xref: /linux/Documentation/gpu/todo.rst (revision ab93e0dd72c37d378dd936f031ffb83ff2bd87ce)
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37implementation (there's lots of outdated locking leftovers in various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
176Get rid of dev->struct_mutex from GEM drivers
177---------------------------------------------
178
179``dev->struct_mutex`` is the Big DRM Lock from legacy days and infested
180everything. Nowadays in modern drivers the only bit where it's mandatory is
181serializing GEM buffer object destruction. Which unfortunately means drivers
182have to keep track of that lock and either call ``unreference`` or
183``unreference_locked`` depending upon context.
184
185Core GEM doesn't have a need for ``struct_mutex`` any more since kernel 4.8,
186and there's a GEM object ``free`` callback for any drivers which are
187entirely ``struct_mutex`` free.
188
189For drivers that need ``struct_mutex`` it should be replaced with a driver-
190private lock. The tricky part is the BO free functions, since those can't
191reliably take that lock any more. Instead state needs to be protected with
192suitable subordinate locks or some cleanup work pushed to a worker thread. For
193performance-critical drivers it might also be better to go with a more
194fine-grained per-buffer object and per-context lockings scheme. Currently only
195the ``msm`` and `i915` drivers use ``struct_mutex``.
196
197Contact: Simona Vetter, respective driver maintainers
198
199Level: Advanced
200
201Move Buffer Object Locking to dma_resv_lock()
202---------------------------------------------
203
204Many drivers have their own per-object locking scheme, usually using
205mutex_lock(). This causes all kinds of trouble for buffer sharing, since
206depending which driver is the exporter and importer, the locking hierarchy is
207reversed.
208
209To solve this we need one standard per-object locking mechanism, which is
210dma_resv_lock(). This lock needs to be called as the outermost lock, with all
211other driver specific per-object locks removed. The problem is that rolling out
212the actual change to the locking contract is a flag day, due to struct dma_buf
213buffer sharing.
214
215Level: Expert
216
217Convert logging to drm_* functions with drm_device parameter
218------------------------------------------------------------
219
220For drivers which could have multiple instances, it is necessary to
221differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
222don't do this, drivers used dev_info/warn/err to make this differentiation. We
223now have drm_* variants of the drm print functions, so we can start to convert
224those drivers back to using drm-formatted specific log messages.
225
226Before you start this conversion please contact the relevant maintainers to make
227sure your work will be merged - not everyone agrees that the DRM dmesg macros
228are better.
229
230Contact: Sean Paul, Maintainer of the driver you plan to convert
231
232Level: Starter
233
234Convert drivers to use simple modeset suspend/resume
235----------------------------------------------------
236
237Most drivers (except i915 and nouveau) that use
238drm_atomic_helper_suspend/resume() can probably be converted to use
239drm_mode_config_helper_suspend/resume(). Also there's still open-coded version
240of the atomic suspend/resume code in older atomic modeset drivers.
241
242Contact: Maintainer of the driver you plan to convert
243
244Level: Intermediate
245
246Reimplement functions in drm_fbdev_fb_ops without fbdev
247-------------------------------------------------------
248
249A number of callback functions in drm_fbdev_fb_ops could benefit from
250being rewritten without dependencies on the fbdev module. Some of the
251helpers could further benefit from using struct iosys_map instead of
252raw pointers.
253
254Contact: Thomas Zimmermann <tzimmermann@suse.de>, Simona Vetter
255
256Level: Advanced
257
258Benchmark and optimize blitting and format-conversion function
259--------------------------------------------------------------
260
261Drawing to display memory quickly is crucial for many applications'
262performance.
263
264On at least x86-64, sys_imageblit() is significantly slower than
265cfb_imageblit(), even though both use the same blitting algorithm and
266the latter is written for I/O memory. It turns out that cfb_imageblit()
267uses movl instructions, while sys_imageblit apparently does not. This
268seems to be a problem with gcc's optimizer. DRM's format-conversion
269helpers might be subject to similar issues.
