xref: /linux/Documentation/gpu/todo.rst (revision f6e8dc9edf963dbc99085e54f6ced6da9daa6100)
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37implementation (there's lots of outdated locking leftovers in various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: Simona 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: Simona 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, Simona 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: Simona 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: Simona Vetter, respective driver maintainers
120
121Level: Advanced
122
123Rename drm_atomic_state
124-----------------------
125
126The KMS framework uses two slightly different definitions for the ``state``
127concept. For a given object (plane, CRTC, encoder, etc., so
128``drm_$OBJECT_state``), the state is the entire state of that object. However,
129at the device level, ``drm_atomic_state`` refers to a state update for a
130limited number of objects.
131
132The state isn't the entire device state, but only the full state of some
133objects in that device. This is confusing to newcomers, and
134``drm_atomic_state`` should be renamed to something clearer like
135``drm_atomic_commit``.
136
137In addition to renaming the structure itself, it would also imply renaming some
138related functions (``drm_atomic_state_alloc``, ``drm_atomic_state_get``,
139``drm_atomic_state_put``, ``drm_atomic_state_init``,
140``__drm_atomic_state_free``, etc.).
141
142Contact: Maxime Ripard <mripard@kernel.org>
143
144Level: Advanced
145
146Fallout from atomic KMS
147-----------------------
148
149``drm_atomic_helper.c`` provides a batch of functions which implement legacy
150IOCTLs on top of the new atomic driver interface. Which is really nice for
151gradual conversion of drivers, but unfortunately the semantic mismatches are
152a bit too severe. So there's some follow-up work to adjust the function
153interfaces to fix these issues:
154
155* atomic needs the lock acquire context. At the moment that's passed around
156  implicitly with some horrible hacks, and it's also allocate with
157  ``GFP_NOFAIL`` behind the scenes. All legacy paths need to start allocating
158  the acquire context explicitly on stack and then also pass it down into
159  drivers explicitly so that the legacy-on-atomic functions can use them.
160
161  Except for some driver code this is done. This task should be finished by
162  adding WARN_ON(!drm_drv_uses_atomic_modeset) in drm_modeset_lock_all().
163
164* A bunch of the vtable hooks are now in the wrong place: DRM has a split
165  between core vfunc tables (named ``drm_foo_funcs``), which are used to
166  implement the userspace ABI. And then there's the optional hooks for the
167  helper libraries (name ``drm_foo_helper_funcs``), which are purely for
168  internal use. Some of these hooks should be move from ``_funcs`` to
169  ``_helper_funcs`` since they are not part of the core ABI. There's a
170  ``FIXME`` comment in the kerneldoc for each such case in ``drm_crtc.h``.
171
172Contact: Simona Vetter
173
174Level: Intermediate
175
176Move Buffer Object Locking to dma_resv_lock()
177---------------------------------------------
178
179Many drivers have their own per-object locking scheme, usually using
180mutex_lock(). This causes all kinds of trouble for buffer sharing, since
181depending which driver is the exporter and importer, the locking hierarchy is
182reversed.
183
184To solve this we need one standard per-object locking mechanism, which is
185dma_resv_lock(). This lock needs to be called as the outermost lock, with all
186other driver specific per-object locks removed. The problem is that rolling out
187the actual change to the locking contract is a flag day, due to struct dma_buf
188buffer sharing.
189
190Level: Expert
191
192Convert logging to drm_* functions with drm_device parameter
193------------------------------------------------------------
194
195For drivers which could have multiple instances, it is necessary to
196differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
197don't do this, drivers used dev_info/warn/err to make this differentiation. We
198now have drm_* variants of the drm print functions, so we can start to convert
199those drivers back to using drm-formatted specific log messages.
200
201Before you start this conversion please contact the relevant maintainers to make
202sure your work will be merged - not everyone agrees that the DRM dmesg macros
203are better.
