0000496: Stream incompatible with Syabas PCH - VDR-Developer.org Bugtracking

(0000744)
schmirl
11-12-08 13:12

Problem could be that PMT must be broadcasted before actual stream data.

patpmt_first.diff is a crude hack to accomplish this. First packet will be the PAT. The patch doesn't guarantee that the PMT will be the second packet. The first PMT could even get filtered. A clean solution would send the PMT from within cStreamdevPatFilter::Process(). This however requires repacking and I don't want to do this without the ability to test if the packet is ok afterwards (I can't test ATM). Maybe someone can contribute a patch here.

If the PMT is filtered, try modifying

bool Syncing() const { return seen != (SI_PMT | SI_PAT); }

into

bool Syncing() const { return (seen & SI_PAT) == 0; }

So we only wait for the PAT to occur, PMT will not be the first packet, but it should show up earlier as without patch.

(0000747)
schmirl
11-12-08 15:34
edited on: 11-12-08 15:35

Refined patch patpmt_first_v2.diff guarantees that first packet is PAT, second is PMT. Untested!

(0000748)
schmirl
11-13-08 09:51

Previous patches didn't work. Patpmt_first_v3.diff hopefully fixes the problem (Untested!)

(0000750)
schmirl
11-14-08 14:02

patpmt_first_v4 fixes PMT generator
Minor fixes for PAT generator (reserved bits must be '1')

(0000751)
schmirl
11-18-08 08:43

Success with the following modifications on top of v4 patch:

*** Retransmit PAT (will give an other continuity error)
Two modifications in livestreamer.c: Around line 214, change

  if (Tid == 0x00 && !pmtPid) {
into
  if (Tid == 0x00) {
    int pmtPidBak = pmtPid;

And around line 275 change

  m_Streamer->SetPids(pmtPid);
  Add(pmtPid, 0x02);
  pmtVersion = -1;

into

  if (pmtPid != pmtPidBak) {
    m_Streamer->SetPids(pmtPid);
    Add(pmtPid, 0x02);
    pmtVersion = -1;
  }

(0000752)
rofafor
11-18-08 11:24

The pat repacker should utilize correct version number instead of zero:

--- livestreamer.c.orig 2008-11-18 12:23:51.000000000 +0200
+++ livestreamer.c 2008-11-18 12:26:24.000000000 +0200
@@ -250,7 +250,7 @@
                                 tspat_buf[7] = 12 + 1; // Section length (12 bit): PAT_TABLE_LEN + 1
                                 tspat_buf[8] = (ts_id >> 8); // Transport stream ID (bits 8-15)
                                 tspat_buf[9] = (ts_id & 0xff); // Transport stream ID (bits 0-7)
- tspat_buf[10] = 0xc1; // Version number 0, Current next indicator bit set
+ tspat_buf[10] = ((pat.getVersionNumber() << 1) & 0x3e) | 0xc1; // Version number, Current next indicator bit set
                                 tspat_buf[11] = 0x0; // Section number
                                 tspat_buf[12] = 0x0; // Last section number
                                 tspat_buf[13] = (pmtSid >> 8); // Program number (bits 8-15)

(0000753)
schmirl
11-18-08 11:39

Perfect - thanks. We should probably copy the current next indicator as well:

- tspat_buf[10] = 0xc1; // Version number 0, Current next indicator bit set
+ tspat_buf[10] = ((pat.getVersionNumber() << 1) & 0x3e) | pat.getCurrentNextIndicator() | 0xc0; // Version number, Current next indicator

(0000754)
rofafor
11-18-08 14:26

Ack for current next indicator modification, but I would move the "seen |= SI_PMT;" statement into inside of "if" block.

(0000755)
schmirl
11-18-08 14:34

That whole "seen" stuff was just a dirty hack to find out if PCH requires PAT/PMT at the very beginning of the stream. Odds are that this isn't necessary at all. Jori will test this tonight.

(0000756)
schmirl
11-21-08 09:19

Added patch v5 which contains the minimum changes I expect to be necessary (no feedback from Jori yet). With this patch the stream will no longer start with PAT and PMT. Basically all it does is send out a repacked PAT everytime one is received (plus some minor PAT fixups).

Added TS continuity counter to TS PAT packets:
+ static uint8_t ccounter = 0;
+ ccounter = (ccounter + 1) % 16;
...
+ tspat_buf[3] = 0x10 | ccounter; // Set payload flag, Continuity counter

