From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: David Miller <davem@davemloft.net>,
kaber@trash.net, torvalds@linux-foundation.org,
shemminger@vyatta.com, dada1@cosmosbay.com,
jeff.chua.linux@gmail.com, paulus@samba.org, mingo@elte.hu,
laijs@cn.fujitsu.com, jengelh@medozas.de, r000n@r000n.net,
linux-kernel@vger.kernel.org, netfilter-devel@vger.kernel.org,
netdev@vger.kernel.org, benh@kernel.crashing.org,
mathieu.desnoyers@polymtl.ca
Subject: Re: [PATCH] netfilter: use per-cpu spinlock rather than RCU (v3)
Date: Fri, 17 Apr 2009 09:33:52 -0700 [thread overview]
Message-ID: <20090417163351.GC6742@linux.vnet.ibm.com> (raw)
In-Reply-To: <1239948734.23397.4052.camel@laptop>
On Fri, Apr 17, 2009 at 08:12:14AM +0200, Peter Zijlstra wrote:
> On Thu, 2009-04-16 at 18:28 -0700, Paul E. McKenney wrote:
> > On Thu, Apr 16, 2009 at 04:49:55PM -0700, Paul E. McKenney wrote:
> > > On Thu, Apr 16, 2009 at 03:33:54PM -0700, David Miller wrote:
> > > > From: Patrick McHardy <kaber@trash.net>
> > > > Date: Thu, 16 Apr 2009 15:11:31 +0200
> > > >
> > > > > Linus Torvalds wrote:
> > > > >> On Wed, 15 Apr 2009, Stephen Hemminger wrote:
> > > > >>> The counters are the bigger problem, otherwise we could just free
> > > > >>> table
> > > > >>> info via rcu. Do we really have to support: replace where the counter
> > > > >>> values coming out to user space are always exactly accurate, or is it
> > > > >>> allowed to replace a rule and maybe lose some counter ticks (worst
> > > > >>> case
> > > > >>> NCPU-1).
> > > > >> Why not just read the counters fromt he old one at RCU free time (they
> > > > >> are guaranteed to be stable at that point, since we're all done with
> > > > >> those entries), and apply them at that point to the current setup?
> > > > >
> > > > > We need the counters immediately to copy them to userspace, so waiting
> > > > > for an asynchronous RCU free is not going to work.
> > > >
> > > > It just occurred to me that since all netfilter packet handling
> > > > goes through one place, we could have a sort-of "netfilter RCU"
> > > > of sorts to solve this problem.
> > >
> > > OK, I am putting one together...
> > >
> > > It will be needed sooner or later, though I suspect per-CPU locking
> > > would work fine in this case.
> >
> > And here is a crude first cut. Untested, probably does not even compile.
> >
> > Straight conversion of Mathieu Desnoyers's user-space RCU implementation
> > at git://lttng.org/userspace-rcu.git to the kernel (and yes, I did help
> > a little, but he must bear the bulk of the guilt). Pick on srcu.h
> > and srcu.c out of sheer laziness. User-space testing gives deep
> > sub-microsecond grace-period latencies, so should be fast enough, at
> > least if you don't mind two smp_call_function() invocations per grace
> > period and spinning on each instance of a per-CPU variable.
> >
> > Again, I believe per-CPU locking should work fine for the netfilter
> > counters, but I guess "friends don't let friends use hashed locks".
> > (I would not know for sure, never having used them myself, except of
> > course to protect hash tables.)
> >
> > Most definitely -not- for inclusion at this point. Next step is to hack
> > up the relevant rcutorture code and watch it explode on contact. ;-)
>
> One comment, its again a global thing..
>
> I've been playing with the idea for a while now to make all RCU
> implementations into proper objects so that you can do things like:
>
> struct atomic_rcu_domain my_rcu_domain = create_atomic_rcu();
>
> atomic_rcu_read_lock(&my_rcu_domain());
> ...
>
> atomic_rcu_read_unlock(&my_rcu_domain());
>
> and
>
> call_atomic_rcu(&my_rcu_domain, &my_obj->rcu_head, do_something);
>
> etc..
