7 This section contains a list of smaller janitorial tasks in the kernel DRM
8 graphics subsystem useful as newbie projects. Or for slow rainy days.
13 To make it easier task are categorized into different levels:
15 Starter: Good tasks to get started with the DRM subsystem.
17 Intermediate: Tasks which need some experience with working in the DRM
18 subsystem, or some specific GPU/display graphics knowledge. For debugging issue
19 it's good to have the relevant hardware (or a virtual driver set up) available
22 Advanced: Tricky tasks that need fairly good understanding of the DRM subsystem
23 and graphics topics. Generally need the relevant hardware for development and
26 Subsystem-wide refactorings
27 ===========================
29 Remove custom dumb_map_offset implementations
30 ---------------------------------------------
32 All GEM based drivers should be using drm_gem_create_mmap_offset() instead.
33 Audit each individual driver, make sure it'll work with the generic
34 implementation (there's lots of outdated locking leftovers in various
35 implementations), and then remove it.
37 Contact: Daniel Vetter, respective driver maintainers
41 Convert existing KMS drivers to atomic modesetting
42 --------------------------------------------------
44 3.19 has the atomic modeset interfaces and helpers, so drivers can now be
45 converted over. Modern compositors like Wayland or Surfaceflinger on Android
46 really want an atomic modeset interface, so this is all about the bright
49 There is a conversion guide for atomic and all you need is a GPU for a
50 non-converted driver (again virtual HW drivers for KVM are still all
53 As part of this drivers also need to convert to universal plane (which means
54 exposing primary & cursor as proper plane objects). But that's much easier to
55 do by directly using the new atomic helper driver callbacks.
57 Contact: Daniel Vetter, respective driver maintainers
61 Clean up the clipped coordination confusion around planes
62 ---------------------------------------------------------
64 We have a helper to get this right with drm_plane_helper_check_update(), but
65 it's not consistently used. This should be fixed, preferrably in the atomic
66 helpers (and drivers then moved over to clipped coordinates). Probably the
67 helper should also be moved from drm_plane_helper.c to the atomic helpers, to
68 avoid confusion - the other helpers in that file are all deprecated legacy
71 Contact: Ville Syrjälä, Daniel Vetter, driver maintainers
75 Improve plane atomic_check helpers
76 ----------------------------------
78 Aside from the clipped coordinates right above there's a few suboptimal things
79 with the current helpers:
81 - drm_plane_helper_funcs->atomic_check gets called for enabled or disabled
82 planes. At best this seems to confuse drivers, worst it means they blow up
83 when the plane is disabled without the CRTC. The only special handling is
84 resetting values in the plane state structures, which instead should be moved
85 into the drm_plane_funcs->atomic_duplicate_state functions.
87 - Once that's done, helpers could stop calling ->atomic_check for disabled
90 - Then we could go through all the drivers and remove the more-or-less confused
91 checks for plane_state->fb and plane_state->crtc.
93 Contact: Daniel Vetter
97 Convert early atomic drivers to async commit helpers
98 ----------------------------------------------------
100 For the first year the atomic modeset helpers didn't support asynchronous /
101 nonblocking commits, and every driver had to hand-roll them. This is fixed
102 now, but there's still a pile of existing drivers that easily could be
103 converted over to the new infrastructure.
105 One issue with the helpers is that they require that drivers handle completion
106 events for atomic commits correctly. But fixing these bugs is good anyway.
108 Somewhat related is the legacy_cursor_update hack, which should be replaced with
109 the new atomic_async_check/commit functionality in the helpers in drivers that
110 still look at that flag.
112 Contact: Daniel Vetter, respective driver maintainers
116 Fallout from atomic KMS
117 -----------------------
119 ``drm_atomic_helper.c`` provides a batch of functions which implement legacy
120 IOCTLs on top of the new atomic driver interface. Which is really nice for
121 gradual conversion of drivers, but unfortunately the semantic mismatches are
122 a bit too severe. So there's some follow-up work to adjust the function
123 interfaces to fix these issues:
125 * atomic needs the lock acquire context. At the moment that's passed around
126 implicitly with some horrible hacks, and it's also allocate with
127 ``GFP_NOFAIL`` behind the scenes. All legacy paths need to start allocating
128 the acquire context explicitly on stack and then also pass it down into
129 drivers explicitly so that the legacy-on-atomic functions can use them.
