Re: A true story of a crash.

wolff@cardit.et.tudelft.nl, Rogier Wolff (wolff@dutepp0.et.tudelft.nl)
Sun, 16 Aug 1998 20:12:11 +0200 (MEST)

linker@z.ml.org wrote:
>
> In linux when a process allocates say a meg of memory, the kernel gives it
> 256k pages of Copy on write zero page mappings.
>
> When the process actually writes to these pages, the kernel actually
> allcates the memory. This gives increadble performance and memory saving
> benifits.. The problem is that you can have 50 apps open and be WAY over
> commited.. Then you run out of memory and process (who are not trying any
> more allocations are are simply using the ram they allocated at start)
> start dieing as they use those COW pages.

Well, this overcommitment is fine in some cases, and not in others. A
kernel option that enables the no overcommit case would be nice for
some people.

The argument FOR overcommitting memory is:

Almost nobody check the malloc return values. And if they do, all
they do is bomb out with "out of memory". If you keep such a
process, it might even run to completion.

The argument AGAINST overcommitting is:

I happent to have a couple of important applications that DO know
what to do in the "malloc-returns-NULL" case.

Regards,

Roger.

-- 
| The secret of success is sincerity.  Once you can |  wolff@cardit.et.tudelft.nl 
| fake that, you've got it made.  -- Jean Giraudoux |       T: +31-15-2137555 
- Custom Linux device drivers for sale! Call for a quote. - F: +31-15-2138217


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html