Initial NetBSD/UHID native backend implementation #612

ghost · 2023-08-17T09:29:08Z

This is the culmination of the research I did for uhid on NetBSD. It should be as complete as a native backend can be given the limitations of NetBSD's kernel and uhid as a whole. See related issue, #608.

CMakeLists.txt

ghost · 2023-08-18T22:45:45Z

Minor code cleanup. I noticed some improvements I could make when reusing some of the code for another project of mine.

mcuee · 2023-08-21T13:36:40Z

Nice, the first simple test seems to work fine. I need to modify hidapitester to test this more.

I am using NetBSD 9.3 VM running under Proxmox and I attach a generic HID device to the VM.

localhost$ cmake -B build -D HIDAPI_BUILD_HIDTEST=1
-- The C compiler identification is GNU 7.5.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- hidapi: v0.14.0
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE
-- Found PkgConfig: /usr/pkg/bin/pkg-config (found version "1.9.4")
-- Checking for module 'libusb-1.0>=1.0.9'
--   Found libusb-1.0, version 1.0.24
-- Check for Iconv
-- Performing Test Iconv_IS_BUILT_IN
-- Performing Test Iconv_IS_BUILT_IN - Success
-- Found Iconv: built in to C library
-- Performing Test HIDAPI_ICONV_CONST
-- Performing Test HIDAPI_ICONV_CONST - Success
-- Building hidtest
-- Configuring done
-- Generating done
-- Build files have been written to: /home/mcuee/build/hidapi_netbsd/build
localhost$ cmake --build build
[ 16%] Building C object src/netbsd/CMakeFiles/hidapi_netbsd.dir/hid.c.o
[ 33%] Linking C shared library libhidapi-netbsd.so
[ 33%] Built target hidapi_netbsd
[ 50%] Building C object src/libusb/CMakeFiles/hidapi_libusb.dir/hid.c.o
[ 66%] Linking C shared library libhidapi-libusb.so
[ 66%] Built target hidapi_libusb
[ 83%] Building C object hidtest/CMakeFiles/hidtest.dir/test.c.o
[100%] Linking C executable hidtest
[100%] Built target hidtest
localhost$ sudo ./build/hidtest/hidtest
Password:
hidapi test/example tool. Compiled with hidapi version 0.14.0, runtime version 0.14.0.
Compile-time version matches runtime version of hidapi.

Device Found
  type: 0925 7001
  path: uhidev0
  serial_number:
  Manufacturer: Microchip Technology Inc.
  Product:      Generic HID
  Release:      1
  Interface:    -1
  Usage (page): 0x1 (0xffa0)
  Bus type: 1 (USB)

  Report Descriptor: (47 bytes)
0x06, 0xa0, 0xff, 0x09, 0x01, 0xa1, 0x01, 0x09, 0x03, 0x15,
0x00, 0x26, 0xff, 0x00, 0x75, 0x08, 0x95, 0x3f, 0x81, 0x02,
0x09, 0x04, 0x15, 0x00, 0x26, 0xff, 0x00, 0x75, 0x08, 0x95,
0x3f, 0x91, 0x02, 0x09, 0x05, 0x15, 0x00, 0x26, 0xff, 0x00,
0x75, 0x08, 0x95, 0x3f, 0xb1, 0x02, 0xc0,
unable to open device

localhost$ uname -a
NetBSD localhost 9.3 NetBSD 9.3 (GENERIC) #0: Thu Aug  4 15:30:37 UTC 2022  mkrepro@mkrepro.NetBSD.org:/usr/src/sys/arch/amd64/compile/GENERIC amd64

localhost$ sudo usbdevs
addr 1: UHCI root hub, NetBSD
addr 0: xHCI root hub, NetBSD
addr 0: xHCI root hub, NetBSD
 addr 1: Generic HID, Microchip Technology Inc.

ghost · 2023-08-22T05:18:02Z

Nice, the first simple test seems to work fine. I need to modify hidapitester to test this more.

I am using NetBSD 9.3 VM running under Proxmox and I attach a generic HID device to the VM.

Feel free to refine it further but I have reached the limit of what I can see to do. I am considering writing a FreeBSD backend after NetBSD has been merged. It is probably the last native BSD backend that has enough kernel API functionality to work with the current library abstraction. I may end up creating two versions for FreeBSD due to it having the option of using a hidraw compatibility layer as well as the older uhid one.

mcuee · 2023-08-22T06:43:15Z

I am considering writing a FreeBSD backend after NetBSD has been merged. It is probably the last native BSD backend that has enough kernel API functionality to work with the current library abstraction. I may end up creating two versions for FreeBSD due to it having the option of using a hidraw compatibility layer as well as the older uhid one.