270
271Benchmark and optimize fbdev's sys_() helpers and DRM's format-conversion
272helpers. In cases that can be further optimized, maybe implement a different
273algorithm. For micro-optimizations, use movl/movq instructions explicitly.
274That might possibly require architecture-specific helpers (e.g., storel()
275storeq()).
276
277Contact: Thomas Zimmermann <tzimmermann@suse.de>
278
279Level: Intermediate
280
281drm_framebuffer_funcs and drm_mode_config_funcs.fb_create cleanup
282-----------------------------------------------------------------
283
284A lot more drivers could be switched over to the drm_gem_framebuffer helpers.
285Various hold-ups:
286
287- Need to switch over to the generic dirty tracking code using
288  drm_atomic_helper_dirtyfb first (e.g. qxl).
289
290- Need to switch to drm_fbdev_generic_setup(), otherwise a lot of the custom fb
291  setup code can't be deleted.
292
293- Need to switch to drm_gem_fb_create(), as now drm_gem_fb_create() checks for
294  valid formats for atomic drivers.
295
296- Many drivers subclass drm_framebuffer, we'd need a embedding compatible
297  version of the varios drm_gem_fb_create functions. Maybe called
298  drm_gem_fb_create/_with_dirty/_with_funcs as needed.
299
300Contact: Simona Vetter
301
302Level: Intermediate
303
304Generic fbdev defio support
305---------------------------
306
307The defio support code in the fbdev core has some very specific requirements,
308which means drivers need to have a special framebuffer for fbdev. The main
309issue is that it uses some fields in struct page itself, which breaks shmem
310gem objects (and other things). To support defio, affected drivers require
311the use of a shadow buffer, which may add CPU and memory overhead.
312
313Possible solution would be to write our own defio mmap code in the drm fbdev
314emulation. It would need to fully wrap the existing mmap ops, forwarding
315everything after it has done the write-protect/mkwrite trickery:
316
317- In the drm_fbdev_fb_mmap helper, if we need defio, change the
318  default page prots to write-protected with something like this::
319
320      vma->vm_page_prot = pgprot_wrprotect(vma->vm_page_prot);
321
322- Set the mkwrite and fsync callbacks with similar implementions to the core
323  fbdev defio stuff. These should all work on plain ptes, they don't actually
324  require a struct page.  uff. These should all work on plain ptes, they don't
325  actually require a struct page.
326
327- Track the dirty pages in a separate structure (bitfield with one bit per page
328  should work) to avoid clobbering struct page.
329
330Might be good to also have some igt testcases for this.
331
332Contact: Simona Vetter, Noralf Tronnes
333
334Level: Advanced
335
336connector register/unregister fixes
337-----------------------------------
338
339- For most connectors it's a no-op to call drm_connector_register/unregister
340  directly from driver code, drm_dev_register/unregister take care of this
341  already. We can remove all of them.
342
343- For dp drivers it's a bit more a mess, since we need the connector to be
344  registered when calling drm_dp_aux_register. Fix this by instead calling
345  drm_dp_aux_init, and moving the actual registering into a late_register
346  callback as recommended in the kerneldoc.
347
348Level: Intermediate
349
350Remove load/unload callbacks
351----------------------------
352
353The load/unload callbacks in struct &drm_driver are very much midlayers, plus
354for historical reasons they get the ordering wrong (and we can't fix that)
355between setting up the &drm_driver structure and calling drm_dev_register().
356
357- Rework drivers to no longer use the load/unload callbacks, directly coding the
358  load/unload sequence into the driver's probe function.
359
360- Once all drivers are converted, remove the load/unload callbacks.
361
362Contact: Simona Vetter
363
364Level: Intermediate
365
366Replace drm_detect_hdmi_monitor() with drm_display_info.is_hdmi
367---------------------------------------------------------------
368
369Once EDID is parsed, the monitor HDMI support information is available through
370drm_display_info.is_hdmi. Many drivers still call drm_detect_hdmi_monitor() to
371retrieve the same information, which is less efficient.
372
373Audit each individual driver calling drm_detect_hdmi_monitor() and switch to
374drm_display_info.is_hdmi if applicable.
