The following issue has been ASSIGNED. ====================================================================== <https://bugtrack.alsa-project.org/alsa-bug/view.php?id=2159> ====================================================================== Reported By: nikke Assigned To: Clemens Ladisch ====================================================================== Project: ALSA - driver Issue ID: 2159 Category: PCI - ymfpci Reproducibility: always Severity: minor Priority: normal Status: assigned Distribution: Debian Kernel Version: 2.6.15 ====================================================================== Date Submitted: 05-24-2006 23:44 CEST Last Modified: 06-19-2006 13:11 CEST ====================================================================== Summary: ac3 passthrough not working on recent versions Description: Using the following AC3 passthrough with mplayer works: Linux 2.4.31 alsa-driver 1.0.8-7 (debian stable packaged) Using Linux 2.6.15 (both with alsa-driver 1.10rc3 that ships with the kernel and alsa-driver 1.0.11) it doesn't work. The reciever just says "unlock" and stays quiet. The only change is booting with different kernels, no other config or hardware is touched, so this must be a driver regression. The card is YMF-754 based, connected with coax spdif to my HK reciever. If you need more info, just tell me and I'll dig it up. ====================================================================== ---------------------------------------------------------------------- Raymond - 06-02-06 04:25 ---------------------------------------------------------------------- You may need to revert some patch related to SPDIF to find out the cause http://hg-mirror.alsa-project.org/alsa-kernel?fl=25ac64e78c4c;file=pci/ymfpci/ymfpci_main.c;style=gitweb ---------------------------------------------------------------------- nikke - 06-14-06 22:32 ---------------------------------------------------------------------- I tried using a debian 2.6.8 kernel, and then AC3 passthrough works. /proc/asound/version says ALSA 1.0.4 ... Using a 2.6.14 kernel did not work (ALSA 1.0.10rc1). So, any possibility of getting a list of what patches has been added to ymfpci between v 1.0.4 (that works) and v 1.0.10rc1? It would help narrowing down the candidates quite a bit... ---------------------------------------------------------------------- Raymond - 06-15-06 02:27 ---------------------------------------------------------------------- Please provide output of your YMF-754 lspci -vn cat /proc/asound/cards cat /proc/asound/pcm cat /proc/asound/card0/codec97#0/ac97#0-0+regs The revisions links contain the list of patches. http://hg-mirror.alsa-project.org/alsa-kernel?mf=f43cd9eba10a;path=/pci/ymfpci/;style=gitweb The release date of 1.0.4 is 2004/04/03 and 1.0.10rc1 is 2005/08/20 ---------------------------------------------------------------------- Mimosius - 06-15-06 11:03 ---------------------------------------------------------------------- Hi, I've exactly the same problem. Passthrough since 1.0.10* seems to be broken, 1.0.9 instead works great. output of "cat ac97#0-0+regs (playing AC3, using ac3dec -C)" 0:00 = 6940 0:02 = 9f1f 0:04 = 9f1f 0:06 = 801f 0:08 = 0000 0:0a = 801e 0:0c = 801f 0:0e = 801f 0:10 = 0808 0:12 = 9f1f 0:14 = 9f1f 0:16 = 9f1f 0:18 = 9f1f 0:1a = 0505 0:1c = 0101 0:1e = 0000 0:20 = 0200 0:22 = 0000 0:24 = 0000 0:26 = 800f 0:28 = 0080 0:2a = 0080 0:2c = 0000 0:2e = 0000 0:30 = 0000 0:32 = 0000 0:34 = 0000 