Let me apologize in advance for the length of this message - it is long!
I'm trying to test out IPv6 multicast forwarding on a 2.6.26 kernel
and I'm getting some strange values for the upcall messages from the
kernel. My code is below, but to give an overview, my setup is as
follows:
sender ------ ff15::1 -----> [eth1] linux 2.6.26 [eth0] ------> ...
So my incoming interface is eth1 and my outgoing interface is eth0 and
I am sending traffic to IPv6 multicast group ff15::1
I am seeing that mifs are being created successfully:
[root@localhost mcast-tools]# cat /proc/net/ip6_mr_vif
Interface BytesIn PktsIn BytesOut PktsOut Flags
1 eth1 0 0 0 0 00000
2 eth0 0 0 0 0 00000
And after I start sending traffic I can see that the kernel is
queueing the packets (as it does not yet have a route installed for
this traffic):
[root@localhost mcast-tools]# cat /proc/net/ip6_mr_cache
Group Origin Iif
Pkts Bytes Wrong Oifs
ff15:0000:0000:0000:0000:0000:0000:0001
fec0:0000:0000:0000:0000:0000:0000:0001 65535 1028 4
0
Which is where my code comes in, I am monitoring a raw ICMPV6 socket
so as to receive the kernel upcalls for this traffic. Now, I am
expecting to get a message from the kernel that "appears" to be an MLD
header but is in reality an mrt6msg, the flag being that the icmp6
type is set to 0 so as to indicate to the userland process that this
is an upcall.
However, in my code after I've detected that the raw socket is
readable and I read from it, I never really seem to get data that
looks like an mld_hdr:
...
n = read(raw_icmpv6_sock, raw_recv_buf, sizeof(raw_recv_buf));
if (n < 0) {
perror("read: ");
exit(1);
}
mh = (struct mld_hdr *)raw_recv_buf;
if (mh->mld_icmp6_hdr.icmp6_type == 0) { /* kernel upcall */
...
Except the icmp6_type is *never* 0 - instead it's an assortment of
strange values.
I fired up gdb, examined the contents of raw_recv_buf, right after
read() occurs and mh is set to point to raw_recv_buf:
(gdb) p/x raw_recv_buf[0]@n
$13 = {0x20, 0x0, 0x80, 0x1, 0x20, 0x0, ....
(gdb) p/d mh->mld_icmp6_hdr.icmp6_type
$14 = 32
and then the next packet comes in with different values..
(gdb) p/x raw_recv_buf[0]@n
$15 = {0x86, 0x15, 0x60, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
0xc2, 0x3, 0x18, 0x0, 0x95, 0x1a, 0x7, 0x0, 0x3, 0x15, 0x60, 0x1,
0x86, 0x15, 0x60, 0x1, 0xec, 0x\
ec, 0xe9, 0xe9, 0xe9, 0xe9, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0}
(gdb) p/d mh->mld_icmp6_hdr.icmp6_type
$16 = 134
Now I am examining the traffic sent in another window:
[root@localhost linux]# tcpdump -xnnevi eth1
tcpdump: listening on eth1, link-type EN10MB (Ethernet), capture size 96 bytes
17:59:50.713520 08:00:27:9c:aa:88 > 33:33:00:00:00:01, ethertype IPv6
(0x86dd), length 66: (hlim 3, next-header UDP (17) payload length: 12)
fec0::1.40825 > ff15::1.31337: [udp sum ok] UDP, length 4
0x0000: 6000 0000 000c 1103 fec0 0000 0000 0000
0x0010: 0000 0000 0000 0001 ff15 0000 0000 0000
0x0020: 0000 0000 0000 0001 9f79 7a69 000c 2354
0x0030: 6162 6364
17:59:51.725815 08:00:27:9c:aa:88 > 33:33:00:00:00:01, ethertype IPv6
(0x86dd), length 66: (hlim 3, next-header UDP (17) payload length: 12)
fec0::1.40825 > ff15::1.31337: [udp sum ok] UDP, length 4
0x0000: 6000 0000 000c 1103 fec0 0000 0000 0000
0x0010: 0000 0000 0000 0001 ff15 0000 0000 0000
0x0020: 0000 0000 0000 0001 9f79 7a69 000c 2354
0x0030: 6162 6364
So either I'm misunderstanding the upcall mechanism here (most likely)
or there's a bug... If there's any other information I can include let
me know. Below is the code I am using to monitor the raw socket - it's
not pretty code :-) just what I am using for testing right now!
