1 =========================
2 Kernel Mode Setting (KMS)
3 =========================
5 Drivers must initialize the mode setting core by calling
6 :c:func:`drm_mode_config_init()` on the DRM device. The function
7 initializes the :c:type:`struct drm_device <drm_device>`
8 mode_config field and never fails. Once done, mode configuration must
9 be setup by initializing the following fields.
11 - int min_width, min_height; int max_width, max_height;
12 Minimum and maximum width and height of the frame buffers in pixel
15 - struct drm_mode_config_funcs \*funcs;
16 Mode setting functions.
18 Modeset Base Object Abstraction
19 ===============================
21 .. kernel-doc:: include/drm/drm_mode_object.h
24 .. kernel-doc:: drivers/gpu/drm/drm_mode_object.c
30 .. kernel-doc:: include/drm/drm_crtc.h
36 .. kernel-doc:: drivers/gpu/drm/drm_crtc.c
39 Atomic Mode Setting Function Reference
40 ======================================
42 .. kernel-doc:: drivers/gpu/drm/drm_atomic.c
45 .. kernel-doc:: include/drm/drm_atomic.h
48 Frame Buffer Abstraction
49 ========================
51 .. kernel-doc:: drivers/gpu/drm/drm_framebuffer.c
54 Frame Buffer Functions Reference
55 --------------------------------
57 .. kernel-doc:: drivers/gpu/drm/drm_framebuffer.c
60 .. kernel-doc:: include/drm/drm_framebuffer.h
66 .. kernel-doc:: drivers/gpu/drm/drm_fourcc.c
72 The KMS API doesn't standardize backing storage object creation and
73 leaves it to driver-specific ioctls. Furthermore actually creating a
74 buffer object even for GEM-based drivers is done through a
75 driver-specific ioctl - GEM only has a common userspace interface for
76 sharing and destroying objects. While not an issue for full-fledged
77 graphics stacks that include device-specific userspace components (in
78 libdrm for instance), this limit makes DRM-based early boot graphics
79 unnecessarily complex.
81 Dumb objects partly alleviate the problem by providing a standard API to
82 create dumb buffers suitable for scanout, which can then be used to
83 create KMS frame buffers.
85 To support dumb objects drivers must implement the dumb_create,
86 dumb_destroy and dumb_map_offset operations.
88 - int (\*dumb_create)(struct drm_file \*file_priv, struct
89 drm_device \*dev, struct drm_mode_create_dumb \*args);
90 The dumb_create operation creates a driver object (GEM or TTM
91 handle) suitable for scanout based on the width, height and depth
92 from the struct :c:type:`struct drm_mode_create_dumb
93 <drm_mode_create_dumb>` argument. It fills the argument's
94 handle, pitch and size fields with a handle for the newly created
95 object and its line pitch and size in bytes.
97 - int (\*dumb_destroy)(struct drm_file \*file_priv, struct
98 drm_device \*dev, uint32_t handle);
99 The dumb_destroy operation destroys a dumb object created by
102 - int (\*dumb_map_offset)(struct drm_file \*file_priv, struct
103 drm_device \*dev, uint32_t handle, uint64_t \*offset);
104 The dumb_map_offset operation associates an mmap fake offset with
105 the object given by the handle and returns it. Drivers must use the
106 :c:func:`drm_gem_create_mmap_offset()` function to associate
107 the fake offset as described in ?.
109 Note that dumb objects may not be used for gpu acceleration, as has been
110 attempted on some ARM embedded platforms. Such drivers really must have
111 a hardware-specific ioctl to allocate suitable buffer objects.
