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>
11 <firstname>Jesse</firstname>
12 <surname>Barnes</surname>
13 <contrib>Initial version</contrib>
15 <orgname>Intel Corporation</orgname>
17 <email>jesse.barnes@intel.com</email>
22 <firstname>Laurent</firstname>
23 <surname>Pinchart</surname>
24 <contrib>Driver internals</contrib>
26 <orgname>Ideas on board SPRL</orgname>
28 <email>laurent.pinchart@ideasonboard.com</email>
35 <year>2008-2009</year>
37 <holder>Intel Corporation</holder>
38 <holder>Laurent Pinchart</holder>
43 The contents of this file may be used under the terms of the GNU
44 General Public License version 2 (the "GPL") as distributed in
45 the kernel source COPYING file.
50 <!-- Put document revisions here, newest first. -->
52 <revnumber>1.0</revnumber>
53 <date>2012-07-13</date>
54 <authorinitials>LP</authorinitials>
55 <revremark>Added extensive documentation about driver internals.
65 <chapter id="drmIntroduction">
66 <title>Introduction</title>
68 The Linux DRM layer contains code intended to support the needs
69 of complex graphics devices, usually containing programmable
70 pipelines well suited to 3D graphics acceleration. Graphics
71 drivers in the kernel may make use of DRM functions to make
72 tasks like memory management, interrupt handling and DMA easier,
73 and provide a uniform interface to applications.
76 A note on versions: this guide covers features found in the DRM
77 tree, including the TTM memory manager, output configuration and
78 mode setting, and the new vblank internals, in addition to all
79 the regular features found in current kernels.
82 [Insert diagram of typical DRM stack here]
88 <chapter id="drmInternals">
89 <title>DRM Internals</title>
91 This chapter documents DRM internals relevant to driver authors
92 and developers working to add support for the latest features to
96 First, we go over some typical driver initialization
97 requirements, like setting up command buffers, creating an
98 initial output configuration, and initializing core services.
99 Subsequent sections cover core internals in more detail,
100 providing implementation notes and examples.
103 The DRM layer provides several services to graphics drivers,
104 many of them driven by the application interfaces it provides
105 through libdrm, the library that wraps most of the DRM ioctls.
106 These include vblank event handling, memory
107 management, output management, framebuffer management, command
108 submission & fencing, suspend/resume support, and DMA
112 <!-- Internals: driver init -->
115 <title>Driver Initialization</title>
117 At the core of every DRM driver is a <structname>drm_driver</structname>
118 structure. Drivers typically statically initialize a drm_driver structure,
119 and then pass it to one of the <function>drm_*_init()</function> functions
120 to register it with the DRM subsystem.
123 The <structname>drm_driver</structname> structure contains static
124 information that describes the driver and features it supports, and
125 pointers to methods that the DRM core will call to implement the DRM API.
126 We will first go through the <structname>drm_driver</structname> static
127 information fields, and will then describe individual operations in
128 details as they get used in later sections.
131 <title>Driver Information</title>
133 <title>Driver Features</title>
135 Drivers inform the DRM core about their requirements and supported
136 features by setting appropriate flags in the
137 <structfield>driver_features</structfield> field. Since those flags
138 influence the DRM core behaviour since registration time, most of them
139 must be set to registering the <structname>drm_driver</structname>
142 <synopsis>u32 driver_features;</synopsis>
144 <title>Driver Feature Flags</title>
146 <term>DRIVER_USE_AGP</term>
148 Driver uses AGP interface, the DRM core will manage AGP resources.
152 <term>DRIVER_REQUIRE_AGP</term>
154 Driver needs AGP interface to function. AGP initialization failure
155 will become a fatal error.
159 <term>DRIVER_PCI_DMA</term>
161 Driver is capable of PCI DMA, mapping of PCI DMA buffers to
162 userspace will be enabled. Deprecated.
166 <term>DRIVER_SG</term>
168 Driver can perform scatter/gather DMA, allocation and mapping of
169 scatter/gather buffers will be enabled. Deprecated.
173 <term>DRIVER_HAVE_DMA</term>
175 Driver supports DMA, the userspace DMA API will be supported.
180 <term>DRIVER_HAVE_IRQ</term><term>DRIVER_IRQ_SHARED</term>
182 DRIVER_HAVE_IRQ indicates whether the driver has an IRQ handler
183 managed by the DRM Core. The core will support simple IRQ handler
184 installation when the flag is set. The installation process is
185 described in <xref linkend="drm-irq-registration"/>.</para>
186 <para>DRIVER_IRQ_SHARED indicates whether the device & handler
187 support shared IRQs (note that this is required of PCI drivers).
191 <term>DRIVER_GEM</term>
193 Driver use the GEM memory manager.
197 <term>DRIVER_MODESET</term>
199 Driver supports mode setting interfaces (KMS).
203 <term>DRIVER_PRIME</term>
205 Driver implements DRM PRIME buffer sharing.
211 <title>Major, Minor and Patchlevel</title>
214 int patchlevel;</synopsis>
216 The DRM core identifies driver versions by a major, minor and patch
217 level triplet. The information is printed to the kernel log at
218 initialization time and passed to userspace through the
219 DRM_IOCTL_VERSION ioctl.
222 The major and minor numbers are also used to verify the requested driver
223 API version passed to DRM_IOCTL_SET_VERSION. When the driver API changes
224 between minor versions, applications can call DRM_IOCTL_SET_VERSION to
225 select a specific version of the API. If the requested major isn't equal
226 to the driver major, or the requested minor is larger than the driver
227 minor, the DRM_IOCTL_SET_VERSION call will return an error. Otherwise
228 the driver's set_version() method will be called with the requested
233 <title>Name, Description and Date</title>
234 <synopsis>char *name;
236 char *date;</synopsis>
238 The driver name is printed to the kernel log at initialization time,
239 used for IRQ registration and passed to userspace through
243 The driver description is a purely informative string passed to
244 userspace through the DRM_IOCTL_VERSION ioctl and otherwise unused by
248 The driver date, formatted as YYYYMMDD, is meant to identify the date of
249 the latest modification to the driver. However, as most drivers fail to
250 update it, its value is mostly useless. The DRM core prints it to the
251 kernel log at initialization time and passes it to userspace through the
252 DRM_IOCTL_VERSION ioctl.
257 <title>Driver Load</title>
259 The <methodname>load</methodname> method is the driver and device
260 initialization entry point. The method is responsible for allocating and
261 initializing driver private data, specifying supported performance
262 counters, performing resource allocation and mapping (e.g. acquiring
263 clocks, mapping registers or allocating command buffers), initializing
264 the memory manager (<xref linkend="drm-memory-management"/>), installing
265 the IRQ handler (<xref linkend="drm-irq-registration"/>), setting up
266 vertical blanking handling (<xref linkend="drm-vertical-blank"/>), mode
267 setting (<xref linkend="drm-mode-setting"/>) and initial output
268 configuration (<xref linkend="drm-kms-init"/>).
271 If compatibility is a concern (e.g. with drivers converted over from
272 User Mode Setting to Kernel Mode Setting), care must be taken to prevent
273 device initialization and control that is incompatible with currently
274 active userspace drivers. For instance, if user level mode setting
275 drivers are in use, it would be problematic to perform output discovery
276 & configuration at load time. Likewise, if user-level drivers
277 unaware of memory management are in use, memory management and command
278 buffer setup may need to be omitted. These requirements are
279 driver-specific, and care needs to be taken to keep both old and new
280 applications and libraries working.
282 <synopsis>int (*load) (struct drm_device *, unsigned long flags);</synopsis>
284 The method takes two arguments, a pointer to the newly created
285 <structname>drm_device</structname> and flags. The flags are used to
286 pass the <structfield>driver_data</structfield> field of the device id
287 corresponding to the device passed to <function>drm_*_init()</function>.
288 Only PCI devices currently use this, USB and platform DRM drivers have
289 their <methodname>load</methodname> method called with flags to 0.
292 <title>Driver Private & Performance Counters</title>
294 The driver private hangs off the main
295 <structname>drm_device</structname> structure and can be used for
296 tracking various device-specific bits of information, like register
297 offsets, command buffer status, register state for suspend/resume, etc.
298 At load time, a driver may simply allocate one and set
299 <structname>drm_device</structname>.<structfield>dev_priv</structfield>
300 appropriately; it should be freed and
301 <structname>drm_device</structname>.<structfield>dev_priv</structfield>
302 set to NULL when the driver is unloaded.
305 DRM supports several counters which were used for rough performance
306 characterization. This stat counter system is deprecated and should not
307 be used. If performance monitoring is desired, the developer should
308 investigate and potentially enhance the kernel perf and tracing
309 infrastructure to export GPU related performance information for
310 consumption by performance monitoring tools and applications.
313 <sect3 id="drm-irq-registration">
314 <title>IRQ Registration</title>
316 The DRM core tries to facilitate IRQ handler registration and
317 unregistration by providing <function>drm_irq_install</function> and
318 <function>drm_irq_uninstall</function> functions. Those functions only
319 support a single interrupt per device, devices that use more than one
320 IRQs need to be handled manually.
323 <title>Managed IRQ Registration</title>
325 Both the <function>drm_irq_install</function> and
326 <function>drm_irq_uninstall</function> functions get the device IRQ by
327 calling <function>drm_dev_to_irq</function>. This inline function will
328 call a bus-specific operation to retrieve the IRQ number. For platform
329 devices, <function>platform_get_irq</function>(..., 0) is used to
330 retrieve the IRQ number.
333 <function>drm_irq_install</function> starts by calling the
334 <methodname>irq_preinstall</methodname> driver operation. The operation
335 is optional and must make sure that the interrupt will not get fired by
336 clearing all pending interrupt flags or disabling the interrupt.
339 The IRQ will then be requested by a call to
340 <function>request_irq</function>. If the DRIVER_IRQ_SHARED driver
341 feature flag is set, a shared (IRQF_SHARED) IRQ handler will be
345 The IRQ handler function must be provided as the mandatory irq_handler
346 driver operation. It will get passed directly to
347 <function>request_irq</function> and thus has the same prototype as all
348 IRQ handlers. It will get called with a pointer to the DRM device as the
352 Finally the function calls the optional
353 <methodname>irq_postinstall</methodname> driver operation. The operation
354 usually enables interrupts (excluding the vblank interrupt, which is
355 enabled separately), but drivers may choose to enable/disable interrupts
359 <function>drm_irq_uninstall</function> is similarly used to uninstall an
360 IRQ handler. It starts by waking up all processes waiting on a vblank
361 interrupt to make sure they don't hang, and then calls the optional
362 <methodname>irq_uninstall</methodname> driver operation. The operation
363 must disable all hardware interrupts. Finally the function frees the IRQ
364 by calling <function>free_irq</function>.
368 <title>Manual IRQ Registration</title>
370 Drivers that require multiple interrupt handlers can't use the managed
371 IRQ registration functions. In that case IRQs must be registered and
372 unregistered manually (usually with the <function>request_irq</function>
373 and <function>free_irq</function> functions, or their devm_* equivalent).
376 When manually registering IRQs, drivers must not set the DRIVER_HAVE_IRQ
377 driver feature flag, and must not provide the
378 <methodname>irq_handler</methodname> driver operation. They must set the
379 <structname>drm_device</structname> <structfield>irq_enabled</structfield>
380 field to 1 upon registration of the IRQs, and clear it to 0 after
381 unregistering the IRQs.
