xref: /openbmc/linux/Documentation/gpu/todo.rst (revision ac191bcb)
1.. _todo:
2
3=========
4TODO list
5=========
6
7This section contains a list of smaller janitorial tasks in the kernel DRM
8graphics subsystem useful as newbie projects. Or for slow rainy days.
9
10Difficulty
11----------
12
13To make it easier task are categorized into different levels:
14
15Starter: Good tasks to get started with the DRM subsystem.
16
17Intermediate: Tasks which need some experience with working in the DRM
18subsystem, or some specific GPU/display graphics knowledge. For debugging issue
19it's good to have the relevant hardware (or a virtual driver set up) available
20for testing.
21
22Advanced: Tricky tasks that need fairly good understanding of the DRM subsystem
23and graphics topics. Generally need the relevant hardware for development and
24testing.
25
26Expert: Only attempt these if you've successfully completed some tricky
27refactorings already and are an expert in the specific area
28
29Subsystem-wide refactorings
30===========================
31
32Remove custom dumb_map_offset implementations
33---------------------------------------------
34
35All GEM based drivers should be using drm_gem_create_mmap_offset() instead.
36Audit each individual driver, make sure it'll work with the generic
37implementation (there's lots of outdated locking leftovers in various
38implementations), and then remove it.
39
40Contact: Daniel Vetter, respective driver maintainers
41
42Level: Intermediate
43
44Convert existing KMS drivers to atomic modesetting
45--------------------------------------------------
46
473.19 has the atomic modeset interfaces and helpers, so drivers can now be
48converted over. Modern compositors like Wayland or Surfaceflinger on Android
49really want an atomic modeset interface, so this is all about the bright
50future.
51
52There is a conversion guide for atomic [1]_ and all you need is a GPU for a
53non-converted driver.  The "Atomic mode setting design overview" series [2]_
54[3]_ at LWN.net can also be helpful.
55
56As part of this drivers also need to convert to universal plane (which means
57exposing primary & cursor as proper plane objects). But that's much easier to
58do by directly using the new atomic helper driver callbacks.
59
60  .. [1] https://blog.ffwll.ch/2014/11/atomic-modeset-support-for-kms-drivers.html
61  .. [2] https://lwn.net/Articles/653071/
62  .. [3] https://lwn.net/Articles/653466/
63
64Contact: Daniel Vetter, respective driver maintainers
65
66Level: Advanced
67
68Clean up the clipped coordination confusion around planes
69---------------------------------------------------------
70
71We have a helper to get this right with drm_plane_helper_check_update(), but
72it's not consistently used. This should be fixed, preferably in the atomic
73helpers (and drivers then moved over to clipped coordinates). Probably the
74helper should also be moved from drm_plane_helper.c to the atomic helpers, to
75avoid confusion - the other helpers in that file are all deprecated legacy
76helpers.
77
78Contact: Ville Syrjälä, Daniel Vetter, driver maintainers
79
80Level: Advanced
81
82Improve plane atomic_check helpers
83----------------------------------
84
85Aside from the clipped coordinates right above there's a few suboptimal things
86with the current helpers:
87
88- drm_plane_helper_funcs->atomic_check gets called for enabled or disabled
89  planes. At best this seems to confuse drivers, worst it means they blow up
90  when the plane is disabled without the CRTC. The only special handling is
91  resetting values in the plane state structures, which instead should be moved
92  into the drm_plane_funcs->atomic_duplicate_state functions.
93
94- Once that's done, helpers could stop calling ->atomic_check for disabled
95  planes.
96
97- Then we could go through all the drivers and remove the more-or-less confused
98  checks for plane_state->fb and plane_state->crtc.
99
100Contact: Daniel Vetter
101
102Level: Advanced
103
104Convert early atomic drivers to async commit helpers
105----------------------------------------------------
106
107For the first year the atomic modeset helpers didn't support asynchronous /
108nonblocking commits, and every driver had to hand-roll them. This is fixed
109now, but there's still a pile of existing drivers that easily could be
110converted over to the new infrastructure.
111
112One issue with the helpers is that they require that drivers handle completion
113events for atomic commits correctly. But fixing these bugs is good anyway.
114
115Somewhat related is the legacy_cursor_update hack, which should be replaced with
116the new atomic_async_check/commit functionality in the helpers in drivers that
117still look at that flag.
