GitHub/moto-9609/android_kernel_motorola_exynos9610.git
11 years agodrm/vma: add access management helpers
David Herrmann [Sun, 25 Aug 2013 16:28:57 +0000 (18:28 +0200)]
drm/vma: add access management helpers

The VMA offset manager uses a device-global address-space. Hence, any
user can currently map any offset-node they want. They only need to guess
the right offset. If we wanted per open-file offset spaces, we'd either
need VM_NONLINEAR mappings or multiple "struct address_space" trees. As
both doesn't really scale, we implement access management in the VMA
manager itself.

We use an rb-tree to store open-files for each VMA node. On each mmap
call, GEM, TTM or the drivers must check whether the current user is
allowed to map this file.

We add a separate lock for each node as there is no generic lock available
for the caller to protect the node easily.

As we currently don't know whether an object may be used for mmap(), we
have to do access management for all objects. If it turns out to slow down
handle creation/deletion significantly, we can optimize it in several
ways:
 - Most times only a single filp is added per bo so we could use a static
   "struct file *main_filp" which is checked/added/removed first before we
   fall back to the rbtree+drm_vma_offset_file.
   This could be even done lockless with rcu.
 - Let user-space pass a hint whether mmap() should be supported on the
   bo and avoid access-management if not.
 - .. there are probably more ideas once we have benchmarks ..

v2: add drm_vma_node_verify_access() helper

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge branch 'gma500-next' of git://github.com/patjak/drm-gma500 into drm-next
Dave Airlie [Thu, 22 Aug 2013 00:38:28 +0000 (10:38 +1000)]
Merge branch 'gma500-next' of git://github.com/patjak/drm-gma500 into drm-next

Here's some gma500 unifying and cleanups for drm-next. There is more stuff in
the pipe for 3.12 but I'd like to get these out of the way first.

* 'gma500-next' of git://github.com/patjak/drm-gma500: (35 commits)
  drm/gma500/cdv: Add and hook up chip op for disabling sr
  drm/gma500/cdv: Add and hook up chip op for watermarks
  drm/gma500: Rename psb_intel_encoder to gma_encoder
  drm/gma500: Rename psb_intel_connector to gma_connector
  drm/gma500: Rename psb_intel_crtc to gma_crtc
  drm/gma500/cdv: Convert to generic set_config()
  drm/gma500/psb: Convert to generic set_config()
  drm/gma500: Add generic set_config() function
  drm/gma500/cdv: Convert to generic save/restore
  drm/gma500/psb: Convert to generic save/restore
  drm/gma500: Add generic crtc save/restore funcs
  drm/gma500: Convert to generic encoder funcs
  drm/gma500: Add generic encoder functions
  drm/gma500/psb: Convert to generic cursor funcs
  drm/gma500/cdv: Convert to generic cursor funcs
  drm/gma500: Add generic cursor functions
  drm/gma500/psb: Convert to generic crtc->destroy
  drm/gma500/mdfld: Use identical generic crtc funcs
  drm/gma500/oak: Use identical generic crtc funcs
  drm/gma500/psb: Convert to gma_crtc_dpms()
  ...

11 years agodrm/prime: Always add exported buffers to the handle cache
Daniel Vetter [Wed, 14 Aug 2013 22:02:49 +0000 (00:02 +0200)]
drm/prime: Always add exported buffers to the handle cache

... not only when the dma-buf is freshly created. In contrived
examples someone else could have exported/imported the dma-buf already
and handed us the gem object with a flink name. If such on object gets
reexported as a dma_buf we won't have it in the handle cache already,
which breaks the guarantee that for dma-buf imports we always hand
back an existing handle if there is one.

