Skip to content

Commit

Permalink
bpf: Make 32->64 bounds propagation slightly more robust
Browse files Browse the repository at this point in the history
commit e572ff8 upstream.

Make the bounds propagation in __reg_assign_32_into_64() slightly more
robust and readable by aligning it similarly as we did back in the
__reg_combine_64_into_32() counterpart. Meaning, only propagate or
pessimize them as a smin/smax pair.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  • Loading branch information
borkmann authored and gregkh committed Dec 22, 2021
1 parent e2aad0b commit 279e0bf
Showing 1 changed file with 15 additions and 9 deletions.
24 changes: 15 additions & 9 deletions kernel/bpf/verifier.c
Expand Up @@ -1249,22 +1249,28 @@ static void __reg_bound_offset(struct bpf_reg_state *reg)
reg->var_off = tnum_or(tnum_clear_subreg(var64_off), var32_off);
}

static bool __reg32_bound_s64(s32 a)
{
return a >= 0 && a <= S32_MAX;
}

static void __reg_assign_32_into_64(struct bpf_reg_state *reg)
{
reg->umin_value = reg->u32_min_value;
reg->umax_value = reg->u32_max_value;
/* Attempt to pull 32-bit signed bounds into 64-bit bounds
* but must be positive otherwise set to worse case bounds
* and refine later from tnum.

/* Attempt to pull 32-bit signed bounds into 64-bit bounds but must
* be positive otherwise set to worse case bounds and refine later
* from tnum.
*/
if (reg->s32_min_value >= 0 && reg->s32_max_value >= 0)
reg->smax_value = reg->s32_max_value;
else
reg->smax_value = U32_MAX;
if (reg->s32_min_value >= 0)
if (__reg32_bound_s64(reg->s32_min_value) &&
__reg32_bound_s64(reg->s32_max_value)) {
reg->smin_value = reg->s32_min_value;
else
reg->smax_value = reg->s32_max_value;
} else {
reg->smin_value = 0;
reg->smax_value = U32_MAX;
}
}

static void __reg_combine_32_into_64(struct bpf_reg_state *reg)
Expand Down

0 comments on commit 279e0bf

Please sign in to comment.