Wenpeng Song authored 2 years ago and Yalu Zhang committed 2 years ago

A separate package asterisk-chan-voicemngr has been added to replace chan_brcm.

cherry picked from commit 32337872

- EARLY_OFFHOOK
- EARLY_ONHOOK
- MEDIA

Wenpeng Song authored 2 years ago and Yalu Zhang committed 2 years ago

The commit c6a0ad61 changes sub->parent->context during an outgoing
call that causes future outgoing calls fail.

Wenpeng Song authored 2 years ago and Yalu Zhang committed 2 years ago

In function brcm_start_calling(), the type of sub->parent is struct brcm_pvt*. But it was wrongly
casted to struct ast_channel* so the contents of sub->parent were messed up. 

Hemlata authored 2 years ago and Yalu Zhang committed 2 years ago

Before this commit, a normal dial tone is played if mwi_dialtone_state in "pjsip show endpoint" is
not "off", "congestion" or "special".
After this commit, in the above condition, a stutter dial tone is played if mwi_enable is 1, otherwise
a normal dial tone is played.

Hemlata authored 2 years ago and Yalu Zhang committed 2 years ago

- Include the header in requests and responses as per the RFC
- Play early media if the configuration allows

Grzegorz Sluja authored 2 years ago and Yalu Zhang committed 2 years ago

An incoming SIP request will be silently dropped if the following conditions are all met:
- The request is about to create a new session
- The source IP address of the request is not the address of the configured proxy

Note
- The source port of the request is not checked since we have not found the good way to
  retrieve the port that is used for transmission
- If the configured proxy is a domain name other than IP address, the resolved IP address
  might be different from the one which is being used in existing sessions

(cherry picked from commit c50716b2)

(cherry picked from commit 7c71ebce)

Grzegorz Sluja authored 2 years ago and Yalu Zhang committed 2 years ago

In brcm_request() there were posibilities that we called ast_channel_name(tmp)
if tmp is NULL, it leads to crash. Also fixed other deadlock possibilities where
mutex would not be unlocked in some cases.

For the case if there is more than one contact header in 200 OK to REGISTER and the last Contact has
expires=0, asterisk wrongly set next registration period to default value DEFAULTEXPIRY.

The correct behaviour is to pick the expire value in the Contact which matches the Contact in the
original REGISTER.

(cherry picked from commit 39421f0a)

- R0. If more than one account is registered, and there has another incoming call(ringing) to other accounts,
  it will be rejected as well when the ongoing one doing R0. (A, B registered on DUT; A<=>C(ongoing call);
  D=>B(ringing); E=>A(ringing,cw); A press R0 then both D and E be rejected)
- 3-way conference back to 2-way call, then performing R1 will lead to a crash

"uk"   - the default config which support flash-hook (R) only triggering call waiting and 3-way conference.
         R4 and R5 are for attended and unattended call transfer respectively with a timer.
"etsi" - Using R0, R1, R2, to trigger different ways of handling call waiting. R3 for 3-way conference.
         R4 and R5 are for attended and unattended call transfer respectively without a timer.

- The issue about TELCHAN stuck after call hold/unhold
- Asterisk crashes when call transfer is provided in an invalid scenario: Transferor calls to Transferee,
  then Transferor calls to Transfer Target, i.e. trying to let the caller do the transfer.

The service is triggered by R5, i.e. flash hook + 5.

Sequence numbers received in RTP packet need to be forwarded to brcm endpoint since based on this parameter
some of RTP statistics are calculated. It was wrong to use the locally generated sequence numbers.

Grzegorz Sluja authored 2 years ago and Yalu Zhang committed 2 years ago

SIPSessionID is from Session-ID header field either in INVITE for incoming calls or in
200 OK for outgoing calls.