You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Pixel 7 (panther): a second unit available for testing #342
Hi @nesmeck and @derekross, I have a spare Pixel 7 (panther) I can use for testing the panther port, so there would be a second unit next to nesmeck's.
Phone: Pixel 7, unlocked model, stock Android 17 CP3A.261005.002.A1. The bootloader is still locked; I'm fine unlocking it and wiping the phone. Its NFC doesn't work (a hardware fault), everything else does.
What I can do: flash the test kernels and images you point me to, run the runbooks step by step and report the results with logs (dmesg, journalctl, the smoke tests). I'm used to measuring things one change at a time; I've been doing that on the same phone for DroidDeck (Mali, frame pacing, temperatures, memory).
What I'm most interested in: gaming. Once Wi-Fi and CPU frequency land on panther, I'd like to try gamescope with Steam (arm64 + FEX) on Panthor/PanVK and measure it.
Is this useful, and is there something on panther you'd like checked first?
I write these reports with help from an AI assistant (Claude); the testing and the results are mine.
Welcome, @angelrenefm, and thank you: a second Pixel 7 is very useful. Several of panther's
values come from one phone so far (the CPU power model on linux-omarchy-mobile#184 is one chip's
numbers, and the display and battery work in review has run on @nesmeck's unit only), so a
second unit that reproduces them tells us what's the phone and what's the port.
Before you unlock it: please read CONTRIBUTING §3 (the safety model). Stock Android stays on
slot A as the way back, and we only ever write slot B and userdata. Then the first useful
step needs no flash at all: the read-only stock capture in CONTRIBUTING §5.1, which lets us
compare your phone with @nesmeck's (panel, battery, board revision). After that, a boot of the
current images on slot B with a tools/phone-smoke report, following device/panther/BRINGUP-M1.md, is the best first check. @nesmeck can say which of his open
pull requests he'd most like a second run of.
Gaming is a great goal, and Panthor already renders the desktop on panther. Wi-Fi and CPU
frequency are the prerequisites, as you say, so measuring those first fits nicely. Thanks
for mentioning the broken NFC: we'll leave NFC out of anything we ask you to run.
Broken hardware: NFC and wireless charging. Wired charging works.
Both slots on the same bootloader:getvar all reports no per-slot version, so I did what akita's NOTES describe: the full OTA (post-build ...CP3A.261005.002.A1/16269273) sideloaded twice, and each slot read from Android while running: _b and _a both CP3A.261005.002.A1 with cloudripper-17.0-15902555. Then unlocked.
Install: built on my laptop from main at a08771b: kernel omamobile at 4cf2c8be1 (7.3.0-rc6-omamobile-push1-g4cf2c8be1705), bin/omamobile modules/initramfs/images with DEVICE=panther (dtbo: one entry, id 0x30406, rev 0x10000), and tools/build-rootfs --channel development with [omamobile] taken from pkgs.omamobile.org/development. tools/install-phone --bootloader-slots-match cloudripper-17.0-15902555 passed its checks and flashed slot B and userdata without errors.
What happened: slot B never showed up on USB. Watching lsusb every second after fastboot --set-active=b + reboot, there was nothing from the phone for 72 s, then stock recovery on slot A (reboot,recovery, as in BRINGUP-M1's boot 6 note), with /sys/fs/pstore empty. The first boot right after the install also ended in stock recovery (I wasn't watching USB that time). So the three slot B attempts went by in about a minute without a USB gadget.
The phone is on stock recovery now, with no data reset, and I can boot it again whenever useful. 4cf2c8be1 differs from the last published kernel (2219cc6cda77) only in bluejay and zumapro files, nothing for gs201, so I don't think rebuilding that one would change anything. Is there a known-good image set for panther I could flash to tell my build apart from the phone, and what would you like me to capture on the next attempt?
I'm working on this with AI help (Claude); every value above comes from the phone or the build logs.
One more data point: I rebuilt the slot B images around your published kernel instead of mine (vmlinuz and gs201-panther.dtb from linux-omamobile-7.3.0rc6.r635.g2219cc6cda77-1, same initramfs) and flashed them with bin/omamobile flash. Same result: the screen stays black, nothing on USB, and after 74 s the phone is back in stock recovery on slot A. So it isn't my kernel build; what's left on my side is the initramfs and the image packing, which I built on Fedora with podman instead of docker.
Welcome, @angelrenefm, and thanks for the careful report. The 72 s with no USB and the return to recovery point at AVB, not at your build.
The cause, I think: stock's vbmeta on the Pixel 7 chains boot (rollback index location 2) and init_boot (location 4), exactly as on the Pixel 7 Pro. I checked the factory image's vbmeta.img with avbtool info_image. After your full OTA on both slots, your vbmeta_b is stock's. libavb then reads a footer-less boot or init_boot as invalid metadata, which an unlocked bootloader doesn't let through, so slot B never starts. That's the Pixel 7 Pro's issue #11 (device/cheetah/NOTES.md, 2026-10-28). device/cheetah/device.conf has AVB_FOOTERS for it, and device/panther/device.conf doesn't.
