linux/Documentation/watchdog/watchdog-kernel-api.txt
Hans de Goede e907df3272 watchdog: Add support for dynamically allocated watchdog_device structs
If a driver's watchdog_device struct is part of a dynamically allocated
struct (which it often will be), merely locking the module is not enough,
even with a drivers module locked, the driver can be unbound from the device,
examples:
1) The root user can unbind it through sysfd
2) The i2c bus master driver being unloaded for an i2c watchdog

I will gladly admit that these are corner cases, but we still need to handle
them correctly.

The fix for this consists of 2 parts:
1) Add ref / unref operations, so that the driver can refcount the struct
   holding the watchdog_device struct and delay freeing it until any
   open filehandles referring to it are closed
2) Most driver operations will do IO on the device and the driver should not
   do any IO on the device after it has been unbound. Rather then letting each
   driver deal with this internally, it is better to ensure at the watchdog
   core level that no operations (other then unref) will get called after
   the driver has called watchdog_unregister_device(). This actually is the
   bulk of this patch.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
2012-05-30 07:55:31 +02:00

215 lines
11 KiB
Plaintext

The Linux WatchDog Timer Driver Core kernel API.
===============================================
Last reviewed: 22-May-2012
Wim Van Sebroeck <wim@iguana.be>
Introduction
------------
This document does not describe what a WatchDog Timer (WDT) Driver or Device is.
It also does not describe the API which can be used by user space to communicate
with a WatchDog Timer. If you want to know this then please read the following
file: Documentation/watchdog/watchdog-api.txt .
So what does this document describe? It describes the API that can be used by
WatchDog Timer Drivers that want to use the WatchDog Timer Driver Core
Framework. This framework provides all interfacing towards user space so that
the same code does not have to be reproduced each time. This also means that
a watchdog timer driver then only needs to provide the different routines
(operations) that control the watchdog timer (WDT).
The API
-------
Each watchdog timer driver that wants to use the WatchDog Timer Driver Core
must #include <linux/watchdog.h> (you would have to do this anyway when
writing a watchdog device driver). This include file contains following
register/unregister routines:
extern int watchdog_register_device(struct watchdog_device *);
extern void watchdog_unregister_device(struct watchdog_device *);
The watchdog_register_device routine registers a watchdog timer device.
The parameter of this routine is a pointer to a watchdog_device structure.
This routine returns zero on success and a negative errno code for failure.
The watchdog_unregister_device routine deregisters a registered watchdog timer
device. The parameter of this routine is the pointer to the registered
watchdog_device structure.
The watchdog device structure looks like this:
struct watchdog_device {
int id;
struct cdev cdev;
struct device *dev;
struct device *parent;
const struct watchdog_info *info;
const struct watchdog_ops *ops;
unsigned int bootstatus;
unsigned int timeout;
unsigned int min_timeout;
unsigned int max_timeout;
void *driver_data;
struct mutex lock;
unsigned long status;
};
It contains following fields:
* id: set by watchdog_register_device, id 0 is special. It has both a
/dev/watchdog0 cdev (dynamic major, minor 0) as well as the old
/dev/watchdog miscdev. The id is set automatically when calling
watchdog_register_device.
* cdev: cdev for the dynamic /dev/watchdog<id> device nodes. This
field is also populated by watchdog_register_device.
* dev: device under the watchdog class (created by watchdog_register_device).
* parent: set this to the parent device (or NULL) before calling
watchdog_register_device.
* info: a pointer to a watchdog_info structure. This structure gives some
additional information about the watchdog timer itself. (Like it's unique name)
* ops: a pointer to the list of watchdog operations that the watchdog supports.
* timeout: the watchdog timer's timeout value (in seconds).
* min_timeout: the watchdog timer's minimum timeout value (in seconds).
* max_timeout: the watchdog timer's maximum timeout value (in seconds).
* bootstatus: status of the device after booting (reported with watchdog
WDIOF_* status bits).
* driver_data: a pointer to the drivers private data of a watchdog device.
This data should only be accessed via the watchdog_set_drvdata and
watchdog_get_drvdata routines.
* lock: Mutex for WatchDog Timer Driver Core internal use only.
* status: this field contains a number of status bits that give extra
information about the status of the device (Like: is the watchdog timer
running/active, is the nowayout bit set, is the device opened via
the /dev/watchdog interface or not, ...).
The list of watchdog operations is defined as:
struct watchdog_ops {
struct module *owner;
/* mandatory operations */
int (*start)(struct watchdog_device *);
int (*stop)(struct watchdog_device *);
/* optional operations */
int (*ping)(struct watchdog_device *);
unsigned int (*status)(struct watchdog_device *);
int (*set_timeout)(struct watchdog_device *, unsigned int);
unsigned int (*get_timeleft)(struct watchdog_device *);
void (*ref)(struct watchdog_device *);
void (*unref)(struct watchdog_device *);
long (*ioctl)(struct watchdog_device *, unsigned int, unsigned long);
};
It is important that you first define the module owner of the watchdog timer
driver's operations. This module owner will be used to lock the module when
the watchdog is active. (This to avoid a system crash when you unload the
module and /dev/watchdog is still open).
If the watchdog_device struct is dynamically allocated, just locking the module
is not enough and a driver also needs to define the ref and unref operations to
ensure the structure holding the watchdog_device does not go away.
The simplest (and usually sufficient) implementation of this is to:
