As suggested by Maciej Żenczykowski in [1], extend try_match_pkt_pointers
to allow verification of BPF, generated for C code like below:
if (data + 42 != data_end) { ... }
Also simplify try_match_pkt_pointers to avoid checking both
'pkt <op> pkt_end' and 'pkt_end <op> pkt' conditions,
as suggested by Andrii Nakryiko.
[1] https://lore.kernel.org/bpf/CAHo-Oow5V2u4ZYvzuR8NmJmFDPNYp0pQDJX66rZqUjFHvhx82A@xxxxxxxxxxxxxx/
Eduard Zingerman (3):
bpf: simplify try_match_pkt_pointers()
bpf: infer packet range for 'if pkt ==/!= pkt_end' comparisons
selftests/bpf: test packet range inference for 'if pkt ==/!= pkt_end'
kernel/bpf/verifier.c | 112 ++++----------
.../bpf/progs/verifier_direct_packet_access.c | 138 ++++++++++++++++++
2 files changed, 170 insertions(+), 80 deletions(-)
--
2.43.0