My unit boots without footers because it ran LineageOS before the port. My stock install wrote the bootloader and radio to slot B, but not vbmeta_b, so slot B kept a vbmeta that doesn't chain them. So your phone is the normal case, and mine is the exception.
The fix is on my fork, branch panther-avb-footers. It adds cheetah's three lines to the descriptor, with the values from the factory image's own boot.img and init_boot.img footers:
The sizes are getvar partition-size:boot_b and init_boot_b. Rebuilt from the same kernel output, only boot.img and init_boot.img change, by their padding and footer. vendor_boot, vendor_kernel_boot, dtbo and dtb.img are byte for byte the same. tests/all passes. The package's copy of the descriptor follows, so omamobile-boot-update writes the footers on the phone too.
Could you try it? Your userdata is already installed, so only the boot images need rebuilding:
Check out the branch, or add the three lines to your device/panther/device.conf.
Run DEVICE=panther bin/omamobile images. It needs avbtool (platform/external/avb, or tools/fetch-toolchain). It should print "AVB footer: boot.img, padded to its 67108864-byte partition", and the same for init_boot.img.
Run bin/omamobile flash, which resets slot B's retries.
If it boots, the Omarchy splash and USB should come up within a minute or so. If it still doesn't, these would help:
fastboot getvar all 2>&1 | grep -E 'slot-(retry-count|successful|unbootable)' right after it falls back;
the reboot reason stock reports (adb shell getprop ro.boot.bootreason from recovery, or /proc/bootconfig);
avbtool info_image --image boot.img | head on the image you flashed.
I flashed the same images on my unit, whose vbmeta_b doesn't chain them: it booted as before, same kernel, all services up, and spent one retry. So the footers are harmless on the exception too. I'll open the pull request once your phone confirms the fix.
What a second unit would help with most, once it boots:
the CPU energy model on linux-omarchy-mobile#184, since its values are one chip's: tools/power-ab eas with a video playing, as in the PR;
the display on linux-omarchy-mobile#181 (the screen sleeping with screen_off=2, and 90 Hz later);
the battery and gauge values in device/panther/BATTERY-PLAN.md, to see which come from the pack and which from my phone.
Thank you both. @angelrenefm, that's a model report: parity checked properly with two OTA sideloads, the USB watched second by second, and the reset declined. @nesmeck, I agree with your reading: a stock vbmeta_b chains boot and init_boot, footer-less images are invalid metadata to libavb, and that's the 7 Pro's issue #11 exactly. The three values match both phones' slot A build (CP3A.261005.002.A1), so they're safe for both units.
@angelrenefm, please go ahead with nesmeck's branch. Two notes from the 7 Pro: once slot B has booted with footers, it keeps needing them, so build every later image set with them; and if stock recovery shows again, keep choosing neither option and use Volume Down + Power for fastboot. After the first good boot, a read-only tools/nv-backup is worth doing before any radio work.
@nesmeck, please open the pull request with angelrenefm's result in it; no need to wait for more than that boot.
What I flashed: my tree at a08771b (the same base as the installed userdata) with 81b8d63 from panther-avb-footers cherry-picked on top, nothing else. DEVICE=panther bin/omamobile images with the pinned avbtool 1.4.0 printed both lines: "AVB footer: boot.img, padded to its 67108864-byte partition" and the same for init_boot.img at 8388608. Same kernel as before (7.3.0-rc6-omamobile-push1-g4cf2c8be1705).
Flash:bin/omamobile flash wrote the five slot B images, all OKAY, then set slot B active and rebooted.
Boot: watching lsusb every second, fastboot dropped off at 06:42:32 and 1d6b:0104 Linux Foundation Multifunction Composite Gadget showed up at 06:42:56, 24 s later. It was still there five minutes on, and the host bound it as a cdc_ncm interface. Before the footers there was nothing on USB for 72 s and then stock recovery.
I haven't got SSH yet because my laptop doesn't have the omamobile-phone NetworkManager connection from INSTALL.md. Once it does, I'll run tools/phone-smoke and the read-only tools/nv-backup derekross mentioned, and post both. From now on I'll build every image set with the footers.
I'm working on this with AI help (Claude); the values above come from the build output and from lsusb on my laptop.
cma: Failed to reserve 384 MiB
panthor 28000000.gpu: error -ENODEV: _opp_set_regulators: no regulator (mali) found
cs40l26-vibra 0-0043: error -ENOENT: No cs40l26.wmfw
The vibra one is expected, since I haven't run tools/phone-provision with the factory firmware yet. tools/nv-backup -d panther stops before reading anything ("device.conf names no NV_PARTS"), so there's no NV backup yet. Tell me which partitions to take and I'll run it.
I'm working on this with AI help (Claude); the report is the tool's output, unedited.
