On Fri, Oct 25, 2024 at 12:36:28PM -0700, Christoph Lameter (Ampere) wrote:

Sorry saw this after the other email.

On Fri, 25 Oct 2024, Peter Zijlstra wrote:

quoted

quoted

Could we follow NUMA policies like with other metadata allocations during
systen call processing?

I had a quick look at this, and since the mempolicy stuff is per vma,
and we don't have the vma, this is going to be terribly expensive --
mmap_lock and all that.

There is a memory policy for the task as a whole that is used for slab
allocations and allocations that are not vma bound in current->mempolicy.
Use that.

You can get a node number following the current task mempolicy by calling
mempolicy_slab_node() and keep using that node for the future.

I'll look into the per task thing, which I'm hoping means per-process.
We need something that is mm wide consistent.

But since futexes play in the address space, I was really rather
thinking we ought to use the vma policy.