204
205Contact: Sean Paul, Maintainer of the driver you plan to convert
206
207Level: Starter
208
209Convert drivers to use simple modeset suspend/resume
210----------------------------------------------------
211
212Most drivers (except i915 and nouveau) that use
213drm_atomic_helper_suspend/resume() can probably be converted to use
214drm_mode_config_helper_suspend/resume(). Also there's still open-coded version
215of the atomic suspend/resume code in older atomic modeset drivers.
216
217Contact: Maintainer of the driver you plan to convert
218
219Level: Intermediate
220
221Reimplement functions in drm_fbdev_fb_ops without fbdev
222-------------------------------------------------------
223
224A number of callback functions in drm_fbdev_fb_ops could benefit from
225being rewritten without dependencies on the fbdev module. Some of the
226helpers could further benefit from using struct iosys_map instead of
227raw pointers.
228
229Contact: Thomas Zimmermann <tzimmermann@suse.de>, Simona Vetter
230
231Level: Advanced
232
233Benchmark and optimize blitting and format-conversion function
234--------------------------------------------------------------
235
236Drawing to display memory quickly is crucial for many applications'
237performance.
238
239On at least x86-64, sys_imageblit() is significantly slower than
240cfb_imageblit(), even though both use the same blitting algorithm and
241the latter is written for I/O memory. It turns out that cfb_imageblit()
242uses movl instructions, while sys_imageblit apparently does not. This
243seems to be a problem with gcc's optimizer. DRM's format-conversion
244helpers might be subject to similar issues.
245
246Benchmark and optimize fbdev's sys_() helpers and DRM's format-conversion
247helpers. In cases that can be further optimized, maybe implement a different
248algorithm. For micro-optimizations, use movl/movq instructions explicitly.
249That might possibly require architecture-specific helpers (e.g., storel()
250storeq()).
251
252Contact: Thomas Zimmermann <tzimmermann@suse.de>
253
254Level: Intermediate
255
256drm_framebuffer_funcs and drm_mode_config_funcs.fb_create cleanup
257-----------------------------------------------------------------
258
259A lot more drivers could be switched over to the drm_gem_framebuffer helpers.
260Various hold-ups:
261
262- Need to switch over to the generic dirty tracking code using
263  drm_atomic_helper_dirtyfb first (e.g. qxl).
264
265- Need to switch to drm_fbdev_generic_setup(), otherwise a lot of the custom fb
266  setup code can't be deleted.
267
268- Need to switch to drm_gem_fb_create(), as now drm_gem_fb_create() checks for
269  valid formats for atomic drivers.
270
271- Many drivers subclass drm_framebuffer, we'd need a embedding compatible
272  version of the varios drm_gem_fb_create functions. Maybe called
273  drm_gem_fb_create/_with_dirty/_with_funcs as needed.
274
275Contact: Simona Vetter
276
277Level: Intermediate
278
279Generic fbdev defio support
280---------------------------
281
282The defio support code in the fbdev core has some very specific requirements,
283which means drivers need to have a special framebuffer for fbdev. The main
284issue is that it uses some fields in struct page itself, which breaks shmem
285gem objects (and other things). To support defio, affected drivers require
286the use of a shadow buffer, which may add CPU and memory overhead.
287
288Possible solution would be to write our own defio mmap code in the drm fbdev
289emulation. It would need to fully wrap the existing mmap ops, forwarding
290everything after it has done the write-protect/mkwrite trickery:
291
292- In the drm_fbdev_fb_mmap helper, if we need defio, change the
293  default page prots to write-protected with something like this::
294
295      vma->vm_page_prot = pgprot_wrprotect(vma->vm_page_prot);
296
297- Set the mkwrite and fsync callbacks with similar implementions to the core
298  fbdev defio stuff. These should all work on plain ptes, they don't actually
299  require a struct page.  uff. These should all work on plain ptes, they don't
300  actually require a struct page.
301
302- Track the dirty pages in a separate structure (bitfield with one bit per page
303  should work) to avoid clobbering struct page.
304
305Might be good to also have some igt testcases for this.
