All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v10 0/5] Add support for AAEON SRG-IMX8P MCU
@ 2026-11-20  7:23 Thomas Perrot (Schneider Electric)
  2026-11-20  7:23 ` [PATCH v10 1/5] dt-bindings: vendor-prefixes: Add AAEON vendor prefix Thomas Perrot (Schneider Electric)
                   ` (4 more replies)
  0 siblings, 5 replies; 7+ messages in thread
From: Thomas Perrot (Schneider Electric) @ 2026-11-20  7:23 UTC (permalink / raw)
  To: Rob Herring, Krzysztof Kozlowski, Conor Dooley, Linus Walleij,
	Bartosz Golaszewski, Shawn Guo, Sascha Hauer,
	Pengutronix Kernel Team, Fabio Estevam,
	Jérémie Dautheribes, Wim Van Sebroeck, Guenter Roeck,
	Lee Jones, Frank Li, Jonathan Corbet, Shuah Khan, Randy Dunlap
  Cc: devicetree, linux-kernel, linux-gpio, imx, linux-arm-kernel,
	linux-watchdog, Thomas Petazzoni, Miquel Raynal, mfd, linux-doc,
	Thomas Perrot (Schneider Electric), Krzysztof Kozlowski,
	Conor Dooley, Bartosz Golaszewski

This patch series introduces support for the AAEON SRG-IMX8P embedded
controller (MCU). The MCU is connected via I2C and provides GPIO and
watchdog functionality for the SRG-IMX8P board.

The series includes:
- Device tree binding for the MFD driver
- MFD driver that serves as the core driver for the MCU
- GPIO driver implementing the GPIO functionality
- Watchdog driver for system monitoring
- MAINTAINERS entry for the new drivers

The drivers follow the standard Linux kernel subsystem patterns, with
the MFD driver registering the sub-devices (GPIO and watchdog) which
are then handled by their respective subsystem drivers.

Signed-off-by: Thomas Perrot (Schneider Electric) <thomas.perrot@bootlin.com>
---
Changes in v10:
- dt-bindings: send the AAEON vendor prefix patch again as patch 1; v9
  omitted it, leaving the "aaeon" compatible vendor undocumented
- mfd: reject reads whose register or value size is not the 2-byte
  address and 1-byte value of a MCU command, as writes already do
- mfd: replace the opcode shift and the watchdog status argument with
  named macros, and name the register, value and command lengths
- mfd: embed the two I2C transfer buffers, aligned to ARCH_DMA_MINALIGN,
  in struct aaeon_mcu instead of allocating them separately
- mfd: move the Kconfig entry, which sat between ABX500_CORE and
  AB8500_CORE and broke the menuconfig submenu of the latter
- mfd: include linux/module.h
- dt-bindings: define timeout-sec in the binding (maxItems and
  description) instead of a bare "true"; watchdog.yaml cannot be
  referenced from an embedded-controller node
- gpio: return 0 or 1 from .get() instead of the raw MCU byte, which
  gpiolib warns about
- gpio: use the shared read opcode instead of a duplicate define, and
  build the GPIO register addresses in a helper using MAX_GPOS
- watchdog: use the shared watchdog register macros
- watchdog: document the timeout parameter in watchdog-parameters.rst
- watchdog: trim the Kconfig help and add the module name line, as the
  other MFD-child watchdog drivers do
- watchdog: document that the suspend callback stops the MCU watchdog even
  with nowayout set, and say so in the commit message
- Link to v9: https://lore.kernel.org/r/20261029-dev-b4-aaeon-mcu-driver-v9-0-f9b6d0cc180d@bootlin.com

Changes in v9:
- mfd: hold the I2C segment lock across the MCU command write and the
  response read, so no other transfer on the same I2C segment can land
  in between; the STOP between the two is kept
- mfd: read the response through the driver's DMA-safe buffer instead of
  flagging the regmap caller's buffer as I2C_M_DMA_SAFE; reject writes
  that are not a single 3-byte command
- mfd: add .readable_reg so only the GPIO input and watchdog status
  registers can be read; any other address is a command that a read
  (e.g. through the regmap debugfs registers file) would execute
- mfd: drop "Multi-Function Device (MFD)" from the commit message and
  document the bus locking and the readable registers
- dt-bindings: document the timeout-sec property read by the watchdog
  driver and reword the watchdog description accordingly
- gpio: move GPIO_AAEON_MCU to the "MFD GPIO expanders" Kconfig menu
- watchdog: set min_timeout to the 25 second hardware heartbeat, as
  shorter timeouts cannot be honoured
- watchdog: honour CONFIG_WATCHDOG_NOWAYOUT with watchdog_set_nowayout()
- watchdog: fix the timeout comment and commit message (timeout module
  parameter, not a "watchdog_timeout" boot parameter); align the defines
  with tabs
- MAINTAINERS: keep the F: entries sorted
- Link to v8: https://lore.kernel.org/r/20260731-dev-b4-aaeon-mcu-driver-v8-0-9ab00e80df2a@bootlin.com

