xref: /linux/Documentation/gpu/todo.rst (revision 17cfcb68af3bc7d5e8ae08779b1853310a2949f3)
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
26Subsystem-wide refactorings
27===========================
28
29Remove custom dumb_map_offset implementations
30---------------------------------------------
31
32All GEM based drivers should be using drm_gem_create_mmap_offset() instead.
33Audit each individual driver, make sure it'll work with the generic
34implementation (there's lots of outdated locking leftovers in various
35implementations), and then remove it.
36
37Contact: Daniel Vetter, respective driver maintainers
38
39Level: Intermediate
40
41Convert existing KMS drivers to atomic modesetting
42--------------------------------------------------
43
443.19 has the atomic modeset interfaces and helpers, so drivers can now be
45converted over. Modern compositors like Wayland or Surfaceflinger on Android
46really want an atomic modeset interface, so this is all about the bright
47future.
48
49There is a conversion guide for atomic and all you need is a GPU for a
50non-converted driver (again virtual HW drivers for KVM are still all
51suitable).
52
53As part of this drivers also need to convert to universal plane (which means
54exposing primary & cursor as proper plane objects). But that's much easier to
55do by directly using the new atomic helper driver callbacks.
56
57Contact: Daniel Vetter, respective driver maintainers
58
59Level: Advanced
60
61Clean up the clipped coordination confusion around planes
62---------------------------------------------------------
63
64We have a helper to get this right with drm_plane_helper_check_update(), but
65it's not consistently used. This should be fixed, preferrably in the atomic
66helpers (and drivers then moved over to clipped coordinates). Probably the
67helper should also be moved from drm_plane_helper.c to the atomic helpers, to
68avoid confusion - the other helpers in that file are all deprecated legacy
69helpers.
70
71Contact: Ville Syrjälä, Daniel Vetter, driver maintainers
72
73Level: Advanced
74
75Convert early atomic drivers to async commit helpers
76----------------------------------------------------
77
78For the first year the atomic modeset helpers didn't support asynchronous /
79nonblocking commits, and every driver had to hand-roll them. This is fixed
80now, but there's still a pile of existing drivers that easily could be
81converted over to the new infrastructure.
82
83One issue with the helpers is that they require that drivers handle completion
84events for atomic commits correctly. But fixing these bugs is good anyway.
85
86Contact: Daniel Vetter, respective driver maintainers
87
88Level: Advanced
89
90Fallout from atomic KMS
91-----------------------
92
93``drm_atomic_helper.c`` provides a batch of functions which implement legacy
94IOCTLs on top of the new atomic driver interface. Which is really nice for
95gradual conversion of drivers, but unfortunately the semantic mismatches are
96a bit too severe. So there's some follow-up work to adjust the function
97interfaces to fix these issues:
98
99* atomic needs the lock acquire context. At the moment that's passed around
100  implicitly with some horrible hacks, and it's also allocate with
101  ``GFP_NOFAIL`` behind the scenes. All legacy paths need to start allocating
102  the acquire context explicitly on stack and then also pass it down into
103  drivers explicitly so that the legacy-on-atomic functions can use them.
104
105  Except for some driver code this is done. This task should be finished by
106  adding WARN_ON(!drm_drv_uses_atomic_modeset) in drm_modeset_lock_all().
107
108* A bunch of the vtable hooks are now in the wrong place: DRM has a split
109  between core vfunc tables (named ``drm_foo_funcs``), which are used to
110  implement the userspace ABI. And then there's the optional hooks for the
111  helper libraries (name ``drm_foo_helper_funcs``), which are purely for
112  internal use. Some of these hooks should be move from ``_funcs`` to
113  ``_helper_funcs`` since they are not part of the core ABI. There's a
114  ``FIXME`` comment in the kerneldoc for each such case in ``drm_crtc.h``.
115
116Contact: Daniel Vetter
117
118Level: Intermediate
119
120Get rid of dev->struct_mutex from GEM drivers
121---------------------------------------------
122
123``dev->struct_mutex`` is the Big DRM Lock from legacy days and infested
124everything. Nowadays in modern drivers the only bit where it's mandatory is
125serializing GEM buffer object destruction. Which unfortunately means drivers
126have to keep track of that lock and either call ``unreference`` or
127``unreference_locked`` depending upon context.