375
376Contact: Laurent Pinchart, respective driver maintainers
377
378Level: Intermediate
379
380Consolidate custom driver modeset properties
381--------------------------------------------
382
383Before atomic modeset took place, many drivers where creating their own
384properties. Among other things, atomic brought the requirement that custom,
385driver specific properties should not be used.
386
387For this task, we aim to introduce core helpers or reuse the existing ones
388if available:
389
390A quick, unconfirmed, examples list.
391
392Introduce core helpers:
393- audio (amdgpu, intel, gma500, radeon)
394- brightness, contrast, etc (armada, nouveau) - overlay only (?)
395- broadcast rgb (gma500, intel)
396- colorkey (armada, nouveau, rcar) - overlay only (?)
397- dither (amdgpu, nouveau, radeon) - varies across drivers
398- underscan family (amdgpu, radeon, nouveau)
399
400Already in core:
401- colorspace (sti)
402- tv format names, enhancements (gma500, intel)
403- tv overscan, margins, etc. (gma500, intel)
404- zorder (omapdrm) - same as zpos (?)
405
406
407Contact: Emil Velikov, respective driver maintainers
408
409Level: Intermediate
410
411Use struct iosys_map throughout codebase
412----------------------------------------
413
414Pointers to shared device memory are stored in struct iosys_map. Each
415instance knows whether it refers to system or I/O memory. Most of the DRM-wide
416interface have been converted to use struct iosys_map, but implementations
417often still use raw pointers.
418
419The task is to use struct iosys_map where it makes sense.
420
421* Memory managers should use struct iosys_map for dma-buf-imported buffers.
422* TTM might benefit from using struct iosys_map internally.
423* Framebuffer copying and blitting helpers should operate on struct iosys_map.
424
425Contact: Thomas Zimmermann <tzimmermann@suse.de>, Christian König, Simona Vetter
426
427Level: Intermediate
428
429Review all drivers for setting struct drm_mode_config.{max_width,max_height} correctly
430--------------------------------------------------------------------------------------
431
432The values in struct drm_mode_config.{max_width,max_height} describe the
433maximum supported framebuffer size. It's the virtual screen size, but many
434drivers treat it like limitations of the physical resolution.
435
436The maximum width depends on the hardware's maximum scanline pitch. The
437maximum height depends on the amount of addressable video memory. Review all
438drivers to initialize the fields to the correct values.
439
440Contact: Thomas Zimmermann <tzimmermann@suse.de>
441
442Level: Intermediate
443
444Request memory regions in all fbdev drivers
445--------------------------------------------
446
447Old/ancient fbdev drivers do not request their memory properly.
448Go through these drivers and add code to request the memory regions
449that the driver uses. This requires adding calls to request_mem_region(),
450pci_request_region() or similar functions. Use helpers for managed cleanup
451where possible. Problematic areas include hardware that has exclusive ranges
452like VGA. VGA16fb does not request the range as it is expected.
453Drivers are pretty bad at doing this and there used to be conflicts among
454DRM and fbdev drivers. Still, it's the correct thing to do.
455
456Contact: Thomas Zimmermann <tzimmermann@suse.de>
457
458Level: Starter
459
460Remove driver dependencies on FB_DEVICE
461---------------------------------------
462
463A number of fbdev drivers provide attributes via sysfs and therefore depend
464on CONFIG_FB_DEVICE to be selected. Review each driver and attempt to make
465any dependencies on CONFIG_FB_DEVICE optional. At the minimum, the respective
466code in the driver could be conditionalized via ifdef CONFIG_FB_DEVICE. Not
467all drivers might be able to drop CONFIG_FB_DEVICE.
468
469Contact: Thomas Zimmermann <tzimmermann@suse.de>
470
471Level: Starter
472
473Remove disable/unprepare in remove/shutdown in panel-simple and panel-edp
474-------------------------------------------------------------------------
475
476As of commit d2aacaf07395 ("drm/panel: Check for already prepared/enabled in
477drm_panel"), we have a check in the drm_panel core to make sure nobody
478double-calls prepare/enable/disable/unprepare. Eventually that should probably
479be turned into a WARN_ON() or somehow made louder.