386 <title>Memory Manager Initialization</title>
388 Every DRM driver requires a memory manager which must be initialized at
389 load time. DRM currently contains two memory managers, the Translation
390 Table Manager (TTM) and the Graphics Execution Manager (GEM).
391 This document describes the use of the GEM memory manager only. See
392 <xref linkend="drm-memory-management"/> for details.
396 <title>Miscellaneous Device Configuration</title>
398 Another task that may be necessary for PCI devices during configuration
399 is mapping the video BIOS. On many devices, the VBIOS describes device
400 configuration, LCD panel timings (if any), and contains flags indicating
401 device state. Mapping the BIOS can be done using the pci_map_rom() call,
402 a convenience function that takes care of mapping the actual ROM,
403 whether it has been shadowed into memory (typically at address 0xc0000)
404 or exists on the PCI device in the ROM BAR. Note that after the ROM has
405 been mapped and any necessary information has been extracted, it should
406 be unmapped; on many devices, the ROM address decoder is shared with
407 other BARs, so leaving it mapped could cause undesired behaviour like
408 hangs or memory corruption.
409 <!--!Fdrivers/pci/rom.c pci_map_rom-->
415 <!-- Internals: memory management -->
417 <sect1 id="drm-memory-management">
418 <title>Memory management</title>
420 Modern Linux systems require large amount of graphics memory to store
421 frame buffers, textures, vertices and other graphics-related data. Given
422 the very dynamic nature of many of that data, managing graphics memory
423 efficiently is thus crucial for the graphics stack and plays a central
424 role in the DRM infrastructure.
427 The DRM core includes two memory managers, namely Translation Table Maps
428 (TTM) and Graphics Execution Manager (GEM). TTM was the first DRM memory
429 manager to be developed and tried to be a one-size-fits-them all
430 solution. It provides a single userspace API to accommodate the need of
431 all hardware, supporting both Unified Memory Architecture (UMA) devices
432 and devices with dedicated video RAM (i.e. most discrete video cards).
433 This resulted in a large, complex piece of code that turned out to be
434 hard to use for driver development.
437 GEM started as an Intel-sponsored project in reaction to TTM's
438 complexity. Its design philosophy is completely different: instead of
439 providing a solution to every graphics memory-related problems, GEM
440 identified common code between drivers and created a support library to
441 share it. GEM has simpler initialization and execution requirements than
442 TTM, but has no video RAM management capabitilies and is thus limited to
446 <title>The Translation Table Manager (TTM)</title>
448 TTM design background and information belongs here.
451 <title>TTM initialization</title>
452 <warning><para>This section is outdated.</para></warning>
454 Drivers wishing to support TTM must fill out a drm_bo_driver
455 structure. The structure contains several fields with function
456 pointers for initializing the TTM, allocating and freeing memory,
457 waiting for command completion and fence synchronization, and memory
458 migration. See the radeon_ttm.c file for an example of usage.
461 The ttm_global_reference structure is made up of several fields:
464 struct ttm_global_reference {
465 enum ttm_global_types global_type;
468 int (*init) (struct ttm_global_reference *);
469 void (*release) (struct ttm_global_reference *);
473 There should be one global reference structure for your memory
474 manager as a whole, and there will be others for each object
475 created by the memory manager at runtime. Your global TTM should
476 have a type of TTM_GLOBAL_TTM_MEM. The size field for the global
477 object should be sizeof(struct ttm_mem_global), and the init and
478 release hooks should point at your driver-specific init and
479 release routines, which probably eventually call
480 ttm_mem_global_init and ttm_mem_global_release, respectively.
483 Once your global TTM accounting structure is set up and initialized
484 by calling ttm_global_item_ref() on it,
485 you need to create a buffer object TTM to
486 provide a pool for buffer object allocation by clients and the
487 kernel itself. The type of this object should be TTM_GLOBAL_TTM_BO,
488 and its size should be sizeof(struct ttm_bo_global). Again,
489 driver-specific init and release functions may be provided,
490 likely eventually calling ttm_bo_global_init() and
491 ttm_bo_global_release(), respectively. Also, like the previous
492 object, ttm_global_item_ref() is used to create an initial reference
493 count for the TTM, which will call your initialization function.
498 <title>The Graphics Execution Manager (GEM)</title>
500 The GEM design approach has resulted in a memory manager that doesn't
501 provide full coverage of all (or even all common) use cases in its
502 userspace or kernel API. GEM exposes a set of standard memory-related
503 operations to userspace and a set of helper functions to drivers, and let
504 drivers implement hardware-specific operations with their own private API.
507 The GEM userspace API is described in the
508 <ulink url="http://lwn.net/Articles/283798/"><citetitle>GEM - the Graphics
509 Execution Manager</citetitle></ulink> article on LWN. While slightly
510 outdated, the document provides a good overview of the GEM API principles.
511 Buffer allocation and read and write operations, described as part of the
512 common GEM API, are currently implemented using driver-specific ioctls.
515 GEM is data-agnostic. It manages abstract buffer objects without knowing
516 what individual buffers contain. APIs that require knowledge of buffer
517 contents or purpose, such as buffer allocation or synchronization
518 primitives, are thus outside of the scope of GEM and must be implemented
519 using driver-specific ioctls.
522 On a fundamental level, GEM involves several operations:
524 <listitem>Memory allocation and freeing</listitem>
525 <listitem>Command execution</listitem>
526 <listitem>Aperture management at command execution time</listitem>
528 Buffer object allocation is relatively straightforward and largely
529 provided by Linux's shmem layer, which provides memory to back each
533 Device-specific operations, such as command execution, pinning, buffer
534 read & write, mapping, and domain ownership transfers are left to
535 driver-specific ioctls.
538 <title>GEM Initialization</title>
540 Drivers that use GEM must set the DRIVER_GEM bit in the struct
541 <structname>drm_driver</structname>
542 <structfield>driver_features</structfield> field. The DRM core will
543 then automatically initialize the GEM core before calling the
544 <methodname>load</methodname> operation. Behind the scene, this will
545 create a DRM Memory Manager object which provides an address space
546 pool for object allocation.
549 In a KMS configuration, drivers need to allocate and initialize a
550 command ring buffer following core GEM initialization if required by
551 the hardware. UMA devices usually have what is called a "stolen"
552 memory region, which provides space for the initial framebuffer and
553 large, contiguous memory regions required by the device. This space is
554 typically not managed by GEM, and must be initialized separately into
555 its own DRM MM object.
559 <title>GEM Objects Creation</title>
561 GEM splits creation of GEM objects and allocation of the memory that
562 backs them in two distinct operations.
565 GEM objects are represented by an instance of struct
566 <structname>drm_gem_object</structname>. Drivers usually need to extend
567 GEM objects with private information and thus create a driver-specific
568 GEM object structure type that embeds an instance of struct
569 <structname>drm_gem_object</structname>.
572 To create a GEM object, a driver allocates memory for an instance of its
573 specific GEM object type and initializes the embedded struct
574 <structname>drm_gem_object</structname> with a call to
575 <function>drm_gem_object_init</function>. The function takes a pointer to
576 the DRM device, a pointer to the GEM object and the buffer object size
580 GEM uses shmem to allocate anonymous pageable memory.
581 <function>drm_gem_object_init</function> will create an shmfs file of
582 the requested size and store it into the struct
583 <structname>drm_gem_object</structname> <structfield>filp</structfield>
584 field. The memory is used as either main storage for the object when the
585 graphics hardware uses system memory directly or as a backing store
589 Drivers are responsible for the actual physical pages allocation by
590 calling <function>shmem_read_mapping_page_gfp</function> for each page.
591 Note that they can decide to allocate pages when initializing the GEM
592 object, or to delay allocation until the memory is needed (for instance
593 when a page fault occurs as a result of a userspace memory access or
594 when the driver needs to start a DMA transfer involving the memory).
597 Anonymous pageable memory allocation is not always desired, for instance
598 when the hardware requires physically contiguous system memory as is
599 often the case in embedded devices. Drivers can create GEM objects with
600 no shmfs backing (called private GEM objects) by initializing them with
601 a call to <function>drm_gem_private_object_init</function> instead of
602 <function>drm_gem_object_init</function>. Storage for private GEM
603 objects must be managed by drivers.
606 Drivers that do not need to extend GEM objects with private information
607 can call the <function>drm_gem_object_alloc</function> function to
608 allocate and initialize a struct <structname>drm_gem_object</structname>
609 instance. The GEM core will call the optional driver
610 <methodname>gem_init_object</methodname> operation after initializing
611 the GEM object with <function>drm_gem_object_init</function>.
612 <synopsis>int (*gem_init_object) (struct drm_gem_object *obj);</synopsis>
615 No alloc-and-init function exists for private GEM objects.
619 <title>GEM Objects Lifetime</title>
621 All GEM objects are reference-counted by the GEM core. References can be
622 acquired and release by <function>calling drm_gem_object_reference</function>
623 and <function>drm_gem_object_unreference</function> respectively. The
624 caller must hold the <structname>drm_device</structname>
625 <structfield>struct_mutex</structfield> lock. As a convenience, GEM
626 provides the <function>drm_gem_object_reference_unlocked</function> and
627 <function>drm_gem_object_unreference_unlocked</function> functions that
628 can be called without holding the lock.
631 When the last reference to a GEM object is released the GEM core calls
632 the <structname>drm_driver</structname>
633 <methodname>gem_free_object</methodname> operation. That operation is
634 mandatory for GEM-enabled drivers and must free the GEM object and all
635 associated resources.
638 <synopsis>void (*gem_free_object) (struct drm_gem_object *obj);</synopsis>
639 Drivers are responsible for freeing all GEM object resources, including
640 the resources created by the GEM core. If an mmap offset has been
641 created for the object (in which case
642 <structname>drm_gem_object</structname>::<structfield>map_list</structfield>::<structfield>map</structfield>
643 is not NULL) it must be freed by a call to
644 <function>drm_gem_free_mmap_offset</function>. The shmfs backing store
645 must be released by calling <function>drm_gem_object_release</function>
646 (that function can safely be called if no shmfs backing store has been
651 <title>GEM Objects Naming</title>
653 Communication between userspace and the kernel refers to GEM objects
654 using local handles, global names or, more recently, file descriptors.
655 All of those are 32-bit integer values; the usual Linux kernel limits
656 apply to the file descriptors.
659 GEM handles are local to a DRM file. Applications get a handle to a GEM
660 object through a driver-specific ioctl, and can use that handle to refer
661 to the GEM object in other standard or driver-specific ioctls. Closing a
662 DRM file handle frees all its GEM handles and dereferences the
663 associated GEM objects.
666 To create a handle for a GEM object drivers call
667 <function>drm_gem_handle_create</function>. The function takes a pointer
668 to the DRM file and the GEM object and returns a locally unique handle.
669 When the handle is no longer needed drivers delete it with a call to
670 <function>drm_gem_handle_delete</function>. Finally the GEM object
671 associated with a handle can be retrieved by a call to
672 <function>drm_gem_object_lookup</function>.
675 Handles don't take ownership of GEM objects, they only take a reference
676 to the object that will be dropped when the handle is destroyed. To
677 avoid leaking GEM objects, drivers must make sure they drop the
678 reference(s) they own (such as the initial reference taken at object
679 creation time) as appropriate, without any special consideration for the
680 handle. For example, in the particular case of combined GEM object and
681 handle creation in the implementation of the
682 <methodname>dumb_create</methodname> operation, drivers must drop the
683 initial reference to the GEM object before returning the handle.
686 GEM names are similar in purpose to handles but are not local to DRM
687 files. They can be passed between processes to reference a GEM object
688 globally. Names can't be used directly to refer to objects in the DRM
689 API, applications must convert handles to names and names to handles
690 using the DRM_IOCTL_GEM_FLINK and DRM_IOCTL_GEM_OPEN ioctls
691 respectively. The conversion is handled by the DRM core without any
692 driver-specific support.
