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.
209 <term>DRIVER_RENDER</term>
211 Driver supports dedicated render nodes.
217 <title>Major, Minor and Patchlevel</title>
220 int patchlevel;</synopsis>
222 The DRM core identifies driver versions by a major, minor and patch
223 level triplet. The information is printed to the kernel log at
224 initialization time and passed to userspace through the
225 DRM_IOCTL_VERSION ioctl.
228 The major and minor numbers are also used to verify the requested driver
229 API version passed to DRM_IOCTL_SET_VERSION. When the driver API changes
230 between minor versions, applications can call DRM_IOCTL_SET_VERSION to
231 select a specific version of the API. If the requested major isn't equal
232 to the driver major, or the requested minor is larger than the driver
233 minor, the DRM_IOCTL_SET_VERSION call will return an error. Otherwise
234 the driver's set_version() method will be called with the requested
239 <title>Name, Description and Date</title>
240 <synopsis>char *name;
242 char *date;</synopsis>
244 The driver name is printed to the kernel log at initialization time,
245 used for IRQ registration and passed to userspace through
249 The driver description is a purely informative string passed to
250 userspace through the DRM_IOCTL_VERSION ioctl and otherwise unused by
254 The driver date, formatted as YYYYMMDD, is meant to identify the date of
255 the latest modification to the driver. However, as most drivers fail to
256 update it, its value is mostly useless. The DRM core prints it to the
257 kernel log at initialization time and passes it to userspace through the
258 DRM_IOCTL_VERSION ioctl.
263 <title>Driver Load</title>
265 The <methodname>load</methodname> method is the driver and device
266 initialization entry point. The method is responsible for allocating and
267 initializing driver private data, specifying supported performance
268 counters, performing resource allocation and mapping (e.g. acquiring
269 clocks, mapping registers or allocating command buffers), initializing
270 the memory manager (<xref linkend="drm-memory-management"/>), installing
271 the IRQ handler (<xref linkend="drm-irq-registration"/>), setting up
272 vertical blanking handling (<xref linkend="drm-vertical-blank"/>), mode
273 setting (<xref linkend="drm-mode-setting"/>) and initial output
274 configuration (<xref linkend="drm-kms-init"/>).
277 If compatibility is a concern (e.g. with drivers converted over from
278 User Mode Setting to Kernel Mode Setting), care must be taken to prevent
279 device initialization and control that is incompatible with currently
280 active userspace drivers. For instance, if user level mode setting
281 drivers are in use, it would be problematic to perform output discovery
282 & configuration at load time. Likewise, if user-level drivers
283 unaware of memory management are in use, memory management and command
284 buffer setup may need to be omitted. These requirements are
285 driver-specific, and care needs to be taken to keep both old and new
286 applications and libraries working.
288 <synopsis>int (*load) (struct drm_device *, unsigned long flags);</synopsis>
290 The method takes two arguments, a pointer to the newly created
291 <structname>drm_device</structname> and flags. The flags are used to
292 pass the <structfield>driver_data</structfield> field of the device id
293 corresponding to the device passed to <function>drm_*_init()</function>.
294 Only PCI devices currently use this, USB and platform DRM drivers have
295 their <methodname>load</methodname> method called with flags to 0.
298 <title>Driver Private & Performance Counters</title>
300 The driver private hangs off the main
301 <structname>drm_device</structname> structure and can be used for
302 tracking various device-specific bits of information, like register
303 offsets, command buffer status, register state for suspend/resume, etc.
304 At load time, a driver may simply allocate one and set
305 <structname>drm_device</structname>.<structfield>dev_priv</structfield>
306 appropriately; it should be freed and
307 <structname>drm_device</structname>.<structfield>dev_priv</structfield>
308 set to NULL when the driver is unloaded.
311 DRM supports several counters which were used for rough performance
312 characterization. This stat counter system is deprecated and should not
313 be used. If performance monitoring is desired, the developer should
314 investigate and potentially enhance the kernel perf and tracing
315 infrastructure to export GPU related performance information for
316 consumption by performance monitoring tools and applications.
319 <sect3 id="drm-irq-registration">
320 <title>IRQ Registration</title>
322 The DRM core tries to facilitate IRQ handler registration and
323 unregistration by providing <function>drm_irq_install</function> and
324 <function>drm_irq_uninstall</function> functions. Those functions only
325 support a single interrupt per device, devices that use more than one
326 IRQs need to be handled manually.
329 <title>Managed IRQ Registration</title>
331 Both the <function>drm_irq_install</function> and
332 <function>drm_irq_uninstall</function> functions get the device IRQ by
333 calling <function>drm_dev_to_irq</function>. This inline function will
334 call a bus-specific operation to retrieve the IRQ number. For platform
335 devices, <function>platform_get_irq</function>(..., 0) is used to
336 retrieve the IRQ number.
339 <function>drm_irq_install</function> starts by calling the
340 <methodname>irq_preinstall</methodname> driver operation. The operation
341 is optional and must make sure that the interrupt will not get fired by
342 clearing all pending interrupt flags or disabling the interrupt.
345 The IRQ will then be requested by a call to
346 <function>request_irq</function>. If the DRIVER_IRQ_SHARED driver
347 feature flag is set, a shared (IRQF_SHARED) IRQ handler will be
351 The IRQ handler function must be provided as the mandatory irq_handler
352 driver operation. It will get passed directly to
353 <function>request_irq</function> and thus has the same prototype as all
354 IRQ handlers. It will get called with a pointer to the DRM device as the
358 Finally the function calls the optional
359 <methodname>irq_postinstall</methodname> driver operation. The operation
360 usually enables interrupts (excluding the vblank interrupt, which is
361 enabled separately), but drivers may choose to enable/disable interrupts
365 <function>drm_irq_uninstall</function> is similarly used to uninstall an
366 IRQ handler. It starts by waking up all processes waiting on a vblank
367 interrupt to make sure they don't hang, and then calls the optional
368 <methodname>irq_uninstall</methodname> driver operation. The operation
369 must disable all hardware interrupts. Finally the function frees the IRQ
370 by calling <function>free_irq</function>.
374 <title>Manual IRQ Registration</title>
376 Drivers that require multiple interrupt handlers can't use the managed
377 IRQ registration functions. In that case IRQs must be registered and
378 unregistered manually (usually with the <function>request_irq</function>
379 and <function>free_irq</function> functions, or their devm_* equivalent).
382 When manually registering IRQs, drivers must not set the DRIVER_HAVE_IRQ
383 driver feature flag, and must not provide the
384 <methodname>irq_handler</methodname> driver operation. They must set the
385 <structname>drm_device</structname> <structfield>irq_enabled</structfield>
386 field to 1 upon registration of the IRQs, and clear it to 0 after
387 unregistering the IRQs.
392 <title>Memory Manager Initialization</title>
394 Every DRM driver requires a memory manager which must be initialized at
395 load time. DRM currently contains two memory managers, the Translation
396 Table Manager (TTM) and the Graphics Execution Manager (GEM).
397 This document describes the use of the GEM memory manager only. See
398 <xref linkend="drm-memory-management"/> for details.
402 <title>Miscellaneous Device Configuration</title>
404 Another task that may be necessary for PCI devices during configuration
405 is mapping the video BIOS. On many devices, the VBIOS describes device
406 configuration, LCD panel timings (if any), and contains flags indicating
407 device state. Mapping the BIOS can be done using the pci_map_rom() call,
408 a convenience function that takes care of mapping the actual ROM,
409 whether it has been shadowed into memory (typically at address 0xc0000)
410 or exists on the PCI device in the ROM BAR. Note that after the ROM has
411 been mapped and any necessary information has been extracted, it should
412 be unmapped; on many devices, the ROM address decoder is shared with
413 other BARs, so leaving it mapped could cause undesired behaviour like
414 hangs or memory corruption.
415 <!--!Fdrivers/pci/rom.c pci_map_rom-->
421 <!-- Internals: memory management -->
423 <sect1 id="drm-memory-management">
424 <title>Memory management</title>
426 Modern Linux systems require large amount of graphics memory to store
427 frame buffers, textures, vertices and other graphics-related data. Given
428 the very dynamic nature of many of that data, managing graphics memory
429 efficiently is thus crucial for the graphics stack and plays a central
430 role in the DRM infrastructure.
433 The DRM core includes two memory managers, namely Translation Table Maps
434 (TTM) and Graphics Execution Manager (GEM). TTM was the first DRM memory
435 manager to be developed and tried to be a one-size-fits-them all
436 solution. It provides a single userspace API to accommodate the need of
437 all hardware, supporting both Unified Memory Architecture (UMA) devices
438 and devices with dedicated video RAM (i.e. most discrete video cards).
439 This resulted in a large, complex piece of code that turned out to be
440 hard to use for driver development.
443 GEM started as an Intel-sponsored project in reaction to TTM's
444 complexity. Its design philosophy is completely different: instead of
445 providing a solution to every graphics memory-related problems, GEM
446 identified common code between drivers and created a support library to
447 share it. GEM has simpler initialization and execution requirements than
448 TTM, but has no video RAM management capabitilies and is thus limited to
452 <title>The Translation Table Manager (TTM)</title>
454 TTM design background and information belongs here.
457 <title>TTM initialization</title>
458 <warning><para>This section is outdated.</para></warning>
460 Drivers wishing to support TTM must fill out a drm_bo_driver
461 structure. The structure contains several fields with function
462 pointers for initializing the TTM, allocating and freeing memory,
463 waiting for command completion and fence synchronization, and memory
464 migration. See the radeon_ttm.c file for an example of usage.
467 The ttm_global_reference structure is made up of several fields:
470 struct ttm_global_reference {
471 enum ttm_global_types global_type;
474 int (*init) (struct ttm_global_reference *);
475 void (*release) (struct ttm_global_reference *);
479 There should be one global reference structure for your memory
480 manager as a whole, and there will be others for each object
481 created by the memory manager at runtime. Your global TTM should
482 have a type of TTM_GLOBAL_TTM_MEM. The size field for the global
483 object should be sizeof(struct ttm_mem_global), and the init and
484 release hooks should point at your driver-specific init and
485 release routines, which probably eventually call
486 ttm_mem_global_init and ttm_mem_global_release, respectively.
489 Once your global TTM accounting structure is set up and initialized
490 by calling ttm_global_item_ref() on it,
491 you need to create a buffer object TTM to
492 provide a pool for buffer object allocation by clients and the
493 kernel itself. The type of this object should be TTM_GLOBAL_TTM_BO,
494 and its size should be sizeof(struct ttm_bo_global). Again,
495 driver-specific init and release functions may be provided,
496 likely eventually calling ttm_bo_global_init() and
497 ttm_bo_global_release(), respectively. Also, like the previous
498 object, ttm_global_item_ref() is used to create an initial reference
499 count for the TTM, which will call your initialization function.
504 <title>The Graphics Execution Manager (GEM)</title>
506 The GEM design approach has resulted in a memory manager that doesn't
507 provide full coverage of all (or even all common) use cases in its
508 userspace or kernel API. GEM exposes a set of standard memory-related
509 operations to userspace and a set of helper functions to drivers, and let
510 drivers implement hardware-specific operations with their own private API.
513 The GEM userspace API is described in the
514 <ulink url="http://lwn.net/Articles/283798/"><citetitle>GEM - the Graphics
515 Execution Manager</citetitle></ulink> article on LWN. While slightly
516 outdated, the document provides a good overview of the GEM API principles.
