1Everything you never wanted to know about kobjects, ksets, and ktypes
   3Greg Kroah-Hartman <>
   5Based on an original article by Jon Corbet for written October 1,
   62003 and located at
   8Last updated December 19, 2007
  11Part of the difficulty in understanding the driver model - and the kobject
  12abstraction upon which it is built - is that there is no obvious starting
  13place. Dealing with kobjects requires understanding a few different types,
  14all of which make reference to each other. In an attempt to make things
  15easier, we'll take a multi-pass approach, starting with vague terms and
  16adding detail as we go. To that end, here are some quick definitions of
  17some terms we will be working with.
  19 - A kobject is an object of type struct kobject.  Kobjects have a name
  20   and a reference count.  A kobject also has a parent pointer (allowing
  21   objects to be arranged into hierarchies), a specific type, and,
  22   usually, a representation in the sysfs virtual filesystem.
  24   Kobjects are generally not interesting on their own; instead, they are
  25   usually embedded within some other structure which contains the stuff
  26   the code is really interested in.
  28   No structure should EVER have more than one kobject embedded within it.
  29   If it does, the reference counting for the object is sure to be messed
  30   up and incorrect, and your code will be buggy.  So do not do this.
  32 - A ktype is the type of object that embeds a kobject.  Every structure
  33   that embeds a kobject needs a corresponding ktype.  The ktype controls
  34   what happens to the kobject when it is created and destroyed.
  36 - A kset is a group of kobjects.  These kobjects can be of the same ktype
  37   or belong to different ktypes.  The kset is the basic container type for
  38   collections of kobjects. Ksets contain their own kobjects, but you can
  39   safely ignore that implementation detail as the kset core code handles
  40   this kobject automatically.
  42   When you see a sysfs directory full of other directories, generally each
  43   of those directories corresponds to a kobject in the same kset.
  45We'll look at how to create and manipulate all of these types. A bottom-up
  46approach will be taken, so we'll go back to kobjects.
  49Embedding kobjects
  51It is rare for kernel code to create a standalone kobject, with one major
  52exception explained below.  Instead, kobjects are used to control access to
  53a larger, domain-specific object.  To this end, kobjects will be found
  54embedded in other structures.  If you are used to thinking of things in
  55object-oriented terms, kobjects can be seen as a top-level, abstract class
  56from which other classes are derived.  A kobject implements a set of
  57capabilities which are not particularly useful by themselves, but which are
  58nice to have in other objects.  The C language does not allow for the
  59direct expression of inheritance, so other techniques - such as structure
  60embedding - must be used.
  62So, for example, the UIO code has a structure that defines the memory
  63region associated with a uio device:
  65struct uio_mem {
  66        struct kobject kobj;
  67        unsigned long addr;
  68        unsigned long size;
  69        int memtype;
  70        void __iomem *internal_addr;
  73If you have a struct uio_mem structure, finding its embedded kobject is
  74just a matter of using the kobj member.  Code that works with kobjects will
  75often have the opposite problem, however: given a struct kobject pointer,
  76what is the pointer to the containing structure?  You must avoid tricks
  77(such as assuming that the kobject is at the beginning of the structure)
  78and, instead, use the container_of() macro, found in <linux/kernel.h>:
  80        container_of(pointer, type, member)
  82where pointer is the pointer to the embedded kobject, type is the type of
  83the containing structure, and member is the name of the structure field to
  84which pointer points.  The return value from container_of() is a pointer to
  85the given type. So, for example, a pointer "kp" to a struct kobject
  86embedded within a struct uio_mem could be converted to a pointer to the
  87containing uio_mem structure with:
  89    struct uio_mem *u_mem = container_of(kp, struct uio_mem, kobj);
  91Programmers often define a simple macro for "back-casting" kobject pointers
  92to the containing type.
  95Initialization of kobjects
  97Code which creates a kobject must, of course, initialize that object. Some
  98of the internal fields are setup with a (mandatory) call to kobject_init():
 100    void kobject_init(struct kobject *kobj, struct kobj_type *ktype);
 102The ktype is required for a kobject to be created properly, as every kobject
 103must have an associated kobj_type.  After calling kobject_init(), to
 104register the kobject with sysfs, the function kobject_add() must be called:
 106    int kobject_add(struct kobject *kobj, struct kobject *parent, const char *fmt, ...);
 108This sets up the parent of the kobject and the name for the kobject
 109properly.  If the kobject is to be associated with a specific kset,
 110kobj->kset must be assigned before calling kobject_add().  If a kset is
 111associated with a kobject, then the parent for the kobject can be set to
 112NULL in the call to kobject_add() and then the kobject's parent will be the
 113kset itself.
 115As the name of the kobject is set when it is added to the kernel, the name
 116of the kobject should never be manipulated directly.  If you must change
 117the name of the kobject, call kobject_rename():
 119    int kobject_rename(struct kobject *kobj, const char *new_name);
 121kobject_rename does not perform any locking or have a solid notion of
 122what names are valid so the caller must provide their own sanity checking
 123and serialization.
 125There is a function called kobject_set_name() but that is legacy cruft and
 126is being removed.  If your code needs to call this function, it is
 127incorrect and needs to be fixed.
