2020-02-06 15:28:26

by Johannes Weiner

[permalink] [raw]
Subject: Re: [PATCH 1/2] mm: Charge current memcg when no mm is set

On Wed, Feb 05, 2020 at 02:33:47PM -0800, Dan Schatzberg wrote:
> This modifies the shmem and mm charge logic so that now if there is no
> mm set (as in the case of tmpfs backed loop device), we charge the
> current memcg, if set.
>
> Signed-off-by: Dan Schatzberg <[email protected]>

Acked-by: Johannes Weiner <[email protected]>

It's a dependency for 2/2, but it's also an overdue cleanup IMO: it's
always been a bit weird that memalloc_use_memcg() worked for kernel
allocations but was silently ignored for user pages.

This patch establishes a precedence order for who gets charged:

1. If there is a memcg associated with the page already, that memcg is
charged. This happens during swapin.

2. If an explicit mm is passed, mm->memcg is charged. This happens
during page faults, which can be triggered in remote VMs (eg gup).

3. Otherwise consult the current process context. If it has configured
a current->active_memcg, use that. Otherwise, current->mm->memcg.

Thanks Dan