Are we ARM yet?

HP Chromebook x2 11

Snapdragon 7c · 2021-01 · SC7180

Google's board name for it is CoachZ.

Booting with Omarchy

Bootable
Not yet
Bootloader
Unlockable
Device tree
Not carried
Linux image
None yet

A Chromebook starts through ChromeOS's bootloader, not through UEFI. Omarchy's install image doesn't start on it, so it needs no ACPI tables or hardware IDs, and getting past the bootloader comes first.

Upstream, for every distro

Linux
Linux 5.13

0 of 29 checks are written up for this model on Omarchy, 0 of them working.

What works

Click a check to see what it means, how it's checked and every result.

Notes are what people reported on Omarchy, each checked against its source. They can be old.

Tested is what hw-test found on people's machines, with the date.

Loading test results…

Starting and installing

USB stick and installerNot tested yet, and no notes.Untested

Works means: The install image starts from a USB stick by itself and reaches the installer, with keyboard and touchpad working.

The person testing checks: Boot the current install image from a USB stick with Secure Boot off, and say whether the installer appeared and whether the keyboard and touchpad worked in it.

When it fails, look at: The installer image picks the device tree by the machine's hardware IDs. A black screen right after the boot menu usually means no tree matched. Compare `systemd-analyze chid` with the IDs on the model's page. A keyboard or touchpad that is dead only in the installer means the tree lacks it or its driver isn't in the image. Fixed in `linux-omarchy-dtbs` (tree and hardware IDs).

Disk password promptNot tested yet, and no notes.Untested

Works means: The disk password prompt shows on the built-in screen at startup and the built-in keyboard types into it.

The test measures: On a laptop with an encrypted disk, when the built-in screen and keyboard came up at this boot against when the disk opened, from the kernel's log.

The person testing checks: Restart and say whether the password prompt was visible and whether the built-in keyboard worked in it.

When it fails, look at: The prompt runs from the initramfs, before most drivers load. A prompt that can't be typed into means part of the keyboard's driver chain isn't in the initramfs (i2c-hid-of, the I2C controller, pin control, regulators, or the USB controller for a USB keyboard). No visible prompt means the display's chain isn't. Compare `lsinitcpio` on the image with `lsmod` on the running system.

Screen and input

Built-in screenNot tested yet, and no notes.Untested

Works means: The built-in screen shows the desktop at its native resolution.

The test measures: The built-in panel's connector, its state and its preferred mode.

The person testing checks: Say whether the desktop fills the screen sharply, and whether anything flickers.

When it fails, look at: The built-in panel hangs off the eDP controller (`mdss_dp3` on Snapdragon X), with the panel under its `aux-bus`. Read `journalctl -k -b | grep -iE 'msm|dpu|edp|panel'` and `cat /sys/class/drm/card*-eDP-1/{status,modes}`. A panel the kernel doesn't know gets cautious timings from `panel-edp` and a warning that names its EDID id. A wrong resolution or a black screen after the boot logo points at the tree's panel node, its supply or its enable GPIO.

Graphics accelerationNot tested yet, and no notes.Untested

Works means: The desktop runs on the GPU, not on software rendering.

The test measures: On a Snapdragon, whether a render node from the msm driver exists and the GPU's firmware loaded. On an NVIDIA machine, whether a driver is bound to its GPU. The graphics test asks Vulkan which device programs draw on.

When it fails, look at: The GPU needs the msm kernel driver, Mesa, and three firmware files, which are the SQE and GMU firmware from linux-firmware and a zap shader signed by the computer's maker, named by `firmware-name` under the tree's `gpu` node. `journalctl -k -b | grep -iE 'adreno|a7xx|a8xx|gmu|zap'` shows which is missing. Without the zap shader the desktop runs on software rendering, and `vulkaninfo --summary` shows llvmpipe.

WebGL in the browserNot tested yet, and no notes.Untested

Works means: The default browser gives pages WebGL on the GPU, not on a renderer that runs on the processor.

