rob thijssen 2457df7a37
All checks were successful
build image / build (push) Successful in 53m10s
docs: record what the GPU firmware trace actually proves
Tested on the machine. With the generic Adreno firmware reachable from the
initramfs and the model-signed blob absent, the GPU gets two files in and stops:

    loaded qcom/a630_sqe.fw from new location
    loaded qcom/a630_gmu.bin from new location
    *ERROR* Unable to load qcom/sdm850/LENOVO/81JL/qcdxkmsuc850.mbn
    *ERROR* gpu hw init failed: -2

That settles a question the docs had been hedging: the zap shader is a hard
requirement, not an optimisation, and the generic pieces are necessary but not
sufficient. Getting past sqe and gmu only to die at zap is about as clean a
demonstration as the hardware will give.

Two things fall out of the same trace and are worth writing down, because
nothing upstream documents them for this machine. The driver never falls back
to the installed sdm845/a630_zap.mbn — the device tree pins the model-specific
path — and a fallback would not help anyway, since the shader is signed against
this device's own secure-boot chain. And the display survives the failure
entirely: msm_dpu still binds the panel, so a dead GPU costs acceleration
rather than a picture, and a desktop runs on llvmpipe.

Also record why these files have to travel in the initramfs at all: msm_dpu
probes around six seconds in, while the initramfs is still root, and never
retries.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XWRjNJMistCy6ngXH5aJLS
2026-07-27 20:05:17 +03:00

Fedora for the Lenovo Yoga C630 13Q50

Builds a ready-to-write Fedora aarch64 disk image for the Lenovo Yoga C630 (model 81JL, Qualcomm SDM850), using Gitea Actions.

The approach is borrowed from aarch64-laptops/build, but the heavy lifting that project had to do in 2019 is now unnecessary: mainline Linux has carried sdm850-lenovo-yoga-c630.dts since 5.5, and Fedora ships it in kernel-core. There is no kernel to patch and no GRUB to compile. What is left is assembling a disk image that boots on hardware whose firmware hands Linux no device tree.

What you get

output/fedora-44-<variant>-lenovo-yoga-c630-<date>-<ref>.img.zst — a GPT disk image with an ESP, a /boot partition and an ext4 root. Decompress, write it to a USB stick or microSD card, and boot. The root filesystem grows to fill the medium on first boot.

Two variants:

Variant Size Contents
minimal 8 GiB Console, sshd, and enough tools to debug the machine
workstation 16 GiB GNOME desktop

Default login is fedora / fedora, and the password must be changed at first login. Root is locked.

Hardware status

Works out of the box Needs firmware from Windows Not supported
UFS storage, USB, keyboard, touchpad, touchscreen Graphics (Adreno 630 zap shader) LTE modem
WiFi + Bluetooth (ath10k WCN3990) Audio
Battery and charging Sensor hub — lid switch, accelerometer, auto-rotate
Display (unaccelerated) Hardware video decode (venus)

Fedora ships everything Qualcomm permits to be redistributed, which covers WiFi and Bluetooth outright. Graphics is the awkward case: the generic Adreno pieces (a630_gmu.bin, a630_sqe.fw) are packaged, but the C630's device tree asks for a model-signed zap shader, qcdxkmsuc850.mbn, which exists only in your machine's Windows partition — Fedora's generic sdm845/a630_zap.mbn is not what this device requests. Audio, sensors and video decode are the same story.

Run sudo c630-firmware once after installing — see docs/firmware.md for the full list and a manual fallback.

Building

CI does this on every push to main. To run it yourself:

./build/build-image.sh --variant minimal

You need podman and, on an x86_64 host, aarch64 emulation:

sudo dnf install -y qemu-user-static-aarch64
sudo systemctl restart systemd-binfmt

build/build-image.sh runs on the host and only sets up the container. build/stage2.sh runs inside an aarch64 Fedora container and does everything else: dnf --installroot, the overlay, dracut, and the disk assembly. It builds filesystems from directory trees with mke2fs -d and mcopy rather than mounting loop devices, so it does not need /dev/loop-control — which CI runners generally will not hand out.