This is exercised by igt/prime_self_import/with_one_bo_two_files

Now if we extend the locked sections just a notch more we can also
plug th racy buf/handle cache setup in handle_to_fd:

If evil userspace races a concurrent gem close against a prime export
operation we can end up tearing down the gem handle before the dma buf
handle cache is set up. When handle_to_fd gets around to adding the
handle to the cache there will be no one left to clean it up,
effectily leaking the bo (and the dma-buf, since the handle cache
holds a ref on the dma-buf):

Thread A Thread B

handle_to_fd:

lookup gem object from handle
creates new dma_buf

gem_close on the same handle
obj->dma_buf is set, but file priv buf
handle cache has no entry

obj->handle_count drops to 0

drm_prime_add_buf_handle sets up the handle cache

-> We have a dma-buf reference in the handle cache, but since the
handle_count of the gem object already dropped to 0 no on will clean
it up. When closing the drm device fd we'll hit the WARN_ON in
drm_prime_destroy_file_private.

The important change is to extend the critical section of the
filp->prime.lock to cover the gem handle lookup. This serializes with
a concurrent gem handle close.

This leak is exercised by igt/prime_self_import/export-vs-gem_close-race

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: make drm_prime_lookup_buf_handle static
Daniel Vetter [Wed, 14 Aug 2013 22:02:48 +0000 (00:02 +0200)]
drm/prime: make drm_prime_lookup_buf_handle static

... and move it to the top of the function to avoid a forward
declaration.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: Simplify drm_gem_remove_prime_handles
Daniel Vetter [Wed, 14 Aug 2013 22:02:47 +0000 (00:02 +0200)]
drm/prime: Simplify drm_gem_remove_prime_handles

with the reworking semantics and locking of the obj->dma_buf pointer
this pointer is always set as long as there's still a gem handle
around and a dma_buf associated with this gem object.

Also, the per file-priv lookup-cache for dma-buf importing is also
unified between foreign and native objects.

Hence we don't need to special case the clean any more and can simply
drop the clause which only runs for foreing objects, i.e. with
obj->import_attach set.

Note that with this change (actually with the previous one to always
set up obj->dma_buf even for foreign objects) it is no longer required
to set obj->import_attach when importing a foreing object. So update
comments accordingly, too.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: proper locking+refcounting for obj->dma_buf link
Daniel Vetter [Wed, 14 Aug 2013 22:02:46 +0000 (00:02 +0200)]
drm/prime: proper locking+refcounting for obj->dma_buf link

The export dma-buf cache is semantically similar to an flink name. So
semantically it makes sense to treat it the same and remove the name
(i.e. the dma_buf pointer) and its references when the last gem handle
disappears.

Again we need to be careful, but double so: Not just could someone
race and export with a gem close ioctl (so we need to recheck
obj->handle_count again when assigning the new name), but multiple
exports can also race against each another. This is prevented by
holding the dev->object_name_lock across the entire section which
touches obj->dma_buf.

With the new scheme we also need to reinstate the obj->dma_buf link at
import time (in case the only reference userspace has held in-between
was through the dma-buf fd and not through any native gem handle). For
simplicity we don't check whether it's a native object but
unconditionally set up that link - with the new scheme of removing the
obj->dma_buf reference when the last handle disappears we can do that.

To make it clear that this is not just for exported buffers anymore
als rename it from export_dma_buf to dma_buf.

To make sure that now one can race a fd_to_handle or handle_to_fd with
gem_close we use the same tricks as in flink of extending the
dev->object_name_locking critical section. With this change we finally
have a guaranteed 1:1 relationship (at least for native objects)
between gem objects and dma-bufs, even accounting for races (which can
happen since the dma-buf itself holds a reference while in-flight).

This prevent igt/prime_self_import/export-vs-gem_close-race from
Oopsing the kernel. There is still a leak though since the per-file
priv dma-buf/handle cache handling is racy. That will be fixed in a
later patch.

v2: Remove the bogus dma_buf_put from the export_and_register_object
failure path if we've raced with the handle count dropping to 0.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: completely close gem_open vs. gem_close races
Daniel Vetter [Wed, 14 Aug 2013 22:02:45 +0000 (00:02 +0200)]
drm/gem: completely close gem_open vs. gem_close races

The gem flink name holds a reference onto the object itself, and this
self-reference would prevent an flink'ed object from every being
freed. To break that loop we remove the flink name when the last
userspace handle disappears, i.e. when obj->handle_count reaches 0.

Now in gem_open we drop the dev->object_name_lock between the flink
name lookup and actually adding the handle. This means a concurrent
gem_close of the last handle could result in the flink name getting
reaped right inbetween, i.e.