(0000758)
schmirl
11-24-08 13:12

Commited

Issue History
Date Modified	Username	Field	Change
11-10-08 13:52	schmirl	New Issue
11-12-08 12:42	schmirl	File Added: patpmt_first.diff
11-12-08 12:58	schmirl	File Deleted: patpmt_first.diff
11-12-08 12:59	schmirl	File Added: patpmt_first.diff
11-12-08 13:12	schmirl	Note Added: 0000744
11-12-08 13:12	schmirl	Status	new => feedback
11-12-08 15:34	schmirl	File Added: patpmt_first_v2.diff
11-12-08 15:34	schmirl	Note Added: 0000747
11-12-08 15:35	schmirl	Note Edited: 0000747
11-13-08 09:49	schmirl	File Added: patpmt_first_v3.diff
11-13-08 09:50	schmirl	File Deleted: patpmt_first.diff
11-13-08 09:50	schmirl	File Deleted: patpmt_first_v2.diff
11-13-08 09:51	schmirl	Note Added: 0000748
11-14-08 13:48	schmirl	File Added: patpmt_first_v4.diff
11-14-08 14:02	schmirl	Note Added: 0000750
11-18-08 08:32	schmirl	File Deleted: patpmt_first_v3.diff
11-18-08 08:43	schmirl	Note Added: 0000751
11-18-08 11:24	rofafor	Note Added: 0000752
11-18-08 11:39	schmirl	Note Added: 0000753
11-18-08 14:26	rofafor	Note Added: 0000754
11-18-08 14:34	schmirl	Note Added: 0000755
11-21-08 09:14	schmirl	File Added: patpmt_first_v5.diff
11-21-08 09:19	schmirl	Note Added: 0000756
11-24-08 13:12	schmirl	Status	feedback => resolved
11-24-08 13:12	schmirl	Resolution	open => fixed
11-24-08 13:12	schmirl	Assigned To	=> schmirl
11-24-08 13:12	schmirl	Note Added: 0000758
11-24-08 13:21	bdecord	Issue Monitored: bdecord
11-24-08 13:21	bdecord	Issue End Monitor: bdecord

Notes
(0000744) schmirl 11-12-08 13:12	Problem could be that PMT must be broadcasted before actual stream data. patpmt_first.diff is a crude hack to accomplish this. First packet will be the PAT. The patch doesn't guarantee that the PMT will be the second packet. The first PMT could even get filtered. A clean solution would send the PMT from within cStreamdevPatFilter::Process(). This however requires repacking and I don't want to do this without the ability to test if the packet is ok afterwards (I can't test ATM). Maybe someone can contribute a patch here. If the PMT is filtered, try modifying bool Syncing() const { return seen != (SI_PMT \| SI_PAT); } into bool Syncing() const { return (seen & SI_PAT) == 0; } So we only wait for the PAT to occur, PMT will not be the first packet, but it should show up earlier as without patch.

(0000747) schmirl 11-12-08 15:34 edited on: 11-12-08 15:35	Refined patch patpmt_first_v2.diff guarantees that first packet is PAT, second is PMT. Untested!

(0000748) schmirl 11-13-08 09:51	Previous patches didn't work. Patpmt_first_v3.diff hopefully fixes the problem (Untested!)

(0000750) schmirl 11-14-08 14:02	patpmt_first_v4 fixes PMT generator Minor fixes for PAT generator (reserved bits must be '1')

(0000751) schmirl 11-18-08 08:43	Success with the following modifications on top of v4 patch: *** Retransmit PAT (will give an other continuity error) Two modifications in livestreamer.c: Around line 214, change if (Tid == 0x00 && !pmtPid) { into if (Tid == 0x00) { int pmtPidBak = pmtPid; And around line 275 change m_Streamer->SetPids(pmtPid); Add(pmtPid, 0x02); pmtVersion = -1; into if (pmtPid != pmtPidBak) { m_Streamer->SetPids(pmtPid); Add(pmtPid, 0x02); pmtVersion = -1; }

(0000752) rofafor 11-18-08 11:24	The pat repacker should utilize correct version number instead of zero: --- livestreamer.c.orig 2008-11-18 12:23:51.000000000 +0200 +++ livestreamer.c 2008-11-18 12:26:24.000000000 +0200 @@ -250,7 +250,7 @@ tspat_buf[7] = 12 + 1; // Section length (12 bit): PAT_TABLE_LEN + 1 tspat_buf[8] = (ts_id >> 8); // Transport stream ID (bits 8-15) tspat_buf[9] = (ts_id & 0xff); // Transport stream ID (bits 0-7) - tspat_buf[10] = 0xc1; // Version number 0, Current next indicator bit set + tspat_buf[10] = ((pat.getVersionNumber() << 1) & 0x3e) \| 0xc1; // Version number, Current next indicator bit set tspat_buf[11] = 0x0; // Section number tspat_buf[12] = 0x0; // Last section number tspat_buf[13] = (pmtSid >> 8); // Program number (bits 8-15)

(0000753) schmirl 11-18-08 11:39	Perfect - thanks. We should probably copy the current next indicator as well: - tspat_buf[10] = 0xc1; // Version number 0, Current next indicator bit set + tspat_buf[10] = ((pat.getVersionNumber() << 1) & 0x3e) \| pat.getCurrentNextIndicator() \| 0xc0; // Version number, Current next indicator

(0000754) rofafor 11-18-08 14:26	Ack for current next indicator modification, but I would move the "seen \|= SI_PMT;" statement into inside of "if" block.

(0000755) schmirl 11-18-08 14:34	That whole "seen" stuff was just a dirty hack to find out if PCH requires PAT/PMT at the very beginning of the stream. Odds are that this isn't necessary at all. Jori will test this tonight.

(0000756) schmirl 11-21-08 09:19	Added patch v5 which contains the minimum changes I expect to be necessary (no feedback from Jori yet). With this patch the stream will no longer start with PAT and PMT. Basically all it does is send out a repacked PAT everytime one is received (plus some minor PAT fixups). Added TS continuity counter to TS PAT packets: + static uint8_t ccounter = 0; + ccounter = (ccounter + 1) % 16; ... + tspat_buf[3] = 0x10 \| ccounter; // Set payload flag, Continuity counter

(0000758) schmirl 11-24-08 13:12	Commited

Relationships