695 Similar to global names, GEM file descriptors are also used to share GEM
696 objects across processes. They offer additional security: as file
697 descriptors must be explicitly sent over UNIX domain sockets to be shared
698 between applications, they can't be guessed like the globally unique GEM
702 Drivers that support GEM file descriptors, also known as the DRM PRIME
703 API, must set the DRIVER_PRIME bit in the struct
704 <structname>drm_driver</structname>
705 <structfield>driver_features</structfield> field, and implement the
706 <methodname>prime_handle_to_fd</methodname> and
707 <methodname>prime_fd_to_handle</methodname> operations.
710 <synopsis>int (*prime_handle_to_fd)(struct drm_device *dev,
711 struct drm_file *file_priv, uint32_t handle,
712 uint32_t flags, int *prime_fd);
713 int (*prime_fd_to_handle)(struct drm_device *dev,
714 struct drm_file *file_priv, int prime_fd,
715 uint32_t *handle);</synopsis>
716 Those two operations convert a handle to a PRIME file descriptor and
717 vice versa. Drivers must use the kernel dma-buf buffer sharing framework
718 to manage the PRIME file descriptors.
721 While non-GEM drivers must implement the operations themselves, GEM
722 drivers must use the <function>drm_gem_prime_handle_to_fd</function>
723 and <function>drm_gem_prime_fd_to_handle</function> helper functions.
724 Those helpers rely on the driver
725 <methodname>gem_prime_export</methodname> and
726 <methodname>gem_prime_import</methodname> operations to create a dma-buf
727 instance from a GEM object (dma-buf exporter role) and to create a GEM
728 object from a dma-buf instance (dma-buf importer role).
731 <synopsis>struct dma_buf * (*gem_prime_export)(struct drm_device *dev,
732 struct drm_gem_object *obj,
734 struct drm_gem_object * (*gem_prime_import)(struct drm_device *dev,
735 struct dma_buf *dma_buf);</synopsis>
736 These two operations are mandatory for GEM drivers that support DRM
740 <title>DRM PRIME Helper Functions Reference</title>
741 !Pdrivers/gpu/drm/drm_prime.c PRIME Helpers
744 <sect3 id="drm-gem-objects-mapping">
745 <title>GEM Objects Mapping</title>
747 Because mapping operations are fairly heavyweight GEM favours
748 read/write-like access to buffers, implemented through driver-specific
749 ioctls, over mapping buffers to userspace. However, when random access
750 to the buffer is needed (to perform software rendering for instance),
751 direct access to the object can be more efficient.
754 The mmap system call can't be used directly to map GEM objects, as they
755 don't have their own file handle. Two alternative methods currently
756 co-exist to map GEM objects to userspace. The first method uses a
757 driver-specific ioctl to perform the mapping operation, calling
758 <function>do_mmap</function> under the hood. This is often considered
759 dubious, seems to be discouraged for new GEM-enabled drivers, and will
760 thus not be described here.
763 The second method uses the mmap system call on the DRM file handle.
764 <synopsis>void *mmap(void *addr, size_t length, int prot, int flags, int fd,
765 off_t offset);</synopsis>
766 DRM identifies the GEM object to be mapped by a fake offset passed
767 through the mmap offset argument. Prior to being mapped, a GEM object
768 must thus be associated with a fake offset. To do so, drivers must call
769 <function>drm_gem_create_mmap_offset</function> on the object. The
770 function allocates a fake offset range from a pool and stores the
771 offset divided by PAGE_SIZE in
772 <literal>obj->map_list.hash.key</literal>. Care must be taken not to
773 call <function>drm_gem_create_mmap_offset</function> if a fake offset
774 has already been allocated for the object. This can be tested by
775 <literal>obj->map_list.map</literal> being non-NULL.
778 Once allocated, the fake offset value
779 (<literal>obj->map_list.hash.key << PAGE_SHIFT</literal>)
780 must be passed to the application in a driver-specific way and can then
781 be used as the mmap offset argument.
784 The GEM core provides a helper method <function>drm_gem_mmap</function>
785 to handle object mapping. The method can be set directly as the mmap
786 file operation handler. It will look up the GEM object based on the
787 offset value and set the VMA operations to the
788 <structname>drm_driver</structname> <structfield>gem_vm_ops</structfield>
789 field. Note that <function>drm_gem_mmap</function> doesn't map memory to
790 userspace, but relies on the driver-provided fault handler to map pages
794 To use <function>drm_gem_mmap</function>, drivers must fill the struct
795 <structname>drm_driver</structname> <structfield>gem_vm_ops</structfield>
796 field with a pointer to VM operations.
799 <synopsis>struct vm_operations_struct *gem_vm_ops
801 struct vm_operations_struct {
802 void (*open)(struct vm_area_struct * area);
803 void (*close)(struct vm_area_struct * area);
804 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
808 The <methodname>open</methodname> and <methodname>close</methodname>
809 operations must update the GEM object reference count. Drivers can use
810 the <function>drm_gem_vm_open</function> and
811 <function>drm_gem_vm_close</function> helper functions directly as open
815 The fault operation handler is responsible for mapping individual pages
816 to userspace when a page fault occurs. Depending on the memory
817 allocation scheme, drivers can allocate pages at fault time, or can
818 decide to allocate memory for the GEM object at the time the object is
822 Drivers that want to map the GEM object upfront instead of handling page
823 faults can implement their own mmap file operation handler.
827 <title>Dumb GEM Objects</title>
829 The GEM API doesn't standardize GEM objects creation and leaves it to
830 driver-specific ioctls. While not an issue for full-fledged graphics
831 stacks that include device-specific userspace components (in libdrm for
832 instance), this limit makes DRM-based early boot graphics unnecessarily
836 Dumb GEM objects partly alleviate the problem by providing a standard
837 API to create dumb buffers suitable for scanout, which can then be used
838 to create KMS frame buffers.
841 To support dumb GEM objects drivers must implement the
842 <methodname>dumb_create</methodname>,
843 <methodname>dumb_destroy</methodname> and
844 <methodname>dumb_map_offset</methodname> operations.
848 <synopsis>int (*dumb_create)(struct drm_file *file_priv, struct drm_device *dev,
849 struct drm_mode_create_dumb *args);</synopsis>
851 The <methodname>dumb_create</methodname> operation creates a GEM
852 object suitable for scanout based on the width, height and depth
853 from the struct <structname>drm_mode_create_dumb</structname>
854 argument. It fills the argument's <structfield>handle</structfield>,
855 <structfield>pitch</structfield> and <structfield>size</structfield>
856 fields with a handle for the newly created GEM object and its line
857 pitch and size in bytes.
861 <synopsis>int (*dumb_destroy)(struct drm_file *file_priv, struct drm_device *dev,
862 uint32_t handle);</synopsis>
864 The <methodname>dumb_destroy</methodname> operation destroys a dumb
865 GEM object created by <methodname>dumb_create</methodname>.
869 <synopsis>int (*dumb_map_offset)(struct drm_file *file_priv, struct drm_device *dev,
870 uint32_t handle, uint64_t *offset);</synopsis>
872 The <methodname>dumb_map_offset</methodname> operation associates an
873 mmap fake offset with the GEM object given by the handle and returns
874 it. Drivers must use the
875 <function>drm_gem_create_mmap_offset</function> function to
876 associate the fake offset as described in
877 <xref linkend="drm-gem-objects-mapping"/>.
883 <title>Memory Coherency</title>
885 When mapped to the device or used in a command buffer, backing pages
886 for an object are flushed to memory and marked write combined so as to
887 be coherent with the GPU. Likewise, if the CPU accesses an object
888 after the GPU has finished rendering to the object, then the object
889 must be made coherent with the CPU's view of memory, usually involving
890 GPU cache flushing of various kinds. This core CPU<->GPU
891 coherency management is provided by a device-specific ioctl, which
892 evaluates an object's current domain and performs any necessary
893 flushing or synchronization to put the object into the desired
894 coherency domain (note that the object may be busy, i.e. an active
895 render target; in that case, setting the domain blocks the client and
896 waits for rendering to complete before performing any necessary
897 flushing operations).
901 <title>Command Execution</title>
903 Perhaps the most important GEM function for GPU devices is providing a
904 command execution interface to clients. Client programs construct
905 command buffers containing references to previously allocated memory
906 objects, and then submit them to GEM. At that point, GEM takes care to
907 bind all the objects into the GTT, execute the buffer, and provide
908 necessary synchronization between clients accessing the same buffers.
909 This often involves evicting some objects from the GTT and re-binding
910 others (a fairly expensive operation), and providing relocation
911 support which hides fixed GTT offsets from clients. Clients must take
912 care not to submit command buffers that reference more objects than
913 can fit in the GTT; otherwise, GEM will reject them and no rendering
914 will occur. Similarly, if several objects in the buffer require fence
915 registers to be allocated for correct rendering (e.g. 2D blits on
916 pre-965 chips), care must be taken not to require more fence registers
917 than are available to the client. Such resource management should be
918 abstracted from the client in libdrm.
924 <!-- Internals: mode setting -->
926 <sect1 id="drm-mode-setting">
927 <title>Mode Setting</title>
929 Drivers must initialize the mode setting core by calling
930 <function>drm_mode_config_init</function> on the DRM device. The function
931 initializes the <structname>drm_device</structname>
932 <structfield>mode_config</structfield> field and never fails. Once done,
933 mode configuration must be setup by initializing the following fields.
937 <synopsis>int min_width, min_height;
938 int max_width, max_height;</synopsis>
940 Minimum and maximum width and height of the frame buffers in pixel
945 <synopsis>struct drm_mode_config_funcs *funcs;</synopsis>
946 <para>Mode setting functions.</para>
950 <title>Frame Buffer Creation</title>
951 <synopsis>struct drm_framebuffer *(*fb_create)(struct drm_device *dev,
952 struct drm_file *file_priv,
953 struct drm_mode_fb_cmd2 *mode_cmd);</synopsis>
955 Frame buffers are abstract memory objects that provide a source of
956 pixels to scanout to a CRTC. Applications explicitly request the
957 creation of frame buffers through the DRM_IOCTL_MODE_ADDFB(2) ioctls and
958 receive an opaque handle that can be passed to the KMS CRTC control,
959 plane configuration and page flip functions.
962 Frame buffers rely on the underneath memory manager for low-level memory
963 operations. When creating a frame buffer applications pass a memory
964 handle (or a list of memory handles for multi-planar formats) through
965 the <parameter>drm_mode_fb_cmd2</parameter> argument. This document
966 assumes that the driver uses GEM, those handles thus reference GEM
970 Drivers must first validate the requested frame buffer parameters passed
971 through the mode_cmd argument. In particular this is where invalid
972 sizes, pixel formats or pitches can be caught.
975 If the parameters are deemed valid, drivers then create, initialize and
976 return an instance of struct <structname>drm_framebuffer</structname>.
977 If desired the instance can be embedded in a larger driver-specific
978 structure. Drivers must fill its <structfield>width</structfield>,
979 <structfield>height</structfield>, <structfield>pitches</structfield>,
980 <structfield>offsets</structfield>, <structfield>depth</structfield>,
981 <structfield>bits_per_pixel</structfield> and
982 <structfield>pixel_format</structfield> fields from the values passed
983 through the <parameter>drm_mode_fb_cmd2</parameter> argument. They
984 should call the <function>drm_helper_mode_fill_fb_struct</function>
985 helper function to do so.