480
481At the moment, we expect that we may still encounter the warnings in the
482drm_panel core when using panel-simple and panel-edp. Since those panel
483drivers are used with a lot of different DRM modeset drivers they still
484make an extra effort to disable/unprepare the panel themsevles at shutdown
485time. Specifically we could still encounter those warnings if the panel
486driver gets shutdown() _before_ the DRM modeset driver and the DRM modeset
487driver properly calls drm_atomic_helper_shutdown() in its own shutdown()
488callback. Warnings could be avoided in such a case by using something like
489device links to ensure that the panel gets shutdown() after the DRM modeset
490driver.
491
492Once all DRM modeset drivers are known to shutdown properly, the extra
493calls to disable/unprepare in remove/shutdown in panel-simple and panel-edp
494should be removed and this TODO item marked complete.
495
496Contact: Douglas Anderson <dianders@chromium.org>
497
498Level: Intermediate
499
500Transition away from using mipi_dsi_*_write_seq()
501-------------------------------------------------
502
503The macros mipi_dsi_generic_write_seq() and mipi_dsi_dcs_write_seq() are
504non-intuitive because, if there are errors, they return out of the *caller's*
505function. We should move all callers to use mipi_dsi_generic_write_seq_multi()
506and mipi_dsi_dcs_write_seq_multi() macros instead.
507
508Once all callers are transitioned, the macros and the functions that they call,
509mipi_dsi_generic_write_chatty() and mipi_dsi_dcs_write_buffer_chatty(), can
510probably be removed. Alternatively, if people feel like the _multi() variants
511are overkill for some use cases, we could keep the mipi_dsi_*_write_seq()
512variants but change them not to return out of the caller.
513
514Contact: Douglas Anderson <dianders@chromium.org>
515
516Level: Starter
517
518Remove devm_drm_put_bridge()
519----------------------------
520
521Due to how the panel bridge handles the drm_bridge object lifetime, special
522care must be taken to dispose of the drm_bridge object when the
523panel_bridge is removed. This is currently managed using
524devm_drm_put_bridge(), but that is an unsafe, temporary workaround. To fix
525that, the DRM panel lifetime needs to be reworked. After the rework is
526done, remove devm_drm_put_bridge() and the TODO in
527drm_panel_bridge_remove().
528
529Contact: Maxime Ripard <mripard@kernel.org>,
530         Luca Ceresoli <luca.ceresoli@bootlin.com>
531
532Level: Intermediate
533
534Core refactorings
535=================
536
537Make panic handling work
538------------------------
539
540This is a really varied tasks with lots of little bits and pieces:
541
542* The panic path can't be tested currently, leading to constant breaking. The
543  main issue here is that panics can be triggered from hardirq contexts and
544  hence all panic related callback can run in hardirq context. It would be
545  awesome if we could test at least the fbdev helper code and driver code by
546  e.g. trigger calls through drm debugfs files. hardirq context could be
547  achieved by using an IPI to the local processor.
548
549* There's a massive confusion of different panic handlers. DRM fbdev emulation
550  helpers had their own (long removed), but on top of that the fbcon code itself
551  also has one. We need to make sure that they stop fighting over each other.
552  This is worked around by checking ``oops_in_progress`` at various entry points
553  into the DRM fbdev emulation helpers. A much cleaner approach here would be to
554  switch fbcon to the `threaded printk support
555  <https://lwn.net/Articles/800946/>`_.
556
557* ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
558  isn't a full solution for panic paths. We need to make sure that it only
559  returns true if there's a panic going on for real, and fix up all the
560  fallout.
561
562* The panic handler must never sleep, which also means it can't ever
563  ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
564  even spinlocks (because NMI and hardirq can panic too). We need to either
565  make sure to not call such paths, or trylock everything. Really tricky.
566
567* A clean solution would be an entirely separate panic output support in KMS,
568  bypassing the current fbcon support. See `[PATCH v2 0/3] drm: Add panic handling
569  <https://lore.kernel.org/dri-devel/20190311174218.51899-1-noralf@tronnes.org/>`_.
