Re: [PATCH RESEND v2] fs/epoll: Remove unnecessary wakeups of nested epoll that in ET mode

[Jason Baron]

On 9/20/19 12:00 PM, Jason Baron wrote:
> On 9/19/19 5:24 AM, hev wrote:
>> From: Heiher <r@xxxxxx>
>>
>> Take the case where we have:
>>
>>         t0
>>          | (ew)
>>         e0
>>          | (et)
>>         e1
>>          | (lt)
>>         s0
>>
>> t0: thread 0
>> e0: epoll fd 0
>> e1: epoll fd 1
>> s0: socket fd 0
>> ew: epoll_wait
>> et: edge-trigger
>> lt: level-trigger
>>
>> When s0 fires an event, e1 catches the event, and then e0 catches an event from
>> e1. After this, There is a thread t0 do epoll_wait() many times on e0, it should
>> only get one event in total, because e1 is a dded to e0 in edge-triggered mode.
>>
>> This patch only allows the wakeup(&ep->poll_wait) in ep_scan_ready_list under
>> two conditions:
>>
>>  1. depth == 0.
>>  2. There have event is added to ep->ovflist during processing.
>>
>> Test code:
>>  #include <unistd.h>
>>  #include <sys/epoll.h>
>>  #include <sys/socket.h>
>>
>>  int main(int argc, char *argv[])
>>  {
>>  	int sfd[2];
>>  	int efd[2];
>>  	struct epoll_event e;
>>
>>  	if (socketpair(AF_UNIX, SOCK_STREAM, 0, sfd) < 0)
>>  		goto out;
>>
>>  	efd[0] = epoll_create(1);
>>  	if (efd[0] < 0)
>>  		goto out;
>>
>>  	efd[1] = epoll_create(1);
>>  	if (efd[1] < 0)
>>  		goto out;
>>
>>  	e.events = EPOLLIN;
>>  	if (epoll_ctl(efd[1], EPOLL_CTL_ADD, sfd[0], &e) < 0)
>>  		goto out;
>>
>>  	e.events = EPOLLIN | EPOLLET;
>>  	if (epoll_ctl(efd[0], EPOLL_CTL_ADD, efd[1], &e) < 0)
>>  		goto out;
>>
>>  	if (write(sfd[1], "w", 1) != 1)
>>  		goto out;
>>
>>  	if (epoll_wait(efd[0], &e, 1, 0) != 1)
>>  		goto out;
>>
>>  	if (epoll_wait(efd[0], &e, 1, 0) != 0)
>>  		goto out;
>>
>>  	close(efd[0]);
>>  	close(efd[1]);
>>  	close(sfd[0]);
>>  	close(sfd[1]);
>>
>>  	return 0;
>>
>>  out:
>>  	return -1;
>>  }
>>
>> More tests:
>>  https://github.com/heiher/epoll-wakeup
>>
>> Cc: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
>> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
>> Cc: Davide Libenzi <davidel@xxxxxxxxxxxxxxx>
>> Cc: Davidlohr Bueso <dave@xxxxxxxxxxxx>
>> Cc: Dominik Brodowski <linux@xxxxxxxxxxxxxxxxxxxx>
>> Cc: Eric Wong <e@xxxxxxxxx>
>> Cc: Jason Baron <jbaron@xxxxxxxxxx>
>> Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
>> Cc: Roman Penyaev <rpenyaev@xxxxxxx>
>> Cc: Sridhar Samudrala <sridhar.samudrala@xxxxxxxxx>
>> Cc: linux-kernel@xxxxxxxxxxxxxxx
>> Cc: linux-fsdevel@xxxxxxxxxxxxxxx
>> Signed-off-by: hev <r@xxxxxx>
>> ---
>>  fs/eventpoll.c | 5 ++++-
>>  1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/eventpoll.c b/fs/eventpoll.c
>> index c4159bcc05d9..fa71468dbd51 100644
>> --- a/fs/eventpoll.c
>> +++ b/fs/eventpoll.c
>> @@ -685,6 +685,9 @@ static __poll_t ep_scan_ready_list(struct eventpoll *ep,
>>  	if (!ep_locked)
>>  		mutex_lock_nested(&ep->mtx, depth);
>>  
>> +	if (!depth || list_empty_careful(&ep->rdllist))
>> +		pwake++;
>> +
>>  	/*
>>  	 * Steal the ready list, and re-init the original one to the
>>  	 * empty list. Also, set ep->ovflist to NULL so that events
>> @@ -755,7 +758,7 @@ static __poll_t ep_scan_ready_list(struct eventpoll *ep,
>>  		mutex_unlock(&ep->mtx);
>>  
>>  	/* We have to call this outside the lock */
>> -	if (pwake)
>> +	if (pwake == 2)
>>  		ep_poll_safewake(&ep->poll_wait);
>>  
>>  	return res;
>>
> 
> 
> Hi,
> 
> I was thinking more like the following. I tried it using your test-suite
> and it seems to work. What do you think?
> 
> Thanks,
> 
> -Jason
> 
> 
> diff --git a/fs/eventpoll.c b/fs/eventpoll.c
> index d7f1f50..662136b 100644
> --- a/fs/eventpoll.c
> +++ b/fs/eventpoll.c
> @@ -712,6 +712,15 @@ static __poll_t ep_scan_ready_list(struct eventpoll
> *ep,
>         for (nepi = READ_ONCE(ep->ovflist); (epi = nepi) != NULL;
>              nepi = epi->next, epi->next = EP_UNACTIVE_PTR) {
>                 /*
> +                * We only need to wakeup nested epoll fds if
> +                * if something has been queued to the overflow list,
> +                * since the ep_poll() traverses the rdllist during
> +                * recursive poll and thus events on the overflow list
> +                * may not be visible yet.
> +                */
> +               if (!pwake)
> +                       pwake++;
> +               /*
>                  * We need to check if the item is already in the list.
>                  * During the "sproc" callback execution time, items are
>                  * queued into ->ovflist but the "txlist" might already
> @@ -755,7 +764,7 @@ static __poll_t ep_scan_ready_list(struct eventpoll *ep,
>                 mutex_unlock(&ep->mtx);
> 
>         /* We have to call this outside the lock */
> -       if (pwake)
> +       if (pwake == 2)
>                 ep_poll_safewake(&ep->poll_wait);
> 
>         return res;
> 
> 


