Mike, I just realize I missed reply-to-all once again.
In case you were wondering, I was looking for a sync_ep event right in the start of the loading of snd-usb-audio. It happens inside pcm.c. If you see one of those specifically let me know.
With my Pioneer device, the DDJ-SR2, snd-usb-audio recognized the device's implicit feedback EP as sync_ep and that prevented JACK from starting.
Em Qua, 7 de abr de 2021 23:47, Mike Oliphant oliphant@nostatic.org
escreveu:
Hi Geraldo - I had the same thought and I have tried enabling playback feedback while turning off capture feedback without my endpoint patch. It doesn't work - at least not on the GT-1. Playback/capture fails to start.
Would you mind sharing the dyndbg logs for the failure case from the point you connect the device, and trying to start JACK?
Sometimes failures elucidate more than success cases.
Speaking of testing, if you (or anyone else following this discussion) have one of these Roland/BOSS devices, it would be great if you could try testing with playback feedback enabled and my endpoint.c workaround.
I don't have the hardware, but I'm sure it works. The problem is that endpoint.c code really shouldn't be disabled unless all the other options fail, and even then, I'm not so sure the maintainers will be so willing to disable it. So let's keep some other options on the table by debugging a little further.
Thank you, Geraldo
Mike
On Wed, Apr 7, 2021 at 7:21 PM Geraldo Nascimento < geraldogabriel@gmail.com> wrote:
Thanks, Mike, one thing I can tell you right off the bat is that if you turn implicit feedback sync on for both endpoints at the same time like you mentioned, without your workaround inside endpoint.c, they most likely will stall because they would both be waiting for each other to start.
Have you tried engaging the generic implicit feedback sync just for the playback, thereby commenting out the capture implicit feedback entries, and re-enabling snd_usb_endpoint_implicit_feedback_sink(ep) inside endpoint.c? Does playback still fail to start?
If you choose to perform this simple test please share with us the test's dyndbg logs right from the moment you turn on the device, not just after you start jackd.
Sorry for the endless do-this, do-that. Such is the pace of teledebugging
Em Qua, 7 de abr de 2021 21:38, Mike Oliphant oliphant@nostatic.org escreveu:
Geraldo - here is the dmesg output when starting up jackd for capture/playback:
With my patch:
[ 136.734081] usb 3-1.2: Open EP 0x8e, iface=2:1, idx=0 [ 136.734088] usb 3-1.2: channels=2, rate=44100, format=S32_LE, period_bytes=256, periods=2, implicit_fb=0 [ 136.734092] usb 3-1.2: Setting usb interface 2:0 for EP 0x8e [ 136.734233] usb 3-1.2: Setting usb interface 2:1 for EP 0x8e [ 136.734381] usb 3-1.2: 2:1 Set sample rate 44100, clock 0 [ 136.734388] usb 3-1.2: Setting params for data EP 0x8e, pipe 0x70480 [ 136.734394] usb 3-1.2: Set up 12 URBS, ret=0 [ 136.734466] usb 3-1.2: Open EP 0xd, iface=1:1, idx=0 [ 136.734468] usb 3-1.2: channels=2, rate=44100, format=S32_LE, period_bytes=256, periods=2, implicit_fb=1 [ 136.734471] usb 3-1.2: Reopened EP 0x8e (count 1) [ 136.734473] usb 3-1.2: Setting usb interface 1:0 for EP 0xd [ 136.734605] usb 3-1.2: Setting usb interface 1:1 for EP 0xd [ 136.734762] usb 3-1.2: 1:1 Set sample rate 44100, clock 0 [ 136.734774] usb 3-1.2: Setting params for data EP 0xd, pipe 0x68400 [ 136.734781] usb 3-1.2: Set up 12 URBS, ret=0 [ 136.734830] usb 3-1.2: Starting data EP 0xd (running 0) [ 136.734873] usb 3-1.2: 12 URBs submitted for EP 0xd [ 136.734876] usb 3-1.2: Starting data EP 0x8e (running 0) [ 136.734906] usb 3-1.2: 12 URBs submitted for EP 0x8e [ 136.740627] usb 3-1.2: Starting data EP 0x8e (running 1) [ 136.740634] usb 3-1.2: 2:1 Start Capture PCM [ 136.740635] usb 3-1.2: 1:1 Start Playback PCM
With the unpatched kernel:
