Properly determine libusb read size for large reports (Fixes #274) #728

sudobash1 · 2025-03-25T20:27:04Z

Currently the libusb version of hidapi simply reads up to wMaxPacketSize bytes as the report. This is problematic when reports are longer than wMaxPacketSize. The current behavior will split that report up.

The proper solution is to review the report descriptor to find the longest input report and use that as the length of the libusb_fill_interrupt_transfer buffer. (Note: there is no need to manually get multiple USB packets and concatenate them together to fit the report length. USB already handles that for us.)

This will still work for HID devices when some input reports are shorter than others. The HID device will just send a short packet terminator and libusb will give us the shorter buffer.

The substance of these changes is in the get_max_input_size method. It uses the same basic report descriptor parsing as get_usage. I considered changing the code so that there could be shared parsing code, but I decided that was overkill for this.

Fixes #274

sudobash1 · 2025-03-25T22:08:47Z

PR updated to remove different signedness comparison compiler warning when compiling with -Wall.

Be-ing · 2025-03-26T01:07:51Z

Thank you for finally taking care of this longstanding bug.

Youw

Thanks for this, I'll perform a thorrow review a bit later.
@Be-ing can you confirm it fixes the issue with long reports too?

libusb/hid.c

Fixes libusb#274

JoergAtGithub · 2025-03-26T18:21:25Z

libusb/hid.c

+   Requires an opened device with *claimed interface*.
+
+   The return value is the size on success and -1 on failure. */
+static size_t get_max_input_size(libusb_device_handle *handle, int interface_num, uint16_t expected_report_descriptor_size)


This function seems to be the functional the same as InputReportByteLength from https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/hidpi/ns-hidpi-_hidp_caps on Windows.

We have a lot of test data for the Windows unit test, which stores the result of this function:

hidapi/windows/test/data/045E_02FF_0005_0001.pp_data

Line 21 in 0ab6c14

pp_data->caps_info[0]->ReportByteLength = 16

together with the real ReportDescriptor https://github.com/libusb/hidapi/blob/master/windows/test/data/045E_02FF_0005_0001_real.rpt_desc

What do you think about creating a unit test for the new get_max_input_size function using the _real.rpt_desc files as input and compare the result with the value stored in the .pp_data.

Nice! That does look like a great dataset. I'll see if I can figure a clean way to make a unit test, and see how it fares with that data.

Currently the get_max_input_size is a static function, but I need a way to use it externally in the unit test. It seems that the windows library has some extensions prefixed with winapi. Should I rename the get max function to hid_libusbapi_get_max_input_report_size and mark it HID_API_EXPORT_CALL?

I will also need to change how it gets the report descriptor, but that won't be a problem.

Making it with HID_API_EXPORT_CALL would make it a public (at least on binary level) function, which is undesirable, if it is required for tests only. Maybe export it only when building unit-tests etc. or try to avoid it by using static linking or smth else (or have it in an internal header file?)

Have a look at the Windows unit test ( https://github.com/libusb/hidapi/tree/master/windows/test ) for the report descriptor reconstructor. There we had exactly the same challenges, hid_winapi_descriptor_reconstruct_pp_data was the internal function to test there.

@JoergAtGithub I did actually look at hid_winapi_descriptor_reconstruct_pp_data, and it is marked as HID_API_EXPORT_CALL in the header file:

hidapi/windows/hidapi_winapi.h

Line 68 in 0ab6c14

int HID_API_EXPORT_CALL hid_winapi_descriptor_reconstruct_pp_data(void *hidp_preparsed_data, unsigned char *buf, size_t buf_size);

Although, now that I am looking, it is not marked in the C file:

hidapi/windows/hidapi_descriptor_reconstruct.c

Line 181 in 0ab6c14

int hid_winapi_descriptor_reconstruct_pp_data(void *preparsed_data, unsigned char *buf, size_t buf_size)

WRT @Youw's suggestion. I could simplify the function a bit so that it doesn't depend on anything in hid.c and then move it to hid_max_input_report.h. If that is acceptable, it is probably the easiest solution.

hid_winapi_descriptor_reconstruct_pp_data, and it is marked as HID_API_EXPORT_CALL in the header file

Right - that is a public API function. That is also the reason why there is a winapi prefix in the name.

it is not marked in the C file

Not required, if it is marked in the header.

I could simplify the function a bit so that it doesn't depend on anything in hid.c and then move it to hid_max_input_report.h. If that is acceptable, it is probably the easiest solution.

Yes, sounds like simples solution for now. Go for it.

I have a test, but it isn't passing. It could be that I am misunderstanding the Windows HID structures. I have only a vague understanding of them.

For example, in the 046A_0011_0006_0001 test data, it gives a max input size of 9 bytes:

hidapi/windows/test/data/046A_0011_0006_0001.pp_data

Lines 16 to 19 in 0ab6c14

pp_data->caps_info[0]->FirstCap = 0

pp_data->caps_info[0]->LastCap = 2

pp_data->caps_info[0]->NumberOfCaps = 2

pp_data->caps_info[0]->ReportByteLength = 9

But two input caps (cap[0] and cap[1]) are:

hidapi/windows/test/data/046A_0011_0006_0001.pp_data

Lines 37 to 38 in 0ab6c14

pp_data->cap[0]->BitSize = 1

pp_data->cap[0]->ReportCount = 8

hidapi/windows/test/data/046A_0011_0006_0001.pp_data

Lines 83 to 84 in 0ab6c14

pp_data->cap[1]->BitSize = 8

pp_data->cap[1]->ReportCount = 6

According to the documentation I read, the report size is just BitSize * ReportCount (plus a byte for the report number I assume).

This gives report sizes of (1*8)/8 + 1 = 2 and (8*6)/8 + 1 = 7. So I don't know where the ReportByteLength = 9 value is coming from.

