The ICMPv6 Echo Request/Reply sequence number and id as well as the IPv6 header length are two byte long fields and therefore might need a conversion on a little endian system. Otherwise the output will be broken on such a machine.
Fixes: 35b37756f4a3 ("add IPv6 support to tcpdump parser") Signed-off-by: Linus Lüssing linus.luessing@c0d3.blue --- tcpdump.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/tcpdump.c b/tcpdump.c index db93681..b9edc20 100644 --- a/tcpdump.c +++ b/tcpdump.c @@ -589,13 +589,15 @@ static void dump_ipv6(unsigned char *packet_buff, ssize_t buff_len, break; case ICMP6_ECHO_REQUEST: printf(" echo request, id: %d, seq: %d, length: %hu\n", - icmphdr->icmp6_id, icmphdr->icmp6_seq, - iphdr->ip6_plen); + ntohs(icmphdr->icmp6_id), + ntohs(icmphdr->icmp6_seq), + ntohs(iphdr->ip6_plen)); break; case ICMP6_ECHO_REPLY: printf(" echo reply, id: %d, seq: %d, length: %hu\n", - icmphdr->icmp6_id, icmphdr->icmp6_seq, - iphdr->ip6_plen); + ntohs(icmphdr->icmp6_id), + ntohs(icmphdr->icmp6_seq), + ntohs(iphdr->ip6_plen)); break; case ICMP6_TIME_EXCEEDED: printf(" time exceeded in-transit, length %zu\n",
I wonder about the "_might_ need a conversion" as most cpu architectures run little endian. Especially x86.
Wouldn't it be broken since years if it needed big → little conversion?
kind regards txt.file
On 9/13/20 11:30 PM, Linus Lüssing wrote:
The ICMPv6 Echo Request/Reply sequence number and id as well as the IPv6 header length are two byte long fields and therefore might need a conversion on a little endian system. Otherwise the output will be broken on such a machine.
Fixes: 35b37756f4a3 ("add IPv6 support to tcpdump parser") Signed-off-by: Linus Lüssing linus.luessing@c0d3.blue
tcpdump.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/tcpdump.c b/tcpdump.c index db93681..b9edc20 100644 --- a/tcpdump.c +++ b/tcpdump.c @@ -589,13 +589,15 @@ static void dump_ipv6(unsigned char *packet_buff, ssize_t buff_len, break; case ICMP6_ECHO_REQUEST: printf(" echo request, id: %d, seq: %d, length: %hu\n",
icmphdr->icmp6_id, icmphdr->icmp6_seq,iphdr->ip6_plen);
ntohs(icmphdr->icmp6_id),ntohs(icmphdr->icmp6_seq),ntohs(iphdr->ip6_plen)); break;
icmphdr->icmp6_id, icmphdr->icmp6_seq,iphdr->ip6_plen);
ntohs(icmphdr->icmp6_id),ntohs(icmphdr->icmp6_seq),ntohs(iphdr->ip6_plen)); break;
On Mon, Sep 14, 2020 at 12:07:41AM +0200, txt.file@txtfile.eu wrote:
I wonder about the "_might_ need a conversion" as most cpu architectures run little endian. Especially x86.
Wouldn't it be broken since years if it needed big → little conversion?
The tcpdump output for IPv6 echo request/reply was for sure broken for years on little endian targets. Note that most old-school router platforms are big endian, hence it worked well on the common ubiquiti and tp-link devices.
kind regards txt.file
On 9/13/20 11:30 PM, Linus Lüssing wrote:
The ICMPv6 Echo Request/Reply sequence number and id as well as the IPv6 header length are two byte long fields and therefore might need a conversion on a little endian system. Otherwise the output will be broken on such a machine.
Fixes: 35b37756f4a3 ("add IPv6 support to tcpdump parser") Signed-off-by: Linus Lüssing linus.luessing@c0d3.blue
tcpdump.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/tcpdump.c b/tcpdump.c index db93681..b9edc20 100644 --- a/tcpdump.c +++ b/tcpdump.c @@ -589,13 +589,15 @@ static void dump_ipv6(unsigned char *packet_buff, ssize_t buff_len, break; case ICMP6_ECHO_REQUEST: printf(" echo request, id: %d, seq: %d, length: %hu\n",
icmphdr->icmp6_id, icmphdr->icmp6_seq,iphdr->ip6_plen);
ntohs(icmphdr->icmp6_id),ntohs(icmphdr->icmp6_seq),ntohs(iphdr->ip6_plen)); break;
icmphdr->icmp6_id, icmphdr->icmp6_seq,iphdr->ip6_plen);
ntohs(icmphdr->icmp6_id),ntohs(icmphdr->icmp6_seq),ntohs(iphdr->ip6_plen)); break;
b.a.t.m.a.n@lists.open-mesh.org
-
Daniel Golle -
Linus Lüssing -
txt.file＠txtfile.eu