Haptics: the "new" probe error in my smoke report was cs40l26-vibra with no firmware. With cs40l26.wmfw/.bin
from this phone's OTA (tools/extract-vendor-firmware -d panther), it runs (v7.2.91) and vibrates (felt on the
phone). After a reboot the firmware loads on its own ("running" 5 s in); vibration not re-tested after that boot. The panthor regulator line is the known one; sorry for the noise.
Battery pack: this unit has the Lishen 3 kOhm pack, not the Coslight one: the gauge reads DesignCap
0x0891 (4386 mAh), exactly lsn-a1-3k's in stock's panther MP overlay (board id 0x30406; cos-a1-1k is
0x0873). TempCo reads 0x1402 (Lishen's stock value is 0x1802, so not an exact match). FullCapRep 0x0754,
POR clear (Status 0x2000), and DesignCap is not the Coslight model's, so it looks like the kernel never loaded its model
here and the gauge keeps what stock wrote (inferred from those two reads). If this
gauge ever goes through POR, the current DT would load the Coslight model on a Lishen pack.
devinfo (for Marking slot B successful: the plan after #125's hold, and for the Pixel 8 Pro #252 / L-36): layout 3.15. In the same state, Linux read bytes 0x30-0x37 01 08 00 00 01 04 00 00
and fastboot said A retry 1 / B retry 1, none successful, current B: they match, also after an ABL change
(B 2→1). After fastboot --set-active=b and one boot: 01 08 00 00 02 0c 00 00 (B's flags gained bit 3).
Stock panther MP overlay vs cheetah MP (from this phone's OTA dtbo): the MAX77759 PMIC interrupt (gpa9-4),
the Wi-Fi lines (gph2-4/5, gpa6-1) and Bluetooth lines (gpp16-2/3, gpa6-6) are the same; the charger's
temperature, CC and CV tables are identical; only chg-battery-capacity differs (4277 vs 4926 mAh).
Wi-Fi and Bluetooth come up on the Pixel 7 with cheetah's nodes. Stock's panther MP overlay has the same lines
(WL_REG_ON gph2-4, WL_DEV_WAKE gph2-5, WL_HOST_WAKE gpa6-1, BT_REG_ON gpp16-2, BT_DEV_WAKE gpp16-3, BT_HOST_WAKE
gpa6-6), so the commit only copies them with brcm,board-type = "google,gs201-panther". Firmware from this phone's
OTA with tools/extract-vendor-firmware -d panther (Wi-Fi 20.101.91; BT patch ES1PX_GG_P10, not cheetah's C10).
Result, kernel = omamobile + derekross/linux-omarchy-mobile#181 + that commit: PCIe Gen2 x1 link up on try 1, BCM4389/3, wlp1s0, a scan sees
33 networks; hci0 up at boot and powered. Not connected to a network yet, and nothing measured on power. @nesmeck, is this something you'd like as a kernel PR from me, or would you rather take it into your series?
The package side would be cheetah's Wi-Fi/BT files copied into device-panther.
Thank you, @angelrenefm, for a very productive day on the second unit: the footers, a smoke report, a backup with a passing --check, haptics, the pack identified, and Wi-Fi and Bluetooth up. A few things, roughly in order of importance.
Slot A on your phone. You wrote that fastboot showed slot A at one retry and not successful. After the install, the fallbacks to A went into stock recovery and never finished an Android boot, so ABL never marked A. That means stock has only one try left as your way back. The next time you're in stock anyway, let Android boot to the lock screen once so it gets marked, and then take a fresh tools/nv-backup, because a stock boot rewrites efs, persist and modem_userdata. No need to do it just for this, but please do it before any radio or PMIC work.
The Lishen pack. That's an important find. The panther board file on omamobile carries the Coslight 1 kOhm model, and the gauge driver writes that model whenever the gauge reports a power-on reset. On your unit POR is clear, so nothing has been written so far. But if the battery ever runs completely flat, or the pack is disconnected, the next boot under our kernel would load the Coslight model onto a Lishen pack. Until the model is chosen by the pack's ID, please don't let that unit run flat. @nesmeck, could this go into device/panther/BATTERY-PLAN.md? It also argues for picking the model by the ID resistor, as stock does, or skipping the load for a pack we don't recognise.
NV_PARTS for panther: yes please, as a pull request. Both copies of the descriptor need the lines (the package test compares them). Leave the version to me: #396 takes 0.1.6 and #281 0.1.7, so I'll set the next free one when I merge.
Wi-Fi and Bluetooth: great to see a scan of 33 networks on the first try. Please open the kernel PR yourself, and agree with @nesmeck on where it sits in his series. Before it goes in:
The Wi-Fi node needs cheetah's restricted DMA pool (restricted-dma-pool in gs201-cheetah.dts) and its memory-region. On gs201 we only run Wi-Fi with that pool, because the HSI2 IOMMU can't keep PCIe apart from UFS.
Keep the firmware out of the repositories, as you did (each owner extracts it from their own OTA).
The package side (cheetah's Wi-Fi start script with its hang marker, and the Bluetooth address unit) is a separate omarchy-mobile PR, and it has to merge with or before the kernel change that turns Wi-Fi on.
