On Mon, Mar 2, 2020 at 2:37 PM Alexei Starovoitov
<alexei.starovoitov@xxxxxxxxx> wrote:
>
> On Sat, Feb 29, 2020 at 10:24:03PM -0800, Andrii Nakryiko wrote:
> > Switch BPF UAPI constants, previously defined as #define macro, to anonymous
> > enum values. This preserves constants values and behavior in expressions, but
> > has added advantaged of being captured as part of DWARF and, subsequently, BTF
> > type info. Which, in turn, greatly improves usefulness of generated vmlinux.h
> > for BPF applications, as it will not require BPF users to copy/paste various
> > flags and constants, which are frequently used with BPF helpers.
> >
> > Signed-off-by: Andrii Nakryiko <andriin@xxxxxx>
> > ---
> > include/uapi/linux/bpf.h | 272 +++++++++++++++----------
> > include/uapi/linux/bpf_common.h | 86 ++++----
> > include/uapi/linux/btf.h | 60 +++---
> > tools/include/uapi/linux/bpf.h | 274 ++++++++++++++++----------
> > tools/include/uapi/linux/bpf_common.h | 86 ++++----
> > tools/include/uapi/linux/btf.h | 60 +++---
> > 6 files changed, 497 insertions(+), 341 deletions(-)
>
> I see two reasons why converting #define to enum is useful:
> 1. bpf progs can use them from vmlinux.h as evident in patch 3.
> 2. "bpftool feature probe" can be replaced with
> bpftool btf dump file /sys/kernel/btf/vmlinux |grep BPF_CGROUP_SETSOCKOPT
>
> The second use case is already possible, since bpf_prog_type,
> bpf_attach_type, bpf_cmd, bpf_func_id are all enums.
> So kernel is already self describing most bpf features.
> Does kernel support bpf_probe_read_user() ? Answer is:
> bpftool btf dump file /sys/kernel/btf/vmlinux | grep BPF_FUNC_probe_read_user
>
> The only bit missing is supported kernel flags and instructions.
Yep, my motivation was primarily the former, but I can see benefits
from the latter as well.
>
> I think for now I would only convert flags that are going to be
> used from bpf program and see whether 1st use case works well.
> Later we can convert flags that are used out of user space too.
>
> In other words:
>
> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > index 8e98ced0963b..03e08f256bd1 100644
> > --- a/include/uapi/linux/bpf.h
> > +++ b/include/uapi/linux/bpf.h
> > @@ -14,34 +14,36 @@
> > /* Extended instruction set based on top of classic BPF */
> >
> > /* instruction classes */
> > -#define BPF_JMP32 0x06 /* jmp mode in word width */
> > -#define BPF_ALU64 0x07 /* alu mode in double word width */
> > +enum {
> > + BPF_JMP32 = 0x06, /* jmp mode in word width */
> > + BPF_ALU64 = 0x07, /* alu mode in double word width */
>
> not those.
makes sense
>
> > -#define BPF_F_ALLOW_OVERRIDE (1U << 0)
> > -#define BPF_F_ALLOW_MULTI (1U << 1)
> > -#define BPF_F_REPLACE (1U << 2)
> > +enum {
> > + BPF_F_ALLOW_OVERRIDE = (1U << 0),
> > + BPF_F_ALLOW_MULTI = (1U << 1),
> > + BPF_F_REPLACE = (1U << 2),
> > +};
>
> not those either. These are the flags for user space. Not for the prog.
yep...
>
> > /* flags for BPF_MAP_UPDATE_ELEM command */
> > -#define BPF_ANY 0 /* create new element or update existing */
> > -#define BPF_NOEXIST 1 /* create new element if it didn't exist */
> > -#define BPF_EXIST 2 /* update existing element */
> > -#define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */
> > +enum {
> > + BPF_ANY = 0, /* create new element or update existing */
> > + BPF_NOEXIST = 1, /* create new element if it didn't exist */
> > + BPF_EXIST = 2, /* update existing element */
> > + BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
> > +};
>
> yes to these.
yep, these and below are the most important ones...
[...]
>
> In all such cases I don't think we need #define FOO FOO
> trick. These are the flags used within bpf program.
> I don't think any user is doing #ifdef logic there.
> I cannot come up with a use case of anything useful this way.
Sounds good, I'll revert non-BPF helper flags cases and will post v2, thanks!
