Message ID | 20170816170156.11594-1-fw@strlen.de |
---|---|
State | Accepted |
Delegated to: | Florian Westphal |
Headers | show |
On Wed, Aug 16, 2017 at 07:01:55PM +0200, Florian Westphal wrote: > nft add rule .. ip ttl set 64 > > erronously mangles ip protocol instead of ttl. > > Because the kernel can't deal with odd-sized data (ttl is one byte) when > doing checksum fixups, so the write to 'ttl' is turned into > > [ payload load 2b @ network header + 8 => reg 1 ] > [ bitwise reg 1 = (reg=1 & 0x000000ff ) ^ $new_value ] > [ payload write reg 1 => 2b @ network header + 8 csum_type 1 csum_off 10 csum_flags 0x0 ] > > While doing so, we did fail to shift the imm value, i.e. > we clear the wrong half of the u16 (protocol) instead of csum. > > The correct mask is 0xff00, and $new_value needs to be shifted > so we leave the protocol value (which is next to ttl) alone. > > Fixes: f9069cefdf ("netlink: make checksum fixup work with odd-sized header fields") > Signed-off-by: Florian Westphal <fw@strlen.de> Acked-by: Pablo Neira Ayuso <pablo@netfilter.org> -- To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/src/evaluate.c b/src/evaluate.c index 0fad0913488d..f52a0843a0c0 100644 --- a/src/evaluate.c +++ b/src/evaluate.c @@ -1869,6 +1869,20 @@ static int stmt_evaluate_payload(struct eval_ctx *ctx, struct stmt *stmt) shift_imm = expr_offset_shift(payload, payload->payload.offset, &extra_len); + payload_byte_size = round_up(payload->len, BITS_PER_BYTE) / BITS_PER_BYTE; + payload_byte_size += (extra_len / BITS_PER_BYTE); + + if (need_csum && payload_byte_size & 1) { + payload_byte_size++; + + if (payload_byte_offset & 1) { /* prefer 16bit aligned fetch */ + payload_byte_offset--; + assert(payload->payload.offset >= BITS_PER_BYTE); + } else { + shift_imm += BITS_PER_BYTE; + } + } + if (shift_imm) { struct expr *off; @@ -1885,17 +1899,6 @@ static int stmt_evaluate_payload(struct eval_ctx *ctx, struct stmt *stmt) stmt->payload.val = binop; } - payload_byte_size = round_up(payload->len, BITS_PER_BYTE) / BITS_PER_BYTE; - payload_byte_size += (extra_len / BITS_PER_BYTE); - - if (need_csum && payload_byte_size & 1) { - payload_byte_size++; - - if (payload_byte_offset & 1) { /* prefer 16bit aligned fetch */ - payload_byte_offset--; - assert(payload->payload.offset >= BITS_PER_BYTE); - } - } masklen = payload_byte_size * BITS_PER_BYTE; mpz_init_bitmask(ff, masklen);
nft add rule .. ip ttl set 64 erronously mangles ip protocol instead of ttl. Because the kernel can't deal with odd-sized data (ttl is one byte) when doing checksum fixups, so the write to 'ttl' is turned into [ payload load 2b @ network header + 8 => reg 1 ] [ bitwise reg 1 = (reg=1 & 0x000000ff ) ^ $new_value ] [ payload write reg 1 => 2b @ network header + 8 csum_type 1 csum_off 10 csum_flags 0x0 ] While doing so, we did fail to shift the imm value, i.e. we clear the wrong half of the u16 (protocol) instead of csum. The correct mask is 0xff00, and $new_value needs to be shifted so we leave the protocol value (which is next to ttl) alone. Fixes: f9069cefdf ("netlink: make checksum fixup work with odd-sized header fields") Signed-off-by: Florian Westphal <fw@strlen.de> --- src/evaluate.c | 25 ++++++++++++++----------- 1 file changed, 14 insertions(+), 11 deletions(-)