118
119Contact: Daniel Vetter, respective driver maintainers
120
121Level: Advanced
122
123Fallout from atomic KMS
124-----------------------
125
126``drm_atomic_helper.c`` provides a batch of functions which implement legacy
127IOCTLs on top of the new atomic driver interface. Which is really nice for
128gradual conversion of drivers, but unfortunately the semantic mismatches are
129a bit too severe. So there's some follow-up work to adjust the function
130interfaces to fix these issues:
131
132* atomic needs the lock acquire context. At the moment that's passed around
133  implicitly with some horrible hacks, and it's also allocate with
134  ``GFP_NOFAIL`` behind the scenes. All legacy paths need to start allocating
135  the acquire context explicitly on stack and then also pass it down into
136  drivers explicitly so that the legacy-on-atomic functions can use them.
137
138  Except for some driver code this is done. This task should be finished by
139  adding WARN_ON(!drm_drv_uses_atomic_modeset) in drm_modeset_lock_all().
140
141* A bunch of the vtable hooks are now in the wrong place: DRM has a split
142  between core vfunc tables (named ``drm_foo_funcs``), which are used to
143  implement the userspace ABI. And then there's the optional hooks for the
144  helper libraries (name ``drm_foo_helper_funcs``), which are purely for
145  internal use. Some of these hooks should be move from ``_funcs`` to
146  ``_helper_funcs`` since they are not part of the core ABI. There's a
147  ``FIXME`` comment in the kerneldoc for each such case in ``drm_crtc.h``.
148
149Contact: Daniel Vetter
150
151Level: Intermediate
152
153Get rid of dev->struct_mutex from GEM drivers
154---------------------------------------------
155
156``dev->struct_mutex`` is the Big DRM Lock from legacy days and infested
157everything. Nowadays in modern drivers the only bit where it's mandatory is
158serializing GEM buffer object destruction. Which unfortunately means drivers
159have to keep track of that lock and either call ``unreference`` or
160``unreference_locked`` depending upon context.
161
162Core GEM doesn't have a need for ``struct_mutex`` any more since kernel 4.8,
163and there's a GEM object ``free`` callback for any drivers which are
164entirely ``struct_mutex`` free.
165
166For drivers that need ``struct_mutex`` it should be replaced with a driver-
167private lock. The tricky part is the BO free functions, since those can't
168reliably take that lock any more. Instead state needs to be protected with
169suitable subordinate locks or some cleanup work pushed to a worker thread. For
170performance-critical drivers it might also be better to go with a more
171fine-grained per-buffer object and per-context lockings scheme. Currently only
172the ``msm`` and `i915` drivers use ``struct_mutex``.
173
174Contact: Daniel Vetter, respective driver maintainers
175
176Level: Advanced
177
178Move Buffer Object Locking to dma_resv_lock()
179---------------------------------------------
180
181Many drivers have their own per-object locking scheme, usually using
182mutex_lock(). This causes all kinds of trouble for buffer sharing, since
183depending which driver is the exporter and importer, the locking hierarchy is
184reversed.
185
186To solve this we need one standard per-object locking mechanism, which is
187dma_resv_lock(). This lock needs to be called as the outermost lock, with all
188other driver specific per-object locks removed. The problem is that rolling out
189the actual change to the locking contract is a flag day, due to struct dma_buf
190buffer sharing.
191
192Level: Expert
193
194Convert logging to drm_* functions with drm_device parameter
195------------------------------------------------------------
196
197For drivers which could have multiple instances, it is necessary to
198differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
199don't do this, drivers used dev_info/warn/err to make this differentiation. We
200now have drm_* variants of the drm print functions, so we can start to convert
201those drivers back to using drm-formatted specific log messages.
202
203Before you start this conversion please contact the relevant maintainers to make
204sure your work will be merged - not everyone agrees that the DRM dmesg macros
205are better.
206
207Contact: Sean Paul, Maintainer of the driver you plan to convert
208
209Level: Starter
210
211Convert drivers to use simple modeset suspend/resume
212----------------------------------------------------
213
214Most drivers (except i915 and nouveau) that use
215drm_atomic_helper_suspend/resume() can probably be converted to use
216drm_mode_config_helper_suspend/resume(). Also there's still open-coded version
217of the atomic suspend/resume code in older atomic modeset drivers.