0:36 = 0000 0:38 = 9f9f 0:3a = 0000 0:3c = 0000 0:3e = 0000 0:40 = 0000 0:42 = 0000 0:44 = 0000 0:46 = 0000 0:48 = 0000 0:4a = 0000 0:4c = 0000 0:4e = 0000 0:50 = 0000 0:52 = 0000 0:54 = 0000 0:56 = 0000 0:58 = 0000 0:5a = 0000 0:5c = 0000 0:5e = 0000 0:60 = 0000 0:62 = 0000 0:64 = 0000 0:66 = 0000 0:68 = 0000 0:6a = 0000 0:6c = 0000 0:6e = 0004 0:70 = 0000 0:72 = 8000 0:74 = 0380 0:76 = abba 0:78 = 9381 0:7a = 0000 0:7c = 8384 0:7e = 7608 output of "cat /proc/asound/card0/ymfpci (playing AC3, using ac3dec -C)" YMFPCI 0000: 0002 0004: 0000 0008: 0000 000c: 0000 0010: 0000 0014: 0000 0018: 0003 001c: 8202 0020: 8202 0024: 8202 0028: 80008202 002c: 0000 0030: 0000 0034: 0004 0038: 0000 003c: 000c 0040: 0000 0044: 0000 0048: 0000 004c: 0000 0050: 0000 0054: ff00ff 0058: 555500ff 005c: 0000 0060: 807e800f 0064: 7e7608 0068: 7fffff 006c: 0000 0070: 0012 0074: 0000 0078: 0000 007c: 0000 0080: 0000 0084: 3fff3fff 0088: 0000 008c: 0000 0090: 3fff3fff 0094: 0000 0098: 0000 009c: 0000 00a0: 0000 00a4: 3fff3fff 00a8: 0000 00ac: 3fff3fff 00b0: 0000 00b4: 0000 00b8: 0000 00bc: 0000 00c0: 0000 00c4: 5fff 00c8: 0000 00cc: 0001 00d0: 0000 00d4: 0000 00d8: 0000 00dc: 0000 00e0: 0000 00e4: 0000 00e8: 0000 00ec: 0000 00f0: 0000 00f4: 0000 00f8: 0000 00fc: 0000 0100: 0003 0104: 0001 0108: 40000003 010c: 4200043 0110: 002a 0114: 0001 0118: 0000 011c: 0000 0120: 0000 0124: 0000 0128: 0000 012c: 0000 0130: 0000 0134: 0000 0138: 0000 013c: 0000 0140: 001e 0144: 0004 0148: 0004 014c: 0100 0150: 0000 0154: 0018 0158: 3c468000 015c: 3c46be00 0160: 3c46bf00 0164: 3c46c000 there is no ac97 register change playing MP3 then AC3 diff -u <(cat /proc/asound/card0/codec97#0/ac97#0-0+regs) <(sleep 5; cat /proc/asound/card0/codec97#0/ac97#0-0+regs) doesn't output anything register change of /proc/asound/card0/ymfpci (playing MP3 then AC3) diff -u <(cat /proc/asound/card0/ymfpci) <(sleep 5; cat /proc/asound/card0/ymfpci) --- /dev/fd/63 2006-06-15 10:49:46.356867225 +0200 +++ /dev/fd/62 2006-06-15 10:49:46.348866702 +0200 @@ -6,11 +6,11 @@ 000c: 0000 0010: 0000 0014: 0000 -0018: 0001 -001c: 8200 -0020: 0200 -0024: 0200 -0028: 80000200 +0018: 0003 +001c: 8202 +0020: 8202 +0024: 8202 +0028: 80008202 002c: 0000 0030: 0000 0034: 0004 @@ -28,7 +28,7 @@ 0064: 7e7608 0068: 7fffff 006c: 0000 -0070: 0002 +0070: 0012 0074: 0000 0078: 0000 007c: 0000 @@ -64,9 +64,9 @@ 00f4: 0000 00f8: 0000 00fc: 0000 -0100: 0000 -0104: 0000 -0108: 0000 +0100: 0003 +0104: 0001 +0108: 40000003 010c: 4200043 0110: 002a 0114: 0001 ---------------------------------------------------------------------- nikke - 06-15-06 23:43 ---------------------------------------------------------------------- This is on a 2.6.15 kernel with ALSA 1.0.11. lspci -vn: 0000:02:06.0 0401: 1073:0012 Subsystem: 1073:0012 Flags: bus master, medium devsel, latency 64, IRQ 177 Memory at faff8000 (32-bit, non-prefetchable) [size=32K] I/O ports at ecc0 [size=64] I/O ports at ecb8 [size=4] Capabilities: [50] Power Management version 1 cat /proc/asound/cards: 0 [YMF754 ]: YMF754 - Yamaha DS-1E (YMF754) Yamaha DS-1E (YMF754) at 0xfaff8000, irq 177 cat /proc/asound/pcm: 00-03: YMFPCI - PCM2 : YMFPCI - Direct Recording : capture 1 00-02: YMFPCI - Rear : YMFPCI - Rear PCM : playback 1 00-01: YMFPCI - IEC958 : YMFPCI - IEC958 : playback 1 00-00: YMFPCI : YMFPCI : playback 32 : capture 1 cat /proc/asound/card0/codec97#0/ac97#0-0+regs (identical output regardless of idle, playing pcm or playing ac3): 0:00 = 6940 0:02 = 0303 0:04 = 9f1f 0:06 = 801f 0:08 = 0000 0:0a = 801e 0:0c = 801f 0:0e = 801f 0:10 = 9f1f 0:12 = 9f1f 0:14 = 9f1f 0:16 = 9f1f 0:18 = 0303 0:1a = 0000 0:1c = 0000 0:1e = 0000 0:20 = 0000 0:22 = 0000 0:24 = 0000 0:26 = 800f 0:28 = 0080 0:2a = 0080 0:2c = 0000 0:2e = 0000 0:30 = 0000 0:32 = 0000 0:34 = 0000 0:36 = 0000 0:38 = 9f9f 0:3a = 0000 0:3c = 0000 0:3e = 0000 0:40 = 0000 0:42 = 0000 0:44 = 0000 0:46 = 0000 0:48 = 0000 0:4a = 0000 0:4c = 0000 0:4e = 0000 0:50 = 0000 0:52 = 0000 0:54 = 0000 0:56 = 0000 0:58 = 0000 0:5a = 0000 0:5c = 0000 0:5e = 0000 0:60 = 0000 0:62 = 0000 0:64 = 0000 0:66 = 0000 0:68 = 0000 0:6a = 0000 0:6c = 0000 0:6e = 0000 0:70 = 0000 0:72 = 8000 0:74 = 0380 0:76 = abba 0:78 = 9381 0:7a = 0000 0:7c = 8384 0:7e = 7608 ---------------------------------------------------------------------- Mimosius - 06-16-06 13:16 ---------------------------------------------------------------------- I've started from 1.0.9b and applied patch by patch up to 1.0.11. I figured out, that this patch from Clemens Ladisch http://hg-mirror.alsa-project.org/alsa-kernel?cmd=changeset;node=218164651eb8;style=gitweb breaks the AC3-passthrough. playing mp3 or ac3 the AC97 register are filled with the same values, but there is a difference between version 1.0.11, which comes with my SUSE 10.1 Kernel, and my patched 1.0.9b without Clemens patch: --- ac97-1.0.11-ac3 2006-06-15 22:17:15.000000000 +0200 +++ ac97-1.0.9_patched-ac3 2006-06-16 12:49:49.000000000 +0200 @@ -6,18 +6,18 @@ 0:0a = 801e 0:0c = 801f 0:0e = 801f -0:10 = 0808 +0:10 = 9f1f 0:12 = 9f1f 0:14 = 9f1f 0:16 = 9f1f 0:18 = 9f1f -0:1a = 0505 -0:1c = 0101 +0:1a = 0000 +0:1c = 0000 0:1e = 0000 -0:20 = 0200 +0:20 = 0000 0:22 = 0000 0:24 = 0000 -0:26 = 800f +0:26 = 000f 0:28 = 0080 0:2a = 0080 0:2c = 0000 @@ -53,9 +53,9 @@ 0:68 = 0000 0:6a = 0000 0:6c = 0000 -0:6e = 0004 +0:6e = 0008 0:70 = 0000 -0:72 = 8000 +0:72 = ffef 0:74 = 0380 0:76 = abba 0:78 = 9381 and here the diff of the ymfpci registers from 1.0.11 and 1.0.9b(patched) --- ymfpci-1.0.11-ac3 2006-06-15 22:17:20.000000000 +0200 +++ ymfpci-1.0.9_patched-ac3 2006-06-16 12:47:54.000000000 +0200 @@ -24,7 +24,7 @@ 0054: ff00ff 0058: 555500ff 005c: 0000 -0060: 807e0808 +0060: 807e0000 0064: 7e7608 0068: 7fffff 006c: 0000 @@ -32,26 +32,26 @@ 0074: 0000 0078: 0000 007c: 0000 -0080: 0000 +0080: 3fff3fff 0084: 3fff3fff -0088: 0000 +0088: 3fff3fff 008c: 0000 -0090: 3fff3fff +0090: 0000 0094: 0000 0098: 0000 009c: 0000 00a0: 0000 00a4: 3fff3fff -00a8: 0000 +00a8: 3fff3fff 00ac: 3fff3fff 00b0: 0000 00b4: 0000 00b8: 0000 00bc: 0000 00c0: 0000 -00c4: 45a7 +00c4: 5fff 00c8: 0000 -00cc: 0002 +00cc: 0001 00d0: 0000 00d4: 0000 00d8: 0000 @@ -65,7 +65,7 @@ 00f8: 0000 00fc: 0000 0100: 0003 -0104: 0001 +0104: 0000 0108: 40000003 010c: 25a01 0110: 0001 @@ -86,7 +86,7 @@ 014c: 0100 0150: 0000 0154: 0018 -0158: 3db70000 -015c: 3db73e00 -0160: 3db73f00 -0164: 3db74000 +0158: 13810000 +015c: 13813e00 +0160: 13813f00 +0164: 13814000 so AC3 works for me at the moment :) , but thats not the solution. I hope Clemens will review his patch. ---------------------------------------------------------------------- Raymond - 06-17-06 02:43 ---------------------------------------------------------------------- Are you really sure that the patch "ymfpci: add per-voice volume controls" break AC3 passthrough, the purpose of that patch is to implement pcm volume control per substream using the hardware mixing of ymfpci Please note that STAC9708 