feat: add usb host uac driver and examples (IEC-99)

[leeebo]

The PR aims to add UAC Host Driver (Based on ESP-IDF USB Host Driver), featuring:

• Compatibility with UAC 1.0 specification
• Independent IN/OUT streaming control
• Independent Volume/Mute control
• Multiple devices/instances for future HUB support

Checklist:

• [x] (No mock test, can only be test locally) Speaker and Mic works as normal - [x] example: Play MP3 as normal I have removed the examples to esp-iot-solution/examples, reducing the work of review here. - [ ] example: Record voice to SDcard

• [x] thread-safe protect (ringbuffer safe delete, critical section protect)

• [x] refactor open/close device as @roma-jam suggest

• [x] Support Parse USB Audio device with MIXER_UNIT see

[leeebo]

@tore-espressif @roma-jam @Dazza0 PTAL!

[roma-jam]

Hey @leeebo, Wow, a lot of code, great job! I briefly checked the logic, and it seems that it is the same as we have right now in hid_driver v1.0.2 (correct me if I am wrong).

The main problem there - is that if device has at least one interface, driver opens the device with usb_host_device_open() and hold the device opened, even if the interface is not claimed (opened) by user. This happened during the client_event_cb which we should fulfill as fast as we can.

In that case, it is better to "touch" the new device, checking the uac interface presence and add it to the list. I made the changes in hid driver here (but they are not ready to be merge yet): https://github.com/espressif/idf-extra-components/pull/214

But I explained the idea in https://github.com/espressif/idf-extra-components/pull/214#issuecomment-1761262534

Please, read the document and share your opinion.

[leeebo]

@roma-jam Yes, you are right, the device handling logic and data structure is same as HID v1.0.2. Thanks for yor advice, I will check it.

[peter-marcisovsky]

@leeebo regarding the test app:

I took a look at the failing CI logs, could you please try renaming the pytest_usb_host_hid.py to pytest_usb_host_uac.py in your test_app folder? So we will have passing CI for this PR.

[leeebo]

@leeebo regarding the test app:

I took a look at the failing CI logs, could you please try renaming the pytest_usb_host_hid.py to pytest_usb_host_uac.py in your test_app folder? So we will have passing CI for this PR.

Hi @peter-marcisovsky , as we currently have no UAC 1.0 device for the mock test (tinyusb supports UAC2.0 only) , I just test it using my local UAC device, any advice for the test?

[peter-marcisovsky]

@leeebo sorry I was not aware that the tinyusb supports only UAC2.0.

In this case, please enable at least build in CI by adding a path to your new component to .idf_build_apps.toml file in the root dir.

[leeebo]

Another topic is whether we can adjust ep's MPS before claiming the interface, which, although not standard, makes it easier to provide workaround when the FIFO size is limited.

In this Headset case, the speaker supports sampling rate 96KHz 16bit 2ch at maximum, so the OUT MPS is 96*2*2 = 384. It is used when the user selects 96KHz, and will not be touched in most time. It is sufficient to declare an interface with MPS=192~193 when selecting 48KHz.

Full descriptor honor_headset_pcapng.zip

[roma-jam]

Hi @leeebo, I have only two UAC devices I could use. One of them fails with:

I (3253) UAC_TEST: USB Host installed
I (3253) UAC_TEST: UAC Class Driver installed
E (128523) HCD DWC: EP MPS (200) exceeds supported limit (128)
E (128523) uac-host: uac_host_interface_claim_and_prepare_transfer(897): Unable to claim Interface
E (128528) uac-host: uac_host_device_start(1930): Unable to claim Interface

Which seems OK, as this device has two audio interfaces, where mps for EP is 200 B.

Another one fails with:

 (13845) UAC_TEST: USB Host installed
I (13845) UAC_TEST: UAC Class Driver installed
W (19301) uac-host: Feature Unit ID 2 not corresponding to the input terminal
W (19301) uac-host: Feature Unit ID 7 not corresponding to the input terminal

assert failed: block_trim_free tlsf.c:496 (block_is_free(block) && "block must be free")
HINT: CORRUPT HEAP: heap metadata corrupted resulting in TLSF malfunction.
Make sure you are not making out of bound writing on the memory you allocate in your application.
Make sure you are not writing on freed memory.
For more information run 'idf.py docs -sp api-reference/system/heap_debug.html'.