The test measures: The graphics test opens a page in the default browser without a window and reads the renderer its WebGL context reports. The result names the browser. Chromium stands in when the default is not Chrome, Chromium or one built on it.

When it fails, look at: Each browser on Omarchy starts with switches of its own, Chrome with `~/.config/chrome-flags.conf` and Chromium with `~/.config/chromium-flags.conf`. The result names the browser that was tested, so change that one's file. Its `chrome://gpu` page says what it uses and why a feature is off. WebGL on SwiftShader means the browser found no usable GPU, so fix the `gpu` check first.

WebGPU in the browserNot tested yet, and no notes.Untested

Works means: The default browser gives pages WebGPU on the GPU without being started in a special way.

The test measures: The graphics test asks the default browser for a WebGPU adapter and device, as it comes and again with the switches that turn WebGPU on. Working only with the switches counts as partly. The result names the browser.

When it fails, look at: Chrome and Chromium on Linux don't turn WebGPU on by themselves on most ARM GPUs. Vulkan has to work first, which `vulkaninfo --summary` shows by naming the GPU and not llvmpipe. The result names the browser that was tested, and each reads its own file, Chrome `~/.config/chrome-flags.conf` and Chromium `~/.config/chromium-flags.conf`. The test's first run starts the browser with that file, so switches set there count as working. A switch set for one browser does nothing for the other. To try it for Chromium on this machine, keep a copy of its file, then `omarchy-chromium-add-feature Vulkan` and `omarchy-chromium-add-feature WebGPU` (they add to the `--enable-features` line that is there, where a second such line would replace it), and put `--enable-unsafe-webgpu` on a line of its own. A fix for everyone is a per-user setting in omacom/omarchy (the guide's table says where). Whether Omarchy wants that switch on by default is the team's call, so say in the pull request that it is a proposal.

Screen brightnessNot tested yet, and no notes.Untested

Works means: The brightness keys change the screen's brightness over its whole range.

The test measures: Whether a backlight device exists, and its range.

The person testing checks: Press the brightness keys and say whether the screen gets darker and brighter.

When it fails, look at: `ls /sys/class/backlight` should list one device. OLED panels and some LCDs are dimmed over the eDP link, where `panel-edp` registers the backlight. Others take a PWM from the PMIC (`pwm-backlight` with an enable GPIO). Brightness that reads back but changes nothing usually means the wrong PWM channel or a missing enable GPIO in the tree.

Built-in keyboardNot tested yet, and no notes.Untested

Works means: Every key types, including the function row.

The test measures: Whether a built-in keyboard device exists.

The person testing checks: Type on it and try the function row (volume, brightness, mute).

When it fails, look at: Most of these keyboards are HID over I2C, which the tree describes as a node under one of the tree's I2C buses with `compatible = "hid-over-i2c"`, the descriptor address, an interrupt GPIO and its supply. `cat /sys/bus/i2c/devices/*/name` and `journalctl -k -b | grep -i i2c_hid` show whether it was found. Bus, address and interrupt come from the device's `_CRS` in the ACPI DSDT. Some models put the keyboard on an internal USB port or behind the embedded controller instead.

TouchpadNot tested yet, and no notes.Untested

Works means: Pointer, tap or click, two-finger scrolling and gestures work.

The test measures: Whether a built-in touchpad device exists.

The person testing checks: Move, click, scroll with two fingers and try a three-finger swipe.

When it fails, look at: Like the keyboard it is usually HID over I2C, described in the tree with bus, address, interrupt GPIO and supply taken from the DSDT. A touchpad that is found but doesn't move the pointer often has the wrong interrupt pin or polarity. Its line in `/proc/interrupts` should count up while a finger moves.

Wireless

Wi-FiNot tested yet, and no notes.Untested

Works means: Wi-Fi connects and stays connected at full speed, also after sleep.

The test measures: The Wi-Fi interface, its driver and whether it's connected. The wireless test reads the link's signal and speed.