>
> We would have:
>
> atomic_rcu -- 'classic' non preemptible RCU (treercu these days)
> sleep_rcu -- 'preemptible' RCU
>
> Then have 3 default domains:
>
> sched_rcu -- always atomic_rcu
This is the call_rcu_sched() variant.
> rcu -- depends on PREEMPT_RCU
This is the call_rcu() variant.
> preempt_rcu -- always sleep_rcu
I guess that this one could allow sleeping on mutexes... Does anyone
need to do that?
> This would allow generic code to:
> 1) use preemptible RCU for those cases where needed
> 2) create smaller RCU domains where needed, such as in this case
> 3) mostly do away with SRCU
#3 would be good! But...
At an API level, there are two differences between SRCU and the other
RCU implementations:
a. The return value from srcu_read_lock() is passed to
srcu_read_unlock().
b. There is a control block passed in to each SRCU primitive.
Difference (a) could potentially be taken care of with a few tricks I
am trying in the process of getting preemptrcu merged into treercu.
Your approach to (b) certainly makes it uniform, there are >500
occurrences of rcu_read_lock() and rcu_read_unlock() each, but only
a very few occurrences of srcu_read_lock() and srcu_read_unlock()
(like exactly one each!). So adding an argument to rcu_read_lock()
does not sound at all reasonable.
> Now I realize that the presented RCU implementation has a different
> grace period method than the existing ones that use the timer tick to
> drive the state machine, so 2) might not be too relevant here. But maybe
> we can do something with different grace periods too.
>
> Anyway, just an idea because I always get a little offended at the hard
> coded global variables in all these RCU implementations :-)
I am thinking in terms of adding a synchronize_rcu_bh() with the desired
properties. That way we avoid yet another RCU flavor. (What can I say?
I got carried away!) Also, since the rcu-bh flavor is used only by
networking, we have a fair amount of freedom to tweak it. It will take
longer than introducing a new flavor, but Steve Hemminger has a good
solution already, and RCU really isn't the thing to do quick hacks on.
Thanx, Paul
next prev parent reply other threads:[~2009-04-17 16:33 UTC|newest]
Thread overview: 215+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <Pine.LNX.4.64.0904101656190.2093@boston.corp.fedex.com>
[not found] ` <20090410095246.4fdccb56@s6510>
2009-04-11 1:25 ` iptables very slow after commit784544739a25c30637397ace5489eeb6e15d7d49 David Miller
2009-04-11 1:39 ` iptables very slow after commit 784544739a25c30637397ace5489eeb6e15d7d49 Linus Torvalds
2009-04-11 4:15 ` Paul E. McKenney
2009-04-11 5:14 ` Jan Engelhardt
2009-04-11 5:42 ` Paul E. McKenney
2009-04-11 6:00 ` David Miller
2009-04-11 18:12 ` Kyle Moffett
2009-04-11 18:32 ` Arkadiusz Miskiewicz
2009-04-12 0:54 ` david
2009-04-12 5:05 ` Kyle Moffett
2009-04-12 12:30 ` Harald Welte
2009-04-12 16:38 ` Jan Engelhardt
2009-04-11 15:07 ` Stephen Hemminger
2009-04-11 16:05 ` Jeff Chua
2009-04-11 17:51 ` Linus Torvalds
2009-04-11 7:08 ` Ingo Molnar
2009-04-11 15:05 ` Stephen Hemminger
