Re: [PATCH bpf-next v3 1/2] bpf: Skip bounds adjustment for conditional jumps on same scalar register

[Alexei Starovoitov]

On Fri, Oct 31, 2025 at 8:44 AM KaFai Wan <kafai.wan@xxxxxxxxx> wrote:
>
> When conditional jumps are performed on the same scalar register
> (e.g., r0 <= r0, r0 > r0, r0 < r0), the BPF verifier incorrectly
> attempts to adjust the register's min/max bounds. This leads to
> invalid range bounds and triggers a BUG warning.
>
> The problematic BPF program:
>    0: call bpf_get_prandom_u32
>    1: w8 = 0x80000000
>    2: r0 &= r8
>    3: if r0 > r0 goto <exit>
>
> The instruction 3 triggers kernel warning:
>    3: if r0 > r0 goto <exit>
>    true_reg1: range bounds violation u64=[0x1, 0x0] s64=[0x1, 0x0] u32=[0x1, 0x0] s32=[0x1, 0x0] var_off=(0x0, 0x0)
>    true_reg2: const tnum out of sync with range bounds u64=[0x0, 0xffffffffffffffff] s64=[0x8000000000000000, 0x7fffffffffffffff] var_off=(0x0, 0x0)
>
> Comparing a register with itself should not change its bounds and
> for most comparison operations, comparing a register with itself has
> a known result (e.g., r0 == r0 is always true, r0 < r0 is always false).
>
> Fix this by:
> 1. Enhance is_scalar_branch_taken() to properly handle branch direction
>    computation for same register comparisons across all BPF jump operations
> 2. Adds early return in reg_set_min_max() to avoid bounds adjustment
>    for unknown branch directions (e.g., BPF_JSET) on the same register
>
> The fix ensures that unnecessary bounds adjustments are skipped, preventing
> the verifier bug while maintaining correct branch direction analysis.
>
> Reported-by: Kaiyan Mei <M202472210@xxxxxxxxxxx>
> Reported-by: Yinhao Hu <dddddd@xxxxxxxxxxx>
> Closes: https://lore.kernel.org/all/1881f0f5.300df.199f2576a01.Coremail.kaiyanm@xxxxxxxxxxx/
> Signed-off-by: KaFai Wan <kafai.wan@xxxxxxxxx>
> ---
>  kernel/bpf/verifier.c | 33 +++++++++++++++++++++++++++++++++
>  1 file changed, 33 insertions(+)
>
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index 542e23fb19c7..a571263f4ebe 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -15995,6 +15995,8 @@ static int is_scalar_branch_taken(struct bpf_reg_state *reg1, struct bpf_reg_sta
>
>         switch (opcode) {
>         case BPF_JEQ:
> +               if (reg1 == reg2)
> +                       return 1;
>                 /* constants, umin/umax and smin/smax checks would be
>                  * redundant in this case because they all should match
>                  */
> @@ -16021,6 +16023,8 @@ static int is_scalar_branch_taken(struct bpf_reg_state *reg1, struct bpf_reg_sta
>                 }
>                 break;
>         case BPF_JNE:
> +               if (reg1 == reg2)
> +                       return 0;
>                 /* constants, umin/umax and smin/smax checks would be
>                  * redundant in this case because they all should match
>                  */
> @@ -16047,6 +16051,12 @@ static int is_scalar_branch_taken(struct bpf_reg_state *reg1, struct bpf_reg_sta
>                 }
>                 break;
>         case BPF_JSET:
> +               if (reg1 == reg2) {
> +                       if (tnum_is_const(t1))
> +                               return t1.value != 0;
> +                       else
> +                               return (smin1 <= 0 && smax1 >= 0) ? -1 : 1;
> +               }
>                 if (!is_reg_const(reg2, is_jmp32)) {
>                         swap(reg1, reg2);
>                         swap(t1, t2);
> @@ -16059,48 +16069,64 @@ static int is_scalar_branch_taken(struct bpf_reg_state *reg1, struct bpf_reg_sta
>                         return 0;
>                 break;
>         case BPF_JGT:
> +               if (reg1 == reg2)
> +                       return 0;
>                 if (umin1 > umax2)
>                         return 1;
>                 else if (umax1 <= umin2)
>                         return 0;
>                 break;
>         case BPF_JSGT:
> +               if (reg1 == reg2)
> +                       return 0;

This is uglier than the previous version.
reg1 == reg2 is a syzbot territory.
We shouldn't uglify the code everywhere because of it.

pw-bot: cr