128
129Core GEM doesn't have a need for ``struct_mutex`` any more since kernel 4.8,
130and there's a ``gem_free_object_unlocked`` callback for any drivers which are
131entirely ``struct_mutex`` free.
132
133For drivers that need ``struct_mutex`` it should be replaced with a driver-
134private lock. The tricky part is the BO free functions, since those can't
135reliably take that lock any more. Instead state needs to be protected with
136suitable subordinate locks or some cleanup work pushed to a worker thread. For
137performance-critical drivers it might also be better to go with a more
138fine-grained per-buffer object and per-context lockings scheme. Currently only the
139``msm`` driver still use ``struct_mutex``.
140
141Contact: Daniel Vetter, respective driver maintainers
142
143Level: Advanced
144
145Convert instances of dev_info/dev_err/dev_warn to their DRM_DEV_* equivalent
146----------------------------------------------------------------------------
147
148For drivers which could have multiple instances, it is necessary to
149differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
150don't do this, drivers used dev_info/warn/err to make this differentiation. We
151now have DRM_DEV_* variants of the drm print macros, so we can start to convert
152those drivers back to using drm-formwatted specific log messages.
153
154Before you start this conversion please contact the relevant maintainers to make
155sure your work will be merged - not everyone agrees that the DRM dmesg macros
156are better.
157
158Contact: Sean Paul, Maintainer of the driver you plan to convert
159
160Level: Starter
161
162Convert drivers to use simple modeset suspend/resume
163----------------------------------------------------
164
165Most drivers (except i915 and nouveau) that use
166drm_atomic_helper_suspend/resume() can probably be converted to use
167drm_mode_config_helper_suspend/resume(). Also there's still open-coded version
168of the atomic suspend/resume code in older atomic modeset drivers.
169
170Contact: Maintainer of the driver you plan to convert
171
172Level: Intermediate
173
174Convert drivers to use drm_fbdev_generic_setup()
175------------------------------------------------
176
177Most drivers can use drm_fbdev_generic_setup(). Driver have to implement
178atomic modesetting and GEM vmap support. Current generic fbdev emulation
179expects the framebuffer in system memory (or system-like memory).
180
181Contact: Maintainer of the driver you plan to convert
182
183Level: Intermediate
184
185drm_framebuffer_funcs and drm_mode_config_funcs.fb_create cleanup
186-----------------------------------------------------------------
187
188A lot more drivers could be switched over to the drm_gem_framebuffer helpers.
189Various hold-ups:
190
191- Need to switch over to the generic dirty tracking code using
192  drm_atomic_helper_dirtyfb first (e.g. qxl).
193
194- Need to switch to drm_fbdev_generic_setup(), otherwise a lot of the custom fb
195  setup code can't be deleted.
196
197- Many drivers wrap drm_gem_fb_create() only to check for valid formats. For
198  atomic drivers we could check for valid formats by calling
199  drm_plane_check_pixel_format() against all planes, and pass if any plane
200  supports the format. For non-atomic that's not possible since like the format
201  list for the primary plane is fake and we'd therefor reject valid formats.
202
203- Many drivers subclass drm_framebuffer, we'd need a embedding compatible
204  version of the varios drm_gem_fb_create functions. Maybe called
205  drm_gem_fb_create/_with_dirty/_with_funcs as needed.
206
207Contact: Daniel Vetter
208
209Level: Intermediate
210
211Clean up mmap forwarding
212------------------------
213
214A lot of drivers forward gem mmap calls to dma-buf mmap for imported buffers.
215And also a lot of them forward dma-buf mmap to the gem mmap implementations.
216There's drm_gem_prime_mmap() for this now, but still needs to be rolled out.
217
218Contact: Daniel Vetter
219
220Level: Intermediate
221
222Generic fbdev defio support
223---------------------------
224
225The defio support code in the fbdev core has some very specific requirements,
226which means drivers need to have a special framebuffer for fbdev. The main
227issue is that it uses some fields in struct page itself, which breaks shmem
228gem objects (and other things). To support defio, affected drivers require
229the use of a shadow buffer, which may add CPU and memory overhead.
