Hi Sylvain,
if you are using ISC DHCP try to recompile it to use sockets and not LPF :
+++ dhcp-3.0pl2-enri/includes/cf/linux.h Fri Jan 17 22:02:28 2003
@@ -122,7 +122,7 @@
#if (LINUX_MAJOR >= 2)
# if (LINUX_MINOR >= 1)
# if defined (USE_DEFAULT_NETWORK)
-# define USE_LPF
+# undef USE_LPF
# endif
# if !defined (__sparc__) /* XXX hopefully this will be fixed
someday */
# define SIOCGIFCONF_ZERO_PROBE
Sylvain Beucler wrote:
> Hi,
>
> On Thu, Sep 14, 2006 at 05:17:28PM +0200, Peter Stuge wrote:
>
>> On Thu, Sep 14, 2006 at 12:25:43PM +0200, Sylvain Beucler wrote:
>>
>>> Shouldn't it work?
>>>
>> I think it should.
>>
>> I would check what is actually on the wire, to confirm that nothing
>> is really leaving the box. (Use a hub.)
>>
>> I would also investigate what dhclient actually does by running it
>> under strace.
>>
>> I would also try other DHCP clients. Personally I like dhcpcd but
>> there's also pump and udhcpc.
>>
>
> Thanks for the advice.
>
>
> It turns out that something goes out of the wire, the DHCP server
> machine receives it, yet does not reply (I doubt the DHCP server
> daemon itself sees it). Or, the reply does not show on tcpdump.
>
>
> This is strange really: tcpdump -vv shows that the initial broadcast
> paquet is nearly identical for eth0 and eth0.5. Only a 'xid' field
> differs, which is one-char shorter on the vlan:
> * pump -i eth0.5
> 22:14:24.039510 IP (tos 0x0, ttl 64, id 0, offset 0, flags [none], length: 576) 0.0.0.0.bootpc > 255.255.255.255.bootps: BOOTP/DHCP, Request from 00:40:ca:d5:1c:57, length: 548, xid:0xb4c6a13, flags: [none] (0x0000)
> Client Ethernet Address: 00:40:ca:d5:1c:57 [|bootp]
> [nothing]
>
> * pump -i eth0
> 22:14:26.495647 IP (tos 0x0, ttl 64, id 0, offset 0, flags [none], length: 576) 0.0.0.0.bootpc > 255.255.255.255.bootps: BOOTP/DHCP, Request from 00:40:ca:d5:1c:57, length: 548, xid:0xd54c6b50, flags: [none] (0x0000)
> Client Ethernet Address: 00:40:ca:d5:1c:57 [|bootp]
> [DHCP offer/request follows]
>
>
> The behavior is the same with all DHCP clients I tries (dhclient,
> dhclient3, pump, dhcpcd). I also tried dhcp3-server and udhcpd (and my
> ISP's box's) :/
>
>
> As for the strace, dhclient3 times out on a select:
>
> time([1158267744]) = 1158267744
> time([1158267744]) = 1158267744
> rt_sigaction(SIGPIPE, {0xa7ea57e0, [], 0}, {SIG_DFL}, 8) = 0
> send(3, "<30>Sep 14 23:02:24 dhclient: DH"..., 91, 0) = 91
> rt_sigaction(SIGPIPE, {SIG_DFL}, NULL, 8) = 0
> write(2, "DHCPDISCOVER on eth0.5 to 255.25"..., 61DHCPDISCOVER on eth0.5 to 255.255.255.255 port 67 interval 10) = 61
> write(2, "\n", 1
> ) = 1
> sendto(8, "\377\377\377\377\377\377\0@\312\325\34W\10\0E\20\1H\0\0"..., 342, 0, {sa_family=AF_PACKET, proto=0x6574, if892219496, pkttype=PACKET_HOST, addr(0)={0, }, 16) = 342
> select(9, [7 8], [], [], {10, 0}) = 0 (Timeout)
>
> time([1158267754]) = 1158267754
> time([1158267754]) = 1158267754
> [...]
>
>
> A (newbie) question though: theoretically how could the eth0.5 grab a
> DHCP offer? Doesn't 'eth0' steal it first?
>
> Thanks,
>
>
