Re: [PATCH] failslab - failmalloc for slab allocator

[Andrew Morton <akpm@osdl.org>]

On Sun, 13 Aug 2006 18:22:19 +0800
Akinobu Mita <mita@miraclelinux.com> wrote:

> This patch is not intended for inclusion. But I could find
> several interesting crashes.
> 
> The idea behind failslab is to demonstrate what really happens if
> slab allocation fails. The idea of failslab is completely taken
> from failmalloc (http://www.nongnu.org/failmalloc/).
> 
> boot option:
> 
> failslab=<probability>,<interval>,<times>,<space>
> 
> <probability>
> 	specifies how often it should fail in percent.
> <interval>
> 	specifies the interval of failures.
> <times>
> 	specifies how many times failures may happen at most.
> <space>
> 	specifies the size of free space where memory can be allocated
> 	safely in bytes.
> 
> examples:
> 
> failslab=100,10,-1,0
> 
> slab allocation (kmalloc, kmem_cache_alloc,..) fails once per 10 times.

We would benefit from having some faul-injection capabilities in the
mainline kernel.

- kmalloc failures

- alloc_pages() failures

- disk IO errors (there are rumours of a DM module for this, but I
  haven't seen it).

They would need to be lightweight, clean and enabled/configured at runtime,
not at boot time.

This would end up being a fairly complex project.