Do either of you know what I am missing here?

mcuee · 2025-03-27T05:12:38Z

BTW, there is a typo here.
@Youw and @JoergAtGithub.

hidapi/windows/test/data/046A_0011_0006_0001.pp_data

Line 4 in 0ab6c14

    
           dev->manufacturer_string = "dev->product_string      = "dev->release_number      = 0x0100

dev->manufacturer_string = "dev->product_string      = "dev->release_number      = 0x0100

JoergAtGithub · 2025-03-27T06:17:16Z

I put 046A_0011_0006_0001_real.rpt_desc into https://eleccelerator.com/usbdescreqparser/ and got:

0x05, 0x01,        // Usage Page (Generic Desktop Ctrls)
0x09, 0x06,        // Usage (Keyboard)
0xA1, 0x01,        // Collection (Application)
0x05, 0x07,        //   Usage Page (Kbrd/Keypad)
0x19, 0xE0,        //   Usage Minimum (0xE0)
0x29, 0xE7,        //   Usage Maximum (0xE7)
0x15, 0x00,        //   Logical Minimum (0)
0x25, 0x01,        //   Logical Maximum (1)
0x75, 0x01,        //   Report Size (1)
0x95, 0x08,        //   Report Count (8)
0x81, 0x02,        //   Input (Data,Var,Abs,No Wrap,Linear,Preferred State,No Null Position)
0x75, 0x08,        //   Report Size (8)
0x95, 0x01,        //   Report Count (1)
0x81, 0x03,        //   Input (Const,Var,Abs,No Wrap,Linear,Preferred State,No Null Position)
0x19, 0x00,        //   Usage Minimum (0x00)
0x29, 0xDD,        //   Usage Maximum (0xDD)
0x15, 0x00,        //   Logical Minimum (0)
0x26, 0xDD, 0x00,  //   Logical Maximum (221)
0x75, 0x08,        //   Report Size (8)
0x95, 0x06,        //   Report Count (6)
0x81, 0x00,        //   Input (Data,Array,Abs,No Wrap,Linear,Preferred State,No Null Position)
0x05, 0x08,        //   Usage Page (LEDs)
0x19, 0x01,        //   Usage Minimum (Num Lock)
0x29, 0x03,        //   Usage Maximum (Scroll Lock)
0x15, 0x00,        //   Logical Minimum (0)
0x25, 0x01,        //   Logical Maximum (1)
0x75, 0x01,        //   Report Size (1)
0x95, 0x03,        //   Report Count (3)
0x91, 0x02,        //   Output (Data,Var,Abs,No Wrap,Linear,Preferred State,No Null Position,Non-volatile)
0x75, 0x05,        //   Report Size (5)
0x95, 0x01,        //   Report Count (1)
0x91, 0x03,        //   Output (Const,Var,Abs,No Wrap,Linear,Preferred State,No Null Position,Non-volatile)
0xC0,              // End Collection

The InputReport contains:

8x1bit data
1x8bit constant padding
6x8bit data
This sums to 8byte, but InputReportByteLength is defined as follows:
Specifies the maximum size, in bytes, of all the input reports. Includes the report ID, which is prepended to the report data. If report ID is not used, the ID value is zero.
So we've to add a byte for the ReportID and are at 9 bytes.

JoergAtGithub · 2025-03-27T06:20:44Z

BTW, there is a typo here. @Youw and @JoergAtGithub.

hidapi/windows/test/data/046A_0011_0006_0001.pp_data

Line 4 in 0ab6c14

dev->manufacturer_string = "dev->product_string = "dev->release_number = 0x0100
dev->manufacturer_string = "dev->product_string      = "dev->release_number      = 0x0100

There is something missing, as this is autogenerated by pp_data_dump, I guess there was something in the manufacturer_string , that pp_data_dump couldn't handle. Could you please open a dedicated issue for this, as this is unrelated to this PR.

mcuee · 2025-03-27T07:34:55Z

There is something missing, as this is autogenerated by pp_data_dump, I guess there was something in the manufacturer_string , that pp_data_dump couldn't handle. Could you please open a dedicated issue for this, as this is unrelated to this PR.

pp_data_dump issue for 046A_0011_0006_0001 -- Lower Priority #729

JoergAtGithub · 2025-03-27T19:48:22Z

libusb/test/CMakeLists.txt

+     if(NOT EXISTS "${TEST_PP_DATA}")
+          message(FATAL_ERROR "Missing '${TEST_PP_DATA}' file for '${TEST_CASE}' test case")
+     endif()
+     set(TEST_EXPECTED_DESCRIPTOR "${CMAKE_CURRENT_LIST_DIR}/../../windows/test/data/${TEST_CASE}_expected.rpt_desc")


You need to parse the _real.rpt_desc, not the _expected.rpt_desc. The real is what is dumped from the device, and the expected is what the Windows ReportDescriptor-Reconstructor generates out of the .pp_data.

I see. I assumed that the real and expected should contain equivalent reports.

The _real.rpt_desc files do not follow a consistent format, so should I parse them manually (should be pretty doable with a bit of regex) into new files and add them to the repo?

Maybe you can put them into https://eleccelerator.com/usbdescreqparser/ to unify the format. That makes them also human readable.

JoergAtGithub · 2025-03-27T19:49:30Z

libusb/hid.c

+			if (report_count < 0 || report_size < 0) {
+				/* We are missing size or count. That isn't good. */
+				return 0;


Suggested change

if (report_count < 0 || report_size < 0) {

/* We are missing size or count. That isn't good. */

return 0;

if (report_count < 0 || report_size < 0) {

/* We are missing size or count. That isn't good. */

return -1;

This would mean a corrupt ReportDescriptor

I wasn't consistent with whether I was using size_t or ssize_t. In my last commit, I settled on size_t, but perhaps I should change it back to ssize_t so that we can more clearly differentiate between errors and a zero value (if there are no feature reports, it will return 0, which is not an error).

Makes sense!

sudobash1 · 2025-03-27T20:00:14Z

Thanks @JoergAtGithub for walking me through that. I was misreading the descriptor.

I have added tests for libusb using the same data as the windows tests. And I extended the functionality of the libusb method to be able to calculate the maximum output and feature report sizes as well. This has no functional use currently, but it lets us run three times as many tests since the pp_data files have all three max sizes available.

JoergAtGithub · 2025-03-27T20:10:00Z

And I extended the functionality of the libusb method to be able to calculate the maximum output and feature report sizes as well. This has no functional use currently, but it lets us run three times as many tests since the pp_data files have all three max sizes available.

