xref: /openbmc/linux/Documentation/gpu/todo.rst (revision 1504b6f9)
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37implementation (there's lots of outdated locking leftovers in various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 and all you need is a GPU for a
53non-converted driver (again virtual HW drivers for KVM are still all
54suitable).
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
60Contact: Daniel Vetter, respective driver maintainers
61
62Level: Advanced
63
64Clean up the clipped coordination confusion around planes
65---------------------------------------------------------
66
67We have a helper to get this right with drm_plane_helper_check_update(), but
68it's not consistently used. This should be fixed, preferrably in the atomic
69helpers (and drivers then moved over to clipped coordinates). Probably the
70helper should also be moved from drm_plane_helper.c to the atomic helpers, to
71avoid confusion - the other helpers in that file are all deprecated legacy
72helpers.
73
74Contact: Ville Syrjälä, Daniel Vetter, driver maintainers
75
76Level: Advanced
77
78Improve plane atomic_check helpers
79----------------------------------
80
81Aside from the clipped coordinates right above there's a few suboptimal things
82with the current helpers:
83
84- drm_plane_helper_funcs->atomic_check gets called for enabled or disabled
85  planes. At best this seems to confuse drivers, worst it means they blow up
86  when the plane is disabled without the CRTC. The only special handling is
87  resetting values in the plane state structures, which instead should be moved
88  into the drm_plane_funcs->atomic_duplicate_state functions.
89
90- Once that's done, helpers could stop calling ->atomic_check for disabled
91  planes.
92
93- Then we could go through all the drivers and remove the more-or-less confused
94  checks for plane_state->fb and plane_state->crtc.
95
96Contact: Daniel Vetter
97
98Level: Advanced
99
100Convert early atomic drivers to async commit helpers
101----------------------------------------------------
102
103For the first year the atomic modeset helpers didn't support asynchronous /
104nonblocking commits, and every driver had to hand-roll them. This is fixed
105now, but there's still a pile of existing drivers that easily could be
106converted over to the new infrastructure.
107
108One issue with the helpers is that they require that drivers handle completion
109events for atomic commits correctly. But fixing these bugs is good anyway.
110
111Somewhat related is the legacy_cursor_update hack, which should be replaced with
112the new atomic_async_check/commit functionality in the helpers in drivers that
113still look at that flag.
114
115Contact: Daniel Vetter, respective driver maintainers
116
117Level: Advanced
118
119Fallout from atomic KMS
120-----------------------
121
122``drm_atomic_helper.c`` provides a batch of functions which implement legacy
123IOCTLs on top of the new atomic driver interface. Which is really nice for
124gradual conversion of drivers, but unfortunately the semantic mismatches are
125a bit too severe. So there's some follow-up work to adjust the function
126interfaces to fix these issues:
127
128* atomic needs the lock acquire context. At the moment that's passed around
129  implicitly with some horrible hacks, and it's also allocate with
130  ``GFP_NOFAIL`` behind the scenes. All legacy paths need to start allocating
131  the acquire context explicitly on stack and then also pass it down into
132  drivers explicitly so that the legacy-on-atomic functions can use them.
133
134  Except for some driver code this is done. This task should be finished by
135  adding WARN_ON(!drm_drv_uses_atomic_modeset) in drm_modeset_lock_all().
136
137* A bunch of the vtable hooks are now in the wrong place: DRM has a split
138  between core vfunc tables (named ``drm_foo_funcs``), which are used to
139  implement the userspace ABI. And then there's the optional hooks for the
140  helper libraries (name ``drm_foo_helper_funcs``), which are purely for
141  internal use. Some of these hooks should be move from ``_funcs`` to
142  ``_helper_funcs`` since they are not part of the core ABI. There's a
143  ``FIXME`` comment in the kerneldoc for each such case in ``drm_crtc.h``.