230
231Possible solution would be to write our own defio mmap code in the drm fbdev
232emulation. It would need to fully wrap the existing mmap ops, forwarding
233everything after it has done the write-protect/mkwrite trickery:
234
235- In the drm_fbdev_fb_mmap helper, if we need defio, change the
236  default page prots to write-protected with something like this::
237
238      vma->vm_page_prot = pgprot_wrprotect(vma->vm_page_prot);
239
240- Set the mkwrite and fsync callbacks with similar implementions to the core
241  fbdev defio stuff. These should all work on plain ptes, they don't actually
242  require a struct page.  uff. These should all work on plain ptes, they don't
243  actually require a struct page.
244
245- Track the dirty pages in a separate structure (bitfield with one bit per page
246  should work) to avoid clobbering struct page.
247
248Might be good to also have some igt testcases for this.
249
250Contact: Daniel Vetter, Noralf Tronnes
251
252Level: Advanced
253
254idr_init_base()
255---------------
256
257DRM core&drivers uses a lot of idr (integer lookup directories) for mapping
258userspace IDs to internal objects, and in most places ID=0 means NULL and hence
259is never used. Switching to idr_init_base() for these would make the idr more
260efficient.
261
262Contact: Daniel Vetter
263
264Level: Starter
265
266struct drm_gem_object_funcs
267---------------------------
268
269GEM objects can now have a function table instead of having the callbacks on the
270DRM driver struct. This is now the preferred way and drivers can be moved over.
271
272We also need a 2nd version of the CMA define that doesn't require the
273vmapping to be present (different hook for prime importing). Plus this needs to
274be rolled out to all drivers using their own implementations, too.
275
276Level: Intermediate
277
278Use DRM_MODESET_LOCK_ALL_* helpers instead of boilerplate
279---------------------------------------------------------
280
281For cases where drivers are attempting to grab the modeset locks with a local
282acquire context. Replace the boilerplate code surrounding
283drm_modeset_lock_all_ctx() with DRM_MODESET_LOCK_ALL_BEGIN() and
284DRM_MODESET_LOCK_ALL_END() instead.
285
286This should also be done for all places where drm_modest_lock_all() is still
287used.
288
289As a reference, take a look at the conversions already completed in drm core.
290
291Contact: Sean Paul, respective driver maintainers
292
293Level: Starter
294
295Rename CMA helpers to DMA helpers
296---------------------------------
297
298CMA (standing for contiguous memory allocator) is really a bit an accident of
299what these were used for first, a much better name would be DMA helpers. In the
300text these should even be called coherent DMA memory helpers (so maybe CDM, but
301no one knows what that means) since underneath they just use dma_alloc_coherent.
302
303Contact: Laurent Pinchart, Daniel Vetter
304
305Level: Intermediate (mostly because it is a huge tasks without good partial
306milestones, not technically itself that challenging)
307
308Convert direct mode.vrefresh accesses to use drm_mode_vrefresh()
309----------------------------------------------------------------
310
311drm_display_mode.vrefresh isn't guaranteed to be populated. As such, using it
312is risky and has been known to cause div-by-zero bugs. Fortunately, drm core
313has helper which will use mode.vrefresh if it's !0 and will calculate it from
314the timings when it's 0.
315
316Use simple search/replace, or (more fun) cocci to replace instances of direct
317vrefresh access with a call to the helper. Check out
318https://lists.freedesktop.org/archives/dri-devel/2019-January/205186.html for
319inspiration.
320
321Once all instances of vrefresh have been converted, remove vrefresh from
322drm_display_mode to avoid future use.
323
324Contact: Sean Paul
325
326Level: Starter
327
328Remove drm_display_mode.hsync
329-----------------------------
330
331We have drm_mode_hsync() to calculate this from hsync_start/end, since drivers
332shouldn't/don't use this, remove this member to avoid any temptations to use it
333in the future. If there is any debug code using drm_display_mode.hsync, convert
334it to use drm_mode_hsync() instead.
335
336Contact: Sean Paul
337
338Level: Starter
339
340drm_fb_helper tasks
341-------------------
342
343- drm_fb_helper_restore_fbdev_mode_unlocked() should call restore_fbdev_mode()
344  not the _force variant so it can bail out if there is a master. But first
345  these igt tests need to be fixed: kms_fbcon_fbt@psr and
346  kms_fbcon_fbt@psr-suspend.
347
348- The max connector argument for drm_fb_helper_init() isn't used anymore and
349  can be removed.
350
351- The helper doesn't keep an array of connectors anymore so these can be
352  removed: drm_fb_helper_single_add_all_connectors(),
353  drm_fb_helper_add_one_connector() and drm_fb_helper_remove_one_connector().
354
355Level: Intermediate
356
357connector register/unregister fixes
358-----------------------------------
359
360- For most connectors it's a no-op to call drm_connector_register/unregister
361  directly from driver code, drm_dev_register/unregister take care of this
362  already. We can remove all of them.
363
364- For dp drivers it's a bit more a mess, since we need the connector to be
365  registered when calling drm_dp_aux_register. Fix this by instead calling
366  drm_dp_aux_init, and moving the actual registering into a late_register
367  callback as recommended in the kerneldoc.
368
369Level: Intermediate
370
371Remove load/unload callbacks from all non-DRIVER_LEGACY drivers
372---------------------------------------------------------------
373
374The load/unload callbacks in struct &drm_driver are very much midlayers, plus
375for historical reasons they get the ordering wrong (and we can't fix that)
376between setting up the &drm_driver structure and calling drm_dev_register().
377
378- Rework drivers to no longer use the load/unload callbacks, directly coding the
379  load/unload sequence into the driver's probe function.
380
381- Once all non-DRIVER_LEGACY drivers are converted, disallow the load/unload
382  callbacks for all modern drivers.
383
384Contact: Daniel Vetter
385
386Level: Intermediate
387
388Core refactorings
389=================
390
391Make panic handling work
392------------------------
393
394This is a really varied tasks with lots of little bits and pieces:
395
396* The panic path can't be tested currently, leading to constant breaking. The
397  main issue here is that panics can be triggered from hardirq contexts and
398  hence all panic related callback can run in hardirq context. It would be
399  awesome if we could test at least the fbdev helper code and driver code by
400  e.g. trigger calls through drm debugfs files. hardirq context could be
401  achieved by using an IPI to the local processor.
402
403* There's a massive confusion of different panic handlers. DRM fbdev emulation
404  helpers have one, but on top of that the fbcon code itself also has one. We
405  need to make sure that they stop fighting over each another.
406
407* ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
408  isn't a full solution for panic paths. We need to make sure that it only
409  returns true if there's a panic going on for real, and fix up all the
410  fallout.
411
412* The panic handler must never sleep, which also means it can't ever
413  ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
414  even spinlocks (because NMI and hardirq can panic too). We need to either
415  make sure to not call such paths, or trylock everything. Really tricky.
416
417* For the above locking troubles reasons it's pretty much impossible to
418  attempt a synchronous modeset from panic handlers. The only thing we could
419  try to achive is an atomic ``set_base`` of the primary plane, and hope that
420  it shows up. Everything else probably needs to be delayed to some worker or
421  something else which happens later on. Otherwise it just kills the box
422  harder, prevent the panic from going out on e.g. netconsole.
423
424* There's also proposal for a simplied DRM console instead of the full-blown
425  fbcon and DRM fbdev emulation. Any kind of panic handling tricks should
426  obviously work for both console, in case we ever get kmslog merged.
427
428Contact: Daniel Vetter
429
430Level: Advanced
431
432Clean up the debugfs support
433----------------------------
434
435There's a bunch of issues with it:
436
437- The drm_info_list ->show() function doesn't even bother to cast to the drm
438  structure for you. This is lazy.
439
440- We probably want to have some support for debugfs files on crtc/connectors and
441  maybe other kms objects directly in core. There's even drm_print support in
442  the funcs for these objects to dump kms state, so it's all there. And then the
443  ->show() functions should obviously give you a pointer to the right object.
444
445- The drm_info_list stuff is centered on drm_minor instead of drm_device. For
446  anything we want to print drm_device (or maybe drm_file) is the right thing.
447
448- The drm_driver->debugfs_init hooks we have is just an artifact of the old
449  midlayered load sequence. DRM debugfs should work more like sysfs, where you
450  can create properties/files for an object anytime you want, and the core
451  takes care of publishing/unpuplishing all the files at register/unregister
452  time. Drivers shouldn't need to worry about these technicalities, and fixing
453  this (together with the drm_minor->drm_device move) would allow us to remove
454  debugfs_init.
455
456- Drop the return code and error checking from all debugfs functions. Greg KH is
457  working on this already.