Independent of this PR, I think it would generally make sense to store these 3 values in the device structure. On Windows we would have to use the values InputReportByteLength, OutputReportByteLength and FeatureReportByteLength from https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/hidpi/ns-hidpi-_hidp_caps . On all other backends we could use your parser to determine them - and with the testcase you proved that they are the same as on Windows.
In this way, we were able to programmatically determine the required buffer sizes for read/write operations on all platforms. @Youw What do you think about this?

Youw · 2025-03-28T10:31:54Z

CMakeLists.txt

@@ -70,8 +70,8 @@ if(HIDAPI_ENABLE_ASAN)
    endif()
 endif()

-if(WIN32)
-    # so far only Windows has tests
+if(WIN32 OR HIDAPI_WITH_LIBUSB)


also need to update builds.yml

libusb/test/CMakeLists.txt

Youw · 2025-03-28T10:35:37Z

libusb/test/CMakeLists.txt

+set(CMAKE_VERSION_SUPPORTS_ENVIRONMENT_MODIFICATION "3.22")
+
+foreach(TEST_CASE ${HID_DESCRIPTOR_RECONSTRUCT_TEST_CASES})
+     set(TEST_PP_DATA "${CMAKE_CURRENT_LIST_DIR}/../../windows/test/data/${TEST_CASE}.pp_data")


maybe move the test data to <root>/test_data ?

Youw

I don't have a strong opinion on test implementation (I trust @JoergAtGithub on this one).
libusb implementation seem fine.

Lets make sure it runs with CI on Github Actions and we're good to go here.

Youw · 2025-03-28T10:39:49Z

And I extended the functionality of the libusb method to be able to calculate the maximum output and feature report sizes as well. This has no functional use currently, but it lets us run three times as many tests since the pp_data files have all three max sizes available.

Independent of this PR, I think it would generally make sense to store these 3 values in the device structure. On Windows we would have to use the values InputReportByteLength, OutputReportByteLength and FeatureReportByteLength from https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/hidpi/ns-hidpi-_hidp_caps . On all other backends we could use your parser to determine them - and with the testcase you proved that they are the same as on Windows. In this way, we were able to programmatically determine the required buffer sizes for read/write operations on all platforms. @Youw What do you think about this?

Lets continue here: #731

mcuee · 2025-03-28T14:46:09Z

This PR does not seem to work.

Test device is the same as the one used in Issue #274. The FW is a mod of Jan Axelson's FX2HID example and codes are included in the following #274 comment.

HID reports of more than 64 bytes and libusb backend #274 (comment)

I can reproduce the issue reported in #274 with hidapi git libusb backend, hidraw backend is okay.

mcuee@UbuntuSwift3 ~/build/hid/hidapitester (master)$ sudo ./hidapitester_hidraw_git --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 129-bytes...wrote 129 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
 80
Reading 129-byte input report 0, 250 msec timeout...read 128 bytes:
 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 20
 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F 40
 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F 60
 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F 80
 00
Closing device

mcuee@UbuntuSwift3 ~/build/hid/hidapitester (master)$ sudo ./hidapitester_libusb_git --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 129-bytes...wrote 129 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
 80
Reading 129-byte input report 0, 250 msec timeout...read 64 bytes:
 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 20
 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F 40
 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 00
Closing device

With this PR, no change in hidraw backend behavior, which is good.

mcuee@UbuntuSwift3 ~/build/hid/hidapitester (master)$ sudo ./hidapitester_hidraw_pr728 --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 129-bytes...wrote 129 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
 80
Reading 129-byte input report 0, 250 msec timeout...read 128 bytes:
 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 20
 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F 40
 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F 60
 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F 80
 00
Closing device

But the libusb backend fix is not working.

mcuee@UbuntuSwift3 ~/build/hid/hidapitester (master)$ make -f Makefile_pr728.libusb 
cc -I/usr/local/include/libusb-1.0 -I ../hidapi_pr728/hidapi -c ../hidapi_pr728/libusb/hid.c -o ../hidapi_pr728/libusb/hid.o
cc -I/usr/local/include/libusb-1.0 -I ../hidapi_pr728/hidapi -c hidapitester.c -o hidapitester.o
cc -I/usr/local/include/libusb-1.0 -I ../hidapi_pr728/hidapi ../hidapi_pr728/libusb/hid.o hidapitester.o -o hidapitester -L/usr/local/lib -lusb-1.0 -pthread

mcuee@UbuntuSwift3 ~/build/hid/hidapitester (master)$ mv hidapitester hidapitester_libusb_pr728 

mcuee@UbuntuSwift3 ~/build/hid/hidapitester (master)$ sudo ./hidapitester_libusb_pr728 --vidpid 0925:1234 --open --buflen 256 -l 128 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 128-bytes...wrote 128 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
Reading 128-byte input report 0, 250 msec timeout...read 0 bytes:
Closing device

mcuee · 2025-03-28T14:50:44Z

HID Report Descriptor.

ReportDscr:

	db 06h, 0A0h, 0FFh ;    Usage Page (FFA0h, vendor defined)
	db 09h, 01h     ;       Usage (vendor defined)
	db 0A1h, 01h    ;       Collection (Application)
	db 09h, 02h     ;       Usage (vendor defined)
	db 0A1h, 00h    ;       Collection (Physical)
	db 06h, 0A1h, 0FFh ;    Usage Page (vendor defined)

;; The Input report
	db 09h, 03h     ;       Usage (vendor defined)
	db 09h, 04h     ;       Usage (vendor defined)
	db 15h, 80h	;	Logical minimum (80h or -128)
	db 25h, 7Fh	;	Logical maximum (7Fh or 127)
	db 35h, 00h	;	Physical minimum (0)
	db 45h, 0FFh	;	Physical maximum (255)
	db 75h, 08h	;	Report size (8 bits)
	db 95h, 80h	;	Report count (128 fields)
	db 81h, 02h	;	Input (data, variable, absolute)

;; The Output report
	db 09h, 05h     ;       Usage (vendor defined)
	db 09h, 06h     ;       Usage (vendor defined)
	db 15h, 80h	;	Logical minimum (80h or -128)
	db 25h, 7Fh	;	Logical maximum (7Fh or 127)
	db 35h, 00h	;	Physical minimum (0)
	db 45h, 0FFh	;	Physical maximum (255)
	db 75h, 08h	;	Report size (8 bits)
	db 95h, 80h	;	Report count (128 fields)
	db 91h, 02h	;	Output (data, variable, absolute)

	db 0C0h         ;       End Collection (Physical)
	db 0C0h         ;       End Collection (Application)

ReportDscrEnd:

hidtest is okay.

mcuee@UbuntuSwift3 ~/build/hid/hidapi (master)$ sudo ./hidtest/hidtest-libusb 
hidapi test/example tool. Compiled with hidapi version 0.15.0, runtime version 0.15.0.
Compile-time version matches runtime version of hidapi.