144
145Contact: Daniel Vetter
146
147Level: Intermediate
148
149Get rid of dev->struct_mutex from GEM drivers
150---------------------------------------------
151
152``dev->struct_mutex`` is the Big DRM Lock from legacy days and infested
153everything. Nowadays in modern drivers the only bit where it's mandatory is
154serializing GEM buffer object destruction. Which unfortunately means drivers
155have to keep track of that lock and either call ``unreference`` or
156``unreference_locked`` depending upon context.
157
158Core GEM doesn't have a need for ``struct_mutex`` any more since kernel 4.8,
159and there's a GEM object ``free`` callback for any drivers which are
160entirely ``struct_mutex`` free.
161
162For drivers that need ``struct_mutex`` it should be replaced with a driver-
163private lock. The tricky part is the BO free functions, since those can't
164reliably take that lock any more. Instead state needs to be protected with
165suitable subordinate locks or some cleanup work pushed to a worker thread. For
166performance-critical drivers it might also be better to go with a more
167fine-grained per-buffer object and per-context lockings scheme. Currently only
168the ``msm`` and `i915` drivers use ``struct_mutex``.
169
170Contact: Daniel Vetter, respective driver maintainers
171
172Level: Advanced
173
174Move Buffer Object Locking to dma_resv_lock()
175---------------------------------------------
176
177Many drivers have their own per-object locking scheme, usually using
178mutex_lock(). This causes all kinds of trouble for buffer sharing, since
179depending which driver is the exporter and importer, the locking hierarchy is
180reversed.
181
182To solve this we need one standard per-object locking mechanism, which is
183dma_resv_lock(). This lock needs to be called as the outermost lock, with all
184other driver specific per-object locks removed. The problem is tha rolling out
185the actual change to the locking contract is a flag day, due to struct dma_buf
186buffer sharing.
187
188Level: Expert
189
190Convert logging to drm_* functions with drm_device paramater
191------------------------------------------------------------
192
193For drivers which could have multiple instances, it is necessary to
194differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
195don't do this, drivers used dev_info/warn/err to make this differentiation. We
196now have drm_* variants of the drm print functions, so we can start to convert
197those drivers back to using drm-formatted specific log messages.
198
199Before you start this conversion please contact the relevant maintainers to make
200sure your work will be merged - not everyone agrees that the DRM dmesg macros
201are better.
202
203Contact: Sean Paul, Maintainer of the driver you plan to convert
204
205Level: Starter
206
207Convert drivers to use simple modeset suspend/resume
208----------------------------------------------------
209
210Most drivers (except i915 and nouveau) that use
211drm_atomic_helper_suspend/resume() can probably be converted to use
212drm_mode_config_helper_suspend/resume(). Also there's still open-coded version
213of the atomic suspend/resume code in older atomic modeset drivers.
214
215Contact: Maintainer of the driver you plan to convert
216
217Level: Intermediate
218
219Convert drivers to use drm_fbdev_generic_setup()
220------------------------------------------------
221
222Most drivers can use drm_fbdev_generic_setup(). Driver have to implement
223atomic modesetting and GEM vmap support. Historically, generic fbdev emulation
224expected the framebuffer in system memory or system-like memory. By employing
225struct iosys_map, drivers with frambuffers in I/O memory can be supported
226as well.
227
228Contact: Maintainer of the driver you plan to convert
229
230Level: Intermediate
231
232Reimplement functions in drm_fbdev_fb_ops without fbdev
233-------------------------------------------------------
234
235A number of callback functions in drm_fbdev_fb_ops could benefit from
236being rewritten without dependencies on the fbdev module. Some of the
237helpers could further benefit from using struct iosys_map instead of
238raw pointers.
239
240Contact: Thomas Zimmermann <tzimmermann@suse.de>, Daniel Vetter
241
242Level: Advanced
243
244Benchmark and optimize blitting and format-conversion function
245--------------------------------------------------------------
246
247Drawing to dispay memory quickly is crucial for many applications'
248performance.
249
250On at least x86-64, sys_imageblit() is significantly slower than
251cfb_imageblit(), even though both use the same blitting algorithm and
252the latter is written for I/O memory. It turns out that cfb_imageblit()
253uses movl instructions, while sys_imageblit apparently does not. This
254seems to be a problem with gcc's optimizer. DRM's format-conversion
255helpers might be subject to similar issues.
