public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Eric Dumazet <dada1@cosmosbay.com>
To: "Ulrich Drepper" <drepper@gmail.com>
Cc: "Nick Piggin" <nickpiggin@yahoo.com.au>,
	"Andi Kleen" <ak@suse.de>,
	"Ravikiran G Thirumalai" <kiran@scalex86.org>,
	"Shai Fultheim (Shai@scalex86.org)" <shai@scalex86.org>,
	"pravin b shelar" <pravin.shelar@calsoftinc.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [RFC] NUMA futex hashing
Date: Tue, 8 Aug 2006 19:08:17 +0200	[thread overview]
Message-ID: <200608081908.18157.dada1@cosmosbay.com> (raw)
In-Reply-To: <a36005b50608080958n192e9324jb9d5a7a59b365eae@mail.gmail.com>

On Tuesday 08 August 2006 18:58, Ulrich Drepper wrote:
> On 8/8/06, Eric Dumazet <dada1@cosmosbay.com> wrote:
> > So we really can... but for 'private futexes' which are the vast majority
> > of futexes needed by typical program (using POSIX pshared thread mutex
> > attribute PTHREAD_PROCESS_PRIVATE, currently not used by NPTL glibc)
>
> Nonsense.  Mutexes are by default always private.  They explicitly
> have to be marked as sharable.  This happens using the
> pthread_mutexattr_setpshared function which takes
> PTHREAD_PROCESS_PRIVATE or PTHREAD_PROCESS_SHARED in the second
> parameter.  So the former _is_ clearly used.
>

I was saying that PTHREAD_PROCESS_PRIVATE or PTHREAD_PROCESS_SHARED info is 
not provided to the kernel (because futex api/implementation dont need to). 
It was not an attack on glibc.

> > Of course we would need a new syscall, and to change glibc to be able to
> > actually use this new private_futex syscall.
>
> No, why?  The kernel already does recognize private mutexes.  It just
> checks whether the pages used to store it are private or mapped.  This
> requires some interaction with the memory subsystem but as long as no
> crashes happen the data can change underneath.  It's the program's
> fault if it does.

But if you let futex code doing the vma walk to check the private/shared 
status, you still need the mmap_sem locking.

Moreover, a program can mmap() a file (shared in terms of VMA), and continue 
to use a  PTHREAD_PROCESS_PRIVATE mutex lying in this shared zone
(Example : shmem or hugetlb mapping, wich API might always give a 'shared' 
vma)

>
> On the waker side you would search the local futex hash table/tree
> first and if this doesn't yield a match, search the global table.
> Wakeup calls without any waiters are usually rare.

If the two searches touch two different cache lines in the hash table, we 
might have a performance regression.
Of course we might chose a hash function so that the same slot is accessed.

Eric


  reply	other threads:[~2006-08-08 17:08 UTC|newest]

