Mol Freeze (cont.)

Francois Taiani mol-devel@lists.maconlinux.org
Mon, 17 Jun 2002 18:27:06 +0200


Hi everyone,

some weeks ago (05-27-02), Samuel proposed a VM bug fix to prevent MOL 
from freezing. I've been trying the patch for some weeks now, and alas I 
must say mol still freezes now and then (I'm using a 2.4.18 kernel with 
the powerpc patch on a G3 White & Blue, with 192 Mb RAM). The freeze 
happens most frequently when I save some file within word or powerpoint.

Since then I've disabled the VM in MacOS, and allocated more RAM to mol 
in the molrc file (125 Mb). It works fine, and mol does not seem to use 
that much more memory (top returns 32 Mb).

In fact, I now wonder what's the interest of Virtual Memory management 
within mol, since Linux already has its own virtual address spaces whose 
pages are physically allocated on demand (AFAIK). Indeed having mol 
being granted 125 Mb does seems to really hamper my system. Would 
anybody know more ?

Cheers

Francois

Francois Taiani wrote:
> Hi Samuel,
> 
> the 2 problems are solved now ! I could compile the new kernel module, 
> and it works. I'm going to work for a while with it, and I'll come back 
> to you to tell if the situation has improved.
> 
> Thanks a lot !
> 
> Francois
> 
> samuel@ibrium.se wrote:
> 
>> On Mon, May 27, 2002 at 10:29:52AM +0200, Francois Taiani wrote:
>>
>>> I've tried to compile the new module version, and here are 2 errors I 
>>> encountered:
>>>
>>> 1) The first problem is due to "cpu/mregs.h" not being found from 
>>> within "src/shared/mac_registers.h" (I already had that problem 
>>> before, but forgot to report it):
>>> My work-around is to create a symbolic link :
>>>
>>> # cd src/shared/
>>> # ln -s ppc/ cpu
>>
>>
>>
>> This is the correct thing to do. The symlink is currently created
>> automatically if you just do a 'make' instead of 'make modules'.
>> I'll fix this behaviour and create the symlink at the ./configure step 
>> instead.
>>
>>
>>> 2) The second problem is more mysterious to me:
>>
>>
>>
>> Sorry... I failed to push some of the changes. It should
>> work now.
>>
>> /Samuel
>>
> 


-- 
Francois Taiani              LAAS-CNRS (http://www.laas.fr)
+33 (0) 561 336 406          Dependable Computing
http://www.laas.fr/~ftaiani  and Fault Tolerance