Documentation/gpu/todo.rst

   1 .. _todo:
   2
   3 =========
   4 TODO list
   5 =========
   6
   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.
   9
  10 Subsystem-wide refactorings
  11 ===========================
  12
  13 De-midlayer drivers
  14 -------------------
  15
  16 With the recent ``drm_bus`` cleanup patches for 3.17 it is no longer required
  17 to 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``
  19 and ``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
  21 the new two-stage ``drm_device`` setup/teardown.
  22
  23 Once all existing drivers are converted we can also remove those bus support
  24 files for USB and platform devices.
  25
  26 All you need is a GPU for a non-converted driver (currently almost all of
  27 them, but also all the virtual ones used by KVM, so everyone qualifies).
  28
  29 Contact: Daniel Vetter, Thierry Reding, respective driver maintainers
  30
  31 Switch from reference/unreference to get/put
  32 --------------------------------------------
  33
  34 For some reason DRM core uses ``reference``/``unreference`` suffixes for
  35 refcounting functions, but kernel uses ``get``/``put`` (e.g.
  36 ``kref_get``/``put()``). It would be good to switch over for consistency, and
  37 it's shorter. Needs to be done in 3 steps for each pair of functions:
  38
  39 * Create new ``get``/``put`` functions, define the old names as compatibility
  40   wrappers
  41 * Switch over each file/driver using a cocci-generated spatch.
  42 * Once all users of the old names are gone, remove them.
  43
  44 This way drivers/patches in the progress of getting merged won't break.
  45
  46 Contact: Daniel Vetter
  47
  48 Convert existing KMS drivers to atomic modesetting
  49 --------------------------------------------------
  50
  51 3.19 has the atomic modeset interfaces and helpers, so drivers can now be
  52 converted over. Modern compositors like Wayland or Surfaceflinger on Android
  53 really want an atomic modeset interface, so this is all about the bright
  54 future.
  55
  56 There is a conversion guide for atomic and all you need is a GPU for a
  57 non-converted driver (again virtual HW drivers for KVM are still all
  58 suitable).
  59
  60 As part of this drivers also need to convert to universal plane (which means
  61 exposing primary & cursor as proper plane objects). But that's much easier to
  62 do by directly using the new atomic helper driver callbacks.
  63
  64 Contact: Daniel Vetter, respective driver maintainers
  65
  66 Clean up the clipped coordination confusion around planes
  67 ---------------------------------------------------------
  68
  69 We have a helper to get this right with drm_plane_helper_check_update(), but
  70 it's not consistently used. This should be fixed, preferrably in the atomic
  71 helpers (and drivers then moved over to clipped coordinates). Probably the
  72 helper should also be moved from drm_plane_helper.c to the atomic helpers, to
  73 avoid confusion - the other helpers in that file are all deprecated legacy
  74 helpers.
  75
  76 Contact: Ville Syrjälä, Daniel Vetter, driver maintainers
  77
  78 Implement deferred fbdev setup in the helper
  79 --------------------------------------------
  80
  81 Many (especially embedded drivers) want to delay fbdev setup until there's a
  82 real screen plugged in. This is to avoid the dreaded fallback to the low-res
  83 fbdev default. Many drivers have a hacked-up (and often broken) version of this,
  84 better to do it once in the shared helpers. Thierry has a patch series, but that
  85 one needs to be rebased and final polish applied.
  86
  87 Contact: Thierry Reding, Daniel Vetter, driver maintainers
  88
  89 Convert early atomic drivers to async commit helpers
  90 ----------------------------------------------------
  91
  92 For the first year the atomic modeset helpers didn't support asynchronous /
  93 nonblocking commits, and every driver had to hand-roll them. This is fixed
  94 now, but there's still a pile of existing drivers that easily could be
  95 converted over to the new infrastructure.
  96
  97 One issue with the helpers is that they require that drivers handle completion
  98 events for atomic commits correctly. But fixing these bugs is good anyway.
  99
 100 Contact: Daniel Vetter, respective driver maintainers
 101
 102 Better manual-upload support for atomic
 103 ---------------------------------------
 104
 105 This would be especially useful for tinydrm:
 106
 107 - Add a struct drm_rect dirty_clip to drm_crtc_state. When duplicating the
 108   crtc state, clear that to the max values, x/y = 0 and w/h = MAX_INT, in
 109   __drm_atomic_helper_crtc_duplicate_state().
 110
 111 - Move tinydrm_merge_clips into drm_framebuffer.c, dropping the tinydrm_
 112   prefix ofc and using drm_fb_. drm_framebuffer.c makes sense since this
 113   is a function useful to implement the fb->dirty function.
