the poll callback can start hitting

	 * the new item.

	 */

	revents = ep_item_poll(epi, &epq.pt);

	revents = ep_item_poll(epi, &epq.pt, 1);



	/*

	 * We have to check if something went wrong during the poll wait queue
@@ -1593,7 +1586,7 @@ static int ep_modify(struct eventpoll *ep, struct epitem *epi, struct epoll_even 
	 * Get current event bits. We can safely use the file* here because

	 * its usage count has been increased by the caller of this function.

	 */

	revents = ep_item_poll(epi, &pt);

	revents = ep_item_poll(epi, &pt, 1);



	/*

	 * If the item is "hot" and it is not registered inside the ready
@@ -1661,7 +1654,7 @@ static int ep_send_events_proc(struct eventpoll *ep, struct list_head *head, 


		list_del_init(&epi->rdllink);



		revents = ep_item_poll(epi, &pt);

		revents = ep_item_poll(epi, &pt, 1);



		/*

		 * If the event mask intersect the caller-requested one,
@@ -2307,9 +2300,6 @@ static int __init eventpoll_init(void) 
	ep_nested_calls_init(&poll_safewake_ncalls);

#endif



	/* Initialize the structure used to perform file's f_op->poll() calls */

	ep_nested_calls_init(&poll_readywalk_ncalls);



	/*

	 * We can have many thousands of epitems, so prevent this from

	 * using an extra cache line on 64-bit (and smaller) CPUs

CRA Git | Maintained and supported by SUSTech CRA and CCSE