Chrome Version: 
OS: all ?

What steps will reproduce the problem?
(1) Attach LG UltraFine 4K Display
(2) In a Chrome.extension, open its HID device (available on its USB interface with VID:PID 043e:9a40)
(3) read 'maxFeatureReportSize' in HidDeviceInfo

What is the expected result?

maxFeatureReportSize is equal to 6 and the screen answers 6-byte (including the current brightness in the 4 first) when using  chrome.hid.receiveFeatureReport(connectionId, 0, callback)

What happens instead?

maxFeatureReportSize is equal to 2 and the USB transfer triggered by chrome.hid.receiveFeatureReport() fails.

The chrome://device-log shows "Failed to get feature report: Value too large for defined data type (75)"
Meaning that in Chrome, the ioctl(fd_.get(), HIDIOCGFEATURE(buffer->size()), buffer->data()) returns EOVERFLOW.
On this chromebook, looking at the kernel, this likely means that the actual USB transfer fails due to a 'babble' of the USB device.
so, it might be because we ask for 2 bytes and the screen tries to return more (which is not compliant with USB spec too ...).


internal discussion is on https://goo.gl/oc11wc,
summary here:
The HID report descriptor of the screen is the following:
0x05, 0x80,        // Usage Page (Monitor Pages)
0x09, 0x01,        // Usage (0x01)
0xA1, 0x01,        // Collection (Application)
0x06, 0x82, 0x00,  //   Usage Page (Monitor Pages)
0x09, 0x10,        //   Usage (0x10)
0x15, 0x04,        //   Logical Minimum (4)
0x26, 0x1C, 0x02,  //   Logical Maximum (540)
0x67, 0xE1, 0x00, 0x00, 0x01,  //   Unit (System: SI Linear, Luminous Intensity: Candela)
0x55, 0x0E,        //   Unit Exponent (-2)
0x75, 0x20,        //   Report Size (32)
0x95, 0x01,        //   Report Count (1)
0xB1, 0x42,        //   Feature (Data,Var,Abs,No Wrap,Linear,Preferred State,Null State,Non-volatile)
0x06, 0x82, 0x00,  //   Usage Page (Monitor Pages)
0x09, 0x10,        //   Usage (0x10)
0x67, 0xE1, 0x00, 0x00, 0x01,  //   Unit (System: SI Linear, Luminous Intensity: Candela)
0x55, 0x0E,        //   Unit Exponent (-2)
0x75, 0x20,        //   Report Size (32)
0x95, 0x01,        //   Report Count (1)
0x81, 0x02,        //   Input (Data,Var,Abs,No Wrap,Linear,Preferred State,No Null Position)
0x05, 0x0F,        //   Usage Page (PID Page)
0x09, 0x50,        //   Usage (0x50)
0x66, 0x10, 0x01,  //   Unit (Length: Centimeter, Mass: Gram)
0x55, 0x0D,        //   Unit Exponent (-3)
0x75, 0x10,        //   Report Size (16)
0xB1, 0x42,        //   Feature (Data,Var,Abs,No Wrap,Linear,Preferred State,Null State,Non-volatile)
0xC0,              // End Collection

When I run it through device/hid/hid_report_descriptor_unittest.cc,
I'm getting: 
actual_max_input_report_size == 0
actual_max_output_report_size == 0
actual_max_feature_report_size == 2
while I would have expected (by reading the report descriptor):
actual_max_input_report_size == 4
actual_max_output_report_size == 0
actual_max_feature_report_size == 6

When tracing the unittest execution, it's getting lost on 0x67, 0xE1, 0x00, 0x00, 0x01, 
The tag starting with 0x67 is a 'short item' for the HID spec (with bTag == 6 aka Unit, bType == 1 aka Global, bSize == 3)
But the trick is that according to the HID 1.1 spec: bSize == 3 means that *4* bytes are following (see picture of the spec attached)

Trying the following fix:
--- a/device/hid/hid_report_descriptor_item.cc
+++ b/device/hid/hid_report_descriptor_item.cc
@@ -37,7 +37,7 @@ HidReportDescriptorItem::HidReportDescriptorItem(
     if (size >= 2)
       payload_size_ = bytes[1];
   } else {
-    payload_size_ = header->size;
+    payload_size_ = header->size == 0x3 ? 4 : header->size;
     DCHECK(payload_size_ <= sizeof(shortData_));
     if (GetHeaderSize() + payload_size() <= size)
       memcpy(&shortData_, &bytes[GetHeaderSize()], payload_size());

I'm getting the expected:
actual_max_input_report_size == 4
actual_max_output_report_size == 0
actual_max_feature_report_size == 6

 

Deleted: HID_short_items.png 44.4 KB

	HID_short_items.png 44.4 KB View Download