From: "Rafael J. Wysocki" <rjw@sisk.pl>
To: Jiri Slaby <jslaby@suse.cz>
Cc: "Arve Hjønnevåg" <arve@android.com>, NeilBrown <neilb@suse.de>,
LKML <linux-kernel@vger.kernel.org>,
"Jiri Slaby" <jirislaby@gmail.com>,
"Linux PM list" <linux-pm@vger.kernel.org>
Subject: Re: [-next regression] TCP window full with EPOLLWAKEUP
Date: Sun, 20 May 2012 14:53:11 +0200 [thread overview]
Message-ID: <201205201453.11794.rjw@sisk.pl> (raw)
In-Reply-To: <201205201423.08334.rjw@sisk.pl>
On Sunday, May 20, 2012, Rafael J. Wysocki wrote:
> On Sunday, May 20, 2012, Jiri Slaby wrote:
> > Hi,
> >
> > a bisection shows that with the following commit from -next:
> > commit 4d7e30d98939a0340022ccd49325a3d70f7e0238
> > Author: Arve Hjønnevåg <arve@android.com>
> > Date: Tue May 1 21:33:34 2012 +0200
> >
> > epoll: Add a flag, EPOLLWAKEUP, to prevent suspend while epoll
> > events are ready
> >
> > ====
> >
> > one of mono programs I use stops receiving data from the network.
> > Wireshark shows that the TCP window of a connection is filled. This
> > means the program does not read the data fast enough after requesting
> > the data.
> >
> > If I revert that commit on the top of -next (20120518), everything works
> > as expected.
>
> Hmm. I suppose that the failing program doesn't set EPOLLWAKEUP by mistake,
> does it?
If it doesn't, we can assume that epi-ws is always NULL and all of the added
overhead comes from the function calls. So, I wonder if the appended patch
makes any difference?
Rafael
---
fs/eventpoll.c | 35 ++++++++++++++++++++++++-----------
1 file changed, 24 insertions(+), 11 deletions(-)
Index: linux/fs/eventpoll.c
===================================================================
--- linux.orig/fs/eventpoll.c
+++ linux/fs/eventpoll.c
@@ -597,7 +597,8 @@ static int ep_scan_ready_list(struct eve
*/
if (!ep_is_linked(&epi->rdllink)) {
list_add_tail(&epi->rdllink, &ep->rdllist);
- __pm_stay_awake(epi->ws);
+ if (epi->ws)
+ __pm_stay_awake(epi->ws);
}
}
/*
@@ -611,7 +612,8 @@ static int ep_scan_ready_list(struct eve
* Quickly re-inject items left on "txlist".
*/
list_splice(&txlist, &ep->rdllist);
- __pm_relax(ep->ws);
+ if (ep->ws)
+ __pm_relax(ep->ws);
if (!list_empty(&ep->rdllist)) {
/*
@@ -750,7 +752,9 @@ static int ep_read_events_proc(struct ev
* callback, but it's not actually ready, as far as
* caller requested events goes. We can remove it here.
*/
- __pm_relax(epi->ws);
+ if (epi->ws)
+ __pm_relax(epi->ws);
+
list_del_init(&epi->rdllink);
}
}
@@ -956,7 +960,8 @@ static int ep_poll_callback(wait_queue_t
/* If this file is already in the ready list we exit soon */
if (!ep_is_linked(&epi->rdllink)) {
list_add_tail(&epi->rdllink, &ep->rdllist);
- __pm_stay_awake(epi->ws);
+ if (epi->ws)
+ __pm_stay_awake(epi->ws);
}
/*
@@ -1219,7 +1224,8 @@ static int ep_insert(struct eventpoll *e
/* If the file is already "ready" we drop it inside the ready list */
if ((revents & event->events) && !ep_is_linked(&epi->rdllink)) {
list_add_tail(&epi->rdllink, &ep->rdllist);
- __pm_stay_awake(epi->ws);
+ if (epi->ws)
+ __pm_stay_awake(epi->ws);
/* Notify waiting tasks that events are available */
if (waitqueue_active(&ep->wq))
@@ -1309,7 +1315,8 @@ static int ep_modify(struct eventpoll *e
spin_lock_irq(&ep->lock);
if (!ep_is_linked(&epi->rdllink)) {
list_add_tail(&epi->rdllink, &ep->rdllist);
- __pm_stay_awake(epi->ws);
+ if (epi->ws)
+ __pm_stay_awake(epi->ws);
/* Notify waiting tasks that events are available */
if (waitqueue_active(&ep->wq))
@@ -1357,9 +1364,12 @@ static int ep_send_events_proc(struct ev
* instead, but then epi->ws would temporarily be out of sync
* with ep_is_linked().
*/
- if (epi->ws && epi->ws->active)
- __pm_stay_awake(ep->ws);
- __pm_relax(epi->ws);
+ if (epi->ws) {
+ if (epi->ws->active)
+ __pm_stay_awake(ep->ws);
+
+ __pm_relax(epi->ws);
+ }
list_del_init(&epi->rdllink);
pt._key = epi->event.events;
@@ -1376,7 +1386,9 @@ static int ep_send_events_proc(struct ev
if (__put_user(revents, &uevent->events) ||
__put_user(epi->event.data, &uevent->data)) {
list_add(&epi->rdllink, head);
- __pm_stay_awake(epi->ws);
+ if (epi->ws)
+ __pm_stay_awake(epi->ws);
+
return eventcnt ? eventcnt : -EFAULT;
}
eventcnt++;
@@ -1396,7 +1408,8 @@ static int ep_send_events_proc(struct ev
* poll callback will queue them in ep->ovflist.
*/
list_add_tail(&epi->rdllink, &ep->rdllist);
- __pm_stay_awake(epi->ws);
+ if (epi->ws)
+ __pm_stay_awake(epi->ws);
}
}
}
next prev parent reply other threads:[~2012-05-20 12:48 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-05-19 22:06 [-next regression] TCP window full with EPOLLWAKEUP Jiri Slaby
2012-05-20 12:23 ` Rafael J. Wysocki
2012-05-20 12:53 ` Rafael J. Wysocki [this message]
2012-05-20 18:34 ` Rafael J. Wysocki
2012-05-21 13:55 ` Jiri Slaby
2012-05-21 19:11 ` Rafael J. Wysocki
2012-05-21 19:28 ` [PATCH] epoll: Fix user space breakage related to EPOLLWAKEUP (was: Re: [-next regression] TCP window full with EPOLLWAKEUP) Rafael J. Wysocki
2012-05-21 22:11 ` Arve Hjønnevåg
2012-05-22 15:56 ` Rafael J. Wysocki
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=201205201453.11794.rjw@sisk.pl \
--to=rjw@sisk.pl \
--cc=arve@android.com \
--cc=jirislaby@gmail.com \
--cc=jslaby@suse.cz \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pm@vger.kernel.org \
--cc=neilb@suse.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.