218
219Contact: Maintainer of the driver you plan to convert
220
221Level: Intermediate
222
223Convert drivers to use drm_fbdev_generic_setup()
224------------------------------------------------
225
226Most drivers can use drm_fbdev_generic_setup(). Driver have to implement
227atomic modesetting and GEM vmap support. Historically, generic fbdev emulation
228expected the framebuffer in system memory or system-like memory. By employing
229struct iosys_map, drivers with frambuffers in I/O memory can be supported
230as well.
231
232Contact: Maintainer of the driver you plan to convert
233
234Level: Intermediate
235
236Reimplement functions in drm_fbdev_fb_ops without fbdev
237-------------------------------------------------------
238
239A number of callback functions in drm_fbdev_fb_ops could benefit from
240being rewritten without dependencies on the fbdev module. Some of the
241helpers could further benefit from using struct iosys_map instead of
242raw pointers.
243
244Contact: Thomas Zimmermann <tzimmermann@suse.de>, Daniel Vetter
245
246Level: Advanced
247
248Benchmark and optimize blitting and format-conversion function
249--------------------------------------------------------------
250
251Drawing to display memory quickly is crucial for many applications'
252performance.
253
254On at least x86-64, sys_imageblit() is significantly slower than
255cfb_imageblit(), even though both use the same blitting algorithm and
256the latter is written for I/O memory. It turns out that cfb_imageblit()
257uses movl instructions, while sys_imageblit apparently does not. This
258seems to be a problem with gcc's optimizer. DRM's format-conversion
259helpers might be subject to similar issues.
260
261Benchmark and optimize fbdev's sys_() helpers and DRM's format-conversion
262helpers. In cases that can be further optimized, maybe implement a different
263algorithm. For micro-optimizations, use movl/movq instructions explicitly.
264That might possibly require architecture-specific helpers (e.g., storel()
265storeq()).
266
267Contact: Thomas Zimmermann <tzimmermann@suse.de>
268
269Level: Intermediate
270
271drm_framebuffer_funcs and drm_mode_config_funcs.fb_create cleanup
272-----------------------------------------------------------------
273
274A lot more drivers could be switched over to the drm_gem_framebuffer helpers.
275Various hold-ups:
276
277- Need to switch over to the generic dirty tracking code using
278  drm_atomic_helper_dirtyfb first (e.g. qxl).
279
280- Need to switch to drm_fbdev_generic_setup(), otherwise a lot of the custom fb
281  setup code can't be deleted.
282
283- Need to switch to drm_gem_fb_create(), as now drm_gem_fb_create() checks for
284  valid formats for atomic drivers.
285
286- Many drivers subclass drm_framebuffer, we'd need a embedding compatible
287  version of the varios drm_gem_fb_create functions. Maybe called
288  drm_gem_fb_create/_with_dirty/_with_funcs as needed.
289
290Contact: Daniel Vetter
291
292Level: Intermediate
293
294Generic fbdev defio support
295---------------------------
296
297The defio support code in the fbdev core has some very specific requirements,
298which means drivers need to have a special framebuffer for fbdev. The main
299issue is that it uses some fields in struct page itself, which breaks shmem
300gem objects (and other things). To support defio, affected drivers require
301the use of a shadow buffer, which may add CPU and memory overhead.
302
303Possible solution would be to write our own defio mmap code in the drm fbdev
304emulation. It would need to fully wrap the existing mmap ops, forwarding
305everything after it has done the write-protect/mkwrite trickery:
306
307- In the drm_fbdev_fb_mmap helper, if we need defio, change the
308  default page prots to write-protected with something like this::
309
310      vma->vm_page_prot = pgprot_wrprotect(vma->vm_page_prot);
311
312- Set the mkwrite and fsync callbacks with similar implementions to the core
313  fbdev defio stuff. These should all work on plain ptes, they don't actually
314  require a struct page.  uff. These should all work on plain ptes, they don't
315  actually require a struct page.
316
317- Track the dirty pages in a separate structure (bitfield with one bit per page
318  should work) to avoid clobbering struct page.
319
320Might be good to also have some igt testcases for this.
321
322Contact: Daniel Vetter, Noralf Tronnes
323
324Level: Advanced
325
326connector register/unregister fixes
327-----------------------------------
328
329- For most connectors it's a no-op to call drm_connector_register/unregister
330  directly from driver code, drm_dev_register/unregister take care of this
331  already. We can remove all of them.
332
333- For dp drivers it's a bit more a mess, since we need the connector to be
334  registered when calling drm_dp_aux_register. Fix this by instead calling
335  drm_dp_aux_init, and moving the actual registering into a late_register
336  callback as recommended in the kerneldoc.
