From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ingo Molnar Subject: Re: [PATCH] poll: Avoid extra wakeups in select/poll Date: Wed, 29 Apr 2009 12:27:34 +0200 Message-ID: <20090429102734.GC2373@elte.hu> References: <49F3308B.1030507@cosmosbay.com> <20090426.020411.157511269.davem@davemloft.net> <49F43B8F.2050907@cosmosbay.com> <87ab60rh8t.fsf@basil.nowhere.org> <49F71B63.8010503@cosmosbay.com> <20090429091637.GA29874@elte.hu> <49F81FB9.50504@cosmosbay.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: linux kernel , Andi Kleen , David Miller , cl@linux.com, jesse.brandeburg@intel.com, netdev@vger.kernel.org, haoki@redhat.com, mchan@broadcom.com, davidel@xmailserver.org To: Eric Dumazet Return-path: Received: from mx3.mail.elte.hu ([157.181.1.138]:53103 "EHLO mx3.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751929AbZD2K2F (ORCPT ); Wed, 29 Apr 2009 06:28:05 -0400 Content-Disposition: inline In-Reply-To: <49F81FB9.50504@cosmosbay.com> Sender: netdev-owner@vger.kernel.org List-ID: * Eric Dumazet wrote: > Ingo Molnar a =E9crit : > > * Eric Dumazet wrote: > >=20 > >> @@ -418,8 +429,16 @@ int do_select(int n, fd_set_bits *fds, struct= timespec *end_time) > >> if (file) { > >> f_op =3D file->f_op; > >> mask =3D DEFAULT_POLLMASK; > >> - if (f_op && f_op->poll) > >> + if (f_op && f_op->poll) { > >> + if (wait) { > >> + wait->key =3D POLLEX_SET; > >> + if (in & bit) > >> + wait->key |=3D POLLIN_SET; > >> + if (out & bit) > >> + wait->key |=3D POLLOUT_SET; > >> + } > >> mask =3D (*f_op->poll)(file, retval ? NULL : wait); > >> + } > >> fput_light(file, fput_needed); > >> if ((mask & POLLIN_SET) && (in & bit)) { > >> res_in |=3D bit; > >=20 > > Please factor this whole 'if (file)' branch out into a helper.=20 > > Typical indentation levels go from 1 to 3 tabs - 4 should be avoide= d=20 > > if possible and 5 is pretty excessive already. This goes to eight. > >=20 >=20 > Thanks Ingo, >=20 > Here is v3 of patch, with your Acked-by included :) >=20 > This is IMHO clearer since helper immediatly follows POLLIN_SET / POL= LOUT_SET / > POLLEX_SET defines. >=20 > [PATCH] poll: Avoid extra wakeups in select/poll >=20 > After introduction of keyed wakeups Davide Libenzi did on epoll, we > are able to avoid spurious wakeups in poll()/select() code too. >=20 > For example, typical use of poll()/select() is to wait for incoming > network frames on many sockets. But TX completion for UDP/TCP=20 > frames call sock_wfree() which in turn schedules thread. >=20 > When scheduled, thread does a full scan of all polled fds and > can sleep again, because nothing is really available. If number > of fds is large, this cause significant load. >=20 > This patch makes select()/poll() aware of keyed wakeups and > useless wakeups are avoided. This reduces number of context > switches by about 50% on some setups, and work performed > by sofirq handlers. >=20 > Signed-off-by: Eric Dumazet > Acked-by: David S. Miller > Acked-by: Andi Kleen > Acked-by: Ingo Molnar > --- > fs/select.c | 40 ++++++++++++++++++++++++++++++++++++---- > include/linux/poll.h | 3 +++ > 2 files changed, 39 insertions(+), 4 deletions(-) >=20 > diff --git a/fs/select.c b/fs/select.c > index 0fe0e14..ba068ad 100644 > --- a/fs/select.c > +++ b/fs/select.c > @@ -168,7 +168,7 @@ static struct poll_table_entry *poll_get_entry(st= ruct poll_wqueues *p) > return table->entry++; > } > =20 > -static int pollwake(wait_queue_t *wait, unsigned mode, int sync, voi= d *key) > +static int __pollwake(wait_queue_t *wait, unsigned mode, int sync, v= oid *key) > { > struct poll_wqueues *pwq =3D wait->private; > DECLARE_WAITQUEUE(dummy_wait, pwq->polling_task); > @@ -194,6 +194,16 @@ static int pollwake(wait_queue_t *wait, unsigned= mode, int sync, void *key) > return default_wake_function(&dummy_wait, mode, sync, key); > } > =20 > +static int pollwake(wait_queue_t *wait, unsigned mode, int sync, voi= d *key) > +{ > + struct poll_table_entry *entry; > + > + entry =3D container_of(wait, struct poll_table_entry, wait); > + if (key && !((unsigned long)key & entry->key)) > + return 0; > + return __pollwake(wait, mode, sync, key); > +} > + > /* Add a new entry */ > static void __pollwait(struct file *filp, wait_queue_head_t *wait_ad= dress, > poll_table *p) > @@ -205,6 +215,7 @@ static void __pollwait(struct file *filp, wait_qu= eue_head_t *wait_address, > get_file(filp); > entry->filp =3D filp; > entry->wait_address =3D wait_address; > + entry->key =3D p->key; > init_waitqueue_func_entry(&entry->wait, pollwake); > entry->wait.private =3D pwq; > add_wait_queue(wait_address, &entry->wait); > @@ -362,6 +373,18 @@ get_max: > #define POLLOUT_SET (POLLWRBAND | POLLWRNORM | POLLOUT | POLLERR) > #define POLLEX_SET (POLLPRI) > =20 > +static void wait_key_set(poll_table *wait, unsigned long in, > + unsigned long out, unsigned long bit) > +{ > + if (wait) { > + wait->key =3D POLLEX_SET; > + if (in & bit) > + wait->key |=3D POLLIN_SET; > + if (out & bit) > + wait->key |=3D POLLOUT_SET; > + } > +} should be inline perhaps? > + > int do_select(int n, fd_set_bits *fds, struct timespec *end_time) > { > ktime_t expire, *to =3D NULL; > @@ -418,20 +441,25 @@ int do_select(int n, fd_set_bits *fds, struct t= imespec *end_time) > if (file) { > f_op =3D file->f_op; > mask =3D DEFAULT_POLLMASK; > - if (f_op && f_op->poll) > - mask =3D (*f_op->poll)(file, retval ? NULL : wait); > + if (f_op && f_op->poll) { > + wait_key_set(wait, in, out, bit); > + mask =3D (*f_op->poll)(file, wait); > + } > fput_light(file, fput_needed); > if ((mask & POLLIN_SET) && (in & bit)) { > res_in |=3D bit; > retval++; > + wait =3D NULL; > } > if ((mask & POLLOUT_SET) && (out & bit)) { > res_out |=3D bit; > retval++; > + wait =3D NULL; > } > if ((mask & POLLEX_SET) && (ex & bit)) { > res_ex |=3D bit; > retval++; > + wait =3D NULL; > } > } > } Looks much nicer now! [ I'd still suggest to factor out the guts of=20 do_select() as its nesting is excessive that hurts its reviewability=20 quite a bit - but now your patch does not make the situation any=20 worse. ] Even-More-Acked-by: Ingo Molnar Ingo