From: Mikulas Patocka <mpatocka@redhat.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
Ingo Molnar <mingo@elte.hu>,
Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
Ananth N Mavinakayanahalli <ananth@in.ibm.com>,
Anton Arapov <anton@redhat.com>,
linux-kernel@vger.kernel.org,
Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [PATCH 1/2] brw_mutex: big read-write mutex
Date: Fri, 19 Oct 2012 18:54:41 -0400 (EDT) [thread overview]
Message-ID: <Pine.LNX.4.64.1210191758570.12753@file.rdu.redhat.com> (raw)
In-Reply-To: <1350668451.2768.60.camel@twins>
On Fri, 19 Oct 2012, Peter Zijlstra wrote:
> > Yes, I tried this approach - it involves doing LOCK instruction on read
> > lock, remembering the cpu and doing another LOCK instruction on read
> > unlock (which will hopefully be on the same CPU, so no cacheline bouncing
> > happens in the common case). It was slower than the approach without any
> > LOCK instructions (43.3 seconds seconds for the implementation with
> > per-cpu LOCKed access, 42.7 seconds for this implementation without atomic
> > instruction; the benchmark involved doing 512-byte direct-io reads and
> > writes on a ramdisk with 8 processes on 8-core machine).
>
> So why is that a problem? Surely that's already tons better then what
> you've currently got.
Percpu rw-semaphores do not improve performance at all. I put them there
to avoid performance regression, not to improve performance.
All Linux kernels have a race condition - when you change block size of a
block device and you read or write the device at the same time, a crash
may happen. This bug is there since ever. Recently, this bug started to
cause major trouble - multiple high profile business sites report crashes
because of this race condition.
You can fix this race by using a read lock around I/O paths and write lock
around block size changing, but normal rw semaphore cause cache line
bouncing when taken for read by multiple processors and I/O performance
degradation because of it is measurable.
So I put this percpu-rw-semaphore there to fix the crashes and minimize
performance impact - on x86 it doesn't take any interlocked instructions
in the read path.
I don't quite understand why are people opposing to this and what do they
want to do instead? If you pull percpu-rw-semaphores out of the kernel,
you introduce a performance regression (raw device i/o will be slower on
3.7 than on 3.6, because on 3.6 it doesn't take any lock at all and on 3.7
it takes a read lock).
So you have options:
1) don't lock i/o just like on 3.6 and previous versions - you get a fast
kernel that randomly crashes
2) lock i/o with normal rw semaphore - you get a kernel that doesn't
crash, but that is slower than previous versions
3) lock i/o with percpu rw semaphore - you get kernel that is almost as
fast as previous kernels and that doesn't crash
For the users, the option 3) is the best. The users don't care whether it
looks ugly or not, they care about correctness and performance, that's
all.
Obviously, you can improve rw semaphores by adding lockdep annotations, or
by other things (turning rcu_read_lock/sychronize_rcu into
preempt_disable/synchronize_sched, by using barrier()-synchronize_sched()
instead of smp_mb()...), but I don't see a reason why do you want to hurt
users' experience by pulling it out and reverting to state 1) or 2) and
then, two kernel cycles later, come up with percpu-rw-semaphores again.
Mikulas
next prev parent reply other threads:[~2012-10-19 22:55 UTC|newest]
Thread overview: 103+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-10-15 19:09 [RFC PATCH 0/2] uprobes: register/unregister can race with fork Oleg Nesterov
2012-10-15 19:10 ` [PATCH 1/2] brw_mutex: big read-write mutex Oleg Nesterov
2012-10-15 23:28 ` Paul E. McKenney
2012-10-16 15:56 ` Oleg Nesterov
2012-10-16 18:58 ` Paul E. McKenney
2012-10-17 16:37 ` Oleg Nesterov
2012-10-17 22:28 ` Paul E. McKenney
2012-10-16 19:56 ` Linus Torvalds
2012-10-17 16:59 ` Oleg Nesterov
2012-10-17 22:44 ` Paul E. McKenney
2012-10-18 16:24 ` Oleg Nesterov
2012-10-18 16:38 ` Paul E. McKenney
2012-10-18 17:57 ` Oleg Nesterov
2012-10-18 19:28 ` Mikulas Patocka
2012-10-19 12:38 ` Peter Zijlstra
2012-10-19 15:32 ` Mikulas Patocka
2012-10-19 17:40 ` Peter Zijlstra
