[pulseaudio-discuss] incomplete IPv6 support in pa_socket_client_new_string()

[Hajime Fujita]

Tanu Kaskinen wrote:
> On Thu, 2014-11-06 at 22:26 -0600, Hajime Fujita wrote:
>> Tanu Kaskinen wrote:
>>> On Wed, 2014-11-05 at 22:29 -0600, Hajime Fujita wrote:
>>>> Hi Tanu,
>>>>
>>>> Thank you for your comment and sorry for the late reply.
>>>>
>>>> Tanu Kaskinen wrote:
>>>>> On Tue, 2014-10-21 at 00:49 -0500, Hajime Fujita wrote:
>>>>>> Hello,
>>>>>>
>>>>>> I'm currently working on IPv6 support for the raop module [1].
>>>>>>
>>>>>> During the work I found a potential issue in
>>>>>> pa_socket_client_new_string() (in pulsecore/socket-client.c) as pointed
>>>>>> out in [1].
>>>>>>
>>>>>> It does not work as expected when an IPv6 address string like
>>>>>> "2001:db8::1" is passed as the "name" parameter. This is because the
>>>>>> name parameter is passed to pa_parse_address(), which thinks the last
>>>>>> colon as a separator between hostname (or address) and a port number. To
>>>>>> prevent pa_parse_address() from doing this, an IPv6 address must be
>>>>>> bracketed with "[]" (e.g. "[2001:db8::1]").
>>>>>>
>>>>>> I'm wondering what would be the best solution for this situation.
>>>>>> a. Modify callers: callers of pa_socket_client_new_string() must add
>>>>>> brackets to IPv6 addresses.
>>>>>> b. Modify pa_socket_client_new_string(): if an IPv6 address is given as
>>>>>> the name parameter, it will be bracketed before being passed to
>>>>>> pa_parse_address().
>>>>>>
>>>>>> Any opinions or suggestions?
>>>>>
>>>>> I think it makes more sense to fix this in the callers of
>>>>> pa_socket_client_new_string(). It would be good to also add a comment in
>>>>> socket-client.h saying that the name parameter can include also the port
>>>>> using the usual syntax, and for that reason IPv6 addresses must be
>>>>> enclosed in brackets.
>>>>>
>>>>
>>>> Do you have any justification behind this?
>>>
>>> Yes, sorry for leaving it out from my answer. The justification: I think
>>> you can't actually implement suggestion b, because there's no way you
>>> can figure out in pa_socket_client_new_string() whether
>>> "1234::1234:1234" contains only an IP address or an IP address and a
>>> port.
>>
>> Well my justification for option b was:
>> if the caller wants to pass "addr:port" style string to
>> pa_socket_client_new_string(), "addr" must be surrounded by brackets for
>> the very reason you described.
>> So, if pa_socket_client_new_string() sees a string like
>> "1234::1234:1234", I thought it would be safe to interpret it as an IPv6
>> address alone, not including port.
>
> Ok, now I understand (and looking back, I feel I should have understood
> already from the first mail...). I think pa_parse_address() should be
> changed so that "1234::1234:1234" is parsed as an IP address, not IP
> address + port. That way the problem is fixed for all callers of
> pa_parse_address(), not only for pa_socket_client_new_string().
>

Thanks, this makes sense. I'll send a patch for this.


Best,
Hajime