Re: 2.4.20pre11aa1

[Srihari Vijayaraghavan <harisri@bigpond.com>]

Hello Andrea,

On Saturday 19 October 2002 02:00, Andrea Arcangeli wrote:
> the corrupted module tables rings a bell. I fixed the wrong locking in
> the module code that could corrupt these tables (they were relying on
> the bkl but the bkl means nothing if you copy_user in the middle of the
> loop like the module code does, so I replaced the bkl with a semaphore
> and that should fix things), but I wonder if I broken something else
> with these fixes.
>
> Here's the patch that I'm talking about, you may want to start the
> binary search backing this out and see if the problem goes away. if it
> goes away I clearly need to double check it ;)

Unfortunately removing that change off kernel/module.c did not help.

I may be wrong but considering in my case the kernel is crashing whether 
agpgart/radeon are compiled as modules or built-in, I suspect that this issue 
is larger than just modules sub-system.

Anyway I will start applying the patches from 00* on-wards from your tree to 
see if I can reliably prove where the problem is.

Thanks.
-- 
Hari
harisri@bigpond.com