(usagi-users 03393) Re: MLDv2 linux implementation questions

[Hoerdt Mickael <hoerdt@xxxxxxxxxxxxxxxxxxxxx>]

Hi,

Another issue, still with test_msf, doing these commands :

./test_msf ff3e::1234

test_msf > r 2001:6::2 0
For the interface index 0, you want to join the source specific group
(2001:6::2,ff3e::1234).
Join SSM group (ff3e::1234, 2001:6::2) on interface index 0 : OK.
test_msf > r 2001:6::3 0
For the interface index 0, you want to join the source specific group
(2001:6::3,ff3e::1234).
Join SSM group (ff3e::1234, 2001:6::3) on interface index 0 : OK.
test_msf > p 2001:6::2 0
For the interface index 0, you want to leave the source specific group
(2001:6::2, ff3e::1234).
join_leave_ssm setsockopt : Cannot assign requested address
Leave SSM group (ff3e::1234, 2001:6::2) on interface index 0 : problem

the kernel returns me an error where it should be ok and the filter on
the socket
should be INCLUDE(2001:6::3) for ff3e::1234

Hoerdt Mickaël

David Stevens wrote:

>Yes, the patch is incomplete. EADDRINUSE is (on other systems)
>the error returned in the regular join case (with no filters) when the
>group has already been joined. Linux didn't have that, which
>caused the socket buffer filling you saw, but the SSM code needs
>to do the join but ignore the error return. (ie, join the first time, and
>only edit the filter on subsequent calls).
>
>I'll send a revised patch after I've had some sleep. :-)
>
>                                        +-DLS
>
>  
>