306
307Contact: Simona Vetter, Noralf Tronnes
308
309Level: Advanced
310
311connector register/unregister fixes
312-----------------------------------
313
314- For most connectors it's a no-op to call drm_connector_register/unregister
315  directly from driver code, drm_dev_register/unregister take care of this
316  already. We can remove all of them.
317
318- For dp drivers it's a bit more a mess, since we need the connector to be
319  registered when calling drm_dp_aux_register. Fix this by instead calling
320  drm_dp_aux_init, and moving the actual registering into a late_register
321  callback as recommended in the kerneldoc.
322
323Level: Intermediate
324
325Remove load/unload callbacks
326----------------------------
327
328The load/unload callbacks in struct &drm_driver are very much midlayers, plus
329for historical reasons they get the ordering wrong (and we can't fix that)
330between setting up the &drm_driver structure and calling drm_dev_register().
331
332- Rework drivers to no longer use the load/unload callbacks, directly coding the
333  load/unload sequence into the driver's probe function.
334
335- Once all drivers are converted, remove the load/unload callbacks.
336
337Contact: Simona Vetter
338
339Level: Intermediate
340
341Replace drm_detect_hdmi_monitor() with drm_display_info.is_hdmi
342---------------------------------------------------------------
343
344Once EDID is parsed, the monitor HDMI support information is available through
345drm_display_info.is_hdmi. Many drivers still call drm_detect_hdmi_monitor() to
346retrieve the same information, which is less efficient.
347
348Audit each individual driver calling drm_detect_hdmi_monitor() and switch to
349drm_display_info.is_hdmi if applicable.
350
351Contact: Laurent Pinchart, respective driver maintainers
352
353Level: Intermediate
354
355Consolidate custom driver modeset properties
356--------------------------------------------
357
358Before atomic modeset took place, many drivers where creating their own
359properties. Among other things, atomic brought the requirement that custom,
360driver specific properties should not be used.
361
362For this task, we aim to introduce core helpers or reuse the existing ones
363if available:
364
365A quick, unconfirmed, examples list.
366
367Introduce core helpers:
368- audio (amdgpu, intel, gma500, radeon)
369- brightness, contrast, etc (armada, nouveau) - overlay only (?)
370- broadcast rgb (gma500, intel)
371- colorkey (armada, nouveau, rcar) - overlay only (?)
372- dither (amdgpu, nouveau, radeon) - varies across drivers
373- underscan family (amdgpu, radeon, nouveau)
374
375Already in core:
376- colorspace (sti)
377- tv format names, enhancements (gma500, intel)
378- tv overscan, margins, etc. (gma500, intel)
379- zorder (omapdrm) - same as zpos (?)
380
381
382Contact: Emil Velikov, respective driver maintainers
383
384Level: Intermediate
385
386Use struct iosys_map throughout codebase
387----------------------------------------
388
389Pointers to shared device memory are stored in struct iosys_map. Each
390instance knows whether it refers to system or I/O memory. Most of the DRM-wide
391interface have been converted to use struct iosys_map, but implementations
392often still use raw pointers.
393
394The task is to use struct iosys_map where it makes sense.
395
396* Memory managers should use struct iosys_map for dma-buf-imported buffers.
397* TTM might benefit from using struct iosys_map internally.
398* Framebuffer copying and blitting helpers should operate on struct iosys_map.
399
400Contact: Thomas Zimmermann <tzimmermann@suse.de>, Christian König, Simona Vetter
401
402Level: Intermediate
403
404Review all drivers for setting struct drm_mode_config.{max_width,max_height} correctly
405--------------------------------------------------------------------------------------
406
407The values in struct drm_mode_config.{max_width,max_height} describe the
408maximum supported framebuffer size. It's the virtual screen size, but many
409drivers treat it like limitations of the physical resolution.
410
411The maximum width depends on the hardware's maximum scanline pitch. The
412maximum height depends on the amount of addressable video memory. Review all
413drivers to initialize the fields to the correct values.