256
257Benchmark and optimize fbdev's sys_() helpers and DRM's format-conversion
258helpers. In cases that can be further optimized, maybe implement a different
259algorithm. For micro-optimizations, use movl/movq instructions explicitly.
260That might possibly require architecture-specific helpers (e.g., storel()
261storeq()).
262
263Contact: Thomas Zimmermann <tzimmermann@suse.de>
264
265Level: Intermediate
266
267drm_framebuffer_funcs and drm_mode_config_funcs.fb_create cleanup
268-----------------------------------------------------------------
269
270A lot more drivers could be switched over to the drm_gem_framebuffer helpers.
271Various hold-ups:
272
273- Need to switch over to the generic dirty tracking code using
274  drm_atomic_helper_dirtyfb first (e.g. qxl).
275
276- Need to switch to drm_fbdev_generic_setup(), otherwise a lot of the custom fb
277  setup code can't be deleted.
278
279- Many drivers wrap drm_gem_fb_create() only to check for valid formats. For
280  atomic drivers we could check for valid formats by calling
281  drm_plane_check_pixel_format() against all planes, and pass if any plane
282  supports the format. For non-atomic that's not possible since like the format
283  list for the primary plane is fake and we'd therefor reject valid formats.
284
285- Many drivers subclass drm_framebuffer, we'd need a embedding compatible
286  version of the varios drm_gem_fb_create functions. Maybe called
287  drm_gem_fb_create/_with_dirty/_with_funcs as needed.
288
289Contact: Daniel Vetter
290
291Level: Intermediate
292
293Generic fbdev defio support
294---------------------------
295
296The defio support code in the fbdev core has some very specific requirements,
297which means drivers need to have a special framebuffer for fbdev. The main
298issue is that it uses some fields in struct page itself, which breaks shmem
299gem objects (and other things). To support defio, affected drivers require
300the use of a shadow buffer, which may add CPU and memory overhead.
301
302Possible solution would be to write our own defio mmap code in the drm fbdev
303emulation. It would need to fully wrap the existing mmap ops, forwarding
304everything after it has done the write-protect/mkwrite trickery:
305
306- In the drm_fbdev_fb_mmap helper, if we need defio, change the
307  default page prots to write-protected with something like this::
308
309      vma->vm_page_prot = pgprot_wrprotect(vma->vm_page_prot);
310
311- Set the mkwrite and fsync callbacks with similar implementions to the core
312  fbdev defio stuff. These should all work on plain ptes, they don't actually
313  require a struct page.  uff. These should all work on plain ptes, they don't
314  actually require a struct page.
315
316- Track the dirty pages in a separate structure (bitfield with one bit per page
317  should work) to avoid clobbering struct page.
318
319Might be good to also have some igt testcases for this.
320
321Contact: Daniel Vetter, Noralf Tronnes
322
323Level: Advanced
324
325struct drm_gem_object_funcs
326---------------------------
327
328GEM objects can now have a function table instead of having the callbacks on the
329DRM driver struct. This is now the preferred way. Callbacks in drivers have been
330converted, except for struct drm_driver.gem_prime_mmap.
331
332Level: Intermediate
333
334connector register/unregister fixes
335-----------------------------------
336
337- For most connectors it's a no-op to call drm_connector_register/unregister
338  directly from driver code, drm_dev_register/unregister take care of this
339  already. We can remove all of them.
340
341- For dp drivers it's a bit more a mess, since we need the connector to be
342  registered when calling drm_dp_aux_register. Fix this by instead calling
343  drm_dp_aux_init, and moving the actual registering into a late_register
344  callback as recommended in the kerneldoc.
345
346Level: Intermediate
347
348Remove load/unload callbacks from all non-DRIVER_LEGACY drivers
349---------------------------------------------------------------
350
351The load/unload callbacks in struct &drm_driver are very much midlayers, plus
352for historical reasons they get the ordering wrong (and we can't fix that)
353between setting up the &drm_driver structure and calling drm_dev_register().
