docs: the firmware is downloadable — WOA-Project reference drivers

The blobs turned out to be a public download rather than something to recover
from a lost backup. WOA-Project/Qualcomm-Reference-Drivers collects the
Windows-on-ARM driver cabinets by SoC, and three of them cover this machine
completely: qcdx850.cab has the zap shader and venus, qcipa850.cab has
ipa_fws.elf, and qcsubsys850.cab has the four DSP images plus WLANMDSP.MBN —
the WiFi firmware, which is not in the device tree's firmware-name list at all
and so was never on our list of things to look for.

Correct three claims elsewhere in the file that this disproves. `file` reports
these as ELF QUALCOMM DSP6 images, not as `data` — worth stating precisely,
since it is the quickest way to tell a real blob from a download that returned
an HTML error page. The sizes span 14KB to 60MB rather than "a few hundred KB
to a few MB". And the heading claiming they must come off the Windows partition
is no longer true of the easiest route.

Also narrow what c630-firmware puts in the initramfs. It globbed every .mbn in
the firmware directory, which would have pulled in the 60MB qcdsp2850.mbn and
inflated the initramfs roughly twentyfold. Only the GPU needs to be there —
msm_dpu probes at six seconds, while the initramfs is still root; the ADSP,
CDSP, SLPI, venus and modem all probe around forty seconds in and load from the
real root perfectly happily.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XWRjNJMistCy6ngXH5aJLS

docs/firmware.md

@@ -16,11 +16,11 @@ installs them:
 Networking therefore works on a freshly written image. Graphics does not: see
 below.
 
-## What has to come off the Windows partition
+## What you have to supply
 
-These are signed against per-model keys and Fedora does not ship them, so on a
-stock install they exist only inside the machine's Windows partition. Nothing
-about them is unredistributable in principle — the equivalent files for the
+These are signed against per-model keys and Fedora does not ship them, so a
+stock install has none of them. Nothing about them is unredistributable in
+principle — the equivalent files for the
 ThinkPad X13s *are* in linux-firmware — it is that nobody has obtained
 permission for this model. See [If you have no
 backup](#if-you-have-no-backup).
@@ -91,10 +91,13 @@ msm_dpu ae01000.display-controller: Direct firmware load for qcom/a630_sqe.fw fa
 ```
 
 `overlay/etc/dracut.conf.d/10-c630.conf` puts the Adreno files in the initramfs
-for this reason, and `c630-firmware` adds the extracted DSP blobs to it after
-extraction — the ADSP has to be up before the root filesystem is mounted, and
-dracut rejects `install_items` globs that match nothing, so the entry can only
-be written once the files exist.
+for this reason, and `c630-firmware` adds `qcdxkmsuc850.mbn` once it exists.
+
+Only the GPU needs this. The ADSP, CDSP, SLPI, venus and modem all probe around
+forty seconds in, well after the real root is mounted, and load from it
+perfectly happily. That distinction is worth keeping: `qcdsp2850.mbn` alone is
+60 MB, so sweeping the whole firmware directory into the initramfs would inflate
+it roughly twentyfold for no benefit.
 
 ## Finding them on a backup
 
@@ -115,8 +118,14 @@ files turn up nothing, look for those directories wholesale:
 find /mnt/backup -type d -iname 'qc*.inf_arm64_*'
 ```
 
-Expect a few hundred KB to a few MB each. `file` reports them as `data` — they
-are signed Qualcomm MBN containers, not ELF, apart from `ipa_fws.elf`.
+Sizes range from 14 KB (`qcdxkmsuc850.mbn`) to 60 MB (`qcdsp2850.mbn`).
+Despite the `.mbn` extension they are ELF images, and `file` should say so:
+
+```
+qcdxkmsuc850.mbn: ELF 32-bit LSB executable, QUALCOMM DSP6, version 1 (SYSV)
+```
+
+Anything reporting as text or HTML is a download that went wrong.
 
 Lenovo does not appear to publish a driver package for the 81JL containing
 these. If the backups turn up nothing, see [If you have no
@@ -147,19 +156,37 @@ licensing question, not a technical one.
 
