Skip to content

tools/extract-vendor-firmware: the Pixel 9a's firmware - #158

Merged
derekross merged 2 commits into
derekross:mainfrom
W4HO-ham:tools-firmware-tegu
Oct 30, 2026
Merged

derekross merged 2 commits into
derekross:mainfrom
W4HO-ham:tools-firmware-tegu

Conversation

@W4HO-ham

Copy link
Copy Markdown
Contributor

This gives tools/extract-vendor-firmware a -d tegu mode, so a fresh install on the palmtop gets its firmware from the owner's own factory image the way the 7 Pro and the 6 Pro do. It answers the firmware-list question from the 9a's package inventory: which stock files the mainline drivers need and under which names.

What it stages for the 9a (board name google,zumapro-tegu):

stock (vendor partition) installed under /usr/lib/firmware/updates
firmware/fw_bcmdhd.bin brcm/brcmfmac4383-pcie.google,zumapro-tegu.bin
firmware/bcmdhd.cal_4383_a3 brcm/brcmfmac4383-pcie.google,zumapro-tegu.txt
firmware/bcmdhd_clm.blob_4383_a3 brcm/brcmfmac4383-pcie.google,zumapro-tegu.clm_blob
firmware/bcmdhd_txcap.blob_4383_a3 brcm/brcmfmac4383-pcie.google,zumapro-tegu.txcap_blob
firmware/brcm/BCM.hcd brcm/BCM.google,zumapro-tegu.hcd
firmware/aoc.bin aoc-tegu.bin
firmware/cs40l26{.wmfw,.bin,-svc.bin,-dvl.bin,-calib.wmfw} the same five names
etc/sensors/registry/ /usr/share/omamobile/sensors/registry/

The 9a's radio is a BCM4383 A3; stock's _4383_a3 files are the production silicon's (its unsuffixed fw_bcmdhd.bin is the same file as the _4383_a3 one, and the _EVT/_PROTO NVRAMs are pre-production boards'). brcmfmac requests the txcap blob on this chip through the same board-name lookup as the other three, so all four get the board suffix, as on cheetah: linux-firmware ships nothing for the 4383, and a generic file that lands there later must not shadow the owner's. The AoC firmware is named aoc-tegu.bin because that is the name the 9a's AoC start script requests (GSA-PLAN F1); it keeps a 7 Pro aoc.bin from ever reaching the 9a's GSA.

Two things the palmtop loads are not in the map on purpose, because they come from linux-firmware, not the image: the Mali CSF firmware (arm/mali/arch11.8/mali_csffw.bin, byte-identical to what the palmtop runs) and the TAS2572 amplifiers' RCA profile (ti/audio/tas257x/tas2572-1-1amp-reg.bin, installed as Top-/Bottom-tas2572RCA1.bin). Stock's tismartpa_driver_tuning.bin is in a different format and AUDIO-PLAN says not to rename it into an RCA. The header documents both. The modem's and GNSS's files stay with tools/extract-vendor-chroot.

For cheetah and raven nothing changes in the staged tree or the messages: the CS35L41 checks now run only for the phones that have those amplifiers, the CS40L26 checks for cheetah and tegu, and the output directories come from the map instead of a fixed brcm/cirrus pair.

Run on the owner's bp3a factory zip (on a build VM, nothing on a phone):

==> Unpacking tegu-bp3a.250905.014/image-tegu-bp3a.250905.014.zip from the factory zip
==> Unpacking vendor.img
==> vendor.img is ext4; reading it with debugfs (no mount)
==> Vendor partition: google/tegu/tegu:16/BP3A.250905.014/13873947:user/release-keys
  /vendor/firmware/fw_bcmdhd.bin       -> /usr/lib/firmware/updates/brcm/brcmfmac4383-pcie.google,zumapro-tegu.bin
  /vendor/firmware/bcmdhd.cal_4383_a3  -> /usr/lib/firmware/updates/brcm/brcmfmac4383-pcie.google,zumapro-tegu.txt
  /vendor/firmware/bcmdhd_clm.blob_4383_a3 -> /usr/lib/firmware/updates/brcm/brcmfmac4383-pcie.google,zumapro-tegu.clm_blob
  /vendor/firmware/bcmdhd_txcap.blob_4383_a3 -> /usr/lib/firmware/updates/brcm/brcmfmac4383-pcie.google,zumapro-tegu.txcap_blob
  /vendor/firmware/brcm/BCM.hcd        -> /usr/lib/firmware/updates/brcm/BCM.google,zumapro-tegu.hcd
  /vendor/firmware/aoc.bin             -> /usr/lib/firmware/updates/aoc-tegu.bin
  /vendor/firmware/cs40l26.wmfw        -> /usr/lib/firmware/updates/cs40l26.wmfw
  /vendor/firmware/cs40l26.bin         -> /usr/lib/firmware/updates/cs40l26.bin
  /vendor/firmware/cs40l26-svc.bin     -> /usr/lib/firmware/updates/cs40l26-svc.bin
  /vendor/firmware/cs40l26-dvl.bin     -> /usr/lib/firmware/updates/cs40l26-dvl.bin
  /vendor/firmware/cs40l26-calib.wmfw  -> /usr/lib/firmware/updates/cs40l26-calib.wmfw
  /vendor/etc/sensors/registry/        -> /usr/share/omamobile/sensors/registry/
==> Wi-Fi firmware: Version=20.123.13 (wlan=r1102095 a3 release25)
==> Bluetooth patch: BCM4383A3 REL M25 TG4 FW:75964737ae CFG:0c4fb8ce44
==> Haptics firmware: cs40l26.wmfw (57716 bytes) and its wavetable
==> AoC firmware: aoc-tegu.bin (29242240 bytes)
==> Staged in .../out (checksums in .../out/vendor-firmware.sha256)

The four Wi-Fi files are byte-identical to the ones the palmtop runs today, and aoc-tegu.bin matches the firmware GSA-PLAN F1 records. tools/pr-check passes. Nothing here runs on a phone; tools/phone-provision gets its tegu flow with the 9a's device package, since that is where the units to enable and the modem's staging are decided.

