Re: [PATCH] mm/khugepaged: alloc_charge_hpage() take care of mem charge errors

[Johannes Weiner]

On Wed, Feb 22, 2023 at 12:38:36PM -0500, Peter Xu wrote:
> On Wed, Feb 22, 2023 at 12:06:20PM -0500, Johannes Weiner wrote:
> > Hello,
> > 
> > On Tue, Feb 21, 2023 at 04:43:44PM -0500, Peter Xu wrote:
> > > If memory charge failed, the caller shouldn't call mem_cgroup_uncharge().
> > > Let alloc_charge_hpage() handle the error itself and clear hpage properly
> > > if mem charge fails.
> > 
> > I'm a bit confused by this patch.
> > 
> > There isn't anything wrong with calling mem_cgroup_uncharge() on an
> > uncharged page, functionally. It checks and bails out.
> 
> Indeed, I didn't really notice there's zero side effect of calling that,
> sorry.  In that case both "Fixes" and "Cc: stable" do not apply.
> 
> > 
> > It's an unnecessary call of course, but since it's an error path it's
> > also not a cost issue, either.
> > 
> > I could see an argument for improving the code, but this is actually
> > more code, and the caller still has the uncharge-and-put branch anyway
> > for when the collapse fails later on.
> > 
> > So I'm not sure I understand the benefit of this change.
> 
> Yes, the benefit is having a clear interface for alloc_charge_hpage() with
> no prone to leaking huge page.
> 
> The patch comes from a review for David's other patch here:
> 
> https://lore.kernel.org/all/Y%2FU9fBxVJdhxiZ1v@x1n/

Ah, that makes sense. Thanks for the context.

> From 0595acbd688b60ff7b2821a073c0fe857a4ae0ee Mon Sep 17 00:00:00 2001
> From: Peter Xu <peterx@xxxxxxxxxx>
> Date: Tue, 21 Feb 2023 16:43:44 -0500
> Subject: [PATCH] mm/khugepaged: alloc_charge_hpage() take care of mem charge
>  errors
> 
> If memory charge failed, instead of returning the hpage but with an error,
> allow the function to cleanup the folio properly, which is normally what a
> function should do in this case - either return successfully, or return
> with no side effect of partial runs with an indicated error.
> 
> This will also avoid the caller calling mem_cgroup_uncharge() unnecessarily
> with either anon or shmem path (even if it's safe to do so).
> 
> Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
> Cc: Yang Shi <shy828301@xxxxxxxxx>
> Cc: David Stevens <stevensd@xxxxxxxxxxxx>
> Signed-off-by: Peter Xu <peterx@xxxxxxxxxx>

Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx>