mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Davide Libenzi <davidel@xmailserver.org>
To: Ulrich Drepper <drepper@gmail.com>
Cc: Davi Arnaut <davi@haxent.com.br>,
	Eric Dumazet <dada1@cosmosbay.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [patch 14/22] pollfs: pollable futex
Date: Thu, 3 May 2007 11:24:48 -0700 (PDT)	[thread overview]
Message-ID: <Pine.LNX.4.64.0705031111320.6502@alien.or.mcafeemobile.com> (raw)
In-Reply-To: <a36005b50705030646v5b729838pe8dc7b91dc9bee2b@mail.gmail.com>


I thought you were talking about the poll/epoll interface in general, and 
the approach on how to extend it for the very few cases that ppl asks for. 
but I see we're focusing on futexes ...


On Thu, 3 May 2007, Ulrich Drepper wrote:

> On 5/2/07, Davide Libenzi <davidel@xmailserver.org> wrote:
> > 99% of the fds you'll find inside an event loop you care to scale about,
> > are *already* fd based.
> 
> You are missing the point.  To get acceptable behavior of the wakeup
> it is necessary with this approach to open one descriptor _per thread_
> for a futex.  Otherwise all threads get woken upon FUTEX_WAKE.
> 
> This also means you need individual epoll sets for each thread.  You
> cannot share them anymore among all the threads in the process.

I'm not sure if futexes are the best approach to do that, but a way for 
the user to signal an event into a main event loop is needed.



> > On top of that, those fds are very cheap in terms of memory
> 
> They might be when they are counted in dozens.  But here we are
> talking about the possible need to use thousands of additional file
> descriptors.  If they are so cheap to allow thousands of descriptors
> with ease, why would the rlimit for files default to a small number
> (1024 on Fedora right now)?

Right now, ppl do that using pipes. That costs 2 file descriptors and at 
least 4KB of kernel data (plus an inode, a dentry and a file). This just 
to have a way to signal to an event loop dispatcher. The patches I posted 
a few weeks ago introduce an eventfd, that reduces the amount of kernel 
memory to basically a dentry and a file (plus uses only one file 
descriptor, and its 2-3 times faster than pipes. Add to that cost, about 
200 lines of code in fs/eventfd.c.



> > And this approach is not bound to a completely new and monolitic interface.
> 
> So?  It's stil additional, new code for an approach which will have to
> be superceded real soon.  That's just pure overhead to me.

IMO it is better to leave futexes alone. They are great for syncronizing 
MT apps, but do not properly fit an fd-based solution. For that, something 
like eventfd is enough.



- Davide



  reply	other threads:[~2007-05-03 18:24 UTC|newest]

