LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Buffer overflow when listing large number of VLS's?

To: Wensong Zhang <wensong@xxxxxxxxxxxx>
Subject: Re: Buffer overflow when listing large number of VLS's?
Cc: linux-virtualserver@xxxxxxxxxxxx
From: Rok Sosic <rok@xxxxxxxxxxxxxx>
Date: Wed, 30 Jun 1999 19:05:49 +0200 (MEST)
On Wed, 30 Jun 1999, Wensong Zhang wrote:

> 
> 
> Ted Pavlic -- List Mail Account wrote:
> 
> > I'm using the VS 0.5 patch for a Linux 2.2.9 kernel.
> >
> > Recently I configured 254 virtual servers on my LinuxDirector. This worked
> > fine, but whenever I tried to list them... random things would occur.
> 
> Wah, this is the biggest LVS that I have ever heard!
> 
> BTW, what are you using it for? could we visit it?
> 
> >
> > Sometimes the kernel would give me a GPF... sometimes the machine would
> > completely lockup... Sometimes my shell would core dump... etc... etc...
> >
> > It sounds like the ipvsadm -L is not checking the size of the buffer its
> > dumping its list into before it prints it... Memory is being overwritten
> > outside of the buffer and causing problems. (I haven't actually looked at
> > the code, but I'm guessing this is the problem... buffer overflows are
> > pretty common)
> >
> > But things (the actual load balancing) work fine... as long as I don't view
> > my VLS's. :)
> >
> > Can anyone else confirm these findings?
> >
> 
> Sorry that I have left this overflow bug in the code. Tomorrow I will
> release a new patch.

Hi Wensong,

I suggest that you take a look at ip_vs.c:ip_vs_procinfo(). The routine
indeed is not checking its buffer size. :-( We did not expect such large
configurations...

Best,
Rok
--
Rok Sosic
TurboLinux development team
rok@xxxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>
  • Re: Buffer overflow when listing large number of VLS's?, Rok Sosic <=