dvbapi unable to switch to another reader when one is falling down

[knasson]

Revision

Version: 1.10rc-svn Rev. 6089

Issue Description

For a channel available on 2 readers, if dvbapi is requesting ecm on reader 1 ,if reader 1 falls down (reader in timeout )reader 2 is not requested and you get a freeze.

When the issue occurs

this issue appears when the reader selected is disabled (or in timeout).

How the issue is reproducable

To reproduce that just choose a channel available on both readers and disable it.You will see a black screen.

<Don't forget to ATTACH (NOT post here as text) a log file of oscam in debug mode (start oscam with -d255) and to remove the brackets around the fields! Otherwise, your ticket will be closed as you seem to be unable to read.>

[knasson]

this switching of dvbapi works if you use request_mode=1 but there is several unconvenients to do that:
priority of oscam.dvbapi no more taken into account
freeze of dvbapi when zapping if the channel requested is longer to be obtained
lb_mode no more working correctly.

[BurnMasterRecords]

Loadbalancing only works for Same CAID:Provider & not mixed Cards.

[Bit]

I asked corsair already about this problem.
It's not a loadbalancer problem, but a dvbapi problem.
If one reader is not available anymore, the dvbapi should switch to another reader which should take in account all possible CAID:IDENTs for the actual channel. This does not happen actually.

[knasson]

test made with svn 6551:

The switch is freeze free but dvbapi looks for systematicaly the same CAID:ident as the one which has been lost.

exemple: if reader 1 is answering with 0500:032830 when it is disabled dvbapi looks immediatly for the same CAID:ident on an other reader.

--> oscam.dvbapi is no longer taken into account
--> loadbalancing seems to work but only between the readers which own the answers

[JrCs]

Same problem report in ticket #2479

[pooyair]

could confirm it.

[lareq78]

please try with >= r7051.

[puya]

r7108

[pooyair]

fixed in r7123

[knasson]

Try this 7123 version ,but nothing changed (request_mode=0)

1) i loose the oscam.dvbapi priority
2) the dvbapi does not change of reader when one is blocked

I added reopenonzap=1 but no effect...

@+
knasson

[knasson]

Replying to JrCs:

Same problem report in ticket #2479

it seems a different problem, in my case dvbapi keeps the same CAID:ident to look for a new DW.

The true question is how to make dvbapi changing of CAID:IDENT when the one used is no longer available...

[BurnMasterRecords]

Can Confirm this Bug.
After one CAID is gone / blocked OSCam don't switch to another CAID.

[hk2k2]

confirmed too.

also in request mode = 1 , if a ecm got timeout, no other caid is requested. :/

[avs1978]

Does this problem only occur for the DVBAPI user or any user?

[avs1978]

Could the problem be in reader-common.c?

If so, I suspect the functions reader_card_inserted and/or reader_checkhealth.

But it could also be reader_ecm in reader-common.c or reader_get_ecm in oscam-reader.c.

Does this help?

[zywo]

Is this issue solved in r7638 with the new feature decodeforever = 1 ?