354
355- Rework drivers to no longer use the load/unload callbacks, directly coding the
356  load/unload sequence into the driver's probe function.
357
358- Once all non-DRIVER_LEGACY drivers are converted, disallow the load/unload
359  callbacks for all modern drivers.
360
361Contact: Daniel Vetter
362
363Level: Intermediate
364
365Replace drm_detect_hdmi_monitor() with drm_display_info.is_hdmi
366---------------------------------------------------------------
367
368Once EDID is parsed, the monitor HDMI support information is available through
369drm_display_info.is_hdmi. Many drivers still call drm_detect_hdmi_monitor() to
370retrieve the same information, which is less efficient.
371
372Audit each individual driver calling drm_detect_hdmi_monitor() and switch to
373drm_display_info.is_hdmi if applicable.
374
375Contact: Laurent Pinchart, respective driver maintainers
376
377Level: Intermediate
378
379Consolidate custom driver modeset properties
380--------------------------------------------
381
382Before atomic modeset took place, many drivers where creating their own
383properties. Among other things, atomic brought the requirement that custom,
384driver specific properties should not be used.
385
386For this task, we aim to introduce core helpers or reuse the existing ones
387if available:
388
389A quick, unconfirmed, examples list.
390
391Introduce core helpers:
392- audio (amdgpu, intel, gma500, radeon)
393- brightness, contrast, etc (armada, nouveau) - overlay only (?)
394- broadcast rgb (gma500, intel)
395- colorkey (armada, nouveau, rcar) - overlay only (?)
396- dither (amdgpu, nouveau, radeon) - varies across drivers
397- underscan family (amdgpu, radeon, nouveau)
398
399Already in core:
400- colorspace (sti)
401- tv format names, enhancements (gma500, intel)
402- tv overscan, margins, etc. (gma500, intel)
403- zorder (omapdrm) - same as zpos (?)
404
405
406Contact: Emil Velikov, respective driver maintainers
407
408Level: Intermediate
409
410Use struct iosys_map throughout codebase
411----------------------------------------
412
413Pointers to shared device memory are stored in struct iosys_map. Each
414instance knows whether it refers to system or I/O memory. Most of the DRM-wide
415interface have been converted to use struct iosys_map, but implementations
416often still use raw pointers.
417
418The task is to use struct iosys_map where it makes sense.
419
420* Memory managers should use struct iosys_map for dma-buf-imported buffers.
421* TTM might benefit from using struct iosys_map internally.
422* Framebuffer copying and blitting helpers should operate on struct iosys_map.
423
424Contact: Thomas Zimmermann <tzimmermann@suse.de>, Christian König, Daniel Vetter
425
426Level: Intermediate
427
428Review all drivers for setting struct drm_mode_config.{max_width,max_height} correctly
429--------------------------------------------------------------------------------------
430
431The values in struct drm_mode_config.{max_width,max_height} describe the
432maximum supported framebuffer size. It's the virtual screen size, but many
433drivers treat it like limitations of the physical resolution.
434
435The maximum width depends on the hardware's maximum scanline pitch. The
436maximum height depends on the amount of addressable video memory. Review all
437drivers to initialize the fields to the correct values.
438
439Contact: Thomas Zimmermann <tzimmermann@suse.de>
440
441Level: Intermediate
442
443Request memory regions in all drivers
444-------------------------------------
445
446Go through all drivers and add code to request the memory regions that the
447driver uses. This requires adding calls to request_mem_region(),
448pci_request_region() or similar functions. Use helpers for managed cleanup
449where possible.
450
451Drivers are pretty bad at doing this and there used to be conflicts among
452DRM and fbdev drivers. Still, it's the correct thing to do.
453
454Contact: Thomas Zimmermann <tzimmermann@suse.de>
455
456Level: Starter
457
458
459Core refactorings
460=================
461
462Make panic handling work
463------------------------
464
465This is a really varied tasks with lots of little bits and pieces:
466
467* The panic path can't be tested currently, leading to constant breaking. The
468  main issue here is that panics can be triggered from hardirq contexts and
469  hence all panic related callback can run in hardirq context. It would be
470  awesome if we could test at least the fbdev helper code and driver code by
471  e.g. trigger calls through drm debugfs files. hardirq context could be
472  achieved by using an IPI to the local processor.