has no SPDIF control registers, there is no need to compare ac97 registers iecset ---------------------------------------------------------------------- Mimosius - 06-16-06 13:37 ---------------------------------------------------------------------- Hi Raymond, yes, I'm really sure this patch breaks AC3 passthrough. A comparison of the /proc/asound/card0/ymfpci registers against 1.0.9b (patched up to 1.0.11 without Clemens patch) and 1.0.11 (native) is included in my previous note. ---------------------------------------------------------------------- Raymond - 06-16-06 16:25 ---------------------------------------------------------------------- I don't have this card. The DAC resolution of STAC9708 is 18-bit , this may mean that there is 2-bit gain when 16-bit audio pass through the hardware mixer. But AC3 stream (passthrough) don't need any gain/attenutation. ---------------------------------------------------------------------- nikke - 06-17-06 01:17 ---------------------------------------------------------------------- I tried to do a quick revert of just that patch from 1.0.11, but there was too many conflicts... However, I noticed that the rear-channel-swapping-thingie shows up in the same function that this patch modifies the most. And since AC3 passthrough works in 1.0.9 this suggests that the patch really breaks things that the rear-channel-swap-patch tries to fix... Mimosius: Can you make a patch between stock 1.0.11 and your 1.0.11-with-working-ac3 so I can test too, and other people really can see that it's the mentioned patch that breaks things? ---------------------------------------------------------------------- Raymond - 06-17-06 10:24 ---------------------------------------------------------------------- Which application change those undocumented registers of STAC9708 ? 0:68 = 0000 0:6a = 0000 0:6c = 0000 -0:6e = 0004 +0:6e = 0008 0:70 = 0000 -0:72 = 8000 +0:72 = ffef 0:74 = 0380 0:76 = abba 0:78 = 9381 Are all channels correct when you run speaker-test -Dsurround40 -c 4 -t wav ? ---------------------------------------------------------------------- Raymond - 06-17-06 11:02 ---------------------------------------------------------------------- Mimosius There is a patch which supposed to fix AC3 output http://hg-mirror.alsa-project.org/alsa-kernel?cmd=changeset;node=fac92b1b0699;style=gitweb Do this patch fix the problem ? However there is another patch which add an module parameter MODULE_PARM_DESC(rear_swap, "Swap rear channels (must be enabled for correct IEC958 (S/PDIF)) output"); http://hg-mirror.alsa-project.org/alsa-kernel?cmd=changeset;node=2e21f4940a0f;style=gitweb ---------------------------------------------------------------------- Raymond - 06-17-06 14:25 ---------------------------------------------------------------------- Are you sure that your patch in https://bugtrack.alsa-project.org/alsa-bug/view.php?id=1447 does not break AC3 pass through ? It seem that there are 3 sources for SPDIF out 1) AC3 2) Digital Mixer 3) SPDIF in ---------------------------------------------------------------------- nikke - 06-17-06 15:16 ---------------------------------------------------------------------- Being less tired than yesterday, I tried the following: Using Linux 2.6.8: ALSA 1.0.9b AC3 works, 1.0.10 AC3 does NOT work. When reverting the "ymfpci: add per-voice volume controls" patch from 1.0.10 AC3 passthrough started to work. Since 1.0.10 compiles on Linux 2.6.15 I have been able