linux-kernel - Re: PROBLEM: Epoll frequently fails to notify connects at connect bursts

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <alpine.DEB.2.00.0908260802520.6557@makko.or.mcafeemobile.com>
Date:	Wed, 26 Aug 2009 08:04:32 -0700 (PDT)
From:	Davide Libenzi <davidel@...ilserver.org>
To:	Will Brown <dr.hfuhruhurr@...il.com>
cc:	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: PROBLEM: Epoll frequently fails to notify connects at connect 
 bursts

On Wed, 26 Aug 2009, Will Brown wrote:

> On Wed, Aug 26, 2009 at 3:24 AM, Davide Libenzi<davidel@...ilserver.org> wrote:
> > On Tue, 25 Aug 2009, Davide Libenzi wrote:
> >
> >> Multiple quasi-simultaneous events for the same fd are merged. This is why
> >> on POLLIN on accepting sockets you want to do something like:
> >>
> >>       while ((afd = accept(sfd, ...)) != -1)
> >>               add_client(afd);
> >
> > Of course, with  sfd  in non-blocking mode. Alternatively, you drop EPOLLET
> > for the  sfd  socket.
> >
> 
> Aha! Thank you immensely Davide. It cost me one week of frustration already.
> 
> 1. Where should I have picked up this gem of essential information? I
> haven't seen it in the man-pages for either 'epoll' or 'accept'.
> 
> 2. Is it better to use a level-triggered listener socket, or to use
> the accept-loop?

EPOLLET can be more efficent, but it can also be a little more trickier to 
grasp. LT mode maps more closely to poll/select, that makes it more 
familiar to users.


- Davide