116 .. kernel-doc:: drivers/gpu/drm/drm_plane.c
119 Plane Functions Reference
120 -------------------------
122 .. kernel-doc:: include/drm/drm_plane.h
125 .. kernel-doc:: drivers/gpu/drm/drm_plane.c
128 Display Modes Function Reference
129 ================================
131 .. kernel-doc:: include/drm/drm_modes.h
134 .. kernel-doc:: drivers/gpu/drm/drm_modes.c
137 Connector Abstraction
138 =====================
140 .. kernel-doc:: drivers/gpu/drm/drm_connector.c
143 Connector Functions Reference
144 -----------------------------
146 .. kernel-doc:: include/drm/drm_connector.h
149 .. kernel-doc:: drivers/gpu/drm/drm_connector.c
155 .. kernel-doc:: drivers/gpu/drm/drm_encoder.c
158 Encoder Functions Reference
159 ---------------------------
161 .. kernel-doc:: include/drm/drm_encoder.h
164 .. kernel-doc:: drivers/gpu/drm/drm_encoder.c
167 KMS Initialization and Cleanup
168 ==============================
170 A KMS device is abstracted and exposed as a set of planes, CRTCs,
171 encoders and connectors. KMS drivers must thus create and initialize all
172 those objects at load time after initializing mode setting.
174 CRTCs (:c:type:`struct drm_crtc <drm_crtc>`)
175 --------------------------------------------
177 A CRTC is an abstraction representing a part of the chip that contains a
178 pointer to a scanout buffer. Therefore, the number of CRTCs available
179 determines how many independent scanout buffers can be active at any
180 given time. The CRTC structure contains several fields to support this:
181 a pointer to some video memory (abstracted as a frame buffer object), a
182 display mode, and an (x, y) offset into the video memory to support
183 panning or configurations where one piece of video memory spans multiple
189 A KMS device must create and register at least one struct
190 :c:type:`struct drm_crtc <drm_crtc>` instance. The instance is
191 allocated and zeroed by the driver, possibly as part of a larger
192 structure, and registered with a call to :c:func:`drm_crtc_init()`
193 with a pointer to CRTC functions.
199 The DRM core manages its objects' lifetime. When an object is not needed
200 anymore the core calls its destroy function, which must clean up and
201 free every resource allocated for the object. Every
202 :c:func:`drm_\*_init()` call must be matched with a corresponding
203 :c:func:`drm_\*_cleanup()` call to cleanup CRTCs
204 (:c:func:`drm_crtc_cleanup()`), planes
205 (:c:func:`drm_plane_cleanup()`), encoders
206 (:c:func:`drm_encoder_cleanup()`) and connectors
207 (:c:func:`drm_connector_cleanup()`). Furthermore, connectors that
208 have been added to sysfs must be removed by a call to
209 :c:func:`drm_connector_unregister()` before calling
210 :c:func:`drm_connector_cleanup()`.
212 Connectors state change detection must be cleanup up with a call to
213 :c:func:`drm_kms_helper_poll_fini()`.
215 Output discovery and initialization example
216 -------------------------------------------
220 void intel_crt_init(struct drm_device *dev)
222 struct drm_connector *connector;
223 struct intel_output *intel_output;
225 intel_output = kzalloc(sizeof(struct intel_output), GFP_KERNEL);
229 connector = &intel_output->base;
230 drm_connector_init(dev, &intel_output->base,
231 &intel_crt_connector_funcs, DRM_MODE_CONNECTOR_VGA);
233 drm_encoder_init(dev, &intel_output->enc, &intel_crt_enc_funcs,
234 DRM_MODE_ENCODER_DAC);
236 drm_mode_connector_attach_encoder(&intel_output->base,
239 /* Set up the DDC bus. */
240 intel_output->ddc_bus = intel_i2c_create(dev, GPIOA, "CRTDDC_A");
241 if (!intel_output->ddc_bus) {
242 dev_printk(KERN_ERR, &dev->pdev->dev, "DDC bus registration "
247 intel_output->type = INTEL_OUTPUT_ANALOG;
248 connector->interlace_allowed = 0;
249 connector->doublescan_allowed = 0;
251 drm_encoder_helper_add(&intel_output->enc, &intel_crt_helper_funcs);
252 drm_connector_helper_add(connector, &intel_crt_connector_helper_funcs);
254 drm_connector_register(connector);
257 In the example above (taken from the i915 driver), a CRTC, connector and
258 encoder combination is created. A device-specific i2c bus is also
259 created for fetching EDID data and performing monitor detection. Once
260 the process is complete, the new connector is registered with sysfs to
261 make its properties available to applications.