989 The initailization of the new framebuffer instance is finalized with a
990 call to <function>drm_framebuffer_init</function> which takes a pointer
991 to DRM frame buffer operations (struct
992 <structname>drm_framebuffer_funcs</structname>). Note that this function
993 publishes the framebuffer and so from this point on it can be accessed
994 concurrently from other threads. Hence it must be the last step in the
995 driver's framebuffer initialization sequence. Frame buffer operations
999 <synopsis>int (*create_handle)(struct drm_framebuffer *fb,
1000 struct drm_file *file_priv, unsigned int *handle);</synopsis>
1002 Create a handle to the frame buffer underlying memory object. If
1003 the frame buffer uses a multi-plane format, the handle will
1004 reference the memory object associated with the first plane.
1007 Drivers call <function>drm_gem_handle_create</function> to create
1012 <synopsis>void (*destroy)(struct drm_framebuffer *framebuffer);</synopsis>
1014 Destroy the frame buffer object and frees all associated
1015 resources. Drivers must call
1016 <function>drm_framebuffer_cleanup</function> to free resources
1017 allocated by the DRM core for the frame buffer object, and must
1018 make sure to unreference all memory objects associated with the
1019 frame buffer. Handles created by the
1020 <methodname>create_handle</methodname> operation are released by
1025 <synopsis>int (*dirty)(struct drm_framebuffer *framebuffer,
1026 struct drm_file *file_priv, unsigned flags, unsigned color,
1027 struct drm_clip_rect *clips, unsigned num_clips);</synopsis>
1029 This optional operation notifies the driver that a region of the
1030 frame buffer has changed in response to a DRM_IOCTL_MODE_DIRTYFB
1037 The lifetime of a drm framebuffer is controlled with a reference count,
1038 drivers can grab additional references with
1039 <function>drm_framebuffer_reference</function> </para> and drop them
1040 again with <function>drm_framebuffer_unreference</function>. For
1041 driver-private framebuffers for which the last reference is never
1042 dropped (e.g. for the fbdev framebuffer when the struct
1043 <structname>drm_framebuffer</structname> is embedded into the fbdev
1044 helper struct) drivers can manually clean up a framebuffer at module
1046 <function>drm_framebuffer_unregister_private</function>.
1049 <title>Output Polling</title>
1050 <synopsis>void (*output_poll_changed)(struct drm_device *dev);</synopsis>
1052 This operation notifies the driver that the status of one or more
1053 connectors has changed. Drivers that use the fb helper can just call the
1054 <function>drm_fb_helper_hotplug_event</function> function to handle this
1059 <title>Locking</title>
1061 Beside some lookup structures with their own locking (which is hidden
1062 behind the interface functions) most of the modeset state is protected
1063 by the <code>dev-<mode_config.lock</code> mutex and additionally
1064 per-crtc locks to allow cursor updates, pageflips and similar operations
1065 to occur concurrently with background tasks like output detection.
1066 Operations which cross domains like a full modeset always grab all
1067 locks. Drivers there need to protect resources shared between crtcs with
1068 additional locking. They also need to be careful to always grab the
1069 relevant crtc locks if a modset functions touches crtc state, e.g. for
1070 load detection (which does only grab the <code>mode_config.lock</code>
1071 to allow concurrent screen updates on live crtcs).
1076 <!-- Internals: kms initialization and cleanup -->
1078 <sect1 id="drm-kms-init">
1079 <title>KMS Initialization and Cleanup</title>
1081 A KMS device is abstracted and exposed as a set of planes, CRTCs, encoders
1082 and connectors. KMS drivers must thus create and initialize all those
1083 objects at load time after initializing mode setting.
1086 <title>CRTCs (struct <structname>drm_crtc</structname>)</title>
1088 A CRTC is an abstraction representing a part of the chip that contains a
1089 pointer to a scanout buffer. Therefore, the number of CRTCs available
1090 determines how many independent scanout buffers can be active at any
1091 given time. The CRTC structure contains several fields to support this:
1092 a pointer to some video memory (abstracted as a frame buffer object), a
1093 display mode, and an (x, y) offset into the video memory to support
1094 panning or configurations where one piece of video memory spans multiple
1098 <title>CRTC Initialization</title>
1100 A KMS device must create and register at least one struct
1101 <structname>drm_crtc</structname> instance. The instance is allocated
1102 and zeroed by the driver, possibly as part of a larger structure, and
1103 registered with a call to <function>drm_crtc_init</function> with a
1104 pointer to CRTC functions.
1108 <title>CRTC Operations</title>
1110 <title>Set Configuration</title>
1111 <synopsis>int (*set_config)(struct drm_mode_set *set);</synopsis>
1113 Apply a new CRTC configuration to the device. The configuration
1114 specifies a CRTC, a frame buffer to scan out from, a (x,y) position in
1115 the frame buffer, a display mode and an array of connectors to drive
1116 with the CRTC if possible.
1119 If the frame buffer specified in the configuration is NULL, the driver
1120 must detach all encoders connected to the CRTC and all connectors
1121 attached to those encoders and disable them.
1124 This operation is called with the mode config lock held.
1127 FIXME: How should set_config interact with DPMS? If the CRTC is
1128 suspended, should it be resumed?
1132 <title>Page Flipping</title>
1133 <synopsis>int (*page_flip)(struct drm_crtc *crtc, struct drm_framebuffer *fb,
1134 struct drm_pending_vblank_event *event);</synopsis>
1136 Schedule a page flip to the given frame buffer for the CRTC. This
1137 operation is called with the mode config mutex held.
1140 Page flipping is a synchronization mechanism that replaces the frame
1141 buffer being scanned out by the CRTC with a new frame buffer during
1142 vertical blanking, avoiding tearing. When an application requests a page
1143 flip the DRM core verifies that the new frame buffer is large enough to
1144 be scanned out by the CRTC in the currently configured mode and then
1145 calls the CRTC <methodname>page_flip</methodname> operation with a
1146 pointer to the new frame buffer.
1149 The <methodname>page_flip</methodname> operation schedules a page flip.
1150 Once any pending rendering targeting the new frame buffer has
1151 completed, the CRTC will be reprogrammed to display that frame buffer
1152 after the next vertical refresh. The operation must return immediately
1153 without waiting for rendering or page flip to complete and must block
1154 any new rendering to the frame buffer until the page flip completes.
1157 If a page flip can be successfully scheduled the driver must set the
1158 <code>drm_crtc-<fb</code> field to the new framebuffer pointed to
1159 by <code>fb</code>. This is important so that the reference counting
1160 on framebuffers stays balanced.
1163 If a page flip is already pending, the
1164 <methodname>page_flip</methodname> operation must return
1165 -<errorname>EBUSY</errorname>.
1168 To synchronize page flip to vertical blanking the driver will likely
1169 need to enable vertical blanking interrupts. It should call
1170 <function>drm_vblank_get</function> for that purpose, and call
1171 <function>drm_vblank_put</function> after the page flip completes.
1174 If the application has requested to be notified when page flip completes
1175 the <methodname>page_flip</methodname> operation will be called with a
1176 non-NULL <parameter>event</parameter> argument pointing to a
1177 <structname>drm_pending_vblank_event</structname> instance. Upon page
1178 flip completion the driver must call <methodname>drm_send_vblank_event</methodname>
1179 to fill in the event and send to wake up any waiting processes.
1180 This can be performed with
1181 <programlisting><![CDATA[
1182 spin_lock_irqsave(&dev->event_lock, flags);
1184 drm_send_vblank_event(dev, pipe, event);
1185 spin_unlock_irqrestore(&dev->event_lock, flags);
1186 ]]></programlisting>
1189 FIXME: Could drivers that don't need to wait for rendering to complete
1190 just add the event to <literal>dev->vblank_event_list</literal> and
1191 let the DRM core handle everything, as for "normal" vertical blanking
1195 While waiting for the page flip to complete, the
1196 <literal>event->base.link</literal> list head can be used freely by
1197 the driver to store the pending event in a driver-specific list.
1200 If the file handle is closed before the event is signaled, drivers must
1201 take care to destroy the event in their
1202 <methodname>preclose</methodname> operation (and, if needed, call
1203 <function>drm_vblank_put</function>).
1207 <title>Miscellaneous</title>
1210 <synopsis>void (*set_property)(struct drm_crtc *crtc,
1211 struct drm_property *property, uint64_t value);</synopsis>
1213 Set the value of the given CRTC property to
1214 <parameter>value</parameter>. See <xref linkend="drm-kms-properties"/>
1215 for more information about properties.
1219 <synopsis>void (*gamma_set)(struct drm_crtc *crtc, u16 *r, u16 *g, u16 *b,
1220 uint32_t start, uint32_t size);</synopsis>
1222 Apply a gamma table to the device. The operation is optional.
1226 <synopsis>void (*destroy)(struct drm_crtc *crtc);</synopsis>
1228 Destroy the CRTC when not needed anymore. See
1229 <xref linkend="drm-kms-init"/>.
1237 <title>Planes (struct <structname>drm_plane</structname>)</title>
1239 A plane represents an image source that can be blended with or overlayed
1240 on top of a CRTC during the scanout process. Planes are associated with
1241 a frame buffer to crop a portion of the image memory (source) and
1242 optionally scale it to a destination size. The result is then blended
1243 with or overlayed on top of a CRTC.
1246 <title>Plane Initialization</title>
1248 Planes are optional. To create a plane, a KMS drivers allocates and
1249 zeroes an instances of struct <structname>drm_plane</structname>
1250 (possibly as part of a larger structure) and registers it with a call
1251 to <function>drm_plane_init</function>. The function takes a bitmask
1252 of the CRTCs that can be associated with the plane, a pointer to the
1253 plane functions and a list of format supported formats.
1257 <title>Plane Operations</title>
1260 <synopsis>int (*update_plane)(struct drm_plane *plane, struct drm_crtc *crtc,
1261 struct drm_framebuffer *fb, int crtc_x, int crtc_y,
1262 unsigned int crtc_w, unsigned int crtc_h,
1263 uint32_t src_x, uint32_t src_y,
1264 uint32_t src_w, uint32_t src_h);</synopsis>
1266 Enable and configure the plane to use the given CRTC and frame buffer.
1269 The source rectangle in frame buffer memory coordinates is given by
1270 the <parameter>src_x</parameter>, <parameter>src_y</parameter>,
1271 <parameter>src_w</parameter> and <parameter>src_h</parameter>
1272 parameters (as 16.16 fixed point values). Devices that don't support
1273 subpixel plane coordinates can ignore the fractional part.
1276 The destination rectangle in CRTC coordinates is given by the
1277 <parameter>crtc_x</parameter>, <parameter>crtc_y</parameter>,
1278 <parameter>crtc_w</parameter> and <parameter>crtc_h</parameter>
1279 parameters (as integer values). Devices scale the source rectangle to
1280 the destination rectangle. If scaling is not supported, and the source
1281 rectangle size doesn't match the destination rectangle size, the
1282 driver must return a -<errorname>EINVAL</errorname> error.
1286 <synopsis>int (*disable_plane)(struct drm_plane *plane);</synopsis>
1288 Disable the plane. The DRM core calls this method in response to a
1289 DRM_IOCTL_MODE_SETPLANE ioctl call with the frame buffer ID set to 0.
1290 Disabled planes must not be processed by the CRTC.
1294 <synopsis>void (*destroy)(struct drm_plane *plane);</synopsis>
1296 Destroy the plane when not needed anymore. See
1297 <xref linkend="drm-kms-init"/>.
