Re: A patch for the file kernel/fork.c

["André Pereira de Almeida" <andre@cachola.com.br>]

Valdis.Kletnieks@vt.edu wrote:

>On Wed, 04 May 2005 15:26:44 -0300, =?ISO-8859-1?Q?Andr=E9_Pereira_de_Almeida?= said:
>
>  
>
>>In a preemptible kernel with the serport module and a serial port try to 
>>run the following program:
>>    
>>
>
>  
>
>>and kill it.
>>In my case it will hang the computer. I think this is a problem with the 
>>serport module. With this patch, the serial mouse stop working, but the 
>>computer don't hang.
>>    
>>
>
>The fact that the mouse stops working is indicative that this patch doesn't
>actually fix the problem, it's just pushing it around in the kernel - sooner
>or later something *else* is going to go pear-shaped on the null *mm.  The right
>fix is to figure out why mm is bogus and fix that issue.
>  
>
This patch is not to fix this problem, but to minimize a bug in other 
modules (in this case I think it's in the serport module) and to keep 
the system working, at least to make a clean shutdown or terminate some 
important process cleanly.

This patch is in the mm_release that should return a null *mm anyway.

The bogus mm came from a pagefault in the do_exit function in 
kernel/exit.c that occured after the exit_mm call. This pagefault, in my 
case, come from a filp_close call that called tty_wakeup. I'm trying to 
identify this problem, but a pagefault like this may came from several 
different modules.
André