 129To properly access the name of the kobject, use the function
 132    const char *kobject_name(const struct kobject * kobj);
 134There is a helper function to both initialize and add the kobject to the
 135kernel at the same time, called supprisingly enough kobject_init_and_add():
 137    int kobject_init_and_add(struct kobject *kobj, struct kobj_type *ktype,
 138                             struct kobject *parent, const char *fmt, ...);
 140The arguments are the same as the individual kobject_init() and
 141kobject_add() functions described above.
 146After a kobject has been registered with the kobject core, you need to
 147announce to the world that it has been created.  This can be done with a
 148call to kobject_uevent():
 150    int kobject_uevent(struct kobject *kobj, enum kobject_action action);
 152Use the KOBJ_ADD action for when the kobject is first added to the kernel.
 153This should be done only after any attributes or children of the kobject
 154have been initialized properly, as userspace will instantly start to look
 155for them when this call happens.
 157When the kobject is removed from the kernel (details on how to do that is
 158below), the uevent for KOBJ_REMOVE will be automatically created by the
 159kobject core, so the caller does not have to worry about doing that by
 163Reference counts
 165One of the key functions of a kobject is to serve as a reference counter
 166for the object in which it is embedded. As long as references to the object
 167exist, the object (and the code which supports it) must continue to exist.
 168The low-level functions for manipulating a kobject's reference counts are:
 170    struct kobject *kobject_get(struct kobject *kobj);
 171    void kobject_put(struct kobject *kobj);
 173A successful call to kobject_get() will increment the kobject's reference
 174counter and return the pointer to the kobject.
 176When a reference is released, the call to kobject_put() will decrement the
 177reference count and, possibly, free the object. Note that kobject_init()
 178sets the reference count to one, so the code which sets up the kobject will
 179need to do a kobject_put() eventually to release that reference.
 181Because kobjects are dynamic, they must not be declared statically or on
 182the stack, but instead, always allocated dynamically.  Future versions of
 183the kernel will contain a run-time check for kobjects that are created
 184statically and will warn the developer of this improper usage.
 186If all that you want to use a kobject for is to provide a reference counter
 187for your structure, please use the struct kref instead; a kobject would be
 188overkill.  For more information on how to use struct kref, please see the
 189file Documentation/kref.txt in the Linux kernel source tree.
 192Creating "simple" kobjects
 194Sometimes all that a developer wants is a way to create a simple directory
 195in the sysfs hierarchy, and not have to mess with the whole complication of
 196ksets, show and store functions, and other details.  This is the one
 197exception where a single kobject should be created.  To create such an
 198entry, use the function:
 200    struct kobject *kobject_create_and_add(char *name, struct kobject *parent);
 202This function will create a kobject and place it in sysfs in the location
 203underneath the specified parent kobject.  To create simple attributes
 204associated with this kobject, use:
 206    int sysfs_create_file(struct kobject *kobj, struct attribute *attr);
 208    int sysfs_create_group(struct kobject *kobj, struct attribute_group *grp);
 210Both types of attributes used here, with a kobject that has been created
 211with the kobject_create_and_add(), can be of type kobj_attribute, so no
 212special custom attribute is needed to be created.
 214See the example module, samples/kobject/kobject-example.c for an
 215implementation of a simple kobject and attributes.
 219ktypes and release methods
 221One important thing still missing from the discussion is what happens to a
 222kobject when its reference count reaches zero. The code which created the
 223kobject generally does not know when that will happen; if it did, there
 224would be little point in using a kobject in the first place. Even
 225predictable object lifecycles become more complicated when sysfs is brought
 226in as other portions of the kernel can get a reference on any kobject that
 227is registered in the system.
 229The end result is that a structure protected by a kobject cannot be freed
 230before its reference count goes to zero. The reference count is not under
 231the direct control of the code which created the kobject. So that code must
 232be notified asynchronously whenever the last reference to one of its
 233kobjects goes away.
 235Once you registered your kobject via kobject_add(), you must never use
 236kfree() to free it directly. The only safe way is to use kobject_put(). It
 237is good practice to always use kobject_put() after kobject_init() to avoid
 238errors creeping in.
 240This notification is done through a kobject's release() method. Usually
 241such a method has a form like:
 243    void my_object_release(struct kobject *kobj)
 244    {
 245            struct my_object *mine = container_of(kobj, struct my_object, kobj);
 247            /* Perform any additional cleanup on this object, then... */
 248            kfree(mine);
 249    }
 251One important point cannot be overstated: every kobject must have a
 252release() method, and the kobject must persist (in a consistent state)
 253until that method is called. If these constraints are not met, the code is
 254flawed.  Note that the kernel will warn you if you forget to provide a
 255release() method.  Do not try to get rid of this warning by providing an
 256"empty" release function; you will be mocked mercilessly by the kobject
 257maintainer if you attempt this.
 259Note, the name of the kobject is available in the release function, but it
 260must NOT be changed within this callback.  Otherwise there will be a memory
 261leak in the kobject core, which makes people unhappy.