570
571* Encoding the actual oops and preceding dmesg in a QR might help with the
572  dread "important stuff scrolled away" problem. See `[RFC][PATCH] Oops messages
573  transfer using QR codes
574  <https://lore.kernel.org/lkml/1446217392-11981-1-git-send-email-alexandru.murtaza@intel.com/>`_
575  for some example code that could be reused.
576
577Contact: Simona Vetter
578
579Level: Advanced
580
581Clean up the debugfs support
582----------------------------
583
584There's a bunch of issues with it:
585
586- Convert drivers to support the drm_debugfs_add_files() function instead of
587  the drm_debugfs_create_files() function.
588
589- Improve late-register debugfs by rolling out the same debugfs pre-register
590  infrastructure for connector and crtc too. That way, the drivers won't need to
591  split their setup code into init and register anymore.
592
593- We probably want to have some support for debugfs files on crtc/connectors and
594  maybe other kms objects directly in core. There's even drm_print support in
595  the funcs for these objects to dump kms state, so it's all there. And then the
596  ->show() functions should obviously give you a pointer to the right object.
597
598- The drm_driver->debugfs_init hooks we have is just an artifact of the old
599  midlayered load sequence. DRM debugfs should work more like sysfs, where you
600  can create properties/files for an object anytime you want, and the core
601  takes care of publishing/unpuplishing all the files at register/unregister
602  time. Drivers shouldn't need to worry about these technicalities, and fixing
603  this (together with the drm_minor->drm_device move) would allow us to remove
604  debugfs_init.
605
606Contact: Simona Vetter
607
608Level: Intermediate
609
610Object lifetime fixes
611---------------------
612
613There's two related issues here
614
615- Cleanup up the various ->destroy callbacks, which often are all the same
616  simple code.
617
618- Lots of drivers erroneously allocate DRM modeset objects using devm_kzalloc,
619  which results in use-after free issues on driver unload. This can be serious
620  trouble even for drivers for hardware integrated on the SoC due to
621  EPROBE_DEFERRED backoff.
622
623Both these problems can be solved by switching over to drmm_kzalloc(), and the
624various convenience wrappers provided, e.g. drmm_crtc_alloc_with_planes(),
625drmm_universal_plane_alloc(), ... and so on.
626
627Contact: Simona Vetter
628
629Level: Intermediate
630
631Remove automatic page mapping from dma-buf importing
632----------------------------------------------------
633
634When importing dma-bufs, the dma-buf and PRIME frameworks automatically map
635imported pages into the importer's DMA area. drm_gem_prime_fd_to_handle() and
636drm_gem_prime_handle_to_fd() require that importers call dma_buf_attach()
637even if they never do actual device DMA, but only CPU access through
638dma_buf_vmap(). This is a problem for USB devices, which do not support DMA
639operations.
640
641To fix the issue, automatic page mappings should be removed from the
642buffer-sharing code. Fixing this is a bit more involved, since the import/export
643cache is also tied to &drm_gem_object.import_attach. Meanwhile we paper over
644this problem for USB devices by fishing out the USB host controller device, as
645long as that supports DMA. Otherwise importing can still needlessly fail.
646
647Contact: Thomas Zimmermann <tzimmermann@suse.de>, Simona Vetter
648
649Level: Advanced
650
651
652Better Testing
653==============
654
655Add unit tests using the Kernel Unit Testing (KUnit) framework
656--------------------------------------------------------------
657
658The `KUnit <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
659provides a common framework for unit tests within the Linux kernel. Having a
660test suite would allow to identify regressions earlier.
661
662A good candidate for the first unit tests are the format-conversion helpers in
663``drm_format_helper.c``.
664
665Contact: Javier Martinez Canillas <javierm@redhat.com>
666
667Level: Intermediate
668
669Clean up and document former selftests suites
670---------------------------------------------
671
672Some KUnit test suites (drm_buddy, drm_cmdline_parser, drm_damage_helper,
673drm_format, drm_framebuffer, drm_dp_mst_helper, drm_mm, drm_plane_helper and
674drm_rect) are former selftests suites that have been converted over when KUnit
675was first introduced.
