Pi 4: mouse interface of a Logitech Bolt receiver gets no interrupt service while the receiver's other interfaces are all polled (VL805, 4 x full-speed bInterval 1 endpoints)

Aperta
#7,760 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
5/5
Tempo stimato
Più di una settimana
Idoneità per principianti
8/100
Tipo di issue
Bug
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Stack tecnologico
c

Direzione di ricerca

Reproduction needs this exact hardware; begin with the usbmon capture described in the issue to confirm that 0x82 completes nothing while a URB is pending on 0x83. The scheduling at fault lives in the xHCI host driver behind the VL805, so read drivers/usb/host/xhci-ring.c and xhci-mem.c for how full-speed bInterval 1 interrupt endpoints are packed into the periodic frame list, and compare with the usbhid side that submits the URBs. Done means endpoint 0x82 keeps completing at roughly 125/s while interface 2's hidraw node is held open, shown the same way, with interface 3 still serviced.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Describe the bug

Disclosure: this issue was found, investigated and written by an AI assistant (Claude, made by Anthropic) working on my behalf. I reviewed it and am posting it myself. Every measurement below comes from real hardware.

A Logitech Logi Bolt receiver (046d:c548) has four HID interfaces, each with one interrupt IN endpoint, all 64 bytes at bInterval 1 (full speed):

Interface Endpoint Purpose Driver
0 0x81 boot keyboard usbhid / hid-generic
1 0x82 boot mouse usbhid / hid-generic
2 0x83 vendor (HID++ reports 0x10/0x11) usbhid / hid-generic
3 0x84 touchpad/multitouch usbhid / hid-multitouch

On a Raspberry Pi 4, as soon as interface 2's interrupt endpoint is polled (for example by opening its hidraw node, which is what tools such as Solaar or logiops do), the mouse on interface 1 stops moving completely. The moment the hidraw node is closed, the mouse works again. Nothing appears in dmesg, and the device is not reset.

usbmon shows that while an URB is pending on 0x83, endpoint 0x82 gets no completions at all, even though the mouse is moving. When the 0x83 URB is cancelled (hidraw closed), 0x82 resumes immediately.

If interface 3 is disabled (echo 0 > /sys/bus/usb/devices/<port>:1.3/authorized), endpoint 0x82 keeps working normally with 0x83 polled. So three of the receiver's four 1 ms interrupt endpoints are serviced, and the fourth makes one of the others starve.

The same receiver and mouse on an x86 PC (kernel 7.3.4, receiver also bound to hid-generic) keep working with the hidraw node open, so this looks specific to the Pi 4's USB controller or its periodic scheduling.

Possibly related: #5201 (mouse freezes on a Pi 4 while a USB audio input is captured), which may be the same limit reached with a different combination of periodic endpoints.

Steps to reproduce the behaviour
  1. Plug a Logi Bolt receiver with a paired mouse (here an MX Master 3S) into a Pi 4, either directly or through a USB 2.0 hub (both behave the same).
  2. Make sure interfaces 0, 1 and 3 are being polled. In my case a userspace program (libinput) has their evdev nodes open, so usbhid keeps their interrupt URBs submitted.
  3. Move the mouse and open interface 2's hidraw node. Find it with grep -l 046D:0000C548 /sys/class/hidraw/*/device/uevent (it's the node whose report descriptor contains report IDs 0x10 and 0x11). Then hold it open for 15 seconds:
    python3 -c 'import os,select,time; fd=os.open("/dev/hidrawN", os.O_RDWR|os.O_NONBLOCK); t=time.monotonic()+15
    while time.monotonic()<t: select.select([fd],[],[],0.2)
    os.close(fd)'
    
  4. Observe: the pointer freezes for those 15 seconds and moves again as soon as the script exits. Nothing needs to be written to the node.
  5. Disable interface 3 (echo 0 | sudo tee /sys/bus/usb/devices/<port>:1.3/authorized) and repeat step 3: the mouse keeps moving.

usbmon capture (bus 1, receiver address only, mouse moved throughout). Counts are interrupt completions on endpoint 2 (mouse) per second:

Second All four interfaces active Interface 3 disabled
0-3 (hidraw closed) 11, 62, 27, 4 125, 131, 127, 130
4 (hidraw opened, URB submitted on ep 3) 0 128
5-15 (hidraw open) 0 every second 89-131 every second
16-17 (hidraw closed, ep 3 URB completes with -2) 35, 51 72, 91

Things that did not help:

  • Moving the receiver from a USB 2.0 hub (Genesys 05e3:0610) to one of the Pi's own ports (behind the VL805's internal hub 2109:3431).
  • Updating kernel, firmware and the bootloader/VL805 EEPROM (first seen on 6.18.34+rpt with the 2026-05-17 bootloader, still there on the versions below).
  • usbhid.mousepoll=2 set at runtime, then rebinding interface 1 (no change; the 0x82 endpoint still gets no service while 0x83 is polled).

Workaround: a udev rule that disables the unused interface 3:

ACTION=="add", SUBSYSTEM=="usb", ENV{DEVTYPE}=="usb_interface", ATTRS{idVendor}=="046d", ATTRS{idProduct}=="c548", ATTR{bInterfaceNumber}=="03", ATTR{authorized}="0"
Device (s)

Raspberry Pi 4 Mod. B

System
  • Raspberry Pi 4 Model B Rev 1.5
  • OS: Raspberry Pi reference 2026-06-18 (pi-gen ca8aeed0ae300c2a89f55ce9617d5f96a27e99e5, stage2), Debian GNU/Linux 13 (trixie)
  • Firmware (vcgencmd version): Sep 14 2026 14:49:30, version a86983925695a7e63166327d7c002d64040ed31d (clean) (release) (start)
  • Bootloader 2026-10-30 (1793402435), VL805 firmware 000138c0
  • Kernel (uname -a): Linux 6.18.60+rpt-rpi-v8 #1 SMP PREEMPT Debian 1:6.18.60-1+rpt1 (2026-11-03) aarch64 GNU/Linux
  • Receiver: 046d:c548 "USB Receiver", bcdUSB 2.00, bcdDevice 5.03, full speed
  • Other USB devices on the same bus: a full-speed USB audio headset with capture active (isochronous IN), another 2.4 GHz keyboard receiver, both behind the USB 2.0 hub. I have not tested with the headset unplugged.
Logs

Nothing relevant in dmesg when the freeze starts or ends. The usbmon counts are above. The full capture of the receiver's address is available if useful.

Additional context
  • The hid-logitech-dj driver keeps interface 2 open permanently on receivers it supports, so any Bolt support in that driver (it was added during 7.3 and reverted) would likely freeze the mouse on a Pi 4 the same way unless interface 3 is idle.
  • I found this while making a Logitech mouse's extra buttons usable on a Pi 4 that acts as a shared keyboard/mouse server.
Lingua principale
C
Stelle
13.3k
Fork
5.5k
Merge medio
4g 10h
PR unite (30g)
31

Preparare l'ambiente

Questo progetto non fornisce container di sviluppo, Dockerfile né guida per i contributori, quindi l'ambiente è a tuo carico: parti dal suo README e consulta la nostra guida al primo contributo per i passaggi generali.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di raspberrypi/linux

Tutte le issue di raspberrypi/linux

Issue simili

Altre issue su C

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.