1 <?xml version="1.0" encoding="UTF-8"?>
2 <!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook XML V4.1.2//EN"
3 "http://www.oasis-open.org/docbook/xml/4.1.2/docbookx.dtd" []>
5 <book id="drmDevelopersGuide">
7 <title>Linux DRM Developer's Guide</title>
10 <year>2008-2009</year>
12 Intel Corporation (Jesse Barnes <jesse.barnes@intel.com>)
18 The contents of this file may be used under the terms of the GNU
19 General Public License version 2 (the "GPL") as distributed in
20 the kernel source COPYING file.
29 <chapter id="drmIntroduction">
30 <title>Introduction</title>
32 The Linux DRM layer contains code intended to support the needs
33 of complex graphics devices, usually containing programmable
34 pipelines well suited to 3D graphics acceleration. Graphics
35 drivers in the kernel can make use of DRM functions to make
36 tasks like memory management, interrupt handling and DMA easier,
37 and provide a uniform interface to applications.
40 A note on versions: this guide covers features found in the DRM
41 tree, including the TTM memory manager, output configuration and
42 mode setting, and the new vblank internals, in addition to all
43 the regular features found in current kernels.
46 [Insert diagram of typical DRM stack here]
52 <chapter id="drmInternals">
53 <title>DRM Internals</title>
55 This chapter documents DRM internals relevant to driver authors
56 and developers working to add support for the latest features to
60 First, we'll go over some typical driver initialization
61 requirements, like setting up command buffers, creating an
62 initial output configuration, and initializing core services.
63 Subsequent sections will cover core internals in more detail,
64 providing implementation notes and examples.
67 The DRM layer provides several services to graphics drivers,
68 many of them driven by the application interfaces it provides
69 through libdrm, the library that wraps most of the DRM ioctls.
70 These include vblank event handling, memory
71 management, output management, framebuffer management, command
72 submission & fencing, suspend/resume support, and DMA
76 The core of every DRM driver is struct drm_device. Drivers
77 will typically statically initialize a drm_device structure,
78 then pass it to drm_init() at load time.
81 <!-- Internals: driver init -->
84 <title>Driver initialization</title>
86 Before calling the DRM initialization routines, the driver must
87 first create and fill out a struct drm_device structure.
90 static struct drm_driver driver = {
91 /* don't use mtrr's here, the Xserver or user space app should
92 * deal with them for intel hardware.
95 DRIVER_USE_AGP | DRIVER_REQUIRE_AGP |
96 DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | DRIVER_MODESET,
97 .load = i915_driver_load,
98 .unload = i915_driver_unload,
99 .firstopen = i915_driver_firstopen,
100 .lastclose = i915_driver_lastclose,
101 .preclose = i915_driver_preclose,
103 .restore = i915_restore,
104 .device_is_agp = i915_driver_device_is_agp,
105 .get_vblank_counter = i915_get_vblank_counter,
106 .enable_vblank = i915_enable_vblank,
107 .disable_vblank = i915_disable_vblank,
108 .irq_preinstall = i915_driver_irq_preinstall,
109 .irq_postinstall = i915_driver_irq_postinstall,
110 .irq_uninstall = i915_driver_irq_uninstall,
111 .irq_handler = i915_driver_irq_handler,
112 .reclaim_buffers = drm_core_reclaim_buffers,
113 .get_map_ofs = drm_core_get_map_ofs,
114 .get_reg_ofs = drm_core_get_reg_ofs,
115 .fb_probe = intelfb_probe,
116 .fb_remove = intelfb_remove,
117 .fb_resize = intelfb_resize,
118 .master_create = i915_master_create,
119 .master_destroy = i915_master_destroy,
120 #if defined(CONFIG_DEBUG_FS)
121 .debugfs_init = i915_debugfs_init,
122 .debugfs_cleanup = i915_debugfs_cleanup,
124 .gem_init_object = i915_gem_init_object,
125 .gem_free_object = i915_gem_free_object,
126 .gem_vm_ops = &i915_gem_vm_ops,
127 .ioctls = i915_ioctls,
129 .owner = THIS_MODULE,
131 .release = drm_release,
135 .fasync = drm_fasync,
137 .compat_ioctl = i915_compat_ioctl,
139 .llseek = noop_llseek,
143 .id_table = pciidlist,
145 .remove = __devexit_p(drm_cleanup_pci),
150 .major = DRIVER_MAJOR,
151 .minor = DRIVER_MINOR,
152 .patchlevel = DRIVER_PATCHLEVEL,
156 In the example above, taken from the i915 DRM driver, the driver
157 sets several flags indicating what core features it supports.