Please list "not connected to a network, no power measured" as untested.
devinfo: thank you, that's exactly the read-only comparison #252 needed (the bytes next to getvar in the same state, and again after the bootloader changed them). The bit that appeared on slot B after --set-active=b and a boot is the bootloader's own. Nothing in our panther package writes devinfo.
The stock overlay comparison (the same MAX77759, Wi-Fi and Bluetooth lines as the 7 Pro, identical charger tables) would be useful in panther's NOTES too.
@angelrenefm, great results, and thanks for the D3 run on #181 as well. Here's my proposal for Wi-Fi and Bluetooth, so neither of us redoes the other's work:
Bluetooth on its own (kernel), from me. I have a local commit, "arm64: dts: exynos: google: gs201-panther: add Bluetooth", already signed. It adds cmu_apm, usi_apm1, serial_apm1 with brcm,bcm4389-bt, and the three pin groups. It was tested step by step in my device/panther/BLUETOOTH-PLAN.md (BT0 to BT3): the patch download (ES1PX_GG_P10, the same as yours), the bootloader's address, the shell's panel, a BLE mouse, and headphones over A2DP and AVRCP. I'll open it against omamobile as its own PR.
Wi-Fi from you, stacked on it. That's the PCIe link, the Wi-Fi node with cheetah's restricted-dma-pool and memory-region as derekross asked, and the WLAN pins. If your commit also has the Bluetooth nodes, please drop them and base it on mine, or on omamobile once mine has merged.
Packages, split the same way.
Bluetooth's files (omamobile-bt-addr, its unit and udev rule, and the btbcm option) come with my device-panther update, which follows the merges ahead of it.
Cheetah's Wi-Fi start script with its hang marker goes with your Wi-Fi package PR, together with or before your kernel PR.
One thing to keep in mind when copying cheetah's files: the 80-omamobile-bt-addr.rules rule matches hci[0-9]*, which also catches the per-connection devices (hci0:N). So every Bluetooth connection leaves a failed omamobile-bt-addr@hci0:N unit; the script refuses them, so nothing changes. I'm preparing a fix for cheetah, raven and tegu (ENV{DEVTYPE}=="host", which only the controller has), and panther's copy will carry it.
If you'd rather split it another way, say so, and I'll adapt.
On the Lishen pack: I'll write it into device/panther/BATTERY-PLAN.md. That covers what your unit reads, why a power-on reset would load the wrong model today, and the two options derekross mentions (picking the model by the pack's ID resistor, as stock does, or not loading one for a pack we don't recognise). Please keep that unit from running flat until then.
@nesmeck, agreed, thanks for laying it out. I've reshaped both of mine to match:
Kernel, Wi-Fi only:arm64: dts: exynos: google: gs201-panther: add Wi-Fi linux-omarchy-mobile#195 now carries just the Wi-Fi nodes (the WL_REG_ON regulator, cheetah's restricted-dma-pool with the node's memory-region, PCIe GEN4A_1, its PHY and System MMU window, and the WLAN pins). The Bluetooth nodes are gone. It sits on omamobile for now; I'll rebase it onto your Bluetooth commit once that PR is open, or onto omamobile once yours has merged.
Package, Wi-Fi only:device-panther: Wi-Fi, with the Pixel 7 Pro's files #419 now has only cheetah's Wi-Fi start script with its hang marker, its unit, the Wi-Fi autoload blacklist and the MAC helper with its udev rule. omamobile-bt-addr, its unit and rule, and the btbcm option are out, for your device-panther update (with your ENV{DEVTYPE}=="host" fix).
Both still pass tools/kernel-pr-check (with dtbs_check) and test/all. The Wi-Fi-only builds haven't been booted on the phone yet; what ran there had the Bluetooth nodes too, and the PRs say so.
Thanks for taking the Lishen pack into BATTERY-PLAN. The unit won't run flat in the meantime.
Thank you both for sorting out the split so quickly. It's exactly what I'd hoped for, and it makes the reviews much simpler. I've left a review on each pull request; here is the order I plan to merge them in, so it's in one place:
linux-omarchy-mobile#195 (Wi-Fi) after device-panther: Wi-Fi, with the Pixel 7 Pro's files #419 is published, so every Pixel 7 has the autoload blacklist before a kernel with the Wi-Fi node. It merges cleanly next to linux-omarchy-mobile#196, so no rebase is needed.
@nesmeck, thanks for taking the Lishen pack into BATTERY-PLAN. Until the model is chosen by the pack's ID, that note is the only warning a Pixel 7 owner with that pack gets, so I'd like it in a pull request soon.
Activity
derekross commentedon Nov 2, 2026
Welcome, @angelrenefm, and thank you: a second Pixel 7 is very useful. Several of panther's
values come from one phone so far (the CPU power model on linux-omarchy-mobile#184 is one chip's
numbers, and the display and battery work in review has run on @nesmeck's unit only), so a
second unit that reproduces them tells us what's the phone and what's the port.
