linux-kernel - Re: Unexpected cascaded epoll behavior

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 for Android: free password hash cracker in your pocket

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

Message-ID: <alpine.DEB.1.10.0901251300290.22000@alien.or.mcafeemobile.com>
Date:	Sun, 25 Jan 2009 13:02:22 -0800 (PST)
From:	Davide Libenzi <davidel@...ilserver.org>
To:	Pavel Pisa <pisa@....felk.cvut.cz>
cc:	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: Unexpected cascaded epoll behavior - my mistake or kernel bug

On Sat, 24 Jan 2009, Davide Libenzi wrote:

> On Fri, 23 Jan 2009, Pavel Pisa wrote:
> 
> > Hello Davide and all others,
> > 
> > I have got to implementing yet another event library and I experience
> > something strange when level triggered epoll_wait() monitors another
> > level triggered epoll set. Top level epoll_wait() return information
> > about one event pending. The event points correctly to the monitored
> > second level epoll fd, but epoll_wait() on this fd with 0 or small timeout
> > returns zero/no events and top level epoll reporting is not reset at this
> > point so program enters busy loop. If there are events, they are processed
> > correctly. More interresting is, that the ill behavior is prevented if I
> > change number of debug writes into unrelated fd or if the application
> > is slowed down by strace to standard output. The strace to the
> > file does not (fortunatelly) hide ill behavior to the strange behavior
> > of my code or kernel can be documented in attached straces. The both
> > traces documents the busy loop cause for case, where there is created
> > epoll set with fd #3, that set monitors fd #0. When character arrives
> > at fd #0, the new epoll fd set with fd #4 is created ad fd #3 is added
> > into this new top level fd set.
> > 
> > epoll_create(8)                         = 3
> > epoll_ctl(3, EPOLL_CTL_ADD, 0, {EPOLLIN, {u32=6353264, u64=6353264}}) = 0
> > epoll_wait(3, {{EPOLLIN, {u32=6353264, u64=6353264}}}, 8, 20000) = 1
> > read(0, "E"..., 1)                      = 1
> > ...
> > epoll_create(8)                         = 4
> > epoll_ctl(4, EPOLL_CTL_ADD, 3, {EPOLLIN, {u32=6353664, u64=6353664}}) = 0
> > epoll_wait(4, {{EPOLLIN, {u32=6353664, u64=6353664}}}, 8, 19998) = 1
> > epoll_wait(3, {{EPOLLIN, {u32=6353264, u64=6353264}}}, 8, 0) = 1
> > read(0, "\n"..., 1)                     = 1
> > epoll_wait(4, {{EPOLLIN, {u32=6353664, u64=6353664}}}, 8, 19998) = 1
> > epoll_wait(3, {}, 8, 0)                 = 0
> > epoll_wait(4, {{EPOLLIN, {u32=6353664, u64=6353664}}}, 8, 19998) = 1
> > epoll_wait(3, {}, 8, 0)                 = 0
> > 
> > You may ask, why that terrible clam starts to writting new event library code
> > and why he needs to cascade event handling mechanisms. I try to defend myself
> > a little. If you do not like beat me, you can skip next section
> 
> Pavel, can you give the patch below a spin? It's working fine in my two 
> boxes, but it'd be great if you could confirm it too.
> Attached there's also a simple test program to check epoll behavior, 
> especially the one related to recursion.

Please use the attached epoll_test.c since the one I sent you before 
wasn't actually triggering the loop condition it wanted to check.



- Davide


View attachment "epoll_test.c" of type "TEXT/x-csrc" (8504 bytes)