158 We'll go over the individual callbacks in later sections. Since
159 flags indicate which features your driver supports to the DRM
160 core, you need to set most of them prior to calling drm_init(). Some,
161 like DRIVER_MODESET can be set later based on user supplied parameters,
162 but that's the exception rather than the rule.
165 <title>Driver flags</title>
167 <term>DRIVER_USE_AGP</term>
169 Driver uses AGP interface
173 <term>DRIVER_REQUIRE_AGP</term>
175 Driver needs AGP interface to function.
179 <term>DRIVER_USE_MTRR</term>
182 Driver uses MTRR interface for mapping memory. Deprecated.
187 <term>DRIVER_PCI_DMA</term>
189 Driver is capable of PCI DMA. Deprecated.
193 <term>DRIVER_SG</term>
195 Driver can perform scatter/gather DMA. Deprecated.
199 <term>DRIVER_HAVE_DMA</term>
200 <listitem><para>Driver supports DMA. Deprecated.</para></listitem>
203 <term>DRIVER_HAVE_IRQ</term><term>DRIVER_IRQ_SHARED</term>
206 DRIVER_HAVE_IRQ indicates whether the driver has a IRQ
207 handler, DRIVER_IRQ_SHARED indicates whether the device &
208 handler support shared IRQs (note that this is required of
214 <term>DRIVER_DMA_QUEUE</term>
217 If the driver queues DMA requests and completes them
218 asynchronously, this flag should be set. Deprecated.
223 <term>DRIVER_FB_DMA</term>
226 Driver supports DMA to/from the framebuffer. Deprecated.
231 <term>DRIVER_MODESET</term>
234 Driver supports mode setting interfaces.
240 In this specific case, the driver requires AGP and supports
241 IRQs. DMA, as we'll see, is handled by device specific ioctls
242 in this case. It also supports the kernel mode setting APIs, though
243 unlike in the actual i915 driver source, this example unconditionally
244 exports KMS capability.
248 <!-- Internals: driver load -->
251 <title>Driver load</title>
253 In the previous section, we saw what a typical drm_driver
254 structure might look like. One of the more important fields in
255 the structure is the hook for the load function.
258 static struct drm_driver driver = {
260 .load = i915_driver_load,
265 The load function has many responsibilities: allocating a driver
266 private structure, specifying supported performance counters,
267 configuring the device (e.g. mapping registers & command
268 buffers), initializing the memory manager, and setting up the
269 initial output configuration.
272 Note that the tasks performed at driver load time must not
273 conflict with DRM client requirements. For instance, if user
274 level mode setting drivers are in use, it would be problematic
275 to perform output discovery & configuration at load time.
276 Likewise, if pre-memory management aware user level drivers are
277 in use, memory management and command buffer setup may need to
278 be omitted. These requirements are driver specific, and care
279 needs to be taken to keep both old and new applications and
280 libraries working. The i915 driver supports the "modeset"
281 module parameter to control whether advanced features are
282 enabled at load time or in legacy fashion. If compatibility is
283 a concern (e.g. with drivers converted over to the new interfaces
284 from the old ones), care must be taken to prevent incompatible
285 device initialization and control with the currently active
290 <title>Driver private & performance counters</title>
292 The driver private hangs off the main drm_device structure and
293 can be used for tracking various device specific bits of
294 information, like register offsets, command buffer status,
295 register state for suspend/resume, etc. At load time, a
296 driver can simply allocate one and set drm_device.dev_priv
297 appropriately; at unload the driver can free it and set
298 drm_device.dev_priv to NULL.
301 The DRM supports several counters which can be used for rough
302 performance characterization. Note that the DRM stat counter
303 system is not often used by applications, and supporting
304 additional counters is completely optional.
307 These interfaces are deprecated and should not be used. If performance
308 monitoring is desired, the developer should investigate and
309 potentially enhance the kernel perf and tracing infrastructure to export
310 GPU related performance information to performance monitoring
311 tools and applications.
