Re: [PATCH bpf-next v1 7/8] bpf: support for void/primitive __arg_untrusted global func params

[Eduard Zingerman <eddyz87@xxxxxxxxx>]

On Wed, 2025-07-02 at 20:20 -0700, Alexei Starovoitov wrote:
> On Wed, Jul 2, 2025 at 3:42 PM Eduard Zingerman <eddyz87@xxxxxxxxx> wrote:
> > 
> > Allow specifying __arg_untrusted for void */char */int */long *
> > parameters. Treat such parameters as
> > PTR_TO_MEM|MEM_RDONLY|PTR_UNTRUSTED of size zero.
> > Intended usage is as follows:
> > 
> >   int memcmp(char *a __arg_untrusted, char *b __arg_untrusted, size_t n) {
> >     bpf_for(i, 0, n) {
> >       if (a[i] - b[i])      // load at any offset is allowed
> >         return a[i] - b[i];
> >     }
> >     return 0;
> >   }
> 
> ...
> 
> > +bool btf_type_is_primitive(const struct btf_type *t)
> > +{
> > +       return (btf_type_is_int(t) && btf_type_int_is_regular(t)) ||
> > +              btf_is_any_enum(t);
> > +}
> 
> Should array of primitive types be allowed as well ?
> Since in C
>    int memcmp(char a[] __arg_untrusted, char b[] __arg_untrusted, size_t n) {
>      bpf_for(i, 0, n) {
>        if (a[i] - b[i])      // load at any offset is allowed
>          return a[i] - b[i];
> 
> will work just like 'char *'.

I agree in general, but compiler converts arrays to pointers for
function parameters, e.g.:

  [~/tmp]
  $ cat test-array-btf.c 
  int foo(int a[], char b[3]) {
    return 0;
  }
  [~/tmp]
  $ clang --target=bpf -c -g -O2 test-array-btf.c -o test-array-btf.o
  [~/tmp]
  $ bpftool btf dump file test-array-btf.o
  [1] PTR '(anon)' type_id=2
  [2] INT 'int' size=4 bits_offset=0 nr_bits=32 encoding=SIGNED
  [3] PTR '(anon)' type_id=4
  [4] INT 'char' size=1 bits_offset=0 nr_bits=8 encoding=SIGNED
  [5] FUNC_PROTO '(anon)' ret_type_id=2 vlen=2
          'a' type_id=1
          'b' type_id=3
  [6] FUNC 'foo' type_id=5 linkage=global

So, I'm inclined to skip this for now.