Repository navigation
kernel config: the Pixel 9a's USB at boot, battery and charger, camera regulator, Wi-Fi host and speaker amplifiers - #227
Conversation
…s powered through A kernel built from omamobile.config booted the 9a without USB: the eUSB2 PHY was a module, and the PHY's rails come from the S2MPG14 PMIC through ACPM and the APM mailbox, also modules. The initramfs loads no modules, so USB never came up. With the four built in, a build of linux-omarchy-mobile#96 boots with USB networking and SSH.
Both drivers are merged in the kernel but were off in the fragment, so a kernel built from it had no battery or charger on the 9a. With them as modules, a boot of the charger device tree shows both supplies, and charge inhibit works.
|
Added a second commit: |
The main camera's supply chain includes an RT6160, whose driver was off in the fragment. With it, linux-omarchy-mobile#164's cameras power up and all three stream.
|
Third commit: |
linux-omarchy-mobile#168's pcie_zumapro, as a module: the device package's hang-safe start loads it, and built in it would probe at boot without the marker.
The tas2781 I2C driver drives the 9a's two TAS2572s; the device package's AoC start loads it once the AoC is online (linux-omarchy-mobile#162).
|
Two more commits, both tested on the palmtop today:
So the fragment now builds everything the 9a's tested rows need: USB, the battery and charger, the cameras, Wi-Fi and the speakers. |
|
Thank you for tracking this down. A kernel built from our own fragment booting the 9a without USB is exactly what this file should prevent, and the comparison against your integration config was a good way to find the four options. I built the config on today's |
…its boots (#240) The second Pixel 9a's record (the phone behind #74): a dated NOTES entry with slot A's state stated once and plainly (stock CP3A.261005.005, live images matching Google's by hash, a recovery boot verified, a full Android boot not), the corrected "not a qualified rollback" line and the `reboot-to bootloader` note, plus `device/tegu/kernel.config`, the fragment those boots ran. The maintainer marked the fragment bring-up only (its shared lines go once #227 lands; touch as a module is a test choice) and gave it the MIT licence line the other fragments carry. Closes #240 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Retitled to cover all five commits, with a line per commit at the top of the body. Thanks for booting the 7 Pro on it. |
|
Thanks for the retitle and the line per commit. The title now matches what lands, and the body reads well. Nothing more is needed from you. The one thing left is mine: building the ACPM path and the APM mailbox in changes when ACPM probes on the Pixel 7 Pro (its CPU clocks, thermal sensors and the AoC's rails go through it), so I'll boot my phone once on a kernel built from this fragment before merging. I'll post the result here. For later, not for this PR: h4rkl's #240 has merged, so |
|
Booted on my Pixel 7 Pro with this fragment (7.3.0-rc6-omamobile-push1-g291da0df9327, the phone's integration kernel plus #164's and #56's shared-driver commits): |
… 13 notes (#270) #227 added CONFIG_PCIE_ZUMAPRO=m, a symbol no published or integration kernel has (it belonged to kernel #168's driver, now superseded by #178's shared zuma-family host), so merge_config dropped it silently; remove it. #227 also made sec_acpm built in for the Pixel 9a's USB PHY. On the 7 Pro that registers the S2MPG13's GNSS rails before the regulator core's 30 s cleanup, and turns the G1 blacklist into a no-op. That is harmless (the rails are GNSS-only), but AGENTS and GNSS-PLAN described load timing the shared kernel can no longer give; they now say so. device/dodge/NOTES.md still called kernel #115 "in review"; it is closed and the kernel stays in the contributor's fork, as DEVICES says (#264). Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The shared config builds them in since derekross#227, for the Pixel 9a's USB PHY. Raven's initramfs loads them in order from INITRAMFS_MODULES, so with the shared config raven's initramfs no longer builds (no exynos-mailbox.ko), and every raven kernel so far has run them as modules. Built in, the regulator core's unused-rail cleanup would act on raven's S2MPG10/11 rails before their consumers load; moving raven to the built-in drivers is its own change, with its own boot.
Since #227 the shared kernel builds sec_acpm, acpm-protocol and the Exynos mailbox in, so raven's INITRAMFS_MODULES stopped "omamobile initramfs" with "no exynos-mailbox.ko" (#391). A module listed in the build's modules.builtin has nothing to load: it is reported and left out. A module that is neither a .ko nor built in still stops the build. The copy loop moves into pack_initramfs_modules so tests/initramfs-steps can run it on a fake modules tree, as it already does with the init's steps. Fixes #391 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The shared config builds them in since derekross#227, for the Pixel 9a's USB PHY. Raven's initramfs loads them in order from INITRAMFS_MODULES, so with the shared config raven's initramfs no longer builds (no exynos-mailbox.ko), and every raven kernel so far has run them as modules. Built in, the regulator core's unused-rail cleanup would act on raven's S2MPG10/11 rails before their consumers load; moving raven to the built-in drivers is its own change, with its own boot.