Thread 1 Thread 2
gem_open gem_close

flink -> obj lookup
handle_count drops to 0
remove flink name
create_handle
handle_count++

If someone now flinks this object again, we'll get a new flink name.

We can close this race by removing the lock dropping and making the
entire lookup+handle_create sequence atomic. Unfortunately to still be
able to share the handle_create logic this requires a
handle_create_tail function which drops the lock - we can't hold the
object_name_lock while calling into a driver's ->gem_open callback.

Note that for flink fixing this race isn't really important, since
racing gem_open against gem_close is clearly a userspace bug. And no
matter how the race ends, we won't leak any references.

But with dma-buf where the userspace dma-buf fd itself is refcounted
this is a valid sequence and hence we should fix it. Therefore this
patch here is just a warm-up exercise (and for consistency between
flink buffer sharing and dma-buf buffer sharing with self-imports).

Also note that this extension of the critical section in gem_open
protected by dev->object_name_lock only works because it's now a
mutex: A spinlock would conflict with the potential memory allocation
in idr_preload().

This is exercises by igt/gem_flink_race/flink_name.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: switch dev->object_name_lock to a mutex
Daniel Vetter [Wed, 14 Aug 2013 22:02:44 +0000 (00:02 +0200)]
drm/gem: switch dev->object_name_lock to a mutex

I want to wrap the creation of a dma-buf from a gem object in it,
so that the obj->export_dma_buf cache can be atomically filled in.

Instead of creating a new mutex just for that variable I've figured
I can reuse the existing dev->object_name_lock, especially since
the new semantics will exactly mirror the flink obj->name already
protected by that lock.

v2: idr_preload/idr_preload_end is now an atomic section, so need to
move the mutex locking outside.

[airlied: fix up conflict with patch to make debugfs use lock]

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: clarify logic a bit in drm_gem_prime_fd_to_handle
Daniel Vetter [Wed, 14 Aug 2013 22:02:43 +0000 (00:02 +0200)]
drm/prime: clarify logic a bit in drm_gem_prime_fd_to_handle

if (!ret) implies that ret == 0, so no need to clear it again. And
explicitly check for ret == 0 to indicate that we're checking an errno
integer.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: shrink critical section protected by prime lock
Daniel Vetter [Wed, 14 Aug 2013 22:02:42 +0000 (00:02 +0200)]
drm/prime: shrink critical section protected by prime lock

When exporting a gem object as a dma-buf the critical section for the
per-fd prime lock is just the adding (and in case of errors, removing)
of the handle to the per-fd lookup cache.

So restrict the critical section to just that part of the function.

This simplifies later reordering.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: use proper pointer in drm_gem_prime_handle_to_fd
Daniel Vetter [Wed, 14 Aug 2013 22:02:41 +0000 (00:02 +0200)]
drm/prime: use proper pointer in drm_gem_prime_handle_to_fd

Part of the function uses the properly-typed dmabuf variable, the
other an untyped void *buf. Kill the later.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: make drm_gem_object_handle_unreference_unlocked static
Daniel Vetter [Wed, 14 Aug 2013 22:02:39 +0000 (00:02 +0200)]
drm/gem: make drm_gem_object_handle_unreference_unlocked static

No one outside of drm should use this, the official interfaces are
drm_gem_handle_create and drm_gem_handle_delete. The handle refcounting
is purely an implementation detail of gem.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: fix error path in drm_gem_prime_fd_to_handle
Daniel Vetter [Wed, 14 Aug 2013 22:02:38 +0000 (00:02 +0200)]
drm/prime: fix error path in drm_gem_prime_fd_to_handle

handle_unreference only clears up the obj->name and the reference,
but would leave a dangling handle in the idr. The right thing
to do is to call handle_delete.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: fix up flink name create race
Daniel Vetter [Wed, 14 Aug 2013 22:02:37 +0000 (00:02 +0200)]
drm/gem: fix up flink name create race

This is the 2nd attempt, I've always been a bit dissatisified with the
tricky nature of the first one:

http://lists.freedesktop.org/archives/dri-devel/2012-July/025451.html

The issue is that the flink ioctl can race with calling gem_close on
the last gem handle. In that case we'll end up with a zero handle
count, but an flink name (and it's corresponding reference). Which
results in a neat space leak.