2012-10-19 17:57 ` Oleg Nesterov
2012-10-19 22:54 ` Mikulas Patocka [this message]
2012-10-24 3:08 ` Dave Chinner
2012-10-25 14:09 ` Mikulas Patocka
2012-10-25 23:40 ` Dave Chinner
2012-10-26 12:06 ` Oleg Nesterov
2012-10-26 13:22 ` Mikulas Patocka
2012-10-26 14:12 ` Oleg Nesterov
2012-10-26 15:23 ` mark_files_ro && sb_end_write Oleg Nesterov
2012-10-26 16:09 ` [PATCH 1/2] brw_mutex: big read-write mutex Mikulas Patocka
2012-10-19 17:49 ` Oleg Nesterov
2012-10-22 23:09 ` Mikulas Patocka
2012-10-23 15:12 ` Oleg Nesterov
2012-10-19 19:28 ` Paul E. McKenney
2012-10-22 23:36 ` [PATCH 0/2] fix and improvements for percpu-rw-semaphores (was: brw_mutex: big read-write mutex) Mikulas Patocka
2012-10-22 23:37 ` [PATCH 1/2] percpu-rw-semaphores: use light/heavy barriers Mikulas Patocka
2012-10-22 23:39 ` [PATCH 2/2] percpu-rw-semaphores: use rcu_read_lock_sched Mikulas Patocka
2012-10-24 16:16 ` Paul E. McKenney
2012-10-24 17:18 ` Oleg Nesterov
2012-10-24 18:20 ` Paul E. McKenney
2012-10-24 18:43 ` Oleg Nesterov
2012-10-24 19:43 ` Paul E. McKenney
2012-10-25 14:54 ` Mikulas Patocka
2012-10-25 15:07 ` Paul E. McKenney
2012-10-25 16:15 ` Mikulas Patocka
2012-10-23 16:59 ` [PATCH 1/2] percpu-rw-semaphores: use light/heavy barriers Oleg Nesterov
2012-10-23 18:05 ` Paul E. McKenney
2012-10-23 18:27 ` Oleg Nesterov
2012-10-23 18:41 ` Oleg Nesterov
2012-10-23 20:29 ` Paul E. McKenney
2012-10-23 20:32 ` Paul E. McKenney
2012-10-23 21:39 ` Mikulas Patocka
2012-10-24 16:23 ` Paul E. McKenney
2012-10-24 20:22 ` Mikulas Patocka
2012-10-24 20:36 ` Paul E. McKenney
2012-10-24 20:44 ` Mikulas Patocka
2012-10-24 23:57 ` Paul E. McKenney
2012-10-25 12:39 ` Paul E. McKenney
2012-10-25 13:48 ` Mikulas Patocka
2012-10-23 19:23 ` Oleg Nesterov
2012-10-23 20:45 ` Peter Zijlstra
2012-10-23 20:57 ` Peter Zijlstra
2012-10-24 15:11 ` Oleg Nesterov
2012-10-23 21:26 ` Mikulas Patocka
2012-10-23 20:32 ` Peter Zijlstra
2012-10-30 18:48 ` [PATCH 0/2] fix and improvements for percpu-rw-semaphores (was: brw_mutex: big read-write mutex) Oleg Nesterov
2012-10-31 19:41 ` [PATCH 0/1] percpu_rw_semaphore: reimplement to not block the readers unnecessarily Oleg Nesterov
2012-10-31 19:41 ` [PATCH 1/1] " Oleg Nesterov
2012-11-01 15:10 ` Linus Torvalds
2012-11-01 15:34 ` Oleg Nesterov
2012-11-02 18:06 ` [PATCH v2 0/1] " Oleg Nesterov
2012-11-02 18:06 ` [PATCH v2 1/1] " Oleg Nesterov
2012-11-07 17:04 ` [PATCH v3 " Mikulas Patocka
2012-11-07 17:47 ` Oleg Nesterov
2012-11-07 19:17 ` Mikulas Patocka
2012-11-08 13:42 ` Oleg Nesterov
2012-11-08 1:23 ` Paul E. McKenney
2012-11-08 1:16 ` [PATCH v2 " Paul E. McKenney
2012-11-08 13:33 ` Oleg Nesterov
2012-11-08 16:27 ` Paul E. McKenney
2012-11-08 13:48 ` [PATCH RESEND v2 0/1] " Oleg Nesterov
2012-11-08 13:48 ` [PATCH RESEND v2 1/1] " Oleg Nesterov
2012-11-08 20:07 ` Andrew Morton
2012-11-08 21:08 ` Paul E. McKenney
2012-11-08 23:41 ` Mikulas Patocka
2012-11-09 0:41 ` Paul E. McKenney
2012-11-09 3:23 ` Paul E. McKenney
2012-11-09 16:35 ` Oleg Nesterov
2012-11-09 16:59 ` Paul E. McKenney
2012-11-09 12:47 ` Mikulas Patocka
2012-11-09 15:46 ` Oleg Nesterov
2012-11-09 17:01 ` Paul E. McKenney
2012-11-09 18:10 ` Oleg Nesterov
2012-11-09 18:19 ` Oleg Nesterov
2012-11-10 0:55 ` Paul E. McKenney
2012-11-11 15:45 ` Oleg Nesterov
2012-11-12 18:38 ` Paul E. McKenney
2012-11-11 18:27 ` [PATCH -mm] percpu_rw_semaphore-reimplement-to-not-block-the-readers-unnecessari ly.fix Oleg Nesterov
2012-11-12 18:31 ` Paul E. McKenney
2012-11-16 23:22 ` Andrew Morton
2012-11-18 19:32 ` Oleg Nesterov
2012-11-01 15:43 ` [PATCH 1/1] percpu_rw_semaphore: reimplement to not block the readers unnecessarily Paul E. McKenney
2012-11-01 18:33 ` Oleg Nesterov
2012-11-02 16:18 ` Oleg Nesterov
2012-10-15 19:10 ` [PATCH 2/2] uprobes: Use brw_mutex to fix register/unregister vs dup_mmap() race Oleg Nesterov
2012-10-18 7:03 ` Srikar Dronamraju
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=Pine.LNX.4.64.1210191758570.12753@file.rdu.redhat.com \
--to=mpatocka@redhat.com \
--cc=ananth@in.ibm.com \
--cc=anton@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=oleg@redhat.com \
--cc=paulmck@linux.vnet.ibm.com \
--cc=peterz@infradead.org \
--cc=srikar@linux.vnet.ibm.com \
--cc=tglx@linutronix.de \
--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).