Device Found
  type: 0925 1234
  path: 3-2.1:1.0
  serial_number: (null)
  Manufacturer: CYPRESS
  Product:      EZ-USB FX2 HID USBHIDIO
  Release:      0
  Interface:    0
  Usage (page): 0x0 (0x0)
  Bus type: 1 (USB)

  Report Descriptor: (52 bytes)
0x06, 0xa0, 0xff, 0x09, 0x01, 0xa1, 0x01, 0x09, 0x02, 0xa1, 
0x00, 0x06, 0xa1, 0xff, 0x09, 0x03, 0x09, 0x04, 0x15, 0x80, 
0x25, 0x7f, 0x35, 0x00, 0x45, 0xff, 0x75, 0x08, 0x95, 0x80, 
0x81, 0x02, 0x09, 0x05, 0x09, 0x06, 0x15, 0x80, 0x25, 0x7f, 
0x35, 0x00, 0x45, 0xff, 0x75, 0x08, 0x95, 0x80, 0x91, 0x02, 
0xc0, 0xc0, 
unable to open device

mcuee · 2025-06-16T06:13:26Z

@aokblast

Would you please run your test device against this PR for hidspi libusb backend under your FreeBSD (as well as other OS you have access to)? Thanks.

mcuee · 2025-06-17T00:16:53Z

@cederom and @dl8dtl

Since you two are the FreeBSD power user I know of, please help to check if you can give this PR a try.

FreeBSD HIDAPI is now based on libusb, so it is affected by the Issue #274.

HID reports of more than 64 bytes and libusb backend #274

#274 has an impact on avrdude as well.

cederom · 2025-06-17T02:06:25Z

@mcuee builds okay, could you please provide test steps, its a long thread :-P

mcuee · 2025-06-17T05:31:28Z

@mcuee builds okay, could you please provide test steps, its a long thread :-P

Great.

Test steps: using the normal libusb backend of HIDAPI and this PR.

Test with your normal generic USB HID device using hidapi's built-in hidtest and hidapitester. This should be okay.
https://github.com/todbot/hidapitester
Test with USB Full Speed HID device with >64 Bytes HID report size. This may fail.
Test with USB High Speed HID device with >64 Bytes HID report side. This may fail.

I have mentioned my test device in the comments above (EZ-USB FX2LP) but you can use other test device if you can.

Properly determine libusb read size for large reports (Fixes #274) #728 (comment)

sudobash1 · 2025-06-17T19:20:06Z

@mcuee

FW binary and source codes are here.

Thanks. I finally got my EZ-USB FX2LP dev board from ebay. I've never worked with these before. How do I flash the hex file onto them? This is as far as I got:

stephen@taurus:~$ lsusb | grep FX2
Bus 001 Device 017: ID 04b4:8613 Cypress Semiconductor Corp. CY7C68013 EZ-USB FX2 USB 2.0 Development Kit
stephen@taurus:~$ sudo fxload -D /dev/bus/usb/001/017 -I fx2hid_128bytes/fx2hid.hex -v
microcontroller type: fx
single stage:  load on-chip memory
open RAM hexfile image fx2hid_128bytes/fx2hid.hex
stop CPU
write on-chip, addr 0x0600 len  234 (0x00ea)
write on-chip, addr 0x09e5 len   10 (0x000a)
write on-chip, addr 0x0380 len  428 (0x01ac)
write on-chip, addr 0x0080 len  768 (0x0300)
write on-chip, addr 0x0033 len    3 (0x0003)
write on-chip, addr 0x09ff len    7 (0x0007)
write on-chip, addr 0x07b8 len   92 (0x005c)
write on-chip, addr 0x0851 len   59 (0x003b)
write on-chip, addr 0x05fe len    2 (0x0002)
write on-chip, addr 0x0a07 len    4 (0x0004)
write on-chip, addr 0x09ef len    8 (0x0008)
write on-chip, addr 0x09b0 len   18 (0x0012)
write on-chip, addr 0x09f7 len    8 (0x0008)
write on-chip, addr 0x09c2 len   18 (0x0012)
write on-chip, addr 0x0a0b len    6 (0x0006)
write on-chip, addr 0x091d len   40 (0x0028)
write on-chip, addr 0x096a len   24 (0x0018)
write on-chip, addr 0x06ea len   22 (0x0016)
write on-chip, addr 0x099a len   22 (0x0016)
write on-chip, addr 0x088c len   54 (0x0036)
write on-chip, addr 0x0982 len   24 (0x0018)
write on-chip, addr 0x0814 len   61 (0x003d)
write on-chip, addr 0x0a11 len   36 (0x0024)
write on-chip, addr 0x08f1 len   44 (0x002c)
write on-chip, addr 0x08c2 len   47 (0x002f)
write on-chip, addr 0x0945 len   20 (0x0014)
write on-chip, addr 0x05b8 len   70 (0x0046)
write on-chip, addr 0x0959 len   17 (0x0011)
write on-chip, addr 0x0043 len    3 (0x0003)
write on-chip, addr 0x0053 len    3 (0x0003)
write on-chip, addr 0x0700 len  184 (0x00b8)
write on-chip, addr 0x0000 len    3 (0x0003)
write on-chip, addr 0x052c len   12 (0x000c)
write on-chip, addr 0x09d4 len   17 (0x0011)
write on-chip, addr 0x0538 len  128 (0x0080)
write on-chip, addr 0x0a06 len    1 (0x0001)
... WROTE: 2497 bytes, 36 segments, avg 69
reset CPU

No errors, but nothing seems to happen. No hid device shows up & there is no USB re-enumeration. (Fedora 42, Linux kernel 6.14.11)

cederom · 2025-06-17T19:59:53Z

I also ordered EZ-USB FX2LP will arrive next week will report back test results :-)

mcuee · 2025-06-17T23:25:37Z

@mcuee

How do I flash the hex file onto them? This is as far as I got:

stephen@taurus:~$ lsusb | grep FX2
Bus 001 Device 017: ID 04b4:8613 Cypress Semiconductor Corp. CY7C68013 EZ-USB FX2 USB 2.0 Development Kit
stephen@taurus:~$ sudo fxload -D /dev/bus/usb/001/017 -I fx2hid_128bytes/fx2hid.hex -v
microcontroller type: fx
...
No errors, but nothing seems to happen. No hid device shows up & there is no USB re-enumeration. (Fedora 42, Linux kernel 6.14.11)

You are almost there. Just need to add -t fx2lp in the command line to specify the type to be EZ-USB FX2LP.