414
415Contact: Thomas Zimmermann <tzimmermann@suse.de>
416
417Level: Intermediate
418
419Request memory regions in all fbdev drivers
420--------------------------------------------
421
422Old/ancient fbdev drivers do not request their memory properly.
423Go through these drivers and add code to request the memory regions
424that the driver uses. This requires adding calls to request_mem_region(),
425pci_request_region() or similar functions. Use helpers for managed cleanup
426where possible. Problematic areas include hardware that has exclusive ranges
427like VGA. VGA16fb does not request the range as it is expected.
428Drivers are pretty bad at doing this and there used to be conflicts among
429DRM and fbdev drivers. Still, it's the correct thing to do.
430
431Contact: Thomas Zimmermann <tzimmermann@suse.de>
432
433Level: Starter
434
435Remove driver dependencies on FB_DEVICE
436---------------------------------------
437
438A number of fbdev drivers provide attributes via sysfs and therefore depend
439on CONFIG_FB_DEVICE to be selected. Review each driver and attempt to make
440any dependencies on CONFIG_FB_DEVICE optional. At the minimum, the respective
441code in the driver could be conditionalized via ifdef CONFIG_FB_DEVICE. Not
442all drivers might be able to drop CONFIG_FB_DEVICE.
443
444Contact: Thomas Zimmermann <tzimmermann@suse.de>
445
446Level: Starter
447
448Remove disable/unprepare in remove/shutdown in panel-simple and panel-edp
449-------------------------------------------------------------------------
450
451As of commit d2aacaf07395 ("drm/panel: Check for already prepared/enabled in
452drm_panel"), we have a check in the drm_panel core to make sure nobody
453double-calls prepare/enable/disable/unprepare. Eventually that should probably
454be turned into a WARN_ON() or somehow made louder.
455
456At the moment, we expect that we may still encounter the warnings in the
457drm_panel core when using panel-simple and panel-edp. Since those panel
458drivers are used with a lot of different DRM modeset drivers they still
459make an extra effort to disable/unprepare the panel themsevles at shutdown
460time. Specifically we could still encounter those warnings if the panel
461driver gets shutdown() _before_ the DRM modeset driver and the DRM modeset
462driver properly calls drm_atomic_helper_shutdown() in its own shutdown()
463callback. Warnings could be avoided in such a case by using something like
464device links to ensure that the panel gets shutdown() after the DRM modeset
465driver.
466
467Once all DRM modeset drivers are known to shutdown properly, the extra
468calls to disable/unprepare in remove/shutdown in panel-simple and panel-edp
469should be removed and this TODO item marked complete.
470
471Contact: Douglas Anderson <dianders@chromium.org>
472
473Level: Intermediate
474
475Transition away from using deprecated MIPI DSI functions
476--------------------------------------------------------
477
478There are many functions defined in ``drm_mipi_dsi.c`` which have been
479deprecated. Each deprecated function was deprecated in favor of its `multi`
480variant (e.g. `mipi_dsi_generic_write()` and `mipi_dsi_generic_write_multi()`).
481The `multi` variant of a function includes improved error handling and logic
482which makes it more convenient to make several calls in a row, as most MIPI
483drivers do.
484
485Drivers should be updated to use undeprecated functions. Once all usages of the
486deprecated MIPI DSI functions have been removed, their definitions may be
487removed from ``drm_mipi_dsi.c``.
488
489Contact: Douglas Anderson <dianders@chromium.org>
490
491Level: Starter
492
493Remove devm_drm_put_bridge()
494----------------------------
495
496Due to how the panel bridge handles the drm_bridge object lifetime, special
497care must be taken to dispose of the drm_bridge object when the
498panel_bridge is removed. This is currently managed using
499devm_drm_put_bridge(), but that is an unsafe, temporary workaround. To fix
500that, the DRM panel lifetime needs to be reworked. After the rework is
501done, remove devm_drm_put_bridge() and the TODO in
502drm_panel_bridge_remove().