2009-04-11 17:48 ` Paul E. McKenney
2009-04-12 10:54 ` Ingo Molnar
2009-04-12 11:34 ` Paul Mackerras
2009-04-12 17:31 ` Paul E. McKenney
2009-04-13 1:13 ` David Miller
2009-04-13 4:04 ` Paul E. McKenney
2009-04-13 16:53 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU Stephen Hemminger
2009-04-13 17:40 ` Eric Dumazet
2009-04-13 18:11 ` Stephen Hemminger
2009-04-13 19:06 ` Martin Josefsson
2009-04-13 19:17 ` Linus Torvalds
2009-04-13 22:24 ` Andrew Morton
2009-04-13 23:20 ` Stephen Hemminger
2009-04-13 23:26 ` Andrew Morton
2009-04-13 23:37 ` Linus Torvalds
2009-04-13 23:52 ` Ingo Molnar
2009-04-14 12:27 ` Patrick McHardy
2009-04-14 14:23 ` Eric Dumazet
2009-04-14 14:45 ` Stephen Hemminger
2009-04-14 15:49 ` Eric Dumazet
2009-04-14 16:51 ` Jeff Chua
2009-04-14 18:17 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU (v2) Stephen Hemminger
2009-04-14 19:28 ` Eric Dumazet
2009-04-14 21:11 ` Stephen Hemminger
2009-04-14 21:13 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU (v3) Stephen Hemminger
2009-04-14 21:40 ` Eric Dumazet
2009-04-15 10:59 ` Patrick McHardy
2009-04-15 16:31 ` Stephen Hemminger
2009-04-15 20:55 ` Stephen Hemminger
2009-04-15 21:07 ` Eric Dumazet
2009-04-15 21:55 ` Jan Engelhardt
2009-04-16 12:12 ` Patrick McHardy
2009-04-16 12:24 ` Jan Engelhardt
2009-04-16 12:31 ` Patrick McHardy
2009-04-15 21:57 ` [PATCH] netfilter: use per-cpu rwlock rather than RCU (v4) Stephen Hemminger
2009-04-15 23:48 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU (v3) David Miller
2009-04-16 0:01 ` Stephen Hemminger
2009-04-16 0:05 ` David Miller
2009-04-16 12:28 ` Patrick McHardy
2009-04-16 0:10 ` Linus Torvalds
2009-04-16 0:45 ` [PATCH] netfilter: use per-cpu spinlock and RCU (v5) Stephen Hemminger
2009-04-16 5:01 ` Eric Dumazet
2009-04-16 13:53 ` Patrick McHardy
2009-04-16 14:47 ` Paul E. McKenney
2009-04-16 16:10 ` [PATCH] netfilter: use per-cpu recursive spinlock (v6) Eric Dumazet
2009-04-16 16:20 ` Eric Dumazet
2009-04-16 16:37 ` Linus Torvalds
2009-04-16 16:59 ` Patrick McHardy
2009-04-16 17:58 ` Paul E. McKenney
2009-04-16 18:41 ` Eric Dumazet
2009-04-16 20:49 ` [PATCH[] netfilter: use per-cpu reader-writer lock (v0.7) Stephen Hemminger
2009-04-16 21:02 ` Linus Torvalds
2009-04-16 23:04 ` Ingo Molnar
2009-04-17 0:13 ` [PATCH] netfilter: use per-cpu recursive spinlock (v6) Paul E. McKenney
2009-04-16 13:11 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU (v3) Patrick McHardy
2009-04-16 22:33 ` David Miller
2009-04-16 23:49 ` Paul E. McKenney
2009-04-16 23:52 ` [PATCH] netfilter: per-cpu spin-lock with recursion (v0.8) Stephen Hemminger
2009-04-17 0:15 ` Jeff Chua
2009-04-17 5:55 ` Peter Zijlstra
2009-04-17 6:03 ` Eric Dumazet
2009-04-17 6:14 ` Eric Dumazet
2009-04-17 17:08 ` Peter Zijlstra
2009-04-17 11:17 ` Patrick McHardy
2009-04-17 1:28 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU (v3) Paul E. McKenney
2009-04-17 2:19 ` Mathieu Desnoyers
2009-04-17 5:05 ` Paul E. McKenney
2009-04-17 5:44 ` Mathieu Desnoyers
2009-04-17 14:51 ` Paul E. McKenney
