usb_hid: honor a boot protocol request that arrives after startup

[mikeysklar]

What

send_report() now drops the report ID while the host has the interface in boot protocol.

Why

I went looking for old issues. #1136 is eight years old, 69 comments.

The bug	usb_hid_setup_devices() swaps in the boot device, whose report ID is 0, but runs only from usb_setup_with_vm() at VM start
Effect	A SET_PROTOCOL(boot) arriving while code.py runs is missed until the next restart, so reports keep the prefix while the host reads 8-byte boot reports
Symptom	The 0x01 lands in the modifier byte, a held left Ctrl. Matches bitboy85's get_boot_device() returning 1 alongside a phantom Ctrl

Hardware tested

Metro RP2040 on Linux. Not tested: real BIOS, GRUB, KVM, Windows, non-RP2 ports, zephyr-cp.

How I tested it

Two devices, so report IDs stay in use:

usb_hid.enable((usb_hid.Device.KEYBOARD, usb_hid.Device.CONSUMER_CONTROL), boot_device=1)

code.py sent 'a' once per second, read from USB DPRAM over SWD.

'a' press report	Baseline d897c15f24	This branch
Report protocol, before	01 00 00 04 ..., 9 bytes	01 00 00 04 ..., 9 bytes
Boot protocol, after	unchanged, 9 bytes	00 00 04 ..., 8 bytes
Default config, no boot.py	IDs 1/2/3	byte-identical to baseline

Scope

Item	Status
The host request	Simulated by setting TinyUSB's protocol_mode over SWD, not sent by a real host
BIOS, GRUB, KVM	Not tested, none available
macOS pre-boot	Unrelated. An M2 Startup Options screen behaves the same with and without this change, so the macOS reports in #1136 have another cause
zephyr-cp	Has the same defect, unchanged here. No available board enables its USB device stack. Happy to mirror it
Keyboard LED reports	Still dropped in boot protocol with multiple devices. IN direction only
Dropped from this PR	Report-ID-less descriptor variants. The hand-rolled descriptors in the thread would mask this timing bug too, so they are not evidence that a host skipped SET_PROTOCOL

AI assistance

Written with an LLM agent (Claude). I ran the boards myself.

Prompt	Root-cause #1136 from the thread and source, fix it, verify on hardware with controls
Numbers above	From the SWD instrumentation, not the model

[dhalbert]

I worked on this a lot years ago, as you can tell. I am surprised you need a new report descriptor without the Report ID slot. Part of the point of boot keyboard and boot mouse is that the descriptor sent by the device is ignored. Instead the host assumes the standard descriptors, as described in https://www.usb.org/sites/default/files/hid1_12.pdf in the boot devices sections

So is this a work-around for something that Macs are not doing properly in their "BIOS"? I think the original code solved the issue on PC's.

[mikeysklar]

You're right about the descriptor. The bug is timing: the swap runs only at VM start, so a later SET_PROTOCOL is missed. Not Mac specific. I'll cut it to that.