In my first attempt I've solved this by rechecking the handle count.
But fundamentally the issue is that ->handle_count isn't your usual
refcount - it can be resurrected from 0 among other things.

For those special beasts atomic_t often suggest way more ordering that
it actually guarantees. To prevent being tricked by those hairy
semantics take the easy way out and simply protect the handle with the
existing dev->object_name_lock.

With that change implemented it's dead easy to fix the flink vs. gem
close reace: When we try to create the name we simply have to check
whether there's still officially a gem handle around and if not refuse
to create the flink name. Since the handle count decrement and flink
name destruction is now also protected by that lock the reace is gone
and we can't ever leak the flink reference again.

Outside of the drm core only the exynos driver looks at the handle
count, and tbh I have no idea why (it's just for debug dmesg output
luckily).

I've considered inlining the drm_gem_object_handle_free, but I plan to
add more name-like things (like the exported dma_buf) to this scheme,
so it's clearer to leave the handle freeing in its own function.

This is exercised by the new gem_flink_race i-g-t testcase, which on
my snb leaks gem objects at a rate of roughly 1k objects/s.

v2: Fix up the error path handling in handle_create and make it more
robust by simply calling object_handle_unreference.

v3: Fix up the handle_unreference logic bug - atomic_dec_and_test
retursn 1 for 0. Oops.

v4: Squash in inlining of drm_gem_object_handle_reference as suggested
by Dave Airlie and add a note that we now have a testcase.

Cc: Dave Airlie <airlied@gmail.com>
Cc: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge tag 'drm-intel-next-2013-08-09' of git://people.freedesktop.org/~danvet/drm...
Dave Airlie [Wed, 21 Aug 2013 02:48:59 +0000 (12:48 +1000)]
Merge tag 'drm-intel-next-2013-08-09' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

Daniel writes:
New pile of stuff for -next:
- Cleanup of the old crtc helper callbacks, all encoders are now converted
  to the i915 modeset infrastructure.
- Massive amount of wm patches from Ville for ilk, snb, ivb, hsw, this is
  prep work to eventually get things going for nuclear pageflips where we
  need to adjust watermarks on the fly.
- More vm/vma patches from Ben. This refactoring isn't yet fully rolled
  out, we miss the execbuf conversion and some of the low-level
  bind/unbind support code.
- Convert our hdmi infoframe code to use the new common helper functions
  (Damien). This contains some bugfixes for the common infoframe helpers.
- Some cruft removal from Damien.
- Various smaller bits&pieces all over, as usual.

* tag 'drm-intel-next-2013-08-09' of git://people.freedesktop.org/~danvet/drm-intel: (105 commits)
  drm/i915: Fix FB WM for HSW
  drm/i915: expose HDMI connectors on port C on BYT
  drm/i915: fix a limit check in hsw_compute_wm_results()
  drm/i915: unbreak i915_gem_object_ggtt_unbind()
  drm/i915: Make intel_set_mode() static
  drm/i915: Remove intel_modeset_disable()
  drm/i915: Make intel_encoder_dpms() static
  drm/i915: Make i915_hangcheck_elapsed() static
  drm/i915: Fix #endif comment
  drm/i915: Remove i915_gem_object_check_coherency()
  drm/i915: Remove stale prototypes
  drm/i915: List objects allocated from stolen memory in debugfs
  drm/i915: Always call intel_update_sprite_watermarks() when disabling a plane
  drm/i915: Pass plane and crtc to intel_update_sprite_watermarks
  drm/i915: Don't try to disable plane if it's already disabled
  drm/i915: Pass crtc to our update/disable_plane hooks
  drm/i915: Split plane watermark parameters into a separate struct
  drm/i915: Pull some watermarks state into a separate structure
  drm/i915: Calculate max watermark levels for ILK+
  drm/i915: Rename hsw_lp_wm_result to intel_wm_level
  ...