131 Except for some driver code this is done. This task should be finished by
132 adding WARN_ON(!drm_drv_uses_atomic_modeset) in drm_modeset_lock_all().
134 * A bunch of the vtable hooks are now in the wrong place: DRM has a split
135 between core vfunc tables (named ``drm_foo_funcs``), which are used to
136 implement the userspace ABI. And then there's the optional hooks for the
137 helper libraries (name ``drm_foo_helper_funcs``), which are purely for
138 internal use. Some of these hooks should be move from ``_funcs`` to
139 ``_helper_funcs`` since they are not part of the core ABI. There's a
140 ``FIXME`` comment in the kerneldoc for each such case in ``drm_crtc.h``.
142 Contact: Daniel Vetter
146 Get rid of dev->struct_mutex from GEM drivers
147 ---------------------------------------------
149 ``dev->struct_mutex`` is the Big DRM Lock from legacy days and infested
150 everything. Nowadays in modern drivers the only bit where it's mandatory is
151 serializing GEM buffer object destruction. Which unfortunately means drivers
152 have to keep track of that lock and either call ``unreference`` or
153 ``unreference_locked`` depending upon context.
155 Core GEM doesn't have a need for ``struct_mutex`` any more since kernel 4.8,
156 and there's a GEM object ``free`` callback for any drivers which are
157 entirely ``struct_mutex`` free.
159 For drivers that need ``struct_mutex`` it should be replaced with a driver-
160 private lock. The tricky part is the BO free functions, since those can't
161 reliably take that lock any more. Instead state needs to be protected with
162 suitable subordinate locks or some cleanup work pushed to a worker thread. For
163 performance-critical drivers it might also be better to go with a more
164 fine-grained per-buffer object and per-context lockings scheme. Currently only
165 the ``msm`` and `i915` drivers use ``struct_mutex``.
167 Contact: Daniel Vetter, respective driver maintainers
171 Convert logging to drm_* functions with drm_device paramater
172 ------------------------------------------------------------
174 For drivers which could have multiple instances, it is necessary to
175 differentiate between which is which in the logs. Since DRM_INFO/WARN/ERROR
176 don't do this, drivers used dev_info/warn/err to make this differentiation. We
177 now have drm_* variants of the drm print functions, so we can start to convert
178 those drivers back to using drm-formatted specific log messages.
180 Before you start this conversion please contact the relevant maintainers to make
181 sure your work will be merged - not everyone agrees that the DRM dmesg macros
184 Contact: Sean Paul, Maintainer of the driver you plan to convert
188 Convert drivers to use simple modeset suspend/resume
189 ----------------------------------------------------
191 Most drivers (except i915 and nouveau) that use
192 drm_atomic_helper_suspend/resume() can probably be converted to use
193 drm_mode_config_helper_suspend/resume(). Also there's still open-coded version
194 of the atomic suspend/resume code in older atomic modeset drivers.
196 Contact: Maintainer of the driver you plan to convert
200 Convert drivers to use drm_fbdev_generic_setup()
201 ------------------------------------------------
203 Most drivers can use drm_fbdev_generic_setup(). Driver have to implement
204 atomic modesetting and GEM vmap support. Historically, generic fbdev emulation
205 expected the framebuffer in system memory or system-like memory. By employing
206 struct dma_buf_map, drivers with frambuffers in I/O memory can be supported
209 Contact: Maintainer of the driver you plan to convert
213 Reimplement functions in drm_fbdev_fb_ops without fbdev
214 -------------------------------------------------------
216 A number of callback functions in drm_fbdev_fb_ops could benefit from
217 being rewritten without dependencies on the fbdev module. Some of the
218 helpers could further benefit from using struct dma_buf_map instead of
221 Contact: Thomas Zimmermann <tzimmermann@suse.de>, Daniel Vetter
226 drm_framebuffer_funcs and drm_mode_config_funcs.fb_create cleanup
227 -----------------------------------------------------------------
229 A lot more drivers could be switched over to the drm_gem_framebuffer helpers.
232 - Need to switch over to the generic dirty tracking code using
233 drm_atomic_helper_dirtyfb first (e.g. qxl).
