Re: [PATCH bpf v2] bpf: Fix a verifier bug due to incorrect branch offset comparison with cpu=v4

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

On 11/29/23 7:19 PM, Andrii Nakryiko wrote:
> On Wed, Nov 29, 2023 at 4:15 PM Yonghong Song <yonghong.song@linux.dev> wrote:
>> Bpf cpu=v4 support is introduced in [1] and Commit 4cd58e9af8b9
>> ("bpf: Support new 32bit offset jmp instruction") added support for new
>> 32bit offset jmp instruction. Unfortunately, in function
>> bpf_adj_delta_to_off(), for new branch insn with 32bit offset, the offset
>> (plus/minor a small delta) compares to 16-bit offset bound
>> [S16_MIN, S16_MAX], which caused the following verification failure:
>>    $ ./test_progs-cpuv4 -t verif_scale_pyperf180
>>    ...
>>    insn 10 cannot be patched due to 16-bit range
>>    ...
>>    libbpf: failed to load object 'pyperf180.bpf.o'
>>    scale_test:FAIL:expect_success unexpected error: -12 (errno 12)
>>    #405     verif_scale_pyperf180:FAIL
>>
>> Note that due to recent llvm18 development, the patch [2] (already applied
>> in bpf-next) needs to be applied to bpf tree for testing purpose.
>>
>> The fix is rather simple. For 32bit offset branch insn, the adjusted
>> offset compares to [S32_MIN, S32_MAX] and then verification succeeded.
>>
>>    [1] https://lore.kernel.org/all/20230728011143.3710005-1-yonghong.song@linux.dev
>>    [2] https://lore.kernel.org/bpf/20231110193644.3130906-1-yonghong.song@linux.dev
>>
>> Fixes: 4cd58e9af8b9 ("bpf: Support new 32bit offset jmp instruction")
>> Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
>> ---
>>   kernel/bpf/core.c | 11 +++++++----
>>   1 file changed, 7 insertions(+), 4 deletions(-)
>>
>> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
>> index cd3afe57ece3..beff7e1d7fd0 100644
>> --- a/kernel/bpf/core.c
>> +++ b/kernel/bpf/core.c
>> @@ -371,14 +371,17 @@ static int bpf_adj_delta_to_imm(struct bpf_insn *insn, u32 pos, s32 end_old,
>>   static int bpf_adj_delta_to_off(struct bpf_insn *insn, u32 pos, s32 end_old,
>>                                  s32 end_new, s32 curr, const bool probe_pass)
>>   {
>> -       const s32 off_min = S16_MIN, off_max = S16_MAX;
>> +       s64 off_min = S16_MIN, off_max = S16_MAX;
>>          s32 delta = end_new - end_old;
>> -       s32 off;
>> +       s64 off;
>>
>> -       if (insn->code == (BPF_JMP32 | BPF_JA))
>> +       if (insn->code == (BPF_JMP32 | BPF_JA)) {
>>                  off = insn->imm;
>> -       else
>> +               off_min = S32_MIN;
>> +               off_max = S32_MAX;
>> +       } else {
> nit: it would be more symmetrical and easier to follow if you set
> S16_{MIN,MAX} in this branch, instead of using variable initialization
> approach

I tried to minimize the code change but probably not worth it.
If no further errors in this patch, should I send v3 with better
coding style or Maintainers could help do the change? Either
way, please let me know.

>
>>                  off = insn->off;
>> +       }
>>
>>          if (curr < pos && curr + off + 1 >= end_old)
>>                  off += delta;
>> --
>> 2.34.1
>>