Re: OOM behavior in constrained memory situations

[Andrew Morton <akpm@osdl.org>]

Christoph Lameter <clameter@engr.sgi.com> wrote:
>
> There are situations in which memory allocations are restricted by policy, 
> by a cpuset or by type of allocation. 
> 
> I propose that we need different OOM behavior for the cases in which the
> user has imposed a limit on what type of memory to be allocated. In that 
> case the application should be terminate with OOM. The OOM killer should 
> not run.
> 
> The huge page allocator has already been modified to return a Bus Error
> because it would otherwise trigger the OOM killer. Its a bit strange
> if an app returns a Bus Error because it its out of memory.
> 
> Could we modify the system so that the application requesting 
> memory is terminated with an out of memory condition if
> 
> 1. No huge pages are available anymore.
> 
> 2. The application has set a policy that restricts allocation to
>    certain nodes.
> 
> 3. An application is restricted by a cpuset to certain nodes.
> 
> 4. An application has requested large amounts of memory and the 
>    allocation fails.
> 
> That should avoid the OOM killer in most situations.

Do we really want to kill the application?  A more convetional response
would be to return NULL from the page allocator and let that trickle back.

The hugepage thing is special, because it's a pagefault, not a syscall.