xref: /openbmc/linux/Documentation/gpu/todo.rst (revision e907bf3c)
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
10Subsystem-wide refactorings
11===========================
12
13De-midlayer drivers
14-------------------
15
16With the recent ``drm_bus`` cleanup patches for 3.17 it is no longer required
17to have a ``drm_bus`` structure set up. Drivers can directly set up the
18``drm_device`` structure instead of relying on bus methods in ``drm_usb.c``
19and ``drm_pci.c``. The goal is to get rid of the driver's ``->load`` /
20``->unload`` callbacks and open-code the load/unload sequence properly, using
21the new two-stage ``drm_device`` setup/teardown.
22
23Once all existing drivers are converted we can also remove those bus support
24files for USB and platform devices.
25
26All you need is a GPU for a non-converted driver (currently almost all of
27them, but also all the virtual ones used by KVM, so everyone qualifies).
28
29Contact: Daniel Vetter, Thierry Reding, respective driver maintainers
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37implementation (there's lots of outdated locking leftovers in various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
42Convert existing KMS drivers to atomic modesetting
43--------------------------------------------------
44
453.19 has the atomic modeset interfaces and helpers, so drivers can now be
46converted over. Modern compositors like Wayland or Surfaceflinger on Android
47really want an atomic modeset interface, so this is all about the bright
48future.
49
50There is a conversion guide for atomic and all you need is a GPU for a
51non-converted driver (again virtual HW drivers for KVM are still all
52suitable).
53
54As part of this drivers also need to convert to universal plane (which means
55exposing primary & cursor as proper plane objects). But that's much easier to
56do by directly using the new atomic helper driver callbacks.
57
58Contact: Daniel Vetter, respective driver maintainers
59
60Clean up the clipped coordination confusion around planes
61---------------------------------------------------------
62
63We have a helper to get this right with drm_plane_helper_check_update(), but
64it's not consistently used. This should be fixed, preferrably in the atomic
65helpers (and drivers then moved over to clipped coordinates). Probably the
66helper should also be moved from drm_plane_helper.c to the atomic helpers, to
67avoid confusion - the other helpers in that file are all deprecated legacy
68helpers.
69
70Contact: Ville Syrjälä, Daniel Vetter, driver maintainers
71
72Convert early atomic drivers to async commit helpers
73----------------------------------------------------
74
75For the first year the atomic modeset helpers didn't support asynchronous /
76nonblocking commits, and every driver had to hand-roll them. This is fixed
77now, but there's still a pile of existing drivers that easily could be
78converted over to the new infrastructure.
79
80One issue with the helpers is that they require that drivers handle completion
81events for atomic commits correctly. But fixing these bugs is good anyway.
82
83Contact: Daniel Vetter, respective driver maintainers
84
85Better manual-upload support for atomic
86---------------------------------------
87
88This would be especially useful for tinydrm:
89
90- Add a struct drm_rect dirty_clip to drm_crtc_state. When duplicating the
91  crtc state, clear that to the max values, x/y = 0 and w/h = MAX_INT, in
92  __drm_atomic_helper_crtc_duplicate_state().
93
94- Move tinydrm_merge_clips into drm_framebuffer.c, dropping the tinydrm\_
95  prefix ofc and using drm_fb\_. drm_framebuffer.c makes sense since this
96  is a function useful to implement the fb->dirty function.
97
98- Create a new drm_fb_dirty function which does essentially what e.g.
99  mipi_dbi_fb_dirty does. You can use e.g. drm_atomic_helper_update_plane as the
100  template. But instead of doing a simple full-screen plane update, this new
101  helper also sets crtc_state->dirty_clip to the right coordinates. And of
102  course it needs to check whether the fb is actually active (and maybe where),
103  so there's some book-keeping involved. There's also some good fun involved in
104  scaling things appropriately. For that case we might simply give up and
105  declare the entire area covered by the plane as dirty.
106
107Contact: Noralf Trønnes, Daniel Vetter
108
109Fallout from atomic KMS
110-----------------------
111
112``drm_atomic_helper.c`` provides a batch of functions which implement legacy
113IOCTLs on top of the new atomic driver interface. Which is really nice for
114gradual conversion of drivers, but unfortunately the semantic mismatches are
115a bit too severe. So there's some follow-up work to adjust the function
116interfaces to fix these issues:
117
118* atomic needs the lock acquire context. At the moment that's passed around
119  implicitly with some horrible hacks, and it's also allocate with
120  ``GFP_NOFAIL`` behind the scenes. All legacy paths need to start allocating
121  the acquire context explicitly on stack and then also pass it down into
122  drivers explicitly so that the legacy-on-atomic functions can use them.
123
124  Except for some driver code this is done. This task should be finished by
125  adding WARN_ON(!drm_drv_uses_atomic_modeset) in drm_modeset_lock_all().
126
127* A bunch of the vtable hooks are now in the wrong place: DRM has a split
128  between core vfunc tables (named ``drm_foo_funcs``), which are used to
129  implement the userspace ABI. And then there's the optional hooks for the
130  helper libraries (name ``drm_foo_helper_funcs``), which are purely for
131  internal use. Some of these hooks should be move from ``_funcs`` to
132  ``_helper_funcs`` since they are not part of the core ABI. There's a
133  ``FIXME`` comment in the kerneldoc for each such case in ``drm_crtc.h``.
134
135Contact: Daniel Vetter
136
137Get rid of dev->struct_mutex from GEM drivers
138---------------------------------------------
139
140``dev->struct_mutex`` is the Big DRM Lock from legacy days and infested
141everything. Nowadays in modern drivers the only bit where it's mandatory is
142serializing GEM buffer object destruction. Which unfortunately means drivers
143have to keep track of that lock and either call ``unreference`` or
144``unreference_locked`` depending upon context.
145
146Core GEM doesn't have a need for ``struct_mutex`` any more since kernel 4.8,
147and there's a ``gem_free_object_unlocked`` callback for any drivers which are
148entirely ``struct_mutex`` free.
149
150For drivers that need ``struct_mutex`` it should be replaced with a driver-
151private lock. The tricky part is the BO free functions, since those can't
152reliably take that lock any more. Instead state needs to be protected with
153suitable subordinate locks or some cleanup work pushed to a worker thread. For
154performance-critical drivers it might also be better to go with a more
155fine-grained per-buffer object and per-context lockings scheme. Currently only the
156``msm`` driver still use ``struct_mutex``.
157
158Contact: Daniel Vetter, respective driver maintainers
159
160Convert instances of dev_info/dev_err/dev_warn to their DRM_DEV_* equivalent
161----------------------------------------------------------------------------
162
163For drivers which could have multiple instances, it is necessary to
164differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
165don't do this, drivers used dev_info/warn/err to make this differentiation. We
166now have DRM_DEV_* variants of the drm print macros, so we can start to convert
167those drivers back to using drm-formwatted specific log messages.
168
169Before you start this conversion please contact the relevant maintainers to make
170sure your work will be merged - not everyone agrees that the DRM dmesg macros
171are better.
172
173Contact: Sean Paul, Maintainer of the driver you plan to convert
174
175Convert drivers to use simple modeset suspend/resume
176----------------------------------------------------
177
178Most drivers (except i915 and nouveau) that use
179drm_atomic_helper_suspend/resume() can probably be converted to use
180drm_mode_config_helper_suspend/resume(). Also there's still open-coded version
181of the atomic suspend/resume code in older atomic modeset drivers.
182
183Contact: Maintainer of the driver you plan to convert
184
185Convert drivers to use drm_fb_helper_fbdev_setup/teardown()
186-----------------------------------------------------------
187
188Most drivers can use drm_fb_helper_fbdev_setup() except maybe:
189
190- amdgpu which has special logic to decide whether to call
191  drm_helper_disable_unused_functions()
192
193- armada which isn't atomic and doesn't call
194  drm_helper_disable_unused_functions()
195
196- i915 which calls drm_fb_helper_initial_config() in a worker
197
198Drivers that use drm_framebuffer_remove() to clean up the fbdev framebuffer can
199probably use drm_fb_helper_fbdev_teardown().
200
201Contact: Maintainer of the driver you plan to convert
202
203Clean up mmap forwarding
204------------------------
205
206A lot of drivers forward gem mmap calls to dma-buf mmap for imported buffers.
207And also a lot of them forward dma-buf mmap to the gem mmap implementations.
208Would be great to refactor this all into a set of small common helpers.
209
210Contact: Daniel Vetter
211
212Put a reservation_object into drm_gem_object
213--------------------------------------------
214
215This would remove the need for the ->gem_prime_res_obj callback. It would also
216allow us to implement generic helpers for waiting for a bo, allowing for quite a
217bit of refactoring in the various wait ioctl implementations.
218
219Contact: Daniel Vetter
220
221idr_init_base()
222---------------
223
224DRM core&drivers uses a lot of idr (integer lookup directories) for mapping
225userspace IDs to internal objects, and in most places ID=0 means NULL and hence
226is never used. Switching to idr_init_base() for these would make the idr more
227efficient.
228
229Contact: Daniel Vetter
230
231Defaults for .gem_prime_import and export
232-----------------------------------------
233
234Most drivers don't need to set drm_driver->gem_prime_import and
235->gem_prime_export now that drm_gem_prime_import() and drm_gem_prime_export()
236are the default.