473
474* There's a massive confusion of different panic handlers. DRM fbdev emulation
475  helpers had their own (long removed), but on top of that the fbcon code itself
476  also has one. We need to make sure that they stop fighting over each other.
477  This is worked around by checking ``oops_in_progress`` at various entry points
478  into the DRM fbdev emulation helpers. A much cleaner approach here would be to
479  switch fbcon to the `threaded printk support
480  <https://lwn.net/Articles/800946/>`_.
481
482* ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
483  isn't a full solution for panic paths. We need to make sure that it only
484  returns true if there's a panic going on for real, and fix up all the
485  fallout.
486
487* The panic handler must never sleep, which also means it can't ever
488  ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
489  even spinlocks (because NMI and hardirq can panic too). We need to either
490  make sure to not call such paths, or trylock everything. Really tricky.
491
492* A clean solution would be an entirely separate panic output support in KMS,
493  bypassing the current fbcon support. See `[PATCH v2 0/3] drm: Add panic handling
494  <https://lore.kernel.org/dri-devel/20190311174218.51899-1-noralf@tronnes.org/>`_.
495
496* Encoding the actual oops and preceding dmesg in a QR might help with the
497  dread "important stuff scrolled away" problem. See `[RFC][PATCH] Oops messages
498  transfer using QR codes
499  <https://lore.kernel.org/lkml/1446217392-11981-1-git-send-email-alexandru.murtaza@intel.com/>`_
500  for some example code that could be reused.
501
502Contact: Daniel Vetter
503
504Level: Advanced
505
506Clean up the debugfs support
507----------------------------
508
509There's a bunch of issues with it:
510
511- The drm_info_list ->show() function doesn't even bother to cast to the drm
512  structure for you. This is lazy.
513
514- We probably want to have some support for debugfs files on crtc/connectors and
515  maybe other kms objects directly in core. There's even drm_print support in
516  the funcs for these objects to dump kms state, so it's all there. And then the
517  ->show() functions should obviously give you a pointer to the right object.
518
519- The drm_info_list stuff is centered on drm_minor instead of drm_device. For
520  anything we want to print drm_device (or maybe drm_file) is the right thing.
521
522- The drm_driver->debugfs_init hooks we have is just an artifact of the old
523  midlayered load sequence. DRM debugfs should work more like sysfs, where you
524  can create properties/files for an object anytime you want, and the core
525  takes care of publishing/unpuplishing all the files at register/unregister
526  time. Drivers shouldn't need to worry about these technicalities, and fixing
527  this (together with the drm_minor->drm_device move) would allow us to remove
528  debugfs_init.
529
530Previous RFC that hasn't landed yet: https://lore.kernel.org/dri-devel/20200513114130.28641-2-wambui.karugax@gmail.com/
531
532Contact: Daniel Vetter
533
534Level: Intermediate
535
536Object lifetime fixes
537---------------------
538
539There's two related issues here
540
541- Cleanup up the various ->destroy callbacks, which often are all the same
542  simple code.
543
544- Lots of drivers erroneously allocate DRM modeset objects using devm_kzalloc,
545  which results in use-after free issues on driver unload. This can be serious
546  trouble even for drivers for hardware integrated on the SoC due to
547  EPROBE_DEFERRED backoff.
548
549Both these problems can be solved by switching over to drmm_kzalloc(), and the
550various convenience wrappers provided, e.g. drmm_crtc_alloc_with_planes(),
551drmm_universal_plane_alloc(), ... and so on.
552
553Contact: Daniel Vetter
554
555Level: Intermediate
556
557Remove automatic page mapping from dma-buf importing
558----------------------------------------------------
559
560When importing dma-bufs, the dma-buf and PRIME frameworks automatically map
561imported pages into the importer's DMA area. drm_gem_prime_fd_to_handle() and
562drm_gem_prime_handle_to_fd() require that importers call dma_buf_attach()
563even if they never do actual device DMA, but only CPU access through
564dma_buf_vmap(). This is a problem for USB devices, which do not support DMA
565operations.
566
567To fix the issue, automatic page mappings should be removed from the
568buffer-sharing code. Fixing this is a bit more involved, since the import/export
569cache is also tied to &drm_gem_object.import_attach. Meanwhile we paper over
570this problem for USB devices by fishing out the USB host controller device, as
571long as that supports DMA. Otherwise importing can still needlessly fail.
572
573Contact: Thomas Zimmermann <tzimmermann@suse.de>, Daniel Vetter
574
575Level: Advanced
576
577
578Better Testing
579==============
580
581Add unit tests using the Kernel Unit Testing (KUnit) framework
582--------------------------------------------------------------
583
584The `KUnit <https://www.kernel.org/doc/html/latest/dev-tools/kunit/index.html>`_
585provides a common framework for unit tests within the Linux kernel. Having a
586test suite would allow to identify regressions earlier.
587
588A good candidate for the first unit tests are the format-conversion helpers in
589``drm_format_helper.c``.
590
591Contact: Javier Martinez Canillas <javierm@redhat.com>
592
593Level: Intermediate
594
595Enable trinity for DRM
596----------------------
597
598And fix up the fallout. Should be really interesting ...
599
600Level: Advanced
601
602Make KMS tests in i-g-t generic
603-------------------------------
604
605The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
606including tons of testcases for corner-cases in the modesetting API. It would
607be awesome if those tests (at least the ones not relying on Intel-specific GEM
608features) could be made to run on any KMS driver.
609
610Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
611converting things over. For modeset tests we also first need a bit of
612infrastructure to use dumb buffers for untiled buffers, to be able to run all
613the non-i915 specific modeset tests.
614
615Level: Advanced
616
617Extend virtual test driver (VKMS)
618---------------------------------
619
620See the documentation of :ref:`VKMS <vkms>` for more details. This is an ideal
621internship task, since it only requires a virtual machine and can be sized to
622fit the available time.
623
624Level: See details
625
626Backlight Refactoring
627---------------------
628
629Backlight drivers have a triple enable/disable state, which is a bit overkill.
630Plan to fix this:
631
6321. Roll out backlight_enable() and backlight_disable() helpers everywhere. This
633   has started already.
6342. In all, only look at one of the three status bits set by the above helpers.
6353. Remove the other two status bits.
636
637Contact: Daniel Vetter
638
639Level: Intermediate
640
641Driver Specific
642===============
643
644AMD DC Display Driver
645---------------------
646
647AMD DC is the display driver for AMD devices starting with Vega. There has been
648a bunch of progress cleaning it up but there's still plenty of work to be done.
649
650See drivers/gpu/drm/amd/display/TODO for tasks.
651
652Contact: Harry Wentland, Alex Deucher
653
654vmwgfx: Replace hashtable with Linux' implementation
655----------------------------------------------------
656
657The vmwgfx driver uses its own hashtable implementation. Replace the
658code with Linux' implementation and update the callers. It's mostly a
659refactoring task, but the interfaces are different.
660
661Contact: Zack Rusin, Thomas Zimmermann <tzimmermann@suse.de>
662
663Level: Intermediate
664
665Bootsplash
666==========
667
668There is support in place now for writing internal DRM clients making it
669possible to pick up the bootsplash work that was rejected because it was written
670for fbdev.
671
672- [v6,8/8] drm/client: Hack: Add bootsplash example
673  https://patchwork.freedesktop.org/patch/306579/
674
675- [RFC PATCH v2 00/13] Kernel based bootsplash
676  https://lore.kernel.org/r/20171213194755.3409-1-mstaudt@suse.de
677
678Contact: Sam Ravnborg
679
680Level: Advanced
681
682Brightness handling on devices with multiple internal panels
683============================================================
684
685On x86/ACPI devices there can be multiple backlight firmware interfaces:
686(ACPI) video, vendor specific and others. As well as direct/native (PWM)
687register programming by the KMS driver.