Todd H
#include <stdio.h>
#include <stdlib.h>
#include <strings.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/uio.h>
#include <sys/socket.h>
#include <sys/select.h>
#include <sys/param.h>
#include <net/route.h>
#include <netinet/in.h>
#include <netinet/ip6.h>
#include <netinet/icmp6.h>
#include <linux/mroute6.h>
#include <arpa/inet.h>
#include <unistd.h>
#define IVIF 1
#define OVIF 2
static int raw_icmpv6_sock = -1;
static char raw_recv_buf[8192];
int raw_sock_init()
{
struct icmp6_filter filt;
raw_icmpv6_sock = socket(AF_INET6, SOCK_RAW, IPPROTO_ICMPV6);
if (raw_icmpv6_sock < 0) {
perror("socket: ");
return -1;
}
ICMP6_FILTER_SETBLOCKALL(&filt);
if (setsockopt(raw_icmpv6_sock, IPPROTO_ICMPV6, ICMP6_FILTER, &filt,
sizeof(filt)) < 0) {
perror("setsockopt ICMP6_FILTER");
return -1;
}
return 0;
}
int mrt6_init()
{
int v;
if (raw_sock_init() < 0)
return -1;
v = 1;
if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_INIT,
(void *)&v, sizeof(int)) < 0) {
perror("setsockopt: ");
return -1;
}
return 0;
}
int mrt6_done()
{
int v;
v = 1;
if (setsockopt(raw_icmpv6_sock, IPPROTO_IP, MRT6_DONE,
(void *)&v, sizeof(int)) < 0) {
perror("setsockopt: ");
return -1;
}
return 0;
}
int mrt6_add_mfc(struct sockaddr_in6 src, struct sockaddr_in6 grp)
{
struct mf6cctl mc;
bzero(&mc, sizeof(mc));
mc.mf6cc_origin = src;
mc.mf6cc_mcastgrp = grp;
mc.mf6cc_parent = IVIF;
IF_ZERO(&mc.mf6cc_ifset);
IF_SET(OVIF, &mc.mf6cc_ifset);
if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_ADD_MFC,
(void *)&mc, sizeof(mc))) {
perror("setsockopt: ");
return -1;
}
return 0;
}
int mrt6_add_mif()
{
struct mif6ctl mc;
bzero(&mc, sizeof(mc));
mc.mif6c_mifi = IVIF;
mc.mif6c_pifi = 3;
if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_ADD_MIF,
(void *)&mc, sizeof(mc))) {
perror("setsockopt: ");
return -1;
}
bzero(&mc, sizeof(mc));
mc.mif6c_mifi = OVIF;
mc.mif6c_pifi = 2;
if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_ADD_MIF,
(void *)&mc, sizeof(mc))) {
perror("setsockopt: ");
return -1;
}
return 0;
}
int main()
{
if (mrt6_init() < 0)
exit(1);
if (mrt6_add_mif() < 0)
exit(1);
/* wait for upcall */
while (1) {
int n = 0;
fd_set rset;
struct mld_hdr * mh = NULL;
struct mrt6msg * upcall = NULL;
FD_ZERO(&rset);
FD_SET(raw_icmpv6_sock, &rset);
n = select(raw_icmpv6_sock + 1, &rset, NULL, NULL, NULL);
if (n < 0) {
perror("select: ");
exit(1);
}
n = read(raw_icmpv6_sock, raw_recv_buf, sizeof(raw_recv_buf));
if (n < 0) {
perror("read: ");
exit(1);
}
mh = (struct mld_hdr *)raw_recv_buf;
if (mh->mld_icmp6_hdr.icmp6_type == 0) { /* kernel upcall */
upcall = (struct mrt6msg *)raw_recv_buf;
if (upcall->im6_msgtype == MRT6MSG_NOCACHE) {
char sbuf[64], dbuf[64];
struct sockaddr_in6 src, grp;
fprintf(stderr, "(%s,%s)\n",
inet_ntop(AF_INET6, &upcall->im6_src, sbuf, sizeof(sbuf)),
inet_ntop(AF_INET6, &upcall->im6_dst, dbuf, sizeof(dbuf)));
bzero(&src, sizeof(src));
bzero(&grp, sizeof(grp));
src.sin6_addr = upcall->im6_src;
grp.sin6_addr = upcall->im6_dst;
mrt6_add_mfc(src, grp);
}
}
}
if (mrt6_done() < 0)
exit(1);
exit(0);
}
El Thu, 23 Oct 2008 18:11:50 -0400
"Todd Hayton" <[email protected]> escribió:
[ CCing [email protected] ]
> Let me apologize in advance for the length of this message - it is long!