237
238struct drm_gem_object_funcs
239---------------------------
240
241GEM objects can now have a function table instead of having the callbacks on the
242DRM driver struct. This is now the preferred way and drivers can be moved over.
243
244Use DRM_MODESET_LOCK_ALL_* helpers instead of boilerplate
245---------------------------------------------------------
246
247For cases where drivers are attempting to grab the modeset locks with a local
248acquire context. Replace the boilerplate code surrounding
249drm_modeset_lock_all_ctx() with DRM_MODESET_LOCK_ALL_BEGIN() and
250DRM_MODESET_LOCK_ALL_END() instead.
251
252This should also be done for all places where drm_modest_lock_all() is still
253used.
254
255As a reference, take a look at the conversions already completed in drm core.
256
257Contact: Sean Paul, respective driver maintainers
258
259Core refactorings
260=================
261
262Clean up the DRM header mess
263----------------------------
264
265The DRM subsystem originally had only one huge global header, ``drmP.h``. This
266is now split up, but many source files still include it. The remaining part of
267the cleanup work here is to replace any ``#include <drm/drmP.h>`` by only the
268headers needed (and fixing up any missing pre-declarations in the headers).
269
270In the end no .c file should need to include ``drmP.h`` anymore.
271
272Contact: Daniel Vetter
273
274Add missing kerneldoc for exported functions
275--------------------------------------------
276
277The DRM reference documentation is still lacking kerneldoc in a few areas. The
278task would be to clean up interfaces like moving functions around between
279files to better group them and improving the interfaces like dropping return
280values for functions that never fail. Then write kerneldoc for all exported
281functions and an overview section and integrate it all into the drm book.
282
283See https://dri.freedesktop.org/docs/drm/ for what's there already.
284
285Contact: Daniel Vetter
286
287Make panic handling work
288------------------------
289
290This is a really varied tasks with lots of little bits and pieces:
291
292* The panic path can't be tested currently, leading to constant breaking. The
293  main issue here is that panics can be triggered from hardirq contexts and
294  hence all panic related callback can run in hardirq context. It would be
295  awesome if we could test at least the fbdev helper code and driver code by
296  e.g. trigger calls through drm debugfs files. hardirq context could be
297  achieved by using an IPI to the local processor.
298
299* There's a massive confusion of different panic handlers. DRM fbdev emulation
300  helpers have one, but on top of that the fbcon code itself also has one. We
301  need to make sure that they stop fighting over each another.
302
303* ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
304  isn't a full solution for panic paths. We need to make sure that it only
305  returns true if there's a panic going on for real, and fix up all the
306  fallout.
307
308* The panic handler must never sleep, which also means it can't ever
309  ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
310  even spinlocks (because NMI and hardirq can panic too). We need to either
311  make sure to not call such paths, or trylock everything. Really tricky.
312
313* For the above locking troubles reasons it's pretty much impossible to
314  attempt a synchronous modeset from panic handlers. The only thing we could
315  try to achive is an atomic ``set_base`` of the primary plane, and hope that
316  it shows up. Everything else probably needs to be delayed to some worker or
317  something else which happens later on. Otherwise it just kills the box
318  harder, prevent the panic from going out on e.g. netconsole.
319
320* There's also proposal for a simplied DRM console instead of the full-blown
321  fbcon and DRM fbdev emulation. Any kind of panic handling tricks should
322  obviously work for both console, in case we ever get kmslog merged.
323
324Contact: Daniel Vetter
325
326Clean up the debugfs support
327----------------------------
328
329There's a bunch of issues with it:
330
331- The drm_info_list ->show() function doesn't even bother to cast to the drm
332  structure for you. This is lazy.
333
334- We probably want to have some support for debugfs files on crtc/connectors and
335  maybe other kms objects directly in core. There's even drm_print support in
336  the funcs for these objects to dump kms state, so it's all there. And then the
337  ->show() functions should obviously give you a pointer to the right object.
338
339- The drm_info_list stuff is centered on drm_minor instead of drm_device. For
340  anything we want to print drm_device (or maybe drm_file) is the right thing.
341
342- The drm_driver->debugfs_init hooks we have is just an artifact of the old
343  midlayered load sequence. DRM debugfs should work more like sysfs, where you
344  can create properties/files for an object anytime you want, and the core
345  takes care of publishing/unpuplishing all the files at register/unregister
346  time. Drivers shouldn't need to worry about these technicalities, and fixing
347  this (together with the drm_minor->drm_device move) would allow us to remove
348  debugfs_init.
349
350Contact: Daniel Vetter
351
352KMS cleanups
353------------
354
355Some of these date from the very introduction of KMS in 2008 ...