458
459Contact: Daniel Vetter
460
461Level: Intermediate
462
463KMS cleanups
464------------
465
466Some of these date from the very introduction of KMS in 2008 ...
467
468- Make ->funcs and ->helper_private vtables optional. There's a bunch of empty
469  function tables in drivers, but before we can remove them we need to make sure
470  that all the users in helpers and drivers do correctly check for a NULL
471  vtable.
472
473- Cleanup up the various ->destroy callbacks. A lot of them just wrapt the
474  drm_*_cleanup implementations and can be removed. Some tack a kfree() at the
475  end, for which we could add drm_*_cleanup_kfree(). And then there's the (for
476  historical reasons) misnamed drm_primary_helper_destroy() function.
477
478Level: Intermediate
479
480Better Testing
481==============
482
483Enable trinity for DRM
484----------------------
485
486And fix up the fallout. Should be really interesting ...
487
488Level: Advanced
489
490Make KMS tests in i-g-t generic
491-------------------------------
492
493The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
494including tons of testcases for corner-cases in the modesetting API. It would
495be awesome if those tests (at least the ones not relying on Intel-specific GEM
496features) could be made to run on any KMS driver.
497
498Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
499converting things over. For modeset tests we also first need a bit of
500infrastructure to use dumb buffers for untiled buffers, to be able to run all
501the non-i915 specific modeset tests.
502
503Level: Advanced
504
505Extend virtual test driver (VKMS)
506---------------------------------
507
508See the documentation of :ref:`VKMS <vkms>` for more details. This is an ideal
509internship task, since it only requires a virtual machine and can be sized to
510fit the available time.
511
512Contact: Daniel Vetter
513
514Level: See details
515
516Backlight Refactoring
517---------------------
518
519Backlight drivers have a triple enable/disable state, which is a bit overkill.
520Plan to fix this:
521
5221. Roll out backlight_enable() and backlight_disable() helpers everywhere. This
523   has started already.
5242. In all, only look at one of the three status bits set by the above helpers.
5253. Remove the other two status bits.
526
527Contact: Daniel Vetter
528
529Level: Intermediate
530
531Driver Specific
532===============
533
534AMD DC Display Driver
535---------------------
536
537AMD DC is the display driver for AMD devices starting with Vega. There has been
538a bunch of progress cleaning it up but there's still plenty of work to be done.
539
540See drivers/gpu/drm/amd/display/TODO for tasks.
541
542Contact: Harry Wentland, Alex Deucher
543
544Bootsplash
545==========
546
547There is support in place now for writing internal DRM clients making it
548possible to pick up the bootsplash work that was rejected because it was written
549for fbdev.
550
551- [v6,8/8] drm/client: Hack: Add bootsplash example
552  https://patchwork.freedesktop.org/patch/306579/
553
554- [RFC PATCH v2 00/13] Kernel based bootsplash
555  https://lkml.org/lkml/2017/12/13/764
556
557Contact: Sam Ravnborg
558
559Level: Advanced
560
561Outside DRM
562===========
563
564Convert fbdev drivers to DRM
565----------------------------
566
567There are plenty of fbdev drivers for older hardware. Some hwardware has
568become obsolete, but some still provides good(-enough) framebuffers. The
569drivers that are still useful should be converted to DRM and afterwards
570removed from fbdev.
571
572Very simple fbdev drivers can best be converted by starting with a new
573DRM driver. Simple KMS helpers and SHMEM should be able to handle any
574existing hardware. The new driver's call-back functions are filled from
575existing fbdev code.
576
577More complex fbdev drivers can be refactored step-by-step into a DRM
578driver with the help of the DRM fbconv helpers. [1] These helpers provide
579the transition layer between the DRM core infrastructure and the fbdev
580driver interface. Create a new DRM driver on top of the fbconv helpers,
581copy over the fbdev driver, and hook it up to the DRM code. Examples for
582several fbdev drivers are available at [1] and a tutorial of this process
583available at [2]. The result is a primitive DRM driver that can run X11
584and Weston.
585
586 - [1] https://gitlab.freedesktop.org/tzimmermann/linux/tree/fbconv
587 - [2] https://gitlab.freedesktop.org/tzimmermann/linux/blob/fbconv/drivers/gpu/drm/drm_fbconv_helper.c
588
589Contact: Thomas Zimmermann <tzimmermann@suse.de>
590
591Level: Advanced
592