That would be wonderful. I tend to think the 1hidraw` backend is the way to go for FreeBSD.

FreeBSD hidraw backend to replace libusb backend #543

ghost · 2023-08-23T13:22:53Z

That would be wonderful. I tend to think the 1hidraw` backend is the way to go for FreeBSD.

Preliminary investigation suggests this has its own problems due to how new this backend is. It isn't enabled by default on fresh installations. Of the two, uhid and hidraw, only one can be attached to any particular device.

uhid is the default if hw.usb.usbhid.enable is set to 0. hidraw can only be used if hw.usb.usbhid.enable is set to 1 and the hidraw module is loaded explicitly. If hw.usb.usbhid.enable is set to 1 but hidraw is not available, there won't be anything setup to provide userspace HID access.

Any solution would probably have to be prepared to use whichever API the FreeBSD environment is configured to use as I don't think it would be a good practice to forcefully enable hidraw against the wishes of the end user.

Youw · 2023-08-23T16:01:54Z

I don't think it would be a good practice to forcefully enable hidraw against the wishes of the end user.

I agree. I'd say that even if we will get the hidraw backend implementation for FreeBSD - that is absolutely fine that it will be functional only if user explicitly enables the hidraw subsystem/driver in his environmnet on his own.

mcuee · 2023-08-24T03:27:52Z

That would be wonderful. I tend to think the 1hidraw` backend is the way to go for FreeBSD.

Preliminary investigation suggests this has its own problems due to how new this backend is. It isn't enabled by default on fresh installations. Of the two, uhid and hidraw, only one can be attached to any particular device.

uhid is the default if hw.usb.usbhid.enable is set to 0. hidraw can only be used if hw.usb.usbhid.enable is set to 1 and the hidraw module is loaded explicitly. If hw.usb.usbhid.enable is set to 1 but hidraw is not available, there won't be anything setup to provide userspace HID access.

Any solution would probably have to be prepared to use whichever API the FreeBSD environment is configured to use as I don't think it would be a good practice to forcefully enable hidraw against the wishes of the end user.

Thanks for the analysis.

However, for the FreeBSD OS, there is already a libusb backend which will be the default as it already works fine (other than common libusb issues: https://github.com/libusb/hidapi/issues?q=is%3Aopen+is%3Aissue+label%3Alibusb). So libusb backend will be the preferred backend in the foreseeable future.

I believe ugen driver is also attached to generic USB HID device along with uhid (or hidraw). That is why it works.

I see this as similar to Linux side last time -- libusb was the preferred and hidraw was secondary as there was too many limitations previously. Now hidraw is the preferred.

But of course you can develop uhid backend first as it is more common for now. However, I tend to believe hidraw may be the future for FreeBSD as well (and maybe NetBSD/OpenBSD).

mcuee · 2023-08-26T02:02:25Z

@Youw

This PR does not support the existing auto-tools build system. Do we want to keep the auto-tools support?

ghost · 2023-08-26T02:04:58Z

This PR does not support the existing auto-tools build system. Do we want to keep the auto-tools support?

I left it out because I found comments saying that it was deprecated so I assumed it was going to be removed in the near future.

Youw · 2023-08-26T07:54:51Z

Do we want to keep the auto-tools support?

No.

I left it out because I found comments saying that it was deprecated so I assumed it was going to be removed in the near future.

This is exactly why.

BTW: I've reviewed all CMake changes - very consistent with what we already have. Very nice, thanks.

Youw

Overall very consistent implementation.
I didn't check all of the implementation/logic (specially what consernes the uhid API/usage) but overall looks very good.
Just a few comments/improvements to consider.

netbsd/hid.c

This also modifies an internal routine for walking the device tree to accept a function pointer for the string comparison so it can be reused in more places than its originally intended one.

ghost · 2023-08-27T03:37:42Z

Re-based on latest master branch.

ghost · 2023-08-27T05:51:00Z

Ok. That should handle the requested changes. Anything else?