503
504Contact: Maxime Ripard <mripard@kernel.org>,
505         Luca Ceresoli <luca.ceresoli@bootlin.com>
506
507Level: Intermediate
508
509Core refactorings
510=================
511
512Make panic handling work
513------------------------
514
515This is a really varied tasks with lots of little bits and pieces:
516
517* The panic path can't be tested currently, leading to constant breaking. The
518  main issue here is that panics can be triggered from hardirq contexts and
519  hence all panic related callback can run in hardirq context. It would be
520  awesome if we could test at least the fbdev helper code and driver code by
521  e.g. trigger calls through drm debugfs files. hardirq context could be
522  achieved by using an IPI to the local processor.
523
524* There's a massive confusion of different panic handlers. DRM fbdev emulation
525  helpers had their own (long removed), but on top of that the fbcon code itself
526  also has one. We need to make sure that they stop fighting over each other.
527  This is worked around by checking ``oops_in_progress`` at various entry points
528  into the DRM fbdev emulation helpers. A much cleaner approach here would be to
529  switch fbcon to the `threaded printk support
530  <https://lwn.net/Articles/800946/>`_.
531
532* ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
533  isn't a full solution for panic paths. We need to make sure that it only
534  returns true if there's a panic going on for real, and fix up all the
535  fallout.
536
537* The panic handler must never sleep, which also means it can't ever
538  ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
539  even spinlocks (because NMI and hardirq can panic too). We need to either
540  make sure to not call such paths, or trylock everything. Really tricky.
541
542* A clean solution would be an entirely separate panic output support in KMS,
543  bypassing the current fbcon support. See `[PATCH v2 0/3] drm: Add panic handling
544  <https://lore.kernel.org/dri-devel/20190311174218.51899-1-noralf@tronnes.org/>`_.
545
546* Encoding the actual oops and preceding dmesg in a QR might help with the
547  dread "important stuff scrolled away" problem. See `[RFC][PATCH] Oops messages
548  transfer using QR codes
549  <https://lore.kernel.org/lkml/1446217392-11981-1-git-send-email-alexandru.murtaza@intel.com/>`_
550  for some example code that could be reused.
551
552Contact: Simona Vetter
553
554Level: Advanced
555
556Clean up the debugfs support
557----------------------------
558
559There's a bunch of issues with it:
560
561- Convert drivers to support the drm_debugfs_add_files() function instead of
562  the drm_debugfs_create_files() function.
563
564- Improve late-register debugfs by rolling out the same debugfs pre-register
565  infrastructure for connector and crtc too. That way, the drivers won't need to
566  split their setup code into init and register anymore.
567
568- We probably want to have some support for debugfs files on crtc/connectors and
569  maybe other kms objects directly in core. There's even drm_print support in
570  the funcs for these objects to dump kms state, so it's all there. And then the
571  ->show() functions should obviously give you a pointer to the right object.
572
573- The drm_driver->debugfs_init hooks we have is just an artifact of the old
574  midlayered load sequence. DRM debugfs should work more like sysfs, where you
575  can create properties/files for an object anytime you want, and the core
576  takes care of publishing/unpuplishing all the files at register/unregister
577  time. Drivers shouldn't need to worry about these technicalities, and fixing
578  this (together with the drm_minor->drm_device move) would allow us to remove
579  debugfs_init.
580
581Contact: Simona Vetter
582
583Level: Intermediate
584
585Object lifetime fixes
586---------------------
587
588There's two related issues here
589
590- Cleanup up the various ->destroy callbacks, which often are all the same
591  simple code.
592
593- Lots of drivers erroneously allocate DRM modeset objects using devm_kzalloc,
594  which results in use-after free issues on driver unload. This can be serious
595  trouble even for drivers for hardware integrated on the SoC due to
596  EPROBE_DEFERRED backoff.
597
598Both these problems can be solved by switching over to drmm_kzalloc(), and the
599various convenience wrappers provided, e.g. drmm_crtc_alloc_with_planes(),
600drmm_universal_plane_alloc(), ... and so on.