>
> I'm trying to test out IPv6 multicast forwarding on a 2.6.26 kernel
> and I'm getting some strange values for the upcall messages from the
> kernel. My code is below, but to give an overview, my setup is as
> follows:
>
> sender ------ ff15::1 -----> [eth1] linux 2.6.26 [eth0] ------> ...
>
> So my incoming interface is eth1 and my outgoing interface is eth0 and
> I am sending traffic to IPv6 multicast group ff15::1
>
> I am seeing that mifs are being created successfully:
>
> [root@localhost mcast-tools]# cat /proc/net/ip6_mr_vif
> Interface BytesIn PktsIn BytesOut PktsOut Flags
> 1 eth1 0 0 0 0 00000
> 2 eth0 0 0 0 0 00000
>
> And after I start sending traffic I can see that the kernel is
> queueing the packets (as it does not yet have a route installed for
> this traffic):
>
> [root@localhost mcast-tools]# cat /proc/net/ip6_mr_cache
> Group Origin Iif
> Pkts Bytes Wrong Oifs
> ff15:0000:0000:0000:0000:0000:0000:0001
> fec0:0000:0000:0000:0000:0000:0000:0001 65535 1028 4
> 0
>
> Which is where my code comes in, I am monitoring a raw ICMPV6 socket
> so as to receive the kernel upcalls for this traffic. Now, I am
> expecting to get a message from the kernel that "appears" to be an MLD
> header but is in reality an mrt6msg, the flag being that the icmp6
> type is set to 0 so as to indicate to the userland process that this
> is an upcall.
>
> However, in my code after I've detected that the raw socket is
> readable and I read from it, I never really seem to get data that
> looks like an mld_hdr:
>
> ...
> n = read(raw_icmpv6_sock, raw_recv_buf, sizeof(raw_recv_buf));
> if (n < 0) {
> perror("read: ");
> exit(1);
> }
> mh = (struct mld_hdr *)raw_recv_buf;
> if (mh->mld_icmp6_hdr.icmp6_type == 0) { /* kernel upcall */
> ...
>
> Except the icmp6_type is *never* 0 - instead it's an assortment of
> strange values.
>
> I fired up gdb, examined the contents of raw_recv_buf, right after
> read() occurs and mh is set to point to raw_recv_buf:
>
> (gdb) p/x raw_recv_buf[0]@n
> $13 = {0x20, 0x0, 0x80, 0x1, 0x20, 0x0, ....
> (gdb) p/d mh->mld_icmp6_hdr.icmp6_type
> $14 = 32
>
> and then the next packet comes in with different values..
>
> (gdb) p/x raw_recv_buf[0]@n
> $15 = {0x86, 0x15, 0x60, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
> 0xc2, 0x3, 0x18, 0x0, 0x95, 0x1a, 0x7, 0x0, 0x3, 0x15, 0x60, 0x1,
> 0x86, 0x15, 0x60, 0x1, 0xec, 0x\
> ec, 0xe9, 0xe9, 0xe9, 0xe9, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0}
> (gdb) p/d mh->mld_icmp6_hdr.icmp6_type
> $16 = 134
>
> Now I am examining the traffic sent in another window:
>
> [root@localhost linux]# tcpdump -xnnevi eth1
> tcpdump: listening on eth1, link-type EN10MB (Ethernet), capture size 96 bytes
> 17:59:50.713520 08:00:27:9c:aa:88 > 33:33:00:00:00:01, ethertype IPv6
> (0x86dd), length 66: (hlim 3, next-header UDP (17) payload length: 12)
> fec0::1.40825 > ff15::1.31337: [udp sum ok] UDP, length 4
> 0x0000: 6000 0000 000c 1103 fec0 0000 0000 0000
> 0x0010: 0000 0000 0000 0001 ff15 0000 0000 0000
> 0x0020: 0000 0000 0000 0001 9f79 7a69 000c 2354
> 0x0030: 6162 6364
> 17:59:51.725815 08:00:27:9c:aa:88 > 33:33:00:00:00:01, ethertype IPv6
> (0x86dd), length 66: (hlim 3, next-header UDP (17) payload length: 12)
> fec0::1.40825 > ff15::1.31337: [udp sum ok] UDP, length 4
> 0x0000: 6000 0000 000c 1103 fec0 0000 0000 0000
> 0x0010: 0000 0000 0000 0001 ff15 0000 0000 0000
> 0x0020: 0000 0000 0000 0001 9f79 7a69 000c 2354
> 0x0030: 6162 6364
>
> So either I'm misunderstanding the upcall mechanism here (most likely)
> or there's a bug... If there's any other information I can include let
> me know. Below is the code I am using to monitor the raw socket - it's
> not pretty code :-) just what I am using for testing right now!