FX --> EZ-USB FX
FX2 --> EZ-USB FX2
FX2LP --> EZ-USB FX2LP
FX3 --> EZ-USB FX3

[mcuee@CachyN100 fx2hid_128bytes]$ yay -S yay

[mcuee@CachyN100 fx2hid_128bytes]$ lsusb | grep FX2
Bus 003 Device 011: ID 04b4:8613 Cypress Semiconductor Corp. CY7C68013 EZ-USB FX2 USB 2.0 Development Kit

[mcuee@CachyN100 fx2hid_128bytes]$ fxload
no firmware specified!

Usage: fxload [-v] [-V] [-t type] [-d vid:pid] [-p bus,addr] [-s loader] -i firmware
  -i <path>       -- Firmware to upload
  -s <path>       -- Second stage loader
  -t <type>       -- Target type: an21, fx, fx2, fx2lp, fx3
  -d <vid:pid>    -- Target device, as an USB VID:PID
  -p <bus,addr>   -- Target device, as a libusb bus number and device address path
  -v              -- Increase verbosity
  -q              -- Decrease verbosity (silent mode)
  -V              -- Print program version

[mcuee@CachyN100 fx2hid_128bytes]$ sudo fxload -t fx2lp -i ./fx2hid.hex 
found device 'Cypress EZ-USB FX2LP (68013A/68014A/68015A/68016A)' [04b4:8613] (3,11)
microcontroller type: fx2lp
./fx2hid.hex: type Intel HEX
stop CPU
... WROTE: 2497 bytes, 36 segments, avg 69
reset CPU

[mcuee@CachyN100 fx2hid_128bytes]$ lsusb | grep FX2
Bus 003 Device 014: ID 0925:1234 Lakeview Research EZ-USB FX2 HID USBHIDIO

mcuee · 2025-06-17T23:41:17Z

@sudobash1

Fedora 42's fxload (kind of original version of fxload) is a bit different and it does not support FX3. But it does support FX2LP.

mcuee@fedoran100:~/build/cypress/fx2hid_128bytes$ fxload
no device specified!
usage: fxload [-vV] [-l] [-t type] [-D devpath]
		[-I firmware_hexfile] [-s loader] [-c config_byte]
		[-L link] [-m mode]
... [-D devpath] overrides DEVICE= in env
... device types:  one of an21, fx, fx2, fx2lp
... at least one of -I, -L, -m is required

mcuee@fedoran100:~/build/cypress/fx2hid_128bytes$ lsusb | grep FX2
Bus 003 Device 009: ID 04b4:8613 Cypress Semiconductor Corp. CY7C68013 EZ-USB FX2 USB 2.0 Development Kit

mcuee@fedoran100:~/build/cypress/fx2hid_128bytes$ sudo fxload -D /dev/bus/usb/003/009 -t fx2lp -I ./fx2hid.hex -v
microcontroller type: fx2lp
single stage:  load on-chip memory
open RAM hexfile image ./fx2hid.hex
stop CPU
write on-chip, addr 0x0600 len  234 (0x00ea)
write on-chip, addr 0x09e5 len   10 (0x000a)
write on-chip, addr 0x0380 len  428 (0x01ac)
write on-chip, addr 0x0080 len  768 (0x0300)
write on-chip, addr 0x0033 len    3 (0x0003)
write on-chip, addr 0x09ff len    7 (0x0007)
write on-chip, addr 0x07b8 len   92 (0x005c)
write on-chip, addr 0x0851 len   59 (0x003b)
write on-chip, addr 0x05fe len    2 (0x0002)
write on-chip, addr 0x0a07 len    4 (0x0004)
write on-chip, addr 0x09ef len    8 (0x0008)
write on-chip, addr 0x09b0 len   18 (0x0012)
write on-chip, addr 0x09f7 len    8 (0x0008)
write on-chip, addr 0x09c2 len   18 (0x0012)
write on-chip, addr 0x0a0b len    6 (0x0006)
write on-chip, addr 0x091d len   40 (0x0028)
write on-chip, addr 0x096a len   24 (0x0018)
write on-chip, addr 0x06ea len   22 (0x0016)
write on-chip, addr 0x099a len   22 (0x0016)
write on-chip, addr 0x088c len   54 (0x0036)
write on-chip, addr 0x0982 len   24 (0x0018)
write on-chip, addr 0x0814 len   61 (0x003d)
write on-chip, addr 0x0a11 len   36 (0x0024)
write on-chip, addr 0x08f1 len   44 (0x002c)
write on-chip, addr 0x08c2 len   47 (0x002f)
write on-chip, addr 0x0945 len   20 (0x0014)
write on-chip, addr 0x05b8 len   70 (0x0046)
write on-chip, addr 0x0959 len   17 (0x0011)
write on-chip, addr 0x0043 len    3 (0x0003)
write on-chip, addr 0x0053 len    3 (0x0003)
write on-chip, addr 0x0700 len  184 (0x00b8)
write on-chip, addr 0x0000 len    3 (0x0003)
write on-chip, addr 0x052c len   12 (0x000c)
write on-chip, addr 0x09d4 len   17 (0x0011)
write on-chip, addr 0x0538 len  128 (0x0080)
write on-chip, addr 0x0a06 len    1 (0x0001)
... WROTE: 2497 bytes, 36 segments, avg 69
reset CPU

mcuee@fedoran100:~/build/cypress/fx2hid_128bytes$ lsusb | grep FX2
Bus 003 Device 010: ID 0925:1234 Lakeview Research EZ-USB FX2 HID USBHIDIO

mcuee · 2025-06-17T23:43:44Z

Interestingly Arch Linux AUR fxload is actually packing libusb project's fxload example.
https://github.com/libusb/libusb/tree/master/examples

https://aur.archlinux.org/packages/fxload
Git Clone URL: https://aur.archlinux.org/fxload.git (read-only, click to copy)
Package Base: fxload
Description: fxload firmware loader. Used with udev or devfs/hotplug.
Upstream URL: http://libusb.info/

mcuee · 2025-06-17T23:58:37Z

Interestingly there is a Fedora request to package libusb's fxload example (probably as libusb-fxload) but it is not implemented in the end.
https://bugzilla.redhat.com/show_bug.cgi?id=1907078