316 <title>Configuring the device</title>
318 Obviously, device configuration will be device specific.
319 However, there are several common operations: finding a
320 device's PCI resources, mapping them, and potentially setting
324 Finding & mapping resources is fairly straightforward. The
325 DRM wrapper functions, drm_get_resource_start() and
326 drm_get_resource_len() can be used to find BARs on the given
327 drm_device struct. Once those values have been retrieved, the
328 driver load function can call drm_addmap() to create a new
329 mapping for the BAR in question. Note you'll probably want a
330 drm_local_map_t in your driver private structure to track any
332 <!-- !Fdrivers/gpu/drm/drm_bufs.c drm_get_resource_* -->
333 <!-- !Finclude/drm/drmP.h drm_local_map_t -->
336 if compatibility with other operating systems isn't a concern
337 (DRM drivers can run under various BSD variants and OpenSolaris),
338 native Linux calls can be used for the above, e.g. pci_resource_*
339 and iomap*/iounmap. See the Linux device driver book for more
343 Once you have a register map, you can use the DRM_READn() and
344 DRM_WRITEn() macros to access the registers on your device, or
345 use driver specific versions to offset into your MMIO space
346 relative to a driver specific base pointer (see I915_READ for
350 If your device supports interrupt generation, you may want to
351 setup an interrupt handler at driver load time as well. This
352 is done using the drm_irq_install() function. If your device
353 supports vertical blank interrupts, it should call
354 drm_vblank_init() to initialize the core vblank handling code before
355 enabling interrupts on your device. This ensures the vblank related
356 structures are allocated and allows the core to handle vblank events.
358 <!--!Fdrivers/char/drm/drm_irq.c drm_irq_install-->
360 Once your interrupt handler is registered (it'll use your
361 drm_driver.irq_handler as the actual interrupt handling
362 function), you can safely enable interrupts on your device,
363 assuming any other state your interrupt handler uses is also
367 Another task that may be necessary during configuration is
368 mapping the video BIOS. On many devices, the VBIOS describes
369 device configuration, LCD panel timings (if any), and contains
370 flags indicating device state. Mapping the BIOS can be done
371 using the pci_map_rom() call, a convenience function that
372 takes care of mapping the actual ROM, whether it has been
373 shadowed into memory (typically at address 0xc0000) or exists
374 on the PCI device in the ROM BAR. Note that once you've
375 mapped the ROM and extracted any necessary information, be
376 sure to unmap it; on many devices the ROM address decoder is
377 shared with other BARs, so leaving it mapped can cause
378 undesired behavior like hangs or memory corruption.
379 <!--!Fdrivers/pci/rom.c pci_map_rom-->
384 <title>Memory manager initialization</title>
386 In order to allocate command buffers, cursor memory, scanout
387 buffers, etc., as well as support the latest features provided
388 by packages like Mesa and the X.Org X server, your driver
389 should support a memory manager.
392 If your driver supports memory management (it should!), you'll
393 need to set that up at load time as well. How you initialize
394 it depends on which memory manager you're using, TTM or GEM.
397 <title>TTM initialization</title>
399 TTM (for Translation Table Manager) manages video memory and
400 aperture space for graphics devices. TTM supports both UMA devices
401 and devices with dedicated video RAM (VRAM), i.e. most discrete
402 graphics devices. If your device has dedicated RAM, supporting
403 TTM is desirable. TTM also integrates tightly with your
404 driver specific buffer execution function. See the radeon
408 The core TTM structure is the ttm_bo_driver struct. It contains
409 several fields with function pointers for initializing the TTM,
410 allocating and freeing memory, waiting for command completion
411 and fence synchronization, and memory migration. See the
412 radeon_ttm.c file for an example of usage.
415 The ttm_global_reference structure is made up of several fields:
418 struct ttm_global_reference {
419 enum ttm_global_types global_type;
422 int (*init) (struct ttm_global_reference *);
423 void (*release) (struct ttm_global_reference *);
427 There should be one global reference structure for your memory
428 manager as a whole, and there will be others for each object
429 created by the memory manager at runtime. Your global TTM should
430 have a type of TTM_GLOBAL_TTM_MEM. The size field for the global
431 object should be sizeof(struct ttm_mem_global), and the init and
432 release hooks should point at your driver specific init and
433 release routines, which will probably eventually call
434 ttm_mem_global_init and ttm_mem_global_release respectively.
437 Once your global TTM accounting structure is set up and initialized
438 (done by calling ttm_global_item_ref on the global object you
439 just created), you'll need to create a buffer object TTM to
440 provide a pool for buffer object allocation by clients and the
441 kernel itself. The type of this object should be TTM_GLOBAL_TTM_BO,
442 and its size should be sizeof(struct ttm_bo_global). Again,
443 driver specific init and release functions can be provided,
444 likely eventually calling ttm_bo_global_init and
445 ttm_bo_global_release, respectively. Also like the previous
446 object, ttm_global_item_ref is used to create an initial reference
447 count for the TTM, which will call your initialization function.