676
677These suites were fairly undocumented, and with different goals than what unit
678tests can be. Trying to identify what each test in these suites actually test
679for, whether that makes sense for a unit test, and either remove it if it
680doesn't or document it if it does would be of great help.
681
682Contact: Maxime Ripard <mripard@kernel.org>
683
684Level: Intermediate
685
686Enable trinity for DRM
687----------------------
688
689And fix up the fallout. Should be really interesting ...
690
691Level: Advanced
692
693Make KMS tests in i-g-t generic
694-------------------------------
695
696The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
697including tons of testcases for corner-cases in the modesetting API. It would
698be awesome if those tests (at least the ones not relying on Intel-specific GEM
699features) could be made to run on any KMS driver.
700
701Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
702converting things over. For modeset tests we also first need a bit of
703infrastructure to use dumb buffers for untiled buffers, to be able to run all
704the non-i915 specific modeset tests.
705
706Level: Advanced
707
708Extend virtual test driver (VKMS)
709---------------------------------
710
711See the documentation of :ref:`VKMS <vkms>` for more details. This is an ideal
712internship task, since it only requires a virtual machine and can be sized to
713fit the available time.
714
715Level: See details
716
717Backlight Refactoring
718---------------------
719
720Backlight drivers have a triple enable/disable state, which is a bit overkill.
721Plan to fix this:
722
7231. Roll out backlight_enable() and backlight_disable() helpers everywhere. This
724   has started already.
7252. In all, only look at one of the three status bits set by the above helpers.
7263. Remove the other two status bits.
727
728Contact: Simona Vetter
729
730Level: Intermediate
731
732Driver Specific
733===============
734
735AMD DC Display Driver
736---------------------
737
738AMD DC is the display driver for AMD devices starting with Vega. There has been
739a bunch of progress cleaning it up but there's still plenty of work to be done.
740
741See drivers/gpu/drm/amd/display/TODO for tasks.
742
743Contact: Harry Wentland, Alex Deucher
744
745Bootsplash
746==========
747
748There is support in place now for writing internal DRM clients making it
749possible to pick up the bootsplash work that was rejected because it was written
750for fbdev.
751
752- [v6,8/8] drm/client: Hack: Add bootsplash example
753  https://patchwork.freedesktop.org/patch/306579/
754
755- [RFC PATCH v2 00/13] Kernel based bootsplash
756  https://lore.kernel.org/r/20171213194755.3409-1-mstaudt@suse.de
757
758Contact: Sam Ravnborg
759
760Level: Advanced
761
762Brightness handling on devices with multiple internal panels
763============================================================
764
765On x86/ACPI devices there can be multiple backlight firmware interfaces:
766(ACPI) video, vendor specific and others. As well as direct/native (PWM)
767register programming by the KMS driver.
768
769To deal with this backlight drivers used on x86/ACPI call
770acpi_video_get_backlight_type() which has heuristics (+quirks) to select
771which backlight interface to use; and backlight drivers which do not match
772the returned type will not register themselves, so that only one backlight
773device gets registered (in a single GPU setup, see below).
774
775At the moment this more or less assumes that there will only
776be 1 (internal) panel on a system.
777
778On systems with 2 panels this may be a problem, depending on
779what interface acpi_video_get_backlight_type() selects:
780
7811. native: in this case the KMS driver is expected to know which backlight
782   device belongs to which output so everything should just work.
7832. video: this does support controlling multiple backlights, but some work
784   will need to be done to get the output <-> backlight device mapping
785
786The above assumes both panels will require the same backlight interface type.
787Things will break on systems with multiple panels where the 2 panels need
788a different type of control. E.g. one panel needs ACPI video backlight control,
789where as the other is using native backlight control. Currently in this case
790only one of the 2 required backlight devices will get registered, based on
791the acpi_video_get_backlight_type() return value.
792
793If this (theoretical) case ever shows up, then supporting this will need some
794work. A possible solution here would be to pass a device and connector-name
795to acpi_video_get_backlight_type() so that it can deal with this.
