Re: Mbuf double-free guilty party detection patch

From: Thierry Herbelot <thierry_at_herbelot.com>
Date: Sun, 26 Jun 2005 10:49:39 +0200
Le Sunday 26 June 2005 08:43, Andrey Chernov a écrit :
> On Sun, Jun 26, 2005 at 01:20:47AM -0500, Mike Silbersack wrote:
> > Yes and no.  Your results tell me that it's probably not a simple use
> > after free, but rather something smashing all over memory for some
> > reason.
>
> It sounds nasty, but I observe no strange behaviour of TCP connections at
> all for a long time. I have a lots of them actively running sshd,
> sendmail, popper, ftpd, bind, httpd and cvsupd servers.

still no good luck : after using the second patch, no panic, but the debug 
messages seem incomplete : (last freed by: /0/)

This memory last freed by: 0
Memory modified after free 0xc15d5800(256) val=800 _at_ 0xc15d583c
This memory last freed by: 0
Memory modified after free 0xc15d5800(256) val=0 _at_ 0xc15d5840
This memory last freed by: 0
Memory modified after free 0xc15d5800(256) val=3 _at_ 0xc15d5844

the test case is : building the kernel while tar-ing the src tree over two 
separate ssh session.

	TfH

PS : what is puzzling is that I've got another machine running a more recent 
-current, with no ill effects (but it uses an ed(4) I/F)
Received on Sun Jun 26 2005 - 06:50:00 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:37 UTC