Re: [PATCH v3 bpf-next 1/3] bpf: add mmap() support for BPF_MAP_TYPE_ARRAY

[Andrii Nakryiko <andrii.nakryiko@gmail.com>]

On Wed, Nov 13, 2019 at 12:06 PM John Fastabend
<john.fastabend@gmail.com> wrote:
>
> Andrii Nakryiko wrote:
> > Add ability to memory-map contents of BPF array map. This is extremely useful
> > for working with BPF global data from userspace programs. It allows to avoid
> > typical bpf_map_{lookup,update}_elem operations, improving both performance
> > and usability.
> >
> > There had to be special considerations for map freezing, to avoid having
> > writable memory view into a frozen map. To solve this issue, map freezing and
> > mmap-ing is happening under mutex now:
> >   - if map is already frozen, no writable mapping is allowed;
> >   - if map has writable memory mappings active (accounted in map->writecnt),
> >     map freezing will keep failing with -EBUSY;
> >   - once number of writable memory mappings drops to zero, map freezing can be
> >     performed again.
> >
> > Only non-per-CPU plain arrays are supported right now. Maps with spinlocks
> > can't be memory mapped either.
> >
> > For BPF_F_MMAPABLE array, memory allocation has to be done through vmalloc()
> > to be mmap()'able. We also need to make sure that array data memory is
> > page-sized and page-aligned, so we over-allocate memory in such a way that
> > struct bpf_array is at the end of a single page of memory with array->value
> > being aligned with the start of the second page. On deallocation we need to
> > accomodate this memory arrangement to free vmalloc()'ed memory correctly.
> >
> > Cc: Rik van Riel <riel@surriel.com>
> > Cc: Johannes Weiner <hannes@cmpxchg.org>
> > Acked-by: Song Liu <songliubraving@fb.com>
> > Signed-off-by: Andrii Nakryiko <andriin@fb.com>
>
> [...]
>
> > @@ -102,10 +106,20 @@ static struct bpf_map *array_map_alloc(union bpf_attr *attr)
> >       }
> >
> >       array_size = sizeof(*array);
> > -     if (percpu)
> > +     if (percpu) {
> >               array_size += (u64) max_entries * sizeof(void *);
> > -     else
> > -             array_size += (u64) max_entries * elem_size;
> > +     } else {
> > +             /* rely on vmalloc() to return page-aligned memory and
> > +              * ensure array->value is exactly page-aligned
> > +              */
> > +             if (attr->map_flags & BPF_F_MMAPABLE) {
> > +                     array_size = round_up(array_size, PAGE_SIZE);
> > +                     array_size += (u64) max_entries * elem_size;
> > +                     array_size = round_up(array_size, PAGE_SIZE);
> > +             } else {
> > +                     array_size += (u64) max_entries * elem_size;
> > +             }
> > +     }
>
> Thought about this chunk for a bit, assuming we don't end up with lots of
> small mmap arrays it should be OK. So userspace will probably need to try and
> optimize this to create as few mmaps as possible.

I think typically most explicitly declared maps won't be
BPF_F_MMAPABLE, unless user really expects to mmap() it for use from
user-space. For global data, though, the benefits are really great
from being able to mmap(), which is why I'm defaulting them to
BPF_F_MMAPABLE by default, if possible.

>
> [...]
>
> Acked-by: John Fastabend <john.fastabend@gmail.com>