517 Buffer allocation and read and write operations, described as part of the
518 common GEM API, are currently implemented using driver-specific ioctls.
521 GEM is data-agnostic. It manages abstract buffer objects without knowing
522 what individual buffers contain. APIs that require knowledge of buffer
523 contents or purpose, such as buffer allocation or synchronization
524 primitives, are thus outside of the scope of GEM and must be implemented
525 using driver-specific ioctls.
528 On a fundamental level, GEM involves several operations:
530 <listitem>Memory allocation and freeing</listitem>
531 <listitem>Command execution</listitem>
532 <listitem>Aperture management at command execution time</listitem>
534 Buffer object allocation is relatively straightforward and largely
535 provided by Linux's shmem layer, which provides memory to back each
539 Device-specific operations, such as command execution, pinning, buffer
540 read & write, mapping, and domain ownership transfers are left to
541 driver-specific ioctls.
544 <title>GEM Initialization</title>
546 Drivers that use GEM must set the DRIVER_GEM bit in the struct
547 <structname>drm_driver</structname>
548 <structfield>driver_features</structfield> field. The DRM core will
549 then automatically initialize the GEM core before calling the
550 <methodname>load</methodname> operation. Behind the scene, this will
551 create a DRM Memory Manager object which provides an address space
552 pool for object allocation.
555 In a KMS configuration, drivers need to allocate and initialize a
556 command ring buffer following core GEM initialization if required by
557 the hardware. UMA devices usually have what is called a "stolen"
558 memory region, which provides space for the initial framebuffer and
559 large, contiguous memory regions required by the device. This space is
560 typically not managed by GEM, and must be initialized separately into
561 its own DRM MM object.
565 <title>GEM Objects Creation</title>
567 GEM splits creation of GEM objects and allocation of the memory that
568 backs them in two distinct operations.
571 GEM objects are represented by an instance of struct
572 <structname>drm_gem_object</structname>. Drivers usually need to extend
573 GEM objects with private information and thus create a driver-specific
574 GEM object structure type that embeds an instance of struct
575 <structname>drm_gem_object</structname>.
578 To create a GEM object, a driver allocates memory for an instance of its
579 specific GEM object type and initializes the embedded struct
580 <structname>drm_gem_object</structname> with a call to
581 <function>drm_gem_object_init</function>. The function takes a pointer to
582 the DRM device, a pointer to the GEM object and the buffer object size
586 GEM uses shmem to allocate anonymous pageable memory.
587 <function>drm_gem_object_init</function> will create an shmfs file of
588 the requested size and store it into the struct
589 <structname>drm_gem_object</structname> <structfield>filp</structfield>
590 field. The memory is used as either main storage for the object when the
591 graphics hardware uses system memory directly or as a backing store
595 Drivers are responsible for the actual physical pages allocation by
596 calling <function>shmem_read_mapping_page_gfp</function> for each page.
597 Note that they can decide to allocate pages when initializing the GEM
598 object, or to delay allocation until the memory is needed (for instance
599 when a page fault occurs as a result of a userspace memory access or
600 when the driver needs to start a DMA transfer involving the memory).
603 Anonymous pageable memory allocation is not always desired, for instance
604 when the hardware requires physically contiguous system memory as is
605 often the case in embedded devices. Drivers can create GEM objects with
606 no shmfs backing (called private GEM objects) by initializing them with
607 a call to <function>drm_gem_private_object_init</function> instead of
608 <function>drm_gem_object_init</function>. Storage for private GEM
609 objects must be managed by drivers.
612 Drivers that do not need to extend GEM objects with private information
613 can call the <function>drm_gem_object_alloc</function> function to
614 allocate and initialize a struct <structname>drm_gem_object</structname>
615 instance. The GEM core will call the optional driver
616 <methodname>gem_init_object</methodname> operation after initializing
617 the GEM object with <function>drm_gem_object_init</function>.
618 <synopsis>int (*gem_init_object) (struct drm_gem_object *obj);</synopsis>
621 No alloc-and-init function exists for private GEM objects.
625 <title>GEM Objects Lifetime</title>
627 All GEM objects are reference-counted by the GEM core. References can be
628 acquired and release by <function>calling drm_gem_object_reference</function>
629 and <function>drm_gem_object_unreference</function> respectively. The
630 caller must hold the <structname>drm_device</structname>
631 <structfield>struct_mutex</structfield> lock. As a convenience, GEM
632 provides the <function>drm_gem_object_reference_unlocked</function> and
633 <function>drm_gem_object_unreference_unlocked</function> functions that
634 can be called without holding the lock.
637 When the last reference to a GEM object is released the GEM core calls
638 the <structname>drm_driver</structname>
639 <methodname>gem_free_object</methodname> operation. That operation is
640 mandatory for GEM-enabled drivers and must free the GEM object and all
641 associated resources.
644 <synopsis>void (*gem_free_object) (struct drm_gem_object *obj);</synopsis>
645 Drivers are responsible for freeing all GEM object resources, including
646 the resources created by the GEM core. If an mmap offset has been
647 created for the object (in which case
648 <structname>drm_gem_object</structname>::<structfield>map_list</structfield>::<structfield>map</structfield>
649 is not NULL) it must be freed by a call to
650 <function>drm_gem_free_mmap_offset</function>. The shmfs backing store
651 must be released by calling <function>drm_gem_object_release</function>
652 (that function can safely be called if no shmfs backing store has been
657 <title>GEM Objects Naming</title>
659 Communication between userspace and the kernel refers to GEM objects
660 using local handles, global names or, more recently, file descriptors.
661 All of those are 32-bit integer values; the usual Linux kernel limits
662 apply to the file descriptors.
665 GEM handles are local to a DRM file. Applications get a handle to a GEM
666 object through a driver-specific ioctl, and can use that handle to refer
667 to the GEM object in other standard or driver-specific ioctls. Closing a
668 DRM file handle frees all its GEM handles and dereferences the
669 associated GEM objects.
672 To create a handle for a GEM object drivers call
673 <function>drm_gem_handle_create</function>. The function takes a pointer
674 to the DRM file and the GEM object and returns a locally unique handle.
675 When the handle is no longer needed drivers delete it with a call to
676 <function>drm_gem_handle_delete</function>. Finally the GEM object
677 associated with a handle can be retrieved by a call to
678 <function>drm_gem_object_lookup</function>.
681 Handles don't take ownership of GEM objects, they only take a reference
682 to the object that will be dropped when the handle is destroyed. To
683 avoid leaking GEM objects, drivers must make sure they drop the
684 reference(s) they own (such as the initial reference taken at object
685 creation time) as appropriate, without any special consideration for the
686 handle. For example, in the particular case of combined GEM object and
687 handle creation in the implementation of the
688 <methodname>dumb_create</methodname> operation, drivers must drop the
689 initial reference to the GEM object before returning the handle.
692 GEM names are similar in purpose to handles but are not local to DRM
693 files. They can be passed between processes to reference a GEM object
694 globally. Names can't be used directly to refer to objects in the DRM
695 API, applications must convert handles to names and names to handles
696 using the DRM_IOCTL_GEM_FLINK and DRM_IOCTL_GEM_OPEN ioctls
697 respectively. The conversion is handled by the DRM core without any
698 driver-specific support.
701 Similar to global names, GEM file descriptors are also used to share GEM
702 objects across processes. They offer additional security: as file
703 descriptors must be explicitly sent over UNIX domain sockets to be shared
704 between applications, they can't be guessed like the globally unique GEM
708 Drivers that support GEM file descriptors, also known as the DRM PRIME
709 API, must set the DRIVER_PRIME bit in the struct
710 <structname>drm_driver</structname>
711 <structfield>driver_features</structfield> field, and implement the
712 <methodname>prime_handle_to_fd</methodname> and
713 <methodname>prime_fd_to_handle</methodname> operations.
716 <synopsis>int (*prime_handle_to_fd)(struct drm_device *dev,
717 struct drm_file *file_priv, uint32_t handle,
718 uint32_t flags, int *prime_fd);
719 int (*prime_fd_to_handle)(struct drm_device *dev,
720 struct drm_file *file_priv, int prime_fd,
721 uint32_t *handle);</synopsis>
722 Those two operations convert a handle to a PRIME file descriptor and
723 vice versa. Drivers must use the kernel dma-buf buffer sharing framework
724 to manage the PRIME file descriptors.
727 While non-GEM drivers must implement the operations themselves, GEM
728 drivers must use the <function>drm_gem_prime_handle_to_fd</function>
729 and <function>drm_gem_prime_fd_to_handle</function> helper functions.
730 Those helpers rely on the driver
731 <methodname>gem_prime_export</methodname> and
732 <methodname>gem_prime_import</methodname> operations to create a dma-buf
733 instance from a GEM object (dma-buf exporter role) and to create a GEM
734 object from a dma-buf instance (dma-buf importer role).
737 <synopsis>struct dma_buf * (*gem_prime_export)(struct drm_device *dev,
738 struct drm_gem_object *obj,
740 struct drm_gem_object * (*gem_prime_import)(struct drm_device *dev,
741 struct dma_buf *dma_buf);</synopsis>
742 These two operations are mandatory for GEM drivers that support DRM
746 <title>DRM PRIME Helper Functions Reference</title>
747 !Pdrivers/gpu/drm/drm_prime.c PRIME Helpers
750 <sect3 id="drm-gem-objects-mapping">
751 <title>GEM Objects Mapping</title>
753 Because mapping operations are fairly heavyweight GEM favours
754 read/write-like access to buffers, implemented through driver-specific
755 ioctls, over mapping buffers to userspace. However, when random access
756 to the buffer is needed (to perform software rendering for instance),
757 direct access to the object can be more efficient.
760 The mmap system call can't be used directly to map GEM objects, as they
761 don't have their own file handle. Two alternative methods currently
762 co-exist to map GEM objects to userspace. The first method uses a
763 driver-specific ioctl to perform the mapping operation, calling
764 <function>do_mmap</function> under the hood. This is often considered
765 dubious, seems to be discouraged for new GEM-enabled drivers, and will
766 thus not be described here.
769 The second method uses the mmap system call on the DRM file handle.
770 <synopsis>void *mmap(void *addr, size_t length, int prot, int flags, int fd,
771 off_t offset);</synopsis>
772 DRM identifies the GEM object to be mapped by a fake offset passed
773 through the mmap offset argument. Prior to being mapped, a GEM object
774 must thus be associated with a fake offset. To do so, drivers must call
775 <function>drm_gem_create_mmap_offset</function> on the object. The
776 function allocates a fake offset range from a pool and stores the
777 offset divided by PAGE_SIZE in
778 <literal>obj->map_list.hash.key</literal>. Care must be taken not to
779 call <function>drm_gem_create_mmap_offset</function> if a fake offset
780 has already been allocated for the object. This can be tested by
781 <literal>obj->map_list.map</literal> being non-NULL.
784 Once allocated, the fake offset value
785 (<literal>obj->map_list.hash.key << PAGE_SHIFT</literal>)
786 must be passed to the application in a driver-specific way and can then
787 be used as the mmap offset argument.
