Simulating an upgrade — running kernel-core's %posttrans command by hand —
turned up two faults that had nothing to do with the full /boot, and would
have broken every future kernel install on their own.
95-c630-devicetree trusted $KERNEL_INSTALL_BOOT_ROOT. `kernel-install inspect`
reports "Layout: other, Boot Root: /boot/efi" here, because the ESP is all it
recognises, while Fedora's 20-grub.install writes BLS entries to
/boot/loader/entries regardless. So the hook looked for the device tree on the
ESP, did not find it, and skipped the patch — leaving an entry with no
devicetree line, which on this machine cannot boot. Paths in a BLS entry are
relative to the filesystem holding the entry, so the only correct boot root is
the one carrying both loader/entries and dtb-<kver>. Search for that instead.
dracut-config-rescue sets dracut_rescue_image=yes, so each install also built
a second ~210 MiB initramfs, which is what refilled /boot straight after it
was cleared. That image cannot boot this machine either: the rescue entry is
<machine-id>-0-rescue.conf, carrying no kernel version, so the devicetree hook
never matches it. Turned off.
Verified by rerunning the simulation: both kernels now have a devicetree line
and a present initrd, no rescue image is generated, snd-soc-wsa881x is rebuilt
automatically, and /boot holds at 72%.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011XgGF5wfxLDAybVnNz6eNQ
Rebuilding the initramfs on the machine rather than in the build container
pulled qcom_q6v5_pas into it. There it probes while the initramfs is still the
root filesystem, asks for qcadsp850.mbn, gets -2, and never retries — the same
race 10-c630.conf already documents for the Adreno firmware. The ADSP hosts
the whole audio path, so the machine comes up with no sound card and a sound
device deferring on "error getting cpu dai name", which points nowhere near
firmware. The blobs cannot go in the initramfs to fix it: they are ~20 MiB and
are extracted from Windows, so they need not exist at build time. Omitting the
driver instead means it loads once the real root is there. The modem driver
goes with it, since it is not needed early either.
Separately, the WCD9340 decides what is plugged in by measuring impedance, so
an amplifier on the headphone socket reads as an open circuit and the jack
control never leaves 'off'. PipeWire marks the port not available, will not
keep it as the default sink, and GNOME hides it — an output that works when
driven directly becomes unreachable from the desktop. Dropping JackControl
from the Headphones device leaves availability unknown rather than no, and the
port becomes permanently selectable.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011XgGF5wfxLDAybVnNz6eNQ
A dnf upgrade to 7.1.8 installed the kernel rpm cleanly, then ran /boot out
of space. dracut wrote no initramfs, and because kernel-install stops at the
first failing plugin, 95-c630-devicetree never ran either — leaving a boot
entry with neither an initrd nor a devicetree line, which on this machine can
never boot. dnf reported success and nothing retried.
A kernel costs ~336 MiB here: a 210 MiB hostonly=no initramfs, a 98 MiB
dtb-<kver> directory carrying every board's device tree, plus vmlinuz and
System.map. Three of those cannot fit 1 GiB, so /boot goes to 2 GiB and
installonly_limit drops to 2.
Also fixes the quieter half of the same trap. snd-soc-wsa881x lives outside
the kernel package, so the speakers go silent after any kernel update with
nothing in the logs to explain it. c630-wsa881x rebuilds it and
96-c630-wsa881x.install calls it on each kernel-install add — always exiting
0, since a plugin failure is precisely what caused the damage above.
grub.cfg gains the next_entry one-shot block that has been carried by hand all
along, so testing a kernel costs a power cycle rather than a rescue.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011XgGF5wfxLDAybVnNz6eNQ
Fedora runs rmtfs with -r, which starts the modem remoteproc at boot. Since
the modem never gets past RF init, every C630 sits in a fatal crash and
recovery cycle every ~3.4 seconds from boot onwards — and is already a dozen
crashes deep by the time anyone logs in, which quietly confuses any manual
experiment. Drop the -r so the modem stays offline until asked for.