Take note libusb's fxload does not support EEPROM. So the suggeston from @jwrdegoede is good in the above discussion. libusb's fxload example cannot replace the original fxload, it is better to have both packages in Fedora.

Initial commit in libusb project
libusb/libusb@0597533

FX3 support:
libusb/libusb@b74b7f7

sudobash1 · 2025-06-18T00:18:36Z

@mcuee Thanks, I have it running now. I am experiencing the same things you are. I added a quick printf to libusb/hid.c to check the dev->max_input_report_size. It reports 129:

stephen@taurus:~/src/hidapitester/hidapitester$ git -C ../hidapi-fork/ diff
diff --git i/libusb/hid.c w/libusb/hid.c
index aee5782..05a3370 100644
--- i/libusb/hid.c
+++ w/libusb/hid.c
@@ -1313,6 +1313,7 @@ static int hidapi_initialize_device(hid_device *dev, const struct libusb_interfa
        int desc_size = hid_get_report_descriptor_libusb(dev->device_handle, dev->interface, dev->report_descriptor_size, report_descriptor, sizeof(report>
        if (desc_size > 0) {
                dev->max_input_report_size = get_max_report_size(report_descriptor, desc_size, REPORT_DESCR_INPUT);
+               printf("computed max input report size (rounded): %zu\n", dev->max_input_report_size);
        } else {
                dev->max_input_report_size = 0;
        }
stephen@taurus:~/src/hidapitester/hidapitester$ sudo ./hidapitester --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input | grep "computed"
computed max input report size (rounded): 129

So, this PR is computing what I would have expected for the max input report size from the device descriptor. But I notice with hidraw, it is only reading 128 bytes (not 129). So as a quick kludge test, I removed the + 1 from the report size, and it works:

stephen@taurus:~/src/hidapitester/hidapitester$ git -C ../hidapi-fork/ diff
diff --git i/libusb/hid.c w/libusb/hid.c
index aee5782..7d28b17 100644
--- i/libusb/hid.c
+++ w/libusb/hid.c
@@ -356,7 +356,7 @@ static size_t get_max_report_size(uint8_t * report_descriptor, int desc_size, en
        } else {
                /* report_size is in bits. Determine the total size convert to bytes
                (rounded up), and add one byte for the report number. */
-               return ((max_size + 7) / 8) + 1;
+               return ((max_size + 7) / 8);
        }
 }
 
stephen@taurus:~/src/hidapitester/hidapitester$ sudo ./hidapitester --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 129-bytes...wrote 129 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
 80
Reading up to 129-byte input report, 250 msec timeout...read 128 bytes:
 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 20
 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F 40
 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F 60
 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F 80
 00
Closing device

It looks like the input report doesn't use a report number, so there are only 128 bytes sent. Right now, I'm guessing that the issue is something to do with the fact that 128 is an even multiple of the packet size, so USB isn't sending any short packet.

I've got a cold right now, so my head has been a bit fuzzy. I will have to revisit this at a later time with a clearer head.

tl;dr

It looks like the code in this PR to parse the report descriptor is working, but libusb isn't reading a 128 byte message when it expected a 129-byte one.

Anybody have more insight into this?

Cheers! 🍻

JoergAtGithub · 2025-06-18T02:00:57Z

For interupt transfers, which are send without requests from the device to the computer, the ReportId byte ist only prepend to the report data, if the device uses ReportIDs.
Therefore you need to determine from the report descriptor, if the device uses ReportIDs. If anywhere in the report descriptor REPORT_ID occurs, you need, to increase all transfer sizes by one byte for the ReportID.

mcuee · 2025-06-18T02:38:02Z

@sudobash1

I think you are close to find the root cause of the issue.

https://github.com/sudobash1/hidapi/blob/libusb_input_size/libusb/hid.c

	if (max_size == 0) {
		// No matching reports found
		return 0;
	} else {
		/* report_size is in bits. Determine the total size convert to bytes
		(rounded up), and add one byte for the report number. */
		return ((max_size + 7) / 8) + 1;
	}

It is kind of a complicated topic. Please refer to the following discussion.

hidapi documentation about number of bytes written and read #589

Good comments in that issue by @JoergAtGithub

hidapi documentation about number of bytes written and read #589 (comment)

0 is a placeholder and not a Report ID. This is because the HID class specification specifies "If a Report ID tag was used in the Report descriptor, all reports include a single byte ID prefix. If the Report ID tag was not used, all values are returned in a single report and a prefix ID is not included in that report." and further about the representation in the ReportDescriptor "Set Report ID to 0 (zero) if Report IDs are not used."

Therefore only values from 1 to 255 can be Report IDs. The value 0 can be used in the ReportDescriptor, to define that the reports do not contain a ReportID.

Windows implements it as specified in the class specification and adds the Report ID prefix byte only for devices that use Report IDs.

I know that we need to keep backward compatibility, but from a clean API I would expect a clear seperation between address and data in two arguments.

You can also refer to my comment.

hidapi documentation about number of bytes written and read #589 (comment)

That's exactly my point. Having report ID at the beginning of the buffer - means not extra buffer "prepending" on HIDAPI side.

There are two aspects of the issue.

1. Buffer handling -- platform specific.

I think the current codes already do that. For example, Windows codes have already done that, at least for Output Report and Feature Report (first byte of the report is the report ID, 0 when there is no report ID). And I believe all platforms are doing that for Feature report, based on my testing results.

