Re: [PATCH bpf-next v2 5/6] selftests/bpf: Add tests for pointer type merge at a shared load

On Fri, 2026-08-14 at 23:52 +0200, Daniel Borkmann wrote:
> Cover the ways in which the type recorded for a shared load used to lose
> the BPF_PROBE_MEM rewrite which would then trigger a NULL deref if not
> handled properly.
> 
>   # LDLIBS=-static PKG_CONFIG='pkg-config --static' ./vmtest.sh -- ./test_progs -t mem_rdonly_untrusted
>   [...]
>   #238/1   mem_rdonly_untrusted/btf_id_to_ptr_mem:OK
>   #238/2   mem_rdonly_untrusted/ldx_is_ok_bad_addr:OK
>   #238/3   mem_rdonly_untrusted/ldx_is_ok_good_addr:OK
>   #238/4   mem_rdonly_untrusted/offset_not_tracked:OK
>   #238/5   mem_rdonly_untrusted/stx_not_ok:OK
>   #238/6   mem_rdonly_untrusted/atomic_not_ok:OK
>   #238/7   mem_rdonly_untrusted/atomic_rmw_not_ok:OK
>   #238/8   mem_rdonly_untrusted/kfunc_param_not_ok:OK
>   #238/9   mem_rdonly_untrusted/mixed_mem_type:OK
>   #238/10  mem_rdonly_untrusted/mixed_mem_untrusted_btf_id_type:OK
>   #238/11  mem_rdonly_untrusted/mixed_mem_btf_id_type:OK
>   #238/12  mem_rdonly_untrusted/mixed_rdonly_mem_btf_id_type:OK
>   #238/13  mem_rdonly_untrusted/mixed_mem_mem_type:OK
>   #238/14  mem_rdonly_untrusted/mixed_map_value_mem_type:OK
>   #238/15  mem_rdonly_untrusted/mixed_stack_mem_type:OK
>   #238/16  mem_rdonly_untrusted/diff_size_access:OK
>   #238/17  mem_rdonly_untrusted/misaligned_access:OK
>   #238/18  mem_rdonly_untrusted/null_check:OK
>   #238/19  mem_rdonly_untrusted/ldx_is_ok_commuted_addr:OK
>   #238/20  mem_rdonly_untrusted/helper_param_not_ok:OK
>   #238     mem_rdonly_untrusted:OK
>   Summary: 1/20 PASSED, 0 SKIPPED, 0/0 FAILED
> 
> Signed-off-by: Daniel Borkmann <[email protected]>
> ---

Hi Daniel,

GCC-BPF fails for the series [1]: because of the difference in
codegen some __msg patterns do not match. How would you like to handle
this, land first 4 patches and do the tests as a follow-up or as a v3?

[1] https://github.com/kernel-patches/bpf/actions/runs/31970050902/job/95222785689

...