1) Add a kref struct to the same structure which is holding the watchdog_device
2) Define a release callback for the kref which frees the struct holding both
3) Call kref_init on this kref *before* calling watchdog_register_device()
4) Define a ref operation calling kref_get on this kref
5) Define a unref operation calling kref_put on this kref
6) When it is time to cleanup:
* Do not kfree() the struct holding both, the last kref_put will do this!
* *After* calling watchdog_unregister_device() call kref_put on the kref
Some operations are mandatory and some are optional. The mandatory operations
are:
* start: this is a pointer to the routine that starts the watchdog timer
device.
The routine needs a pointer to the watchdog timer device structure as a
parameter. It returns zero on success or a negative errno code for failure.
* stop: with this routine the watchdog timer device is being stopped.
The routine needs a pointer to the watchdog timer device structure as a
parameter. It returns zero on success or a negative errno code for failure.
Some watchdog timer hardware can only be started and not be stopped. The
driver supporting this hardware needs to make sure that a start and stop
routine is being provided. This can be done by using a timer in the driver
that regularly sends a keepalive ping to the watchdog timer hardware.
Not all watchdog timer hardware supports the same functionality. That's why
all other routines/operations are optional. They only need to be provided if
they are supported. These optional routines/operations are:
* ping: this is the routine that sends a keepalive ping to the watchdog timer
hardware.
The routine needs a pointer to the watchdog timer device structure as a
parameter. It returns zero on success or a negative errno code for failure.
Most hardware that does not support this as a separate function uses the
start function to restart the watchdog timer hardware. And that's also what
the watchdog timer driver core does: to send a keepalive ping to the watchdog
timer hardware it will either use the ping operation (when available) or the
start operation (when the ping operation is not available).
(Note: the WDIOC_KEEPALIVE ioctl call will only be active when the
WDIOF_KEEPALIVEPING bit has been set in the option field on the watchdog's
info structure).
* status: this routine checks the status of the watchdog timer device. The
status of the device is reported with watchdog WDIOF_* status flags/bits.
* set_timeout: this routine checks and changes the timeout of the watchdog
timer device. It returns 0 on success, -EINVAL for "parameter out of range"
and -EIO for "could not write value to the watchdog". On success this
routine should set the timeout value of the watchdog_device to the
achieved timeout value (which may be different from the requested one
because the watchdog does not necessarily has a 1 second resolution).
(Note: the WDIOF_SETTIMEOUT needs to be set in the options field of the
watchdog's info structure).
* get_timeleft: this routines returns the time that's left before a reset.
* ref: the operation that calls kref_get on the kref of a dynamically
allocated watchdog_device struct.
* unref: the operation that calls kref_put on the kref of a dynamically
allocated watchdog_device struct.
* ioctl: if this routine is present then it will be called first before we do
our own internal ioctl call handling. This routine should return -ENOIOCTLCMD
if a command is not supported. The parameters that are passed to the ioctl
call are: watchdog_device, cmd and arg.
The status bits should (preferably) be set with the set_bit and clear_bit alike
bit-operations. The status bits that are defined are:
* WDOG_ACTIVE: this status bit indicates whether or not a watchdog timer device
is active or not. When the watchdog is active after booting, then you should
set this status bit (Note: when you register the watchdog timer device with
this bit set, then opening /dev/watchdog will skip the start operation)
* WDOG_DEV_OPEN: this status bit shows whether or not the watchdog device
was opened via /dev/watchdog.
(This bit should only be used by the WatchDog Timer Driver Core).
* WDOG_ALLOW_RELEASE: this bit stores whether or not the magic close character
has been sent (so that we can support the magic close feature).
(This bit should only be used by the WatchDog Timer Driver Core).
* WDOG_NO_WAY_OUT: this bit stores the nowayout setting for the watchdog.
If this bit is set then the watchdog timer will not be able to stop.
* WDOG_UNREGISTERED: this bit gets set by the WatchDog Timer Driver Core
after calling watchdog_unregister_device, and then checked before calling
any watchdog_ops, so that you can be sure that no operations (other then
unref) will get called after unregister, even if userspace still holds a
reference to /dev/watchdog
To set the WDOG_NO_WAY_OUT status bit (before registering your watchdog
timer device) you can either:
* set it statically in your watchdog_device struct with
.status = WATCHDOG_NOWAYOUT_INIT_STATUS,
(this will set the value the same as CONFIG_WATCHDOG_NOWAYOUT) or
* use the following helper function:
static inline void watchdog_set_nowayout(struct watchdog_device *wdd, int nowayout)
Note: The WatchDog Timer Driver Core supports the magic close feature and
the nowayout feature. To use the magic close feature you must set the
WDIOF_MAGICCLOSE bit in the options field of the watchdog's info structure.
The nowayout feature will overrule the magic close feature.
To get or set driver specific data the following two helper functions should be
used:
static inline void watchdog_set_drvdata(struct watchdog_device *wdd, void *data)
static inline void *watchdog_get_drvdata(struct watchdog_device *wdd)
The watchdog_set_drvdata function allows you to add driver specific data. The
arguments of this function are the watchdog device where you want to add the
driver specific data to and a pointer to the data itself.
The watchdog_get_drvdata function allows you to retrieve driver specific data.
The argument of this function is the watchdog device where you want to retrieve
data from. The function returns the pointer to the driver specific data.