summaryrefslogtreecommitdiffstats
path: root/Documentation/user/watchdog.rst
blob: 0220965598edec0b99817a27015db1f52862bb67 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
Watchdog Support
================

Barebox Watchdog Functionality
------------------------------

In some cases we are not able to influence the hardware design anymore or while
developing one needs to be able to feed the watchdog to disable it from within
the bootloader. For these scenarios barebox provides the watchdog framework
with the following functionality and at least ``CONFIG_WATCHDOG`` should be
enabled.

Disabling for development
~~~~~~~~~~~~~~~~~~~~~~~~~

The shorthand command ``wd -x`` will disable all watchdogs.
If hardware (or driver) doesn't support turning off the watchdog,
an autpoller will be registered to periodically feed watchdogs.
This should only be needed for development.
See :ref:`boot-watchdog-timeout` for how to use the watchdog in the field.

Polling
~~~~~~~

Watchdog polling/feeding allows to feed the watchdog and keep it running on one
side and to not reset the system on the other side. It is needed on hardware
with short-time watchdogs. For example the Atheros ar9331 watchdog has a
maximal timeout of 7 seconds, so it may reset even on netboot.
Or it can be used on systems where the watchdog is already running and can't be
disabled, an example for that is the watchdog of the i.MX2 series.
This functionally can be seen as a threat, since in error cases barebox will
continue to feed the watchdog even if that is not desired. So, depending on
your needs ``CONFIG_WATCHDOG_POLLER`` can be enabled or disabled at compile
time. Even if barebox was built with watchdog polling support, it is not
enabled by default. To start polling from command line run:

.. code-block:: console

  wdog0.autoping=1

**NOTE** Using this feature might have the effect that the watchdog is
effectively disabled. In case barebox is stuck in a loop that includes feeding
the watchdog, then the watchdog will never trigger. Only use this feature
during development or when a bad watchdog design (Short watchdog timeout
enabled as boot default) doesn't give you another choice.

The poller interval is not configurable, but fixed at 500ms and the watchdog
timeout is configured by default to the maximum of the supported values by
hardware. To change the timeout used by the poller, run:

.. code-block:: console

  wdog0.timeout_cur=7

To read the current watchdog's configuration, run:

.. code-block:: console

  devinfo wdog0

The output may look as follows where ``timeout_cur`` and ``timeout_max`` are
measured in seconds:

.. code-block:: console

  barebox@DPTechnics DPT-Module:/ devinfo wdog0
  Parameters:
    autoping: 1 (type: bool)
    priority: 100 (type: uint32)
    running: 1 (type: enum) (values: "unknown", "1", "0")
    seconds_to_expire: 7 (type: uint32)
    timeout_cur: 7 (type: uint32)
    timeout_max: 10 (type: uint32)

Use barebox' environment to persist these changes between reboots:

.. code-block:: console

  nv dev.wdog0.autoping=1
  nv dev.wdog0.timeout_cur=7

Watchdog State
~~~~~~~~~~~~~~

To check whether a watchdog is currently running, the ``running`` device
parameter can be consulted. Not all watchdog devices (or their drivers)
provide the information whether a watchdog was running prior to barebox.
In that case, the parameter will contain the value ``unknown``.

Watchdogs started by barebox can be monitored using the
``seconds_to_expire`` parameter. A well-behaving system of watchdog
device, watchdog driver and clocksource should reset as soon as the
count down reaches zero.

To manually start a watchdog, :ref:`command_wd` can be used.

Default Watchdog
~~~~~~~~~~~~~~~~

barebox supports multiple concurrent watchdogs. The default watchdog used
with :ref:`command_wd`, ``boot.watchdog_timeout`` and :ref:`command_boot`'s
``-w`` option is the one with the highest positive priority.
If multiple watchdogs share the same priority, only one will be affected.

The priority is initially set by drivers and can be overridden in the
device tree or via the ``priority`` device parameter. Normally, watchdogs
that have a wider effect should be given the higher priority (e.g.
PMIC watchdog resetting the board vs. SoC's watchdog resetting only itself).

.. _boot-watchdog-timeout:

Boot Watchdog Timeout
~~~~~~~~~~~~~~~~~~~~~

With this functionality barebox may start a watchdog or update the timeout of
an already-running one, just before kicking the boot image. It can be
configured temporarily via

.. code-block:: console

  global boot.watchdog_timeout=10

or persistently by

.. code-block:: console

  nv boot.watchdog_timeout=10

where the used value again is measured in seconds.
Only the default watchdog will be started.