11 years agodrm: Make drm_get_platform_dev() static
Lespiau, Damien [Mon, 19 Aug 2013 23:53:10 +0000 (00:53 +0100)]
drm: Make drm_get_platform_dev() static

It's only used in drm_platform.c.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Remove unused PCI ids
Lespiau, Damien [Mon, 19 Aug 2013 23:53:09 +0000 (00:53 +0100)]
drm: Remove unused PCI ids

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Make drm_fb_cma_describe() static
Lespiau, Damien [Mon, 19 Aug 2013 23:53:08 +0000 (00:53 +0100)]
drm: Make drm_fb_cma_describe() static

This function is only used in drm_fb_cma_helper.c.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Remove 2 unused defines
Lespiau, Damien [Mon, 19 Aug 2013 23:53:07 +0000 (00:53 +0100)]
drm: Remove 2 unused defines

These were introduced in the very first DRM commit:

  commit f453ba0460742ad027ae0c4c7d61e62817b3e7ef
  Author: Dave Airlie <airlied@redhat.com>
  Date:   Fri Nov 7 14:05:41 2008 -0800

      DRM: add mode setting support

      Add mode setting support to the DRM layer.

But are unused.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Make drm_mode_remove() static
Lespiau, Damien [Mon, 19 Aug 2013 23:53:06 +0000 (00:53 +0100)]
drm: Make drm_mode_remove() static

It's only used in drm_crtc.c.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Remove drm_mode_list_concat()
Lespiau, Damien [Mon, 19 Aug 2013 23:53:05 +0000 (00:53 +0100)]
drm: Remove drm_mode_list_concat()

The last user was removed in

  commit 575dc34ee0de867ba83abf25998e0963bff451fa
  Author: Dave Airlie <airlied@redhat.com>
  Date:   Mon Sep 7 18:43:26 2009 +1000

      drm/kms: remove old std mode fallback code.

      The new code adds modes in the helper, which makes more sense
      I disliked the non-driver code adding modes.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Remove drm_mode_create_dithering_property()
Lespiau, Damien [Mon, 19 Aug 2013 23:53:04 +0000 (00:53 +0100)]
drm: Remove drm_mode_create_dithering_property()

This was last used by nouveau, replaced by a driver-specific property
in:

  commit de69185573586302ada2e59ba41835df36986277
  Author: Ben Skeggs <bskeggs@redhat.com>
  Date:   Mon Oct 17 12:23:41 2011 +1000

      drm/nouveau: improve dithering properties, and implement proper auto mode

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Remove stale prototypes
Lespiau, Damien [Mon, 19 Aug 2013 23:53:03 +0000 (00:53 +0100)]
drm: Remove stale prototypes

A few prototypes have been left in the headers, their function friends
long gone.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: move dev data clearing from drm_setup to lastclose
Daniel Vetter [Thu, 8 Aug 2013 13:41:35 +0000 (15:41 +0200)]
drm: move dev data clearing from drm_setup to lastclose

We kzalloc this structure, and for real kms devices we should never
loose track of things really.

But ums/legacy drivers rely on the drm core to clean up a bit of cruft
between lastclose and firstopen (i.e. when X is being restarted), so
keep this around. But give it a clear drm_legacy_ prefix and
conditionalize the code on !DRIVER_MODESET.

Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: remove procfs code, take 2
Daniel Vetter [Thu, 8 Aug 2013 13:41:34 +0000 (15:41 +0200)]
drm: remove procfs code, take 2

So almost two years ago I've tried to nuke the procfs code already
once before:

http://lists.freedesktop.org/archives/dri-devel/2011-October/015707.html

The conclusion was that userspace drivers (specifically libdrm device
node detection) stopped relying on procfs in 2001. But after some
digging it turned out that the drmstat tool in libdrm is still using
those files (but only when certain options are set). So we've decided
to keep profcs.

