Re: [PATCH bpf-next] bpf: make the verifier trace the "not qeual" for regs

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

On 12/10/23 5:00 AM, Menglong Dong wrote:
> We can derive some new information for BPF_JNE in regs_refine_cond_op().
> Take following code for example:
>
>    /* The type of "a" is u16 */
>    if (a > 0 && a < 100) {
>      /* the range of the register for a is [0, 99], not [1, 99],
>       * and will cause the following error:
>       *
>       *   invalid zero-sized read
>       *
>       * as a can be 0.
>       */
>      bpf_skb_store_bytes(skb, xx, xx, a, 0);
>    }

Could you have a C test to demonstrate this example?
Also, you should have a set of inline asm code (progs/verifier*.c)
to test various cases as in mark_reg32_not_equal() and
mark_reg_not_equal().

>
> In the code above, "a > 0" will be compiled to "jmp xxx if a == 0". In the
> TRUE branch, the dst_reg will be marked as known to 0. However, in the
> fallthrough(FALSE) branch, the dst_reg will not be handled, which makes
> the [min, max] for a is [0, 99], not [1, 99].
>
> For BPF_JNE, we can reduce the range of the dst reg if the src reg is a
> const and is exactly the edge of the dst reg.
>
> Signed-off-by: Menglong Dong <menglong8.dong@gmail.com>
> ---
>   kernel/bpf/verifier.c | 45 ++++++++++++++++++++++++++++++++++++++++++-
>   1 file changed, 44 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
> index 727a59e4a647..7b074ac93190 100644
> --- a/kernel/bpf/verifier.c
> +++ b/kernel/bpf/verifier.c
> @@ -1764,6 +1764,40 @@ static void __mark_reg_const_zero(struct bpf_reg_state *reg)
>   	reg->type = SCALAR_VALUE;
>   }
>   
> +#define CHECK_REG_MIN(value)			\
> +do {						\
> +	if ((value) == (typeof(value))imm)	\
> +		value++;			\
> +} while (0)
> +
> +#define CHECK_REG_MAX(value)			\
> +do {						\
> +	if ((value) == (typeof(value))imm)	\
> +		value--;			\
> +} while (0)
> +
> +static void mark_reg32_not_equal(struct bpf_reg_state *reg, u64 imm)
> +{

What if reg->s32_min_value == imm and reg->s32_max_value == imm?
Has this been handled in previous verifier logic?

> +		CHECK_REG_MIN(reg->s32_min_value);
> +		CHECK_REG_MAX(reg->s32_max_value);
> +		CHECK_REG_MIN(reg->u32_min_value);
> +		CHECK_REG_MAX(reg->u32_max_value);
> +}
> +
> +static void mark_reg_not_equal(struct bpf_reg_state *reg, u64 imm)
> +{
> +		CHECK_REG_MIN(reg->smin_value);
> +		CHECK_REG_MAX(reg->smax_value);
> +
> +		CHECK_REG_MIN(reg->umin_value);
> +		CHECK_REG_MAX(reg->umax_value);
> +
> +		CHECK_REG_MIN(reg->s32_min_value);
> +		CHECK_REG_MAX(reg->s32_max_value);
> +		CHECK_REG_MIN(reg->u32_min_value);
> +		CHECK_REG_MAX(reg->u32_max_value);
> +}
> +
>   static void mark_reg_known_zero(struct bpf_verifier_env *env,
>   				struct bpf_reg_state *regs, u32 regno)
>   {
> @@ -14332,7 +14366,16 @@ static void regs_refine_cond_op(struct bpf_reg_state *reg1, struct bpf_reg_state
>   		}
>   		break;
>   	case BPF_JNE:
> -		/* we don't derive any new information for inequality yet */
> +		/* try to recompute the bound of reg1 if reg2 is a const and
> +		 * is exactly the edge of reg1.
> +		 */
> +		if (is_reg_const(reg2, is_jmp32)) {
> +			val = reg_const_value(reg2, is_jmp32);
> +			if (is_jmp32)
> +				mark_reg32_not_equal(reg1, val);
> +			else
> +				mark_reg_not_equal(reg1, val);
> +		}
>   		break;
>   	case BPF_JSET:
>   		if (!is_reg_const(reg2, is_jmp32))