790 The GEM core provides a helper method <function>drm_gem_mmap</function>
791 to handle object mapping. The method can be set directly as the mmap
792 file operation handler. It will look up the GEM object based on the
793 offset value and set the VMA operations to the
794 <structname>drm_driver</structname> <structfield>gem_vm_ops</structfield>
795 field. Note that <function>drm_gem_mmap</function> doesn't map memory to
796 userspace, but relies on the driver-provided fault handler to map pages
800 To use <function>drm_gem_mmap</function>, drivers must fill the struct
801 <structname>drm_driver</structname> <structfield>gem_vm_ops</structfield>
802 field with a pointer to VM operations.
805 <synopsis>struct vm_operations_struct *gem_vm_ops
807 struct vm_operations_struct {
808 void (*open)(struct vm_area_struct * area);
809 void (*close)(struct vm_area_struct * area);
810 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
814 The <methodname>open</methodname> and <methodname>close</methodname>
815 operations must update the GEM object reference count. Drivers can use
816 the <function>drm_gem_vm_open</function> and
817 <function>drm_gem_vm_close</function> helper functions directly as open
821 The fault operation handler is responsible for mapping individual pages
822 to userspace when a page fault occurs. Depending on the memory
823 allocation scheme, drivers can allocate pages at fault time, or can
824 decide to allocate memory for the GEM object at the time the object is
828 Drivers that want to map the GEM object upfront instead of handling page
829 faults can implement their own mmap file operation handler.
833 <title>Dumb GEM Objects</title>
835 The GEM API doesn't standardize GEM objects creation and leaves it to
836 driver-specific ioctls. While not an issue for full-fledged graphics
837 stacks that include device-specific userspace components (in libdrm for
838 instance), this limit makes DRM-based early boot graphics unnecessarily
842 Dumb GEM objects partly alleviate the problem by providing a standard
843 API to create dumb buffers suitable for scanout, which can then be used
844 to create KMS frame buffers.
847 To support dumb GEM objects drivers must implement the
848 <methodname>dumb_create</methodname>,
849 <methodname>dumb_destroy</methodname> and
850 <methodname>dumb_map_offset</methodname> operations.
854 <synopsis>int (*dumb_create)(struct drm_file *file_priv, struct drm_device *dev,
855 struct drm_mode_create_dumb *args);</synopsis>
857 The <methodname>dumb_create</methodname> operation creates a GEM
858 object suitable for scanout based on the width, height and depth
859 from the struct <structname>drm_mode_create_dumb</structname>
860 argument. It fills the argument's <structfield>handle</structfield>,
861 <structfield>pitch</structfield> and <structfield>size</structfield>
862 fields with a handle for the newly created GEM object and its line
863 pitch and size in bytes.
867 <synopsis>int (*dumb_destroy)(struct drm_file *file_priv, struct drm_device *dev,
868 uint32_t handle);</synopsis>
870 The <methodname>dumb_destroy</methodname> operation destroys a dumb
871 GEM object created by <methodname>dumb_create</methodname>.
875 <synopsis>int (*dumb_map_offset)(struct drm_file *file_priv, struct drm_device *dev,
876 uint32_t handle, uint64_t *offset);</synopsis>
878 The <methodname>dumb_map_offset</methodname> operation associates an
879 mmap fake offset with the GEM object given by the handle and returns
880 it. Drivers must use the
881 <function>drm_gem_create_mmap_offset</function> function to
882 associate the fake offset as described in
883 <xref linkend="drm-gem-objects-mapping"/>.
889 <title>Memory Coherency</title>
891 When mapped to the device or used in a command buffer, backing pages
892 for an object are flushed to memory and marked write combined so as to
893 be coherent with the GPU. Likewise, if the CPU accesses an object
894 after the GPU has finished rendering to the object, then the object
895 must be made coherent with the CPU's view of memory, usually involving
896 GPU cache flushing of various kinds. This core CPU<->GPU
897 coherency management is provided by a device-specific ioctl, which
898 evaluates an object's current domain and performs any necessary
899 flushing or synchronization to put the object into the desired
900 coherency domain (note that the object may be busy, i.e. an active
901 render target; in that case, setting the domain blocks the client and
902 waits for rendering to complete before performing any necessary
903 flushing operations).
907 <title>Command Execution</title>
909 Perhaps the most important GEM function for GPU devices is providing a
910 command execution interface to clients. Client programs construct
911 command buffers containing references to previously allocated memory
912 objects, and then submit them to GEM. At that point, GEM takes care to
913 bind all the objects into the GTT, execute the buffer, and provide
914 necessary synchronization between clients accessing the same buffers.
915 This often involves evicting some objects from the GTT and re-binding
916 others (a fairly expensive operation), and providing relocation
917 support which hides fixed GTT offsets from clients. Clients must take
918 care not to submit command buffers that reference more objects than
919 can fit in the GTT; otherwise, GEM will reject them and no rendering
920 will occur. Similarly, if several objects in the buffer require fence
921 registers to be allocated for correct rendering (e.g. 2D blits on
922 pre-965 chips), care must be taken not to require more fence registers
923 than are available to the client. Such resource management should be
924 abstracted from the client in libdrm.
930 <!-- Internals: mode setting -->
932 <sect1 id="drm-mode-setting">
933 <title>Mode Setting</title>
935 Drivers must initialize the mode setting core by calling
936 <function>drm_mode_config_init</function> on the DRM device. The function
937 initializes the <structname>drm_device</structname>
938 <structfield>mode_config</structfield> field and never fails. Once done,
939 mode configuration must be setup by initializing the following fields.
943 <synopsis>int min_width, min_height;
944 int max_width, max_height;</synopsis>
946 Minimum and maximum width and height of the frame buffers in pixel
951 <synopsis>struct drm_mode_config_funcs *funcs;</synopsis>
952 <para>Mode setting functions.</para>
956 <title>Frame Buffer Creation</title>
957 <synopsis>struct drm_framebuffer *(*fb_create)(struct drm_device *dev,
958 struct drm_file *file_priv,
959 struct drm_mode_fb_cmd2 *mode_cmd);</synopsis>
961 Frame buffers are abstract memory objects that provide a source of
962 pixels to scanout to a CRTC. Applications explicitly request the
963 creation of frame buffers through the DRM_IOCTL_MODE_ADDFB(2) ioctls and
964 receive an opaque handle that can be passed to the KMS CRTC control,
965 plane configuration and page flip functions.
968 Frame buffers rely on the underneath memory manager for low-level memory
969 operations. When creating a frame buffer applications pass a memory
970 handle (or a list of memory handles for multi-planar formats) through
971 the <parameter>drm_mode_fb_cmd2</parameter> argument. This document
972 assumes that the driver uses GEM, those handles thus reference GEM
976 Drivers must first validate the requested frame buffer parameters passed
977 through the mode_cmd argument. In particular this is where invalid
978 sizes, pixel formats or pitches can be caught.
981 If the parameters are deemed valid, drivers then create, initialize and
982 return an instance of struct <structname>drm_framebuffer</structname>.
983 If desired the instance can be embedded in a larger driver-specific
984 structure. Drivers must fill its <structfield>width</structfield>,
985 <structfield>height</structfield>, <structfield>pitches</structfield>,
986 <structfield>offsets</structfield>, <structfield>depth</structfield>,
987 <structfield>bits_per_pixel</structfield> and
988 <structfield>pixel_format</structfield> fields from the values passed
989 through the <parameter>drm_mode_fb_cmd2</parameter> argument. They
990 should call the <function>drm_helper_mode_fill_fb_struct</function>
991 helper function to do so.
995 The initailization of the new framebuffer instance is finalized with a
996 call to <function>drm_framebuffer_init</function> which takes a pointer
997 to DRM frame buffer operations (struct
998 <structname>drm_framebuffer_funcs</structname>). Note that this function
999 publishes the framebuffer and so from this point on it can be accessed
1000 concurrently from other threads. Hence it must be the last step in the
1001 driver's framebuffer initialization sequence. Frame buffer operations
1005 <synopsis>int (*create_handle)(struct drm_framebuffer *fb,
1006 struct drm_file *file_priv, unsigned int *handle);</synopsis>
1008 Create a handle to the frame buffer underlying memory object. If
1009 the frame buffer uses a multi-plane format, the handle will
1010 reference the memory object associated with the first plane.
1013 Drivers call <function>drm_gem_handle_create</function> to create
1018 <synopsis>void (*destroy)(struct drm_framebuffer *framebuffer);</synopsis>
1020 Destroy the frame buffer object and frees all associated
1021 resources. Drivers must call
1022 <function>drm_framebuffer_cleanup</function> to free resources
1023 allocated by the DRM core for the frame buffer object, and must
1024 make sure to unreference all memory objects associated with the
1025 frame buffer. Handles created by the
1026 <methodname>create_handle</methodname> operation are released by
1031 <synopsis>int (*dirty)(struct drm_framebuffer *framebuffer,
1032 struct drm_file *file_priv, unsigned flags, unsigned color,
1033 struct drm_clip_rect *clips, unsigned num_clips);</synopsis>
1035 This optional operation notifies the driver that a region of the
1036 frame buffer has changed in response to a DRM_IOCTL_MODE_DIRTYFB
1043 The lifetime of a drm framebuffer is controlled with a reference count,
1044 drivers can grab additional references with
1045 <function>drm_framebuffer_reference</function> </para> and drop them
1046 again with <function>drm_framebuffer_unreference</function>. For
1047 driver-private framebuffers for which the last reference is never
1048 dropped (e.g. for the fbdev framebuffer when the struct
1049 <structname>drm_framebuffer</structname> is embedded into the fbdev
1050 helper struct) drivers can manually clean up a framebuffer at module
1052 <function>drm_framebuffer_unregister_private</function>.
1055 <title>Output Polling</title>
1056 <synopsis>void (*output_poll_changed)(struct drm_device *dev);</synopsis>
1058 This operation notifies the driver that the status of one or more
1059 connectors has changed. Drivers that use the fb helper can just call the
1060 <function>drm_fb_helper_hotplug_event</function> function to handle this
1065 <title>Locking</title>
1067 Beside some lookup structures with their own locking (which is hidden
1068 behind the interface functions) most of the modeset state is protected
1069 by the <code>dev-<mode_config.lock</code> mutex and additionally
1070 per-crtc locks to allow cursor updates, pageflips and similar operations
1071 to occur concurrently with background tasks like output detection.
1072 Operations which cross domains like a full modeset always grab all
1073 locks. Drivers there need to protect resources shared between crtcs with
1074 additional locking. They also need to be careful to always grab the
1075 relevant crtc locks if a modset functions touches crtc state, e.g. for
1076 load detection (which does only grab the <code>mode_config.lock</code>
1077 to allow concurrent screen updates on live crtcs).
1082 <!-- Internals: kms initialization and cleanup -->
1084 <sect1 id="drm-kms-init">
1085 <title>KMS Initialization and Cleanup</title>
1087 A KMS device is abstracted and exposed as a set of planes, CRTCs, encoders
1088 and connectors. KMS drivers must thus create and initialize all those
1089 objects at load time after initializing mode setting.