796
797Note in a way we already have a case where userspace sees 2 panels,
798in dual GPU laptop setups with a mux. On those systems we may see
799either 2 native backlight devices; or 2 native backlight devices.
800
801Userspace already has code to deal with this by detecting if the related
802panel is active (iow which way the mux between the GPU and the panels
803points) and then uses that backlight device. Userspace here very much
804assumes a single panel though. It picks only 1 of the 2 backlight devices
805and then only uses that one.
806
807Note that all userspace code (that I know off) is currently hardcoded
808to assume a single panel.
809
810Before the recent changes to not register multiple (e.g. video + native)
811/sys/class/backlight devices for a single panel (on a single GPU laptop),
812userspace would see multiple backlight devices all controlling the same
813backlight.
814
815To deal with this userspace had to always picks one preferred device under
816/sys/class/backlight and will ignore the others. So to support brightness
817control on multiple panels userspace will need to be updated too.
818
819There are plans to allow brightness control through the KMS API by adding
820a "display brightness" property to drm_connector objects for panels. This
821solves a number of issues with the /sys/class/backlight API, including not
822being able to map a sysfs backlight device to a specific connector. Any
823userspace changes to add support for brightness control on devices with
824multiple panels really should build on top of this new KMS property.
825
826Contact: Hans de Goede
827
828Level: Advanced
829
830Buffer age or other damage accumulation algorithm for buffer damage
831===================================================================
832
833Drivers that do per-buffer uploads, need a buffer damage handling (rather than
834frame damage like drivers that do per-plane or per-CRTC uploads), but there is
835no support to get the buffer age or any other damage accumulation algorithm.
836
837For this reason, the damage helpers just fallback to a full plane update if the
838framebuffer attached to a plane has changed since the last page-flip. Drivers
839set &drm_plane_state.ignore_damage_clips to true as indication to
840drm_atomic_helper_damage_iter_init() and drm_atomic_helper_damage_iter_next()
841helpers that the damage clips should be ignored.
842
843This should be improved to get damage tracking properly working on drivers that
844do per-buffer uploads.
845
846More information about damage tracking and references to learning materials can
847be found in :ref:`damage_tracking_properties`.
848
849Contact: Javier Martinez Canillas <javierm@redhat.com>
850
851Level: Advanced
852
853Querying errors from drm_syncobj
854================================
855
856The drm_syncobj container can be used by driver independent code to signal
857complection of submission.
858
859One minor feature still missing is a generic DRM IOCTL to query the error
860status of binary and timeline drm_syncobj.
861
862This should probably be improved by implementing the necessary kernel interface
863and adding support for that in the userspace stack.
864
865Contact: Christian König
866
867Level: Starter
868
869Outside DRM
870===========
871
872Convert fbdev drivers to DRM
873----------------------------
874
875There are plenty of fbdev drivers for older hardware. Some hardware has
876become obsolete, but some still provides good(-enough) framebuffers. The
877drivers that are still useful should be converted to DRM and afterwards
878removed from fbdev.
879
880Very simple fbdev drivers can best be converted by starting with a new
881DRM driver. Simple KMS helpers and SHMEM should be able to handle any
882existing hardware. The new driver's call-back functions are filled from
883existing fbdev code.
884
885More complex fbdev drivers can be refactored step-by-step into a DRM
886driver with the help of the DRM fbconv helpers [4]_. These helpers provide
887the transition layer between the DRM core infrastructure and the fbdev
888driver interface. Create a new DRM driver on top of the fbconv helpers,
889copy over the fbdev driver, and hook it up to the DRM code. Examples for
890several fbdev drivers are available in Thomas Zimmermann's fbconv tree
891[4]_, as well as a tutorial of this process [5]_. The result is a primitive
892DRM driver that can run X11 and Weston.
893
894 .. [4] https://gitlab.freedesktop.org/tzimmermann/linux/tree/fbconv
895 .. [5] https://gitlab.freedesktop.org/tzimmermann/linux/blob/fbconv/drivers/gpu/drm/drm_fbconv_helper.c
896
897Contact: Thomas Zimmermann <tzimmermann@suse.de>
898
899Level: Advanced
900