1304 <title>Encoders (struct <structname>drm_encoder</structname>)</title>
1306 An encoder takes pixel data from a CRTC and converts it to a format
1307 suitable for any attached connectors. On some devices, it may be
1308 possible to have a CRTC send data to more than one encoder. In that
1309 case, both encoders would receive data from the same scanout buffer,
1310 resulting in a "cloned" display configuration across the connectors
1311 attached to each encoder.
1314 <title>Encoder Initialization</title>
1316 As for CRTCs, a KMS driver must create, initialize and register at
1317 least one struct <structname>drm_encoder</structname> instance. The
1318 instance is allocated and zeroed by the driver, possibly as part of a
1322 Drivers must initialize the struct <structname>drm_encoder</structname>
1323 <structfield>possible_crtcs</structfield> and
1324 <structfield>possible_clones</structfield> fields before registering the
1325 encoder. Both fields are bitmasks of respectively the CRTCs that the
1326 encoder can be connected to, and sibling encoders candidate for cloning.
1329 After being initialized, the encoder must be registered with a call to
1330 <function>drm_encoder_init</function>. The function takes a pointer to
1331 the encoder functions and an encoder type. Supported types are
1334 DRM_MODE_ENCODER_DAC for VGA and analog on DVI-I/DVI-A
1337 DRM_MODE_ENCODER_TMDS for DVI, HDMI and (embedded) DisplayPort
1340 DRM_MODE_ENCODER_LVDS for display panels
1343 DRM_MODE_ENCODER_TVDAC for TV output (Composite, S-Video, Component,
1347 DRM_MODE_ENCODER_VIRTUAL for virtual machine displays
1352 Encoders must be attached to a CRTC to be used. DRM drivers leave
1353 encoders unattached at initialization time. Applications (or the fbdev
1354 compatibility layer when implemented) are responsible for attaching the
1355 encoders they want to use to a CRTC.
1359 <title>Encoder Operations</title>
1362 <synopsis>void (*destroy)(struct drm_encoder *encoder);</synopsis>
1364 Called to destroy the encoder when not needed anymore. See
1365 <xref linkend="drm-kms-init"/>.
1369 <synopsis>void (*set_property)(struct drm_plane *plane,
1370 struct drm_property *property, uint64_t value);</synopsis>
1372 Set the value of the given plane property to
1373 <parameter>value</parameter>. See <xref linkend="drm-kms-properties"/>
1374 for more information about properties.
1381 <title>Connectors (struct <structname>drm_connector</structname>)</title>
1383 A connector is the final destination for pixel data on a device, and
1384 usually connects directly to an external display device like a monitor
1385 or laptop panel. A connector can only be attached to one encoder at a
1386 time. The connector is also the structure where information about the
1387 attached display is kept, so it contains fields for display data, EDID
1388 data, DPMS & connection status, and information about modes
1389 supported on the attached displays.
1392 <title>Connector Initialization</title>
1394 Finally a KMS driver must create, initialize, register and attach at
1395 least one struct <structname>drm_connector</structname> instance. The
1396 instance is created as other KMS objects and initialized by setting the
1401 <term><structfield>interlace_allowed</structfield></term>
1403 Whether the connector can handle interlaced modes.
1407 <term><structfield>doublescan_allowed</structfield></term>
1409 Whether the connector can handle doublescan.
1413 <term><structfield>display_info
1414 </structfield></term>
1416 Display information is filled from EDID information when a display
1417 is detected. For non hot-pluggable displays such as flat panels in
1418 embedded systems, the driver should initialize the
1419 <structfield>display_info</structfield>.<structfield>width_mm</structfield>
1421 <structfield>display_info</structfield>.<structfield>height_mm</structfield>
1422 fields with the physical size of the display.
1426 <term id="drm-kms-connector-polled"><structfield>polled</structfield></term>
1428 Connector polling mode, a combination of
1431 <term>DRM_CONNECTOR_POLL_HPD</term>
1433 The connector generates hotplug events and doesn't need to be
1434 periodically polled. The CONNECT and DISCONNECT flags must not
1435 be set together with the HPD flag.
1439 <term>DRM_CONNECTOR_POLL_CONNECT</term>
1441 Periodically poll the connector for connection.
1445 <term>DRM_CONNECTOR_POLL_DISCONNECT</term>
1447 Periodically poll the connector for disconnection.
1451 Set to 0 for connectors that don't support connection status
1457 The connector is then registered with a call to
1458 <function>drm_connector_init</function> with a pointer to the connector
1459 functions and a connector type, and exposed through sysfs with a call to
1460 <function>drm_sysfs_connector_add</function>.
1463 Supported connector types are
1465 <listitem>DRM_MODE_CONNECTOR_VGA</listitem>
1466 <listitem>DRM_MODE_CONNECTOR_DVII</listitem>
1467 <listitem>DRM_MODE_CONNECTOR_DVID</listitem>
1468 <listitem>DRM_MODE_CONNECTOR_DVIA</listitem>
1469 <listitem>DRM_MODE_CONNECTOR_Composite</listitem>
1470 <listitem>DRM_MODE_CONNECTOR_SVIDEO</listitem>
1471 <listitem>DRM_MODE_CONNECTOR_LVDS</listitem>
1472 <listitem>DRM_MODE_CONNECTOR_Component</listitem>
1473 <listitem>DRM_MODE_CONNECTOR_9PinDIN</listitem>
1474 <listitem>DRM_MODE_CONNECTOR_DisplayPort</listitem>
1475 <listitem>DRM_MODE_CONNECTOR_HDMIA</listitem>
1476 <listitem>DRM_MODE_CONNECTOR_HDMIB</listitem>
1477 <listitem>DRM_MODE_CONNECTOR_TV</listitem>
1478 <listitem>DRM_MODE_CONNECTOR_eDP</listitem>
1479 <listitem>DRM_MODE_CONNECTOR_VIRTUAL</listitem>
1483 Connectors must be attached to an encoder to be used. For devices that
1484 map connectors to encoders 1:1, the connector should be attached at
1485 initialization time with a call to
1486 <function>drm_mode_connector_attach_encoder</function>. The driver must
1487 also set the <structname>drm_connector</structname>
1488 <structfield>encoder</structfield> field to point to the attached
1492 Finally, drivers must initialize the connectors state change detection
1493 with a call to <function>drm_kms_helper_poll_init</function>. If at
1494 least one connector is pollable but can't generate hotplug interrupts
1495 (indicated by the DRM_CONNECTOR_POLL_CONNECT and
1496 DRM_CONNECTOR_POLL_DISCONNECT connector flags), a delayed work will
1497 automatically be queued to periodically poll for changes. Connectors
1498 that can generate hotplug interrupts must be marked with the
1499 DRM_CONNECTOR_POLL_HPD flag instead, and their interrupt handler must
1500 call <function>drm_helper_hpd_irq_event</function>. The function will
1501 queue a delayed work to check the state of all connectors, but no
1502 periodic polling will be done.
1506 <title>Connector Operations</title>
1508 Unless otherwise state, all operations are mandatory.
1512 <synopsis>void (*dpms)(struct drm_connector *connector, int mode);</synopsis>
1514 The DPMS operation sets the power state of a connector. The mode
1517 <listitem><para>DRM_MODE_DPMS_ON</para></listitem>
1518 <listitem><para>DRM_MODE_DPMS_STANDBY</para></listitem>
1519 <listitem><para>DRM_MODE_DPMS_SUSPEND</para></listitem>
1520 <listitem><para>DRM_MODE_DPMS_OFF</para></listitem>
1524 In all but DPMS_ON mode the encoder to which the connector is attached
1525 should put the display in low-power mode by driving its signals
1526 appropriately. If more than one connector is attached to the encoder
1527 care should be taken not to change the power state of other displays as
1528 a side effect. Low-power mode should be propagated to the encoders and
1529 CRTCs when all related connectors are put in low-power mode.
1533 <title>Modes</title>
1534 <synopsis>int (*fill_modes)(struct drm_connector *connector, uint32_t max_width,
1535 uint32_t max_height);</synopsis>
1537 Fill the mode list with all supported modes for the connector. If the
1538 <parameter>max_width</parameter> and <parameter>max_height</parameter>
1539 arguments are non-zero, the implementation must ignore all modes wider
1540 than <parameter>max_width</parameter> or higher than
1541 <parameter>max_height</parameter>.
1544 The connector must also fill in this operation its
1545 <structfield>display_info</structfield>
1546 <structfield>width_mm</structfield> and
1547 <structfield>height_mm</structfield> fields with the connected display
1548 physical size in millimeters. The fields should be set to 0 if the value
1549 isn't known or is not applicable (for instance for projector devices).
1553 <title>Connection Status</title>
1555 The connection status is updated through polling or hotplug events when
1556 supported (see <xref linkend="drm-kms-connector-polled"/>). The status
1557 value is reported to userspace through ioctls and must not be used
1558 inside the driver, as it only gets initialized by a call to
1559 <function>drm_mode_getconnector</function> from userspace.
1561 <synopsis>enum drm_connector_status (*detect)(struct drm_connector *connector,
1562 bool force);</synopsis>
1564 Check to see if anything is attached to the connector. The
1565 <parameter>force</parameter> parameter is set to false whilst polling or
1566 to true when checking the connector due to user request.
1567 <parameter>force</parameter> can be used by the driver to avoid
1568 expensive, destructive operations during automated probing.
1571 Return connector_status_connected if something is connected to the
1572 connector, connector_status_disconnected if nothing is connected and
1573 connector_status_unknown if the connection state isn't known.
1576 Drivers should only return connector_status_connected if the connection
1577 status has really been probed as connected. Connectors that can't detect
1578 the connection status, or failed connection status probes, should return
1579 connector_status_unknown.
1583 <title>Miscellaneous</title>
1586 <synopsis>void (*set_property)(struct drm_connector *connector,
1587 struct drm_property *property, uint64_t value);</synopsis>
1589 Set the value of the given connector property to
1590 <parameter>value</parameter>. See <xref linkend="drm-kms-properties"/>
1591 for more information about properties.
1595 <synopsis>void (*destroy)(struct drm_connector *connector);</synopsis>
1597 Destroy the connector when not needed anymore. See
1598 <xref linkend="drm-kms-init"/>.
1606 <title>Cleanup</title>
1608 The DRM core manages its objects' lifetime. When an object is not needed
1609 anymore the core calls its destroy function, which must clean up and
1610 free every resource allocated for the object. Every
1611 <function>drm_*_init</function> call must be matched with a
1612 corresponding <function>drm_*_cleanup</function> call to cleanup CRTCs
1613 (<function>drm_crtc_cleanup</function>), planes
1614 (<function>drm_plane_cleanup</function>), encoders
1615 (<function>drm_encoder_cleanup</function>) and connectors
1616 (<function>drm_connector_cleanup</function>). Furthermore, connectors
1617 that have been added to sysfs must be removed by a call to
1618 <function>drm_sysfs_connector_remove</function> before calling
1619 <function>drm_connector_cleanup</function>.
1622 Connectors state change detection must be cleanup up with a call to
1623 <function>drm_kms_helper_poll_fini</function>.