1092 <title>CRTCs (struct <structname>drm_crtc</structname>)</title>
1094 A CRTC is an abstraction representing a part of the chip that contains a
1095 pointer to a scanout buffer. Therefore, the number of CRTCs available
1096 determines how many independent scanout buffers can be active at any
1097 given time. The CRTC structure contains several fields to support this:
1098 a pointer to some video memory (abstracted as a frame buffer object), a
1099 display mode, and an (x, y) offset into the video memory to support
1100 panning or configurations where one piece of video memory spans multiple
1104 <title>CRTC Initialization</title>
1106 A KMS device must create and register at least one struct
1107 <structname>drm_crtc</structname> instance. The instance is allocated
1108 and zeroed by the driver, possibly as part of a larger structure, and
1109 registered with a call to <function>drm_crtc_init</function> with a
1110 pointer to CRTC functions.
1114 <title>CRTC Operations</title>
1116 <title>Set Configuration</title>
1117 <synopsis>int (*set_config)(struct drm_mode_set *set);</synopsis>
1119 Apply a new CRTC configuration to the device. The configuration
1120 specifies a CRTC, a frame buffer to scan out from, a (x,y) position in
1121 the frame buffer, a display mode and an array of connectors to drive
1122 with the CRTC if possible.
1125 If the frame buffer specified in the configuration is NULL, the driver
1126 must detach all encoders connected to the CRTC and all connectors
1127 attached to those encoders and disable them.
1130 This operation is called with the mode config lock held.
1133 FIXME: How should set_config interact with DPMS? If the CRTC is
1134 suspended, should it be resumed?
1138 <title>Page Flipping</title>
1139 <synopsis>int (*page_flip)(struct drm_crtc *crtc, struct drm_framebuffer *fb,
1140 struct drm_pending_vblank_event *event);</synopsis>
1142 Schedule a page flip to the given frame buffer for the CRTC. This
1143 operation is called with the mode config mutex held.
1146 Page flipping is a synchronization mechanism that replaces the frame
1147 buffer being scanned out by the CRTC with a new frame buffer during
1148 vertical blanking, avoiding tearing. When an application requests a page
1149 flip the DRM core verifies that the new frame buffer is large enough to
1150 be scanned out by the CRTC in the currently configured mode and then
1151 calls the CRTC <methodname>page_flip</methodname> operation with a
1152 pointer to the new frame buffer.
1155 The <methodname>page_flip</methodname> operation schedules a page flip.
1156 Once any pending rendering targeting the new frame buffer has
1157 completed, the CRTC will be reprogrammed to display that frame buffer
1158 after the next vertical refresh. The operation must return immediately
1159 without waiting for rendering or page flip to complete and must block
1160 any new rendering to the frame buffer until the page flip completes.
1163 If a page flip can be successfully scheduled the driver must set the
1164 <code>drm_crtc-<fb</code> field to the new framebuffer pointed to
1165 by <code>fb</code>. This is important so that the reference counting
1166 on framebuffers stays balanced.
1169 If a page flip is already pending, the
1170 <methodname>page_flip</methodname> operation must return
1171 -<errorname>EBUSY</errorname>.
1174 To synchronize page flip to vertical blanking the driver will likely
1175 need to enable vertical blanking interrupts. It should call
1176 <function>drm_vblank_get</function> for that purpose, and call
1177 <function>drm_vblank_put</function> after the page flip completes.
1180 If the application has requested to be notified when page flip completes
1181 the <methodname>page_flip</methodname> operation will be called with a
1182 non-NULL <parameter>event</parameter> argument pointing to a
1183 <structname>drm_pending_vblank_event</structname> instance. Upon page
1184 flip completion the driver must call <methodname>drm_send_vblank_event</methodname>
1185 to fill in the event and send to wake up any waiting processes.
1186 This can be performed with
1187 <programlisting><![CDATA[
1188 spin_lock_irqsave(&dev->event_lock, flags);
1190 drm_send_vblank_event(dev, pipe, event);
1191 spin_unlock_irqrestore(&dev->event_lock, flags);
1192 ]]></programlisting>
1195 FIXME: Could drivers that don't need to wait for rendering to complete
1196 just add the event to <literal>dev->vblank_event_list</literal> and
1197 let the DRM core handle everything, as for "normal" vertical blanking
1201 While waiting for the page flip to complete, the
1202 <literal>event->base.link</literal> list head can be used freely by
1203 the driver to store the pending event in a driver-specific list.
1206 If the file handle is closed before the event is signaled, drivers must
1207 take care to destroy the event in their
1208 <methodname>preclose</methodname> operation (and, if needed, call
1209 <function>drm_vblank_put</function>).
1213 <title>Miscellaneous</title>
1216 <synopsis>void (*set_property)(struct drm_crtc *crtc,
1217 struct drm_property *property, uint64_t value);</synopsis>
1219 Set the value of the given CRTC property to
1220 <parameter>value</parameter>. See <xref linkend="drm-kms-properties"/>
1221 for more information about properties.
1225 <synopsis>void (*gamma_set)(struct drm_crtc *crtc, u16 *r, u16 *g, u16 *b,
1226 uint32_t start, uint32_t size);</synopsis>
1228 Apply a gamma table to the device. The operation is optional.
1232 <synopsis>void (*destroy)(struct drm_crtc *crtc);</synopsis>
1234 Destroy the CRTC when not needed anymore. See
1235 <xref linkend="drm-kms-init"/>.
1243 <title>Planes (struct <structname>drm_plane</structname>)</title>
1245 A plane represents an image source that can be blended with or overlayed
1246 on top of a CRTC during the scanout process. Planes are associated with
1247 a frame buffer to crop a portion of the image memory (source) and
1248 optionally scale it to a destination size. The result is then blended
1249 with or overlayed on top of a CRTC.
1252 <title>Plane Initialization</title>
1254 Planes are optional. To create a plane, a KMS drivers allocates and
1255 zeroes an instances of struct <structname>drm_plane</structname>
1256 (possibly as part of a larger structure) and registers it with a call
1257 to <function>drm_plane_init</function>. The function takes a bitmask
1258 of the CRTCs that can be associated with the plane, a pointer to the
1259 plane functions and a list of format supported formats.
1263 <title>Plane Operations</title>
1266 <synopsis>int (*update_plane)(struct drm_plane *plane, struct drm_crtc *crtc,
1267 struct drm_framebuffer *fb, int crtc_x, int crtc_y,
1268 unsigned int crtc_w, unsigned int crtc_h,
1269 uint32_t src_x, uint32_t src_y,
1270 uint32_t src_w, uint32_t src_h);</synopsis>
1272 Enable and configure the plane to use the given CRTC and frame buffer.
1275 The source rectangle in frame buffer memory coordinates is given by
1276 the <parameter>src_x</parameter>, <parameter>src_y</parameter>,
1277 <parameter>src_w</parameter> and <parameter>src_h</parameter>
1278 parameters (as 16.16 fixed point values). Devices that don't support
1279 subpixel plane coordinates can ignore the fractional part.
1282 The destination rectangle in CRTC coordinates is given by the
1283 <parameter>crtc_x</parameter>, <parameter>crtc_y</parameter>,
1284 <parameter>crtc_w</parameter> and <parameter>crtc_h</parameter>
1285 parameters (as integer values). Devices scale the source rectangle to
1286 the destination rectangle. If scaling is not supported, and the source
1287 rectangle size doesn't match the destination rectangle size, the
1288 driver must return a -<errorname>EINVAL</errorname> error.
1292 <synopsis>int (*disable_plane)(struct drm_plane *plane);</synopsis>
1294 Disable the plane. The DRM core calls this method in response to a
1295 DRM_IOCTL_MODE_SETPLANE ioctl call with the frame buffer ID set to 0.
1296 Disabled planes must not be processed by the CRTC.
1300 <synopsis>void (*destroy)(struct drm_plane *plane);</synopsis>
1302 Destroy the plane when not needed anymore. See
1303 <xref linkend="drm-kms-init"/>.
1310 <title>Encoders (struct <structname>drm_encoder</structname>)</title>
1312 An encoder takes pixel data from a CRTC and converts it to a format
1313 suitable for any attached connectors. On some devices, it may be
1314 possible to have a CRTC send data to more than one encoder. In that
1315 case, both encoders would receive data from the same scanout buffer,
1316 resulting in a "cloned" display configuration across the connectors
1317 attached to each encoder.
1320 <title>Encoder Initialization</title>
1322 As for CRTCs, a KMS driver must create, initialize and register at
1323 least one struct <structname>drm_encoder</structname> instance. The
1324 instance is allocated and zeroed by the driver, possibly as part of a
1328 Drivers must initialize the struct <structname>drm_encoder</structname>
1329 <structfield>possible_crtcs</structfield> and
1330 <structfield>possible_clones</structfield> fields before registering the
1331 encoder. Both fields are bitmasks of respectively the CRTCs that the
1332 encoder can be connected to, and sibling encoders candidate for cloning.
1335 After being initialized, the encoder must be registered with a call to
1336 <function>drm_encoder_init</function>. The function takes a pointer to
1337 the encoder functions and an encoder type. Supported types are
1340 DRM_MODE_ENCODER_DAC for VGA and analog on DVI-I/DVI-A
1343 DRM_MODE_ENCODER_TMDS for DVI, HDMI and (embedded) DisplayPort
1346 DRM_MODE_ENCODER_LVDS for display panels
1349 DRM_MODE_ENCODER_TVDAC for TV output (Composite, S-Video, Component,
1353 DRM_MODE_ENCODER_VIRTUAL for virtual machine displays
1358 Encoders must be attached to a CRTC to be used. DRM drivers leave
1359 encoders unattached at initialization time. Applications (or the fbdev
1360 compatibility layer when implemented) are responsible for attaching the
1361 encoders they want to use to a CRTC.
1365 <title>Encoder Operations</title>
1368 <synopsis>void (*destroy)(struct drm_encoder *encoder);</synopsis>
1370 Called to destroy the encoder when not needed anymore. See
1371 <xref linkend="drm-kms-init"/>.
1375 <synopsis>void (*set_property)(struct drm_plane *plane,
1376 struct drm_property *property, uint64_t value);</synopsis>
1378 Set the value of the given plane property to
1379 <parameter>value</parameter>. See <xref linkend="drm-kms-properties"/>
1380 for more information about properties.
1387 <title>Connectors (struct <structname>drm_connector</structname>)</title>
1389 A connector is the final destination for pixel data on a device, and
1390 usually connects directly to an external display device like a monitor
1391 or laptop panel. A connector can only be attached to one encoder at a
1392 time. The connector is also the structure where information about the
1393 attached display is kept, so it contains fields for display data, EDID
1394 data, DPMS & connection status, and information about modes
1395 supported on the attached displays.
1398 <title>Connector Initialization</title>
1400 Finally a KMS driver must create, initialize, register and attach at
1401 least one struct <structname>drm_connector</structname> instance. The
1402 instance is created as other KMS objects and initialized by setting the
1407 <term><structfield>interlace_allowed</structfield></term>
1409 Whether the connector can handle interlaced modes.
1413 <term><structfield>doublescan_allowed</structfield></term>
1415 Whether the connector can handle doublescan.
1419 <term><structfield>display_info
1420 </structfield></term>
1422 Display information is filled from EDID information when a display
1423 is detected. For non hot-pluggable displays such as flat panels in
1424 embedded systems, the driver should initialize the
1425 <structfield>display_info</structfield>.<structfield>width_mm</structfield>
1427 <structfield>display_info</structfield>.<structfield>height_mm</structfield>
1428 fields with the physical size of the display.