2009-04-17 4:50 ` Stephen Hemminger
2009-04-17 5:08 ` Paul E. McKenney
2009-04-17 5:16 ` Eric Dumazet
2009-04-17 5:40 ` Paul E. McKenney
2009-04-17 8:07 ` David Miller
2009-04-17 15:00 ` Paul E. McKenney
2009-04-17 17:22 ` Peter Zijlstra
2009-04-17 17:32 ` Linus Torvalds
2009-04-17 6:12 ` Peter Zijlstra
2009-04-17 16:33 ` Paul E. McKenney [this message]
2009-04-17 16:51 ` Peter Zijlstra
2009-04-17 21:29 ` Paul E. McKenney
2009-04-18 9:40 ` Evgeniy Polyakov
2009-04-18 14:14 ` Paul E. McKenney
2009-04-20 17:34 ` [PATCH] netfilter: use per-cpu recursive lock (v10) Stephen Hemminger
2009-04-20 18:21 ` Paul E. McKenney
2009-04-20 18:25 ` Eric Dumazet
2009-04-20 20:32 ` Stephen Hemminger
2009-04-20 20:42 ` Stephen Hemminger
2009-04-20 21:05 ` Paul E. McKenney
2009-04-20 21:23 ` Paul Mackerras
2009-04-20 21:58 ` Paul E. McKenney
2009-04-20 22:41 ` Paul Mackerras
2009-04-20 23:01 ` [PATCH] netfilter: use per-cpu recursive lock (v11) Stephen Hemminger
2009-04-21 3:41 ` Lai Jiangshan
2009-04-21 3:56 ` Eric Dumazet
2009-04-21 4:15 ` Stephen Hemminger
2009-04-21 5:22 ` Lai Jiangshan
2009-04-21 5:45 ` Stephen Hemminger
2009-04-21 6:52 ` Lai Jiangshan
2009-04-21 8:16 ` Evgeniy Polyakov
2009-04-21 8:42 ` Lai Jiangshan
2009-04-21 8:49 ` David Miller
2009-04-21 8:55 ` Eric Dumazet
2009-04-21 9:22 ` Evgeniy Polyakov
2009-04-21 9:34 ` Lai Jiangshan
2009-04-21 5:34 ` Lai Jiangshan
2009-04-21 4:59 ` Eric Dumazet
2009-04-21 16:37 ` Paul E. McKenney
2009-04-21 5:46 ` Lai Jiangshan
2009-04-21 16:13 ` Linus Torvalds
2009-04-21 16:43 ` Stephen Hemminger
2009-04-21 16:50 ` Linus Torvalds
2009-04-21 18:02 ` Ingo Molnar
2009-04-21 18:15 ` Stephen Hemminger
2009-04-21 19:10 ` Ingo Molnar
2009-04-21 19:46 ` Eric Dumazet
2009-04-22 7:35 ` Ingo Molnar
2009-04-22 8:53 ` Eric Dumazet
2009-04-22 10:13 ` Jarek Poplawski
2009-04-22 11:26 ` Ingo Molnar
2009-04-22 11:39 ` Jarek Poplawski
2009-04-22 11:18 ` Ingo Molnar
2009-04-22 15:19 ` Linus Torvalds
2009-04-22 16:57 ` Eric Dumazet
2009-04-22 17:18 ` Linus Torvalds
2009-04-22 20:46 ` Jarek Poplawski
2009-04-22 17:48 ` Ingo Molnar
2009-04-21 21:04 ` Stephen Hemminger
2009-04-22 8:00 ` Ingo Molnar
2009-04-21 19:39 ` Ingo Molnar
2009-04-21 21:39 ` [PATCH] netfilter: use per-cpu recursive lock (v13) Stephen Hemminger
2009-04-22 4:17 ` Paul E. McKenney
2009-04-22 14:57 ` Eric Dumazet
2009-04-22 15:32 ` Linus Torvalds
2009-04-24 4:09 ` [PATCH] netfilter: use per-CPU recursive lock {XIV} Stephen Hemminger
2009-04-24 4:58 ` Eric Dumazet
2009-04-24 15:33 ` Patrick McHardy
2009-04-24 16:18 ` Stephen Hemminger
2009-04-24 20:43 ` Jarek Poplawski
2009-04-25 20:30 ` [PATCH] netfilter: iptables no lockdep is needed Stephen Hemminger
2009-04-26 8:18 ` Jarek Poplawski
2009-04-26 18:24 ` [PATCH] netfilter: use per-CPU recursive lock {XV} Eric Dumazet
2009-04-26 18:56 ` Mathieu Desnoyers
2009-04-26 21:57 ` Stephen Hemminger
2009-04-26 22:32 ` Mathieu Desnoyers
2009-04-27 17:44 ` Peter Zijlstra
2009-04-27 18:30 ` [PATCH] netfilter: use per-CPU r**ursive " Stephen Hemminger
2009-04-27 18:54 ` Ingo Molnar
2009-04-27 19:06 ` Stephen Hemminger