1627 <title>Output discovery and initialization example</title>
1628 <programlisting><![CDATA[
1629 void intel_crt_init(struct drm_device *dev)
1631 struct drm_connector *connector;
1632 struct intel_output *intel_output;
1634 intel_output = kzalloc(sizeof(struct intel_output), GFP_KERNEL);
1638 connector = &intel_output->base;
1639 drm_connector_init(dev, &intel_output->base,
1640 &intel_crt_connector_funcs, DRM_MODE_CONNECTOR_VGA);
1642 drm_encoder_init(dev, &intel_output->enc, &intel_crt_enc_funcs,
1643 DRM_MODE_ENCODER_DAC);
1645 drm_mode_connector_attach_encoder(&intel_output->base,
1646 &intel_output->enc);
1648 /* Set up the DDC bus. */
1649 intel_output->ddc_bus = intel_i2c_create(dev, GPIOA, "CRTDDC_A");
1650 if (!intel_output->ddc_bus) {
1651 dev_printk(KERN_ERR, &dev->pdev->dev, "DDC bus registration "
1656 intel_output->type = INTEL_OUTPUT_ANALOG;
1657 connector->interlace_allowed = 0;
1658 connector->doublescan_allowed = 0;
1660 drm_encoder_helper_add(&intel_output->enc, &intel_crt_helper_funcs);
1661 drm_connector_helper_add(connector, &intel_crt_connector_helper_funcs);
1663 drm_sysfs_connector_add(connector);
1664 }]]></programlisting>
1666 In the example above (taken from the i915 driver), a CRTC, connector and
1667 encoder combination is created. A device-specific i2c bus is also
1668 created for fetching EDID data and performing monitor detection. Once
1669 the process is complete, the new connector is registered with sysfs to
1670 make its properties available to applications.
1674 <title>KMS API Functions</title>
1675 !Edrivers/gpu/drm/drm_crtc.c
1679 <!-- Internals: kms helper functions -->
1682 <title>Mode Setting Helper Functions</title>
1684 The CRTC, encoder and connector functions provided by the drivers
1685 implement the DRM API. They're called by the DRM core and ioctl handlers
1686 to handle device state changes and configuration request. As implementing
1687 those functions often requires logic not specific to drivers, mid-layer
1688 helper functions are available to avoid duplicating boilerplate code.
1691 The DRM core contains one mid-layer implementation. The mid-layer provides
1692 implementations of several CRTC, encoder and connector functions (called
1693 from the top of the mid-layer) that pre-process requests and call
1694 lower-level functions provided by the driver (at the bottom of the
1695 mid-layer). For instance, the
1696 <function>drm_crtc_helper_set_config</function> function can be used to
1697 fill the struct <structname>drm_crtc_funcs</structname>
1698 <structfield>set_config</structfield> field. When called, it will split
1699 the <methodname>set_config</methodname> operation in smaller, simpler
1700 operations and call the driver to handle them.
1703 To use the mid-layer, drivers call <function>drm_crtc_helper_add</function>,
1704 <function>drm_encoder_helper_add</function> and
1705 <function>drm_connector_helper_add</function> functions to install their
1706 mid-layer bottom operations handlers, and fill the
1707 <structname>drm_crtc_funcs</structname>,
1708 <structname>drm_encoder_funcs</structname> and
1709 <structname>drm_connector_funcs</structname> structures with pointers to
1710 the mid-layer top API functions. Installing the mid-layer bottom operation
1711 handlers is best done right after registering the corresponding KMS object.
1714 The mid-layer is not split between CRTC, encoder and connector operations.
1715 To use it, a driver must provide bottom functions for all of the three KMS
1719 <title>Helper Functions</title>
1722 <synopsis>int drm_crtc_helper_set_config(struct drm_mode_set *set);</synopsis>
1724 The <function>drm_crtc_helper_set_config</function> helper function
1725 is a CRTC <methodname>set_config</methodname> implementation. It
1726 first tries to locate the best encoder for each connector by calling
1727 the connector <methodname>best_encoder</methodname> helper
1731 After locating the appropriate encoders, the helper function will
1732 call the <methodname>mode_fixup</methodname> encoder and CRTC helper
1733 operations to adjust the requested mode, or reject it completely in
1734 which case an error will be returned to the application. If the new
1735 configuration after mode adjustment is identical to the current
1736 configuration the helper function will return without performing any
1740 If the adjusted mode is identical to the current mode but changes to
1741 the frame buffer need to be applied, the
1742 <function>drm_crtc_helper_set_config</function> function will call
1743 the CRTC <methodname>mode_set_base</methodname> helper operation. If
1744 the adjusted mode differs from the current mode, or if the
1745 <methodname>mode_set_base</methodname> helper operation is not
1746 provided, the helper function performs a full mode set sequence by
1747 calling the <methodname>prepare</methodname>,
1748 <methodname>mode_set</methodname> and
1749 <methodname>commit</methodname> CRTC and encoder helper operations,
1754 <synopsis>void drm_helper_connector_dpms(struct drm_connector *connector, int mode);</synopsis>
1756 The <function>drm_helper_connector_dpms</function> helper function
1757 is a connector <methodname>dpms</methodname> implementation that
1758 tracks power state of connectors. To use the function, drivers must
1759 provide <methodname>dpms</methodname> helper operations for CRTCs
1760 and encoders to apply the DPMS state to the device.
1763 The mid-layer doesn't track the power state of CRTCs and encoders.
1764 The <methodname>dpms</methodname> helper operations can thus be
1765 called with a mode identical to the currently active mode.
1769 <synopsis>int drm_helper_probe_single_connector_modes(struct drm_connector *connector,
1770 uint32_t maxX, uint32_t maxY);</synopsis>
1772 The <function>drm_helper_probe_single_connector_modes</function> helper
1773 function is a connector <methodname>fill_modes</methodname>
1774 implementation that updates the connection status for the connector
1775 and then retrieves a list of modes by calling the connector
1776 <methodname>get_modes</methodname> helper operation.
1779 The function filters out modes larger than
1780 <parameter>max_width</parameter> and <parameter>max_height</parameter>
1781 if specified. It then calls the connector
1782 <methodname>mode_valid</methodname> helper operation for each mode in
1783 the probed list to check whether the mode is valid for the connector.
1789 <title>CRTC Helper Operations</title>
1791 <listitem id="drm-helper-crtc-mode-fixup">
1792 <synopsis>bool (*mode_fixup)(struct drm_crtc *crtc,
1793 const struct drm_display_mode *mode,
1794 struct drm_display_mode *adjusted_mode);</synopsis>
1796 Let CRTCs adjust the requested mode or reject it completely. This
1797 operation returns true if the mode is accepted (possibly after being
1798 adjusted) or false if it is rejected.
1801 The <methodname>mode_fixup</methodname> operation should reject the
1802 mode if it can't reasonably use it. The definition of "reasonable"
1803 is currently fuzzy in this context. One possible behaviour would be
1804 to set the adjusted mode to the panel timings when a fixed-mode
1805 panel is used with hardware capable of scaling. Another behaviour
1806 would be to accept any input mode and adjust it to the closest mode
1807 supported by the hardware (FIXME: This needs to be clarified).
1811 <synopsis>int (*mode_set_base)(struct drm_crtc *crtc, int x, int y,
1812 struct drm_framebuffer *old_fb)</synopsis>
1814 Move the CRTC on the current frame buffer (stored in
1815 <literal>crtc->fb</literal>) to position (x,y). Any of the frame
1816 buffer, x position or y position may have been modified.
1819 This helper operation is optional. If not provided, the
1820 <function>drm_crtc_helper_set_config</function> function will fall
1821 back to the <methodname>mode_set</methodname> helper operation.
1824 FIXME: Why are x and y passed as arguments, as they can be accessed
1825 through <literal>crtc->x</literal> and
1826 <literal>crtc->y</literal>?
1830 <synopsis>void (*prepare)(struct drm_crtc *crtc);</synopsis>
1832 Prepare the CRTC for mode setting. This operation is called after
1833 validating the requested mode. Drivers use it to perform
1834 device-specific operations required before setting the new mode.
1838 <synopsis>int (*mode_set)(struct drm_crtc *crtc, struct drm_display_mode *mode,
1839 struct drm_display_mode *adjusted_mode, int x, int y,
1840 struct drm_framebuffer *old_fb);</synopsis>
1842 Set a new mode, position and frame buffer. Depending on the device
1843 requirements, the mode can be stored internally by the driver and
1844 applied in the <methodname>commit</methodname> operation, or
1845 programmed to the hardware immediately.
1848 The <methodname>mode_set</methodname> operation returns 0 on success
1849 or a negative error code if an error occurs.
1853 <synopsis>void (*commit)(struct drm_crtc *crtc);</synopsis>
1855 Commit a mode. This operation is called after setting the new mode.
1856 Upon return the device must use the new mode and be fully
1863 <title>Encoder Helper Operations</title>
1866 <synopsis>bool (*mode_fixup)(struct drm_encoder *encoder,
1867 const struct drm_display_mode *mode,
1868 struct drm_display_mode *adjusted_mode);</synopsis>
1870 Let encoders adjust the requested mode or reject it completely. This
1871 operation returns true if the mode is accepted (possibly after being
1872 adjusted) or false if it is rejected. See the
1873 <link linkend="drm-helper-crtc-mode-fixup">mode_fixup CRTC helper
1874 operation</link> for an explanation of the allowed adjustments.
1878 <synopsis>void (*prepare)(struct drm_encoder *encoder);</synopsis>
1880 Prepare the encoder for mode setting. This operation is called after
1881 validating the requested mode. Drivers use it to perform
1882 device-specific operations required before setting the new mode.
1886 <synopsis>void (*mode_set)(struct drm_encoder *encoder,
1887 struct drm_display_mode *mode,
1888 struct drm_display_mode *adjusted_mode);</synopsis>
1890 Set a new mode. Depending on the device requirements, the mode can
1891 be stored internally by the driver and applied in the
1892 <methodname>commit</methodname> operation, or programmed to the
1893 hardware immediately.
1897 <synopsis>void (*commit)(struct drm_encoder *encoder);</synopsis>
1899 Commit a mode. This operation is called after setting the new mode.
1900 Upon return the device must use the new mode and be fully
1907 <title>Connector Helper Operations</title>
1910 <synopsis>struct drm_encoder *(*best_encoder)(struct drm_connector *connector);</synopsis>
1912 Return a pointer to the best encoder for the connecter. Device that
1913 map connectors to encoders 1:1 simply return the pointer to the
1914 associated encoder. This operation is mandatory.
1918 <synopsis>int (*get_modes)(struct drm_connector *connector);</synopsis>
1920 Fill the connector's <structfield>probed_modes</structfield> list
1921 by parsing EDID data with <function>drm_add_edid_modes</function> or
1922 calling <function>drm_mode_probed_add</function> directly for every
1923 supported mode and return the number of modes it has detected. This
1924 operation is mandatory.
1927 When adding modes manually the driver creates each mode with a call to
1928 <function>drm_mode_create</function> and must fill the following fields.
1931 <synopsis>__u32 type;</synopsis>
1933 Mode type bitmask, a combination of
1936 <term>DRM_MODE_TYPE_BUILTIN</term>
1937 <listitem><para>not used?</para></listitem>
1940 <term>DRM_MODE_TYPE_CLOCK_C</term>
1941 <listitem><para>not used?</para></listitem>
1944 <term>DRM_MODE_TYPE_CRTC_C</term>
1945 <listitem><para>not used?</para></listitem>
1949 DRM_MODE_TYPE_PREFERRED - The preferred mode for the connector
1952 <para>not used?</para>
1956 <term>DRM_MODE_TYPE_DEFAULT</term>
1957 <listitem><para>not used?</para></listitem>
1960 <term>DRM_MODE_TYPE_USERDEF</term>
1961 <listitem><para>not used?</para></listitem>
1964 <term>DRM_MODE_TYPE_DRIVER</term>
1967 The mode has been created by the driver (as opposed to
1968 to user-created modes).