Also, probably better to not have that 'if' in the loop, so how about
the following?


diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index d7f1f50..ed0d8da 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -704,12 +704,21 @@ static __poll_t ep_scan_ready_list(struct
eventpoll *ep,
        res = (*sproc)(ep, &txlist, priv);

        write_lock_irq(&ep->lock);
+       nepi = READ_ONCE(ep->ovflist);
+       /*
+        * We only need to wakeup nested epoll fds if something has been
queued
+        * to the overflow list, since the ep_poll() traverses the rdllist
+        * during recursive poll and thus events on the overflow list
may not be
+        * visible yet.
+        */
+       if (nepi != NULL)
+               pwake++;
        /*
         * During the time we spent inside the "sproc" callback, some
         * other events might have been queued by the poll callback.
         * We re-insert them inside the main ready-list here.
         */
-       for (nepi = READ_ONCE(ep->ovflist); (epi = nepi) != NULL;
+       for (; (epi = nepi) != NULL;
             nepi = epi->next, epi->next = EP_UNACTIVE_PTR) {
                /*
                 * We need to check if the item is already in the list.
@@ -755,7 +764,7 @@ static __poll_t ep_scan_ready_list(struct eventpoll *ep,
                mutex_unlock(&ep->mtx);

        /* We have to call this outside the lock */
-       if (pwake)
+       if (pwake == 2)
                ep_poll_safewake(&ep->poll_wait);

        return res;