235 - Need to switch to drm_fbdev_generic_setup(), otherwise a lot of the custom fb
236 setup code can't be deleted.
238 - Many drivers wrap drm_gem_fb_create() only to check for valid formats. For
239 atomic drivers we could check for valid formats by calling
240 drm_plane_check_pixel_format() against all planes, and pass if any plane
241 supports the format. For non-atomic that's not possible since like the format
242 list for the primary plane is fake and we'd therefor reject valid formats.
244 - Many drivers subclass drm_framebuffer, we'd need a embedding compatible
245 version of the varios drm_gem_fb_create functions. Maybe called
246 drm_gem_fb_create/_with_dirty/_with_funcs as needed.
248 Contact: Daniel Vetter
252 Clean up mmap forwarding
253 ------------------------
255 A lot of drivers forward gem mmap calls to dma-buf mmap for imported buffers.
256 And also a lot of them forward dma-buf mmap to the gem mmap implementations.
257 There's drm_gem_prime_mmap() for this now, but still needs to be rolled out.
259 Contact: Daniel Vetter
263 Generic fbdev defio support
264 ---------------------------
266 The defio support code in the fbdev core has some very specific requirements,
267 which means drivers need to have a special framebuffer for fbdev. The main
268 issue is that it uses some fields in struct page itself, which breaks shmem
269 gem objects (and other things). To support defio, affected drivers require
270 the use of a shadow buffer, which may add CPU and memory overhead.
272 Possible solution would be to write our own defio mmap code in the drm fbdev
273 emulation. It would need to fully wrap the existing mmap ops, forwarding
274 everything after it has done the write-protect/mkwrite trickery:
276 - In the drm_fbdev_fb_mmap helper, if we need defio, change the
277 default page prots to write-protected with something like this::
279 vma->vm_page_prot = pgprot_wrprotect(vma->vm_page_prot);
281 - Set the mkwrite and fsync callbacks with similar implementions to the core
282 fbdev defio stuff. These should all work on plain ptes, they don't actually
283 require a struct page. uff. These should all work on plain ptes, they don't
284 actually require a struct page.
286 - Track the dirty pages in a separate structure (bitfield with one bit per page
287 should work) to avoid clobbering struct page.
289 Might be good to also have some igt testcases for this.
291 Contact: Daniel Vetter, Noralf Tronnes
295 Garbage collect fbdev scrolling acceleration
296 --------------------------------------------
298 Scroll acceleration is disabled in fbcon by hard-wiring p->scrollmode =
299 SCROLL_REDRAW. There's a ton of code this will allow us to remove:
301 - lots of code in fbcon.c
303 - a bunch of the hooks in fbcon_ops, maybe the remaining hooks could be called
304 directly instead of the function table (with a switch on p->rotate)
306 - fb_copyarea is unused after this, and can be deleted from all drivers
308 Note that not all acceleration code can be deleted, since clearing and cursor
309 support is still accelerated, which might be good candidates for further
312 Contact: Daniel Vetter
319 DRM core&drivers uses a lot of idr (integer lookup directories) for mapping
320 userspace IDs to internal objects, and in most places ID=0 means NULL and hence
321 is never used. Switching to idr_init_base() for these would make the idr more
324 Contact: Daniel Vetter
328 struct drm_gem_object_funcs
329 ---------------------------
331 GEM objects can now have a function table instead of having the callbacks on the
332 DRM driver struct. This is now the preferred way. Callbacks in drivers have been
333 converted, except for struct drm_driver.gem_prime_mmap.
