On Wed, Feb 14, 2024 at 9:59 AM Vlastimil Babka <vbabka@xxxxxxx> wrote:
>
> On 2/12/24 22:38, Suren Baghdasaryan wrote:
> > Currently slab pages can store only vectors of obj_cgroup pointers in
> > page->memcg_data. Introduce slabobj_ext structure to allow more data
> > to be stored for each slab object. Wrap obj_cgroup into slabobj_ext
> > to support current functionality while allowing to extend slabobj_ext
> > in the future.
> >
> > Signed-off-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
>
> ...
>
> > +static inline bool need_slab_obj_ext(void)
> > +{
> > + /*
> > + * CONFIG_MEMCG_KMEM creates vector of obj_cgroup objects conditionally
> > + * inside memcg_slab_post_alloc_hook. No other users for now.
> > + */
> > + return false;
> > +}
> > +
> > +static inline struct slabobj_ext *
> > +prepare_slab_obj_exts_hook(struct kmem_cache *s, gfp_t flags, void *p)
> > +{
> > + struct slab *slab;
> > +
> > + if (!p)
> > + return NULL;
> > +
> > + if (!need_slab_obj_ext())
> > + return NULL;
> > +
> > + slab = virt_to_slab(p);
> > + if (!slab_obj_exts(slab) &&
> > + WARN(alloc_slab_obj_exts(slab, s, flags, false),
> > + "%s, %s: Failed to create slab extension vector!\n",
> > + __func__, s->name))
> > + return NULL;
> > +
> > + return slab_obj_exts(slab) + obj_to_index(s, slab, p);
>
> This is called in slab_post_alloc_hook() and the result stored to obj_exts
> but unused. Maybe introduce this only in a later patch where it becomes
> relevant?
Ack. I'll move it into the patch where we start using obj_exts.
>
> > --- a/mm/slab_common.c
> > +++ b/mm/slab_common.c
> > @@ -201,6 +201,54 @@ struct kmem_cache *find_mergeable(unsigned int size, unsigned int align,
> > return NULL;
> > }
> >
> > +#ifdef CONFIG_SLAB_OBJ_EXT
> > +/*
> > + * The allocated objcg pointers array is not accounted directly.
> > + * Moreover, it should not come from DMA buffer and is not readily
> > + * reclaimable. So those GFP bits should be masked off.
> > + */
> > +#define OBJCGS_CLEAR_MASK (__GFP_DMA | __GFP_RECLAIMABLE | \
> > + __GFP_ACCOUNT | __GFP_NOFAIL)
> > +
> > +int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
> > + gfp_t gfp, bool new_slab)
>
> Since you're moving this function between files anyway, could you please
> instead move it to mm/slub.c. I expect we'll eventually (maybe even soon)
> move the rest of performance sensitive kmemcg hooks there as well to make
> inlining possible.
Will do.
>
> > +{
> > + unsigned int objects = objs_per_slab(s, slab);
> > + unsigned long obj_exts;
> > + void *vec;
> > +
> > + gfp &= ~OBJCGS_CLEAR_MASK;
> > + vec = kcalloc_node(objects, sizeof(struct slabobj_ext), gfp,
> > + slab_nid(slab));
> > + if (!vec)
> > + return -ENOMEM;
> > +
> > + obj_exts = (unsigned long)vec;
> > +#ifdef CONFIG_MEMCG
> > + obj_exts |= MEMCG_DATA_OBJEXTS;
> > +#endif
> > + if (new_slab) {
> > + /*
> > + * If the slab is brand new and nobody can yet access its
> > + * obj_exts, no synchronization is required and obj_exts can
> > + * be simply assigned.
> > + */
> > + slab->obj_exts = obj_exts;
> > + } else if (cmpxchg(&slab->obj_exts, 0, obj_exts)) {
> > + /*
> > + * If the slab is already in use, somebody can allocate and
> > + * assign slabobj_exts in parallel. In this case the existing
> > + * objcg vector should be reused.
> > + */
> > + kfree(vec);
> > + return 0;
> > + }
> > +
> > + kmemleak_not_leak(vec);
> > + return 0;
> > +}
> > +#endif /* CONFIG_SLAB_OBJ_EXT */
> > +
> > static struct kmem_cache *create_cache(const char *name,
> > unsigned int object_size, unsigned int align,
> > slab_flags_t flags, unsigned int useroffset,
> > diff --git a/mm/slub.c b/mm/slub.c
> > index 2ef88bbf56a3..1eb1050814aa 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -683,10 +683,10 @@ static inline bool __slab_update_freelist(struct kmem_cache *s, struct slab *sla
> >
> > if (s->flags & __CMPXCHG_DOUBLE) {
> > ret = __update_freelist_fast(slab, freelist_old, counters_old,
> > - freelist_new, counters_new);
> > + freelist_new, counters_new);
> > } else {
> > ret = __update_freelist_slow(slab, freelist_old, counters_old,
> > - freelist_new, counters_new);
> > + freelist_new, counters_new);
> > }
> > if (likely(ret))
> > return true;
> > @@ -710,13 +710,13 @@ static inline bool slab_update_freelist(struct kmem_cache *s, struct slab *slab,
> >
> > if (s->flags & __CMPXCHG_DOUBLE) {
> > ret = __update_freelist_fast(slab, freelist_old, counters_old,
> > - freelist_new, counters_new);
> > + freelist_new, counters_new);
> > } else {
> > unsigned long flags;
> >
> > local_irq_save(flags);
> > ret = __update_freelist_slow(slab, freelist_old, counters_old,
> > - freelist_new, counters_new);
> > + freelist_new, counters_new);
>
> I can see the mixing of tabs and spaces is wrong but perhaps not fix it as
> part of the series?
I'll fix them in the next version.
>
> > local_irq_restore(flags);
> > }
> > if (likely(ret))
> > @@ -1881,13 +1881,25 @@ static inline enum node_stat_item cache_vmstat_idx(struct kmem_cache *s)
> > NR_SLAB_RECLAIMABLE_B : NR_SLAB_UNRECLAIMABLE_B;
> > }
> >
> > -#ifdef CONFIG_MEMCG_KMEM
> > -static inline void memcg_free_slab_cgroups(struct slab *slab)
> > +#ifdef CONFIG_SLAB_OBJ_EXT
> > +static inline void free_slab_obj_exts(struct slab *slab)
>
> Right, freeing is already here, so makes sense put the allocation here as well.
>
> > @@ -3817,6 +3820,7 @@ void slab_post_alloc_hook(struct kmem_cache *s, struct obj_cgroup *objcg,
> > kmemleak_alloc_recursive(p[i], s->object_size, 1,
> > s->flags, init_flags);
> > kmsan_slab_alloc(s, p[i], init_flags);
> > + obj_exts = prepare_slab_obj_exts_hook(s, flags, p[i]);
>
> Yeah here's the hook used. Doesn't it generate a compiler warning? Maybe at
> least postpone the call until the result is further used.
Yes, I'll move that into the patch where we start using it.
Thanks for the review, Vlastimil!
>
> > }
> >
> > memcg_slab_post_alloc_hook(s, objcg, flags, size, p);
>
> --
> To unsubscribe from this group and stop receiving emails from it, send an email to kernel-team+unsubscribe@xxxxxxxxxxx.
>