The person testing checks: Say whether it stays connected through a day of use and after waking from sleep.

When it fails, look at: The card is a WCN7850 (ath12k) or WCN6855 (ath11k) on PCIe. `journalctl -k -b | grep -iE 'ath1[12]k|pci'` tells how far it gets. No PCI device means the tree's PCIe port, its PHY or the card's power sequencer is wrong. 'failed to fetch board data' means linux-firmware's `board-2.bin` has no entry for this machine, and the line names the board ID that is missing. `rfkill list` rules out a soft block.

Fixed Wi-Fi addressNot tested yet, and no notes.Untested

Works means: The Wi-Fi card keeps the same address on every boot, so networks and routers recognise the laptop.

The test measures: Whether the kernel calls the interface's address permanent or randomly made.

When it fails, look at: `cat /sys/class/net/wlan0/addr_assign_type` reads 1 for a made-up address. See whether the address changes over restarts. NetworkManager and iwd can also randomise the address by setting, which isn't a hardware fault. Never report the address itself.

BluetoothNot tested yet, and no notes.Untested

Works means: Bluetooth turns on, finds devices and keeps headphones and mice connected.

The test measures: Whether a Bluetooth controller exists and is powered. The wireless test listens for eight seconds and counts the devices it hears.

The person testing checks: Pair a pair of headphones or a mouse and use it for a few minutes.

When it fails, look at: The Bluetooth half of the Wi-Fi chip sits on a UART, which the tree describes as a `bluetooth` node under one of its serial ports with its supplies. `journalctl -k -b | grep -iE 'bluetooth|hci|qca'` shows its firmware loading from `qca/`, and some machines need an NVM file of their own. A controller without an address shows as unconfigured in `btmgmt info`. `rfkill list` rules out a soft block.

Sound

SpeakersNot tested yet, and no notes.Untested

Works means: Every built-in speaker plays, at a normal volume, without crackling.

The test measures: Whether a sound card with a playback device exists, besides HDMI and USB ones.

The person testing checks: Play music and say whether both sides (and the tweeters, if it has them) play cleanly.

When it fails, look at: Sound runs through the audio DSP. Four things have to hold, in this order. The ADSP must be running (see the `adsp` check), the sound card must exist (`cat /proc/asound/cards`), the topology file the card asks for must be under `/lib/firmware/qcom/` (the journal names it when it's missing), and alsa-ucm-conf needs a profile for the card (`alsaucm listcards`, `wpctl status`). The speaker amplifiers (WSA883x or WSA884x on SoundWire) need their reset GPIOs and supplies in the tree. One silent side is usually one amplifier's GPIO or the channel mapping. `hw-test speakers` measures each side with the microphones. `ls /sys/bus/soundwire/devices/` shows whether every amplifier is there. `dtc -I fs -O dts /proc/device-tree 2>/dev/null | grep -A9 'speaker@'` shows each one's reset GPIO in the running tree, and `tr '\0' '\n' < /proc/device-tree/sound/audio-routing | paste - -` the output each speaker is fed from. On a side with a woofer and a tweeter, the two should come from different outputs (`WSA_SPK1` and `WSA_SPK2`). A quiet or silent side with every amplifier present is usually this routing, not a GPIO. Never raise amplifier or speaker gains above what the sound profile sets. The kernel has no speaker protection on these computers, and a speaker can be damaged.

Built-in microphonesNot tested yet, and no notes.Untested

Works means: The built-in microphones record clearly.

The test measures: Whether a capture device exists on the built-in sound card.

The person testing checks: Record a few seconds and play it back.

When it fails, look at: The built-in microphones are digital and need their supply and pins in the tree and a UCM profile that routes them. `wpctl status` should list an internal microphone source. Silence from a source that exists is usually the microphone supply or the wrong microphone pair in the profile. `hw-test speakers` records known tones and reports the level per pitch.