337 Use DRM_MODESET_LOCK_ALL_* helpers instead of boilerplate
338 ---------------------------------------------------------
340 For cases where drivers are attempting to grab the modeset locks with a local
341 acquire context. Replace the boilerplate code surrounding
342 drm_modeset_lock_all_ctx() with DRM_MODESET_LOCK_ALL_BEGIN() and
343 DRM_MODESET_LOCK_ALL_END() instead.
345 This should also be done for all places where drm_modeset_lock_all() is still
348 As a reference, take a look at the conversions already completed in drm core.
350 Contact: Sean Paul, respective driver maintainers
354 Rename CMA helpers to DMA helpers
355 ---------------------------------
357 CMA (standing for contiguous memory allocator) is really a bit an accident of
358 what these were used for first, a much better name would be DMA helpers. In the
359 text these should even be called coherent DMA memory helpers (so maybe CDM, but
360 no one knows what that means) since underneath they just use dma_alloc_coherent.
362 Contact: Laurent Pinchart, Daniel Vetter
364 Level: Intermediate (mostly because it is a huge tasks without good partial
365 milestones, not technically itself that challenging)
367 connector register/unregister fixes
368 -----------------------------------
370 - For most connectors it's a no-op to call drm_connector_register/unregister
371 directly from driver code, drm_dev_register/unregister take care of this
372 already. We can remove all of them.
374 - For dp drivers it's a bit more a mess, since we need the connector to be
375 registered when calling drm_dp_aux_register. Fix this by instead calling
376 drm_dp_aux_init, and moving the actual registering into a late_register
377 callback as recommended in the kerneldoc.
381 Remove load/unload callbacks from all non-DRIVER_LEGACY drivers
382 ---------------------------------------------------------------
384 The load/unload callbacks in struct &drm_driver are very much midlayers, plus
385 for historical reasons they get the ordering wrong (and we can't fix that)
386 between setting up the &drm_driver structure and calling drm_dev_register().
388 - Rework drivers to no longer use the load/unload callbacks, directly coding the
389 load/unload sequence into the driver's probe function.
391 - Once all non-DRIVER_LEGACY drivers are converted, disallow the load/unload
392 callbacks for all modern drivers.
394 Contact: Daniel Vetter
398 Replace drm_detect_hdmi_monitor() with drm_display_info.is_hdmi
399 ---------------------------------------------------------------
401 Once EDID is parsed, the monitor HDMI support information is available through
402 drm_display_info.is_hdmi. Many drivers still call drm_detect_hdmi_monitor() to
403 retrieve the same information, which is less efficient.
405 Audit each individual driver calling drm_detect_hdmi_monitor() and switch to
406 drm_display_info.is_hdmi if applicable.
408 Contact: Laurent Pinchart, respective driver maintainers
412 Consolidate custom driver modeset properties
413 --------------------------------------------
415 Before atomic modeset took place, many drivers where creating their own
416 properties. Among other things, atomic brought the requirement that custom,
417 driver specific properties should not be used.
419 For this task, we aim to introduce core helpers or reuse the existing ones
422 A quick, unconfirmed, examples list.
424 Introduce core helpers:
425 - audio (amdgpu, intel, gma500, radeon)
426 - brightness, contrast, etc (armada, nouveau) - overlay only (?)
427 - broadcast rgb (gma500, intel)
428 - colorkey (armada, nouveau, rcar) - overlay only (?)
429 - dither (amdgpu, nouveau, radeon) - varies across drivers
430 - underscan family (amdgpu, radeon, nouveau)
434 - tv format names, enhancements (gma500, intel)
435 - tv overscan, margins, etc. (gma500, intel)
436 - zorder (omapdrm) - same as zpos (?)
439 Contact: Emil Velikov, respective driver maintainers
443 Plumb drm_atomic_state all over
444 -------------------------------
446 Currently various atomic functions take just a single or a handful of
447 object states (eg. plane state). While that single object state can
448 suffice for some simple cases, we often have to dig out additional
449 object states for dealing with various dependencies between the individual
450 objects or the hardware they represent. The process of digging out the
451 additional states is rather non-intuitive and error prone.
453 To fix that most functions should rather take the overall
454 drm_atomic_state as one of their parameters. The other parameters
455 would generally be the object(s) we mainly want to interact with.
457 For example, instead of
461 int (*atomic_check)(struct drm_plane *plane, struct drm_plane_state *state);
463 we would have something like
467 int (*atomic_check)(struct drm_plane *plane, struct drm_atomic_state *state);
469 The implementation can then trivially gain access to any required object
470 state(s) via drm_atomic_get_plane_state(), drm_atomic_get_new_plane_state(),
471 drm_atomic_get_old_plane_state(), and their equivalents for
474 Additionally many drivers currently access the object->state pointer
475 directly in their commit functions. That is not going to work if we
476 eg. want to allow deeper commit pipelines as those pointers could
477 then point to the states corresponding to a future commit instead of
478 the current commit we're trying to process. Also non-blocking commits
479 execute locklessly so there are serious concerns with dereferencing
480 the object->state pointers without holding the locks that protect them.
481 Use of drm_atomic_get_new_plane_state(), drm_atomic_get_old_plane_state(),
482 etc. avoids these problems as well since they relate to a specific
483 commit via the passed in drm_atomic_state.
485 Contact: Ville Syrjälä, Daniel Vetter
489 Use struct dma_buf_map throughout codebase
490 ------------------------------------------
492 Pointers to shared device memory are stored in struct dma_buf_map. Each
493 instance knows whether it refers to system or I/O memory. Most of the DRM-wide
494 interface have been converted to use struct dma_buf_map, but implementations
495 often still use raw pointers.
497 The task is to use struct dma_buf_map where it makes sense.
499 * Memory managers should use struct dma_buf_map for dma-buf-imported buffers.
500 * TTM might benefit from using struct dma_buf_map internally.
501 * Framebuffer copying and blitting helpers should operate on struct dma_buf_map.
503 Contact: Thomas Zimmermann <tzimmermann@suse.de>, Christian König, Daniel Vetter
511 Make panic handling work
512 ------------------------
514 This is a really varied tasks with lots of little bits and pieces:
516 * The panic path can't be tested currently, leading to constant breaking. The
517 main issue here is that panics can be triggered from hardirq contexts and
518 hence all panic related callback can run in hardirq context. It would be
519 awesome if we could test at least the fbdev helper code and driver code by
520 e.g. trigger calls through drm debugfs files. hardirq context could be
521 achieved by using an IPI to the local processor.
523 * There's a massive confusion of different panic handlers. DRM fbdev emulation
524 helpers have one, but on top of that the fbcon code itself also has one. We
525 need to make sure that they stop fighting over each another.
527 * ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
528 isn't a full solution for panic paths. We need to make sure that it only
529 returns true if there's a panic going on for real, and fix up all the
532 * The panic handler must never sleep, which also means it can't ever
533 ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
534 even spinlocks (because NMI and hardirq can panic too). We need to either
535 make sure to not call such paths, or trylock everything. Really tricky.
537 * For the above locking troubles reasons it's pretty much impossible to
538 attempt a synchronous modeset from panic handlers. The only thing we could
539 try to achive is an atomic ``set_base`` of the primary plane, and hope that
540 it shows up. Everything else probably needs to be delayed to some worker or
541 something else which happens later on. Otherwise it just kills the box
542 harder, prevent the panic from going out on e.g. netconsole.
544 * There's also proposal for a simplied DRM console instead of the full-blown
545 fbcon and DRM fbdev emulation. Any kind of panic handling tricks should
546 obviously work for both console, in case we ever get kmslog merged.
548 Contact: Daniel Vetter
552 Clean up the debugfs support
553 ----------------------------
555 There's a bunch of issues with it:
557 - The drm_info_list ->show() function doesn't even bother to cast to the drm
558 structure for you. This is lazy.
560 - We probably want to have some support for debugfs files on crtc/connectors and
561 maybe other kms objects directly in core. There's even drm_print support in
562 the funcs for these objects to dump kms state, so it's all there. And then the
563 ->show() functions should obviously give you a pointer to the right object.
565 - The drm_info_list stuff is centered on drm_minor instead of drm_device. For
566 anything we want to print drm_device (or maybe drm_file) is the right thing.
