Re: Oops in 2.5.64

[J Sloan <joe@tmsusa.com>]

Linus Torvalds wrote:

>In article <3E66E782.5010502@tmsusa.com>, J Sloan  <joe@tmsusa.com> wrote:
>  
>
>>2.5.64 was running well, but after a day
>>or so of uptime, in fairly busy use (squid,
>>postfix, dhcp server, iptables, X desktop)
>>I ssh'd in as root, issued an init 3, then
>>a moment later, init 5. A moment after
>>that, the ssh session froze and all internet
>>access stopped as well.
>>
>>The console was frozen, with an oops -
>>    
>>
>
>Are you using DRI? There is some evidence that exiting and restarting X
>will not correctly re-initialize the DRI stuff in the kernel, and
>_massive_ kernel memory corruption can ensure when the new X server
>starts. 
>
Yes, of course, I want altlantis to look _good_

>
>At which point you'll get random oopses etc.
>
That does indeed seem to be the scenario - it
was rock solid and busy for a day, but when X
was killed and restarted everything went right
to hell -

>Looks like you at least have the DRI kernel modules there.
>
>Try to see if the problem goes away if you start X without DRI support
>(ie remove the "Load 'dri'" or whatever from the XF86Config file, or
>start up in a mode that DRI doesn't support, like 8bpp).
>  
>
OK, I'll try it like that -

Joe