But I when I've started to dig around again what exactly this tool
does I've noticed that it tries to read the "mem", "vm", and "vma"
files from procfs. Now as far my git history digging shows "mem" never
did anything useful (at least in the version that first showed up in
upstream history in 2004) and the file was remove in

commit 955b12def42e83287c1bdb1411d99451753c1391
Author: Ben Gamari <bgamari@gmail.com>
Date:   Tue Feb 17 20:08:49 2009 -0500

    drm: Convert proc files to seq_file and introduce debugfs

Which means that for over 4 years drmstat has been broken, and no one
cared. In my opinion that's proof enough that no one is actually using
drmstat, and so that we can savely nuke the procfs support from drm.

While at it fix up the error case cleanup for debugfs in drm_get_minor.

v2: Fix dates, libdrm stopped relying on procfs for drm node detection
in 2001.

v3: fixup compilation warning for !CONFIG_DEBUG_FS, reported by
Fengguang Wu.

Cc: kbuild test robot <fengguang.wu@intel.com>
Cc: Dave Airlie <airlied@linux.ie>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: don't call ->firstopen for KMS drivers
Daniel Vetter [Thu, 8 Aug 2013 13:41:15 +0000 (15:41 +0200)]
drm: don't call ->firstopen for KMS drivers

It has way too much potential for driver writers to do stupid things
like delayed hw setup because the load sequence is somehow racy (e.g.
the imx driver in staging). So don't call it for modesetting drivers,
which reduces the complexity of the drm core -> driver interface a
notch.

v2: Don't forget to update DocBook.

v3: Go with Laurent's slightly more elaborate proposal for the DocBook
update. Add a few words on top of his diff to elaborate a bit on what
KMS drivers should and shouldn't do in lastclose. There was already a
paragraph present talking about restoring properties, I've simply
extended that one.

Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/vmwgfx: remove ->firstopen callback
Daniel Vetter [Thu, 8 Aug 2013 13:41:14 +0000 (15:41 +0200)]
drm/vmwgfx: remove ->firstopen callback

So if we survey kms drivers there's a bunch of things they commonly do
in ->lastclose
- delayed processing of vga switcheroo requests (i915, nouveau,
  radeon)
- force-restoring the fbcon (most)
- resetting a bunch properties to make fbcon work better (omap)
- disabling all outputs (vmwgfx)

In short besides the semantically important vga switcheroo stuff they
all try very hard to keep fbcon working in case X dies.

But none of them try to not do this at driver unload time safe for
vmwgfx, and digging through logs I couldn't find any reason for why
vmwgfx is special.

Since ->firstopen has lots of potential for abuse with kms drivers
(like delaying driver setup to pamper over races in the load sequence)
it's imo very much worth it to remove this logic so that we can
stop using the ->firstopen callback for kms drivers.

Also module unloading is rather a debug feature and developers should
know how to restore the display to a sane configuration.

Cc: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/imx: kill firstopen callback
Daniel Vetter [Thu, 8 Aug 2013 13:41:13 +0000 (15:41 +0200)]
drm/imx: kill firstopen callback

This thing seems to do some kind of delayed setup. Really, real kms
drivers shouldn't do that at all. Either stuff needs to be dynamically
hotplugged or the driver setup sequence needs to be fixed.

This patch here just moves the setup at the very end of the driver
load callback, with the locking adjusted accordingly.

v2: Also move the corresponding put from ->lastclose to ->unload.

Cc: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: fix minor number range calculation
Kristian Høgsberg [Thu, 8 Aug 2013 17:10:21 +0000 (19:10 +0200)]
drm: fix minor number range calculation

Currently, both ranges overlap. Fix the limits so both ranges are mutually
exclusive. Also use the occasion to convert whitespaces to tabs.

Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
(fixed up tabs and adjust commit-msg accordingly)
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: fix locking in gem debugfs/procfs file
Daniel Vetter [Thu, 8 Aug 2013 13:41:33 +0000 (15:41 +0200)]
drm: fix locking in gem debugfs/procfs file

The idr is protected with our spinlock, if we don't hold that nothing
prevents the gem objects from disappearing from under us.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: remove the dma_ioctl special-case
Daniel Vetter [Thu, 8 Aug 2013 13:41:29 +0000 (15:41 +0200)]
drm: remove the dma_ioctl special-case

We might as well have a real ioctl function which checks for the
callbacks. This seems to be a remnant from back in the days when each
drm driver had their own complete ioctl table, with no shared core
drm table at all.