Before you unlock it: please read CONTRIBUTING §3 (the safety model). Stock Android stays on
slot A as the way back, and we only ever write slot B and
userdata. Then the first usefulstep needs no flash at all: the read-only stock capture in CONTRIBUTING §5.1, which lets us
compare your phone with @nesmeck's (panel, battery, board revision). After that, a boot of the
current images on slot B with a
tools/phone-smokereport, followingdevice/panther/BRINGUP-M1.md, is the best first check. @nesmeck can say which of his openpull requests he'd most like a second run of.
Gaming is a great goal, and Panthor already renders the desktop on panther. Wi-Fi and CPU
frequency are the prerequisites, as you say, so measuring those first fits nicely. Thanks
for mentioning the broken NFC: we'll leave NFC out of anything we ask you to run.
angelrenefm commentedon Nov 2, 2026
Thanks @derekross. I read CONTRIBUTING §3 and went ahead. Here is what I have so far, including a first slot B boot that didn't come up.
Stock capture (§5.1, adb, no root), to compare with @nesmeck's unit:
cloudripper-17.0-15902555, basebandg5300q-260604-260807-B-16036236.cdt_hwid0x00030406 00010000 (the same id and rev asdevice.conf), SoC GS201 rev 2.panel-samsung-s6e3fc3-p10, battery design 4,386 mAh.Both slots on the same bootloader:
getvar allreports no per-slot version, so I did what akita's NOTES describe: the full OTA (post-build ...CP3A.261005.002.A1/16269273) sideloaded twice, and each slot read from Android while running:_band_aboth CP3A.261005.002.A1 withcloudripper-17.0-15902555. Then unlocked.Install: built on my laptop from
mainat a08771b: kernelomamobileat 4cf2c8be1 (7.3.0-rc6-omamobile-push1-g4cf2c8be1705),bin/omamobile modules/initramfs/imageswithDEVICE=panther(dtbo: one entry, id 0x30406, rev 0x10000), andtools/build-rootfs --channel developmentwith[omamobile]taken from pkgs.omamobile.org/development.tools/install-phone --bootloader-slots-match cloudripper-17.0-15902555passed its checks and flashed slot B anduserdatawithout errors.What happened: slot B never showed up on USB. Watching
lsusbevery second afterfastboot --set-active=b+reboot, there was nothing from the phone for 72 s, then stock recovery on slot A (reboot,recovery, as in BRINGUP-M1's boot 6 note), with/sys/fs/pstoreempty. The first boot right after the install also ended in stock recovery (I wasn't watching USB that time). So the three slot B attempts went by in about a minute without a USB gadget.The phone is on stock recovery now, with no data reset, and I can boot it again whenever useful. 4cf2c8be1 differs from the last published kernel (2219cc6cda77) only in bluejay and zumapro files, nothing for gs201, so I don't think rebuilding that one would change anything. Is there a known-good image set for panther I could flash to tell my build apart from the phone, and what would you like me to capture on the next attempt?
I'm working on this with AI help (Claude); every value above comes from the phone or the build logs.
angelrenefm commentedon Nov 2, 2026
One more data point: I rebuilt the slot B images around your published kernel instead of mine (
vmlinuzandgs201-panther.dtbfromlinux-omamobile-7.3.0rc6.r635.g2219cc6cda77-1, same initramfs) and flashed them withbin/omamobile flash. Same result: the screen stays black, nothing on USB, and after 74 s the phone is back in stock recovery on slot A. So it isn't my kernel build; what's left on my side is the initramfs and the image packing, which I built on Fedora with podman instead of docker.nesmeck commentedon Nov 3, 2026
Welcome, @angelrenefm, and thanks for the careful report. The 72 s with no USB and the return to recovery point at AVB, not at your build.