Headphone jackNot tested yet, and no notes.Untested

Works means: Headphones play when plugged in, and the speakers go quiet.

The person testing checks: Plug in wired headphones and play something.

When it fails, look at: The jack runs through a WCD938x or WCD939x codec on SoundWire, which also detects the plug. The UCM profile switches the output when a plug is seen, and `wpctl status` then shows the headphones port as available. No switch means jack detection (the codec's interrupt or its settings in the tree). Sound that doesn't come after a switch means the codec's supplies or the profile's routing.

Audio and battery processorNot tested yet, and no notes.Untested

Works means: The audio DSP starts, which sound, the battery readings and USB-C power need.

The test measures: The state of the adsp remote processor and the firmware it loaded.

When it fails, look at: `cat /sys/class/remoteproc/remoteproc*/{name,state,firmware}` shows whether it runs and which file it wants. The file is signed by the computer's maker and comes from the computer's own Windows partition or from its firmware package in omarchy-pkgs. Battery readings, charging, USB-C and sound all depend on this processor.

Battery, sleep and power

Battery readingsNot tested yet, and no notes.Untested

Works means: The battery's charge, state and time left show correctly.

The test measures: Whether a battery reports its charge and state.

When it fails, look at: Battery readings come from the ADSP (`qcom_battmgr` over pmic_glink). No battery under `/sys/class/power_supply` means the ADSP isn't running (see the `adsp` check) or the tree lacks the `pmic-glink` node.

USB-C chargingNot tested yet, and no notes.Untested

Works means: The laptop charges at full power from its own charger, not only at 5 V.

The test measures: The voltage and online state the USB-C power supply reports.

The person testing checks: Plug in the laptop's own charger and say whether the charge rises at the usual speed.

When it fails, look at: USB-C charging is negotiated by the ADSP's firmware and reported through UCSI. Read `cat /sys/class/power_supply/*/{status,online}` and `journalctl -k -b | grep -iE 'ucsi|pmic_glink|battmgr'`. Charging that works with the computer off and not under Linux points at the ADSP not running.

Sleep and wakeNot tested yet, and no notes.Untested

Works means: Closing the lid sleeps, opening it wakes to the lock screen, every time.

The test measures: The sleep mode the kernel uses and how many sleeps succeeded or failed since boot.

The person testing checks: Close the lid for a few minutes, five times in a row, and say whether it woke each time.

When it fails, look at: `cat /sys/power/mem_sleep` shows the sleep mode in brackets, s2idle or deep. After a sleep, `journalctl -k -b | grep -E 'PM:|suspend'` shows which device refused, and the files under `/sys/power/suspend_stats/` count failures and name the last failing device. A restart in place of a wake leaves no line after 'suspend entry' at the end of `journalctl -k -b -1`. Try the other sleep mode to tell a firmware problem from a driver's. `hw-test sleep` measures a lid-close sleep with the owner. For a freeze on wake, check before the long press on power whether the Caps Lock light still toggles and whether the computer answers a ping from another machine, which tells a frozen desktop from a frozen kernel. Record `grep -H . /sys/bus/usb/devices/*/{product,speed}` and `cat /sys/class/drm/card*-*/status` with the same things plugged in. `journalctl --list-boots` lists the boots, and `hw-test note --log boot-N` attaches the kernel's log of the one that froze.

Battery use while asleepNot tested yet, and no notes.Untested

Works means: A night asleep costs only a few percent of the battery.

The person testing checks: Note the charge before closing the lid for the night and after opening it.

When it fails, look at: Compare `cat /sys/class/power_supply/*/energy_now` before and after an hour asleep. High drain in s2idle means the chip never reached its lowest state, and a device kept awake is the usual cause.

HibernationNot tested yet, and no notes.Untested

Works means: The laptop hibernates and resumes where it was.

The test measures: Whether the kernel offers hibernation and has swap to hold it.

The person testing checks: Hibernate from the menu and start it again.