to verify that reverting said patch makes AC3 work on my normal setup too. To conclude, I agree that "ymfpci: add per-voice volume controls" breaks AC3 passthrough. ---------------------------------------------------------------------- Raymond - 06-17-06 16:23 ---------------------------------------------------------------------- It seem that is a bug related to "ymfpci: add per-voice volume controls" It should assign kctl->private_value = substream in snd_ymfpci_playback_open() instead of snd_ymfpci_mixer() I had upload a small program (iface_pcm.c) to test this feature for au88x0 in https://bugtrack.alsa-project.org/alsa-bug/view.php?id=1835 Without a suitable fuction in alsa-lib, this feature is unusable for those hardware mixing sound cards ---------------------------------------------------------------------- rlrevell - 06-17-06 18:22 ---------------------------------------------------------------------- Raymond, Please post your patch to alsa-devel at lists.sourceforge.net so it can be fixed in the next release. ---------------------------------------------------------------------- Raymond - 06-18-06 13:30 ---------------------------------------------------------------------- I don't have this card, you better wait Clemens Ladisch's review. BTW, Is there any reason why dsnoop is not eanbled ? The volume per voice feature is hard to use, especially when different drivers implement this feature in different ways 1) emu10k1 - use index 2) ymfpci - use subdevice Can any of you try speaker-test -Dsurround40 -c 4 -t wav ? https://bugtrack.alsa-project.org/alsa-bug/view.php?id=2026 Issue History Date Modified Username Field Change ====================================================================== 05-24-06 23:44 nikke New Issue 05-24-06 23:44 nikke Distribution => Debian 05-24-06 23:44 nikke Kernel Version => 2.6.15 05-24-06 23:47 nikke Issue Monitored: nikke 06-02-06 04:23 Raymond Note Added: 0010045 06-02-06 04:25 Raymond Note Edited: 0010045 06-12-06 13:14 Mimosius Issue Monitored: Mimosius 06-14-06 22:32 nikke Note Added: 0010216 06-15-06 02:18 Raymond Note Added: 0010221 06-15-06 02:27 Raymond Note Edited: 0010221 06-15-06 11:03 Mimosius Note Added: 0010237 06-15-06 23:43 nikke Note Added: 0010246 06-16-06 13:01 Mimosius Note Added: 0010250 06-16-06 13:16 Mimosius Note Edited: 0010250 06-16-06 13:26 Raymond Note Added: 0010251 06-16-06 13:31 Raymond Note Edited: 0010251 06-16-06 13:37 Mimosius Note Added: 0010252 06-16-06 16:25 Raymond Note Added: 0010255 06-17-06 01:17 nikke Note Added: 0010262 06-17-06 02:43 Raymond Note Edited: 0010251 06-17-06 10:21 Raymond Note Added: 0010266 06-17-06 10:24 Raymond Note Edited: 0010266 06-17-06 11:02 Raymond Note Added: 0010267 06-17-06 14:25 Raymond Note Added: 0010268 06-17-06 15:16 nikke Note Added: 0010269 06-17-06 16:10 Raymond Note Added: 0010270 06-17-06 16:23 Raymond Note Edited: 0010270 06-17-06 18:22 rlrevell Note Added: 0010280 06-18-06 04:03 Raymond Note Added: 0010295 06-18-06 04:08 Raymond Note Edited: 0010295 06-18-06 04:14 Raymond Note Edited: 0010295 06-18-06 13:30 Raymond Note Edited: 0010295 06-19-06 13:11 Clemens LadischStatus new => assigned 06-19-06 13:11 Clemens LadischAssigned To => Clemens Ladisch ====================================================================== _______________________________________________ Alsa-devel mailing list Alsa-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.sourceforge.net/lists/listinfo/alsa-devel
