On Mon, Jun 15, 2020 at 9:08 PM Vlastimil Babka <vbabka@xxxxxxx> wrote:
>
> On 6/14/20 8:38 AM, Muchun Song wrote:
> > When a kmem_cache is initialized with SLAB_ACCOUNT slab flag, we must
> > not call kmem_cache_alloc with __GFP_ACCOUNT GFP flag. In this case,
> > we can be accounted to kmemcg twice. This is not correct. So we add a
>
> Are you sure? How does that happen?
>
> The only place I see these evaluated is this condition in slab_pre_alloc_hook():
>
> if (memcg_kmem_enabled() &&
> ((flags & __GFP_ACCOUNT) || (s->flags & SLAB_ACCOUNT)))
> return memcg_kmem_get_cache(s);
>
> And it doesn't matter if one or both are set? Am I missing something?
>
> > __GFP_ACCOUNT GFP flag check for slab allocation.
> >
> > We also introduce a new helper named fixup_gfp_flags to do that check.
> > We can reuse the fixup_gfp_flags for SLAB/SLUB.
> >
> > Signed-off-by: Muchun Song <songmuchun@xxxxxxxxxxxxx>
> > ---
> > mm/slab.c | 10 +---------
> > mm/slab.h | 21 +++++++++++++++++++++
> > mm/slub.c | 10 +---------
> > 3 files changed, 23 insertions(+), 18 deletions(-)
> >
> > diff --git a/mm/slab.c b/mm/slab.c
> > index 9350062ffc1a..6e0110bef2d6 100644
> > --- a/mm/slab.c
> > +++ b/mm/slab.c
> > @@ -126,8 +126,6 @@
> >
> > #include <trace/events/kmem.h>
> >
> > -#include "internal.h"
> > -
> > #include "slab.h"
> >
> > /*
> > @@ -2579,13 +2577,7 @@ static struct page *cache_grow_begin(struct kmem_cache *cachep,
> > * Be lazy and only check for valid flags here, keeping it out of the
> > * critical path in kmem_cache_alloc().
> > */
> > - if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
> > - gfp_t invalid_mask = flags & GFP_SLAB_BUG_MASK;
> > - flags &= ~GFP_SLAB_BUG_MASK;
> > - pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
> > - invalid_mask, &invalid_mask, flags, &flags);
> > - dump_stack();
> > - }
> > + flags = fixup_gfp_flags(cachep, flags);
> > WARN_ON_ONCE(cachep->ctor && (flags & __GFP_ZERO));
> > local_flags = flags & (GFP_CONSTRAINT_MASK|GFP_RECLAIM_MASK);
> >
> > diff --git a/mm/slab.h b/mm/slab.h
> > index 815e4e9a94cd..0b91f2a7b033 100644
> > --- a/mm/slab.h
> > +++ b/mm/slab.h
> > @@ -109,6 +109,7 @@ struct memcg_cache_params {
> > #include <linux/kmemleak.h>
> > #include <linux/random.h>
> > #include <linux/sched/mm.h>
> > +#include "internal.h"
> >
> > /*
> > * State of the slab allocator.
> > @@ -627,6 +628,26 @@ struct kmem_cache_node {
> >
> > };
> >
> > +static inline gfp_t fixup_gfp_flags(struct kmem_cache *s, gfp_t flags)
> > +{
> > + gfp_t invalid_mask = 0;
> > +
> > + if (unlikely(flags & GFP_SLAB_BUG_MASK))
> > + invalid_mask |= flags & GFP_SLAB_BUG_MASK;
> > +
> > + if (unlikely(flags & __GFP_ACCOUNT && s->flags & SLAB_ACCOUNT))
> > + invalid_mask |= __GFP_ACCOUNT;
> > +
> > + if (unlikely(invalid_mask)) {
> > + flags &= ~invalid_mask;
> > + pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
> > + invalid_mask, &invalid_mask, flags, &flags);
> > + dump_stack();
> > + }
> > +
> > + return flags;
> > +}
> > +
> > static inline struct kmem_cache_node *get_node(struct kmem_cache *s, int node)
> > {
> > return s->node[node];
> > diff --git a/mm/slub.c b/mm/slub.c
> > index b8f798b50d44..49b5cb7da318 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -37,8 +37,6 @@
> >
> > #include <trace/events/kmem.h>
> >
> > -#include "internal.h"
> > -
> > /*
> > * Lock order:
> > * 1. slab_mutex (Global Mutex)
> > @@ -1745,13 +1743,7 @@ static struct page *allocate_slab(struct kmem_cache *s, gfp_t flags, int node)
> >
> > static struct page *new_slab(struct kmem_cache *s, gfp_t flags, int node)
> > {
> > - if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
> > - gfp_t invalid_mask = flags & GFP_SLAB_BUG_MASK;
> > - flags &= ~GFP_SLAB_BUG_MASK;
> > - pr_warn("Unexpected gfp: %#x (%pGg). Fixing up to gfp: %#x (%pGg). Fix your code!\n",
> > - invalid_mask, &invalid_mask, flags, &flags);
> > - dump_stack();
> > - }
> > + flags = fixup_gfp_flags(s, flags);
> >
> > return allocate_slab(s,
> > flags & (GFP_RECLAIM_MASK | GFP_CONSTRAINT_MASK), node);
> >
>
Yeah, you are right. I'm very sorry that I was not thoughtful before.
Please ignore
this patch. Thanks!
--
Yours,
Muchun