 114
 115 - Create a new drm_fb_dirty function which does essentially what e.g.
 116   mipi_dbi_fb_dirty does. You can use e.g. drm_atomic_helper_update_plane as the
 117   template. But instead of doing a simple full-screen plane update, this new
 118   helper also sets crtc_state->dirty_clip to the right coordinates. And of
 119   course it needs to check whether the fb is actually active (and maybe where),
 120   so there's some book-keeping involved. There's also some good fun involved in
 121   scaling things appropriately. For that case we might simply give up and
 122   declare the entire area covered by the plane as dirty.
 123
 124 Contact: Noralf Trønnes, Daniel Vetter
 125
 126 Fallout from atomic KMS
 127 -----------------------
 128
 129 ``drm_atomic_helper.c`` provides a batch of functions which implement legacy
 130 IOCTLs on top of the new atomic driver interface. Which is really nice for
 131 gradual conversion of drivers, but unfortunately the semantic mismatches are
 132 a bit too severe. So there's some follow-up work to adjust the function
 133 interfaces to fix these issues:
 134
 135 * atomic needs the lock acquire context. At the moment that's passed around
 136   implicitly with some horrible hacks, and it's also allocate with
 137   ``GFP_NOFAIL`` behind the scenes. All legacy paths need to start allocating
 138   the acquire context explicitly on stack and then also pass it down into
 139   drivers explicitly so that the legacy-on-atomic functions can use them.
 140
 141 * A bunch of the vtable hooks are now in the wrong place: DRM has a split
 142   between core vfunc tables (named ``drm_foo_funcs``), which are used to
 143   implement the userspace ABI. And then there's the optional hooks for the
 144   helper libraries (name ``drm_foo_helper_funcs``), which are purely for
 145   internal use. Some of these hooks should be move from ``_funcs`` to
 146   ``_helper_funcs`` since they are not part of the core ABI. There's a
 147   ``FIXME`` comment in the kerneldoc for each such case in ``drm_crtc.h``.
 148
 149 * There's a new helper ``drm_atomic_helper_best_encoder()`` which could be
 150   used by all atomic drivers which don't select the encoder for a given
 151   connector at runtime. That's almost all of them, and would allow us to get
 152   rid of a lot of ``best_encoder`` boilerplate in drivers.
 153
 154 Contact: Daniel Vetter
 155
 156 Get rid of dev->struct_mutex from GEM drivers
 157 ---------------------------------------------
 158
 159 ``dev->struct_mutex`` is the Big DRM Lock from legacy days and infested
 160 everything. Nowadays in modern drivers the only bit where it's mandatory is
 161 serializing GEM buffer object destruction. Which unfortunately means drivers
 162 have to keep track of that lock and either call ``unreference`` or
 163 ``unreference_locked`` depending upon context.
 164
 165 Core GEM doesn't have a need for ``struct_mutex`` any more since kernel 4.8,
 166 and there's a ``gem_free_object_unlocked`` callback for any drivers which are
 167 entirely ``struct_mutex`` free.
 168
 169 For drivers that need ``struct_mutex`` it should be replaced with a driver-
 170 private lock. The tricky part is the BO free functions, since those can't
 171 reliably take that lock any more. Instead state needs to be protected with
 172 suitable subordinate locks or some cleanup work pushed to a worker thread. For
 173 performance-critical drivers it might also be better to go with a more
 174 fine-grained per-buffer object and per-context lockings scheme. Currently the
 175 following drivers still use ``struct_mutex``: ``msm``, ``omapdrm`` and
 176 ``udl``.
 177
 178 Contact: Daniel Vetter, respective driver maintainers
 179
 180 Switch to drm_connector_list_iter for any connector_list walking
 181 ----------------------------------------------------------------
 182
 183 Connectors can be hotplugged, and we now have a special list of helpers to walk
 184 the connector_list in a race-free fashion, without incurring deadlocks on
 185 mutexes and other fun stuff.
 186
 187 Unfortunately most drivers are not converted yet. At least all those supporting
 188 DP MST hotplug should be converted, since for those drivers the difference
 189 matters. See drm_for_each_connector_iter() vs. drm_for_each_connector().
 190
 191 Contact: Daniel Vetter
 192
 193 Core refactorings
 194 =================
 195
 196 Use new IDR deletion interface to clean up drm_gem_handle_delete()
 197 ------------------------------------------------------------------
 198
 199 See the "This is gross" comment -- apparently the IDR system now can return an
 200 error code instead of oopsing.