688
689To deal with this backlight drivers used on x86/ACPI call
690acpi_video_get_backlight_type() which has heuristics (+quirks) to select
691which backlight interface to use; and backlight drivers which do not match
692the returned type will not register themselves, so that only one backlight
693device gets registered (in a single GPU setup, see below).
694
695At the moment this more or less assumes that there will only
696be 1 (internal) panel on a system.
697
698On systems with 2 panels this may be a problem, depending on
699what interface acpi_video_get_backlight_type() selects:
700
7011. native: in this case the KMS driver is expected to know which backlight
702   device belongs to which output so everything should just work.
7032. video: this does support controlling multiple backlights, but some work
704   will need to be done to get the output <-> backlight device mapping
705
706The above assumes both panels will require the same backlight interface type.
707Things will break on systems with multiple panels where the 2 panels need
708a different type of control. E.g. one panel needs ACPI video backlight control,
709where as the other is using native backlight control. Currently in this case
710only one of the 2 required backlight devices will get registered, based on
711the acpi_video_get_backlight_type() return value.
712
713If this (theoretical) case ever shows up, then supporting this will need some
714work. A possible solution here would be to pass a device and connector-name
715to acpi_video_get_backlight_type() so that it can deal with this.
716
717Note in a way we already have a case where userspace sees 2 panels,
718in dual GPU laptop setups with a mux. On those systems we may see
719either 2 native backlight devices; or 2 native backlight devices.
720
721Userspace already has code to deal with this by detecting if the related
722panel is active (iow which way the mux between the GPU and the panels
723points) and then uses that backlight device. Userspace here very much
724assumes a single panel though. It picks only 1 of the 2 backlight devices
725and then only uses that one.
726
727Note that all userspace code (that I know off) is currently hardcoded
728to assume a single panel.
729
730Before the recent changes to not register multiple (e.g. video + native)
731/sys/class/backlight devices for a single panel (on a single GPU laptop),
732userspace would see multiple backlight devices all controlling the same
733backlight.
734
735To deal with this userspace had to always picks one preferred device under
736/sys/class/backlight and will ignore the others. So to support brightness
737control on multiple panels userspace will need to be updated too.
738
739There are plans to allow brightness control through the KMS API by adding
740a "display brightness" property to drm_connector objects for panels. This
741solves a number of issues with the /sys/class/backlight API, including not
742being able to map a sysfs backlight device to a specific connector. Any
743userspace changes to add support for brightness control on devices with
744multiple panels really should build on top of this new KMS property.
745
746Contact: Hans de Goede
747
748Level: Advanced
749
750Outside DRM
751===========
752
753Convert fbdev drivers to DRM
754----------------------------
755
756There are plenty of fbdev drivers for older hardware. Some hardware has
757become obsolete, but some still provides good(-enough) framebuffers. The
758drivers that are still useful should be converted to DRM and afterwards
759removed from fbdev.
760
761Very simple fbdev drivers can best be converted by starting with a new
762DRM driver. Simple KMS helpers and SHMEM should be able to handle any
763existing hardware. The new driver's call-back functions are filled from
764existing fbdev code.
765
766More complex fbdev drivers can be refactored step-by-step into a DRM
767driver with the help of the DRM fbconv helpers. [1] These helpers provide
768the transition layer between the DRM core infrastructure and the fbdev
769driver interface. Create a new DRM driver on top of the fbconv helpers,
770copy over the fbdev driver, and hook it up to the DRM code. Examples for
771several fbdev drivers are available at [1] and a tutorial of this process
772available at [2]. The result is a primitive DRM driver that can run X11
773and Weston.
774
775 - [1] https://gitlab.freedesktop.org/tzimmermann/linux/tree/fbconv
776 - [2] https://gitlab.freedesktop.org/tzimmermann/linux/blob/fbconv/drivers/gpu/drm/drm_fbconv_helper.c
777
778Contact: Thomas Zimmermann <tzimmermann@suse.de>
779
780Level: Advanced
781