When it fails, look at: `cat /sys/power/state` must list `disk`, and swap must be at least the size of memory. Where `disk` is missing the kernel offers no hibernation on this chip. That can't be fixed from the machine, so tell the owner and go on to the next part.

Ports and external screens

USB portsNot tested yet, and no notes.Untested

Works means: Every USB-A and USB-C port takes devices at full speed, in both plug orientations.

The test measures: Which USB controllers exist and the fastest speed each offers.

The person testing checks: Plug a USB stick into each port, both ways round on USB-C.

When it fails, look at: Each USB-C port is a controller, a combo PHY and a connector under `pmic-glink` in the tree. USB-A ports hang off the multiport controller or a USB 2 controller, often behind a supply that has to be switched on. `grep -H . /sys/bus/usb/devices/*/speed` shows the speed of each port and device (`lsusb -t` does too, where usbutils is installed) and `journalctl -k -b | grep -iE 'dwc3|xhci|typec|ucsi'` the errors. Full speed in only one orientation of the plug means the port's orientation switch or retimer isn't described. `hw-test usb` reads the speed per port as things are plugged in.

External screenNot tested yet, and no notes.Untested

Works means: A screen on USB-C (or HDMI, if it has one) shows a picture and can be plugged in and out.

The test measures: The external display connectors and whether any is connected.

The person testing checks: Plug in a screen through USB-C, unplug it and plug it in again.

When it fails, look at: DisplayPort over USB-C needs the port's DP controller, the combo PHY, the orientation switch and on many machines a retimer (Parade PS883x) in the tree. `cat /sys/class/drm/card*-DP-*/status` and `journalctl -k -b | grep -iE 'typec|altmode|ps883|msm_dp'` show how far it gets. An HDMI port is a DisplayPort output behind a converter chip with a node and supplies of its own. `hw-test screen` reads the connection and mode when a screen is plugged in.

USB4 and ThunderboltNot tested yet, and no notes.Untested

Works means: USB4 docks and drives run at USB4 speed.

The test measures: Whether the kernel created a USB4 domain.

The person testing checks: Plug in a USB4 or Thunderbolt dock or drive and say what speed it runs at.

When it fails, look at: USB4 and Thunderbolt tunnelling don't work on these chips in Omarchy's kernel yet. Docks work as far as USB 3 and DisplayPort over USB-C carry them, and an empty `/sys/bus/thunderbolt/devices` is expected.

Camera, video and AI

WebcamNot tested yet, and no notes.Untested

Works means: The webcam gives a picture in the browser and in video calls.

The test measures: Whether a camera device from the camera subsystem or USB exists.

The person testing checks: Open a video call test page in Chromium and say whether the picture looks right.

When it fails, look at: The webcam is a MIPI sensor, which the tree describes as a sensor node on a camera I2C bus with its clock, reset GPIO and supplies, linked to the camera subsystem in the tree. `cam -l` from libcamera names the sensor, and `journalctl -k -b | grep -iE 'camss|cci|ov0|imx|hi5'` shows one that didn't answer. `hw-test camera`, run by the owner, takes one frame, and the check works when they answer that the picture is real and its colours look right. A picture that is too dark, pale, green or noisy comes from libcamera's processing for that sensor (its tuning file), not from the driver. Omarchy has no place for a tuning fix yet, so don't open a pull request for one. Write a `workaround` note with what you changed, and open an issue on omacom/omarchy.

Hardware video decodingNot tested yet, and no notes.Untested

Works means: Videos decode on the video engine or the GPU, not on the processor.

The test measures: On a Snapdragon, whether the iris or venus video decoder exists. The video test decodes a two-second H.264 clip and counts the frames, on that decoder, or on an NVIDIA machine on its GPU through NVDEC or VA-API.

The person testing checks: Play a 1080p or 4K video in mpv or Chromium and say whether the processor stays mostly idle.

