[v2,bpf-next,0/5] bpf: Support access to bpf map fields

v1->v2:
- move btf id cache to a new bpf_map_ops.map_btf_id field (Martin, Andrii);
- don't check btf names for collisions (Martin);
- drop btf_find_by_name_kind_next() patch since it was needed only for
  collision check;
- don't fall back to `struct bpf_map` if a map type doesn't specify both
  map_btf_name and map_btf_id;

This patch set adds support to access bpf map fields from bpf programs
using btf_struct_access().

That way a program can cast a pointer to map to either `struct bpf_map *`
or map type specific struct pointer such as `struct bpf_array *` or
`struct bpf_htab *`, and access necessary fields, e.g. map->max_entries.

The fields, in turn, should be defined by a user provided struct with
preserve_access_index attribute or included from vmlinux.h.

Please see patch 3 for more details on the feature and use-cases.

Other patches:

Patch 1 is refactoring to simplify btf_parse_vmlinux().
Patch 2 is a rename to avoid having two different `struct bpf_htab`.

Patch 4 enables access to map fields for all map types.
Patch 5 adds selftests.


Andrey Ignatov (5):
  bpf: Switch btf_parse_vmlinux to btf_find_by_name_kind
  bpf: Rename bpf_htab to bpf_shtab in sock_map
  bpf: Support access to bpf map fields
  bpf: Set map_btf_{name,id} for all map types
  selftests/bpf: Test access to bpf map pointer

 include/linux/bpf.h                           |   9 +
 include/linux/bpf_verifier.h                  |   1 +
 kernel/bpf/arraymap.c                         |  18 +
 kernel/bpf/bpf_struct_ops.c                   |   3 +
 kernel/bpf/btf.c                              |  63 +-
 kernel/bpf/cpumap.c                           |   3 +
 kernel/bpf/devmap.c                           |   6 +
 kernel/bpf/hashtab.c                          |  15 +
 kernel/bpf/local_storage.c                    |   3 +
 kernel/bpf/lpm_trie.c                         |   3 +
 kernel/bpf/queue_stack_maps.c                 |   6 +
 kernel/bpf/reuseport_array.c                  |   3 +
 kernel/bpf/ringbuf.c                          |   3 +
 kernel/bpf/stackmap.c                         |   3 +
 kernel/bpf/verifier.c                         |  82 ++-
 net/core/bpf_sk_storage.c                     |   3 +
 net/core/sock_map.c                           |  88 +--
 net/xdp/xskmap.c                              |   3 +
 .../selftests/bpf/prog_tests/map_ptr.c        |  32 +
 .../selftests/bpf/progs/map_ptr_kern.c        | 686 ++++++++++++++++++
 .../testing/selftests/bpf/verifier/map_ptr.c  |  62 ++
 .../selftests/bpf/verifier/map_ptr_mixing.c   |   2 +-
 22 files changed, 1030 insertions(+), 67 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/map_ptr.c
 create mode 100644 tools/testing/selftests/bpf/progs/map_ptr_kern.c
 create mode 100644 tools/testing/selftests/bpf/verifier/map_ptr.c

On Fri, Jun 19, 2020 at 02:11:40PM -0700, Andrey Ignatov wrote:
> v1->v2:
> - move btf id cache to a new bpf_map_ops.map_btf_id field (Martin, Andrii);
> - don't check btf names for collisions (Martin);
> - drop btf_find_by_name_kind_next() patch since it was needed only for
>   collision check;
> - don't fall back to `struct bpf_map` if a map type doesn't specify both
>   map_btf_name and map_btf_id;
> 
> This patch set adds support to access bpf map fields from bpf programs
> using btf_struct_access().
Acked-by: Martin KaFai Lau <kafai@fb.com>

On 6/19/20 11:11 PM, Andrey Ignatov wrote:
> v1->v2:
> - move btf id cache to a new bpf_map_ops.map_btf_id field (Martin, Andrii);
> - don't check btf names for collisions (Martin);
> - drop btf_find_by_name_kind_next() patch since it was needed only for
>    collision check;
> - don't fall back to `struct bpf_map` if a map type doesn't specify both
>    map_btf_name and map_btf_id;
> 
> This patch set adds support to access bpf map fields from bpf programs
> using btf_struct_access().
> 
> That way a program can cast a pointer to map to either `struct bpf_map *`
> or map type specific struct pointer such as `struct bpf_array *` or
> `struct bpf_htab *`, and access necessary fields, e.g. map->max_entries.
> 
> The fields, in turn, should be defined by a user provided struct with
> preserve_access_index attribute or included from vmlinux.h.
> 
> Please see patch 3 for more details on the feature and use-cases.
> 
> Other patches:
> 
> Patch 1 is refactoring to simplify btf_parse_vmlinux().
> Patch 2 is a rename to avoid having two different `struct bpf_htab`.
> 
> Patch 4 enables access to map fields for all map types.
> Patch 5 adds selftests.
> 

Applied, thanks!