601
602Contact: Simona Vetter
603
604Level: Intermediate
605
606Remove automatic page mapping from dma-buf importing
607----------------------------------------------------
608
609When importing dma-bufs, the dma-buf and PRIME frameworks automatically map
610imported pages into the importer's DMA area. drm_gem_prime_fd_to_handle() and
611drm_gem_prime_handle_to_fd() require that importers call dma_buf_attach()
612even if they never do actual device DMA, but only CPU access through
613dma_buf_vmap(). This is a problem for USB devices, which do not support DMA
614operations.
615
616To fix the issue, automatic page mappings should be removed from the
617buffer-sharing code. Fixing this is a bit more involved, since the import/export
618cache is also tied to &drm_gem_object.import_attach. Meanwhile we paper over
619this problem for USB devices by fishing out the USB host controller device, as
620long as that supports DMA. Otherwise importing can still needlessly fail.
621
622Contact: Thomas Zimmermann <tzimmermann@suse.de>, Simona Vetter
623
624Level: Advanced
625
626Implement a new DUMB_CREATE2 ioctl
627----------------------------------
628
629The current DUMB_CREATE ioctl is not well defined. Instead of a pixel and
630framebuffer format, it only accepts a color mode of vague semantics. Assuming
631a linear framebuffer, the color mode gives an idea of the supported pixel
632format. But userspace effectively has to guess the correct values. It really
633only works reliably with framebuffers in XRGB8888. Userspace has begun to
634workaround these limitations by computing arbitrary format's buffer sizes and
635calculating their sizes in terms of XRGB8888 pixels.
636
637One possible solution is a new ioctl DUMB_CREATE2. It should accept a DRM
638format and a format modifier to resolve the color mode's ambiguity. As
639framebuffers can be multi-planar, the new ioctl has to return the buffer size,
640pitch and GEM handle for each individual color plane.
641
642In the first step, the new ioctl can be limited to the current features of
643the existing DUMB_CREATE. Individual drivers can then be extended to support
644multi-planar formats. Rockchip might require this and would be a good candidate.
645
646It might also be helpful to userspace to query information about the size of
647a potential buffer, if allocated. Userspace would supply geometry and format;
648the kernel would return minimal allocation sizes and scanline pitch. There is
649interest to allocate that memory from another device and provide it to the
650DRM driver (say via dma-buf).
651
652Another requested feature is the ability to allocate a buffer by size, without
653format. Accelators use this for their buffer allocation and it could likely be
654generalized.
655
656In addition to the kernel implementation, there must be user-space support
657for the new ioctl. There's code in Mesa that might be able to use the new
658call.
659
660Contact: Thomas Zimmermann <tzimmermann@suse.de>
661
662Level: Advanced
663
664Better Testing
665==============
666
667Add unit tests using the Kernel Unit Testing (KUnit) framework
668--------------------------------------------------------------
669
670The `KUnit <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
671provides a common framework for unit tests within the Linux kernel. Having a
672test suite would allow to identify regressions earlier.
673
674A good candidate for the first unit tests are the format-conversion helpers in
675``drm_format_helper.c``.
676
677Contact: Javier Martinez Canillas <javierm@redhat.com>
678
679Level: Intermediate
680
681Clean up and document former selftests suites
682---------------------------------------------
683
684Some KUnit test suites (drm_buddy, drm_cmdline_parser, drm_damage_helper,
685drm_format, drm_framebuffer, drm_dp_mst_helper, drm_mm, drm_plane_helper and
686drm_rect) are former selftests suites that have been converted over when KUnit
687was first introduced.
688
689These suites were fairly undocumented, and with different goals than what unit
690tests can be. Trying to identify what each test in these suites actually test
691for, whether that makes sense for a unit test, and either remove it if it
692doesn't or document it if it does would be of great help.