…its boots (derekross#240) The second Pixel 9a's record (the phone behind derekross#74): a dated NOTES entry with slot A's state stated once and plainly (stock CP3A.261005.005, live images matching Google's by hash, a recovery boot verified, a full Android boot not), the corrected "not a qualified rollback" line and the `reboot-to bootloader` note, plus `device/tegu/kernel.config`, the fragment those boots ran. The maintainer marked the fragment bring-up only (its shared lines go once derekross#227 lands; touch as a module is a test choice) and gave it the MIT licence line the other fragments carry. Closes derekross#240 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…a regulator, Wi-Fi host and speaker amplifiers (derekross#227) The Pixel 9a's USB at boot, battery and charger, camera and related symbols in the shared kernel fragment. Booted on the Pixel 7 Pro on 2026-11-01 (7.3.0-rc6-omamobile-push1-g291da0df9327): phone-smoke passes every row.
…e OnePlus 13 notes (derekross#270) derekross#227 added CONFIG_PCIE_ZUMAPRO=m, a symbol no published or integration kernel has (it belonged to kernel derekross#168's driver, now superseded by derekross#178's shared zuma-family host), so merge_config dropped it silently; remove it. derekross#227 also made sec_acpm built in for the Pixel 9a's USB PHY. On the 7 Pro that registers the S2MPG13's GNSS rails before the regulator core's 30 s cleanup, and turns the G1 blacklist into a no-op. That is harmless (the rails are GNSS-only), but AGENTS and GNSS-PLAN described load timing the shared kernel can no longer give; they now say so. device/dodge/NOTES.md still called kernel derekross#115 "in review"; it is closed and the kernel stays in the contributor's fork, as DEVICES says (derekross#264). Claude-Session: https://claude.ai/code/session_01BeEcdVeSfyLCbijF8XQF1d Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…rekross#393) Since derekross#227 the shared kernel builds sec_acpm, acpm-protocol and the Exynos mailbox in, so raven's INITRAMFS_MODULES stopped "omamobile initramfs" with "no exynos-mailbox.ko" (derekross#391). A module listed in the build's modules.builtin has nothing to load: it is reported and left out. A module that is neither a .ko nor built in still stops the build. The copy loop moves into pack_initramfs_modules so tests/initramfs-steps can run it on a fake modules tree, as it already does with the init's steps. Fixes derekross#391 Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The kernel config fragment gets the options the Pixel 9a needs, one commit each:
USB at boot
A kernel built from
kernel/configs/omamobile.configon top of defconfig, asbin/omamobile kerneldoes, boots the Pixel 9a without USB. It came up to the lock screen, but no gadget ever appeared on the host, so there was no SSH and no way back but the buttons.The cause is four options that defconfig builds as modules. The initramfs loads none (the descriptor's
INITRAMFS_MODULESis empty for tegu), and the fragment's own rule is "M1 essentials built in, so the initramfs needs no modules":CONFIG_PHY_SNPS_EUSB2: the 9a's USB 2.0 goes through a Synopsys eUSB2 PHY.CONFIG_EXYNOS_MBOX,CONFIG_EXYNOS_ACPM_PROTOCOL,CONFIG_MFD_SEC_ACPM: the PHY's rails come from the S2MPG14 PMIC, which Linux reaches only through ACPM over the APM mailbox. Without them the PHY waits for its supplies.I found these by comparing against the integration config the palmtop runs every day. Only these four were built in there and built as modules here, among the USB, PHY, PMIC, ACPM and Exynos options.
Tested on the palmtop with linux-omarchy-mobile#96's head (5628f188), built natively with the same steps as
bin/omamobile kernel: defconfig, then this fragment, thenolddefconfig. There is no 9a fragment.7.3.0-rc6-omamobile-push1-g5628f1888904.On that boot the display (DSI-1, 1080x2424) and the GPU (panthor render node) came up through the 9a package's graphics start. The evidence is on #96.
The other four
The testing rows behind each are in #244.
Other phones: the 7 Pro and the 6 Pro already worked with the first four as modules, and building them in costs them only a little kernel size. The ACPM pair is the same code their PMICs use. Once this lands, the same four lines in #240's
device/tegu/kernel.configcan go.