1221837 - sdp without rtpmap for static codec IDs => "Failed to negotiate codec details for all codecs" error.

Status: RESOLVED FIXED

(Core :: WebRTC: Signaling, defect, P2)

Description

[gled]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:42.0) Gecko/20100101 Firefox/42.0
Build ID: 20151029151421

Steps to reproduce:

Install Firefox 42
Browse to a demo service using sip.js
Call a webrtc enabled echo test ( freeswitch for example, with the verbose_sdp param to false, ie default config ).


Actual results:

Firefox rejects the calls and we can see in the logs:

17:33:49.803 Wed Nov 04 2015 17:33:49 GMT-0800 (PST) | sip.transport | received WebSocket text message:

SIP/2.0 200 OK
(...)

v=0
o=#22565356 2232 3510 IN IP4 54.80.250.244
s=Talk
c=IN IP4 54.80.250.244
b=AS:380
t=0 0
m=audio 51618 RTP/SAVPF 0
a=sendrecv
a=rtcp:51618
a=rtcp-mux
a=setup:active
a=fingerprint:sha-1 5D:57:49:A1:29:75:8C:C7:CA:83:4D:34:4B:D9:5E:C8:7A:5F:16:C6
a=ice-ufrag:PhiviwWI
a=ice-pwd:9zIdUvg1fYKqmFl6fY9DGDJtV0
a=candidate:Ji5N9SqYh58Awk9I 1 UDP 2130706431 54.80.250.244 51618 typ host
m=video 51642 RTP/SAVPF 120
a=rtpmap:120 VP8/90000
a=sendrecv
a=rtcp:51642
a=rtcp-mux
a=setup:active
a=fingerprint:sha-1 5D:57:49:A1:29:75:8C:C7:CA:83:4D:34:4B:D9:5E:C8:7A:5F:16:C6
a=ice-ufrag:wURDrzRk
a=ice-pwd:57li34JqH2wU0FxEjf7cf5lJXd
a=candidate:Ji5N9SqYh58Awk9I 1 UDP 2130706431 54.80.250.244 51642 typ host

17:33:49.825 DOMException [InvalidSessionDescriptionError: "Failed to negotiate codec details for all codecs"

That ends with a SIP 488.

This is due to a missing a=rtpmap:0 PCMU/8000 in the sdp, which is a totally valid sdp.


Expected results:

Call should be accepted even if the rtpmap is not present. 0 is a static assigned id, always PCMU.

RFC specifies that rtpmap is only mandatory for dynamic ids.

Comment 1

[Randell Jesup [:jesup] (needinfo me)]

Confirmed.

I think this may be a regression from landing the JSEP rewrite - nils?  

We should support the default mappings for codecs we understand (PCMA/PCMU/G722)

Comment 2

[Nils Ohlmeier [:drno]]

https://reviewboard-hg.mozilla.org/gecko/rev/1053ca7ea2c2

Comment 3

[Nils Ohlmeier [:drno]]

Attachment: MozReview Request: Bug 1221837: Accept hard coded codec numbers without rtpmaps. r?bwc

Bug 1221837: Accept hard coded codec numbers without rtpmaps. r?bwc

Comment 4

[Byron Campen [:bwc]]

Comment on attachment 8688030 [details]
MozReview Request: Bug 1221837: Accept hard coded codec numbers without rtpmaps. r?bwc

https://reviewboard.mozilla.org/r/25273/#review22759

::: media/webrtc/signaling/src/jsep/JsepCodecDescription.h:59
(Diff revision 1)
> +    if (!entry &&
> +        (fmt.compare("0") || fmt.compare("8") || fmt.compare("9"))) {
> +      return true;
> +    }

This is going to cause PCMU to match PCMA, and vice-versa. You probably need to do something like:

```
std::string name;
// ints for clock/channels

if (entry) {
  name = entry->name;
  // etc
} else if(fmt == "0") {
  name = "PCMU";
  // etc
} else if //etc

if (nsCRT::strcasecmp(nName.c_str(), name) &&
    // etc )
```

Comment 5

[Nils Ohlmeier [:drno]]

Comment on attachment 8688030 [details]
MozReview Request: Bug 1221837: Accept hard coded codec numbers without rtpmaps. r?bwc

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/25273/diff/1-2/

Comment 6

[Byron Campen [:bwc]]

Comment on attachment 8688030 [details]
MozReview Request: Bug 1221837: Accept hard coded codec numbers without rtpmaps. r?bwc

https://reviewboard.mozilla.org/r/25273/#review22771

::: media/webrtc/signaling/src/jsep/JsepCodecDescription.h:65
(Diff revisions 1 - 2)
> +    } else if (fmt.compare("9") && mName == "G722") {
> +      return true;
> +    } else if (fmt.compare("0") && mName == "PCMU") {
> +      return true;
> +    } else if (fmt.compare("8") && mName == "PCMA") {
> +      return true;
> +    }

Just to be complete, we probably want to check the clock rate and channels too.

Comment 7

[Nils Ohlmeier [:drno]]

https://reviewboard.mozilla.org/r/25273/#review22771

> Just to be complete, we probably want to check the clock rate and channels too.

But there is nothing to compare them against here. This covers for the case where we only got a codec number in the m-line, but no matching fmtpmap. So we could only compare rate and channels only against our own expectations, which seems rather silly...

Comment 8

[gled]

Just my 2 cents, clock rate for PCM(a|u) and G722 will only be 8khz, and mono anyway.

Well, to be correct, G722 is 16khz, but advertised as 8khz because of RFC1890.

Comment 9

[Byron Campen [:bwc]]

(In reply to Nils Ohlmeier [:drno] from comment #7)
> https://reviewboard.mozilla.org/r/25273/#review22771
> 
> > Just to be complete, we probably want to check the clock rate and channels too.
> 
> But there is nothing to compare them against here. This covers for the case
> where we only got a codec number in the m-line, but no matching fmtpmap. So
> we could only compare rate and channels only against our own expectations,
> which seems rather silly...

   Hmm, I suppose since 8khz mono is an inherent property of these codecs, we can do without.

Comment 10

[Nils Ohlmeier [:drno]]

Try run looks sane

Comment 11

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/12b1bc4e7220

Comment 13

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/12b1bc4e7220