A -d tegu mode, so a fresh install on the palmtop gets its firmware from the owner's own factory image as the 7 Pro and the 6 Pro do: the BCM4383 A3 Wi-Fi firmware, NVRAM, CLM and TX capability blob under brcmfmac's board-suffixed names, the Bluetooth patch for btbcm, the AoC firmware as aoc-tegu.bin (the name the 9a's AoC start script requests, so a 7 Pro aoc.bin can never be fed to its GSA), the CS40L26 haptics files and the AoC's sensor registry. Stock's _4383_a3 files are the production silicon's; fw_bcmdhd.bin is the same file as its _4383_a3 twin.

Two of the 9a's needs come from linux-firmware rather than the image and are left out on purpose: the Mali CSF firmware (arch11.8) and the TAS2572 amplifiers' RCA profile, since stock's tismartpa_driver_tuning.bin is in a different format and must not be renamed into one (device/tegu/AUDIO-PLAN.md). The header says so.

The CS35L41 checks now run only for the phones that have those amplifiers, the CS40L26 checks for cheetah and tegu, and the output directories come from the map instead of a fixed brcm and cirrus pair. For cheetah and raven the staged tree and the messages are unchanged. Run on the owner's bp3a image: all twelve entries found, the four Wi-Fi files identical to the ones the palmtop runs today and the AoC firmware identical to the one GSA-PLAN F1 records.
derekross pushed a commit that referenced this pull request Oct 29, 2026
…he factory image's (#159)

The 10-26 audio entry said the amplifiers' tas2572RCA1.bin was the Pixel 9a's own file from the factory image. It is linux-firmware's ti/audio/tas257x/tas2572-1-1amp-reg.bin byte for byte (the sha the 10-26 run script pinned), and the factory image has no RCA-format profile. Found while writing the 9a's firmware list for tools/extract-vendor-firmware (#158).
# Conflicts:
#	tools/extract-vendor-firmware
@derekross
derekross merged commit cf9dc22 into derekross:main Oct 30, 2026
2 checks passed
@derekross

Copy link
Copy Markdown
Owner

Thank you for this. Getting the 9a's firmware from the owner's own image, with the reasons in the header, is exactly how I'd like each phone added. Main had moved under it (#113's raven haptics and #145's -d panther), and a plain resolution would have dropped panther from the split checks, so I merged main in with panther in each condition and merged. Two small things for a later pass: the Wi-Fi check's message could use $WIFI_CHIP, and the tegu package README still lists the unsuffixed Wi-Fi names.

derekross added a commit that referenced this pull request Oct 31, 2026
… rounds

L-348 to L-361: shared drivers judged against every int-line user (#150),
board pins against every stock user (#151), ownership flags under the
consumer's lock (#152), /chosen children and Google's panel constants
(#154), same-day notes PRs on one device (#208/#209), merge simulations for
DEVICES rows, ours included (#165), stacked PRs that merged main (#115/#117),
titles behind renumbered versions, a contributor merge that dropped the PR
(#99), smoke package lists as provenance and "owner-authorized" writes
(#200), promised fixes and named tags checked on the remote (#153/#157/#158),
and file swaps into old packages (#193). C-23 (usbc-deadman's device scope,
kernel #107) and C-24 (BNEP in a lynx-only fragment, #98).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…he factory image's (derekross#159)

The 10-26 audio entry said the amplifiers' tas2572RCA1.bin was the Pixel 9a's own file from the factory image. It is linux-firmware's ti/audio/tas257x/tas2572-1-1amp-reg.bin byte for byte (the sha the 10-26 run script pinned), and the factory image has no RCA-format profile. Found while writing the 9a's firmware list for tools/extract-vendor-firmware (derekross#158).
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
-d tegu extracts the Pixel 9a's firmware from the owner's own factory image: the A3 production files, the board-suffixed Wi-Fi files (brcmfmac asks for compatible[0] and the .txcap_blob), the CS40L26 haptics firmware and the AoC image as aoc-tegu.bin, so a Pixel 7 Pro AoC image can never reach the 9a's GSA. Merged with main's -d panther and raven haptics: panther stays in every per-device check (CS35L41: cheetah, panther, raven; CS40L26: cheetah, panther, tegu; media deltas: cheetah, panther).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
… rounds

L-348 to L-361: shared drivers judged against every int-line user (derekross#150),
board pins against every stock user (derekross#151), ownership flags under the
consumer's lock (derekross#152), /chosen children and Google's panel constants
(derekross#154), same-day notes PRs on one device (derekross#208/derekross#209), merge simulations for
DEVICES rows, ours included (derekross#165), stacked PRs that merged main (derekross#115/derekross#117),
titles behind renumbered versions, a contributor merge that dropped the PR
(derekross#99), smoke package lists as provenance and "owner-authorized" writes
(derekross#200), promised fixes and named tags checked on the remote (derekross#153/derekross#157/derekross#158),
and file swaps into old packages (derekross#193). C-23 (usbc-deadman's device scope,
kernel derekross#107) and C-24 (BNEP in a lynx-only fragment, derekross#98).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants