xref: /openbmc/linux/Documentation/gpu/tegra.rst (revision da1d9caf)
1===============================================
2 drm/tegra NVIDIA Tegra GPU and display driver
3===============================================
4
5NVIDIA Tegra SoCs support a set of display, graphics and video functions via
6the host1x controller. host1x supplies command streams, gathered from a push
7buffer provided directly by the CPU, to its clients via channels. Software,
8or blocks amongst themselves, can use syncpoints for synchronization.
9
10Up until, but not including, Tegra124 (aka Tegra K1) the drm/tegra driver
11supports the built-in GPU, comprised of the gr2d and gr3d engines. Starting
12with Tegra124 the GPU is based on the NVIDIA desktop GPU architecture and
13supported by the drm/nouveau driver.
14
15The drm/tegra driver supports NVIDIA Tegra SoC generations since Tegra20. It
16has three parts:
17
18  - A host1x driver that provides infrastructure and access to the host1x
19    services.
20
21  - A KMS driver that supports the display controllers as well as a number of
22    outputs, such as RGB, HDMI, DSI, and DisplayPort.
23
24  - A set of custom userspace IOCTLs that can be used to submit jobs to the
25    GPU and video engines via host1x.
26
27Driver Infrastructure
28=====================
29
30The various host1x clients need to be bound together into a logical device in
31order to expose their functionality to users. The infrastructure that supports
32this is implemented in the host1x driver. When a driver is registered with the
33infrastructure it provides a list of compatible strings specifying the devices
34that it needs. The infrastructure creates a logical device and scan the device
35tree for matching device nodes, adding the required clients to a list. Drivers
36for individual clients register with the infrastructure as well and are added
37to the logical host1x device.
38
39Once all clients are available, the infrastructure will initialize the logical
40device using a driver-provided function which will set up the bits specific to
41the subsystem and in turn initialize each of its clients.
42
43Similarly, when one of the clients is unregistered, the infrastructure will
44destroy the logical device by calling back into the driver, which ensures that
45the subsystem specific bits are torn down and the clients destroyed in turn.
46
47Host1x Infrastructure Reference
48-------------------------------
49
50.. kernel-doc:: include/linux/host1x.h
51
52.. kernel-doc:: drivers/gpu/host1x/bus.c
53   :export:
54
55Host1x Syncpoint Reference
56--------------------------
57
58.. kernel-doc:: drivers/gpu/host1x/syncpt.c
59   :export:
60
61KMS driver
62==========
63
64The display hardware has remained mostly backwards compatible over the various
65Tegra SoC generations, up until Tegra186 which introduces several changes that
66make it difficult to support with a parameterized driver.
67
68Display Controllers
69-------------------
70
71Tegra SoCs have two display controllers, each of which can be associated with
72zero or more outputs. Outputs can also share a single display controller, but
73only if they run with compatible display timings. Two display controllers can
74also share a single framebuffer, allowing cloned configurations even if modes
75on two outputs don't match. A display controller is modelled as a CRTC in KMS
76terms.
77
78On Tegra186, the number of display controllers has been increased to three. A
79display controller can no longer drive all of the outputs. While two of these
80controllers can drive both DSI outputs and both SOR outputs, the third cannot
81drive any DSI.
82
83Windows
84~~~~~~~
85
86A display controller controls a set of windows that can be used to composite
87multiple buffers onto the screen. While it is possible to assign arbitrary Z
88ordering to individual windows (by programming the corresponding blending
89registers), this is currently not supported by the driver. Instead, it will
90assume a fixed Z ordering of the windows (window A is the root window, that
91is, the lowest, while windows B and C are overlaid on top of window A). The
92overlay windows support multiple pixel formats and can automatically convert
93from YUV to RGB at scanout time. This makes them useful for displaying video
94content. In KMS, each window is modelled as a plane. Each display controller
95has a hardware cursor that is exposed as a cursor plane.
96
97Outputs
98-------
99
100The type and number of supported outputs varies between Tegra SoC generations.
101All generations support at least HDMI. While earlier generations supported the
102very simple RGB interfaces (one per display controller), recent generations no
103longer do and instead provide standard interfaces such as DSI and eDP/DP.
104
105Outputs are modelled as a composite encoder/connector pair.
106
107RGB/LVDS
108~~~~~~~~
109
110This interface is no longer available since Tegra124. It has been replaced by
111the more standard DSI and eDP interfaces.
112
113HDMI
114~~~~
115
116HDMI is supported on all Tegra SoCs. Starting with Tegra210, HDMI is provided
117by the versatile SOR output, which supports eDP, DP and HDMI. The SOR is able
118to support HDMI 2.0, though support for this is currently not merged.
119
120DSI
121~~~
122
123Although Tegra has supported DSI since Tegra30, the controller has changed in
124several ways in Tegra114. Since none of the publicly available development
125boards prior to Dalmore (Tegra114) have made use of DSI, only Tegra114 and
126later are supported by the drm/tegra driver.
127
128eDP/DP
129~~~~~~
130
131eDP was first introduced in Tegra124 where it was used to drive the display
132panel for notebook form factors. Tegra210 added support for full DisplayPort
133support, though this is currently not implemented in the drm/tegra driver.
134
135Userspace Interface
136===================
137
138The userspace interface provided by drm/tegra allows applications to create
139GEM buffers, access and control syncpoints as well as submit command streams
140to host1x.
141
142GEM Buffers
143-----------
144
145The ``DRM_IOCTL_TEGRA_GEM_CREATE`` IOCTL is used to create a GEM buffer object
146with Tegra-specific flags. This is useful for buffers that should be tiled, or
147that are to be scanned out upside down (useful for 3D content).
148
149After a GEM buffer object has been created, its memory can be mapped by an
150application using the mmap offset returned by the ``DRM_IOCTL_TEGRA_GEM_MMAP``
151IOCTL.
152
153Syncpoints
154----------
155
156The current value of a syncpoint can be obtained by executing the
157``DRM_IOCTL_TEGRA_SYNCPT_READ`` IOCTL. Incrementing the syncpoint is achieved
158using the ``DRM_IOCTL_TEGRA_SYNCPT_INCR`` IOCTL.
159
160Userspace can also request blocking on a syncpoint. To do so, it needs to
161execute the ``DRM_IOCTL_TEGRA_SYNCPT_WAIT`` IOCTL, specifying the value of
162the syncpoint to wait for. The kernel will release the application when the
163syncpoint reaches that value or after a specified timeout.
164
165Command Stream Submission
166-------------------------
167
168Before an application can submit command streams to host1x it needs to open a
169channel to an engine using the ``DRM_IOCTL_TEGRA_OPEN_CHANNEL`` IOCTL. Client
170IDs are used to identify the target of the channel. When a channel is no
171longer needed, it can be closed using the ``DRM_IOCTL_TEGRA_CLOSE_CHANNEL``
172IOCTL. To retrieve the syncpoint associated with a channel, an application
173can use the ``DRM_IOCTL_TEGRA_GET_SYNCPT``.
174
175After opening a channel, submitting command streams is easy. The application
176writes commands into the memory backing a GEM buffer object and passes these
177to the ``DRM_IOCTL_TEGRA_SUBMIT`` IOCTL along with various other parameters,
178such as the syncpoints or relocations used in the job submission.
179