Iterating

Every aarch64 binary runs under emulation, and the dnf transaction is essentially all of the cost — 500-odd packages' worth of rpm scriptlets for minimal, four times that for workstation. Everything after it takes minutes. Since getting this machine to boot will take a few attempts, the build is arranged so you only pay that once:

  • The post-dnf root filesystem is staged under <work>/base, keyed on a hash of the package lists, release and variant. Editing config/device.env, overlay/, or the bootloader config reuses it. Editing config/packages/ invalidates it automatically.
  • The working copy is made with cp --reflink=auto, so on btrfs or xfs it is a copy-on-write clone rather than a real copy.
  • Downloaded rpms live in .cache/dnf, outside the staged tree, so even --fresh re-runs the scriptlets without re-downloading.

In practice a kernel-command-line change rebuilds in a few minutes.

./build/build-image.sh --variant minimal            # reuses the staged base
./build/build-image.sh --variant minimal --fresh    # forces a reinstall
./build/build-image.sh --variant minimal --keep-rootfs   # keep the tree to poke at

CI points --work and --cache at /var/tmp/c630-build so both survive between jobs. That is per-runner, so the first build on a given runner is cold.

The build container defaults to git.lair.cafe/gongfoo/build-fedora-44-aarch64, which ships the assembly tooling so it does not have to be installed under emulation on every run. If that image is not reachable the build falls back to stock Fedora and installs the tooling itself — slower, but it works.

Repository layout

config/device.env            C630 parameters: DTB path, kernel command line, geometry
config/packages/*.pkgs       Package lists — base plus one file per variant
overlay/                     Files copied into the root filesystem (*.in are templated)
build/build-image.sh         Host driver: emulation checks, podman invocation
build/stage2.sh              The actual build, inside an aarch64 container
firmware/local/              Optional drop-in for firmware you extracted yourself (gitignored)
.gitea/workflows/            CI
docs/                        Installation, firmware, runner setup

Documentation

Boot status

Confirmed on hardware, from a USB stick:

  • GRUB loads the kernel and the device tree
  • The kernel comes up and probes the SDM850 — UFS controller, display subsystem, WiFi and IPA all appear as platform devices, so DEVICE_CMDLINE and the DTB are right
  • The framebuffer console works (simple-framebuffer, 240x67)
  • The root filesystem mounts, systemd starts, and it reaches a login prompt
  • Networking works over a USB WiFi dongle, and sshd is reachable
  • build/install-to-disk.sh copies it onto the internal UFS, and it boots from there with the stick removed — 18s to multi-user, no failed units

The RTC cannot be set on this machine and reads 1970 at every boot, so chrony is installed and corrects the clock once the network is up. Until it does, rpm rejects package signatures dated after the (wrong) current time and reports installed packages as missing — see docs/install.md.

Onboard WiFi does not appear at all, and audio, sensors, video decode and accelerated graphics are all absent — every one of them waiting on the model-signed firmware described in docs/firmware.md. On this machine Windows has been wiped, so those blobs are gone; the kernel names each missing file explicitly in dmesg.

The internal drive needs build/install-to-disk.sh rather than dd: its UFS uses 4096-byte logical sectors, which the 512-byte image geometry cannot be written onto directly. See docs/install.md.

SELinux is shipped permissive — see below.

SELinux

The build labels the filesystem offline with setfiles, so enforcing should work. It ships permissive anyway, because the failure mode is unusually punishing: an unlabelled root takes PID 1 down with Failed to allocate manager object before anything can be logged into and repaired, and /.autorelabel cannot save it because nothing survives long enough to act on the flag.

Once the machine is up and you are happy with it:

sudo setenforce 1                 # try it for this boot
sudo sed -i 's/^SELINUX=.*/SELINUX=enforcing/' /etc/selinux/config

Or set SELINUX_MODE=enforcing in config/device.env and rebuild.

Caveats

The kernel command line and firmware layout are taken from Fedora's Snapdragon WoA documentation and the aarch64-laptops project; the parts specific to the C630's older SDM850 are reasoned from those rather than derived from documentation for this machine. Findings belong in this README.

Description
No description provided
Readme 399 KiB
Languages
Shell 100%