266 .. kernel-doc:: drivers/gpu/drm/drm_modeset_lock.c
269 .. kernel-doc:: include/drm/drm_modeset_lock.h
272 .. kernel-doc:: drivers/gpu/drm/drm_modeset_lock.c
278 Property Types and Blob Property Support
279 ----------------------------------------
281 .. kernel-doc:: drivers/gpu/drm/drm_property.c
284 .. kernel-doc:: include/drm/drm_property.h
287 .. kernel-doc:: drivers/gpu/drm/drm_property.c
290 Plane Composition Properties
291 ----------------------------
293 .. kernel-doc:: drivers/gpu/drm/drm_blend.c
296 .. kernel-doc:: drivers/gpu/drm/drm_blend.c
299 Color Management Properties
300 ---------------------------
302 .. kernel-doc:: drivers/gpu/drm/drm_color_mgmt.c
305 .. kernel-doc:: include/drm/drm_color_mgmt.h
308 .. kernel-doc:: drivers/gpu/drm/drm_color_mgmt.c
311 Existing KMS Properties
312 -----------------------
314 The following table gives description of drm properties exposed by
315 various modules/drivers.
319 :file: kms-properties.csv
324 Vertical blanking plays a major role in graphics rendering. To achieve
325 tear-free display, users must synchronize page flips and/or rendering to
326 vertical blanking. The DRM API offers ioctls to perform page flips
327 synchronized to vertical blanking and wait for vertical blanking.
329 The DRM core handles most of the vertical blanking management logic,
330 which involves filtering out spurious interrupts, keeping race-free
331 blanking counters, coping with counter wrap-around and resets and
332 keeping use counts. It relies on the driver to generate vertical
333 blanking interrupts and optionally provide a hardware vertical blanking
334 counter. Drivers must implement the following operations.
336 - int (\*enable_vblank) (struct drm_device \*dev, int crtc); void
337 (\*disable_vblank) (struct drm_device \*dev, int crtc);
338 Enable or disable vertical blanking interrupts for the given CRTC.
340 - u32 (\*get_vblank_counter) (struct drm_device \*dev, int crtc);
341 Retrieve the value of the vertical blanking counter for the given
342 CRTC. If the hardware maintains a vertical blanking counter its value
343 should be returned. Otherwise drivers can use the
344 :c:func:`drm_vblank_count()` helper function to handle this
347 Drivers must initialize the vertical blanking handling core with a call
348 to :c:func:`drm_vblank_init()` in their load operation.
350 Vertical blanking interrupts can be enabled by the DRM core or by
351 drivers themselves (for instance to handle page flipping operations).
352 The DRM core maintains a vertical blanking use count to ensure that the
353 interrupts are not disabled while a user still needs them. To increment
354 the use count, drivers call :c:func:`drm_vblank_get()`. Upon
355 return vertical blanking interrupts are guaranteed to be enabled.
357 To decrement the use count drivers call
358 :c:func:`drm_vblank_put()`. Only when the use count drops to zero
359 will the DRM core disable the vertical blanking interrupts after a delay
360 by scheduling a timer. The delay is accessible through the
361 vblankoffdelay module parameter or the ``drm_vblank_offdelay`` global
362 variable and expressed in milliseconds. Its default value is 5000 ms.
363 Zero means never disable, and a negative value means disable
364 immediately. Drivers may override the behaviour by setting the
365 :c:type:`struct drm_device <drm_device>`
366 vblank_disable_immediate flag, which when set causes vblank interrupts
367 to be disabled immediately regardless of the drm_vblank_offdelay
368 value. The flag should only be set if there's a properly working
369 hardware vblank counter present.
371 When a vertical blanking interrupt occurs drivers only need to call the
372 :c:func:`drm_handle_vblank()` function to account for the
375 Resources allocated by :c:func:`drm_vblank_init()` must be freed
376 with a call to :c:func:`drm_vblank_cleanup()` in the driver unload
379 Vertical Blanking and Interrupt Handling Functions Reference
380 ------------------------------------------------------------
382 .. kernel-doc:: drivers/gpu/drm/drm_irq.c
385 .. kernel-doc:: include/drm/drm_irq.h