To make really sure no mis-guided user in a kms driver pops up again
explicitly check for that in the new ioctl implementation.

v2: Drop the unused variable I've accidentally left in the code,
spotted by David Herrmann.

Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/docs: rip out removed driver flags documentation
Daniel Vetter [Thu, 8 Aug 2013 13:41:28 +0000 (15:41 +0200)]
drm/docs: rip out removed driver flags documentation

I've forgotten this and shuffling all the little pieces into the
respective patches is rather cumbersome ...

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: rip out drm_core_has_MTRR checks
Daniel Vetter [Thu, 8 Aug 2013 13:41:27 +0000 (15:41 +0200)]
drm: rip out drm_core_has_MTRR checks

The new arch_phys_wc_add/del functions do the right thing both with
and without MTRR support in the kernel. So we can drop these
additional checks.

David Herrmann suggest to also kill the DRIVER_USE_MTRR flag since
it's now unused, which spurred me to do a bit a better audit of the
affected drivers. David helped a lot in that. Quoting our mail
discussion:

On Wed, Jul 10, 2013 at 5:41 PM, David Herrmann <dh.herrmann@gmail.com> wrote:
> On Wed, Jul 10, 2013 at 5:22 PM, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
>> On Wed, Jul 10, 2013 at 3:51 PM, David Herrmann <dh.herrmann@gmail.com> wrote:
>>>> -#if __OS_HAS_MTRR
>>>> -static inline int drm_core_has_MTRR(struct drm_device *dev)
>>>> -{
>>>> -       return drm_core_check_feature(dev, DRIVER_USE_MTRR);
>>>> -}
>>>> -#else
>>>> -#define drm_core_has_MTRR(dev) (0)
>>>> -#endif
>>>> -
>>>
>>> That was the last user of DRIVER_USE_MTRR (apart from drivers setting
>>> it in .driver_features). Any reason to keep it around?
>>
>> Yeah, I guess we could rip things out. Which will also force me to
>> properly audit drivers for the eventual behaviour change this could
>> entail (in case there's an x86 driver which did not ask for an mtrr,
>> but iirc there isn't).
>
> david@david-mb ~/dev/kernel/linux $ for i in drivers/gpu/drm/* ; do if
> test -d "$i" ; then if ! grep -q USE_MTRR -r $i ; then echo $i ; fi ;
> fi ; done
> drivers/gpu/drm/exynos
> drivers/gpu/drm/gma500
> drivers/gpu/drm/i2c
> drivers/gpu/drm/nouveau
> drivers/gpu/drm/omapdrm
> drivers/gpu/drm/qxl
> drivers/gpu/drm/rcar-du
> drivers/gpu/drm/shmobile
> drivers/gpu/drm/tilcdc
> drivers/gpu/drm/ttm
> drivers/gpu/drm/udl
> drivers/gpu/drm/vmwgfx
> david@david-mb ~/dev/kernel/linux $
>
> So for x86 gma500,nouveau,qxl,udl,vmwgfx don't set DRIVER_USE_MTRR.
> But I cannot tell whether they break if we call arch_phys_wc_add/del,
> anyway. At least nouveau seemed to work here, but it doesn't use AGP
> or drm_bufs, I guess.

Cool, thanks a lot for stitching together the list of drivers to look
at. So for real KMS drivers it's the drives responsibility to add an
mtrr if it needs one. nouvea, radeon, mgag200, i915 and vmwgfx do that
already. Somehow the savage driver also ends up doing that, I have no
idea why.

Note that gma500 as a pure KMS driver doesn't need MTRR setup since
the platforms that it supports all support PAT. So no MTRRs needed to
get wc iomappings.

The mtrr support in the drm core is all for legacy mappings of garts,
framebuffers and registers. All legacy drivers set the USE_MTRR flag,
so we're good there.

All in all I think we can really just ditch this

/endquote

v2: Also kill DRIVER_USE_MTRR as suggested by David Herrmann

v3: Rebase on top of David Herrmann's agp setup/cleanup changes.