1432 <term id="drm-kms-connector-polled"><structfield>polled</structfield></term>
1434 Connector polling mode, a combination of
1437 <term>DRM_CONNECTOR_POLL_HPD</term>
1439 The connector generates hotplug events and doesn't need to be
1440 periodically polled. The CONNECT and DISCONNECT flags must not
1441 be set together with the HPD flag.
1445 <term>DRM_CONNECTOR_POLL_CONNECT</term>
1447 Periodically poll the connector for connection.
1451 <term>DRM_CONNECTOR_POLL_DISCONNECT</term>
1453 Periodically poll the connector for disconnection.
1457 Set to 0 for connectors that don't support connection status
1463 The connector is then registered with a call to
1464 <function>drm_connector_init</function> with a pointer to the connector
1465 functions and a connector type, and exposed through sysfs with a call to
1466 <function>drm_connector_register</function>.
1469 Supported connector types are
1471 <listitem>DRM_MODE_CONNECTOR_VGA</listitem>
1472 <listitem>DRM_MODE_CONNECTOR_DVII</listitem>
1473 <listitem>DRM_MODE_CONNECTOR_DVID</listitem>
1474 <listitem>DRM_MODE_CONNECTOR_DVIA</listitem>
1475 <listitem>DRM_MODE_CONNECTOR_Composite</listitem>
1476 <listitem>DRM_MODE_CONNECTOR_SVIDEO</listitem>
1477 <listitem>DRM_MODE_CONNECTOR_LVDS</listitem>
1478 <listitem>DRM_MODE_CONNECTOR_Component</listitem>
1479 <listitem>DRM_MODE_CONNECTOR_9PinDIN</listitem>
1480 <listitem>DRM_MODE_CONNECTOR_DisplayPort</listitem>
1481 <listitem>DRM_MODE_CONNECTOR_HDMIA</listitem>
1482 <listitem>DRM_MODE_CONNECTOR_HDMIB</listitem>
1483 <listitem>DRM_MODE_CONNECTOR_TV</listitem>
1484 <listitem>DRM_MODE_CONNECTOR_eDP</listitem>
1485 <listitem>DRM_MODE_CONNECTOR_VIRTUAL</listitem>
1489 Connectors must be attached to an encoder to be used. For devices that
1490 map connectors to encoders 1:1, the connector should be attached at
1491 initialization time with a call to
1492 <function>drm_mode_connector_attach_encoder</function>. The driver must
1493 also set the <structname>drm_connector</structname>
1494 <structfield>encoder</structfield> field to point to the attached
1498 Finally, drivers must initialize the connectors state change detection
1499 with a call to <function>drm_kms_helper_poll_init</function>. If at
1500 least one connector is pollable but can't generate hotplug interrupts
1501 (indicated by the DRM_CONNECTOR_POLL_CONNECT and
1502 DRM_CONNECTOR_POLL_DISCONNECT connector flags), a delayed work will
1503 automatically be queued to periodically poll for changes. Connectors
1504 that can generate hotplug interrupts must be marked with the
1505 DRM_CONNECTOR_POLL_HPD flag instead, and their interrupt handler must
1506 call <function>drm_helper_hpd_irq_event</function>. The function will
1507 queue a delayed work to check the state of all connectors, but no
1508 periodic polling will be done.
1512 <title>Connector Operations</title>
1514 Unless otherwise state, all operations are mandatory.
1518 <synopsis>void (*dpms)(struct drm_connector *connector, int mode);</synopsis>
1520 The DPMS operation sets the power state of a connector. The mode
1523 <listitem><para>DRM_MODE_DPMS_ON</para></listitem>
1524 <listitem><para>DRM_MODE_DPMS_STANDBY</para></listitem>
1525 <listitem><para>DRM_MODE_DPMS_SUSPEND</para></listitem>
1526 <listitem><para>DRM_MODE_DPMS_OFF</para></listitem>
1530 In all but DPMS_ON mode the encoder to which the connector is attached
1531 should put the display in low-power mode by driving its signals
1532 appropriately. If more than one connector is attached to the encoder
1533 care should be taken not to change the power state of other displays as
1534 a side effect. Low-power mode should be propagated to the encoders and
1535 CRTCs when all related connectors are put in low-power mode.
1539 <title>Modes</title>
1540 <synopsis>int (*fill_modes)(struct drm_connector *connector, uint32_t max_width,
1541 uint32_t max_height);</synopsis>
1543 Fill the mode list with all supported modes for the connector. If the
1544 <parameter>max_width</parameter> and <parameter>max_height</parameter>
1545 arguments are non-zero, the implementation must ignore all modes wider
1546 than <parameter>max_width</parameter> or higher than
1547 <parameter>max_height</parameter>.
1550 The connector must also fill in this operation its
1551 <structfield>display_info</structfield>
1552 <structfield>width_mm</structfield> and
1553 <structfield>height_mm</structfield> fields with the connected display
1554 physical size in millimeters. The fields should be set to 0 if the value
1555 isn't known or is not applicable (for instance for projector devices).
1559 <title>Connection Status</title>
1561 The connection status is updated through polling or hotplug events when
1562 supported (see <xref linkend="drm-kms-connector-polled"/>). The status
1563 value is reported to userspace through ioctls and must not be used
1564 inside the driver, as it only gets initialized by a call to
1565 <function>drm_mode_getconnector</function> from userspace.
1567 <synopsis>enum drm_connector_status (*detect)(struct drm_connector *connector,
1568 bool force);</synopsis>
1570 Check to see if anything is attached to the connector. The
1571 <parameter>force</parameter> parameter is set to false whilst polling or
1572 to true when checking the connector due to user request.
1573 <parameter>force</parameter> can be used by the driver to avoid
1574 expensive, destructive operations during automated probing.
1577 Return connector_status_connected if something is connected to the
1578 connector, connector_status_disconnected if nothing is connected and
1579 connector_status_unknown if the connection state isn't known.
1582 Drivers should only return connector_status_connected if the connection
1583 status has really been probed as connected. Connectors that can't detect
1584 the connection status, or failed connection status probes, should return
1585 connector_status_unknown.
1589 <title>Miscellaneous</title>
1592 <synopsis>void (*set_property)(struct drm_connector *connector,
1593 struct drm_property *property, uint64_t value);</synopsis>
1595 Set the value of the given connector property to
1596 <parameter>value</parameter>. See <xref linkend="drm-kms-properties"/>
1597 for more information about properties.
1601 <synopsis>void (*destroy)(struct drm_connector *connector);</synopsis>
1603 Destroy the connector when not needed anymore. See
1604 <xref linkend="drm-kms-init"/>.
1612 <title>Cleanup</title>
1614 The DRM core manages its objects' lifetime. When an object is not needed
1615 anymore the core calls its destroy function, which must clean up and
1616 free every resource allocated for the object. Every
1617 <function>drm_*_init</function> call must be matched with a
1618 corresponding <function>drm_*_cleanup</function> call to cleanup CRTCs
1619 (<function>drm_crtc_cleanup</function>), planes
1620 (<function>drm_plane_cleanup</function>), encoders
1621 (<function>drm_encoder_cleanup</function>) and connectors
1622 (<function>drm_connector_cleanup</function>). Furthermore, connectors
1623 that have been added to sysfs must be removed by a call to
1624 <function>drm_connector_unregister</function> before calling
1625 <function>drm_connector_cleanup</function>.
1628 Connectors state change detection must be cleanup up with a call to
1629 <function>drm_kms_helper_poll_fini</function>.
1633 <title>Output discovery and initialization example</title>
1634 <programlisting><![CDATA[
1635 void intel_crt_init(struct drm_device *dev)
1637 struct drm_connector *connector;
1638 struct intel_output *intel_output;
1640 intel_output = kzalloc(sizeof(struct intel_output), GFP_KERNEL);
1644 connector = &intel_output->base;
1645 drm_connector_init(dev, &intel_output->base,
1646 &intel_crt_connector_funcs, DRM_MODE_CONNECTOR_VGA);
1648 drm_encoder_init(dev, &intel_output->enc, &intel_crt_enc_funcs,
1649 DRM_MODE_ENCODER_DAC);
1651 drm_mode_connector_attach_encoder(&intel_output->base,
1652 &intel_output->enc);
1654 /* Set up the DDC bus. */
1655 intel_output->ddc_bus = intel_i2c_create(dev, GPIOA, "CRTDDC_A");
1656 if (!intel_output->ddc_bus) {
1657 dev_printk(KERN_ERR, &dev->pdev->dev, "DDC bus registration "
1662 intel_output->type = INTEL_OUTPUT_ANALOG;
1663 connector->interlace_allowed = 0;
1664 connector->doublescan_allowed = 0;
1666 drm_encoder_helper_add(&intel_output->enc, &intel_crt_helper_funcs);
1667 drm_connector_helper_add(connector, &intel_crt_connector_helper_funcs);
1669 drm_connector_register(connector);
1670 }]]></programlisting>
1672 In the example above (taken from the i915 driver), a CRTC, connector and
1673 encoder combination is created. A device-specific i2c bus is also
1674 created for fetching EDID data and performing monitor detection. Once
1675 the process is complete, the new connector is registered with sysfs to
1676 make its properties available to applications.
1680 <title>KMS API Functions</title>
1681 !Edrivers/gpu/drm/drm_crtc.c
1685 <!-- Internals: kms helper functions -->
1688 <title>Mode Setting Helper Functions</title>
1690 The CRTC, encoder and connector functions provided by the drivers
1691 implement the DRM API. They're called by the DRM core and ioctl handlers
1692 to handle device state changes and configuration request. As implementing
1693 those functions often requires logic not specific to drivers, mid-layer
1694 helper functions are available to avoid duplicating boilerplate code.
1697 The DRM core contains one mid-layer implementation. The mid-layer provides
1698 implementations of several CRTC, encoder and connector functions (called
1699 from the top of the mid-layer) that pre-process requests and call
1700 lower-level functions provided by the driver (at the bottom of the
1701 mid-layer). For instance, the
1702 <function>drm_crtc_helper_set_config</function> function can be used to
1703 fill the struct <structname>drm_crtc_funcs</structname>
1704 <structfield>set_config</structfield> field. When called, it will split
1705 the <methodname>set_config</methodname> operation in smaller, simpler
1706 operations and call the driver to handle them.
1709 To use the mid-layer, drivers call <function>drm_crtc_helper_add</function>,
1710 <function>drm_encoder_helper_add</function> and
1711 <function>drm_connector_helper_add</function> functions to install their
1712 mid-layer bottom operations handlers, and fill the
1713 <structname>drm_crtc_funcs</structname>,
1714 <structname>drm_encoder_funcs</structname> and
1715 <structname>drm_connector_funcs</structname> structures with pointers to
1716 the mid-layer top API functions. Installing the mid-layer bottom operation
1717 handlers is best done right after registering the corresponding KMS object.