2009-04-27 19:46 ` Linus Torvalds
2009-04-27 19:48 ` Linus Torvalds
2009-04-27 20:36 ` Evgeniy Polyakov
2009-04-27 20:58 ` Linus Torvalds
2009-04-27 21:40 ` Stephen Hemminger
2009-04-27 22:24 ` Linus Torvalds
2009-04-27 23:01 ` Linus Torvalds
2009-04-27 23:03 ` Linus Torvalds
2009-04-28 6:58 ` Eric Dumazet
2009-04-28 11:53 ` David Miller
2009-04-28 12:40 ` Ingo Molnar
2009-04-28 13:43 ` David Miller
2009-04-28 13:52 ` Mathieu Desnoyers
2009-04-28 14:37 ` David Miller
2009-04-28 14:49 ` Mathieu Desnoyers
2009-04-28 15:00 ` David Miller
2009-04-28 16:24 ` [PATCH] netfilter: revised locking for x_tables Stephen Hemminger
2009-04-28 16:50 ` Linus Torvalds
2009-04-28 16:55 ` Linus Torvalds
2009-04-29 5:37 ` David Miller
[not found] ` <20090428.223708.168741998.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2009-04-30 3:26 ` Jeff Chua
[not found] ` <b6a2187b0904292026k7d6107a7vcdc761d4149f40aa-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2009-04-30 3:31 ` David Miller
2009-05-01 8:38 ` [PATCH] netfilter: use likely() in xt_info_rdlock_bh() Eric Dumazet
2009-05-01 16:10 ` David Miller
2009-04-28 15:42 ` [PATCH] netfilter: use per-CPU r**ursive lock {XV} Paul E. McKenney
2009-04-28 17:35 ` Christoph Lameter
2009-04-28 15:09 ` Linus Torvalds
2009-04-27 23:32 ` Linus Torvalds
2009-04-28 7:41 ` Peter Zijlstra
2009-04-28 14:22 ` Paul E. McKenney
2009-04-28 7:42 ` Jan Engelhardt
2009-04-26 19:31 ` [PATCH] netfilter: use per-CPU recursive " Mathieu Desnoyers
2009-04-26 20:55 ` Eric Dumazet
2009-04-26 21:39 ` Mathieu Desnoyers
2009-04-21 18:34 ` [PATCH] netfilter: use per-cpu recursive lock (v11) Paul E. McKenney
2009-04-21 20:14 ` Linus Torvalds
2009-04-20 23:44 ` [PATCH] netfilter: use per-cpu recursive lock (v10) Paul E. McKenney
2009-04-16 0:02 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU (v3) Linus Torvalds
2009-04-16 6:26 ` Eric Dumazet
2009-04-16 14:33 ` Paul E. McKenney
2009-04-15 3:23 ` David Miller
2009-04-14 17:19 ` [PATCH] netfilter: use per-cpu spinlock rather than RCU Stephen Hemminger
2009-04-11 15:50 ` iptables very slow after commit 784544739a25c30637397ace5489eeb6e15d7d49 Stephen Hemminger
2009-04-11 17:43 ` Paul E. McKenney
2009-04-11 18:57 ` Linus Torvalds
2009-04-12 0:34 ` Paul E. McKenney
2009-04-12 7:23 ` Evgeniy Polyakov
2009-04-12 16:06 ` Stephen Hemminger
2009-04-12 17:30 ` Paul E. McKenney
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=20090417163351.GC6742@linux.vnet.ibm.com \
--to=paulmck@linux.vnet.ibm.com \
--cc=benh@kernel.crashing.org \
--cc=dada1@cosmosbay.com \
--cc=davem@davemloft.net \
--cc=jeff.chua.linux@gmail.com \
--cc=jengelh@medozas.de \
--cc=kaber@trash.net \
--cc=laijs@cn.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mathieu.desnoyers@polymtl.ca \
--cc=mingo@elte.hu \
--cc=netdev@vger.kernel.org \
--cc=netfilter-devel@vger.kernel.org \
--cc=paulus@samba.org \
--cc=peterz@infradead.org \
--cc=r000n@r000n.net \
--cc=shemminger@vyatta.com \
--cc=torvalds@linux-foundation.org \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).