[cip-dev] [PATCH 00/10] Backport of watchdog core triggered keepalive infrastructure

[mosipov@ilbers.de (Maxim Yu. Osipov)]

Hello Ben,

This series contains 

* backport of patches of watchdog core infrastructure 
supporting handling watchdog keepalive. 

* imx21_wdt converted to use infrastructure triggered keepalives.

* backported support of WATCHDOG_HANDLE_BOOT_ENABLED option 
 
On some systems it's desirable to have watchdog reboot the system
when it does not come up fast enough. This adds a kernel parameter
to disable the auto-update of watchdog before userspace takes over
and a kernel option to set the default.

One of the important use cases is safe system update.
In that case, the bootloader enables the watchdog, and Linux should only
take it over at the point the whole Linux system is sure to be able to
have a fully working system again, including userspace services that
could take the next update. If we start the trigger the watchdog too
early, like it was so far, we risk to enter a system state where the
kernel works but not all other services we need.

i.MX6 based board was used as a test platform. 

We suppose that watchdog is enabled in bootloader and set to 120 seconds 
(maximum timeout supported by i.MX watchdog timer)

After applying these patches, built-in i.MX watchdog (imx21_wdt) into 
kernel and perform the following test cases:

1) option WATCHDOG_HANDLE_BOOT_ENABLED is turned off in kernel configuration
w/o userspace application triggering the watchdog. 

Make sure that no userspace application is triggering watchdog.

After around 120 seconds (timeout set in bootloader) board will reboot.

2) option WATCHDOG_HANDLE_BOOT_ENABLED is turned off in kernel configuration
but userspace application is triggering watchdog more frequently that watchdog's
timeout (by default set to 60 seconds for imx21_wdt).

Watchdog will not reboot the board until the application re-arms 
the watchdog.

3) option WATCHDOG_HANDLE_BOOT_ENABLED is turned on in kernel configuration
w/o userspace application triggering the watchdog. 

Make sure that no userspace application is triggering watchdog.
Board will not reboot (watchdog keepalive is supported by watchdog core 
infrastructure, not by driver itself)

Kind regards,
Maxim.

Guenter Roeck (6):
  watchdog: Introduce hardware maximum heartbeat in watchdog core
  watchdog: Introduce WDOG_HW_RUNNING flag
  watchdog: Make stop function optional
  watchdog: imx2: Convert to use infrastructure triggered keepalives
  watchdog: core: Fix circular locking dependency
  watchdog: core: Clear WDOG_HW_RUNNING before calling the stop function

Pratyush Anand (1):
  watchdog: skip min and max timeout validity check when
    max_hw_heartbeat_ms is defined

Rasmus Villemoes (1):
  watchdog: change watchdog_need_worker logic

Sebastian Reichel (1):
  watchdog: core: add option to avoid early handling of watchdog

Wei Yongjun (1):
  watchdog: core: Fix error handling of watchdog_dev_init()

 Documentation/watchdog/watchdog-kernel-api.txt |  51 +++++--
 drivers/watchdog/Kconfig                       |  11 ++
 drivers/watchdog/imx2_wdt.c                    |  74 ++-------
 drivers/watchdog/watchdog_core.c               |   4 +-
 drivers/watchdog/watchdog_dev.c                | 198 +++++++++++++++++++++++--
 include/linux/watchdog.h                       |  40 ++++-
 6 files changed, 279 insertions(+), 99 deletions(-)

-- 
2.11.0