693
694Contact: Maxime Ripard <mripard@kernel.org>
695
696Level: Intermediate
697
698Enable trinity for DRM
699----------------------
700
701And fix up the fallout. Should be really interesting ...
702
703Level: Advanced
704
705Make KMS tests in i-g-t generic
706-------------------------------
707
708The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
709including tons of testcases for corner-cases in the modesetting API. It would
710be awesome if those tests (at least the ones not relying on Intel-specific GEM
711features) could be made to run on any KMS driver.
712
713Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
714converting things over. For modeset tests we also first need a bit of
715infrastructure to use dumb buffers for untiled buffers, to be able to run all
716the non-i915 specific modeset tests.
717
718Level: Advanced
719
720Extend virtual test driver (VKMS)
721---------------------------------
722
723See the documentation of :ref:`VKMS <vkms>` for more details. This is an ideal
724internship task, since it only requires a virtual machine and can be sized to
725fit the available time.
726
727Level: See details
728
729Backlight Refactoring
730---------------------
731
732Backlight drivers have a triple enable/disable state, which is a bit overkill.
733Plan to fix this:
734
7351. Roll out backlight_enable() and backlight_disable() helpers everywhere. This
736   has started already.
7372. In all, only look at one of the three status bits set by the above helpers.
7383. Remove the other two status bits.
739
740Contact: Simona Vetter
741
742Level: Intermediate
743
744Driver Specific
745===============
746
747AMD DC Display Driver
748---------------------
749
750AMD DC is the display driver for AMD devices starting with Vega. There has been
751a bunch of progress cleaning it up but there's still plenty of work to be done.
752
753See drivers/gpu/drm/amd/display/TODO for tasks.
754
755Contact: Harry Wentland, Alex Deucher
756
757Bootsplash
758==========
759
760There is support in place now for writing internal DRM clients making it
761possible to pick up the bootsplash work that was rejected because it was written
762for fbdev.
763
764- [v6,8/8] drm/client: Hack: Add bootsplash example
765  https://patchwork.freedesktop.org/patch/306579/
766
767- [RFC PATCH v2 00/13] Kernel based bootsplash
768  https://lore.kernel.org/r/20171213194755.3409-1-mstaudt@suse.de
769
770Contact: Sam Ravnborg
771
772Level: Advanced
773
774Brightness handling on devices with multiple internal panels
775============================================================
776
777On x86/ACPI devices there can be multiple backlight firmware interfaces:
778(ACPI) video, vendor specific and others. As well as direct/native (PWM)
779register programming by the KMS driver.
780
781To deal with this backlight drivers used on x86/ACPI call
782acpi_video_get_backlight_type() which has heuristics (+quirks) to select
783which backlight interface to use; and backlight drivers which do not match
784the returned type will not register themselves, so that only one backlight
785device gets registered (in a single GPU setup, see below).
786
787At the moment this more or less assumes that there will only
788be 1 (internal) panel on a system.
789
790On systems with 2 panels this may be a problem, depending on
791what interface acpi_video_get_backlight_type() selects:
792
7931. native: in this case the KMS driver is expected to know which backlight
794   device belongs to which output so everything should just work.
7952. video: this does support controlling multiple backlights, but some work
796   will need to be done to get the output <-> backlight device mapping
797
798The above assumes both panels will require the same backlight interface type.
799Things will break on systems with multiple panels where the 2 panels need
800a different type of control. E.g. one panel needs ACPI video backlight control,
801where as the other is using native backlight control. Currently in this case
802only one of the 2 required backlight devices will get registered, based on
803the acpi_video_get_backlight_type() return value.
804
805If this (theoretical) case ever shows up, then supporting this will need some
806work. A possible solution here would be to pass a device and connector-name
807to acpi_video_get_backlight_type() so that it can deal with this.
808
809Note in a way we already have a case where userspace sees 2 panels,
810in dual GPU laptop setups with a mux. On those systems we may see
811either 2 native backlight devices; or 2 native backlight devices.