Changes in v8:
- mfd: add .use_single_read = true to regmap config; the MCU produces exactly
  one response byte per command and a bulk read passing val_size > 1 to the
  bus callback would be incorrect
- watchdog: call watchdog_stop_ping_on_suspend() in probe so the kthread ping
  worker is cancelled before platform devices are suspended, preventing I2C
  transfers on a suspended bus
- watchdog: check watchdog_hw_running() alongside watchdog_active() in the
  suspend/resume callbacks so watchdog is stopped before sleep and restarted
  on wake, preventing a spurious MCU reset if suspend exceeds the 25-second
  hardware timeout
- Link to v7: https://lore.kernel.org/r/20260707-dev-b4-aaeon-mcu-driver-v7-0-ca6c59abd672@bootlin.com

Changes in v7:
- mfd: fix select REGMAP Kconfig entry indented with spaces instead of a tab
- mfd: use devm_kcalloc() instead of devm_kzalloc() for the cmd buffer allocation
- mfd: add .use_single_write = true to the regmap config to prevent regcache sync
or debugfs from issuing multi-register writes that would overflow the fixed 3-byte
cmd buffer
- watchdog: remove max_timeout; it is documented as irrelevant when
max_hw_heartbeat_ms is set and is silently ignored by the watchdog core
- watchdog: pass dev->parent to watchdog_init_timeout() so that the timeout-sec
DT property is correctly read from the MCU I2C node; the MFD child platform device
has no of_node of its own
- watchdog: add suspend/resume PM hooks to stop the watchdog before entering sleep
and restart it on resume, preventing spurious resets if suspend lasts longer than
the 25-second hardware heartbeat
- Link to v6: https://lore.kernel.org/r/20260630-dev-b4-aaeon-mcu-driver-v6-0-d66b5fcbd2f0@bootlin.com

Changes in v6:
- mfd: rename local variable to ddata in probe
- mfd: fix driver name from "aaeon_mcu" to "aaeon-mcu"
- mfd: set I2C_M_DMA_SAFE on all i2c_msg flags so the host driver
  skips bounce-buffering the heap-allocated DMA-safe buffers
- mfd: drop COMPILE_TEST
- gpio: replace __set/__clear/__assign_bit with atomic set_bit/
  clear_bit/assign_bit to fix potential races on shared bitmaps
- gpio: write output value before switching pin to output mode to
  avoid a potential glitch on direction_output
- gpio: add MODULE_ALIAS("platform:aaeon-mcu-gpio")
- watchdog: add WDIOF_SETTIMEOUT and watchdog_init_timeout() so the
  software timeout is configurable via ioctl, DT timeout-sec or
  the watchdog_timeout boot parameter
- watchdog: add watchdog_stop_on_reboot() to prevent a spurious
  reset from the external MCU during system shutdown
- watchdog: add MODULE_ALIAS("platform:aaeon-mcu-wdt")
- Link to v5: https://lore.kernel.org/r/20260408-dev-b4-aaeon-mcu-driver-v5-0-ad98bd481668@bootlin.com

Changes in v5:
- mfd: use heap-allocated DMA-safe buffers for I2C transfers, replacing
  stack-allocated buffers in the regmap bus callbacks
- mfd: switch from REGCACHE_NONE to REGCACHE_MAPLE; add volatile_reg
  callback marking GPIO input read registers (opcode 0x72) as volatile;
  add max_register
- mfd: use PLATFORM_DEVID_AUTO instead of PLATFORM_DEVID_NONE
- mfd: use MFD_CELL_BASIC() macro for cell definitions
- mfd: use dev_err_probe() for regmap initialization error
- Link to v4: https://lore.kernel.org/r/20260324-dev-b4-aaeon-mcu-driver-v4-0-afb011df4794@bootlin.com