I tend to believe the current codes are already fine in this aspect, i.e, we do not have issues in terms of getting thngs done (the device will receive the right data, host will also receive the right data).

2. Read/Write lentgh reporting

I think this is the issue now. It is not consistent.

Two ways of reporting -- now we are mixing the two. a) lengh of the buffer used by underlying OS driver -- Feature Report (including the 0 byte for device without Report ID). Currently I believe there is no behaviour differences for Feature Report across platforms.

b) length of the data bytes appearing on the USB Bus -- Input Report (not including the 0 byte for device without Report ID). For Windows, this could be due to the way we handle Input Report differently from Output Report and Feature report. Currently I believe there is no behaviour differences for Input Report across platforms.

c) mix of the two ways -- Output Report (Windows always including the 0 byte for device without Report ID, other platforms may or may not including). So Windows may report one byte more than the other platforms or may report the same length as the other platforms. I think we should at least fix this one.

3. Potential Solution for Issue 2. a) If we really want to unify all platforms in terms of the read/write length report mechanism, then I tend to htink Option b is the way to go for all three types of reports. But then it may require some jobs across all platforms and for Output and Feature reports..

b) If we just want to unify the behavior across platforms, then we only need to fix Output Report write length reporting to the user. I tend to think we only need to make Windows consistent with others. So only one platform (Windows) and one report type (Output Report) needs to be fixed.

sudobash1 · 2025-06-19T00:50:54Z

Thank you @JoergAtGithub and @mcuee. That is clearly the issue. I reviewed the HID spec again and indeed it says:

If a Report ID tag is used anywhere in Report descriptor, all data reports for the device are preceded by a single byte ID field

The Windows behavior when there is no ReportID was tripping me up.

I've pushed a simple commit to address this, and the FX2LP 128-byte report test works fine for me now.

The tests also needed to be fixed for the cases when there is no ReportID (ReportId = 0x00 in the pp_data file).

Am I forgetting anything, or does that finish everything outstanding on this PR (provided CI completes and @mcuee's & @cederom's tests pass)?

cederom · 2025-06-19T00:55:55Z

@sudobash1 great news! I will be able to provide soonest feedback around Monday sorry if you are ready before and want to go ahead then go ahead :-) I am waiting for the board and will have to see how to set things up but then I will be able to help you in other cases :-)

Is there any sort of bad-usb device out there that could provide us various test cases from hardware point of view (i.e. simulate bad descriptors, strange transfers, etc)?

mcuee · 2025-06-19T06:55:31Z

@sudobash1

Other than testing, @Youw will need to review the codes and then approve this PR.

But now the first thing is to fix the CI build -- two builds failed because of the same issue.

[ 37%] Building C object src/libusb/CMakeFiles/hidapi_libusb.dir/hid.c.o
/__w/hidapi/hidapi/hidapisrc/libusb/hid.c: In function ‘get_max_report_size’:
/__w/hidapi/hidapi/hidapisrc/libusb/hid.c:333:29: error: comparison of integer expressions of different signedness: ‘int’ and ‘enum report_descr_type’ [-Werror=sign-compare]
  333 |                 if (key_cmd == report_type) { /* Input / Output / Feature */
      |                             ^~
cc1: all warnings being treated as errors
make[2]: *** [src/libusb/CMakeFiles/hidapi_libusb.dir/build.make:79: src/libusb/CMakeFiles/hidapi_libusb.dir/hid.c.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:206: src/libusb/CMakeFiles/hidapi_libusb.dir/all] Error 2
make: *** [Makefile:[13](https://github.com/libusb/hidapi/actions/runs/15746712859/job/44397056740?pr=728#step:7:14)6: all] Error 2

sudobash1 · 2025-06-19T20:24:51Z

But now the first thing is to fix the CI build -- two builds failed because of the same issue.

Oops. I thought I had fixed that. I guess not. I have pushed e2391b6 to address compiler warnings. It compiles cleanly with -Wall -Wextra on gcc-15 and clang-20 on my Fedora box now. Hopefully all the CI compile now.

mcuee · 2025-06-20T02:08:08Z

Oops. I thought I had fixed that. I guess not. I have pushed e2391b6 to address compiler warnings. It compiles cleanly with -Wall -Wextra on gcc-15 and clang-20 on my Fedora box now. Hopefully all the CI compile now.

Yes, CI builds are okay now.

sudobash1 · 2025-06-20T06:04:21Z

Ah, it looks like there a few comments from @Youw that I need to address too. Most significant is probably updating builds.yml (to make sure the tests are enabled).

I'll look through this on Friday.

JoergAtGithub · 2025-06-20T06:30:04Z

It would be good if it could clarify in the comments what exactly the “report size” returned by get_max_report_size comprises.

Here it is only about the InputReports that are sent unrequested by the HID device via interupt transfers (hid_read / hid_read_timeout).
However, InputReports can also be actively retrieved with hid_get_input_report via ControlTransfers, where the ReportId byte is always included, even if the device does not use ReportIDs (in this case the ReportID-byte contains the value 0).
And get_max_report_size also supports feature and output reports, where it's different too.

In other code of mine, I introduced the unofficial term payload-size for the report size without the ReportID-byte. There I add 1 depending on the context where it is used. There I've also a boolean that indicates, if the device uses ReportIDs.

mcuee · 2025-06-23T00:06:08Z

@sudobash1

First test under Windows -- it seems the reported write/read length is not consistent with native API, one byte higher.

Native Windows HID backend-- wrote 129 bytes, read 128 bytes
libusb backend -- wrote 130 bytes, read 129 bytes

But this is getting close. I will test under Linux and FreeBSD later.