451 <title>GEM initialization</title>
453 GEM is an alternative to TTM, designed specifically for UMA
454 devices. It has simpler initialization and execution requirements
455 than TTM, but has no VRAM management capability. Core GEM
456 initialization is comprised of a basic drm_mm_init call to create
457 a GTT DRM MM object, which provides an address space pool for
458 object allocation. In a KMS configuration, the driver will
459 need to allocate and initialize a command ring buffer following
460 basic GEM initialization. Most UMA devices have a so-called
461 "stolen" memory region, which provides space for the initial
462 framebuffer and large, contiguous memory regions required by the
463 device. This space is not typically managed by GEM, and must
464 be initialized separately into its own DRM MM object.
467 Initialization will be driver specific, and will depend on
468 the architecture of the device. In the case of Intel
469 integrated graphics chips like 965GM, GEM initialization can
470 be done by calling the internal GEM init function,
471 i915_gem_do_init(). Since the 965GM is a UMA device
472 (i.e. it doesn't have dedicated VRAM), GEM will manage
473 making regular RAM available for GPU operations. Memory set
474 aside by the BIOS (called "stolen" memory by the i915
475 driver) will be managed by the DRM memrange allocator; the
476 rest of the aperture will be managed by GEM.
478 /* Basic memrange allocator for stolen space (aka vram) */
479 drm_memrange_init(&dev_priv->vram, 0, prealloc_size);
480 /* Let GEM Manage from end of prealloc space to end of aperture */
481 i915_gem_do_init(dev, prealloc_size, agp_size);
483 <!--!Edrivers/char/drm/drm_memrange.c-->
486 Once the memory manager has been set up, we can allocate the
487 command buffer. In the i915 case, this is also done with a
488 GEM function, i915_gem_init_ringbuffer().
494 <title>Output configuration</title>
496 The final initialization task is output configuration. This involves
497 finding and initializing the CRTCs, encoders and connectors
498 for your device, creating an initial configuration and
499 registering a framebuffer console driver.
502 <title>Output discovery and initialization</title>
504 Several core functions exist to create CRTCs, encoders and
505 connectors, namely drm_crtc_init(), drm_connector_init() and
506 drm_encoder_init(), along with several "helper" functions to
507 perform common tasks.
510 Connectors should be registered with sysfs once they've been
511 detected and initialized, using the
512 drm_sysfs_connector_add() function. Likewise, when they're
513 removed from the system, they should be destroyed with
514 drm_sysfs_connector_remove().
518 void intel_crt_init(struct drm_device *dev)
520 struct drm_connector *connector;
521 struct intel_output *intel_output;
523 intel_output = kzalloc(sizeof(struct intel_output), GFP_KERNEL);
527 connector = &intel_output->base;
528 drm_connector_init(dev, &intel_output->base,
529 &intel_crt_connector_funcs, DRM_MODE_CONNECTOR_VGA);
531 drm_encoder_init(dev, &intel_output->enc, &intel_crt_enc_funcs,
532 DRM_MODE_ENCODER_DAC);
534 drm_mode_connector_attach_encoder(&intel_output->base,
537 /* Set up the DDC bus. */
538 intel_output->ddc_bus = intel_i2c_create(dev, GPIOA, "CRTDDC_A");
539 if (!intel_output->ddc_bus) {
540 dev_printk(KERN_ERR, &dev->pdev->dev, "DDC bus registration "
545 intel_output->type = INTEL_OUTPUT_ANALOG;
546 connector->interlace_allowed = 0;
547 connector->doublescan_allowed = 0;
549 drm_encoder_helper_add(&intel_output->enc, &intel_crt_helper_funcs);
550 drm_connector_helper_add(connector, &intel_crt_connector_helper_funcs);
552 drm_sysfs_connector_add(connector);
557 In the example above (again, taken from the i915 driver), a
558 CRT connector and encoder combination is created. A device
559 specific i2c bus is also created, for fetching EDID data and
560 performing monitor detection. Once the process is complete,
561 the new connector is registered with sysfs, to make its
562 properties available to applications.