1973 Drivers must set the DRM_MODE_TYPE_DRIVER bit for all modes they
1974 create, and set the DRM_MODE_TYPE_PREFERRED bit for the preferred
1979 <synopsis>__u32 clock;</synopsis>
1980 <para>Pixel clock frequency in kHz unit</para>
1983 <synopsis>__u16 hdisplay, hsync_start, hsync_end, htotal;
1984 __u16 vdisplay, vsync_start, vsync_end, vtotal;</synopsis>
1985 <para>Horizontal and vertical timing information</para>
1987 Active Front Sync Back
1989 <-----------------------><----------------><-------------><-------------->
1991 //////////////////////|
1992 ////////////////////// |
1993 ////////////////////// |.................. ................
1996 <----- [hv]display ----->
1997 <------------- [hv]sync_start ------------>
1998 <--------------------- [hv]sync_end --------------------->
1999 <-------------------------------- [hv]total ----------------------------->
2003 <synopsis>__u16 hskew;
2004 __u16 vscan;</synopsis>
2005 <para>Unknown</para>
2008 <synopsis>__u32 flags;</synopsis>
2010 Mode flags, a combination of
2013 <term>DRM_MODE_FLAG_PHSYNC</term>
2015 Horizontal sync is active high
2019 <term>DRM_MODE_FLAG_NHSYNC</term>
2021 Horizontal sync is active low
2025 <term>DRM_MODE_FLAG_PVSYNC</term>
2027 Vertical sync is active high
2031 <term>DRM_MODE_FLAG_NVSYNC</term>
2033 Vertical sync is active low
2037 <term>DRM_MODE_FLAG_INTERLACE</term>
2043 <term>DRM_MODE_FLAG_DBLSCAN</term>
2045 Mode uses doublescan
2049 <term>DRM_MODE_FLAG_CSYNC</term>
2051 Mode uses composite sync
2055 <term>DRM_MODE_FLAG_PCSYNC</term>
2057 Composite sync is active high
2061 <term>DRM_MODE_FLAG_NCSYNC</term>
2063 Composite sync is active low
2067 <term>DRM_MODE_FLAG_HSKEW</term>
2069 hskew provided (not used?)
2073 <term>DRM_MODE_FLAG_BCAST</term>
2079 <term>DRM_MODE_FLAG_PIXMUX</term>
2085 <term>DRM_MODE_FLAG_DBLCLK</term>
2091 <term>DRM_MODE_FLAG_CLKDIV2</term>
2099 Note that modes marked with the INTERLACE or DBLSCAN flags will be
2101 <function>drm_helper_probe_single_connector_modes</function> if
2102 the connector's <structfield>interlace_allowed</structfield> or
2103 <structfield>doublescan_allowed</structfield> field is set to 0.
2107 <synopsis>char name[DRM_DISPLAY_MODE_LEN];</synopsis>
2109 Mode name. The driver must call
2110 <function>drm_mode_set_name</function> to fill the mode name from
2111 <structfield>hdisplay</structfield>,
2112 <structfield>vdisplay</structfield> and interlace flag after
2113 filling the corresponding fields.
2119 The <structfield>vrefresh</structfield> value is computed by
2120 <function>drm_helper_probe_single_connector_modes</function>.
2123 When parsing EDID data, <function>drm_add_edid_modes</function> fill the
2124 connector <structfield>display_info</structfield>
2125 <structfield>width_mm</structfield> and
2126 <structfield>height_mm</structfield> fields. When creating modes
2127 manually the <methodname>get_modes</methodname> helper operation must
2128 set the <structfield>display_info</structfield>
2129 <structfield>width_mm</structfield> and
2130 <structfield>height_mm</structfield> fields if they haven't been set
2131 already (for instance at initilization time when a fixed-size panel is
2132 attached to the connector). The mode <structfield>width_mm</structfield>
2133 and <structfield>height_mm</structfield> fields are only used internally
2134 during EDID parsing and should not be set when creating modes manually.
2138 <synopsis>int (*mode_valid)(struct drm_connector *connector,
2139 struct drm_display_mode *mode);</synopsis>
2141 Verify whether a mode is valid for the connector. Return MODE_OK for
2142 supported modes and one of the enum drm_mode_status values (MODE_*)
2143 for unsupported modes. This operation is mandatory.
2146 As the mode rejection reason is currently not used beside for
2147 immediately removing the unsupported mode, an implementation can
2148 return MODE_BAD regardless of the exact reason why the mode is not
2152 Note that the <methodname>mode_valid</methodname> helper operation is
2153 only called for modes detected by the device, and
2154 <emphasis>not</emphasis> for modes set by the user through the CRTC
2155 <methodname>set_config</methodname> operation.
2161 <title>Modeset Helper Functions Reference</title>
2162 !Edrivers/gpu/drm/drm_crtc_helper.c
2165 <title>fbdev Helper Functions Reference</title>
2166 !Pdrivers/gpu/drm/drm_fb_helper.c fbdev helpers
2167 !Edrivers/gpu/drm/drm_fb_helper.c
2168 !Iinclude/drm/drm_fb_helper.h
2171 <title>Display Port Helper Functions Reference</title>
2172 !Pdrivers/gpu/drm/drm_dp_helper.c dp helpers
2173 !Iinclude/drm/drm_dp_helper.h
2174 !Edrivers/gpu/drm/drm_dp_helper.c
2177 <title>EDID Helper Functions Reference</title>
2178 !Edrivers/gpu/drm/drm_edid.c
2181 <title>Rectangle Utilities Reference</title>
2182 !Pinclude/drm/drm_rect.h rect utils
2183 !Iinclude/drm/drm_rect.h
2184 !Edrivers/gpu/drm/drm_rect.c
2187 <title>Flip-work Helper Reference</title>
2188 !Pinclude/drm/drm_flip_work.h flip utils
2189 !Iinclude/drm/drm_flip_work.h
2190 !Edrivers/gpu/drm/drm_flip_work.c
2193 <title>VMA Offset Manager</title>
2194 !Pdrivers/gpu/drm/drm_vma_manager.c vma offset manager
2195 !Edrivers/gpu/drm/drm_vma_manager.c
2196 !Iinclude/drm/drm_vma_manager.h
2200 <!-- Internals: kms properties -->
2202 <sect1 id="drm-kms-properties">
2203 <title>KMS Properties</title>
2205 Drivers may need to expose additional parameters to applications than
2206 those described in the previous sections. KMS supports attaching
2207 properties to CRTCs, connectors and planes and offers a userspace API to
2208 list, get and set the property values.
2211 Properties are identified by a name that uniquely defines the property
2212 purpose, and store an associated value. For all property types except blob
2213 properties the value is a 64-bit unsigned integer.
2216 KMS differentiates between properties and property instances. Drivers
2217 first create properties and then create and associate individual instances
2218 of those properties to objects. A property can be instantiated multiple
2219 times and associated with different objects. Values are stored in property
2220 instances, and all other property information are stored in the propery
2221 and shared between all instances of the property.
2224 Every property is created with a type that influences how the KMS core
2225 handles the property. Supported property types are
2228 <term>DRM_MODE_PROP_RANGE</term>
2229 <listitem><para>Range properties report their minimum and maximum
2230 admissible values. The KMS core verifies that values set by
2231 application fit in that range.</para></listitem>
2234 <term>DRM_MODE_PROP_ENUM</term>
2235 <listitem><para>Enumerated properties take a numerical value that
2236 ranges from 0 to the number of enumerated values defined by the
2237 property minus one, and associate a free-formed string name to each
2238 value. Applications can retrieve the list of defined value-name pairs
2239 and use the numerical value to get and set property instance values.
2243 <term>DRM_MODE_PROP_BITMASK</term>
2244 <listitem><para>Bitmask properties are enumeration properties that
2245 additionally restrict all enumerated values to the 0..63 range.
2246 Bitmask property instance values combine one or more of the
2247 enumerated bits defined by the property.</para></listitem>
2250 <term>DRM_MODE_PROP_BLOB</term>
2251 <listitem><para>Blob properties store a binary blob without any format
2252 restriction. The binary blobs are created as KMS standalone objects,
2253 and blob property instance values store the ID of their associated
2255 <para>Blob properties are only used for the connector EDID property
2256 and cannot be created by drivers.</para></listitem>
2261 To create a property drivers call one of the following functions depending
2262 on the property type. All property creation functions take property flags
2263 and name, as well as type-specific arguments.
2266 <synopsis>struct drm_property *drm_property_create_range(struct drm_device *dev, int flags,
2268 uint64_t min, uint64_t max);</synopsis>
2269 <para>Create a range property with the given minimum and maximum
2273 <synopsis>struct drm_property *drm_property_create_enum(struct drm_device *dev, int flags,
2275 const struct drm_prop_enum_list *props,
2276 int num_values);</synopsis>
2277 <para>Create an enumerated property. The <parameter>props</parameter>
2278 argument points to an array of <parameter>num_values</parameter>
2279 value-name pairs.</para>
2282 <synopsis>struct drm_property *drm_property_create_bitmask(struct drm_device *dev,
2283 int flags, const char *name,
2284 const struct drm_prop_enum_list *props,
2285 int num_values);</synopsis>
2286 <para>Create a bitmask property. The <parameter>props</parameter>
2287 argument points to an array of <parameter>num_values</parameter>
2288 value-name pairs.</para>
2293 Properties can additionally be created as immutable, in which case they
2294 will be read-only for applications but can be modified by the driver. To
2295 create an immutable property drivers must set the DRM_MODE_PROP_IMMUTABLE
2296 flag at property creation time.
2299 When no array of value-name pairs is readily available at property
2300 creation time for enumerated or range properties, drivers can create
2301 the property using the <function>drm_property_create</function> function
2302 and manually add enumeration value-name pairs by calling the
2303 <function>drm_property_add_enum</function> function. Care must be taken to
2304 properly specify the property type through the <parameter>flags</parameter>
2308 After creating properties drivers can attach property instances to CRTC,
2309 connector and plane objects by calling the
2310 <function>drm_object_attach_property</function>. The function takes a
2311 pointer to the target object, a pointer to the previously created property
2312 and an initial instance value.
2316 <!-- Internals: vertical blanking -->
2318 <sect1 id="drm-vertical-blank">
2319 <title>Vertical Blanking</title>
2321 Vertical blanking plays a major role in graphics rendering. To achieve
2322 tear-free display, users must synchronize page flips and/or rendering to
2323 vertical blanking. The DRM API offers ioctls to perform page flips
2324 synchronized to vertical blanking and wait for vertical blanking.
2327 The DRM core handles most of the vertical blanking management logic, which
2328 involves filtering out spurious interrupts, keeping race-free blanking
2329 counters, coping with counter wrap-around and resets and keeping use
2330 counts. It relies on the driver to generate vertical blanking interrupts
2331 and optionally provide a hardware vertical blanking counter. Drivers must
2332 implement the following operations.
2336 <synopsis>int (*enable_vblank) (struct drm_device *dev, int crtc);
2337 void (*disable_vblank) (struct drm_device *dev, int crtc);</synopsis>
2339 Enable or disable vertical blanking interrupts for the given CRTC.
2343 <synopsis>u32 (*get_vblank_counter) (struct drm_device *dev, int crtc);</synopsis>
2345 Retrieve the value of the vertical blanking counter for the given
2346 CRTC. If the hardware maintains a vertical blanking counter its value
2347 should be returned. Otherwise drivers can use the
2348 <function>drm_vblank_count</function> helper function to handle this
2354 Drivers must initialize the vertical blanking handling core with a call to
2355 <function>drm_vblank_init</function> in their
2356 <methodname>load</methodname> operation. The function will set the struct
2357 <structname>drm_device</structname>
2358 <structfield>vblank_disable_allowed</structfield> field to 0. This will
2359 keep vertical blanking interrupts enabled permanently until the first mode
2360 set operation, where <structfield>vblank_disable_allowed</structfield> is
2361 set to 1. The reason behind this is not clear. Drivers can set the field
2362 to 1 after <function>calling drm_vblank_init</function> to make vertical
2363 blanking interrupts dynamically managed from the beginning.