When it fails, look at: The decoder is the `iris` driver (venus on older chips), with firmware from `qcom/vpu/` or the file the tree's `firmware-name` asks for. `journalctl -k -b | grep -iE 'iris|venus|vpu'` shows a missing file and `v4l2-ctl --list-devices` the decoder. `hw-test video` decodes a clip on it and counts the frames.

Compute DSP (NPU)Not tested yet, and no notes.Untested

Works means: The compute DSP starts with its firmware.

The test measures: The state of the cdsp remote processor and the firmware it loaded.

When it fails, look at: As for the `adsp` check, `cat /sys/class/remoteproc/remoteproc*/{name,state,firmware}` shows its state and the maker-signed file it wants. Nothing else depends on it yet.

Sensors and other hardware

Lid switchNot tested yet, and no notes.Untested

Works means: Closing and opening the lid is noticed, and nothing wakes it by itself.

The test measures: Whether an input device reports the lid switch.

When it fails, look at: The lid is a `gpio-keys` switch in the tree with `linux,code = <SW_LID>` on the hall sensor's GPIO, marked `wakeup-source`. `journalctl -b -u systemd-logind | grep Lid` shows its history. No switch at all means the node is missing. A switch that reads the wrong way round has the wrong GPIO polarity. `hw-test lid` watches it close and open.

Fans and temperaturesNot tested yet, and no notes.Untested

Works means: The temperatures can be read and follow the load, and the fan speeds up when the computer gets hot.

The test measures: Which temperature and fan sensors exist. The temperature test keeps every core and the GPU busy for up to two minutes and reads whether the sensors show the heat. It sees a fan start by its speed, or where no sensor gives that, by the temperature falling while the processor works on at the same rate.

The person testing checks: Press f during the temperature test the moment you hear the fan speed up, or say that the computer has none.

When it fails, look at: Temperatures come from the chip's own sensors and need nothing per model. Fans are run by the embedded controller by itself, and their speed can be read only where the kernel has a driver for that model's controller (`ls /sys/class/hwmon/*/fan*_input`). A missing fan speed is a missing kernel driver, not a broken fan, and writing that driver is beyond a fix from the machine, so tell the owner. `hw-test temperatures` heats the computer and reports how the sensors follow. With the owner there, they press f when they hear the fan, which is what turns this check to works on a machine without a speed sensor.

Own one? Test it

Omarchy's install image doesn't start on Chromebooks and Googlebooks, so there is no installer for this one yet. How to help it get there. Once Omarchy runs on it by other means, the test below works as on any other machine.

On Omarchy, install the hardware test once and run it. It measures what it can by itself (graphics, Wi-Fi and Bluetooth, sound, video decoding, temperatures), tests the rest with you, shows the results and puts them on this page unless you say no. No serial numbers, addresses or names are sent, and a picture or recording it makes never leaves the computer.

omarchy-mise-install github:birkskyum/arewearmyet hw-test
hw-test

Something doesn't work? hw-test fix hands it to your AI agent, with what was measured, where to look and where a fix goes. Any agent can also start from arewearmyet.com/agents.

Device trees

  • sc7180-trogdor-coachz-r1.dts · google,coachz-rev1
  • sc7180-trogdor-coachz-r1.dts · google,coachz-rev2
  • sc7180-trogdor-coachz-r1-lte.dts · google,coachz-rev1-sku0
  • sc7180-trogdor-coachz-r1-lte.dts · google,coachz-rev2-sku0
  • sc7180-trogdor-coachz-r3.dts · google,coachz
  • sc7180-trogdor-coachz-r3-lte.dts · google,coachz-sku0

Firmware its tree loads

  • qcom/sc7180-trogdor/modem-nolte/mba.mbn
  • qcom/sc7180-trogdor/modem-nolte/qdsp6sw.mbn
  • qcom/sc7180-trogdor/modem/mba.mbn
  • qcom/sc7180-trogdor/modem/qdsp6sw.mbn

Machine-readable: /api/models/hp-chromebook-x2-11. Something wrong or missing? Edit the findings or open an issue.