337
338Level: Intermediate
339
340Remove load/unload callbacks
341----------------------------
342
343The load/unload callbacks in struct &drm_driver are very much midlayers, plus
344for historical reasons they get the ordering wrong (and we can't fix that)
345between setting up the &drm_driver structure and calling drm_dev_register().
346
347- Rework drivers to no longer use the load/unload callbacks, directly coding the
348  load/unload sequence into the driver's probe function.
349
350- Once all drivers are converted, remove the load/unload callbacks.
351
352Contact: Daniel Vetter
353
354Level: Intermediate
355
356Replace drm_detect_hdmi_monitor() with drm_display_info.is_hdmi
357---------------------------------------------------------------
358
359Once EDID is parsed, the monitor HDMI support information is available through
360drm_display_info.is_hdmi. Many drivers still call drm_detect_hdmi_monitor() to
361retrieve the same information, which is less efficient.
362
363Audit each individual driver calling drm_detect_hdmi_monitor() and switch to
364drm_display_info.is_hdmi if applicable.
365
366Contact: Laurent Pinchart, respective driver maintainers
367
368Level: Intermediate
369
370Consolidate custom driver modeset properties
371--------------------------------------------
372
373Before atomic modeset took place, many drivers where creating their own
374properties. Among other things, atomic brought the requirement that custom,
375driver specific properties should not be used.
376
377For this task, we aim to introduce core helpers or reuse the existing ones
378if available:
379
380A quick, unconfirmed, examples list.
381
382Introduce core helpers:
383- audio (amdgpu, intel, gma500, radeon)
384- brightness, contrast, etc (armada, nouveau) - overlay only (?)
385- broadcast rgb (gma500, intel)
386- colorkey (armada, nouveau, rcar) - overlay only (?)
387- dither (amdgpu, nouveau, radeon) - varies across drivers
388- underscan family (amdgpu, radeon, nouveau)
389
390Already in core:
391- colorspace (sti)
392- tv format names, enhancements (gma500, intel)
393- tv overscan, margins, etc. (gma500, intel)
394- zorder (omapdrm) - same as zpos (?)
395
396
397Contact: Emil Velikov, respective driver maintainers
398
399Level: Intermediate
400
401Use struct iosys_map throughout codebase
402----------------------------------------
403
404Pointers to shared device memory are stored in struct iosys_map. Each
405instance knows whether it refers to system or I/O memory. Most of the DRM-wide
406interface have been converted to use struct iosys_map, but implementations
407often still use raw pointers.
408
409The task is to use struct iosys_map where it makes sense.
410
411* Memory managers should use struct iosys_map for dma-buf-imported buffers.
412* TTM might benefit from using struct iosys_map internally.
413* Framebuffer copying and blitting helpers should operate on struct iosys_map.
414
415Contact: Thomas Zimmermann <tzimmermann@suse.de>, Christian König, Daniel Vetter
416
417Level: Intermediate
418
419Review all drivers for setting struct drm_mode_config.{max_width,max_height} correctly
420--------------------------------------------------------------------------------------
421
422The values in struct drm_mode_config.{max_width,max_height} describe the
423maximum supported framebuffer size. It's the virtual screen size, but many
424drivers treat it like limitations of the physical resolution.
425
426The maximum width depends on the hardware's maximum scanline pitch. The
427maximum height depends on the amount of addressable video memory. Review all
428drivers to initialize the fields to the correct values.
429
430Contact: Thomas Zimmermann <tzimmermann@suse.de>
431
432Level: Intermediate
433
434Request memory regions in all drivers
435-------------------------------------
436
437Go through all drivers and add code to request the memory regions that the
438driver uses. This requires adding calls to request_mem_region(),
439pci_request_region() or similar functions. Use helpers for managed cleanup
440where possible.
441
442Drivers are pretty bad at doing this and there used to be conflicts among
443DRM and fbdev drivers. Still, it's the correct thing to do.
444
445Contact: Thomas Zimmermann <tzimmermann@suse.de>
446
447Level: Starter
448
449Remove driver dependencies on FB_DEVICE
450---------------------------------------
451
452A number of fbdev drivers provide attributes via sysfs and therefore depend
453on CONFIG_FB_DEVICE to be selected. Review each driver and attempt to make
454any dependencies on CONFIG_FB_DEVICE optional. At the minimum, the respective
455code in the driver could be conditionalized via ifdef CONFIG_FB_DEVICE. Not
456all drivers might be able to drop CONFIG_FB_DEVICE.