[ 344.813203] usb 3-1.2: Open EP 0x8e, iface=2:1, idx=0 [ 344.813210] usb 3-1.2: channels=2, rate=44100, format=S32_LE, period_bytes=256, periods=2, implicit_fb=1 [ 344.813213] usb 3-1.2: Open EP 0xd, iface=1:1, idx=0 [ 344.813215] usb 3-1.2: channels=2, rate=44100, format=S32_LE, period_bytes=256, periods=2, implicit_fb=0 [ 344.813217] usb 3-1.2: Setting usb interface 2:0 for EP 0x8e [ 344.813367] usb 3-1.2: Setting usb interface 2:1 for EP 0x8e [ 344.813494] usb 3-1.2: 2:1 Set sample rate 44100, clock 0 [ 344.813503] usb 3-1.2: Setting params for data EP 0x8e, pipe 0x70580 [ 344.813509] usb 3-1.2: Set up 12 URBS, ret=0 [ 344.813513] usb 3-1.2: Setting usb interface 1:0 for EP 0xd [ 344.813587] usb 3-1.2: Setting usb interface 1:1 for EP 0xd [ 344.813758] usb 3-1.2: 1:1 Set sample rate 44100, clock 0 [ 344.813765] usb 3-1.2: Setting params for data EP 0xd, pipe 0x68500 [ 344.813770] usb 3-1.2: Set up 2 URBS, ret=0 [ 344.813858] usb 3-1.2: Reopened EP 0xd (count 1) [ 344.813883] usb 3-1.2: Starting data EP 0xd (running 0) [ 344.813899] usb 3-1.2: 2 URBs submitted for EP 0xd [ 344.819499] usb 3-1.2: Starting data EP 0x8e (running 0) [ 344.819532] usb 3-1.2: 12 URBs submitted for EP 0x8e [ 344.819535] usb 3-1.2: Starting data EP 0xd (running 1) [ 344.819538] usb 3-1.2: 2:1 Start Capture PCM [ 344.819540] usb 3-1.2: 1:1 Start Playback PCM [ 344.820663] usb 3-1.2: Stopping data EP 0xd (running 2) [ 344.820688] usb 3-1.2: Stopping data EP 0x8e (running 1) [ 344.820693] usb 3-1.2: 2:1 Stop Capture PCM [ 344.820695] usb 3-1.2: Stopping data EP 0xd (running 1) [ 344.820697] usb 3-1.2: 1:1 Stop Playback PCM [ 344.833799] usb 3-1.2: Starting data EP 0xd (running 0) [ 344.833852] usb 3-1.2: 2 URBs submitted for EP 0xd [ 344.833867] usb 3-1.2: Starting data EP 0x8e (running 0) [ 344.833888] usb 3-1.2: 12 URBs submitted for EP 0x8e [ 344.833890] usb 3-1.2: Starting data EP 0xd (running 1) [ 344.833891] usb 3-1.2: 2:1 Start Capture PCM [ 344.833893] usb 3-1.2: 1:1 Start Playback PCM
On Wed, Apr 7, 2021 at 1:30 PM Geraldo Nascimento < geraldogabriel@gmail.com> wrote:
Em Qua, 7 de abr de 2021 17:16, Mike Oliphant oliphant@nostatic.org escreveu:
Hi Geraldo - I don't have that patch applied, but it shouldn't make any behavioral difference - it just seems to be simplifying the code.
Point taken.
The issue is that the BOSS GT-1 *does* need implicit feedback on playback to avoid clock timing issues, and the current behavior is disabling that feedback.
Mike, would you mind posting some dyndbg logs for both the stock behaviour and your endpoint.c/generic quirk modified behaviour?
Just add snd_usb_audio.dyndbg=+p to your kernel options and please share the logs.
Mike
On Wed, Apr 7, 2021 at 1:04 PM Geraldo Nascimento < geraldogabriel@gmail.com> wrote:
> Hey Mike, did you catch the latest patch by Takashi Iwai for capture > quirky devices? > > You can find it here: > https://patchwork.kernel.org/project/alsa-devel/patch/20210406113837.32041-1... > > Em Qua, 7 de abr de 2021 16:55, Mike Oliphant oliphant@nostatic.org > escreveu: > >> I had thought that the recent implicit feedback changes were fully >> working >> on the BOSS GT-1, but it turns out that I just hadn't tested well >> enough. >> >> Audio playback and capture works, but with periodic dropouts. I get >> the >> exact same behavior as I did with the quirk to completely disable >> implicit >> feedback. Without the implicit feedback, you get dropouts from >> clock drift >> - how bad probably varies from card to card. On mine it is every >> second or >> so. >> >> If I switch playback feedback for the GT-1 to generic by doing >> "IMPLICIT_FB_GENERIC_DEV(0x0582, 0x01d6)", I get the previous old >> behavior, >> which is that playback completely fails to start. >> >> With generic playback feedback, and using my previous patch to >> endpoint.c >> to avoid playback waiting on capture mentioned here: >> >> >> https://mailman.alsa-project.org/pipermail/alsa-devel/2020-January/161951.ht... >> >> playback and capture work perfectly for me. >> >