From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: One Thousand Gnomes <gnomes@lxorguk.ukuu.org.uk>
Cc: Peter Hurley <peter@hurleysoftware.com>,
"H. Peter Anvin" <hpa@zytor.com>,
David Laight <David.Laight@ACULAB.COM>,
Jakub Jelinek <jakub@redhat.com>,
"linux-arch@vger.kernel.org" <linux-arch@vger.kernel.org>,
Tony Luck <tony.luck@intel.com>,
"linux-ia64@vger.kernel.org" <linux-ia64@vger.kernel.org>,
Oleg Nesterov <oleg@redhat.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Paul Mackerras <paulus@samba.org>,
"linuxppc-dev@lists.ozlabs.org" <linuxppc-dev@lists.ozlabs.org>,
Miroslav Franc <mfranc@redhat.com>,
Richard Henderson <rth@twiddle.net>,
linux-alpha@vger.kernel.org, ink@jurassic.park.msu.ru,
mattst88@gmail.com
Subject: Re: bit fields && data tearing
Date: Mon, 22 Sep 2014 12:51:06 -0700 [thread overview]
Message-ID: <20140922195106.GR4723@linux.vnet.ibm.com> (raw)
In-Reply-To: <20140915002427.6ce18965@alan.etchedpixels.co.uk>
On Mon, Sep 15, 2014 at 12:24:27AM +0100, One Thousand Gnomes wrote:
> > So a problem that no one has ever complained about on _any_ arch is suddenly
> > a problem on a subset of Alpha cpus, but a problem I know exists on Alpha
> > isn't important because no one's filed a bug about it?
>
> Yes - because if you think about it that tells you that nobody is hitting
> it with the old code and it probably doesn't matter.
>
> > The only Alpha person in this discussion has come out clearly in favor
> > of dropping EV4/5 support.
>
> That's not a statistically valid sample size btw
OK, adding the other two Alpha Port maintainers on CC.
Attempted summary for their benefit:
o There was a bug involving older Alpha CPUs using 32-bit
memory-reference operations to do smaller memory accesses.
The suggested resolution was to use set_bit().
o Peter Hurley called out a number of theoretical issues with
CPUs lacking 8-bit and 16-bit memory-reference instructions,
for example, adjacent 8-bit variables protected by different
locks not being safe on such CPUs.
o Michael Cree pointed out that some of these issues actually
happen in the X server ever since the libpciaccess change.
Michael would like to compile for Alpha with BWX (thus allowing
8-bit and 16-bit memory references, but disallowing pre-EV56
CPUs) in order make the X server (and thus Debian) work
better on newer Alpha CPUs.
Given that Michael Cree maintains the list of Alpha systems
running Linux, I took this as my cue to provide a couple of
patches to that effect.
o Michael Cree also noted that pre-EV56 Alpha CPUs really can
do 8-bit and 16-bit accesses in an SMP-safe manner via LL/SC,
but that this requires some hacking on the compilers.
o Alan Cox argued that we should support pre-EV56 Alpha CPUs
without any special defense against issues that might arise
from their lack of 8-bit and 16-bit memory-reference
instructions, as you can see above.
Richard, Ivan, Matt, thoughts from your perspectives as Alpha Port
maintainers?
> Plus as I pointed out (and you ignored) you are shutting out any future
> processors with this kind of oddity, and you have not audited all the
> embedded devices we support or may support in future.
True enough, but then again, the Alpha architects did feel the need to
add 8-bit and 16-bit memory reference instructions in EV56. In addition,
if there are future processors that don't provide 8-bit and 16-bit memory
reference instructions, atomic instructions can be used as a fallback.
This fallback is in fact similar to the set_bit approach.
> > The fact is that the kernel itself is much more parallel than it was
> > 15 years ago, and that trend is going to continue. Paired with the fact
> > that the Alpha is the least-parallel-friendly arch, makes improving
> > parallelism and correctness even harder within kernel subsystems; harder
> > than it has to be and harder than it should be.
> >
> > Linus has repeatedly stated that non-arch code should be as
> > arch-independent as possible
>
> That's why many many years ago we added set_bit() and the other bit
> functions. On sane processors they are very fast. On insane ones they
> work. They understand byte tearing, they understand store ordering (which
> a simple variable does not so you've got to audit all your memory
> barriers too). In many cases they are faster than using memory barriers
> to guide the compiler because they invalidate less and allow the compiler
> more freedom.
>
> All this started because I suggested you use set_bit and friends and for
> some reason you've decided to try and find another way to do it. We have
> the bit operations for a reason. On x86 they are very very fast, on
> uniprocessor anything they are very fast, on multiprocessor in general
> they are very fast, or you are dealing with boxes that have sanity
> problems of other kinds.