Thread overview: 78+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-08-08  7:07 [RFC] NUMA futex hashing Ravikiran G Thirumalai
2006-08-08  9:14 ` Eric Dumazet
2006-08-08 20:31   ` Ravikiran G Thirumalai
2006-08-08  9:37 ` Jes Sorensen
2006-08-08  9:58   ` Andi Kleen
2006-08-08 10:07     ` Jes Sorensen
2006-08-08  9:57 ` Andi Kleen
2006-08-08 10:10   ` Eric Dumazet
2006-08-08 10:36     ` Andi Kleen
2006-08-08 12:29       ` Eric Dumazet
2006-08-08 12:47         ` Andi Kleen
2006-08-08 12:57           ` Eric Dumazet
2006-08-08 14:39             ` Ulrich Drepper
2006-08-08 15:11               ` Nick Piggin
2006-08-08 15:36                 ` Ulrich Drepper
2006-08-08 16:22                   ` Nick Piggin
2006-08-08 16:26                     ` Nick Piggin
2006-08-08 16:49                     ` Ulrich Drepper
2006-08-08 16:08                 ` Eric Dumazet
2006-08-08 16:34                   ` Nick Piggin
2006-08-08 16:49                     ` Eric Dumazet
2006-08-08 16:59                       ` Eric Dumazet
2006-08-09  1:56                       ` Nick Piggin
2006-08-08 16:58                   ` Ulrich Drepper
2006-08-08 17:08                     ` Eric Dumazet [this message]
2006-08-09  1:58                     ` Nick Piggin
2006-08-09  6:26                       ` Eric Dumazet
2006-08-09  6:43                         ` Eric Dumazet
2007-03-15 19:10                           ` [PATCH 0/3] FUTEX : new PRIVATE futexes, SMP and NUMA improvements Eric Dumazet
2007-03-15 20:15                             ` Nick Piggin
2007-03-16  8:05                             ` Peter Zijlstra
2007-03-16  9:30                               ` Eric Dumazet
2007-03-16 10:10                                 ` Peter Zijlstra
2007-03-16 10:30                                   ` Eric Dumazet
2007-03-16 10:36                                     ` Peter Zijlstra
2007-04-04  7:16                             ` Ulrich Drepper
2007-04-05 17:49                               ` [PATCH] FUTEX : new PRIVATE futexes Eric Dumazet
2007-04-05 20:43                                 ` Ulrich Drepper
2007-04-06  1:19                                 ` Nick Piggin
2007-04-06  5:53                                   ` Eric Dumazet
2007-04-06 11:50                                     ` Nick Piggin
2007-04-06  6:05                                   ` Hugh Dickins
2007-04-06 17:41                                     ` Jan Engelhardt
2007-04-06 12:26                                 ` Shared futexes (was [PATCH] FUTEX : new PRIVATE futexes) Peter Zijlstra
2007-04-06 13:02                                   ` Hugh Dickins
2007-04-06 13:15                                     ` Peter Zijlstra
2007-04-06 13:15                                     ` Nick Piggin
2007-04-06 13:22                                       ` Peter Zijlstra
2007-04-06 13:40                                         ` Nick Piggin
2007-04-06 12:31                                 ` [PATCH] FUTEX : new PRIVATE futexes Peter Zijlstra
2007-04-07  8:43                                 ` [PATCH, take4] " Eric Dumazet
2007-04-07  9:30                                   ` Nick Piggin
2007-04-07 10:00                                     ` Eric Dumazet
2007-04-11  7:22                                       ` Nick Piggin
2007-04-11  8:14                                         ` Eric Dumazet
2007-04-11  9:23                                           ` Nick Piggin
2007-04-11  9:30                                             ` Pierre Peiffer
2007-04-11  9:39                                               ` Nick Piggin
2007-04-11  9:40                                                 ` Nick Piggin
2007-04-11  9:35                                             ` Eric Dumazet
2007-04-12  1:57                                               ` Nick Piggin
2007-04-07 11:18                                   ` Jakub Jelinek
2007-04-07 11:54                                     ` Eric Dumazet
2007-04-07 16:40                                       ` Ulrich Drepper
2007-04-07 22:15                                   ` Andrew Morton
2007-04-10  9:21                                     ` Eric Dumazet
2007-04-11  9:19                                   ` [PATCH, take5] " Eric Dumazet
2007-04-11 12:23                                     ` Rusty Russell
2007-04-26 12:55                                     ` [PATCH, take6] " Eric Dumazet
2007-04-26 13:35                                       ` Pierre Peiffer
2007-03-15 19:13                           ` [PATCH 1/3] FUTEX : introduce PROCESS_PRIVATE semantic Eric Dumazet
2007-03-15 19:16                           ` [PATCH 2/3] FUTEX : introduce private hashtables Eric Dumazet
2007-03-15 20:25                             ` Nick Piggin
2007-03-15 21:09                               ` Ulrich Drepper
2007-03-15 21:29                                 ` Nick Piggin
2007-03-15 22:59                               ` William Lee Irwin III
2007-03-15 19:20                           ` [PATCH 3/3] FUTEX : NUMA friendly global hashtable Eric Dumazet
2006-08-09  0:13     ` [RFC] NUMA futex hashing Ravikiran G Thirumalai

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=200608081908.18157.dada1@cosmosbay.com \
    --to=dada1@cosmosbay.com \
    --cc=ak@suse.de \
    --cc=drepper@gmail.com \
    --cc=kiran@scalex86.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=nickpiggin@yahoo.com.au \
    --cc=pravin.shelar@calsoftinc.com \
    --cc=shai@scalex86.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