 201
 202 Clean up the DRM header mess
 203 ----------------------------
 204
 205 Currently the DRM subsystem has only one global header, ``drmP.h``. This is
 206 used both for functions exported to helper libraries and drivers and functions
 207 only used internally in the ``drm.ko`` module. The goal would be to move all
 208 header declarations not needed outside of ``drm.ko`` into
 209 ``drivers/gpu/drm/drm_*_internal.h`` header files. ``EXPORT_SYMBOL`` also
 210 needs to be dropped for these functions.
 211
 212 This would nicely tie in with the below task to create kerneldoc after the API
 213 is cleaned up. Or with the "hide legacy cruft better" task.
 214
 215 Note that this is well in progress, but ``drmP.h`` is still huge. The updated
 216 plan is to switch to per-file driver API headers, which will also structure
 217 the kerneldoc better. This should also allow more fine-grained ``#include``
 218 directives.
 219
 220 In the end no .c file should need to include ``drmP.h`` anymore.
 221
 222 Contact: Daniel Vetter
 223
 224 Add missing kerneldoc for exported functions
 225 --------------------------------------------
 226
 227 The DRM reference documentation is still lacking kerneldoc in a few areas. The
 228 task would be to clean up interfaces like moving functions around between
 229 files to better group them and improving the interfaces like dropping return
 230 values for functions that never fail. Then write kerneldoc for all exported
 231 functions and an overview section and integrate it all into the drm DocBook.
 232
 233 See https://dri.freedesktop.org/docs/drm/ for what's there already.
 234
 235 Contact: Daniel Vetter
 236
 237 Hide legacy cruft better
 238 ------------------------
 239
 240 Way back DRM supported only drivers which shadow-attached to PCI devices with
 241 userspace or fbdev drivers setting up outputs. Modern DRM drivers take charge
 242 of the entire device, you can spot them with the DRIVER_MODESET flag.
 243
 244 Unfortunately there's still large piles of legacy code around which needs to
 245 be hidden so that driver writers don't accidentally end up using it. And to
 246 prevent security issues in those legacy IOCTLs from being exploited on modern
 247 drivers. This has multiple possible subtasks:
 248
 249 * Extract support code for legacy features into a ``drm-legacy.ko`` kernel
 250   module and compile it only when one of the legacy drivers is enabled.
 251
 252 This is mostly done, the only thing left is to split up ``drm_irq.c`` into
 253 legacy cruft and the parts needed by modern KMS drivers.
 254
 255 Contact: Daniel Vetter
 256
 257 Make panic handling work
 258 ------------------------
 259
 260 This is a really varied tasks with lots of little bits and pieces:
 261
 262 * The panic path can't be tested currently, leading to constant breaking. The
 263   main issue here is that panics can be triggered from hardirq contexts and
 264   hence all panic related callback can run in hardirq context. It would be
 265   awesome if we could test at least the fbdev helper code and driver code by
 266   e.g. trigger calls through drm debugfs files. hardirq context could be
 267   achieved by using an IPI to the local processor.
 268
 269 * There's a massive confusion of different panic handlers. DRM fbdev emulation
 270   helpers have one, but on top of that the fbcon code itself also has one. We
 271   need to make sure that they stop fighting over each another.
 272
 273 * ``drm_can_sleep()`` is a mess. It hides real bugs in normal operations and
 274   isn't a full solution for panic paths. We need to make sure that it only
 275   returns true if there's a panic going on for real, and fix up all the
 276   fallout.
 277
 278 * The panic handler must never sleep, which also means it can't ever
 279   ``mutex_lock()``. Also it can't grab any other lock unconditionally, not
 280   even spinlocks (because NMI and hardirq can panic too). We need to either
 281   make sure to not call such paths, or trylock everything. Really tricky.
 282
 283 * For the above locking troubles reasons it's pretty much impossible to
 284   attempt a synchronous modeset from panic handlers. The only thing we could
 285   try to achive is an atomic ``set_base`` of the primary plane, and hope that
 286   it shows up. Everything else probably needs to be delayed to some worker or
 287   something else which happens later on. Otherwise it just kills the box
 288   harder, prevent the panic from going out on e.g. netconsole.
 289
 290 * There's also proposal for a simplied DRM console instead of the full-blown
 291   fbcon and DRM fbdev emulation. Any kind of panic handling tricks should
 292   obviously work for both console, in case we ever get kmslog merged.
 293
 294 Contact: Daniel Vetter
 295
 296 Clean up the debugfs support
 297 ----------------------------
 298
 299 There's a bunch of issues with it:
 300
 301 - The drm_info_list ->show() function doesn't even bother to cast to the drm
 302   structure for you. This is lazy.