Thread overview: 71+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-05-02  5:22 [patch 00/22] pollfs: filesystem abstraction for pollable objects Davi Arnaut
2007-05-02  5:22 ` [patch 01/22] pollfs: kernel-side API header Davi Arnaut
2007-05-02  5:22 ` [patch 02/22] pollfs: file system operations Davi Arnaut
2007-05-02  5:22 ` [patch 03/22] pollfs: asynchronously wait for a signal Davi Arnaut
2007-05-02  5:22 ` [patch 04/22] pollfs: pollable signal Davi Arnaut
2007-05-02  5:22 ` [patch 05/22] pollfs: pollable signal compat code Davi Arnaut
2007-05-02  5:22 ` [patch 06/22] pollfs: export the plsignal system call Davi Arnaut
2007-05-02  5:22 ` [patch 07/22] pollfs: x86, wire up " Davi Arnaut
2007-05-02  5:22 ` [patch 08/22] pollfs: x86_64, " Davi Arnaut
2007-05-02  5:22 ` [patch 09/22] pollfs: pollable hrtimers Davi Arnaut
2007-05-02 21:16   ` Thomas Gleixner
2007-05-02 23:00     ` Davi Arnaut
2007-05-02  5:22 ` [patch 10/22] pollfs: export the pltimer system call Davi Arnaut
2007-05-02  5:22 ` [patch 11/22] pollfs: x86, wire up " Davi Arnaut
2007-05-02  5:22 ` [patch 12/22] pollfs: x86_64, " Davi Arnaut
2007-05-02  5:22 ` [patch 13/22] pollfs: asynchronous futex wait Davi Arnaut
2007-05-02  5:22 ` [patch 14/22] pollfs: pollable futex Davi Arnaut
2007-05-02  5:54   ` Eric Dumazet
2007-05-02  6:16     ` Davi Arnaut
2007-05-02  6:39       ` Eric Dumazet
2007-05-02  6:54         ` Davi Arnaut
2007-05-02  7:11         ` Davi Arnaut
2007-05-02  7:40   ` Ulrich Drepper
2007-05-02  7:55     ` Eric Dumazet
2007-05-02  8:08       ` Ulrich Drepper
2007-05-02  8:49         ` Eric Dumazet
2007-05-02 16:39           ` Ulrich Drepper
2007-05-02 16:59             ` Davi Arnaut
2007-05-02 17:10               ` Ulrich Drepper
2007-05-02 17:29                 ` Davide Libenzi
2007-05-02 17:53                   ` Ulrich Drepper
2007-05-02 18:21                     ` Davide Libenzi
2007-05-03 13:46                       ` Ulrich Drepper
2007-05-03 18:24                         ` Davide Libenzi [this message]
2007-05-03 19:03                           ` Ulrich Drepper
2007-05-03 22:14                             ` Davide Libenzi
2007-05-04 15:28                               ` Ulrich Drepper
2007-05-04 19:15                                 ` Davide Libenzi
2007-05-04 19:20                                   ` 2.6.20.4 / 2.6.21.1 AT91SAM9260-EK oops Ryan Ordway
2007-05-04 23:38                                   ` [patch 14/22] pollfs: pollable futex Ulrich Drepper
2007-05-05 18:54                                     ` Davide Libenzi
2007-05-06  7:50                                       ` Ulrich Drepper
2007-05-06 19:47                                         ` Davide Libenzi
2007-05-06 19:54                                         ` Andrew Morton
2007-05-06 20:18                                           ` Davide Libenzi
2007-05-06 21:57                                           ` Davi Arnaut
2007-05-07  5:33                                           ` Ulrich Drepper
2007-05-07  5:46                                           ` Ulrich Drepper
2007-05-02 17:37                 ` Davi Arnaut
2007-05-02 17:49                   ` Ulrich Drepper
2007-05-02 18:05                     ` Davi Arnaut
2007-05-03 13:40                       ` Ulrich Drepper
2007-05-02 12:20     ` Davi Arnaut
2007-05-02 12:39     ` Davi Arnaut
2007-05-02 16:46       ` Ulrich Drepper
2007-05-02 17:05         ` Davi Arnaut
2007-05-02  5:22 ` [patch 15/22] pollfs: export the plfutex system call Davi Arnaut
2007-05-02  5:22 ` [patch 16/22] pollfs: x86, wire up " Davi Arnaut
2007-05-02  5:22 ` [patch 17/22] pollfs: x86_64, " Davi Arnaut
2007-05-02  5:22 ` [patch 18/22] pollfs: check if a AIO event ring is empty Davi Arnaut
2007-05-02  5:22 ` [patch 19/22] pollfs: pollable aio Davi Arnaut
2007-05-02  5:22 ` [patch 20/22] pollfs: export the plaio system call Davi Arnaut
2007-05-02  5:22 ` [patch 21/22] pollfs: x86, wire up " Davi Arnaut
2007-05-02  5:22 ` [patch 22/22] pollfs: x86_64, " Davi Arnaut
2007-05-02  6:05 ` [patch 00/22] pollfs: filesystem abstraction for pollable objects Andrew Morton
2007-05-02 17:28   ` Davide Libenzi
2007-05-02 17:47     ` Davi Arnaut
2007-05-02 18:23       ` Davide Libenzi
2007-05-02 18:50         ` Davi Arnaut
2007-05-02 19:42           ` Davide Libenzi
2007-05-02 20:11             ` Davi Arnaut

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.4.64.0705031111320.6502@alien.or.mcafeemobile.com \
    --to=davidel@xmailserver.org \
    --cc=akpm@linux-foundation.org \
    --cc=dada1@cosmosbay.com \
    --cc=davi@haxent.com.br \
    --cc=drepper@gmail.com \
    --cc=linux-kernel@vger.kernel.org \
    --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

all inboxes | Powered by JetHome®