565 <title>Helper functions and core functions</title>
567 Since many PC-class graphics devices have similar display output
568 designs, the DRM provides a set of helper functions to make
569 output management easier. The core helper routines handle
570 encoder re-routing and disabling of unused functions following
571 mode set. Using the helpers is optional, but recommended for
572 devices with PC-style architectures (i.e. a set of display planes
573 for feeding pixels to encoders which are in turn routed to
574 connectors). Devices with more complex requirements needing
575 finer grained management can opt to use the core callbacks
579 [Insert typical diagram here.] [Insert OMAP style config here.]
583 For each encoder, CRTC and connector, several functions must
584 be provided, depending on the object type. Encoder objects
585 need to provide a DPMS (basically on/off) function, mode fixup
586 (for converting requested modes into native hardware timings),
587 and prepare, set and commit functions for use by the core DRM
588 helper functions. Connector helpers need to provide mode fetch and
589 validity functions as well as an encoder matching function for
590 returning an ideal encoder for a given connector. The core
591 connector functions include a DPMS callback, (deprecated)
592 save/restore routines, detection, mode probing, property handling,
593 and cleanup functions.
595 <!--!Edrivers/char/drm/drm_crtc.h-->
596 <!--!Edrivers/char/drm/drm_crtc.c-->
597 <!--!Edrivers/char/drm/drm_crtc_helper.c-->
602 <!-- Internals: vblank handling -->
605 <title>VBlank event handling</title>
607 The DRM core exposes two vertical blank related ioctls:
608 DRM_IOCTL_WAIT_VBLANK and DRM_IOCTL_MODESET_CTL.
609 <!--!Edrivers/char/drm/drm_irq.c-->
612 DRM_IOCTL_WAIT_VBLANK takes a struct drm_wait_vblank structure
613 as its argument, and is used to block or request a signal when a
614 specified vblank event occurs.
617 DRM_IOCTL_MODESET_CTL should be called by application level
618 drivers before and after mode setting, since on many devices the
619 vertical blank counter will be reset at that time. Internally,
620 the DRM snapshots the last vblank count when the ioctl is called
621 with the _DRM_PRE_MODESET command so that the counter won't go
622 backwards (which is dealt with when _DRM_POST_MODESET is used).
625 To support the functions above, the DRM core provides several
626 helper functions for tracking vertical blank counters, and
627 requires drivers to provide several callbacks:
628 get_vblank_counter(), enable_vblank() and disable_vblank(). The
629 core uses get_vblank_counter() to keep the counter accurate
630 across interrupt disable periods. It should return the current
631 vertical blank event count, which is often tracked in a device
632 register. The enable and disable vblank callbacks should enable
633 and disable vertical blank interrupts, respectively. In the
634 absence of DRM clients waiting on vblank events, the core DRM
635 code will use the disable_vblank() function to disable
636 interrupts, which saves power. They'll be re-enabled again when
637 a client calls the vblank wait ioctl above.
640 Devices that don't provide a count register can simply use an
641 internal atomic counter incremented on every vertical blank
642 interrupt, and can make their enable and disable vblank
643 functions into no-ops.
648 <title>Memory management</title>
650 The memory manager lies at the heart of many DRM operations, and
651 is also required to support advanced client features like OpenGL
652 pbuffers. The DRM currently contains two memory managers, TTM
657 <title>The Translation Table Manager (TTM)</title>
659 TTM was developed by Tungsten Graphics, primarily by Thomas
660 Hellström, and is intended to be a flexible, high performance
661 graphics memory manager.
664 Drivers wishing to support TTM must fill out a drm_bo_driver
668 TTM design background and information belongs here.
673 <title>The Graphics Execution Manager (GEM)</title>
675 GEM is an Intel project, authored by Eric Anholt and Keith
676 Packard. It provides simpler interfaces than TTM, and is well
677 suited for UMA devices.
680 GEM-enabled drivers must provide gem_init_object() and
681 gem_free_object() callbacks to support the core memory
682 allocation routines. They should also provide several driver
683 specific ioctls to support command execution, pinning, buffer
684 read & write, mapping, and domain ownership transfers.
687 On a fundamental level, GEM involves several operations: memory
688 allocation and freeing, command execution, and aperture management
689 at command execution time. Buffer object allocation is relatively
690 straightforward and largely provided by Linux's shmem layer, which
691 provides memory to back each object. When mapped into the GTT
692 or used in a command buffer, the backing pages for an object are
693 flushed to memory and marked write combined so as to be coherent
694 with the GPU. Likewise, when the GPU finishes rendering to an object,
695 if the CPU accesses it, it must be made coherent with the CPU's view
696 of memory, usually involving GPU cache flushing of various kinds.