The cause, I think: stock's
vbmetaon the Pixel 7 chainsboot(rollback index location 2) andinit_boot(location 4), exactly as on the Pixel 7 Pro. I checked the factory image'svbmeta.imgwithavbtool info_image. After your full OTA on both slots, yourvbmeta_bis stock's. libavb then reads a footer-lessbootorinit_bootas invalid metadata, which an unlocked bootloader doesn't let through, so slot B never starts. That's the Pixel 7 Pro's issue #11 (device/cheetah/NOTES.md, 2026-10-28).device/cheetah/device.confhasAVB_FOOTERSfor it, anddevice/panther/device.confdoesn't.My unit boots without footers because it ran LineageOS before the port. My stock install wrote the bootloader and radio to slot B, but not
vbmeta_b, so slot B kept a vbmeta that doesn't chain them. So your phone is the normal case, and mine is the exception.The fix is on my fork, branch
panther-avb-footers. It adds cheetah's three lines to the descriptor, with the values from the factory image's ownboot.imgandinit_boot.imgfooters:The sizes are
getvar partition-size:boot_bandinit_boot_b. Rebuilt from the same kernel output, onlyboot.imgandinit_boot.imgchange, by their padding and footer.vendor_boot,vendor_kernel_boot,dtboanddtb.imgare byte for byte the same.tests/allpasses. The package's copy of the descriptor follows, soomamobile-boot-updatewrites the footers on the phone too.Could you try it? Your
userdatais already installed, so only the boot images need rebuilding:device/panther/device.conf.DEVICE=panther bin/omamobile images. It needsavbtool(platform/external/avb, ortools/fetch-toolchain). It should print "AVB footer: boot.img, padded to its 67108864-byte partition", and the same forinit_boot.img.bin/omamobile flash, which resets slot B's retries.If it boots, the Omarchy splash and USB should come up within a minute or so. If it still doesn't, these would help:
fastboot getvar all 2>&1 | grep -E 'slot-(retry-count|successful|unbootable)'right after it falls back;adb shell getprop ro.boot.bootreasonfrom recovery, or/proc/bootconfig);avbtool info_image --image boot.img | headon the image you flashed.I flashed the same images on my unit, whose
vbmeta_bdoesn't chain them: it booted as before, same kernel, all services up, and spent one retry. So the footers are harmless on the exception too. I'll open the pull request once your phone confirms the fix.What a second unit would help with most, once it boots:
tools/power-ab easwith a video playing, as in the PR;screen_off=2, and 90 Hz later);device/panther/BATTERY-PLAN.md, to see which come from the pack and which from my phone.I'm working on this with AI help (Claude).
🤖 Generated with Claude Code
derekross commentedon Nov 3, 2026
Thank you both. @angelrenefm, that's a model report: parity checked properly with two OTA sideloads, the USB watched second by second, and the reset declined. @nesmeck, I agree with your reading: a stock
vbmeta_bchainsbootandinit_boot, footer-less images are invalid metadata to libavb, and that's the 7 Pro's issue #11 exactly. The three values match both phones' slot A build (CP3A.261005.002.A1), so they're safe for both units.@angelrenefm, please go ahead with nesmeck's branch. Two notes from the 7 Pro: once slot B has booted with footers, it keeps needing them, so build every later image set with them; and if stock recovery shows again, keep choosing neither option and use Volume Down + Power for fastboot. After the first good boot, a read-only
tools/nv-backupis worth doing before any radio work.@nesmeck, please open the pull request with angelrenefm's result in it; no need to wait for more than that boot.
angelrenefm commentedon Nov 3, 2026
@nesmeck @derekross it boots. Thanks for finding it so fast.
userdata) with 81b8d63 frompanther-avb-footerscherry-picked on top, nothing else.DEVICE=panther bin/omamobile imageswith the pinned avbtool 1.4.0 printed both lines: "AVB footer: boot.img, padded to its 67108864-byte partition" and the same forinit_boot.imgat 8388608. Same kernel as before (7.3.0-rc6-omamobile-push1-g4cf2c8be1705).bin/omamobile flashwrote the five slot B images, all OKAY, then set slot B active and rebooted.lsusbevery second, fastboot dropped off at 06:42:32 and1d6b:0104 Linux Foundation Multifunction Composite Gadgetshowed up at 06:42:56, 24 s later. It was still there five minutes on, and the host bound it as acdc_ncminterface. Before the footers there was nothing on USB for 72 s and then stock recovery.I haven't got SSH yet because my laptop doesn't have the
omamobile-phoneNetworkManager connection from INSTALL.md. Once it does, I'll runtools/phone-smokeand the read-onlytools/nv-backupderekross mentioned, and post both. From now on I'll build every image set with the footers.I'm working on this with AI help (Claude); the values above come from the build output and from
lsusbon my laptop.nesmeck commentedon Nov 3, 2026
Thanks for testing it so quickly, @angelrenefm. The pull request is #396, with your boot in it.
🤖 Generated with Claude Code
angelrenefm commentedon Nov 3, 2026
The smoke test on my Pixel 7, booted with the AVB footers (#396), kernel and root unchanged from the install:
phone-smoke: panther
Kernel
7.3.0-rc6-omamobile-push1-g4cf2c8be1705, tools atf1a24c4, 2026-11-03T16:52Z.Packages: hyprland 0.57.0-3.1, libcamera 0.7.2-4.22, mesa 1:26.2.4-1.4, omacam 0.3.3-1, omamobile-apps 0.2.0-1, omamobile-device-panther 0.1.5-1, omamobile-firstboot 0.1.3-1, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-omarchy 0.1.28-1, omamobile-owl-widgets 0.1.1-1, omamobile-shell 0.2.79-1, omarchy 4.0.6-1, quickshell 0.3.2.r7.g2b4501a-1
The three new dmesg errors, in full:
The vibra one is expected, since I haven't run
tools/phone-provisionwith the factory firmware yet.tools/nv-backup -d pantherstops before reading anything ("device.conf names no NV_PARTS"), so there's no NV backup yet. Tell me which partitions to take and I'll run it.I'm working on this with AI help (Claude); the report is the tool's output, unedited.
