DT510: BQ25792 battery charger — kernel driver + IRQ/validation (Tier B1)

[ajlennon]

Parent tracking

Sub-issue of #2 (DT510 BSP bring-up plan).

Scope (Tier B1)

Bring the BQ25792 battery charger path from “DT ready, kernel missing” to a working, testable state on linux-lmp-fslc-imx / 6.6.x-class kernels used by LmP.

Current state (from project plan)

• Device tree: battery-dt510 + bq25792@6b okay on I2C3 0x6B; monitored-battery set.
• Kernel: Upstream charger driver for ti,bq25792 is not present in 6.6.x — no CONFIG_* fragment yet; lab validation via i2cget / i2cdetect on i2c-3.
• IRQ: CHGR_INT# pending SSOT / DTS wiring when confirmed.

Deliverables

• Choose integration path: backport in-tree driver, kernel advance to a tree that has ti,bq25792, or documented out-of-tree module — aligned with OTA/boot risk.
• Add CONFIG_* / machine fragment so the driver is built and loaded on DT510 images.
• Wire CHGR_INT# in DTS when SSOT confirms (or document “polled-only” interim if product accepts).
• Acceptance: charger visible in sysfs / power supply ABI (as applicable), basic charge/read paths validated on hardware; no regressions to boot or unrelated subsystems.

References

• docs/DT510-BSP-PROJECT-PLAN.md — Tier B1
• docs/DT510-HARDWARE-AUDIT-CHECKLIST.md

Assignees

Per parent #2: @ajlennon @ohull456

[ajlennon]

Update (2026-04-14) — factory kernel identity + prerequisite commit test

Kernel we are actually using (important)

• Recipe: linux-lmp-fslc-imx (Foundries LmP / NXP linux-fslc line).
• Upstream git: Freescale/linux-fslc branch 6.6-2.2.x-imx.
• Exact revision is not a guess from docs: meta-lmp pins SRCREV_machine. Confirm on any machine with:

  bitbake -e linux-lmp-fslc-imx | grep '^SRCREV'

  (Do not rely on an example SHA from an old plan line alone.)

Prerequisite before the BQ25792 patch series

The lore “[v6] Add support for TI BQ25792” patches are written on top of the mainline BQ25703A / bq257xx stack. That stack entered mainline as 3b1bbfb5fce3ca9fffc92ac1b053b0cfbb1f322b — mfd: bq257xx: Add support for BQ25703A core driver (torvalds/linux).

Stock linux-fslc 6.6.x does not contain drivers/mfd/bq257xx.c at all — so the BQ25792 series cannot be applied alone.

What we tested

Cherry-pick of 3b1bbfb5fce3 onto linux-fslc 6.6-2.2.x-imx tip 97812d71 (reported LINUX_VERSION 6.6.129 at test time — your factory SRCREV may differ):

• Result: Mostly applies (adds MFD core + headers); one merge conflict in drivers/mfd/Makefile (straightforward: add the obj-$(CONFIG_MFD_BQ257XX) line in the right place). After that, still need the rest of the same mainline merge (charger, regulator, power-supply Kconfig/Makefile, bindings) — usually taken as the full patch set from that merge, not only the MFD commit.

Repo notes (BSP)

• recipes-kernel/linux/linux-lmp-fslc-imx/imx8mm-jaguar-dt510/bq25792-charger.cfg — expanded with kernel identity + test notes.
• docs/DT510-BSP-PROJECT-PLAN.md Tier B1 + docs/DT510-HARDWARE-AUDIT-CHECKLIST.md — aligned with the above.

Next integration steps (unchanged scope, clearer order)

1. Pin and document actual SRCREV from bitbake for the DT510 factory line.
2. Backport full bq257xx (BQ25703A) stack from mainline, then apply BQ25792 delta + DT/bindings as needed.
3. Enable CONFIG_MFD_BQ257XX / CONFIG_CHARGER_BQ257XX in the DT510 fragment when the code is in-tree.
4. Hardware validation: sysfs / power_supply ABI on DT510 once the driver probes (ti,bq25792 + IRQ).

cc @ohull456

[ajlennon]

Merged to main — commit e60e302

What landed

• Patch: recipes-kernel/linux/linux-lmp-fslc-imx/imx8mm-jaguar-dt510/0010-mfd-bq257xx-add-bq25703a-core-fslc.patch — BQ25703A MFD core from mainline 3b1bbfb5fce3 with linux-fslc 6.6 Kconfig / Makefile context (no CGBC line).
• Kconfig fragment: bq257xx-mfd-kconfig.cfg → CONFIG_MFD_BQ257XX=m when machine feature is on.
• Machine: MACHINE_FEATURES includes bq25792-charger on imx8mm-jaguar-dt510 (remove feature to skip patch).
• Recipe: linux-lmp-fslc-imx_%.bbappend appends patch + cfg only when bq25792-charger is set.
• Docs / DT: project plan, hardware checklist, bq25792-charger.cfg, DTS updates aligned with Tier B1.

Super-repo

• meta-dynamicdevices bumped submodule to e60e302 (2cdd6d73 on meta-dynamicdevices main).

Foundries

After meta-subscriber-overrides syncs this BSP revision on main-imx8mm-jaguar-dt510, CI should rebuild the kernel with the patch applied. Watch the factory pipeline for virtual/kernel / compile of drivers/mfd/bq257xx.c.

Note: MFD still matches ti,bq25703a only; DT uses ti,bq25792 — further driver/OF work needed for bind. Child charger/regulator drivers not in this commit.

[ajlennon]

Foundries CI triggered (vixdt)

Empty commit pushed to source.foundries.io/factories/vixdt/meta-subscriber-overrides branch main-imx8mm-jaguar-dt510 to start a build after BSP main at e60e302.

• CI job: https://app.foundries.io/factories/vixdt/targets/152/

Confirm in the job logs that linux-lmp-fslc-imx applies 0010-mfd-bq257xx-add-bq25703a-core-fslc.patch and virtual/kernel completes.

[ajlennon]

Update: 0011–0012 landed (c29fa6f)

The stack that had to exist before the BQ25792-specific lore series is now in main:

Patch	Upstream	Contents
0010	3b1bbfb5fce3	MFD bq257xx.c + bq257xx.h
0011	1cc017b7f9c7	drivers/power/supply/bq257xx_charger.c + CHARGER_BQ257XX
0012	981dd162b635	drivers/regulator/bq257xx-regulator.c + REGULATOR_BQ257XX (Kconfig/Makefile hunks adjusted for linux-fslc 6.6)

linux-lmp-fslc-imx_%.bbappend applies 0010 → 0011 → 0012 in order when bq25792-charger is set. bq257xx-mfd-kconfig.cfg now sets CONFIG_MFD_BQ257XX=m, CONFIG_CHARGER_BQ257XX=m, CONFIG_REGULATOR_BQ257XX=m.

BQ25792 variant (different register map, ti,bq25792, binding YAML): still need to add the lore [PATCH v6 11/11] series as follow-on patches (0013+) when you have the mbox — not vendored here (not on torvalds master yet from this environment).

Verified: git am all three on Freescale/linux-fslc 6.6-2.2.x-imx@97812d71 succeeds.