Re: [PATCH] mm: allow huge kvmalloc() calls if they're accounted to memcg

On 18/10/21 17:19, Kees Cook wrote:

Ah, so memcg wasn't doing sanity checks?

Is there a cheap way to resolve the question "does this much memory
exist"? The "__" versions end up lacking context for why they're "__"
versions. I.e. do we want something more descriptive, like
__huge_kvmalloc_node() or __unbounded_kvmalloc_node()?

No problem with that, I think "unbounded" is descriptive enough that we can remove the __ too.  So that would be kvmalloc_node_unbounded / kvmalloc_array_unbounded / kvcallc_unbounded?

Paolo