Also record today's eliminations, each of which closes an avenue: both
published _nm firmware builds are rejected by the boot ROM, not just the
newest; the PEP tables in this machine's own ACPI show Windows gives the
modem seven clocks, one 752 mV rail vote, cx/mx and IPA bandwidth, and
nothing whatsoever for RF — so the host does not power the RF front end on
either OS, and the one clock mainline lacks changes nothing when forced on;
and kernel bracketing is stuck, because 5.10 will not boot here at all.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011XgGF5wfxLDAybVnNz6eNQ
The ADSP front-end advertises S24_LE and then garbles it: playback runs
without an error, consumes data at rate, powers every DAPM widget, and
produces ~-48 dB of near-silence. PipeWire prefers the widest advertised
format, so every PipeWire client was inaudible while aplay and
speaker-test, which default to S16_LE, worked. One wireplumber rule pins
the format until the q6asm handling is fixed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011XgGF5wfxLDAybVnNz6eNQ
The WCN3990 has no burned-in Bluetooth address, so the kernel registers
the controller unconfigured and bluetoothd never sees it. The factory
address is on the machine all along: QCOM/BT.PROVISION on the DPP
partition, a three-byte header followed by the six-byte address.
c630-bt-addr reads it and hands it to btmgmt before bluetooth.service
starts. Verified on hardware: controller configures, powers, scans, and
streams A2DP.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011XgGF5wfxLDAybVnNz6eNQ
Two problems, one of them mine twice over.
The motd is gone entirely. It had been wrong in a way that mattered — telling
anyone who logged in that onboard WiFi worked when it does not — and a login
banner is a bad place to keep claims that need revising every time we learn
something.
The worse issue is what it and the docs said about why the modem fails. I
inferred from the WOA-Project repository being named "Qualcomm-Reference-
Drivers" that its modem image was a generic reference build unsuited to this
machine's radio, and wrote that up as the explanation. That was a guess resting
on a repository name, and the naming is about redistribution rather than the
contents. Removed from both README and docs/firmware.md.
What is actually established stays: the modem boots, loads mpss, and dies at
"RF stuck in QLINK start state", roughly every 42 seconds. The cause is not
known. The machine has a SIM slot and its modem worked under Windows, so the
radio hardware is present — and the same cabinets supply the GPU, DSP and venus
firmware, all of which work, so whatever is wrong is specific to the modem.
The docs now say that and stop there.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XWRjNJMistCy6ngXH5aJLS
The motd was the worst of it — it told anyone logging in that "WiFi and
Bluetooth work with the firmware Fedora ships", which is precisely backwards for
the onboard adapter, and pointed at a Windows partition this machine no longer
has.
What the blobs did fix, verified: the GPU initialises, ADSP/CDSP/SLPI all run,
sensors appear as IIO devices, venus registers /dev/video0-3.
What they did not, and why, so nobody repeats the search:
Onboard WiFi is loaded by the modem — wlan*mdsp* is the WLAN Modem DSP — so
ath10k_snoc binds and then waits silently on a QMI service that never arrives.
The modem boots and dies at "RF stuck in QLINK start state", the RF front-end,
which is board-specific wiring in a way the SoC-level blobs are not. These are
Qualcomm *reference* drivers; good enough for the GPU and DSPs, evidently not
for this machine's radio. Three plausible fixes that do not work are recorded
too: rmtfs (correct to enable, does not help), the _nm firmware variants
(rejected at PBL), and unplugging the USB dongle (different bus entirely — it
was never competing).
Audio gets further than expected. The codec answers over SLIMbus with its chip
id, so hardware and ADSP are fine; what fails is the codec's SoundWire block
declaring two data-in ports where the controller expects six. That is a kernel
mismatch and needs a kernel change.
