From: Jacques Gelinas (jack_at_solucorp.qc.ca)
Date: Sat 23 Mar 2002 - 17:18:39 GMT
On Sat, 23 Mar 2002 16:20:56 -0500, Martin Josefsson wrote
> On Sat, 23 Mar 2002, Sebastian Schaefer wrote:
> > --On Donnerstag, März 21, 2002 18:51:43 +0100 Martin Josefsson
> > <gandalf_at_wlug.westbo.se> wrote:
> > > Hmm, It was I who wrote that part... It's been stable here since I coded
> > > it so I assumed it was stable.
> > > I havn't tested ctx9 yet, I'm still running ctx8 (the first version with
> > > this new socket stuff) here.
> > >
> > > Did you have this problem with ctx8? or is it a ctx9 specific problem?
> > ctx8 works... no problem with netstat ... mmm
> But hmm, I don't see anything that can explain the crashes between ctx8
> and ctx9 so it's possible that it's my netstat stuff that's causing it
> after all. But I fail to see how on earth it can lock up the machine, all
> it does is to skip some sockets in the creation of the socketlists in
> Some prople said that it just locked up and at least one person provided
> information that sait it hung in tcp_get_info which is one of the
> functions I've modified.
> Jacques, do you have any ideas?
The kernel oops does not report it crashes in tcp_get_info, but in __get_free_page
called from proc_file_read.
Note that I have also seen the bug with
so we know it is related to tcp_get_info(), but the oops does not occur there.
Maybe tcp_get_info is doing something weird (I must admit having read and read
the code and I can't see anything weird :-) ) and then some corruption is
produced and the kernel crash later.
Also the oops is clearly initiated by the netstat process. So it is kind of directly
related (or the cat /proc/net/tcp process).
Jacques Gelinas <jack_at_solucorp.qc.ca>
vserver: run general purpose virtual servers on one box, full speed!