 263Interestingly, the release() method is not stored in the kobject itself;
 264instead, it is associated with the ktype. So let us introduce struct
 267    struct kobj_type {
 268            void (*release)(struct kobject *);
 269            struct sysfs_ops    *sysfs_ops;
 270            struct attribute    **default_attrs;
 271    };
 273This structure is used to describe a particular type of kobject (or, more
 274correctly, of containing object). Every kobject needs to have an associated
 275kobj_type structure; a pointer to that structure must be specified when you
 276call kobject_init() or kobject_init_and_add().
 278The release field in struct kobj_type is, of course, a pointer to the
 279release() method for this type of kobject. The other two fields (sysfs_ops
 280and default_attrs) control how objects of this type are represented in
 281sysfs; they are beyond the scope of this document.
 283The default_attrs pointer is a list of default attributes that will be
 284automatically created for any kobject that is registered with this ktype.
 289A kset is merely a collection of kobjects that want to be associated with
 290each other.  There is no restriction that they be of the same ktype, but be
 291very careful if they are not.
 293A kset serves these functions:
 295 - It serves as a bag containing a group of objects. A kset can be used by
 296   the kernel to track "all block devices" or "all PCI device drivers."
 298 - A kset is also a subdirectory in sysfs, where the associated kobjects
 299   with the kset can show up.  Every kset contains a kobject which can be
 300   set up to be the parent of other kobjects; the top-level directories of
 301   the sysfs hierarchy are constructed in this way.
 303 - Ksets can support the "hotplugging" of kobjects and influence how
 304   uevent events are reported to user space.
 306In object-oriented terms, "kset" is the top-level container class; ksets
 307contain their own kobject, but that kobject is managed by the kset code and
 308should not be manipulated by any other user.
 310A kset keeps its children in a standard kernel linked list.  Kobjects point
 311back to their containing kset via their kset field. In almost all cases,
 312the kobjects belonging to a kset have that kset (or, strictly, its embedded
 313kobject) in their parent.
 315As a kset contains a kobject within it, it should always be dynamically
 316created and never declared statically or on the stack.  To create a new
 317kset use:
 318  struct kset *kset_create_and_add(const char *name,
 319                                   struct kset_uevent_ops *u,
 320                                   struct kobject *parent);
 322When you are finished with the kset, call:
 323  void kset_unregister(struct kset *kset);
 324to destroy it.
 326An example of using a kset can be seen in the
 327samples/kobject/kset-example.c file in the kernel tree.
 329If a kset wishes to control the uevent operations of the kobjects
 330associated with it, it can use the struct kset_uevent_ops to handle it:
 332struct kset_uevent_ops {
 333        int (*filter)(struct kset *kset, struct kobject *kobj);
 334        const char *(*name)(struct kset *kset, struct kobject *kobj);
 335        int (*uevent)(struct kset *kset, struct kobject *kobj,
 336                      struct kobj_uevent_env *env);
 340The filter function allows a kset to prevent a uevent from being emitted to
 341userspace for a specific kobject.  If the function returns 0, the uevent
 342will not be emitted.
 344The name function will be called to override the default name of the kset
 345that the uevent sends to userspace.  By default, the name will be the same
 346as the kset itself, but this function, if present, can override that name.
 348The uevent function will be called when the uevent is about to be sent to
 349userspace to allow more environment variables to be added to the uevent.
 351One might ask how, exactly, a kobject is added to a kset, given that no
 352functions which perform that function have been presented.  The answer is
 353that this task is handled by kobject_add().  When a kobject is passed to
 354kobject_add(), its kset member should point to the kset to which the
 355kobject will belong.  kobject_add() will handle the rest.
 357If the kobject belonging to a kset has no parent kobject set, it will be
 358added to the kset's directory.  Not all members of a kset do necessarily
 359live in the kset directory.  If an explicit parent kobject is assigned
 360before the kobject is added, the kobject is registered with the kset, but
 361added below the parent kobject.
 364Kobject removal
 366After a kobject has been registered with the kobject core successfully, it
 367must be cleaned up when the code is finished with it.  To do that, call
 368kobject_put().  By doing this, the kobject core will automatically clean up
 369all of the memory allocated by this kobject.  If a KOBJ_ADD uevent has been
 370sent for the object, a corresponding KOBJ_REMOVE uevent will be sent, and
 371any other sysfs housekeeping will be handled for the caller properly.
 373If you need to do a two-stage delete of the kobject (say you are not
 374allowed to sleep when you need to destroy the object), then call
 375kobject_del() which will unregister the kobject from sysfs.  This makes the
 376kobject "invisible", but it is not cleaned up, and the reference count of
 377the object is still the same.  At a later time call kobject_put() to finish
 378the cleanup of the memory associated with the kobject.
 380kobject_del() can be used to drop the reference to the parent object, if
 381circular references are constructed.  It is valid in some cases, that a
 382parent objects references a child.  Circular references _must_ be broken
 383with an explicit call to kobject_del(), so that a release functions will be
 384called, and the objects in the former circle release each other.
 387Example code to copy from
 389For a more complete example of using ksets and kobjects properly, see the
 390sample/kobject/kset-example.c code.