1720 The mid-layer is not split between CRTC, encoder and connector operations.
1721 To use it, a driver must provide bottom functions for all of the three KMS
1725 <title>Helper Functions</title>
1728 <synopsis>int drm_crtc_helper_set_config(struct drm_mode_set *set);</synopsis>
1730 The <function>drm_crtc_helper_set_config</function> helper function
1731 is a CRTC <methodname>set_config</methodname> implementation. It
1732 first tries to locate the best encoder for each connector by calling
1733 the connector <methodname>best_encoder</methodname> helper
1737 After locating the appropriate encoders, the helper function will
1738 call the <methodname>mode_fixup</methodname> encoder and CRTC helper
1739 operations to adjust the requested mode, or reject it completely in
1740 which case an error will be returned to the application. If the new
1741 configuration after mode adjustment is identical to the current
1742 configuration the helper function will return without performing any
1746 If the adjusted mode is identical to the current mode but changes to
1747 the frame buffer need to be applied, the
1748 <function>drm_crtc_helper_set_config</function> function will call
1749 the CRTC <methodname>mode_set_base</methodname> helper operation. If
1750 the adjusted mode differs from the current mode, or if the
1751 <methodname>mode_set_base</methodname> helper operation is not
1752 provided, the helper function performs a full mode set sequence by
1753 calling the <methodname>prepare</methodname>,
1754 <methodname>mode_set</methodname> and
1755 <methodname>commit</methodname> CRTC and encoder helper operations,
1760 <synopsis>void drm_helper_connector_dpms(struct drm_connector *connector, int mode);</synopsis>
1762 The <function>drm_helper_connector_dpms</function> helper function
1763 is a connector <methodname>dpms</methodname> implementation that
1764 tracks power state of connectors. To use the function, drivers must
1765 provide <methodname>dpms</methodname> helper operations for CRTCs
1766 and encoders to apply the DPMS state to the device.
1769 The mid-layer doesn't track the power state of CRTCs and encoders.
1770 The <methodname>dpms</methodname> helper operations can thus be
1771 called with a mode identical to the currently active mode.
1775 <synopsis>int drm_helper_probe_single_connector_modes(struct drm_connector *connector,
1776 uint32_t maxX, uint32_t maxY);</synopsis>
1778 The <function>drm_helper_probe_single_connector_modes</function> helper
1779 function is a connector <methodname>fill_modes</methodname>
1780 implementation that updates the connection status for the connector
1781 and then retrieves a list of modes by calling the connector
1782 <methodname>get_modes</methodname> helper operation.
1785 The function filters out modes larger than
1786 <parameter>max_width</parameter> and <parameter>max_height</parameter>
1787 if specified. It then calls the connector
1788 <methodname>mode_valid</methodname> helper operation for each mode in
1789 the probed list to check whether the mode is valid for the connector.
1795 <title>CRTC Helper Operations</title>
1797 <listitem id="drm-helper-crtc-mode-fixup">
1798 <synopsis>bool (*mode_fixup)(struct drm_crtc *crtc,
1799 const struct drm_display_mode *mode,
1800 struct drm_display_mode *adjusted_mode);</synopsis>
1802 Let CRTCs adjust the requested mode or reject it completely. This
1803 operation returns true if the mode is accepted (possibly after being
1804 adjusted) or false if it is rejected.
1807 The <methodname>mode_fixup</methodname> operation should reject the
1808 mode if it can't reasonably use it. The definition of "reasonable"
1809 is currently fuzzy in this context. One possible behaviour would be
1810 to set the adjusted mode to the panel timings when a fixed-mode
1811 panel is used with hardware capable of scaling. Another behaviour
1812 would be to accept any input mode and adjust it to the closest mode
1813 supported by the hardware (FIXME: This needs to be clarified).
1817 <synopsis>int (*mode_set_base)(struct drm_crtc *crtc, int x, int y,
1818 struct drm_framebuffer *old_fb)</synopsis>
1820 Move the CRTC on the current frame buffer (stored in
1821 <literal>crtc->fb</literal>) to position (x,y). Any of the frame
1822 buffer, x position or y position may have been modified.
1825 This helper operation is optional. If not provided, the
1826 <function>drm_crtc_helper_set_config</function> function will fall
1827 back to the <methodname>mode_set</methodname> helper operation.
1830 FIXME: Why are x and y passed as arguments, as they can be accessed
1831 through <literal>crtc->x</literal> and
1832 <literal>crtc->y</literal>?
1836 <synopsis>void (*prepare)(struct drm_crtc *crtc);</synopsis>
1838 Prepare the CRTC for mode setting. This operation is called after
1839 validating the requested mode. Drivers use it to perform
1840 device-specific operations required before setting the new mode.
1844 <synopsis>int (*mode_set)(struct drm_crtc *crtc, struct drm_display_mode *mode,
1845 struct drm_display_mode *adjusted_mode, int x, int y,
1846 struct drm_framebuffer *old_fb);</synopsis>
1848 Set a new mode, position and frame buffer. Depending on the device
1849 requirements, the mode can be stored internally by the driver and
1850 applied in the <methodname>commit</methodname> operation, or
1851 programmed to the hardware immediately.
1854 The <methodname>mode_set</methodname> operation returns 0 on success
1855 or a negative error code if an error occurs.
1859 <synopsis>void (*commit)(struct drm_crtc *crtc);</synopsis>
1861 Commit a mode. This operation is called after setting the new mode.
1862 Upon return the device must use the new mode and be fully
1869 <title>Encoder Helper Operations</title>
1872 <synopsis>bool (*mode_fixup)(struct drm_encoder *encoder,
1873 const struct drm_display_mode *mode,
1874 struct drm_display_mode *adjusted_mode);</synopsis>
1876 Let encoders adjust the requested mode or reject it completely. This
1877 operation returns true if the mode is accepted (possibly after being
1878 adjusted) or false if it is rejected. See the
1879 <link linkend="drm-helper-crtc-mode-fixup">mode_fixup CRTC helper
1880 operation</link> for an explanation of the allowed adjustments.
1884 <synopsis>void (*prepare)(struct drm_encoder *encoder);</synopsis>
1886 Prepare the encoder for mode setting. This operation is called after
1887 validating the requested mode. Drivers use it to perform
1888 device-specific operations required before setting the new mode.
1892 <synopsis>void (*mode_set)(struct drm_encoder *encoder,
1893 struct drm_display_mode *mode,
1894 struct drm_display_mode *adjusted_mode);</synopsis>
1896 Set a new mode. Depending on the device requirements, the mode can
1897 be stored internally by the driver and applied in the
1898 <methodname>commit</methodname> operation, or programmed to the
1899 hardware immediately.
1903 <synopsis>void (*commit)(struct drm_encoder *encoder);</synopsis>
1905 Commit a mode. This operation is called after setting the new mode.
1906 Upon return the device must use the new mode and be fully
1913 <title>Connector Helper Operations</title>
1916 <synopsis>struct drm_encoder *(*best_encoder)(struct drm_connector *connector);</synopsis>
1918 Return a pointer to the best encoder for the connecter. Device that
1919 map connectors to encoders 1:1 simply return the pointer to the
1920 associated encoder. This operation is mandatory.
1924 <synopsis>int (*get_modes)(struct drm_connector *connector);</synopsis>
1926 Fill the connector's <structfield>probed_modes</structfield> list
1927 by parsing EDID data with <function>drm_add_edid_modes</function> or
1928 calling <function>drm_mode_probed_add</function> directly for every
1929 supported mode and return the number of modes it has detected. This
1930 operation is mandatory.
1933 When adding modes manually the driver creates each mode with a call to
1934 <function>drm_mode_create</function> and must fill the following fields.
1937 <synopsis>__u32 type;</synopsis>
1939 Mode type bitmask, a combination of
1942 <term>DRM_MODE_TYPE_BUILTIN</term>
1943 <listitem><para>not used?</para></listitem>
1946 <term>DRM_MODE_TYPE_CLOCK_C</term>
1947 <listitem><para>not used?</para></listitem>
1950 <term>DRM_MODE_TYPE_CRTC_C</term>
1951 <listitem><para>not used?</para></listitem>
1955 DRM_MODE_TYPE_PREFERRED - The preferred mode for the connector
1958 <para>not used?</para>
1962 <term>DRM_MODE_TYPE_DEFAULT</term>
1963 <listitem><para>not used?</para></listitem>
1966 <term>DRM_MODE_TYPE_USERDEF</term>
1967 <listitem><para>not used?</para></listitem>
1970 <term>DRM_MODE_TYPE_DRIVER</term>
1973 The mode has been created by the driver (as opposed to
1974 to user-created modes).
1979 Drivers must set the DRM_MODE_TYPE_DRIVER bit for all modes they
1980 create, and set the DRM_MODE_TYPE_PREFERRED bit for the preferred
1985 <synopsis>__u32 clock;</synopsis>
1986 <para>Pixel clock frequency in kHz unit</para>
1989 <synopsis>__u16 hdisplay, hsync_start, hsync_end, htotal;
1990 __u16 vdisplay, vsync_start, vsync_end, vtotal;</synopsis>
1991 <para>Horizontal and vertical timing information</para>
1993 Active Front Sync Back
1995 <-----------------------><----------------><-------------><-------------->
1997 //////////////////////|
1998 ////////////////////// |
1999 ////////////////////// |.................. ................
2002 <----- [hv]display ----->
2003 <------------- [hv]sync_start ------------>
2004 <--------------------- [hv]sync_end --------------------->
2005 <-------------------------------- [hv]total ----------------------------->
2009 <synopsis>__u16 hskew;
2010 __u16 vscan;</synopsis>
2011 <para>Unknown</para>
2014 <synopsis>__u32 flags;</synopsis>
2016 Mode flags, a combination of
2019 <term>DRM_MODE_FLAG_PHSYNC</term>
2021 Horizontal sync is active high
2025 <term>DRM_MODE_FLAG_NHSYNC</term>
2027 Horizontal sync is active low
2031 <term>DRM_MODE_FLAG_PVSYNC</term>
2033 Vertical sync is active high
2037 <term>DRM_MODE_FLAG_NVSYNC</term>
2039 Vertical sync is active low
2043 <term>DRM_MODE_FLAG_INTERLACE</term>
2049 <term>DRM_MODE_FLAG_DBLSCAN</term>
2051 Mode uses doublescan
2055 <term>DRM_MODE_FLAG_CSYNC</term>
2057 Mode uses composite sync
2061 <term>DRM_MODE_FLAG_PCSYNC</term>
2063 Composite sync is active high
2067 <term>DRM_MODE_FLAG_NCSYNC</term>
2069 Composite sync is active low
2073 <term>DRM_MODE_FLAG_HSKEW</term>
2075 hskew provided (not used?)
2079 <term>DRM_MODE_FLAG_BCAST</term>
2085 <term>DRM_MODE_FLAG_PIXMUX</term>
2091 <term>DRM_MODE_FLAG_DBLCLK</term>
2097 <term>DRM_MODE_FLAG_CLKDIV2</term>
2105 Note that modes marked with the INTERLACE or DBLSCAN flags will be
2107 <function>drm_helper_probe_single_connector_modes</function> if
2108 the connector's <structfield>interlace_allowed</structfield> or
2109 <structfield>doublescan_allowed</structfield> field is set to 0.
2113 <synopsis>char name[DRM_DISPLAY_MODE_LEN];</synopsis>
2115 Mode name. The driver must call
2116 <function>drm_mode_set_name</function> to fill the mode name from
2117 <structfield>hdisplay</structfield>,
2118 <structfield>vdisplay</structfield> and interlace flag after
2119 filling the corresponding fields.
2125 The <structfield>vrefresh</structfield> value is computed by
2126 <function>drm_helper_probe_single_connector_modes</function>.
