xref: /linux/Documentation/gpu/introduction.rst (revision 249ebf3f65f8530beb2cbfb91bff1d83ba88d23c)
1============
2Introduction
3============
4
5The Linux DRM layer contains code intended to support the needs of
6complex graphics devices, usually containing programmable pipelines well
7suited to 3D graphics acceleration. Graphics drivers in the kernel may
8make use of DRM functions to make tasks like memory management,
9interrupt handling and DMA easier, and provide a uniform interface to
10applications.
11
12A note on versions: this guide covers features found in the DRM tree,
13including the TTM memory manager, output configuration and mode setting,
14and the new vblank internals, in addition to all the regular features
15found in current kernels.
16
17[Insert diagram of typical DRM stack here]
18
19Style Guidelines
20================
21
22For consistency this documentation uses American English. Abbreviations
23are written as all-uppercase, for example: DRM, KMS, IOCTL, CRTC, and so
24on. To aid in reading, documentations make full use of the markup
25characters kerneldoc provides: @parameter for function parameters,
26@member for structure members (within the same structure), &struct structure to
27reference structures and function() for functions. These all get automatically
28hyperlinked if kerneldoc for the referenced objects exists. When referencing
29entries in function vtables (and structure members in general) please use
30&vtable_name.vfunc. Unfortunately this does not yet yield a direct link to the
31member, only the structure.
32
33Except in special situations (to separate locked from unlocked variants)
34locking requirements for functions aren't documented in the kerneldoc.
35Instead locking should be check at runtime using e.g.
36``WARN_ON(!mutex_is_locked(...));``. Since it's much easier to ignore
37documentation than runtime noise this provides more value. And on top of
38that runtime checks do need to be updated when the locking rules change,
39increasing the chances that they're correct. Within the documentation
40the locking rules should be explained in the relevant structures: Either
41in the comment for the lock explaining what it protects, or data fields
42need a note about which lock protects them, or both.
43
44Functions which have a non-\ ``void`` return value should have a section
45called "Returns" explaining the expected return values in different
46cases and their meanings. Currently there's no consensus whether that
47section name should be all upper-case or not, and whether it should end
48in a colon or not. Go with the file-local style. Other common section
49names are "Notes" with information for dangerous or tricky corner cases,
50and "FIXME" where the interface could be cleaned up.
51
52Also read the :ref:`guidelines for the kernel documentation at large <doc_guide>`.
53
54Documentation Requirements for kAPI
55-----------------------------------
56
57All kernel APIs exported to other modules must be documented, including their
58datastructures and at least a short introductory section explaining the overall
59concepts. Documentation should be put into the code itself as kerneldoc comments
60as much as reasonable.
61
62Do not blindly document everything, but document only what's relevant for driver
63authors: Internal functions of drm.ko and definitely static functions should not
64have formal kerneldoc comments. Use normal C comments if you feel like a comment
65is warranted. You may use kerneldoc syntax in the comment, but it shall not
66start with a /** kerneldoc marker. Similar for data structures, annotate
67anything entirely private with ``/* private: */`` comments as per the
68documentation guide.
69
70Getting Started
71===============
72
73Developers interested in helping out with the DRM subsystem are very welcome.
74Often people will resort to sending in patches for various issues reported by
75checkpatch or sparse. We welcome such contributions.
76
77Anyone looking to kick it up a notch can find a list of janitorial tasks on
78the :ref:`TODO list <todo>`.
79
80Contribution Process
81====================
82
83Mostly the DRM subsystem works like any other kernel subsystem, see :ref:`the
84main process guidelines and documentation <process_index>` for how things work.
85Here we just document some of the specialities of the GPU subsystem.
86
87Feature Merge Deadlines
88-----------------------
89
90All feature work must be in the linux-next tree by the -rc6 release of the
91current release cycle, otherwise they must be postponed and can't reach the next
92merge window. All patches must have landed in the drm-next tree by latest -rc7,
93but if your branch is not in linux-next then this must have happened by -rc6
94already.
95
96After that point only bugfixes (like after the upstream merge window has closed
97with the -rc1 release) are allowed. No new platform enabling or new drivers are
98allowed.
99
100This means that there's a blackout-period of about one month where feature work
101can't be merged. The recommended way to deal with that is having a -next tree
102that's always open, but making sure to not feed it into linux-next during the
103blackout period. As an example, drm-misc works like that.
104
105Code of Conduct
106---------------
107
108As a freedesktop.org project, dri-devel, and the DRM community, follows the
109Contributor Covenant, found at: https://www.freedesktop.org/wiki/CodeOfConduct
110
111Please conduct yourself in a respectful and civilised manner when
112interacting with community members on mailing lists, IRC, or bug
113trackers. The community represents the project as a whole, and abusive
114or bullying behaviour is not tolerated by the project.
