DRM Janitor Tasks

Smaller janitorial tasks in the kernel DRM graphics subsystem useful as newbie projects. Or for slow rainy days.

Subsystem-wide refactorings

De-midlayer drivers

With the recent drm_bus cleanup patches for 3.17 it is no longer required to have a drm_bus structure set up. Drivers can directly set up the drm device structure instead of relying on bus methods in drm_usb.c and drm_platform.c. The goal is to get rid of the driver's ->load/->unload callbacks and open-code the load/unload sequence properly, using the new two-stage drm_device setup/teardown.

Once all existing drivers are converted we can also remove those bus support files for usb/platform devices.

All you need is a gpu for a non-converted driver (currently almost all of them, but also all the virtual ones used by kvm, so everyone qualifies).

Contact: Daniel Vetter, Thierry Redding, respective driver maintainers

Convert existing KMS drivers to atomic modesetting

3.19 has the atomic modeset interfaces and helpers, so drivers can now be converted over. Modern compositors like Wayland or Surfaceflinger on Android really want an atomic modeset interface, so this is all about the bright future.

There is a conversion guide for atomic and all you need is a gpu for a non-converted driver (again virtual hw drivers for kvm are still all suitable).

As part of this drivers also need to convert to universal plane (which means exposing primary&cursor as proper plane objects). But that's much easier to do by directly using the new atomic helper driver callbacks.

Contact: Daniel Vetter, respective driver maintainers

Fallout from atomic kms

drm_atomic_helper.c provides a batch of functions which implement legacy ioctls on top of the new atomic driver interface. Which is really nice for gradual conversion of drivers, but unfortunately the semantic mismatches are a bit too severe. So there's some follow-up work to adjust the function interfaces to fix these issues: * atomic needs the lock acquire context. Atm that's passed around implicitly with some horrible hacks, and it's also allocate with GFP_NOFAIL behind the scenes. All legacy paths need to start allocating the acquire context explicitly on stack and then also pass it down into drivers explicitly so that the legacy-on-atomic functions can use them. * A bunch of the vtable hooks are now in the wrong place: DRM has a split between core vfunc tables (named drm_foo_funcs), which are used to implement the userspace ABI. And then there's the optional hooks for the helper libraries (name drm_foo_helper_funcs), which are purely for internal use. Some of these hooks should be move from funcs to helperfuncs since they're not part of the core ABI. There's a FIXME comment in the kerneldoc for each such case in drmcrtc.h.

Contact: Daniel Vetter

get rid of dev->struct_mutex from gem code and drivers

dev->struct_mutex is the Big DRM Lock from legacy days and infested everything. Nowadays in modern drivers the only bit where it's mandatory is serializing gem buffer object destruction. Which unfortunately means drivers have to keep track of that lock and either call unreference or unreference_locked depending upon context.

The need for dev->struct_mutex in core gem code is only because of the mmap handling, and that can be resolved with a kref_get_unless_zero weak reference like ttm. Then for drivers that don't need dev->struct_mutex (ttm based ones mostly) we could add a new gem_bo_free_unlocked functions which doesn't make a distinction between unref and unref_unlocked.

For drivers that need struct_mutex it should be replaced with a driver-private lock. The tricky part is the bo free functions, since those can't reliably take that lock any more. Instead state needs to be protected with suitable subordinate locks or some cleanup work pushed to a worker thread.

Contact: Daniel Vetter

debugfs interface to intercept DP AUX

If we can redirect the dp aux transactions and inject hotplug events into drivers through debugfs we can write automated tests for the rather complex DP detection code. Especially MST is really complicated and has lots of interesting corner-cases when plugging/unplugging parts of the sink tree.

First steps would be to add the basic interface and a very basic SST DP sink testcase. Follow-up tasks would be to extend that to MST and interesting testcases.

Contact: Daniel Vetter

DRM Core refactorings

Create a drm-internal header

Currently the drm subsystem has only one global header, drmP.h. This is used both for functions exported to helper libraries and drivers and functions only used internally in the drm.ko module. The goal would be to move all header declarations not needed outside of drm.ko into drivers/gpu/drm/drm_*internal.h header files. EXPORTSYMBOL also needs to be dropped for these functions.