 `qcom/sdm850/LENOVO/81JL/` has simply never been contributed. Fedora ships no
 sdm850 model firmware at all, so it will not turn up in a `linux-firmware`
-update. Two routes:
+update. Three routes, easiest first:
+
+**The WOA-Project reference drivers.** <https://github.com/WOA-Project/Qualcomm-Reference-Drivers>
+collects the Windows-on-ARM driver packages by SoC. The C630's are three
+cabinet files, and between them they hold every file this machine asks for:
+
+| Cabinet | Contains |
+|---|---|
+| `qcdx850.cab` | `qcdxkmsuc850.mbn` (zap shader), `qcvss850.mbn` (venus) |
+| `qcipa850.cab` | `ipa_fws.elf` |
+| `qcsubsys850.cab` | `qcadsp850.mbn`, `qccdsp850.mbn`, `qcslpi850.mbn`, `qcdsp1v2850.mbn`, `qcdsp2850.mbn`, and `WLANMDSP.MBN` |
+
+That last one is a bonus — the WiFi firmware, which is not in the device tree's
+`firmware-name` list but is what the onboard ath10k needs.
+
+`cabextract` or `7z x` will open them:
+
+```sh
+for f in qcdx850.cab qcipa850.cab qcsubsys850.cab; do 7z x -o./extracted "$f"; done
+find ./extracted -iname 'qc*850.mbn' -o -iname 'ipa_fws.elf' -o -iname 'WLANMDSP.MBN'
+```
+
+Ignore the `_nm` and `_CLS` variants; the device tree asks for the plain names.
+Verify what you get with `file` — they should report as
+`ELF 32-bit LSB executable, QUALCOMM DSP6`, not as text or HTML.
 
 **Lenovo's Digital Download Recovery Service.** Lenovo builds a Windows
 recovery image against a machine's serial number and lets the owner download
 it. That image contains `Windows/System32/DriverStore/FileRepository`, and
-therefore all eight files. It is your machine, so you are entitled to the
-image, and it yields the exact blobs for your model rather than a stranger's.
-Mount the image and search it the same way as any other backup:
-
-```sh
-find /mnt/recovery -type f \
-    \( -iname 'qc*850.mbn' -o -iname 'ipa_fws.elf' \) -printf '%10s  %p\n'
-```
+therefore all eight files. Slower than the cabinets, but it is unambiguously
+your own machine's firmware.
 
 **Another C630 owner.** The aarch64-laptops project is where these have
 circulated before, and any 81JL's copy is as good as your own.

overlay/usr/local/sbin/c630-firmware

@@ -34,14 +34,19 @@ fi
 qcom-firmware-extract "$@"
 
 echo
-# The ADSP has to be up before the root filesystem is mounted, so these have to
-# travel in the initramfs. Only write the glob now that the files exist —
-# dracut objects to install_items that match nothing.
+# Only the GPU zap shader has to travel in the initramfs. msm_dpu probes around
+# six seconds in, while the initramfs is still the root filesystem, and never
+# retries. Everything else — ADSP, CDSP, SLPI, venus, modem — is probed well
+# after the real root is mounted and loads from there quite happily.
+#
+# This matters: qcdsp2850.mbn alone is 60MB. Globbing the whole directory into
+# the initramfs, which is what this used to do, would have bloated it roughly
+# twentyfold to no purpose.
 FW_DIR=/usr/lib/firmware/updates/qcom/sdm850/LENOVO/81JL
-if compgen -G "${FW_DIR}/*.mbn" >/dev/null || compgen -G "${FW_DIR}/*.elf" >/dev/null; then
-    printf 'install_items+=" %s/*.mbn %s/*.elf "\n' "$FW_DIR" "$FW_DIR" \
-        > /etc/dracut.conf.d/20-c630-extracted-firmware.conf
-    echo "Extracted firmware will be included in the initramfs."
+if [ -f "${FW_DIR}/qcdxkmsuc850.mbn" ]; then
+    printf 'install_items+=" %s/qcdxkmsuc850.mbn "\n' "$FW_DIR" \
+        > /etc/dracut.conf.d/20-c630-zap-shader.conf
+    echo "GPU zap shader will be included in the initramfs."
 fi
 
 echo "Regenerating the initramfs..."