457
458Contact: Thomas Zimmermann <tzimmermann@suse.de>
459
460Level: Starter
461
462Clean up checks for already prepared/enabled in panels
463------------------------------------------------------
464
465In a whole pile of panel drivers, we have code to make the
466prepare/unprepare/enable/disable callbacks behave as no-ops if they've already
467been called. To get some idea of the duplicated code, try::
468
469  git grep 'if.*>prepared' -- drivers/gpu/drm/panel
470  git grep 'if.*>enabled' -- drivers/gpu/drm/panel
471
472In the patch ("drm/panel: Check for already prepared/enabled in drm_panel")
473we've moved this check to the core. Now we can most definitely remove the
474check from the individual panels and save a pile of code.
475
476In adition to removing the check from the individual panels, it is believed
477that even the core shouldn't need this check and that should be considered
478an error if other code ever relies on this check. The check in the core
479currently prints a warning whenever something is relying on this check with
480dev_warn(). After a little while, we likely want to promote this to a
481WARN(1) to help encourage folks not to rely on this behavior.
482
483Contact: Douglas Anderson <dianders@chromium.org>
484
485Level: Starter/Intermediate
486
487
488Core refactorings
489=================
490
491Make panic handling work
492------------------------
493
494This is a really varied tasks with lots of little bits and pieces:
495
496* The panic path can't be tested currently, leading to constant breaking. The
497  main issue here is that panics can be triggered from hardirq contexts and
498  hence all panic related callback can run in hardirq context. It would be
499  awesome if we could test at least the fbdev helper code and driver code by
500  e.g. trigger calls through drm debugfs files. hardirq context could be
501  achieved by using an IPI to the local processor.
502
503* There's a massive confusion of different panic handlers. DRM fbdev emulation
504  helpers had their own (long removed), but on top of that the fbcon code itself
505  also has one. We need to make sure that they stop fighting over each other.
506  This is worked around by checking ``oops_in_progress`` at various entry points
507  into the DRM fbdev emulation helpers. A much cleaner approach here would be to
508  switch fbcon to the `threaded printk support
509  <https://lwn.net/Articles/800946/>`_.
510
511* ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
512  isn't a full solution for panic paths. We need to make sure that it only
513  returns true if there's a panic going on for real, and fix up all the
514  fallout.
515
516* The panic handler must never sleep, which also means it can't ever
517  ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
518  even spinlocks (because NMI and hardirq can panic too). We need to either
519  make sure to not call such paths, or trylock everything. Really tricky.
520
521* A clean solution would be an entirely separate panic output support in KMS,
522  bypassing the current fbcon support. See `[PATCH v2 0/3] drm: Add panic handling
523  <https://lore.kernel.org/dri-devel/20190311174218.51899-1-noralf@tronnes.org/>`_.
524
525* Encoding the actual oops and preceding dmesg in a QR might help with the
526  dread "important stuff scrolled away" problem. See `[RFC][PATCH] Oops messages
527  transfer using QR codes
528  <https://lore.kernel.org/lkml/1446217392-11981-1-git-send-email-alexandru.murtaza@intel.com/>`_
529  for some example code that could be reused.
530
531Contact: Daniel Vetter
532
533Level: Advanced
534
535Clean up the debugfs support
536----------------------------
537
538There's a bunch of issues with it:
539
540- Convert drivers to support the drm_debugfs_add_files() function instead of
541  the drm_debugfs_create_files() function.
542
543- Improve late-register debugfs by rolling out the same debugfs pre-register
544  infrastructure for connector and crtc too. That way, the drivers won't need to
545  split their setup code into init and register anymore.
546
547- We probably want to have some support for debugfs files on crtc/connectors and
548  maybe other kms objects directly in core. There's even drm_print support in
549  the funcs for these objects to dump kms state, so it's all there. And then the
550  ->show() functions should obviously give you a pointer to the right object.
551
552- The drm_driver->debugfs_init hooks we have is just an artifact of the old
553  midlayered load sequence. DRM debugfs should work more like sysfs, where you
554  can create properties/files for an object anytime you want, and the core
555  takes care of publishing/unpuplishing all the files at register/unregister
556  time. Drivers shouldn't need to worry about these technicalities, and fixing
557  this (together with the drm_minor->drm_device move) would allow us to remove
558  debugfs_init.
