On Thu, Mar 13, 2025 at 04:15:38PM +0000, Matthew Wilcox wrote:
> On Thu, Mar 13, 2025 at 12:03:48PM -0400, Johannes Weiner wrote:
> > > static inline bool folio_memcg_charged(struct folio *folio)
> > > {
> > > - if (folio_memcg_kmem(folio))
> > > - return __folio_objcg(folio) != NULL;
> > > - return __folio_memcg(folio) != NULL;
> > > + return folio->memcg_data != 0;
> >
> > I suppose we lose a few DEBUG_VM asserts, but no doubt this function
> > is unnecessarily complicated.
>
> I think we lose two -- folio_test_slab() and
> memcg_data & MEMCG_DATA_OBJEXTS, both of which seem pretty low value.
> Also, one could argue that asking "is this memory charged" should
> return true if we ask it about a slab. So yeah, I didn't think they
> were worth preserving.
>
With separete struct slab, no one should be calling folio_memcg_charged()
with it, so I think these asserts were unnecessary.
>