2129 When parsing EDID data, <function>drm_add_edid_modes</function> fill the
2130 connector <structfield>display_info</structfield>
2131 <structfield>width_mm</structfield> and
2132 <structfield>height_mm</structfield> fields. When creating modes
2133 manually the <methodname>get_modes</methodname> helper operation must
2134 set the <structfield>display_info</structfield>
2135 <structfield>width_mm</structfield> and
2136 <structfield>height_mm</structfield> fields if they haven't been set
2137 already (for instance at initilization time when a fixed-size panel is
2138 attached to the connector). The mode <structfield>width_mm</structfield>
2139 and <structfield>height_mm</structfield> fields are only used internally
2140 during EDID parsing and should not be set when creating modes manually.
2144 <synopsis>int (*mode_valid)(struct drm_connector *connector,
2145 struct drm_display_mode *mode);</synopsis>
2147 Verify whether a mode is valid for the connector. Return MODE_OK for
2148 supported modes and one of the enum drm_mode_status values (MODE_*)
2149 for unsupported modes. This operation is mandatory.
2152 As the mode rejection reason is currently not used beside for
2153 immediately removing the unsupported mode, an implementation can
2154 return MODE_BAD regardless of the exact reason why the mode is not
2158 Note that the <methodname>mode_valid</methodname> helper operation is
2159 only called for modes detected by the device, and
2160 <emphasis>not</emphasis> for modes set by the user through the CRTC
2161 <methodname>set_config</methodname> operation.
2167 <title>Modeset Helper Functions Reference</title>
2168 !Edrivers/gpu/drm/drm_crtc_helper.c
2171 <title>fbdev Helper Functions Reference</title>
2172 !Pdrivers/gpu/drm/drm_fb_helper.c fbdev helpers
2173 !Edrivers/gpu/drm/drm_fb_helper.c
2174 !Iinclude/drm/drm_fb_helper.h
2177 <title>Display Port Helper Functions Reference</title>
2178 !Pdrivers/gpu/drm/drm_dp_helper.c dp helpers
2179 !Iinclude/drm/drm_dp_helper.h
2180 !Edrivers/gpu/drm/drm_dp_helper.c
2183 <title>EDID Helper Functions Reference</title>
2184 !Edrivers/gpu/drm/drm_edid.c
2187 <title>Rectangle Utilities Reference</title>
2188 !Pinclude/drm/drm_rect.h rect utils
2189 !Iinclude/drm/drm_rect.h
2190 !Edrivers/gpu/drm/drm_rect.c
2193 <title>Flip-work Helper Reference</title>
2194 !Pinclude/drm/drm_flip_work.h flip utils
2195 !Iinclude/drm/drm_flip_work.h
2196 !Edrivers/gpu/drm/drm_flip_work.c
2199 <title>VMA Offset Manager</title>
2200 !Pdrivers/gpu/drm/drm_vma_manager.c vma offset manager
2201 !Edrivers/gpu/drm/drm_vma_manager.c
2202 !Iinclude/drm/drm_vma_manager.h
2206 <!-- Internals: kms properties -->
2208 <sect1 id="drm-kms-properties">
2209 <title>KMS Properties</title>
2211 Drivers may need to expose additional parameters to applications than
2212 those described in the previous sections. KMS supports attaching
2213 properties to CRTCs, connectors and planes and offers a userspace API to
2214 list, get and set the property values.
2217 Properties are identified by a name that uniquely defines the property
2218 purpose, and store an associated value. For all property types except blob
2219 properties the value is a 64-bit unsigned integer.
2222 KMS differentiates between properties and property instances. Drivers
2223 first create properties and then create and associate individual instances
2224 of those properties to objects. A property can be instantiated multiple
2225 times and associated with different objects. Values are stored in property
2226 instances, and all other property information are stored in the propery
2227 and shared between all instances of the property.
2230 Every property is created with a type that influences how the KMS core
2231 handles the property. Supported property types are
2234 <term>DRM_MODE_PROP_RANGE</term>
2235 <listitem><para>Range properties report their minimum and maximum
2236 admissible values. The KMS core verifies that values set by
2237 application fit in that range.</para></listitem>
2240 <term>DRM_MODE_PROP_ENUM</term>
2241 <listitem><para>Enumerated properties take a numerical value that
2242 ranges from 0 to the number of enumerated values defined by the
2243 property minus one, and associate a free-formed string name to each
2244 value. Applications can retrieve the list of defined value-name pairs
2245 and use the numerical value to get and set property instance values.
2249 <term>DRM_MODE_PROP_BITMASK</term>
2250 <listitem><para>Bitmask properties are enumeration properties that
2251 additionally restrict all enumerated values to the 0..63 range.
2252 Bitmask property instance values combine one or more of the
2253 enumerated bits defined by the property.</para></listitem>
2256 <term>DRM_MODE_PROP_BLOB</term>
2257 <listitem><para>Blob properties store a binary blob without any format
2258 restriction. The binary blobs are created as KMS standalone objects,
2259 and blob property instance values store the ID of their associated
2261 <para>Blob properties are only used for the connector EDID property
2262 and cannot be created by drivers.</para></listitem>
2267 To create a property drivers call one of the following functions depending
2268 on the property type. All property creation functions take property flags
2269 and name, as well as type-specific arguments.
2272 <synopsis>struct drm_property *drm_property_create_range(struct drm_device *dev, int flags,
2274 uint64_t min, uint64_t max);</synopsis>
2275 <para>Create a range property with the given minimum and maximum
2279 <synopsis>struct drm_property *drm_property_create_enum(struct drm_device *dev, int flags,
2281 const struct drm_prop_enum_list *props,
2282 int num_values);</synopsis>
2283 <para>Create an enumerated property. The <parameter>props</parameter>
2284 argument points to an array of <parameter>num_values</parameter>
2285 value-name pairs.</para>
2288 <synopsis>struct drm_property *drm_property_create_bitmask(struct drm_device *dev,
2289 int flags, const char *name,
2290 const struct drm_prop_enum_list *props,
2291 int num_values);</synopsis>
2292 <para>Create a bitmask property. The <parameter>props</parameter>
2293 argument points to an array of <parameter>num_values</parameter>
2294 value-name pairs.</para>
2299 Properties can additionally be created as immutable, in which case they
2300 will be read-only for applications but can be modified by the driver. To
2301 create an immutable property drivers must set the DRM_MODE_PROP_IMMUTABLE
2302 flag at property creation time.
2305 When no array of value-name pairs is readily available at property
2306 creation time for enumerated or range properties, drivers can create
2307 the property using the <function>drm_property_create</function> function
2308 and manually add enumeration value-name pairs by calling the
2309 <function>drm_property_add_enum</function> function. Care must be taken to
2310 properly specify the property type through the <parameter>flags</parameter>
2314 After creating properties drivers can attach property instances to CRTC,
2315 connector and plane objects by calling the
2316 <function>drm_object_attach_property</function>. The function takes a
2317 pointer to the target object, a pointer to the previously created property
2318 and an initial instance value.
2322 <!-- Internals: vertical blanking -->
2324 <sect1 id="drm-vertical-blank">
2325 <title>Vertical Blanking</title>
2327 Vertical blanking plays a major role in graphics rendering. To achieve
2328 tear-free display, users must synchronize page flips and/or rendering to
2329 vertical blanking. The DRM API offers ioctls to perform page flips
2330 synchronized to vertical blanking and wait for vertical blanking.
2333 The DRM core handles most of the vertical blanking management logic, which
2334 involves filtering out spurious interrupts, keeping race-free blanking
2335 counters, coping with counter wrap-around and resets and keeping use
2336 counts. It relies on the driver to generate vertical blanking interrupts
2337 and optionally provide a hardware vertical blanking counter. Drivers must
2338 implement the following operations.
2342 <synopsis>int (*enable_vblank) (struct drm_device *dev, int crtc);
2343 void (*disable_vblank) (struct drm_device *dev, int crtc);</synopsis>
2345 Enable or disable vertical blanking interrupts for the given CRTC.
2349 <synopsis>u32 (*get_vblank_counter) (struct drm_device *dev, int crtc);</synopsis>
2351 Retrieve the value of the vertical blanking counter for the given
2352 CRTC. If the hardware maintains a vertical blanking counter its value
2353 should be returned. Otherwise drivers can use the
2354 <function>drm_vblank_count</function> helper function to handle this
2360 Drivers must initialize the vertical blanking handling core with a call to
2361 <function>drm_vblank_init</function> in their
2362 <methodname>load</methodname> operation. The function will set the struct
2363 <structname>drm_device</structname>
2364 <structfield>vblank_disable_allowed</structfield> field to 0. This will
2365 keep vertical blanking interrupts enabled permanently until the first mode
2366 set operation, where <structfield>vblank_disable_allowed</structfield> is
2367 set to 1. The reason behind this is not clear. Drivers can set the field
2368 to 1 after <function>calling drm_vblank_init</function> to make vertical
2369 blanking interrupts dynamically managed from the beginning.
2372 Vertical blanking interrupts can be enabled by the DRM core or by drivers
2373 themselves (for instance to handle page flipping operations). The DRM core
2374 maintains a vertical blanking use count to ensure that the interrupts are
2375 not disabled while a user still needs them. To increment the use count,
2376 drivers call <function>drm_vblank_get</function>. Upon return vertical
2377 blanking interrupts are guaranteed to be enabled.
2380 To decrement the use count drivers call
2381 <function>drm_vblank_put</function>. Only when the use count drops to zero
2382 will the DRM core disable the vertical blanking interrupts after a delay
2383 by scheduling a timer. The delay is accessible through the vblankoffdelay
2384 module parameter or the <varname>drm_vblank_offdelay</varname> global
2385 variable and expressed in milliseconds. Its default value is 5000 ms.
2388 When a vertical blanking interrupt occurs drivers only need to call the
2389 <function>drm_handle_vblank</function> function to account for the
2393 Resources allocated by <function>drm_vblank_init</function> must be freed
2394 with a call to <function>drm_vblank_cleanup</function> in the driver
2395 <methodname>unload</methodname> operation handler.
2399 <!-- Internals: open/close, file operations and ioctls -->
2402 <title>Open/Close, File Operations and IOCTLs</title>
2404 <title>Open and Close</title>
2405 <synopsis>int (*firstopen) (struct drm_device *);
2406 void (*lastclose) (struct drm_device *);
2407 int (*open) (struct drm_device *, struct drm_file *);
2408 void (*preclose) (struct drm_device *, struct drm_file *);
2409 void (*postclose) (struct drm_device *, struct drm_file *);</synopsis>
2410 <abstract>Open and close handlers. None of those methods are mandatory.
2413 The <methodname>firstopen</methodname> method is called by the DRM core
2414 for legacy UMS (User Mode Setting) drivers only when an application
2415 opens a device that has no other opened file handle. UMS drivers can
2416 implement it to acquire device resources. KMS drivers can't use the
2417 method and must acquire resources in the <methodname>load</methodname>
2421 Similarly the <methodname>lastclose</methodname> method is called when
2422 the last application holding a file handle opened on the device closes
2423 it, for both UMS and KMS drivers. Additionally, the method is also
2424 called at module unload time or, for hot-pluggable devices, when the
2425 device is unplugged. The <methodname>firstopen</methodname> and
2426 <methodname>lastclose</methodname> calls can thus be unbalanced.