Anyway, it seems that I am not sure how to verify it on a hardware. I am launching test_app, test 5 and then connect the device to the port. Am I doing something wrong?

[leeebo]

W (19301) uac-host: Feature Unit ID 2 not corresponding to the input terminal W (19301) uac-host: Feature Unit ID 7 not corresponding to the input terminal assert failed: block_trim_free tlsf.c:496 (block_is_free(block) && "block must be free")

Hi @roma-jam thanks for your test, for the first device, it related to the topic "should we support a configurable fifo size?",

For the second device, it must be something wrong in the descriptor's parsing logic, would you please share the full descriptors?

[leeebo]

@leeebo sorry I was not aware that the tinyusb supports only UAC2.0.

In this case, please enable at least build in CI by adding a path to your new component to .idf_build_apps.toml file in the root dir.

Hi @peter-marcisovsky would you please point out how to just build the test app without run it? 388bad100efac41d80a11236d2769a49c52999b2 seems not work

[roma-jam]

Sure,

For the second device, it must be something wrong in the descriptor's parsing logic, would you please share the full descriptors?

jabra_ev_hs.txt

[peter-marcisovsky]

Hi @leeebo, to reply to your comment:

Hi @peter-marcisovsky would you please point out how to just build the test app without run it? 388bad1 seems not work

To build a test app, without running in in the CI, you could mark pytest as temporarily disabled in CI

We are using this approach, for example in usb examples in esp-idf

Edit: Sorry, I forgot we are not in esp-idf CI. So this tag unfortunately will not work.

[leeebo]

Sure,

For the second device, it must be something wrong in the descriptor's parsing logic, would you please share the full descriptors?

jabra_ev_hs.txt

Hi @roma-jam, the logic blocks in your UAC device is like the below:

However the current feature unit parsing logic can not work with MIXER_UNIT, I will refactor the parsing logic in this PR to make it more flexible (Replace the one sequential descriptor query with multiple times )

[leeebo]

Hi @roma-jam , the latest commit should fixed the MIXER_UNIT handle bug, please try again if you have time.

BTW, would you please try with ESP32-P4 for the EP MPS (200) exceeds supported limit (128) issue, it should work.

[roma-jam]

@leeebo,

That's the output with Jabra Evolve headset on S3:

Running test uac tx rx loopback with disconnect...
I (2393) UAC_TEST: USB Host installed
I (2393) UAC_TEST: UAC Class Driver installed
W (8311) uac-host: UAC Interface 2->1, Frequency Number 5 exceed the maximum 4
W (8313) UAC_TEST: Speaker channels 2 and microphone channels 1 are not the same
I (8351) uac-host: Set Interface 1-1
I (8351) uac-host: Set EP 82 frequency 8000
I (8426) uac-host: Set Interface 2-1
I (8426) uac-host: Set EP 02 frequency 8000
I (9593) uac-host: Set Interface 1-0
I (9599) uac-host: Set Interface 2-0
I (9736) uac-host: Set Interface 1-1
I (9736) uac-host: Set EP 82 frequency 8000
I (9816) uac-host: Set Interface 2-1
I (9816) uac-host: Set EP 02 frequency 8000
E (10977) USBH: Device 1 gone
E (10977) uac-host: uac_control_transfer(1473): Unable to submit control transfer
W (10978) uac-host: Set Interface 1-0 Failed
I (10983) UAC_TEST: UAC Device disconnected
E (10997) uac-host: uac_control_transfer(1473): Unable to submit control transfer
W (10997) uac-host: Set Interface 2-0 Failed
I (11001) UAC_TEST: UAC Device disconnected
All devices freed
I (11987) UAC_TEST: UAC Driver uninstall
No more clients
I (11987) UAC_TEST: USB Host shutdown
MALLOC_CAP_8BIT: Before 383072 bytes free, After 383056 bytes free (delta -16)
MALLOC_CAP_32BIT: Before 383072 bytes free, After 383056 bytes free (delta -16)
./main/test_host_uac.c:777:test uac tx rx loopback with disconnect:PASS
Test ran in 9759ms

-----------------------
1 Tests 0 Failures 0 Ignored 
OK
Enter next test, or 'enter' to see menu