This would nicely tie in with the below task to create kerneldoc after the api is cleaned up. Or with the "hide legacy cruft better" task.

Contact: Daniel Vetter

This is mostly done now.

Add missing kerneldoc for exported functions

The drm reference documentation is still lacking kerneldoc in a few areas. The task would be to clean up interfaces like moving functions around between files to better group them and improving the interfaces like dropping return values for functions that never fail. Then write kerneldoc for all exported functions and an overview section and integrate it all into the drm DocBook.

See https://www.kernel.org/doc/htmldocs/drm/ for what's there already.

Contact: Daniel Vetter

Hide legacy cruft better

Way back DRM supported only drivers which shadow-attached to pci devices with userspace or fbdev drivers setting up outputs. Modern DRM drives take charge of the entire device, you can spot them with the DRIVER_MODESET flag.

Unfortunately there's still large piles of legacy code around which needs to be hidden so that driver writers don't accidentally end up using it. And to prevent security issues in those legacy ioctls from being exploited on modern drviers. This has multiple possible subtasks:

  • Make sure legacy ioctls can't be used on modern drivers.
  • Extract support code for legacy features into a drm-legacy.ko kernel module and compile it only when one of the legacy drivers is enabled.
  • Extract legacy functions into their own headers and remove it that from the monolithic drmP.h header.
  • Remove any lingering cruft from the OS abstraction layer from modern drivers.

Contact: Daniel Vetter

Make panic handling work

This is a really varied tasks with lots of little bits and pieces:

  • The panic path can't be tested currently, leading to constant breaking. The main issue here is that panics can be triggered from hardirq contexts and hence all panic related callback can run in hardirq context. It would be awesome if we could test at least the fbdev helper code and driver code by e.g. trigger calls through drm debugfs files. hardirq context could be achieved by using an ipi to the local processor.

  • There's a massive confusion of different panic handlers. drm fbdev emulation helpers have one, but on top of that the fbcon code itself also has one. We need to make sure that they stop fighting over each another.

  • drm_can_sleep() is a mess. It hides real bugs in normal operations and isn't a full solution for panic paths. We need to make sure that it only returns true if there's a panic going on for real, and fix up all the fallout.

  • panic handler can't ever sleep, which also means it can't ever mutex_lock. Also it can't grab any other lock unconditionally, not even spinlocks (because nmi and hardirq can panic too). We need to either make sure to not call such paths, or trylock everything. Really tricky.

  • For the above locking troubles reasons it's pretty much impossible to attempt a synchronous modeset from panic handlers. The only thing we could try to achive is an atomic set_base of the primary plane, and hope that it shows up. Everything else probably needs to be delayed to some worker or something else which happens later on. Otherwise it just kills the box harder, prevent the panic from going out on e.g. netconsole.

  • There's also proposal for a simplied drm console instead of the full-blown fbcon and drm fbdev emulation. Any kind of panic handling tricks should obviously work for both console, in case we ever get kmslog merged.

Contact: Daniel Vetter

Better Testing

Enable trinity for drm

And fix up the fallout. Should be really interesting ...

Make KMS tests in i-g-t generic

The i915 driver team maintains an extensive testsuite for the i915 drm driver, including tons of testcases for corner-cases in the modesetting api. It would be awesome if those tests (at least the ones not relying on intel-specific gem features) could be made to run on any kms driver.

Basic work to run igt tests on non-i915 is done, what's now missing is mass-converting things over. For modeset tests we also first need a bit of unfrastructure to use dumb buffers for untiled buffers, to be able to run all the non-i915 specific modeset tests.

Contact: Daniel Vetter

Driver Specific

Outside DRM

Better kerneldoc

This is pretty much done, but there's some advanced topics:

  • Come up with a way to hyperlink to struct members. Currently you can hyperlink to the struct using #struct_name, but not to a member within. Would need buy-in from kerneldoc maintainers, and the big question is how to make it work without totally unsightly drm_foo_bar_really_long_structure->even_longer_memeber all over the text which breaks text flow.

  • Figure out how to integrate the asciidoc support for ascii-diagrams. We have a few of those (e.g. to describe mode timings), and asciidoc supports converting some ascii-art dialect into pngs. Would be really pretty to make that work.

Contact: Daniel Vetter, Thomas Wood