 303
 304 - We probably want to have some support for debugfs files on crtc/connectors and
 305   maybe other kms objects directly in core. There's even drm_print support in
 306   the funcs for these objects to dump kms state, so it's all there. And then the
 307   ->show() functions should obviously give you a pointer to the right object.
 308
 309 - The drm_info_list stuff is centered on drm_minor instead of drm_device. For
 310   anything we want to print drm_device (or maybe drm_file) is the right thing.
 311
 312 - The drm_driver->debugfs_init hooks we have is just an artifact of the old
 313   midlayered load sequence. DRM debugfs should work more like sysfs, where you
 314   can create properties/files for an object anytime you want, and the core
 315   takes care of publishing/unpuplishing all the files at register/unregister
 316   time. Drivers shouldn't need to worry about these technicalities, and fixing
 317   this (together with the drm_minor->drm_device move) would allow us to remove
 318   debugfs_init.
 319
 320 Contact: Daniel Vetter
 321
 322 Better Testing
 323 ==============
 324
 325 Enable trinity for DRM
 326 ----------------------
 327
 328 And fix up the fallout. Should be really interesting ...
 329
 330 Make KMS tests in i-g-t generic
 331 -------------------------------
 332
 333 The i915 driver team maintains an extensive testsuite for the i915 DRM driver,
 334 including tons of testcases for corner-cases in the modesetting API. It would
 335 be awesome if those tests (at least the ones not relying on Intel-specific GEM
 336 features) could be made to run on any KMS driver.
 337
 338 Basic work to run i-g-t tests on non-i915 is done, what's now missing is mass-
 339 converting things over. For modeset tests we also first need a bit of
 340 infrastructure to use dumb buffers for untiled buffers, to be able to run all
 341 the non-i915 specific modeset tests.
 342
 343 Contact: Daniel Vetter
 344
 345 Create a virtual KMS driver for testing (vkms)
 346 ----------------------------------------------
 347
 348 With all the latest helpers it should be fairly simple to create a virtual KMS
 349 driver useful for testing, or for running X or similar on headless machines
 350 (to be able to still use the GPU). This would be similar to vgem, but aimed at
 351 the modeset side.
 352
 353 Once the basics are there there's tons of possibilities to extend it.
 354
 355 Contact: Daniel Vetter
 356
 357 Driver Specific
 358 ===============
 359
 360 tinydrm
 361 -------
 362
 363 Tinydrm is the helper driver for really simple fb drivers. The goal is to make
 364 those drivers as simple as possible, so lots of room for refactoring:
 365
 366 - backlight helpers, probably best to put them into a new drm_backlight.c.
 367   This is because drivers/video is de-facto unmaintained. We could also
 368   move drivers/video/backlight to drivers/gpu/backlight and take it all
 369   over within drm-misc, but that's more work.
 370
 371 - spi helpers, probably best put into spi core/helper code. Thierry said
 372   the spi maintainer is fast&reactive, so shouldn't be a big issue.
 373
 374 - extract the mipi-dbi helper (well, the non-tinydrm specific parts at
 375   least) into a separate helper, like we have for mipi-dsi already. Or follow
 376   one of the ideas for having a shared dsi/dbi helper, abstracting away the
 377   transport details more.
 378
 379 - tinydrm_lastclose could be drm_fb_helper_lastclose. Only thing we need
 380   for that is to store the drm_fb_helper pointer somewhere in
 381   drm_device->mode_config. And then we could roll that out to all the
 382   drivers.
 383
 384 - tinydrm_gem_cma_prime_import_sg_table should probably go into the cma
 385   helpers, as a _vmapped variant (since not every driver needs the vmap).
 386   And tinydrm_gem_cma_free_object could the be merged into
 387   drm_gem_cma_free_object().
 388
 389 - tinydrm_fb_create we could move into drm_simple_pipe, only need to add
 390   the fb_create hook to drm_simple_pipe_funcs, which would again simplify a
 391   bunch of things (since it gives you a one-stop vfunc for simple drivers).
 392
 393 - Quick aside: The unregister devm stuff is kinda getting the lifetimes of
 394   a drm_device wrong. Doesn't matter, since everyone else gets it wrong
 395   too :-)
 396
 397 - With the fbdev pointer in dev->mode_config we could also make
 398   suspend/resume helpers entirely generic, at least if we add a
 399   dev->mode_config.suspend_state. We could even provide a generic pm_ops
 400   structure with those.
 401
 402 - also rework the drm_framebuffer_funcs->dirty hook wire-up, see above.
 403
 404 Contact: Noralf Trønnes, Daniel Vetter
 405
 406 Outside DRM
 407 ===========