ESP32-P4 for the EP MPS (200) exceeds supported limit (128) output (BTW, I don't have test number 5 for P4, hope that how is was planned):

Running test uac rx reading...
I (70774) UAC_TEST: USB Host installed
I (70774) UAC_TEST: UAC Class Driver installed
I (79664) uac-host: Set Interface 2-1
I (79665) uac-host: Set EP 82 frequency 48000
I (79665) UAC_TEST: Start reading data from MIC
I (84687) UAC_TEST: Stop reading data from MIC
I (84687) uac-host: Set Interface 2-0
I (85198) UAC_TEST: UAC Driver uninstall
No more clients
All devices freed
I (85198) UAC_TEST: USB Host shutdown
MALLOC_CAP_8BIT: Before 604376 bytes free, After 604360 bytes free (delta -16)
MALLOC_CAP_32BIT: Before 604376 bytes free, After 604360 bytes free (delta -16)
./main/test_host_uac.c:381:test uac rx reading:PASS
Test ran in 2656ms

-----------------------
1 Tests 0 Failures 0 Ignored 
OK

[peter-marcisovsky]

Hi @leeebo Just tired the test app again. This time I tried running pytest. There is a following issue:

2024-05-22 11:19:41 |    |   \/   _____/\______   \ _/  |_  ____   _______/  |_ 
2024-05-22 11:19:41 |    |   /\_____  \  |    |  _/ \   __\/ __ \ /  ___/\   __\
2024-05-22 11:19:41 |    |  / /        \ |    |   \  |  | \  ___/ \___ \  |  |  
2024-05-22 11:19:41 |______/ /_______  / |______  /  |__|  \___  >____  > |__|  
2024-05-22 11:19:41                  \/         \/             \/     \/        
2024-05-22 11:19:41 
2024-05-22 11:19:41 
2024-05-22 11:19:41 Press ENTER to see the list of tests.
2024-05-22 11:19:41 [uac_host]
2024-05-22 11:19:41 Running tests matching '[uac_host]'...
2024-05-22 11:19:41 Running test uac device handling...
2024-05-22 11:19:41 I (591) UAC_TEST: USB Host installed
2024-05-22 11:19:41 I (591) UAC_TEST: UAC Class Driver installed
2024-05-22 11:19:42 ./main/test_host_uac.c:308:test uac device handling:FAIL: Expected 3 Was 1. Function [uac_host][known_device]MALLOC_CAP_8BIT: Before 383072 bytes free, After 368080 bytes free (delta -14992)
2024-05-22 11:19:42 
2024-05-22 11:19:42 Running test uac rx reading...
2024-05-22 11:19:42 E (998) usb_phy: usb_new_phy(296): selected PHY is in use
2024-05-22 11:19:42 ./main/test_host_uac.c:151:test uac rx reading:FAIL: Expected 0 Was 259. Function [uac_host][rx]MALLOC_CAP_8BIT: Before 368080 bytes free, After 363252 bytes free (delta -4828)
2024-05-22 11:19:42 
2024-05-22 11:19:42 Running test uac tx writing...
2024-05-22 11:19:42 E (1023) usb_phy: usb_new_phy(296): selected PHY is in use
2024-05-22 11:19:42 ./main/test_host_uac.c:151:test uac tx writing:FAIL: Expected 0 Was 259. Function [uac_host][tx]MALLOC_CAP_8BIT: Before 363252 bytes free, After 358424 bytes free (delta -4828)

If a test fails, in my case its the very first test (the known devices error which I already pointed out, and you already explained.. so it's not an issue here), all the following tests collected by pytest will fail as well because of E (998) usb_phy: usb_new_phy(296): selected PHY is in use

Could you please release the PHY, once a test fails? In the esp-idf usb tests for example, we are calling usb_del_phy() during the test teardown.

[leeebo]

@peter-marcisovsky

Could you please release the PHY, once a test fails? In the esp-idf usb tests for example, we are calling usb_del_phy() during the test teardown.

Fixed now, you will get 5 Tests 1 Failures with your device now.

[leeebo]

Hi @tore-espressif @roma-jam @peter-marcisovsky Thanks very much for the review work. I will squash the commits and release the first usb_host_uac v1.0.0.