559
560Contact: Daniel Vetter
561
562Level: Intermediate
563
564Object lifetime fixes
565---------------------
566
567There's two related issues here
568
569- Cleanup up the various ->destroy callbacks, which often are all the same
570  simple code.
571
572- Lots of drivers erroneously allocate DRM modeset objects using devm_kzalloc,
573  which results in use-after free issues on driver unload. This can be serious
574  trouble even for drivers for hardware integrated on the SoC due to
575  EPROBE_DEFERRED backoff.
576
577Both these problems can be solved by switching over to drmm_kzalloc(), and the
578various convenience wrappers provided, e.g. drmm_crtc_alloc_with_planes(),
579drmm_universal_plane_alloc(), ... and so on.
580
581Contact: Daniel Vetter
582
583Level: Intermediate
584
585Remove automatic page mapping from dma-buf importing
586----------------------------------------------------
587
588When importing dma-bufs, the dma-buf and PRIME frameworks automatically map
589imported pages into the importer's DMA area. drm_gem_prime_fd_to_handle() and
590drm_gem_prime_handle_to_fd() require that importers call dma_buf_attach()
591even if they never do actual device DMA, but only CPU access through
592dma_buf_vmap(). This is a problem for USB devices, which do not support DMA
593operations.
594
595To fix the issue, automatic page mappings should be removed from the
596buffer-sharing code. Fixing this is a bit more involved, since the import/export
597cache is also tied to &drm_gem_object.import_attach. Meanwhile we paper over
598this problem for USB devices by fishing out the USB host controller device, as
599long as that supports DMA. Otherwise importing can still needlessly fail.
600
601Contact: Thomas Zimmermann <tzimmermann@suse.de>, Daniel Vetter
602
603Level: Advanced
604
605
606Better Testing
607==============
608
609Add unit tests using the Kernel Unit Testing (KUnit) framework
610--------------------------------------------------------------
611
612The `KUnit <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
613provides a common framework for unit tests within the Linux kernel. Having a
614test suite would allow to identify regressions earlier.
615
616A good candidate for the first unit tests are the format-conversion helpers in
617``drm_format_helper.c``.
618
619Contact: Javier Martinez Canillas <javierm@redhat.com>
620
621Level: Intermediate
622
623Enable trinity for DRM
624----------------------
625
626And fix up the fallout. Should be really interesting ...
627
628Level: Advanced
629
630Make KMS tests in i-g-t generic
631-------------------------------
632
633The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
634including tons of testcases for corner-cases in the modesetting API. It would
635be awesome if those tests (at least the ones not relying on Intel-specific GEM
636features) could be made to run on any KMS driver.
637
638Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
639converting things over. For modeset tests we also first need a bit of
640infrastructure to use dumb buffers for untiled buffers, to be able to run all
641the non-i915 specific modeset tests.
642
643Level: Advanced
644
645Extend virtual test driver (VKMS)
646---------------------------------
647
648See the documentation of :ref:`VKMS <vkms>` for more details. This is an ideal
649internship task, since it only requires a virtual machine and can be sized to
650fit the available time.
651
652Level: See details
653
654Backlight Refactoring
655---------------------
656
657Backlight drivers have a triple enable/disable state, which is a bit overkill.
658Plan to fix this:
659
6601. Roll out backlight_enable() and backlight_disable() helpers everywhere. This
661   has started already.
6622. In all, only look at one of the three status bits set by the above helpers.
6633. Remove the other two status bits.
664
665Contact: Daniel Vetter
666
667Level: Intermediate
668
669Driver Specific
670===============
671
672AMD DC Display Driver
673---------------------
674
675AMD DC is the display driver for AMD devices starting with Vega. There has been
676a bunch of progress cleaning it up but there's still plenty of work to be done.
677
678See drivers/gpu/drm/amd/display/TODO for tasks.
679
680Contact: Harry Wentland, Alex Deucher
681
682Bootsplash
683==========
684
685There is support in place now for writing internal DRM clients making it
686possible to pick up the bootsplash work that was rejected because it was written
687for fbdev.