115
116Simple DRM drivers to use as examples
117=====================================
118
119The DRM subsystem contains a lot of helper functions to ease writing drivers for
120simple graphic devices. For example, the `drivers/gpu/drm/tiny/` directory has a
121set of drivers that are simple enough to be implemented in a single source file.
122
123These drivers make use of the `struct drm_simple_display_pipe_funcs`, that hides
124any complexity of the DRM subsystem and just requires drivers to implement a few
125functions needed to operate the device. This could be used for devices that just
126need a display pipeline with one full-screen scanout buffer feeding one output.
127
128The tiny DRM drivers are good examples to understand how DRM drivers should look
129like. Since are just a few hundreds lines of code, they are quite easy to read.
130
131External References
132===================
133
134Delving into a Linux kernel subsystem for the first time can be an overwhelming
135experience, one needs to get familiar with all the concepts and learn about the
136subsystem's internals, among other details.
137
138To shallow the learning curve, this section contains a list of presentations
139and documents that can be used to learn about DRM/KMS and graphics in general.
140
141There are different reasons why someone might want to get into DRM: porting an
142existing fbdev driver, write a DRM driver for a new hardware, fixing bugs that
143could face when working on the graphics user-space stack, etc. For this reason,
144the learning material covers many aspects of the Linux graphics stack. From an
145overview of the kernel and user-space stacks to very specific topics.
146
147The list is sorted in reverse chronological order, to keep the most up-to-date
148material at the top. But all of them contain useful information, and it can be
149valuable to go through older material to understand the rationale and context
150in which the changes to the DRM subsystem were made.
151
152Conference talks
153----------------
154
155* `An Overview of the Linux and Userspace Graphics Stack <https://www.youtube.com/watch?v=wjAJmqwg47k>`_ - Paul Kocialkowski (2020)
156* `Getting pixels on screen on Linux: introduction to Kernel Mode Setting <https://www.youtube.com/watch?v=haes4_Xnc5Q>`_ - Simon Ser (2020)
157* `Everything Great about Upstream Graphics <https://www.youtube.com/watch?v=kVzHOgt6WGE>`_ - Daniel Vetter (2019)
158* `An introduction to the Linux DRM subsystem <https://www.youtube.com/watch?v=LbDOCJcDRoo>`_ - Maxime Ripard (2017)
159* `Embrace the Atomic (Display) Age <https://www.youtube.com/watch?v=LjiB_JeDn2M>`_ - Daniel Vetter (2016)
160* `Anatomy of an Atomic KMS Driver <https://www.youtube.com/watch?v=lihqR9sENpc>`_ - Laurent Pinchart (2015)
161* `Atomic Modesetting for Drivers <https://www.youtube.com/watch?v=kl9suFgbTc8>`_ - Daniel Vetter (2015)
162* `Anatomy of an Embedded KMS Driver <https://www.youtube.com/watch?v=Ja8fM7rTae4>`_ - Laurent Pinchart (2013)
163
164Slides and articles
165-------------------
166
167* `The Linux graphics stack in a nutshell, part 1 <https://lwn.net/Articles/955376/>`_ - Thomas Zimmermann (2023)
168* `The Linux graphics stack in a nutshell, part 2 <https://lwn.net/Articles/955708/>`_ - Thomas Zimmermann (2023)
169* `Understanding the Linux Graphics Stack <https://bootlin.com/doc/training/graphics/graphics-slides.pdf>`_ - Bootlin (2022)
170* `DRM KMS overview <https://wiki.st.com/stm32mpu/wiki/DRM_KMS_overview>`_ - STMicroelectronics (2021)
171* `Linux graphic stack <https://studiopixl.com/2017-05-13/linux-graphic-stack-an-overview>`_ - Nathan Gauër (2017)
172* `Atomic mode setting design overview, part 1 <https://lwn.net/Articles/653071/>`_ - Daniel Vetter (2015)
173* `Atomic mode setting design overview, part 2 <https://lwn.net/Articles/653466/>`_ - Daniel Vetter (2015)
174* `The DRM/KMS subsystem from a newbie’s point of view <https://bootlin.com/pub/conferences/2014/elce/brezillon-drm-kms/brezillon-drm-kms.pdf>`_ - Boris Brezillon (2014)
175* `A brief introduction to the Linux graphics stack <https://blogs.igalia.com/itoral/2014/07/29/a-brief-introduction-to-the-linux-graphics-stack/>`_ - Iago Toral (2014)
176* `The Linux Graphics Stack <https://blog.mecheye.net/2012/06/the-linux-graphics-stack/>`_ - Jasper St. Pierre (2012)
177