2366 Vertical blanking interrupts can be enabled by the DRM core or by drivers
2367 themselves (for instance to handle page flipping operations). The DRM core
2368 maintains a vertical blanking use count to ensure that the interrupts are
2369 not disabled while a user still needs them. To increment the use count,
2370 drivers call <function>drm_vblank_get</function>. Upon return vertical
2371 blanking interrupts are guaranteed to be enabled.
2374 To decrement the use count drivers call
2375 <function>drm_vblank_put</function>. Only when the use count drops to zero
2376 will the DRM core disable the vertical blanking interrupts after a delay
2377 by scheduling a timer. The delay is accessible through the vblankoffdelay
2378 module parameter or the <varname>drm_vblank_offdelay</varname> global
2379 variable and expressed in milliseconds. Its default value is 5000 ms.
2382 When a vertical blanking interrupt occurs drivers only need to call the
2383 <function>drm_handle_vblank</function> function to account for the
2387 Resources allocated by <function>drm_vblank_init</function> must be freed
2388 with a call to <function>drm_vblank_cleanup</function> in the driver
2389 <methodname>unload</methodname> operation handler.
2393 <!-- Internals: open/close, file operations and ioctls -->
2396 <title>Open/Close, File Operations and IOCTLs</title>
2398 <title>Open and Close</title>
2399 <synopsis>int (*firstopen) (struct drm_device *);
2400 void (*lastclose) (struct drm_device *);
2401 int (*open) (struct drm_device *, struct drm_file *);
2402 void (*preclose) (struct drm_device *, struct drm_file *);
2403 void (*postclose) (struct drm_device *, struct drm_file *);</synopsis>
2404 <abstract>Open and close handlers. None of those methods are mandatory.
2407 The <methodname>firstopen</methodname> method is called by the DRM core
2408 for legacy UMS (User Mode Setting) drivers only when an application
2409 opens a device that has no other opened file handle. UMS drivers can
2410 implement it to acquire device resources. KMS drivers can't use the
2411 method and must acquire resources in the <methodname>load</methodname>
2415 Similarly the <methodname>lastclose</methodname> method is called when
2416 the last application holding a file handle opened on the device closes
2417 it, for both UMS and KMS drivers. Additionally, the method is also
2418 called at module unload time or, for hot-pluggable devices, when the
2419 device is unplugged. The <methodname>firstopen</methodname> and
2420 <methodname>lastclose</methodname> calls can thus be unbalanced.
2423 The <methodname>open</methodname> method is called every time the device
2424 is opened by an application. Drivers can allocate per-file private data
2425 in this method and store them in the struct
2426 <structname>drm_file</structname> <structfield>driver_priv</structfield>
2427 field. Note that the <methodname>open</methodname> method is called
2428 before <methodname>firstopen</methodname>.
2431 The close operation is split into <methodname>preclose</methodname> and
2432 <methodname>postclose</methodname> methods. Drivers must stop and
2433 cleanup all per-file operations in the <methodname>preclose</methodname>
2434 method. For instance pending vertical blanking and page flip events must
2435 be cancelled. No per-file operation is allowed on the file handle after
2436 returning from the <methodname>preclose</methodname> method.
2439 Finally the <methodname>postclose</methodname> method is called as the
2440 last step of the close operation, right before calling the
2441 <methodname>lastclose</methodname> method if no other open file handle
2442 exists for the device. Drivers that have allocated per-file private data
2443 in the <methodname>open</methodname> method should free it here.
2446 The <methodname>lastclose</methodname> method should restore CRTC and
2447 plane properties to default value, so that a subsequent open of the
2448 device will not inherit state from the previous user. It can also be
2449 used to execute delayed power switching state changes, e.g. in
2450 conjunction with the vga-switcheroo infrastructure. Beyond that KMS
2451 drivers should not do any further cleanup. Only legacy UMS drivers might
2452 need to clean up device state so that the vga console or an independent
2453 fbdev driver could take over.
2457 <title>File Operations</title>
2458 <synopsis>const struct file_operations *fops</synopsis>
2459 <abstract>File operations for the DRM device node.</abstract>
2461 Drivers must define the file operations structure that forms the DRM
2462 userspace API entry point, even though most of those operations are
2463 implemented in the DRM core. The <methodname>open</methodname>,
2464 <methodname>release</methodname> and <methodname>ioctl</methodname>
2465 operations are handled by
2467 .owner = THIS_MODULE,
2469 .release = drm_release,
2470 .unlocked_ioctl = drm_ioctl,
2471 #ifdef CONFIG_COMPAT
2472 .compat_ioctl = drm_compat_ioctl,
2477 Drivers that implement private ioctls that requires 32/64bit
2478 compatibility support must provide their own
2479 <methodname>compat_ioctl</methodname> handler that processes private
2480 ioctls and calls <function>drm_compat_ioctl</function> for core ioctls.
2483 The <methodname>read</methodname> and <methodname>poll</methodname>
2484 operations provide support for reading DRM events and polling them. They
2489 .llseek = no_llseek,
2493 The memory mapping implementation varies depending on how the driver
2494 manages memory. Pre-GEM drivers will use <function>drm_mmap</function>,
2495 while GEM-aware drivers will use <function>drm_gem_mmap</function>. See
2496 <xref linkend="drm-gem"/>.
2498 .mmap = drm_gem_mmap,
2502 No other file operation is supported by the DRM API.
2506 <title>IOCTLs</title>
2507 <synopsis>struct drm_ioctl_desc *ioctls;
2508 int num_ioctls;</synopsis>
2509 <abstract>Driver-specific ioctls descriptors table.</abstract>
2511 Driver-specific ioctls numbers start at DRM_COMMAND_BASE. The ioctls
2512 descriptors table is indexed by the ioctl number offset from the base
2513 value. Drivers can use the DRM_IOCTL_DEF_DRV() macro to initialize the
2517 <programlisting>DRM_IOCTL_DEF_DRV(ioctl, func, flags)</programlisting>
2519 <parameter>ioctl</parameter> is the ioctl name. Drivers must define
2520 the DRM_##ioctl and DRM_IOCTL_##ioctl macros to the ioctl number
2521 offset from DRM_COMMAND_BASE and the ioctl number respectively. The
2522 first macro is private to the device while the second must be exposed
2523 to userspace in a public header.
2526 <parameter>func</parameter> is a pointer to the ioctl handler function
2527 compatible with the <type>drm_ioctl_t</type> type.
2528 <programlisting>typedef int drm_ioctl_t(struct drm_device *dev, void *data,
2529 struct drm_file *file_priv);</programlisting>
2532 <parameter>flags</parameter> is a bitmask combination of the following
2533 values. It restricts how the ioctl is allowed to be called.
2536 DRM_AUTH - Only authenticated callers allowed
2539 DRM_MASTER - The ioctl can only be called on the master file
2543 DRM_ROOT_ONLY - Only callers with the SYSADMIN capability allowed
2546 DRM_CONTROL_ALLOW - The ioctl can only be called on a control
2550 DRM_UNLOCKED - The ioctl handler will be called without locking
2551 the DRM global mutex
2560 <title>Command submission & fencing</title>
2562 This should cover a few device-specific command submission
2567 <!-- Internals: suspend/resume -->
2570 <title>Suspend/Resume</title>
2572 The DRM core provides some suspend/resume code, but drivers wanting full
2573 suspend/resume support should provide save() and restore() functions.
2574 These are called at suspend, hibernate, or resume time, and should perform
2575 any state save or restore required by your device across suspend or
2578 <synopsis>int (*suspend) (struct drm_device *, pm_message_t state);
2579 int (*resume) (struct drm_device *);</synopsis>
2581 Those are legacy suspend and resume methods. New driver should use the
2582 power management interface provided by their bus type (usually through
2583 the struct <structname>device_driver</structname> dev_pm_ops) and set
2584 these methods to NULL.
2589 <title>DMA services</title>
2591 This should cover how DMA mapping etc. is supported by the core.
2592 These functions are deprecated and should not be used.
2600 - Document the struct_mutex catch-all lock
2601 - Document connector properties
2603 - Why is the load method optional?
2604 - What are drivers supposed to set the initial display state to, and how?
2605 Connector's DPMS states are not initialized and are thus equal to
2606 DRM_MODE_DPMS_ON. The fbcon compatibility layer calls
2607 drm_helper_disable_unused_functions(), which disables unused encoders and
2608 CRTCs, but doesn't touch the connectors' DPMS state, and
2609 drm_helper_connector_dpms() in reaction to fbdev blanking events. Do drivers
2610 that don't implement (or just don't use) fbcon compatibility need to call
2611 those functions themselves?
2612 - KMS drivers must call drm_vblank_pre_modeset() and drm_vblank_post_modeset()
2613 around mode setting. Should this be done in the DRM core?
2614 - vblank_disable_allowed is set to 1 in the first drm_vblank_post_modeset()
2615 call and never set back to 0. It seems to be safe to permanently set it to 1
2616 in drm_vblank_init() for KMS driver, and it might be safe for UMS drivers as
2617 well. This should be investigated.
2618 - crtc and connector .save and .restore operations are only used internally in
2619 drivers, should they be removed from the core?
2620 - encoder mid-layer .save and .restore operations are only used internally in
2621 drivers, should they be removed from the core?
2622 - encoder mid-layer .detect operation is only used internally in drivers,
2623 should it be removed from the core?
2626 <!-- External interfaces -->
2628 <chapter id="drmExternals">
2629 <title>Userland interfaces</title>
2631 The DRM core exports several interfaces to applications,
2632 generally intended to be used through corresponding libdrm
2633 wrapper functions. In addition, drivers export device-specific
2634 interfaces for use by userspace drivers & device-aware
2635 applications through ioctls and sysfs files.
2638 External interfaces include: memory mapping, context management,
2639 DMA operations, AGP management, vblank control, fence
2640 management, memory management, and output management.
2643 Cover generic ioctls and sysfs layout here. We only need high-level
2644 info, since man pages should cover the rest.
2647 <!-- External: vblank handling -->
2650 <title>VBlank event handling</title>
2652 The DRM core exposes two vertical blank related ioctls:
2655 <term>DRM_IOCTL_WAIT_VBLANK</term>
2658 This takes a struct drm_wait_vblank structure as its argument,
2659 and it is used to block or request a signal when a specified
2660 vblank event occurs.
2665 <term>DRM_IOCTL_MODESET_CTL</term>
2668 This should be called by application level drivers before and
2669 after mode setting, since on many devices the vertical blank
2670 counter is reset at that time. Internally, the DRM snapshots
2671 the last vblank count when the ioctl is called with the
2672 _DRM_PRE_MODESET command, so that the counter won't go backwards
2673 (which is dealt with when _DRM_POST_MODESET is used).
2678 <!--!Edrivers/char/drm/drm_irq.c-->
2684 <!-- API reference -->
2686 <appendix id="drmDriverApi">
2687 <title>DRM Driver API</title>
2689 Include auto-generated API reference here (need to reference it
2690 from paragraphs above too).
projects / platform / adaptation / renesas_rcar / renesas_kernel.git / blob