812
813Userspace already has code to deal with this by detecting if the related
814panel is active (iow which way the mux between the GPU and the panels
815points) and then uses that backlight device. Userspace here very much
816assumes a single panel though. It picks only 1 of the 2 backlight devices
817and then only uses that one.
818
819Note that all userspace code (that I know off) is currently hardcoded
820to assume a single panel.
821
822Before the recent changes to not register multiple (e.g. video + native)
823/sys/class/backlight devices for a single panel (on a single GPU laptop),
824userspace would see multiple backlight devices all controlling the same
825backlight.
826
827To deal with this userspace had to always picks one preferred device under
828/sys/class/backlight and will ignore the others. So to support brightness
829control on multiple panels userspace will need to be updated too.
830
831There are plans to allow brightness control through the KMS API by adding
832a "display brightness" property to drm_connector objects for panels. This
833solves a number of issues with the /sys/class/backlight API, including not
834being able to map a sysfs backlight device to a specific connector. Any
835userspace changes to add support for brightness control on devices with
836multiple panels really should build on top of this new KMS property.
837
838Contact: Hans de Goede
839
840Level: Advanced
841
842Buffer age or other damage accumulation algorithm for buffer damage
843===================================================================
844
845Drivers that do per-buffer uploads, need a buffer damage handling (rather than
846frame damage like drivers that do per-plane or per-CRTC uploads), but there is
847no support to get the buffer age or any other damage accumulation algorithm.
848
849For this reason, the damage helpers just fallback to a full plane update if the
850framebuffer attached to a plane has changed since the last page-flip. Drivers
851set &drm_plane_state.ignore_damage_clips to true as indication to
852drm_atomic_helper_damage_iter_init() and drm_atomic_helper_damage_iter_next()
853helpers that the damage clips should be ignored.
854
855This should be improved to get damage tracking properly working on drivers that
856do per-buffer uploads.
857
858More information about damage tracking and references to learning materials can
859be found in :ref:`damage_tracking_properties`.
860
861Contact: Javier Martinez Canillas <javierm@redhat.com>
862
863Level: Advanced
864
865Querying errors from drm_syncobj
866================================
867
868The drm_syncobj container can be used by driver independent code to signal
869complection of submission.
870
871One minor feature still missing is a generic DRM IOCTL to query the error
872status of binary and timeline drm_syncobj.
873
874This should probably be improved by implementing the necessary kernel interface
875and adding support for that in the userspace stack.
876
877Contact: Christian König
878
879Level: Starter
880
881Outside DRM
882===========
883
884Convert fbdev drivers to DRM
885----------------------------
886
887There are plenty of fbdev drivers for older hardware. Some hardware has
888become obsolete, but some still provides good(-enough) framebuffers. The
889drivers that are still useful should be converted to DRM and afterwards
890removed from fbdev.
891
892Very simple fbdev drivers can best be converted by starting with a new
893DRM driver. Simple KMS helpers and SHMEM should be able to handle any
894existing hardware. The new driver's call-back functions are filled from
895existing fbdev code.
896
897More complex fbdev drivers can be refactored step-by-step into a DRM
898driver with the help of the DRM fbconv helpers [4]_. These helpers provide
899the transition layer between the DRM core infrastructure and the fbdev
900driver interface. Create a new DRM driver on top of the fbconv helpers,
901copy over the fbdev driver, and hook it up to the DRM code. Examples for
902several fbdev drivers are available in Thomas Zimmermann's fbconv tree
903[4]_, as well as a tutorial of this process [5]_. The result is a primitive
904DRM driver that can run X11 and Weston.
905
906 .. [4] https://gitlab.freedesktop.org/tzimmermann/linux/tree/fbconv
907 .. [5] https://gitlab.freedesktop.org/tzimmermann/linux/blob/fbconv/drivers/gpu/drm/drm_fbconv_helper.c
908
909Contact: Thomas Zimmermann <tzimmermann@suse.de>
910
911Level: Advanced
912