Youw · 2023-08-27T08:33:10Z

just one last comment

ghost · 2023-08-27T08:35:33Z

There. That comment should cover it.

libusb/hidapi@c19ae12 libusb/hidapi#612

mcuee added enhancement New feature or request bsd FreeBSD, NetBSD, OpenBSD, etc labels Aug 17, 2023

Youw reviewed Aug 17, 2023

View reviewed changes

CMakeLists.txt Show resolved Hide resolved

Youw reviewed Aug 26, 2023

View reviewed changes

braewoods added 2 commits August 26, 2023 22:36

Initial NetBSD/UHID native backend implementation

bb981cd

netbsd: remove leftover main function from early development

5a1cc8b

This also modifies an internal routine for walking the device tree to accept a function pointer for the string comparison so it can be reused in more places than its originally intended one.

netbsd: incorporate requested changes

0efdc43

Youw approved these changes Aug 27, 2023

View reviewed changes

Youw merged commit c19ae12 into libusb:master Sep 4, 2023
10 checks passed

ghost deleted the netbsd branch September 4, 2023 13:55

sezero mentioned this pull request Nov 11, 2023

hidapi: netbsd uhid native backend in mainstream libsdl-org/SDL#8535

Closed

sezero added a commit to sezero/SDL that referenced this pull request Nov 11, 2023

initial import of netbsd uhid native backend from mainstream

d3307c5

libusb/hidapi@c19ae12 libusb/hidapi#612

sezero added a commit to sezero/SDL that referenced this pull request Nov 11, 2023

initial import of hidapi netbsd uhid native backend from mainstream

c649681

libusb/hidapi@c19ae12 libusb/hidapi#612

sezero added a commit to sezero/SDL that referenced this pull request Nov 14, 2023

initial import of hidapi netbsd uhid native backend from mainstream

a59e9c1

libusb/hidapi@c19ae12 libusb/hidapi#612

sezero added a commit to sezero/SDL that referenced this pull request Nov 24, 2023

initial import of hidapi netbsd uhid native backend from mainstream

51b9199

libusb/hidapi@c19ae12 libusb/hidapi#612

sezero added a commit to sezero/SDL that referenced this pull request Nov 26, 2023

initial import of hidapi netbsd uhid native backend from mainstream

64ee6cf

libusb/hidapi@c19ae12 libusb/hidapi#612

sezero added a commit to sezero/SDL that referenced this pull request Nov 26, 2023

initial import of hidapi netbsd uhid native backend from mainstream

91b639b

libusb/hidapi@c19ae12 libusb/hidapi#612

sezero added a commit to libsdl-org/SDL that referenced this pull request Nov 26, 2023

initial import of hidapi netbsd uhid native backend from mainstream

2f806c8

libusb/hidapi@c19ae12 libusb/hidapi#612

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Initial NetBSD/UHID native backend implementation #612

Initial NetBSD/UHID native backend implementation #612

ghost commented Aug 17, 2023

ghost commented Aug 18, 2023

mcuee commented Aug 21, 2023 •

edited

Loading

ghost commented Aug 22, 2023

mcuee commented Aug 22, 2023 •

edited

Loading

ghost commented Aug 23, 2023 •

edited by ghost

Loading

Youw commented Aug 23, 2023

mcuee commented Aug 24, 2023

mcuee commented Aug 26, 2023

ghost commented Aug 26, 2023 •

edited by ghost

Loading

Youw commented Aug 26, 2023

Youw left a comment

ghost commented Aug 27, 2023

ghost commented Aug 27, 2023

Youw commented Aug 27, 2023

ghost commented Aug 27, 2023

Initial NetBSD/UHID native backend implementation #612

Initial NetBSD/UHID native backend implementation #612

Conversation

ghost commented Aug 17, 2023

ghost commented Aug 18, 2023

mcuee commented Aug 21, 2023 • edited Loading

ghost commented Aug 22, 2023

mcuee commented Aug 22, 2023 • edited Loading

ghost commented Aug 23, 2023 • edited by ghost Loading

Youw commented Aug 23, 2023

mcuee commented Aug 24, 2023

mcuee commented Aug 26, 2023

ghost commented Aug 26, 2023 • edited by ghost Loading

Youw commented Aug 26, 2023

Youw left a comment

Choose a reason for hiding this comment

ghost commented Aug 27, 2023

ghost commented Aug 27, 2023

Youw commented Aug 27, 2023

ghost commented Aug 27, 2023

mcuee commented Aug 21, 2023 •

edited

Loading

mcuee commented Aug 22, 2023 •

edited

Loading

ghost commented Aug 23, 2023 •

edited by ghost

Loading

ghost commented Aug 26, 2023 •

edited by ghost

Loading