Indeed, these email threads do tend to examine alternatives from time
to time. ;-)
Thanx, Paul
next prev parent reply other threads:[~2014-09-22 19:51 UTC|newest]
Thread overview: 103+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-07-12 18:13 bit fields && data tearing Oleg Nesterov
2014-07-12 20:51 ` Oleg Nesterov
2014-07-12 23:34 ` Benjamin Herrenschmidt
2014-07-13 12:29 ` Oleg Nesterov
2014-07-13 13:15 ` Peter Hurley
2014-07-13 22:25 ` Benjamin Herrenschmidt
2014-07-15 13:54 ` Peter Hurley
2014-07-15 15:02 ` Richard Henderson
2014-09-03 22:51 ` Peter Hurley
2014-09-03 23:11 ` Benjamin Herrenschmidt
2014-09-04 8:43 ` David Laight
2014-09-04 9:52 ` Benjamin Herrenschmidt
2014-09-04 22:14 ` H. Peter Anvin
2014-09-05 0:59 ` Peter Hurley
2014-09-05 2:08 ` H. Peter Anvin
2014-09-05 15:31 ` Peter Hurley
2014-09-05 15:41 ` H. Peter Anvin
2014-09-08 17:52 ` One Thousand Gnomes
2014-09-08 17:59 ` H. Peter Anvin
2014-09-08 19:17 ` One Thousand Gnomes
2014-09-09 11:18 ` Peter Hurley
2014-09-08 22:47 ` Peter Hurley
2014-09-09 1:59 ` Paul E. McKenney
2014-09-09 11:14 ` Peter Hurley
2014-09-11 10:04 ` One Thousand Gnomes
2014-09-11 16:16 ` Paul E. McKenney
2014-09-11 20:01 ` Peter Hurley
2014-09-14 23:24 ` One Thousand Gnomes
2014-09-22 19:51 ` Paul E. McKenney [this message]
2014-09-23 18:19 ` Peter Hurley
2014-09-23 18:39 ` One Thousand Gnomes
2014-09-08 18:13 ` James Bottomley
2014-09-10 20:18 ` H. Peter Anvin
2014-09-10 21:10 ` Rob Landley
2014-09-05 2:08 ` H. Peter Anvin
2014-09-05 8:16 ` Michael Cree
2014-09-05 18:09 ` Paul E. McKenney
2014-09-05 18:31 ` Paul E. McKenney
2014-09-05 19:52 ` Peter Zijlstra
2014-09-05 20:01 ` Peter Hurley
2014-09-05 20:12 ` Peter Zijlstra
2014-09-05 20:15 ` H. Peter Anvin
2014-09-05 20:19 ` Paul E. McKenney
2014-09-05 18:50 ` Peter Hurley
2014-09-05 19:05 ` Paul E. McKenney
2014-09-05 19:24 ` Peter Hurley
2014-09-05 20:09 ` Paul E. McKenney
2014-09-05 19:38 ` Marc Gauthier
2014-09-05 20:14 ` Peter Hurley
2014-09-05 20:34 ` H. Peter Anvin
2014-09-05 20:42 ` Michael Cree
2014-09-05 20:43 ` Paul E. McKenney
2014-09-05 20:48 ` Thomas Gleixner
2014-09-05 21:05 ` Paul E. McKenney
2014-09-05 20:39 ` Michael Cree
2014-09-05 21:12 ` Peter Hurley
2014-09-05 21:27 ` Michael Cree
2014-09-05 20:42 ` Paul E. McKenney
2014-09-04 8:57 ` Mikael Pettersson
2014-09-04 9:09 ` Jakub Jelinek
2014-09-04 12:24 ` Peter Hurley
2014-09-04 12:29 ` Jakub Jelinek
2014-09-04 16:50 ` One Thousand Gnomes
2014-09-04 19:42 ` Peter Hurley
2014-09-04 22:16 ` H. Peter Anvin
2014-09-05 0:17 ` Paul E. McKenney
2014-09-05 1:57 ` Peter Hurley
2014-09-05 2:11 ` James Bottomley
2014-09-05 2:47 ` Peter Hurley
2014-09-05 4:06 ` Paul E. McKenney
2014-09-05 8:30 ` David Laight
2014-09-05 12:31 ` Peter Hurley
2014-09-05 12:37 ` David Laight
2014-09-05 16:17 ` Peter Hurley
2014-09-25 16:12 ` Pavel Machek
2014-09-07 5:07 ` James Bottomley
2014-09-07 16:21 ` Paul E. McKenney
2014-09-07 19:04 ` James Bottomley
2014-09-07 20:41 ` Peter Hurley
2014-09-08 5:50 ` James Bottomley
2014-09-08 20:45 ` Chris Metcalf
2014-09-08 22:43 ` James Bottomley
2014-09-09 2:27 ` H. Peter Anvin
2014-09-09 8:11 ` Arnd Bergmann
2014-09-08 23:30 ` Peter Hurley
2014-09-09 2:56 ` James Bottomley
2014-09-09 3:20 ` H. Peter Anvin
2014-09-09 4:30 ` H. Peter Anvin
2014-09-09 10:40 ` Peter Hurley
2014-09-10 21:48 ` James Bottomley
2014-09-10 23:50 ` Peter Hurley
2014-09-11 10:23 ` Will Deacon
2014-09-07 23:00 ` Paul E. McKenney
2014-09-07 23:17 ` H. Peter Anvin
2014-09-07 23:36 ` Paul E. McKenney
2014-09-07 23:39 ` H. Peter Anvin
2014-09-08 5:56 ` James Bottomley
2014-09-08 18:12 ` H. Peter Anvin
2014-09-08 19:09 ` James Bottomley
2014-09-08 19:12 ` H. Peter Anvin
2014-09-08 19:12 ` H. Peter Anvin
2014-09-08 22:39 ` James Bottomley
2014-09-09 2:30 ` H. Peter Anvin
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=20140922195106.GR4723@linux.vnet.ibm.com \
--to=paulmck@linux.vnet.ibm.com \
--cc=David.Laight@ACULAB.COM \
--cc=gnomes@lxorguk.ukuu.org.uk \
--cc=hpa@zytor.com \
--cc=ink@jurassic.park.msu.ru \
--cc=jakub@redhat.com \
--cc=linux-alpha@vger.kernel.org \
--cc=linux-arch@vger.kernel.org \
--cc=linux-ia64@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=mattst88@gmail.com \
--cc=mfranc@redhat.com \
--cc=oleg@redhat.com \
--cc=paulus@samba.org \
--cc=peter@hurleysoftware.com \
--cc=rth@twiddle.net \
--cc=tony.luck@intel.com \
/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).