356
357- drm_mode_config.crtc_idr is misnamed, since it contains all KMS object. Should
358  be renamed to drm_mode_config.object_idr.
359
360- drm_display_mode doesn't need to be derived from drm_mode_object. That's
361  leftovers from older (never merged into upstream) KMS designs where modes
362  where set using their ID, including support to add/remove modes.
363
364- Make ->funcs and ->helper_private vtables optional. There's a bunch of empty
365  function tables in drivers, but before we can remove them we need to make sure
366  that all the users in helpers and drivers do correctly check for a NULL
367  vtable.
368
369- Cleanup up the various ->destroy callbacks. A lot of them just wrapt the
370  drm_*_cleanup implementations and can be removed. Some tack a kfree() at the
371  end, for which we could add drm_*_cleanup_kfree(). And then there's the (for
372  historical reasons) misnamed drm_primary_helper_destroy() function.
373
374Better Testing
375==============
376
377Enable trinity for DRM
378----------------------
379
380And fix up the fallout. Should be really interesting ...
381
382Make KMS tests in i-g-t generic
383-------------------------------
384
385The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
386including tons of testcases for corner-cases in the modesetting API. It would
387be awesome if those tests (at least the ones not relying on Intel-specific GEM
388features) could be made to run on any KMS driver.
389
390Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
391converting things over. For modeset tests we also first need a bit of
392infrastructure to use dumb buffers for untiled buffers, to be able to run all
393the non-i915 specific modeset tests.
394
395Extend virtual test driver (VKMS)
396---------------------------------
397
398See the documentation of :ref:`VKMS <vkms>` for more details. This is an ideal
399internship task, since it only requires a virtual machine and can be sized to
400fit the available time.
401
402Contact: Daniel Vetter
403
404Driver Specific
405===============
406
407tinydrm
408-------
409
410Tinydrm is the helper driver for really simple fb drivers. The goal is to make
411those drivers as simple as possible, so lots of room for refactoring:
412
413- backlight helpers, probably best to put them into a new drm_backlight.c.
414  This is because drivers/video is de-facto unmaintained. We could also
415  move drivers/video/backlight to drivers/gpu/backlight and take it all
416  over within drm-misc, but that's more work. Backlight helpers require a fair
417  bit of reworking and refactoring. A simple example is the enabling of a backlight.
418  Tinydrm has helpers for this. It would be good if other drivers can also use the
419  helper. However, there are various cases we need to consider i.e different
420  drivers seem to have different ways of enabling/disabling a backlight.
421  We also need to consider the backlight drivers (like gpio_backlight). The situation
422  is further complicated by the fact that the backlight is tied to fbdev
423  via fb_notifier_callback() which has complicated logic. For further details, refer
424  to the following discussion thread:
425  https://groups.google.com/forum/#!topic/outreachy-kernel/8rBe30lwtdA
426
427- spi helpers, probably best put into spi core/helper code. Thierry said
428  the spi maintainer is fast&reactive, so shouldn't be a big issue.
429
430- extract the mipi-dbi helper (well, the non-tinydrm specific parts at
431  least) into a separate helper, like we have for mipi-dsi already. Or follow
432  one of the ideas for having a shared dsi/dbi helper, abstracting away the
433  transport details more.
434
435- tinydrm_gem_cma_prime_import_sg_table should probably go into the cma
436  helpers, as a _vmapped variant (since not every driver needs the vmap).
437  And tinydrm_gem_cma_free_object could the be merged into
438  drm_gem_cma_free_object().
439
440- tinydrm_fb_create we could move into drm_simple_pipe, only need to add
441  the fb_create hook to drm_simple_pipe_funcs, which would again simplify a
442  bunch of things (since it gives you a one-stop vfunc for simple drivers).
443
444- Quick aside: The unregister devm stuff is kinda getting the lifetimes of
445  a drm_device wrong. Doesn't matter, since everyone else gets it wrong
446  too :-)
447
448- also rework the drm_framebuffer_funcs->dirty hook wire-up, see above.
449
450Contact: Noralf Trønnes, Daniel Vetter
451
452AMD DC Display Driver
453---------------------
454
455AMD DC is the display driver for AMD devices starting with Vega. There has been
456a bunch of progress cleaning it up but there's still plenty of work to be done.
457
458See drivers/gpu/drm/amd/display/TODO for tasks.
459
460Contact: Harry Wentland, Alex Deucher
461
462i915
463----
464
465- Our early/late pm callbacks could be removed in favour of using
466  device_link_add to model the dependency between i915 and snd_had. See
467  https://dri.freedesktop.org/docs/drm/driver-api/device_link.html
468
469Outside DRM
470===========
471