>
> Todd H
>
> #include <stdio.h>
> #include <stdlib.h>
> #include <strings.h>
> #include <errno.h>
>
> #include <sys/types.h>
> #include <sys/uio.h>
> #include <sys/socket.h>
> #include <sys/select.h>
> #include <sys/param.h>
>
> #include <net/route.h>
>
> #include <netinet/in.h>
> #include <netinet/ip6.h>
> #include <netinet/icmp6.h>
> #include <linux/mroute6.h>
>
> #include <arpa/inet.h>
>
> #include <unistd.h>
>
> #define IVIF 1
> #define OVIF 2
>
> static int raw_icmpv6_sock = -1;
> static char raw_recv_buf[8192];
>
> int raw_sock_init()
> {
> struct icmp6_filter filt;
>
> raw_icmpv6_sock = socket(AF_INET6, SOCK_RAW, IPPROTO_ICMPV6);
> if (raw_icmpv6_sock < 0) {
> perror("socket: ");
> return -1;
> }
>
> ICMP6_FILTER_SETBLOCKALL(&filt);
> if (setsockopt(raw_icmpv6_sock, IPPROTO_ICMPV6, ICMP6_FILTER, &filt,
> sizeof(filt)) < 0) {
> perror("setsockopt ICMP6_FILTER");
> return -1;
> }
>
> return 0;
> }
>
> int mrt6_init()
> {
> int v;
>
> if (raw_sock_init() < 0)
> return -1;
>
> v = 1;
> if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_INIT,
> (void *)&v, sizeof(int)) < 0) {
> perror("setsockopt: ");
> return -1;
> }
> return 0;
> }
>
> int mrt6_done()
> {
> int v;
> v = 1;
> if (setsockopt(raw_icmpv6_sock, IPPROTO_IP, MRT6_DONE,
> (void *)&v, sizeof(int)) < 0) {
> perror("setsockopt: ");
> return -1;
> }
> return 0;
> }
>
> int mrt6_add_mfc(struct sockaddr_in6 src, struct sockaddr_in6 grp)
> {
> struct mf6cctl mc;
>
> bzero(&mc, sizeof(mc));
> mc.mf6cc_origin = src;
> mc.mf6cc_mcastgrp = grp;
> mc.mf6cc_parent = IVIF;
>
> IF_ZERO(&mc.mf6cc_ifset);
> IF_SET(OVIF, &mc.mf6cc_ifset);
>
> if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_ADD_MFC,
> (void *)&mc, sizeof(mc))) {
> perror("setsockopt: ");
> return -1;
> }
>
> return 0;
> }
>
> int mrt6_add_mif()
> {
> struct mif6ctl mc;
>
> bzero(&mc, sizeof(mc));
>
> mc.mif6c_mifi = IVIF;
> mc.mif6c_pifi = 3;
>
> if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_ADD_MIF,
> (void *)&mc, sizeof(mc))) {
> perror("setsockopt: ");
> return -1;
> }
>
> bzero(&mc, sizeof(mc));
>
> mc.mif6c_mifi = OVIF;
> mc.mif6c_pifi = 2;
>
> if (setsockopt(raw_icmpv6_sock, IPPROTO_IPV6, MRT6_ADD_MIF,
> (void *)&mc, sizeof(mc))) {
> perror("setsockopt: ");
> return -1;
> }
> return 0;
> }
>
> int main()
> {
> if (mrt6_init() < 0)
> exit(1);
>
> if (mrt6_add_mif() < 0)
> exit(1);
>
> /* wait for upcall */
> while (1) {
> int n = 0;
> fd_set rset;
> struct mld_hdr * mh = NULL;
> struct mrt6msg * upcall = NULL;
>
> FD_ZERO(&rset);
> FD_SET(raw_icmpv6_sock, &rset);
>
> n = select(raw_icmpv6_sock + 1, &rset, NULL, NULL, NULL);
> if (n < 0) {
> perror("select: ");
> exit(1);
> }
>
> n = read(raw_icmpv6_sock, raw_recv_buf, sizeof(raw_recv_buf));
> if (n < 0) {
> perror("read: ");
> exit(1);
> }
> mh = (struct mld_hdr *)raw_recv_buf;
> if (mh->mld_icmp6_hdr.icmp6_type == 0) { /* kernel upcall */
> upcall = (struct mrt6msg *)raw_recv_buf;
> if (upcall->im6_msgtype == MRT6MSG_NOCACHE) {
> char sbuf[64], dbuf[64];
> struct sockaddr_in6 src, grp;
> fprintf(stderr, "(%s,%s)\n",
> inet_ntop(AF_INET6, &upcall->im6_src, sbuf, sizeof(sbuf)),
> inet_ntop(AF_INET6, &upcall->im6_dst, dbuf, sizeof(dbuf)));
>
> bzero(&src, sizeof(src));
> bzero(&grp, sizeof(grp));
