But the bandwidth taken by MIDI should be very small compared to the one taken by the 48KHz audio stream, so that sounds very strange !
Obviously it's the multiplexing of audio and MIDI which fails, but I am not much into the USB drivers, either audio or core...
The USB root hub should be 2.0, according to lsusb output:
Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Sounds pointless to try other ports since they are all of the same kind...
The important point here is that, in my opinion, a USB core driver bug seems less probable that a USB audio driver bug because the affected range is very wide (4.12 - 5.0) and at the same time the USB core driver has been tested very widely everyday by virtually all Linux users.
What should I do next ?
Guido
On the 22nd of March 2019 at 16.14 Takashi Iwai tiwai@suse.de wrote:
On Fri, 22 Mar 2019 16:05:17 +0100, Guido Trentalancia wrote:
I suppose the Fedora 29 kernels are included in the range I tested as broken, i.e. 4.12 -- 5.0.
So, there is something triggering it on my system and not the other, maybe the USB system as pointed out by Clemens.
You can try different USB ports, too, e.g. if there are both USB 3.0 and 2.0 ports.
I'll make some research on the two USB systems...
I would also like to try USB Audio driver from kernel 3.0.1.
What else do you suggest?
If any, the problem is rather about the USB core side (likely the host driver), not the USB audio driver. I can imagine some bandwidth management or iosc transfer problem.
Takashi
Guido
Il 22 marzo 2019 15:58:23 CET, Takashi Iwai tiwai@suse.de ha scritto:
On Fri, 22 Mar 2019 15:54:03 +0100, Guido Trentalancia wrote: As already explained, I have tested the following kernels: - kernel 5.0.2 -----> BROKEN - kernel 4.17.10 --> BROKEN - kernel 4.12.9 ---> BROKEN So, it's been around for very long. What do you suggest doing? But which kernel is used for another user who can use without the problem? I suppose they use the recent kernel with Fedora? I have found reports on the web about similar problems (with other audio interfaces) with kernels>3.0.0. Such a regression should have been reported earlier, otherwise it becomes more and more difficult to catch up... thanks, Takashi Regards, Guido Il 22 marzo 2019 15:47:36 CET, Takashi Iwai <tiwai@suse.de> ha scritto: On Fri, 22 Mar 2019 14:44:48 +0100, Guido Trentalancia wrote: Hello Takashi. I have carried out the test that you proposed... My reply follows your quoted text. On Fri, 22/03/2019 at 11.12 +0100, Takashi Iwai wrote: On Fri, 22 Mar 2019 11:04:01 +0100, Guido Trentalancia wrote: Hello Takashi, I am using the latest version of everything, including kernel and ALSA userspace library / tools. The other user has exactly the same hardware and has tested same firmware (both 1.36 and latest 1.46), but with Fedora 29 and it is working. Perhaps Fedora 29 has a different version of the ALSA library, I will find out, try to downgrade, test again and report back. Yes, that'd be really helpful. If aligning the software doesn't fix the issue, it's either because of the hardware or the difference of usage patterns. I have tested exactly the same ALSA userspace library and plugins distributed by Fedora 29 and it does NOT work ! As already explained, the usage pattern is exactly the same between me and the user which is not experiencing this severe problem. Also, the hardware is the same: Hercules P32 DJ (with exactly the same firmware version 1.46 which is the latest). So, the conclusion is that it must be a kernel bug ! I was expecting this, as already pointed out in previous messages. Did you test the very same kernel, too? Without that confirmation, no one can conclude that at all... If the kernel makes difference, you can try identify which kernel version starts showing the problem, and at best, do git bisection. thanks, Takashi Can you please help me fix this bug since you wrote the driver and/or are maintaining it ? Thanks, Guido thanks, Takashi Guido Il 22 marzo 2019 10:53:11 CET, Takashi Iwai <tiwai@suse.de> ha scritto: On Fri, 22 Mar 2019 10:17:17 +0100, Guido Trentalancia wrote: It cannot be a firmware bug, as you say, because: - it does NOT happen on Windows! It doesn't mean that the device behaves correctly as advertised. - it does NOT even always happen on Linux: other users (with different kernel / ALSA library) are not experiencing the same problem; - it happens with several firmware versions, including the latest one (1.36 and 1.46). So it is either a bug in ALSA kernel driver, USB sound driver (more likely) or ALSA library. Now, you are in charge of the USB sound driver, can you please double check?? If it doesn't happen for other users with the very same device, you'd need to identify what's the difference between your case and others. For example, if the difference of alsa-lib matters, you can try the very same condition. This kind of bug can't be easily debugged without the actual hardware, unfortunately. Takashi Regards, Guido Il 22 marzo 2019 09:55:52 CET, Takashi Iwai <tiwai@suse.de ha scritto: On Thu, 21 Mar 2019 23:27:46 +0100, Guido Trentalancia wrote: I wonder if this might be due to a bug in the "USB Audio Driver for Alsa"? https://git.kernel.org/pub/scm/linux/kernel/git/t orvalds/linux.git/tree/sound/usb https://git.kernel.org/pub/scm/linux/kernel/git/t iwai/sound.git/tree/sound/usb More likely a buggy firmware of your USB audio device :) From the driver implementation POV, both audio and MIDI devices are handled by individual endpoints, hence they shouldn't conflict. Or another possibility would be some USB host side issue like the bandwidth. But it's a MIDI stream that is very low data rate, so this sounds also unlikely... thanks, Takashi Guido On the 20th of March 2019 at 21.47 Guido Trentalancia <guido@trentalancia.com wrote: Hello. I am hitting a very serious bug (ALSA kernel driver or ALSA library) when using the Hercules P32 DJ audio interface. The sound is severely distorted during MIDI transfers. To reproduce: + start playing something in a first console: console1# AUDIODEV=hw:2,0 play audio.wav + the audio plays fine + now start "amidi" in a second console while the above track is still playing console2# amidi -p hw:2,0,0 -d + the sound is now severely distorted (basically noise, with some hard- to-distinguish features resembling the original track) until "amidi" is interrupted !
Alsa-devel mailing list Alsa-devel@alsa-project.org https://mailman.alsa-project.org/mailman/listinfo/alsa-devel