Re: [bpf-next 1/4] selftests/bpf: map_kptr: force BPF_STX for the scalar store to kptr

[Yonghong Song <yonghong.song@linux.dev>]

On 8/6/26 9:50 AM, Vineet Gupta wrote:
> On 8/6/26 9:37 AM, Yonghong Song wrote:
>> On 8/5/26 12:25 PM, Vineet Gupta wrote:
>>> On 8/5/26 10:47 AM, Yonghong Song wrote:
>>>> On 8/3/26 10:02 AM, Vineet Gupta wrote:
>>>>> reject_scalar_store_to_kptr stores a scalar constant to a kptr field:
>>>>>
>>>>>            *(volatile u64 *)&v->unref_ptr = 0xBADC0DE;
>>>>>
>>>>> Compilers generate one of two encodings for that:
>>>>>
>>>>>     1. Materialize the constant into a register and emit BPF_STX:
>>>>>
>>>>>            r1 = 0xbadc0de
>>>>>            *(u64 *)(r0 + 0x8) = r1
>>>>>
>>>>>     2. Or fold it into a single BPF_ST (store immediate):
>>>>>
>>>>>            *(u64 *)(r0 + 0x8) = 0xbadc0de
>>>>>
>>>>> check_map_kptr_access() rejects both, but through very different
>>>>> checks.
>>>>> BPF_STX goes through map_kptr_match_type(), whose first test is
>>>>> base_type(reg->type) != PTR_TO_BTF_ID - the scalar rejection this
>>>>> test is
>>>>> named for - and which prints "invalid kptr access, R...". BPF_ST
>>>>> only gets
>>>>> the trivial "BPF_ST imm must be 0 when storing to kptr" immediate
>>>>> check and
>>>>> never reaches map_kptr_match_type() at all.
>>>>>
>>>>> So on a compiler that folds the constant - bpf-gcc, and clang from
>>>>> -mcpu=v4, which enabled BPF_ST around v4 support due to historical
>>>>> verifier limitations - the test fails against its expected message.
>>>>>
>>>>> Widening the __msg to accept either message would make it pass
>>>>> again, but
>>>>> on those toolchains it would then only re-test the imm != 0 path, 
>>>>> which
>>>>> verifier/map_kptr.c ("map_kptr: BPF_ST imm != 0") already covers,
>>>>> and the
>>>>> scalar-vs-PTR_TO_BTF_ID check would lose its only test in the tree.
>>>>>
>>>>> Route the value through barrier_var() instead, so the store stays a
>>>>> BPF_STX everywhere and the test keeps asserting what it was 
>>>>> written to
>>>>> assert. clang -mcpu=v1..v4 and bpf-gcc 16.1 all emit the register 
>>>>> form
>>>>> afterwards.
>>>>>
>>>>>      bpf-gcc, before: #229/20 
>>>>> map_kptr/reject_scalar_store_to_kptr:FAIL
>>>>>      bpf-gcc, after : #229/20 map_kptr/reject_scalar_store_to_kptr:OK
>>>>>
>>>>> Signed-off-by: Vineet Gupta <vineet.gupta@linux.dev>
>>>>> ---
>>>>>     tools/testing/selftests/bpf/progs/map_kptr_fail.c | 11 
>>>>> ++++++++++-
>>>>>     1 file changed, 10 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/tools/testing/selftests/bpf/progs/map_kptr_fail.c
>>>>> b/tools/testing/selftests/bpf/progs/map_kptr_fail.c
>>>>> index f11848dfa78f..cb84e23b83c0 100644
>>>>> --- a/tools/testing/selftests/bpf/progs/map_kptr_fail.c
>>>>> +++ b/tools/testing/selftests/bpf/progs/map_kptr_fail.c
>>>>> @@ -390,13 +390,22 @@ __failure __msg("invalid kptr access, R")
>>>>>     int reject_scalar_store_to_kptr(struct __sk_buff *ctx)
>>>>>     {
>>>>>         struct map_value *v;
>>>>> +    u64 val = 0xBADC0DE;
>>>>>         int key = 0;
>>>>>            v = bpf_map_lookup_elem(&array_map, &key);
>>>>>         if (!v)
>>>>>             return 0;
>>>>>     -    *(volatile u64 *)&v->unref_ptr = 0xBADC0DE;
>>>>> +    /*
>>>>> +     * Keep the value in a register so this stays a BPF_STX and 
>>>>> keeps
>>>>> +     * exercising map_kptr_match_type(). Compilers that fold the
>>>>> constant
>>>>> +     * into a BPF_ST (store immediate) instead - bpf-gcc, and clang
>>>>> from
>>>>> +     * -mcpu=v4 - would be rejected by the far weaker "BPF_ST imm
>>>>> must be
>>>>> +     * 0" check, which verifier/map_kptr.c already covers.
>>>>> +     */
>>>>> +    barrier_var(val);
>>>>> +    *(volatile u64 *)&v->unref_ptr = val;
>>>> Could you explain in details why gcc16 (with cpuv4) won't work with
>>>> BPF_ST?
>>>> What code gcc16 (with cpuv4) generates?
>>> GCC generates BPF_ST (and clang -mcpu=v4 does as well)
>>>
>>>     7: (7a) *(u64 *)(r0 +8) = 195936478
>>>     BPF_ST imm must be 0 when storing to kptr at off=8
>>>     processed 7 insns (limit 1000000) max_states_per_insn 0 
>>> total_states
>>>     0 peak_states 0 mark_read 0
>>>     =============
>>>     EXPECTED   SUBSTR: 'invalid kptr access, R'
>>>
>>>
>>> The issue is not codegen or BPF_ST vs. BPF_STX, its a deliberate bad
>>> write to a pointer.
>>> The current __msg in test only matches v3 (STX) form. v4's
>>> store-immediate hits a different verifier check with a different 
>>> message.
>>> My first approach was to support both in the __msg, but then Claude
>>> suggested to not do it that way because it would reduce coverage as
>>> mentioned in the changelog above.
>> For clang side, we can do:
>>
>> diff --git a/tools/testing/selftests/bpf/progs/map_kptr_fail.c 
>> b/tools/testing/selftests/bpf/progs/map_kptr_fail.c
>> index f11848dfa78f..b9f6d6764431 100644
>> --- a/tools/testing/selftests/bpf/progs/map_kptr_fail.c
>> +++ b/tools/testing/selftests/bpf/progs/map_kptr_fail.c
>> @@ -386,7 +386,11 @@ int kptr_xchg_possibly_null(struct __sk_buff *ctx)
>>    }
>>       SEC("?tc")
>> +#ifdef __BPF_FEATURE_ST
>> +__failure __msg("BPF_ST imm must be 0 when storing to kptr at off=8")
>> +#else
>>    __failure __msg("invalid kptr access, R")
>> +#endif
>>    int reject_scalar_store_to_kptr(struct __sk_buff *ctx)
>>    {
>>           struct map_value *v;
>> ...
>>
>> Note that __BPF_FEATURE_ST is only enabled for cpuv4 in clang.
>> Does gcc has similar macro like above __BPF_FEATURE_ST?
>
> Yes it does and also similarly enabled for cpu=v4.
> The suggestion does look cleaner, but feels fragile.
> Both forms of codegen are valid from selftest passing perspective.
> BPF_FEATURE_ST implies compiler supports the feature but the actual 
> generation is subject to optimization level etc.
>
> Your call.

This is just for selftests. Let us have the above change to accommodate both cpuv2/cpuv3 and cpuv4.

>
> Thx,
> -Vineet