angelrenefm commentedon Nov 3, 2026
Follow-up from the second Pixel 7:
NV_PARTS="efs efs_backup modem_userdata persist modem_a modem_b"andNV_BACKUP_EXTRA="devinfo"in panther's descriptor (cheetah's list; each found once by PARTNAME), a read-onlybackup and
--checkagainst the phone both pass. Branch ready if you want it as a PR (version number is yoursto pick, since device-panther 0.1.6: AVB footers on boot and init_boot #396 and device-panther 0.1.5: the screen sleeps with the display off, and the D3 plan #281 both take 0.1.6).
cs40l26.wmfw/.binfrom this phone's OTA (
tools/extract-vendor-firmware -d panther), it runs (v7.2.91) and vibrates (felt on thephone). After a reboot the firmware loads on its own ("running" 5 s in); vibration not re-tested after that boot. The panthor regulator line is the known one; sorry for the noise.
0x0891 (4386 mAh), exactly
lsn-a1-3k's in stock's panther MP overlay (board id 0x30406;cos-a1-1kis0x0873). TempCo reads 0x1402 (Lishen's stock value is 0x1802, so not an exact match). FullCapRep 0x0754,
POR clear (Status 0x2000), and DesignCap is not the Coslight model's, so it looks like the kernel never loaded its model
here and the gauge keeps what stock wrote (inferred from those two reads). If this
gauge ever goes through POR, the current DT would load the Coslight model on a Lishen pack.
01 08 00 00 01 04 00 00and fastboot said A retry 1 / B retry 1, none successful, current B: they match, also after an ABL change
(B 2→1). After
fastboot --set-active=band one boot:01 08 00 00 02 0c 00 00(B's flags gained bit 3).dtbo): the MAX77759 PMIC interrupt (gpa9-4),the Wi-Fi lines (gph2-4/5, gpa6-1) and Bluetooth lines (gpp16-2/3, gpa6-6) are the same; the charger's
temperature, CC and CV tables are identical; only
chg-battery-capacitydiffers (4277 vs 4926 mAh).I'm working on this with AI help (Claude).
angelrenefm commentedon Nov 3, 2026
Wi-Fi and Bluetooth come up on the Pixel 7 with cheetah's nodes. Stock's panther MP overlay has the same lines
(WL_REG_ON gph2-4, WL_DEV_WAKE gph2-5, WL_HOST_WAKE gpa6-1, BT_REG_ON gpp16-2, BT_DEV_WAKE gpp16-3, BT_HOST_WAKE
gpa6-6), so the commit only copies them with
brcm,board-type = "google,gs201-panther". Firmware from this phone'sOTA with
tools/extract-vendor-firmware -d panther(Wi-Fi 20.101.91; BT patchES1PX_GG_P10, not cheetah's C10).Result, kernel = omamobile + derekross/linux-omarchy-mobile#181 + that commit: PCIe Gen2 x1 link up on try 1, BCM4389/3,
wlp1s0, a scan sees33 networks;
hci0up at boot and powered. Not connected to a network yet, and nothing measured on power.@nesmeck, is this something you'd like as a kernel PR from me, or would you rather take it into your series?
The package side would be cheetah's Wi-Fi/BT files copied into device-panther.
I'm working on this with AI help (Claude).
derekross commentedon Nov 3, 2026
Thank you, @angelrenefm, for a very productive day on the second unit: the footers, a smoke report, a backup with a passing
--check, haptics, the pack identified, and Wi-Fi and Bluetooth up. A few things, roughly in order of importance.Slot A on your phone. You wrote that fastboot showed slot A at one retry and not successful. After the install, the fallbacks to A went into stock recovery and never finished an Android boot, so ABL never marked A. That means stock has only one try left as your way back. The next time you're in stock anyway, let Android boot to the lock screen once so it gets marked, and then take a fresh
tools/nv-backup, because a stock boot rewritesefs,persistandmodem_userdata. No need to do it just for this, but please do it before any radio or PMIC work.The Lishen pack. That's an important find. The panther board file on
omamobilecarries the Coslight 1 kOhm model, and the gauge driver writes that model whenever the gauge reports a power-on reset. On your unit POR is clear, so nothing has been written so far. But if the battery ever runs completely flat, or the pack is disconnected, the next boot under our kernel would load the Coslight model onto a Lishen pack. Until the model is chosen by the pack's ID, please don't let that unit run flat. @nesmeck, could this go intodevice/panther/BATTERY-PLAN.md? It also argues for picking the model by the ID resistor, as stock does, or skipping the load for a pack we don't recognise.NV_PARTSfor panther: yes please, as a pull request. Both copies of the descriptor need the lines (the package test compares them). Leave the version to me: #396 takes 0.1.6 and #281 0.1.7, so I'll set the next free one when I merge.Wi-Fi and Bluetooth: great to see a scan of 33 networks on the first try. Please open the kernel PR yourself, and agree with @nesmeck on where it sits in his series. Before it goes in:
restricted-dma-poolings201-cheetah.dts) and itsmemory-region. On gs201 we only run Wi-Fi with that pool, because the HSI2 IOMMU can't keep PCIe apart from UFS.omamobile+ omamobile-omarchy 0.1.17, omamobile-shell 0.2.41: the lock glance after hands-on testing #181 + the commit. Please say so in the PR, or re-run it onomamobileplus the commit alone.devinfo: thank you, that's exactly the read-only comparison #252 needed (the bytes next to
getvarin the same state, and again after the bootloader changed them). The bit that appeared on slot B after--set-active=band a boot is the bootloader's own. Nothing in our panther package writesdevinfo.The stock overlay comparison (the same MAX77759, Wi-Fi and Bluetooth lines as the 7 Pro, identical charger tables) would be useful in panther's NOTES too.