568 - The drm_driver->debugfs_init hooks we have is just an artifact of the old
569 midlayered load sequence. DRM debugfs should work more like sysfs, where you
570 can create properties/files for an object anytime you want, and the core
571 takes care of publishing/unpuplishing all the files at register/unregister
572 time. Drivers shouldn't need to worry about these technicalities, and fixing
573 this (together with the drm_minor->drm_device move) would allow us to remove
576 Contact: Daniel Vetter
583 Some of these date from the very introduction of KMS in 2008 ...
585 - Make ->funcs and ->helper_private vtables optional. There's a bunch of empty
586 function tables in drivers, but before we can remove them we need to make sure
587 that all the users in helpers and drivers do correctly check for a NULL
590 - Cleanup up the various ->destroy callbacks. A lot of them just wrapt the
591 drm_*_cleanup implementations and can be removed. Some tack a kfree() at the
592 end, for which we could add drm_*_cleanup_kfree(). And then there's the (for
593 historical reasons) misnamed drm_primary_helper_destroy() function.
600 Enable trinity for DRM
601 ----------------------
603 And fix up the fallout. Should be really interesting ...
607 Make KMS tests in i-g-t generic
608 -------------------------------
610 The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
611 including tons of testcases for corner-cases in the modesetting API. It would
612 be awesome if those tests (at least the ones not relying on Intel-specific GEM
613 features) could be made to run on any KMS driver.
615 Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
616 converting things over. For modeset tests we also first need a bit of
617 infrastructure to use dumb buffers for untiled buffers, to be able to run all
618 the non-i915 specific modeset tests.
622 Extend virtual test driver (VKMS)
623 ---------------------------------
625 See the documentation of :ref:`VKMS <vkms>` for more details. This is an ideal
626 internship task, since it only requires a virtual machine and can be sized to
627 fit the available time.
629 Contact: Daniel Vetter
633 Backlight Refactoring
634 ---------------------
636 Backlight drivers have a triple enable/disable state, which is a bit overkill.
639 1. Roll out backlight_enable() and backlight_disable() helpers everywhere. This
641 2. In all, only look at one of the three status bits set by the above helpers.
642 3. Remove the other two status bits.
644 Contact: Daniel Vetter
651 AMD DC Display Driver
652 ---------------------
654 AMD DC is the display driver for AMD devices starting with Vega. There has been
655 a bunch of progress cleaning it up but there's still plenty of work to be done.
657 See drivers/gpu/drm/amd/display/TODO for tasks.
659 Contact: Harry Wentland, Alex Deucher
664 There is support in place now for writing internal DRM clients making it
665 possible to pick up the bootsplash work that was rejected because it was written
668 - [v6,8/8] drm/client: Hack: Add bootsplash example
669 https://patchwork.freedesktop.org/patch/306579/
671 - [RFC PATCH v2 00/13] Kernel based bootsplash
672 https://lkml.org/lkml/2017/12/13/764
674 Contact: Sam Ravnborg
681 Convert fbdev drivers to DRM
682 ----------------------------
684 There are plenty of fbdev drivers for older hardware. Some hwardware has
685 become obsolete, but some still provides good(-enough) framebuffers. The
686 drivers that are still useful should be converted to DRM and afterwards
689 Very simple fbdev drivers can best be converted by starting with a new
690 DRM driver. Simple KMS helpers and SHMEM should be able to handle any
691 existing hardware. The new driver's call-back functions are filled from
694 More complex fbdev drivers can be refactored step-by-step into a DRM
695 driver with the help of the DRM fbconv helpers. [1] These helpers provide
696 the transition layer between the DRM core infrastructure and the fbdev
697 driver interface. Create a new DRM driver on top of the fbconv helpers,
698 copy over the fbdev driver, and hook it up to the DRM code. Examples for
699 several fbdev drivers are available at [1] and a tutorial of this process
700 available at [2]. The result is a primitive DRM driver that can run X11
703 - [1] https://gitlab.freedesktop.org/tzimmermann/linux/tree/fbconv
704 - [2] https://gitlab.freedesktop.org/tzimmermann/linux/blob/fbconv/drivers/gpu/drm/drm_fbconv_helper.c
706 Contact: Thomas Zimmermann <tzimmermann@suse.de>
projects / linux-2.6-microblaze.git / blob