Also record the mesa caveat properly: the GPU works, but freedreno's EGL path
segfaults gnome-shell into a login loop, and LIBGL_ALWAYS_SOFTWARE does nothing
because mutter uses EGL. MESA_LOADER_DRIVER_OVERRIDE=kms_swrast is what works.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XWRjNJMistCy6ngXH5aJLS
Validated by running it on the machine, twice, and inspecting the result.
Two properties of this laptop rule out the obvious approach, and the script
exists mainly to encode them. Its internal UFS reports 4096-byte logical
sectors, so the image — built with 512-byte geometry — cannot be dd'd onto it;
the GPT header and every partition offset would land in the wrong place. And
there are no EFI runtime variables, so efibootmgr cannot register a boot entry
and GRUB has to sit at the removable-media path where the firmware looks
unprompted.
Three things the validation runs caught that review would not have:
rsync is not in the image. I had put it in the build container and never in
the package list, so the first run died at the copy. It now falls back to tar
(--xattrs-include='*', or SELinux labels are silently dropped and the result
does not boot), and rsync is in base.pkgs for the progress output.
Copying a live root makes tar exit non-zero — files change underneath it, and
this machine's clock is wrong besides, so every mtime looks like it is in the
future. With pipefail that aborted the install after the root filesystem and
before /boot, leaving a half-installed disk that looked plausible. Warning-level
exits are now tolerated and only a fatal exit 2 stops the run.
systemd-machine-id-setup keeps an existing valid id, and one had just been
copied off the stick, so the installed system was a clone. The file is removed
first now.
Also: msm_dpu probes ~6s in, while the initramfs is still root, and asks for
qcom/a630_sqe.fw before the real filesystem carrying it is reachable. It never
retries. Adding the Adreno firmware to the initramfs is a few tens of
kilobytes. c630-firmware does the same for the DSP blobs once they exist, since
dracut rejects install_items globs that match nothing.
chrony, because the RTC reads 1970 and nothing was correcting it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XWRjNJMistCy6ngXH5aJLS
I had this wrong. The claim that accelerated graphics works on the firmware
Fedora ships does not survive contact with the device tree.
Fedora does package an sdm845/a630_zap.mbn, which is what led me astray, but
that is the generic Snapdragon 845 zap shader. The C630's node in
sdm850-lenovo-yoga-c630.dts names qcom/sdm850/LENOVO/81JL/qcdxkmsuc850.mbn
specifically — model-signed, present only in the machine's Windows partition.
Until it is supplied the display is unaccelerated.
Also add qcslpi850.mbn and qcvss850.mbn, which were missing entirely, and
attribute the sensor hub to slpi_pas rather than cdsp_pas. The list is now
every firmware-name property in the mainline device tree rather than a
recollection of forum posts, so it should be complete: eight files, not five.
WiFi and Bluetooth are unaffected — ath10k WCN3990 including wlanmdsp.mbn is
genuinely redistributable and genuinely shipped.
Add a section on locating the blobs on an old backup, for the case where
Windows is long gone from the machine.
Only comments changed in config/packages/base.pkgs, so the package set hashes
identically and the staged base stays valid.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XWRjNJMistCy6ngXH5aJLS
Assembles a ready-to-write disk image via Gitea Actions. Mainline has carried
sdm850-lenovo-yoga-c630.dts since 5.5 and Fedora ships it in kernel-core, so
unlike aarch64-laptops/build there is no kernel or GRUB to compile — what is
left is producing an image that boots on firmware which hands Linux no device
tree.
The build runs in an aarch64 container under qemu-user and builds filesystems
from directory trees with mke2fs -d and mcopy rather than mounting loop
devices, so it works on runners that will not hand out /dev/loop-control.
A kernel-install hook writes the devicetree line into each BLS entry; without
it the first `dnf update kernel` would produce an unbootable system.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XWRjNJMistCy6ngXH5aJLS