 UCRT64 /c/work/libusb/hidapitester
$ ./hidapitester_git_winapi.exe --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 129-bytes...wrote 129 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
 80
Reading up to 129-byte input report, 250 msec timeout...read 128 bytes:
 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 20
 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F 40
 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F 60
 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F 80
 00
Closing device

$ ./hidapitester_pr728_libusb_new.exe --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 129-bytes...wrote 130 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
 80
Reading up to 129-byte input report, 250 msec timeout...read 129 bytes:
 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 20
 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F 40
 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F 60
 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F 80
 00
Closing device

$ ./hidapitester_pr728_winapi_new.exe --vidpid 0925:1234 --open --buflen 256 -l 129 --send-output 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37
,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,10
3,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128  --read-input
Opening device, vid/pid: 0x0925/0x1234
Writing output report of 129-bytes...wrote 129 bytes:
 00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F
 20 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F
 40 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F
 60 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F
 80
Reading up to 129-byte input report, 250 msec timeout...read 128 bytes:
 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F 10 11 12 13 14 15 16 17 18 19 1A 1B 1C 1D 1E 1F 20
 21 22 23 24 25 26 27 28 29 2A 2B 2C 2D 2E 2F 30 31 32 33 34 35 36 37 38 39 3A 3B 3C 3D 3E 3F 40
 41 42 43 44 45 46 47 48 49 4A 4B 4C 4D 4E 4F 50 51 52 53 54 55 56 57 58 59 5A 5B 5C 5D 5E 5F 60
 61 62 63 64 65 66 67 68 69 6A 6B 6C 6D 6E 6F 70 71 72 73 74 75 76 77 78 79 7A 7B 7C 7D 7E 7F 80
 00
Closing device

Dirty fix to be able to build under Windows.

PS C:\work\libusb\hidapi_pr728_new> git diff
diff --git a/libusb/hid.c b/libusb/hid.c
index 0f06389..c830479 100644
--- a/libusb/hid.c
+++ b/libusb/hid.c
@@ -34,8 +34,8 @@
 #include <unistd.h>
 #include <sys/types.h>
 #include <sys/stat.h>
-#include <sys/ioctl.h>
-#include <sys/utsname.h>
+///#include <sys/ioctl.h>
+//#include <sys/utsname.h>
 #include <fcntl.h>
 #include <wchar.h>

hidapitester Makefile hack and binaries (against hidapi git and this PR, Windows native backend and libusb backend).

hidapitester_win.zip

Co-authored-by: Ihor Dutchak <[email protected]>

sudobash1 mentioned this pull request Mar 25, 2025

HID reports of more than 64 bytes and libusb backend #274

Open

sudobash1 force-pushed the libusb_input_size branch from 068fd62 to 52c6cd9 Compare March 25, 2025 22:06

mcuee added bug Something isn't working libusb Related to libusb backend labels Mar 26, 2025

Youw reviewed Mar 26, 2025

View reviewed changes

libusb/hid.c Outdated Show resolved Hide resolved

Properly determine libusb read size for large reports

fe758f9

Fixes libusb#274

JoergAtGithub reviewed Mar 26, 2025

View reviewed changes

mcuee mentioned this pull request Mar 27, 2025

pp_data_dump issue for 046A_0011_0006_0001 -- Lower Priority #729

Closed

Fixes to report size calculation and added tests

8949182

sudobash1 force-pushed the libusb_input_size branch from 52c6cd9 to 2708264 Compare March 27, 2025 19:36

Allow computation of max report size for other types

3710895

sudobash1 force-pushed the libusb_input_size branch from 2708264 to 3710895 Compare March 27, 2025 19:42

Don't assign -1 to size_t. That's silly

470c9e9

JoergAtGithub suggested changes Mar 27, 2025

View reviewed changes

Youw reviewed Mar 28, 2025

View reviewed changes

libusb/test/CMakeLists.txt Outdated Show resolved Hide resolved

Youw reviewed Mar 28, 2025

View reviewed changes

Youw mentioned this pull request Mar 28, 2025

Store Input/Output/Feature reports sizes for all platforms #731

Open

sudobash1 force-pushed the libusb_input_size branch from f56c2d7 to 40132e9 Compare June 19, 2025 00:46

mcuee requested a review from JoergAtGithub June 19, 2025 06:44

Stephen and others added 2 commits June 19, 2025 12:23

Do not increase max_report_size by one if not using report numbers

ca3e4a2

Fix Apline CI build (libusb#754)

97ce43c

sudobash1 force-pushed the libusb_input_size branch from c06470e to 44328f8 Compare June 19, 2025 19:24

Fix compiler warnings

e2391b6

sudobash1 force-pushed the libusb_input_size branch from 44328f8 to e2391b6 Compare June 19, 2025 20:16

Merge branch 'master' into libusb_input_size

2a1ba37

Remove comment

a4c4da9

Co-authored-by: Ihor Dutchak <[email protected]>

	pp_data->caps_info[0]->FirstCap = 0
	pp_data->caps_info[0]->LastCap = 2
	pp_data->caps_info[0]->NumberOfCaps = 2
	pp_data->caps_info[0]->ReportByteLength = 9

	pp_data->cap[0]->BitSize = 1
	pp_data->cap[0]->ReportCount = 8

	pp_data->cap[1]->BitSize = 8
	pp_data->cap[1]->ReportCount = 6

Properly determine libusb read size for large reports (Fixes #274) #728

Are you sure you want to change the base?

Properly determine libusb read size for large reports (Fixes #274) #728

Uh oh!

Conversation

sudobash1 commented Mar 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sudobash1 commented Mar 25, 2025

Uh oh!

Be-ing commented Mar 26, 2025

Uh oh!

Youw left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mcuee commented Mar 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

JoergAtGithub commented Mar 27, 2025

Uh oh!

JoergAtGithub commented Mar 27, 2025

Uh oh!

mcuee commented Mar 27, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

JoergAtGithub Mar 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sudobash1 commented Mar 27, 2025

Uh oh!

JoergAtGithub commented Mar 27, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Youw left a comment

Choose a reason for hiding this comment

Uh oh!

Youw commented Mar 28, 2025

Uh oh!

mcuee commented Mar 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mcuee commented Mar 28, 2025

Uh oh!

mcuee commented Jun 16, 2025

Uh oh!

mcuee commented Jun 17, 2025

Uh oh!

cederom commented Jun 17, 2025

Uh oh!

sudobash1 commented Mar 25, 2025 •

edited

Loading

mcuee commented Mar 27, 2025 •

edited

Loading

JoergAtGithub Mar 27, 2025 •

edited

Loading

mcuee commented Mar 28, 2025 •

edited

Loading

mcuee commented Jun 17, 2025 •

edited

Loading

mcuee commented Jun 17, 2025 •

edited

Loading

cederom commented Jun 19, 2025 •

edited

Loading

mcuee commented Jun 23, 2025 •

edited

Loading