2429 The <methodname>open</methodname> method is called every time the device
2430 is opened by an application. Drivers can allocate per-file private data
2431 in this method and store them in the struct
2432 <structname>drm_file</structname> <structfield>driver_priv</structfield>
2433 field. Note that the <methodname>open</methodname> method is called
2434 before <methodname>firstopen</methodname>.
2437 The close operation is split into <methodname>preclose</methodname> and
2438 <methodname>postclose</methodname> methods. Drivers must stop and
2439 cleanup all per-file operations in the <methodname>preclose</methodname>
2440 method. For instance pending vertical blanking and page flip events must
2441 be cancelled. No per-file operation is allowed on the file handle after
2442 returning from the <methodname>preclose</methodname> method.
2445 Finally the <methodname>postclose</methodname> method is called as the
2446 last step of the close operation, right before calling the
2447 <methodname>lastclose</methodname> method if no other open file handle
2448 exists for the device. Drivers that have allocated per-file private data
2449 in the <methodname>open</methodname> method should free it here.
2452 The <methodname>lastclose</methodname> method should restore CRTC and
2453 plane properties to default value, so that a subsequent open of the
2454 device will not inherit state from the previous user. It can also be
2455 used to execute delayed power switching state changes, e.g. in
2456 conjunction with the vga-switcheroo infrastructure. Beyond that KMS
2457 drivers should not do any further cleanup. Only legacy UMS drivers might
2458 need to clean up device state so that the vga console or an independent
2459 fbdev driver could take over.
2463 <title>File Operations</title>
2464 <synopsis>const struct file_operations *fops</synopsis>
2465 <abstract>File operations for the DRM device node.</abstract>
2467 Drivers must define the file operations structure that forms the DRM
2468 userspace API entry point, even though most of those operations are
2469 implemented in the DRM core. The <methodname>open</methodname>,
2470 <methodname>release</methodname> and <methodname>ioctl</methodname>
2471 operations are handled by
2473 .owner = THIS_MODULE,
2475 .release = drm_release,
2476 .unlocked_ioctl = drm_ioctl,
2477 #ifdef CONFIG_COMPAT
2478 .compat_ioctl = drm_compat_ioctl,
2483 Drivers that implement private ioctls that requires 32/64bit
2484 compatibility support must provide their own
2485 <methodname>compat_ioctl</methodname> handler that processes private
2486 ioctls and calls <function>drm_compat_ioctl</function> for core ioctls.
2489 The <methodname>read</methodname> and <methodname>poll</methodname>
2490 operations provide support for reading DRM events and polling them. They
2495 .llseek = no_llseek,
2499 The memory mapping implementation varies depending on how the driver
2500 manages memory. Pre-GEM drivers will use <function>drm_mmap</function>,
2501 while GEM-aware drivers will use <function>drm_gem_mmap</function>. See
2502 <xref linkend="drm-gem"/>.
2504 .mmap = drm_gem_mmap,
2508 No other file operation is supported by the DRM API.
2512 <title>IOCTLs</title>
2513 <synopsis>struct drm_ioctl_desc *ioctls;
2514 int num_ioctls;</synopsis>
2515 <abstract>Driver-specific ioctls descriptors table.</abstract>
2517 Driver-specific ioctls numbers start at DRM_COMMAND_BASE. The ioctls
2518 descriptors table is indexed by the ioctl number offset from the base
2519 value. Drivers can use the DRM_IOCTL_DEF_DRV() macro to initialize the
2523 <programlisting>DRM_IOCTL_DEF_DRV(ioctl, func, flags)</programlisting>
2525 <parameter>ioctl</parameter> is the ioctl name. Drivers must define
2526 the DRM_##ioctl and DRM_IOCTL_##ioctl macros to the ioctl number
2527 offset from DRM_COMMAND_BASE and the ioctl number respectively. The
2528 first macro is private to the device while the second must be exposed
2529 to userspace in a public header.
2532 <parameter>func</parameter> is a pointer to the ioctl handler function
2533 compatible with the <type>drm_ioctl_t</type> type.
2534 <programlisting>typedef int drm_ioctl_t(struct drm_device *dev, void *data,
2535 struct drm_file *file_priv);</programlisting>
2538 <parameter>flags</parameter> is a bitmask combination of the following
2539 values. It restricts how the ioctl is allowed to be called.
2542 DRM_AUTH - Only authenticated callers allowed
2545 DRM_MASTER - The ioctl can only be called on the master file
2549 DRM_ROOT_ONLY - Only callers with the SYSADMIN capability allowed
2552 DRM_CONTROL_ALLOW - The ioctl can only be called on a control
2556 DRM_UNLOCKED - The ioctl handler will be called without locking
2557 the DRM global mutex
2566 <title>Command submission & fencing</title>
2568 This should cover a few device-specific command submission
2573 <!-- Internals: suspend/resume -->
2576 <title>Suspend/Resume</title>
2578 The DRM core provides some suspend/resume code, but drivers wanting full
2579 suspend/resume support should provide save() and restore() functions.
2580 These are called at suspend, hibernate, or resume time, and should perform
2581 any state save or restore required by your device across suspend or
2584 <synopsis>int (*suspend) (struct drm_device *, pm_message_t state);
2585 int (*resume) (struct drm_device *);</synopsis>
2587 Those are legacy suspend and resume methods. New driver should use the
2588 power management interface provided by their bus type (usually through
2589 the struct <structname>device_driver</structname> dev_pm_ops) and set
2590 these methods to NULL.
2595 <title>DMA services</title>
2597 This should cover how DMA mapping etc. is supported by the core.
2598 These functions are deprecated and should not be used.
2606 - Document the struct_mutex catch-all lock
2607 - Document connector properties
2609 - Why is the load method optional?
2610 - What are drivers supposed to set the initial display state to, and how?
2611 Connector's DPMS states are not initialized and are thus equal to
2612 DRM_MODE_DPMS_ON. The fbcon compatibility layer calls
2613 drm_helper_disable_unused_functions(), which disables unused encoders and
2614 CRTCs, but doesn't touch the connectors' DPMS state, and
2615 drm_helper_connector_dpms() in reaction to fbdev blanking events. Do drivers
2616 that don't implement (or just don't use) fbcon compatibility need to call
2617 those functions themselves?
2618 - KMS drivers must call drm_vblank_pre_modeset() and drm_vblank_post_modeset()
2619 around mode setting. Should this be done in the DRM core?
2620 - vblank_disable_allowed is set to 1 in the first drm_vblank_post_modeset()
2621 call and never set back to 0. It seems to be safe to permanently set it to 1
2622 in drm_vblank_init() for KMS driver, and it might be safe for UMS drivers as
2623 well. This should be investigated.
2624 - crtc and connector .save and .restore operations are only used internally in
2625 drivers, should they be removed from the core?
2626 - encoder mid-layer .save and .restore operations are only used internally in
2627 drivers, should they be removed from the core?
2628 - encoder mid-layer .detect operation is only used internally in drivers,
2629 should it be removed from the core?
2632 <!-- External interfaces -->
2634 <chapter id="drmExternals">
2635 <title>Userland interfaces</title>
2637 The DRM core exports several interfaces to applications,
2638 generally intended to be used through corresponding libdrm
2639 wrapper functions. In addition, drivers export device-specific
2640 interfaces for use by userspace drivers & device-aware
2641 applications through ioctls and sysfs files.
2644 External interfaces include: memory mapping, context management,
2645 DMA operations, AGP management, vblank control, fence
2646 management, memory management, and output management.
2649 Cover generic ioctls and sysfs layout here. We only need high-level
2650 info, since man pages should cover the rest.
2653 <!-- External: render nodes -->
2656 <title>Render nodes</title>
2658 DRM core provides multiple character-devices for user-space to use.
2659 Depending on which device is opened, user-space can perform a different
2660 set of operations (mainly ioctls). The primary node is always created
2661 and called <term>card<num></term>. Additionally, a currently
2662 unused control node, called <term>controlD<num></term> is also
2663 created. The primary node provides all legacy operations and
2664 historically was the only interface used by userspace. With KMS, the
2665 control node was introduced. However, the planned KMS control interface
2666 has never been written and so the control node stays unused to date.
2669 With the increased use of offscreen renderers and GPGPU applications,
2670 clients no longer require running compositors or graphics servers to
2671 make use of a GPU. But the DRM API required unprivileged clients to
2672 authenticate to a DRM-Master prior to getting GPU access. To avoid this
2673 step and to grant clients GPU access without authenticating, render
2674 nodes were introduced. Render nodes solely serve render clients, that
2675 is, no modesetting or privileged ioctls can be issued on render nodes.
2676 Only non-global rendering commands are allowed. If a driver supports
2677 render nodes, it must advertise it via the <term>DRIVER_RENDER</term>
2678 DRM driver capability. If not supported, the primary node must be used
2679 for render clients together with the legacy drmAuth authentication
2683 If a driver advertises render node support, DRM core will create a
2684 separate render node called <term>renderD<num></term>. There will
2685 be one render node per device. No ioctls except PRIME-related ioctls
2686 will be allowed on this node. Especially <term>GEM_OPEN</term> will be
2687 explicitly prohibited. Render nodes are designed to avoid the
2688 buffer-leaks, which occur if clients guess the flink names or mmap
2689 offsets on the legacy interface. Additionally to this basic interface,
2690 drivers must mark their driver-dependent render-only ioctls as
2691 <term>DRM_RENDER_ALLOW</term> so render clients can use them. Driver
2692 authors must be careful not to allow any privileged ioctls on render
2696 With render nodes, user-space can now control access to the render node
2697 via basic file-system access-modes. A running graphics server which
2698 authenticates clients on the privileged primary/legacy node is no longer
2699 required. Instead, a client can open the render node and is immediately
2700 granted GPU access. Communication between clients (or servers) is done
2701 via PRIME. FLINK from render node to legacy node is not supported. New
2702 clients must not use the insecure FLINK interface.
2705 Besides dropping all modeset/global ioctls, render nodes also drop the
2706 DRM-Master concept. There is no reason to associate render clients with
2707 a DRM-Master as they are independent of any graphics server. Besides,
2708 they must work without any running master, anyway.
2709 Drivers must be able to run without a master object if they support
2710 render nodes. If, on the other hand, a driver requires shared state
2711 between clients which is visible to user-space and accessible beyond
2712 open-file boundaries, they cannot support render nodes.
2716 <!-- External: vblank handling -->
2719 <title>VBlank event handling</title>
2721 The DRM core exposes two vertical blank related ioctls:
2724 <term>DRM_IOCTL_WAIT_VBLANK</term>
2727 This takes a struct drm_wait_vblank structure as its argument,
2728 and it is used to block or request a signal when a specified
2729 vblank event occurs.
2734 <term>DRM_IOCTL_MODESET_CTL</term>
2737 This should be called by application level drivers before and
2738 after mode setting, since on many devices the vertical blank
2739 counter is reset at that time. Internally, the DRM snapshots
2740 the last vblank count when the ioctl is called with the
2741 _DRM_PRE_MODESET command, so that the counter won't go backwards
2742 (which is dealt with when _DRM_POST_MODESET is used).
2747 <!--!Edrivers/char/drm/drm_irq.c-->
2753 <!-- API reference -->
2755 <appendix id="drmDriverApi">
2756 <title>DRM Driver API</title>
2758 Include auto-generated API reference here (need to reference it
2759 from paragraphs above too).
projects / platform / adaptation / renesas_rcar / renesas_kernel.git / blob