688
689- [v6,8/8] drm/client: Hack: Add bootsplash example
690  https://patchwork.freedesktop.org/patch/306579/
691
692- [RFC PATCH v2 00/13] Kernel based bootsplash
693  https://lore.kernel.org/r/20171213194755.3409-1-mstaudt@suse.de
694
695Contact: Sam Ravnborg
696
697Level: Advanced
698
699Brightness handling on devices with multiple internal panels
700============================================================
701
702On x86/ACPI devices there can be multiple backlight firmware interfaces:
703(ACPI) video, vendor specific and others. As well as direct/native (PWM)
704register programming by the KMS driver.
705
706To deal with this backlight drivers used on x86/ACPI call
707acpi_video_get_backlight_type() which has heuristics (+quirks) to select
708which backlight interface to use; and backlight drivers which do not match
709the returned type will not register themselves, so that only one backlight
710device gets registered (in a single GPU setup, see below).
711
712At the moment this more or less assumes that there will only
713be 1 (internal) panel on a system.
714
715On systems with 2 panels this may be a problem, depending on
716what interface acpi_video_get_backlight_type() selects:
717
7181. native: in this case the KMS driver is expected to know which backlight
719   device belongs to which output so everything should just work.
7202. video: this does support controlling multiple backlights, but some work
721   will need to be done to get the output <-> backlight device mapping
722
723The above assumes both panels will require the same backlight interface type.
724Things will break on systems with multiple panels where the 2 panels need
725a different type of control. E.g. one panel needs ACPI video backlight control,
726where as the other is using native backlight control. Currently in this case
727only one of the 2 required backlight devices will get registered, based on
728the acpi_video_get_backlight_type() return value.
729
730If this (theoretical) case ever shows up, then supporting this will need some
731work. A possible solution here would be to pass a device and connector-name
732to acpi_video_get_backlight_type() so that it can deal with this.
733
734Note in a way we already have a case where userspace sees 2 panels,
735in dual GPU laptop setups with a mux. On those systems we may see
736either 2 native backlight devices; or 2 native backlight devices.
737
738Userspace already has code to deal with this by detecting if the related
739panel is active (iow which way the mux between the GPU and the panels
740points) and then uses that backlight device. Userspace here very much
741assumes a single panel though. It picks only 1 of the 2 backlight devices
742and then only uses that one.
743
744Note that all userspace code (that I know off) is currently hardcoded
745to assume a single panel.
746
747Before the recent changes to not register multiple (e.g. video + native)
748/sys/class/backlight devices for a single panel (on a single GPU laptop),
749userspace would see multiple backlight devices all controlling the same
750backlight.
751
752To deal with this userspace had to always picks one preferred device under
753/sys/class/backlight and will ignore the others. So to support brightness
754control on multiple panels userspace will need to be updated too.
755
756There are plans to allow brightness control through the KMS API by adding
757a "display brightness" property to drm_connector objects for panels. This
758solves a number of issues with the /sys/class/backlight API, including not
759being able to map a sysfs backlight device to a specific connector. Any
760userspace changes to add support for brightness control on devices with
761multiple panels really should build on top of this new KMS property.
762
763Contact: Hans de Goede
764
765Level: Advanced
766
767Outside DRM
768===========
769
770Convert fbdev drivers to DRM
771----------------------------
772
773There are plenty of fbdev drivers for older hardware. Some hardware has
774become obsolete, but some still provides good(-enough) framebuffers. The
775drivers that are still useful should be converted to DRM and afterwards
776removed from fbdev.
777
778Very simple fbdev drivers can best be converted by starting with a new
779DRM driver. Simple KMS helpers and SHMEM should be able to handle any
780existing hardware. The new driver's call-back functions are filled from
781existing fbdev code.
782
783More complex fbdev drivers can be refactored step-by-step into a DRM
784driver with the help of the DRM fbconv helpers [4]_. These helpers provide
785the transition layer between the DRM core infrastructure and the fbdev
786driver interface. Create a new DRM driver on top of the fbconv helpers,
787copy over the fbdev driver, and hook it up to the DRM code. Examples for
788several fbdev drivers are available in Thomas Zimmermann's fbconv tree
789[4]_, as well as a tutorial of this process [5]_. The result is a primitive
790DRM driver that can run X11 and Weston.
791
792 .. [4] https://gitlab.freedesktop.org/tzimmermann/linux/tree/fbconv
793 .. [5] https://gitlab.freedesktop.org/tzimmermann/linux/blob/fbconv/drivers/gpu/drm/drm_fbconv_helper.c
794
795Contact: Thomas Zimmermann <tzimmermann@suse.de>
796
797Level: Advanced
798