obexapp get failure

[Maksim Yevmenkin]

On Mon, Oct 25, 2010 at 4:39 PM, Maksim Yevmenkin
<maksim.yevmenkin at gmail.com> wrote:
> On Mon, Oct 25, 2010 at 2:38 PM, Iain Hibbert <plunky at rya-online.net> wrote:
>> On Mon, 25 Oct 2010, Maksim Yevmenkin wrote:
>>
>>> not really. the way i read the spec, connection id header is not
>>> required on 'continue GET' however, wm6 server might require it (for
>>> some reason). i'd like to see PUT dump, i.e. what does wm6 server
>>> sends back in 'continue' when obexapp does multi-PUT
>>
>> Well, that works fine naturally and shows nothing interesting alas. What
>> is more revealing is if I use wm6 to issue a multi-GET from obexapp server
>> (which does work fine)
>>
>>> ACL data: handle 13 flags 0x02 dlen 42
>>    L2CAP(d): cid 0x0044 len 38 [psm 3]
>>      RFCOMM(d): UIH: cr 1 dlci 20 pf 1 ilen 33 fcs 0x2d credits 1
>>        OBEX: Get cmd(f): len 33
>>        Connection ID (0xcb) = 1
>>        Name (0x01) = Unicode length 22
>>        0000: 00 74 00 65 00 73 00 74  00 2e 00 6c 00 61 00 72  .t.e.s.t...l.a.r
>>        0010: 00 67 00 65 00 00                                 .g.e..
>>
>> < ACL data: handle 13 flags 0x02 dlen 58
>>    L2CAP(d): cid 0x00b8 len 54 [psm 3]
>>      RFCOMM(d): UIH: cr 0 dlci 20 pf 0 ilen 50 fcs 0xeb
>>        OBEX: Get rsp(f): status 100 len 4096
>>        Status 100 = Continue
>>        Length (0xc3) = 65529
>>        Body (0x48) = Sequence length 4085
>>        0000: 61 61 61 61 61 61 61 61  61 61 61 61 61 61 61 61  aaaaaaaaaaaaaaaa
>>        0010: 61 61 61 61 61 61 61 61  61 61 61 61 61 61 61 61  aaaaaaaaaaaaaaaa
>>        [...]
>>        0fe0: 61 61 61 61 61 61 61 61  61 61 61 61 61 61 61 61  aaaaaaaaaaaaaaaa
>>        0ff0: 61 61 61 61 61                                    aaaaa
>>
>> so the Get command starts and we return the first segment with status 100
>> (Continued)
>>
>>> ACL data: handle 13 flags 0x02 dlen 17
>>    L2CAP(d): cid 0x0044 len 13 [psm 3]
>>      RFCOMM(d): UIH: cr 1 dlci 20 pf 1 ilen 8 fcs 0x2d credits 2
>>        OBEX: Get cmd(f): len 8 (continue)
>>        Connection ID (0xcb) = 1
>>
>> < ACL data: handle 13 flags 0x02 dlen 58
>>    L2CAP(d): cid 0x00b8 len 54 [psm 3]
>>      RFCOMM(d): UIH: cr 0 dlci 20 pf 0 ilen 50 fcs 0xeb
>>        OBEX: Get rsp(f): status 100 len 4096
>>        Status 100 = Continue
>>        Body (0x48) = Sequence length 4090
>>        0000: 61 61 61 61 61 61 61 61  0a 61 61 61 61 61 61 61  aaaaaaaa.aaaaaaa
>>        0010: 61 61 61 61 61 61 61 61  61 61 61 61 61 61 61 61  aaaaaaaaaaaaaaaa
>>        [...]
>>        0fe0: 61 61 61 61 61 61 61 61  61 61 61 61 61 61 61 61  aaaaaaaaaaaaaaaa
>>        0ff0: 61 61 61 61 61 61 61 61  61 61                    aaaaaaaaaa
>>
>> and wm6 phone issues a Get continuation cmd as it should, but it does
>> always put the Connection ID in the packet. [and this repeats quite a few
>> times]
>
> [...]
>
> i see
>
>> until the final segment where we return 200 as normal
>>
>> So, I think there is a good chance that the phone is losing because there
>> is no Connection ID included..  it seems to be allowed (but not required)
>> by the spec to include such a thing but I find openobex documentation (!)
>> to be more or less incomprehensible and I don't even see how that packet
>> is being generated..
>
> i'm pretty sure that is what it is.  can you please try the patch
> below. this is completely untested :) no idea what it will do.
>
> beetle% cvs diff -u
> cvs diff: Diffing .
> Index: stream.c
> ===================================================================
> RCS file: /usr/local/cvs/ports/obexapp/stream.c,v
> retrieving revision 1.10
> diff -u -r1.10 stream.c
> --- stream.c    22 Oct 2010 17:04:11 -0000      1.10
> +++ stream.c    25 Oct 2010 23:36:11 -0000
> @@ -230,6 +230,12 @@
>
>        context->stotal += length;
>        log_debug("%s(): Wrote %d bytes of stream data", __func__, length);
> +
> +       if (context->connection_id != OBEXAPP_INVALID_CONNECTION_ID) {
> +               hv.bq4 = context->connection_id;
> +               OBEX_ObjectAddHeader(handle, object,
> +                       OBEX_HDR_CONNECTION, hv, sizeof(hv.bq4), 0);
> +       }
>  } /* obexapp_stream_read */
>
>  void
>
> ==
>
> in any case, openobex library generates continue packets.

i'm convinced that there is a bug in obexapp (or, less likely,
openobex library). it appears that "connection id" header is intended
for multiplexing of several data streams. its pretty much analog of
http "host" header. i think its clear what "connection id" header must
be present on all requests from the client, including continue-GET
request. what is not entirely clear if the server responses should
also contain "connection id" header. i think they should as well.
clearly, wm6 server is not sending "connection id" header back in its
responses, so perhaps its a double whammy.

thanks,
max