bpf: Harden and/or/xor value tracking in verifier
This patch addresses a latent unsoundness issue in the scalar(32)_min_max_and/or/xor functions. While it is not a bugfix, it ensures that the functions produce sound outputs for all inputs. The issue occurs in these functions when setting signed bounds. The following example illustrates the issue for scalar_min_max_and(), but it applies to the other functions. In scalar_min_max_and() the following clause is executed when ANDing positive numbers: /* ANDing two positives gives a positive, so safe to * cast result into s64. */ dst_reg->smin_value = dst_reg->umin_value; dst_reg->smax_value = dst_reg->umax_value; However, if umin_value and umax_value of dst_reg cross the sign boundary (i.e., if (s64)dst_reg->umin_value > (s64)dst_reg->umax_value), then we will end up with smin_value > smax_value, which is unsound. Previous works [1, 2] have discovered and reported this issue. Our tool Agni [2, 3] consideres it a false positive. This is because, during the verification of the abstract operator scalar_min_max_and(), Agni restricts its inputs to those passing through reg_bounds_sync(). This mimics real-world verifier behavior, as reg_bounds_sync() is invariably executed at the tail of every abstract operator. Therefore, such behavior is unlikely in an actual verifier execution. However, it is still unsound for an abstract operator to set signed bounds such that smin_value > smax_value. This patch fixes it, making the abstract operator sound for all (well-formed) inputs. It is worth noting that while the previous code updated the signed bounds (using the output unsigned bounds) only when the *input signed* bounds were positive, the new code updates them whenever the *output unsigned* bounds do not cross the sign boundary. An alternative approach to fix this latent unsoundness would be to unconditionally set the signed bounds to unbounded [S64_MIN, S64_MAX], and let reg_bounds_sync() refine the signed bounds using the unsigned bounds and the tnum. We found that our approach produces more precise (tighter) bounds. For example, consider these inputs to BPF_AND: /* dst_reg */ var_off.value: 8608032320201083347 var_off.mask: 615339716653692460 smin_value: 8070450532247928832 smax_value: 8070450532247928832 umin_value: 13206380674380886586 umax_value: 13206380674380886586 s32_min_value: -2110561598 s32_max_value: -133438816 u32_min_value: 4135055354 u32_max_value: 4135055354 /* src_reg */ var_off.value: 8584102546103074815 var_off.mask: 9862641527606476800 smin_value: 2920655011908158522 smax_value: 7495731535348625717 umin_value: 7001104867969363969 umax_value: 8584102543730304042 s32_min_value: -2097116671 s32_max_value: 71704632 u32_min_value: 1047457619 u32_max_value: 4268683090 After going through tnum_and() -> scalar32_min_max_and() -> scalar_min_max_and() -> reg_bounds_sync(), our patch produces the following bounds for s32: s32_min_value: -1263875629 s32_max_value: -159911942 Whereas, setting the signed bounds to unbounded in scalar_min_max_and() produces: s32_min_value: -1263875629 s32_max_value: -1 As observed, our patch produces a tighter s32 bound. We also confirmed using Agni and SMT verification that our patch always produces signed bounds that are equal to or more precise than setting the signed bounds to unbounded in scalar_min_max_and(). [1] https://sanjit-bhat.github.io/assets/pdf/ebpf-verifier-range-analysis22.pdf [2] https://link.springer.com/chapter/10.1007/978-3-031-37709-9_12 [3] https://github.com/bpfverif/agniCo-developed-by:Matan Shachnai <m.shachnai@rutgers.edu> Signed-off-by:
Srinivas Narayana <srinivas.narayana@rutgers.edu> Signed-off-by:
Santosh Nagarakatte <santosh.nagarakatte@rutgers.edu> Signed-off-by:
Harishankar Vishwanathan <harishankar.vishwanathan@gmail.com> Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Acked-by:
Showing
Please register or sign in to comment