RE: [Squid-cvs] ntlm squid/src authenticate.c,1.1.1.3.12.57,1.1.1.3.12.58

From: Robert Collins <robert.collins@dont-contact.us>
Date: Thu, 13 Jun 2002 23:02:01 +1000

> -----Original Message-----
> From: Henrik Nordstrom [mailto:hno@marasystems.com]
> Sent: Thursday, 13 June 2002 6:53 PM
> To: Robert Collins; squid-dev@squid-cache.org
> Subject: Re: [Squid-cvs] ntlm squid/src
> authenticate.c,1.1.1.3.12.57,1.1.1.3.12.58
>
>
> On Wednesday 12 June 2002 22:26, Robert Collins wrote:
>
> > > Nice to see you around again ;-)
> >
> > Thanks, hopefully I'll have a bit more time in the near future.
> > More on that later.
>
> Nice. Am working on the same thing here.. (to increase my Squid time
> by a couple of 100% so it no longer approaches nil).
>
> > Hmm, it could be use-after-free I suppose, but it didn't look like
> > it. It took Kinkie and I about 4 hours of unpleasant debugging to
> > isolate the returned memory as being wrong.
>
> Any details on how to reproduce the problem? If possible to reproduce
> it should show up pretty quick in a memory debugger with pools
> disabled.

Run several megabit of NTLM authenticated traffic through your cache,
and look for the occasional unexpected behaviour. Complain to the
developer (rightfully so), and get told to up the logging rate. Fill up
your hard disks with auth logging, give the logs to the developer, and
let him come back with "looks like foo struct was invalid. Lets try
forcing it clear". Find it works. Sigh, and move onto the next problem
:}.

> I assume this was not during the period of HEAD where chunked
> mempools
> was broken and returned uncleared memory for some allocations.. but
> even then it is quite unlikely to have been caused by this as it
> depended on "even" pool sizes.

It may have been, I couldn't say for certain.

Rob
Received on Thu Jun 13 2002 - 07:01:52 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:15:40 MST