>
> src.sin6_addr = upcall->im6_src;
> grp.sin6_addr = upcall->im6_dst;
>
> mrt6_add_mfc(src, grp);
> }
> }
>
> }
> if (mrt6_done() < 0)
> exit(1);
>
> exit(0);
> }
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
On Fri, 24 Oct 2008, Alejandro Riveira Fernández wrote:
>> Let me apologize in advance for the length of this message - it is long!
>>
>> I'm trying to test out IPv6 multicast forwarding on a 2.6.26 kernel
>> and I'm getting some strange values for the upcall messages from the
>> kernel. My code is below, but to give an overview, my setup is as
>> follows:
>>
>> sender ------ ff15::1 -----> [eth1] linux 2.6.26 [eth0] ------> ...
Maybe this isn't the bug you're looking for but you shouldn't be using
ff1x multicast addresses in a test like this; ff1x means that the
multicast group is of "interface-local scope" and it isn't useful for
multicast forwarding. So the kernel might be correct in not
installing multicast forwarding state for a group address like this
(but if it's a conscious decision, maybe the failure mode should be
better). See S 2.7 of RFC4291.
--
Pekka Savola "You each name yourselves king, yet the
Netcore Oy kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings
On Fri, Oct 24, 2008 at 6:12 AM, Pekka Savola <[email protected]> wrote:
> On Fri, 24 Oct 2008, Alejandro Riveira Fern?ndez wrote:
>>>
>>> Let me apologize in advance for the length of this message - it is long!
>>>
>>> I'm trying to test out IPv6 multicast forwarding on a 2.6.26 kernel
>>> and I'm getting some strange values for the upcall messages from the
>>> kernel. My code is below, but to give an overview, my setup is as
>>> follows:
>>>
>>> sender ------ ff15::1 -----> [eth1] linux 2.6.26 [eth0] ------> ...
>
> Maybe this isn't the bug you're looking for but you shouldn't be using ff1x
> multicast addresses in a test like this; ff1x means that the multicast group
> is of "interface-local scope" and it isn't useful for multicast forwarding.
> So the kernel might be correct in not installing multicast forwarding state
> for a group address like this (but if it's a conscious decision, maybe the
> failure mode should be better). See S 2.7 of RFC4291.
>
Hey there, thanks for the response! I may be misinterpreting the RFC
but I thought ff15::1 was a site-local address. As in I break down
ff15::1 as follows -
0xff : identifies the address as multicast
0x15 : 4-bits flags and 4-bits scope where
flags = 0x1 meaning that the T bit is set as this address is not a
permanently assigned address
scope = 0x5 indicating site-local scope
Todd H
> --
> Pekka Savola "You each name yourselves king, yet the
> Netcore Oy kingdom bleeds."
> Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings
On Fri, 24 Oct 2008, Todd Hayton wrote:
> Hey there, thanks for the response! I may be misinterpreting the RFC
> but I thought ff15::1 was a site-local address. As in I break down
> ff15::1 as follows -
>
> 0xff : identifies the address as multicast
> 0x15 : 4-bits flags and 4-bits scope where
> flags = 0x1 meaning that the T bit is set as this address is not a
> permanently assigned address
> scope = 0x5 indicating site-local scope
You're right. The nibbles got switched in my head. Sorry for that.