Changes in v4:
- mfd: switch to a custom regmap bus; remove aaeon_mcu_i2c_xfer() and the aaeon_mcu_dev struct
- mfd: locking delegated to regmap's built-in mutex; drop explicit mutex
- mfd: remove firmware version reading at probe time
- gpio, watchdog: use regmap_read()/regmap_write() via dev_get_regmap()
- include: replace aaeon_mcu_i2c_xfer() declaration with AAEON_MCU_REG() macro
- dt-bindings: remove unused label from example node
- Link to v3: https://lore.kernel.org/r/20260203-dev-b4-aaeon-mcu-driver-v3-0-0a19432076ac@bootlin.com

Changes in v3:
- Renamed SRG-IMX8PL to SRG-IMX8P
- dt-bindings: add gpio-controller properties as required
- mfd: move struct aaeon_mcu_dev from header to .c file (private)
- mfd: use guard(mutex) and devm_mutex_init() for cleanup
- mfd: firmware version log changed to dev_dbg()
- mfd: add select MFD_CORE to Kconfig
- Kconfig: add || COMPILE_TEST to all three drivers
- watchdog: add comments explaining hardware timeout and WDOG_HW_RUNNING
- watchdog: remove unused platform_set_drvdata()
- watchdog: add a function to query the status
- Link to v2: https://lore.kernel.org/r/20260123-dev-b4-aaeon-mcu-driver-v2-0-9f4c00bfb5cb@bootlin.com

Changes in v2:
- Fold GPIO and watchdog bindings into MFD binding
- Drop OF_GPIO dependency in GPIO Kconfig
- Use __set_bit/__clear_bit/__assign_bit instead of atomic variants
- Various driver cleanups and improvements
- Link to v1: https://lore.kernel.org/r/20251212-dev-b4-aaeon-mcu-driver-v1-0-6bd65bc8ef12@bootlin.com

---
Thomas Perrot (Schneider Electric) (5):
      dt-bindings: vendor-prefixes: Add AAEON vendor prefix
      dt-bindings: mfd: Add AAEON embedded controller
      mfd: aaeon: Add SRG-IMX8P MCU driver
      gpio: aaeon: Add GPIO driver for SRG-IMX8P MCU
      watchdog: aaeon: Add watchdog driver for SRG-IMX8P MCU

 .../bindings/mfd/aaeon,srg-imx8p-mcu.yaml          |  73 +++++++
 .../devicetree/bindings/vendor-prefixes.yaml       |   2 +
 Documentation/watchdog/watchdog-parameters.rst     |   8 +
 MAINTAINERS                                        |  10 +
 drivers/gpio/Kconfig                               |   9 +
 drivers/gpio/Makefile                              |   1 +
 drivers/gpio/gpio-aaeon-mcu.c                      | 235 +++++++++++++++++++++
 drivers/mfd/Kconfig                                |  11 +
 drivers/mfd/Makefile                               |   1 +
 drivers/mfd/aaeon-mcu.c                            | 227 ++++++++++++++++++++
 drivers/watchdog/Kconfig                           |  11 +
 drivers/watchdog/Makefile                          |   1 +
 drivers/watchdog/aaeon_mcu_wdt.c                   | 174 +++++++++++++++
 include/linux/mfd/aaeon-mcu.h                      |  46 ++++
 14 files changed, 809 insertions(+)
---
base-commit: aebb7274476403040d582c61478df8f4a7bf9cd3
change-id: 20251211-dev-b4-aaeon-mcu-driver-e0e89ebf4afb

Best regards,
-- 
Thomas Perrot (Schneider Electric) <thomas.perrot@bootlin.com>


^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2026-11-20  7:36 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-20  7:23 [PATCH v10 0/5] Add support for AAEON SRG-IMX8P MCU Thomas Perrot (Schneider Electric)
2026-11-20  7:23 ` [PATCH v10 1/5] dt-bindings: vendor-prefixes: Add AAEON vendor prefix Thomas Perrot (Schneider Electric)
2026-11-20  7:23 ` [PATCH v10 2/5] dt-bindings: mfd: Add AAEON embedded controller Thomas Perrot (Schneider Electric)
2026-11-20  7:23 ` [PATCH v10 3/5] mfd: aaeon: Add SRG-IMX8P MCU driver Thomas Perrot (Schneider Electric)
2026-11-20  7:36   ` sashiko-bot
2026-11-20  7:23 ` [PATCH v10 4/5] gpio: aaeon: Add GPIO driver for SRG-IMX8P MCU Thomas Perrot (Schneider Electric)
2026-11-20  7:23 ` [PATCH v10 5/5] watchdog: aaeon: Add watchdog " Thomas Perrot (Schneider Electric)

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.