697 This core CPU<->GPU coherency management is provided by the GEM
698 set domain function, which evaluates an object's current domain and
699 performs any necessary flushing or synchronization to put the object
700 into the desired coherency domain (note that the object may be busy,
701 i.e. an active render target; in that case the set domain function
702 will block the client and wait for rendering to complete before
703 performing any necessary flushing operations).
706 Perhaps the most important GEM function is providing a command
707 execution interface to clients. Client programs construct command
708 buffers containing references to previously allocated memory objects
709 and submit them to GEM. At that point, GEM will take care to bind
710 all the objects into the GTT, execute the buffer, and provide
711 necessary synchronization between clients accessing the same buffers.
712 This often involves evicting some objects from the GTT and re-binding
713 others (a fairly expensive operation), and providing relocation
714 support which hides fixed GTT offsets from clients. Clients must
715 take care not to submit command buffers that reference more objects
716 than can fit in the GTT or GEM will reject them and no rendering
717 will occur. Similarly, if several objects in the buffer require
718 fence registers to be allocated for correct rendering (e.g. 2D blits
719 on pre-965 chips), care must be taken not to require more fence
720 registers than are available to the client. Such resource management
721 should be abstracted from the client in libdrm.
727 <!-- Output management -->
729 <title>Output management</title>
731 At the core of the DRM output management code is a set of
732 structures representing CRTCs, encoders and connectors.
735 A CRTC is an abstraction representing a part of the chip that
736 contains a pointer to a scanout buffer. Therefore, the number
737 of CRTCs available determines how many independent scanout
738 buffers can be active at any given time. The CRTC structure
739 contains several fields to support this: a pointer to some video
740 memory, a display mode, and an (x, y) offset into the video
741 memory to support panning or configurations where one piece of
742 video memory spans multiple CRTCs.
745 An encoder takes pixel data from a CRTC and converts it to a
746 format suitable for any attached connectors. On some devices,
747 it may be possible to have a CRTC send data to more than one
748 encoder. In that case, both encoders would receive data from
749 the same scanout buffer, resulting in a "cloned" display
750 configuration across the connectors attached to each encoder.
753 A connector is the final destination for pixel data on a device,
754 and usually connects directly to an external display device like
755 a monitor or laptop panel. A connector can only be attached to
756 one encoder at a time. The connector is also the structure
757 where information about the attached display is kept, so it
758 contains fields for display data, EDID data, DPMS &
759 connection status, and information about modes supported on the
762 <!--!Edrivers/char/drm/drm_crtc.c-->
766 <title>Framebuffer management</title>
768 In order to set a mode on a given CRTC, encoder and connector
769 configuration, clients need to provide a framebuffer object which
770 will provide a source of pixels for the CRTC to deliver to the encoder(s)
771 and ultimately the connector(s) in the configuration. A framebuffer
772 is fundamentally a driver specific memory object, made into an opaque
773 handle by the DRM addfb function. Once an fb has been created this
774 way it can be passed to the KMS mode setting routines for use in
780 <title>Command submission & fencing</title>
782 This should cover a few device specific command submission
788 <title>Suspend/resume</title>
790 The DRM core provides some suspend/resume code, but drivers
791 wanting full suspend/resume support should provide save() and
792 restore() functions. These will be called at suspend,
793 hibernate, or resume time, and should perform any state save or
794 restore required by your device across suspend or hibernate
800 <title>DMA services</title>
802 This should cover how DMA mapping etc. is supported by the core.
803 These functions are deprecated and should not be used.
808 <!-- External interfaces -->
810 <chapter id="drmExternals">
811 <title>Userland interfaces</title>
813 The DRM core exports several interfaces to applications,
814 generally intended to be used through corresponding libdrm
815 wrapper functions. In addition, drivers export device specific
816 interfaces for use by userspace drivers & device aware
817 applications through ioctls and sysfs files.
820 External interfaces include: memory mapping, context management,
821 DMA operations, AGP management, vblank control, fence
822 management, memory management, and output management.
825 Cover generic ioctls and sysfs layout here. Only need high
826 level info, since man pages will cover the rest.
830 <!-- API reference -->
832 <appendix id="drmDriverApi">
833 <title>DRM Driver API</title>
835 Include auto-generated API reference here (need to reference it
836 from paragraphs above too).