nesmeck commentedon Nov 3, 2026
@angelrenefm, great results, and thanks for the D3 run on #181 as well. Here's my proposal for Wi-Fi and Bluetooth, so neither of us redoes the other's work:
cmu_apm,usi_apm1,serial_apm1withbrcm,bcm4389-bt, and the three pin groups. It was tested step by step in mydevice/panther/BLUETOOTH-PLAN.md(BT0 to BT3): the patch download (ES1PX_GG_P10, the same as yours), the bootloader's address, the shell's panel, a BLE mouse, and headphones over A2DP and AVRCP. I'll open it againstomamobileas its own PR.restricted-dma-poolandmemory-regionas derekross asked, and the WLAN pins. If your commit also has the Bluetooth nodes, please drop them and base it on mine, or onomamobileonce mine has merged.omamobile-bt-addr, its unit and udev rule, and thebtbcmoption) come with my device-panther update, which follows the merges ahead of it.80-omamobile-bt-addr.rulesrule matcheshci[0-9]*, which also catches the per-connection devices (hci0:N). So every Bluetooth connection leaves a failedomamobile-bt-addr@hci0:Nunit; the script refuses them, so nothing changes. I'm preparing a fix for cheetah, raven and tegu (ENV{DEVTYPE}=="host", which only the controller has), and panther's copy will carry it.If you'd rather split it another way, say so, and I'll adapt.
On the Lishen pack: I'll write it into
device/panther/BATTERY-PLAN.md. That covers what your unit reads, why a power-on reset would load the wrong model today, and the two options derekross mentions (picking the model by the pack's ID resistor, as stock does, or not loading one for a pack we don't recognise). Please keep that unit from running flat until then.🤖 Generated with Claude Code
angelrenefm commentedon Nov 4, 2026
@nesmeck, agreed, thanks for laying it out. I've reshaped both of mine to match:
restricted-dma-poolwith the node'smemory-region, PCIe GEN4A_1, its PHY and System MMU window, and the WLAN pins). The Bluetooth nodes are gone. It sits onomamobilefor now; I'll rebase it onto your Bluetooth commit once that PR is open, or ontoomamobileonce yours has merged.omamobile-bt-addr, its unit and rule, and thebtbcmoption are out, for your device-panther update (with yourENV{DEVTYPE}=="host"fix).Both still pass
tools/kernel-pr-check(withdtbs_check) andtest/all. The Wi-Fi-only builds haven't been booted on the phone yet; what ran there had the Bluetooth nodes too, and the PRs say so.Thanks for taking the Lishen pack into BATTERY-PLAN. The unit won't run flat in the meantime.
I used AI help (Claude) to work on this.
nesmeck commentedon Nov 4, 2026
@angelrenefm, the Bluetooth commit is up as derekross/linux-omarchy-mobile#196, so #195 can sit on it. The
bt-addrrule fix is #423 (cheetah, raven and tegu). Panther's copy comes with my device-panther update.🤖 Generated with Claude Code
DISPLAY-PLAN (panther): D3 on a second Pixel 7, and device-panther 0.1.7
device-panther 0.1.6: AVB footers on boot and init_boot (#396)
DISPLAY-PLAN (panther): D3 on a second Pixel 7, and device-panther 0.1.7
derekross commentedon Nov 5, 2026
Thank you both for sorting out the split so quickly. It's exactly what I'd hoped for, and it makes the reviews much simpler. I've left a review on each pull request; here is the order I plan to merge them in, so it's in one place:
NV_PARTSand the second unit's notes) and device-panther: Wi-Fi, with the Pixel 7 Pro's files #419 (the Wi-Fi package) as 0.1.8 and 0.1.9 (0.1.7 stays reserved for device-panther 0.1.5: the screen sleeps with the display off, and the D3 plan #281), device-panther: Wi-Fi, with the Pixel 7 Pro's files #419 once it has its sleep change (details on the PR).@nesmeck, thanks for taking the Lishen pack into BATTERY-PLAN. Until the model is chosen by the pack's ID, that note is the only warning a Pixel 7 owner with that pack gets, so I'd like it in a pull request soon.
device-panther 0.1.6: AVB footers on boot and init_boot (derekross#396)