Cc: David Herrmann <dh.herrmann@gmail.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: WARN about unbalanced handle refcounts
Daniel Vetter [Wed, 14 Aug 2013 22:02:36 +0000 (00:02 +0200)]
drm/gem: WARN about unbalanced handle refcounts

Trying to drop a reference we don't have is a pretty serious bug.
Trying to paper over it is an even worse offense.

So scream into dmesg with a big WARN in case that ever happens.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: remove bogus NULL check from drm_gem_object_handle_unreference_unlocked
Daniel Vetter [Wed, 14 Aug 2013 22:02:35 +0000 (00:02 +0200)]
drm/gem: remove bogus NULL check from drm_gem_object_handle_unreference_unlocked

Calling this function with a NULL object is simply a bug, so papering
over a NULL object not a good idea.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: move drm_gem_object_handle_unreference_unlocked into drm_gem.c
Daniel Vetter [Wed, 14 Aug 2013 22:02:34 +0000 (00:02 +0200)]
drm/gem: move drm_gem_object_handle_unreference_unlocked into drm_gem.c

We have three callers of this function now and it's neither
performance critical nor really small. So an inline function feels
like overkill and unecessarily separates the different parts of the
code.

Since all callers of drm_gem_object_handle_free are now in drm_gem.c
we can make that static (and remove the unused EXPORT_SYMBOL). To
avoid a forward declaration move it (and drm_gem_object_free_bug) up a
bit.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: add a bit of documentation about gem_obj->import_attach
Daniel Vetter [Wed, 14 Aug 2013 22:02:33 +0000 (00:02 +0200)]
drm/prime: add a bit of documentation about gem_obj->import_attach

Lifetime rules seem to be solid around ->import_attach. So this patch
just properly documents them.

Note that pointing directly at the attachment might have issues for
devices that have multiple struct device *dev parts constituting the
logical gpu and so might need multiple attachment points. Similarly
for drm devices which don't need a dma attachment at all (like udl).

But fixing that up is material for different patches.

Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/prime: remove cargo-cult locking from map_sg helper
Daniel Vetter [Wed, 14 Aug 2013 22:02:32 +0000 (00:02 +0200)]
drm/prime: remove cargo-cult locking from map_sg helper

I've checked both implementations (radeon/nouveau) and they both grab
the page array from ttm simply by dereferencing it and then wrapping
it up with drm_prime_pages_to_sg in the callback and map it with
dma_map_sg (in the helper).

Only the grabbing of the underlying page array is anything we need to
be concerned about, and either those pages are pinned independently,
or we're screwed no matter what.

And indeed, nouveau/radeon pin the backing storage in their
attach/detach functions.

Since I've created this patch cma prime support for dma_buf was added.
drm_gem_cma_prime_get_sg_table only calls kzalloc and the creates&maps
the sg table with dma_get_sgtable. It doesn't touch any gem object
state otherwise. So the cma helpers also look safe.

The only thing we might claim it does is prevent concurrent mapping of
dma_buf attachments. But a) that's not allowed and b) the current code
is racy already since it checks whether the sg mapping exists _before_
grabbing the lock.

So the dev->struct_mutex locking here does absolutely nothing useful,
but only distracts. Remove it.

This should also help Maarten's work to eventually pin the backing
storage more dynamically by preventing locking inversions around
dev->struct_mutex.

v2: Add analysis for recently added cma helper prime code.

Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/exynos: explicit store base gem object in dma_buf->priv
Inki Dae [Wed, 14 Aug 2013 22:02:31 +0000 (00:02 +0200)]
drm/exynos: explicit store base gem object in dma_buf->priv

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: use common drm_gem_dmabuf_release in i915/exynos drivers
Daniel Vetter [Wed, 14 Aug 2013 22:02:30 +0000 (00:02 +0200)]
drm: use common drm_gem_dmabuf_release in i915/exynos drivers

Note that this is slightly tricky since both drivers store their
native objects in dma_buf->priv. But both also embed the base
drm_gem_object at the first position, so the implicit cast is ok.

To use the release helper we need to export it, too.

Cc: Inki Dae <inki.dae@samsung.com>
Cc: Intel Graphics Development <intel-gfx@lists.freedesktop.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/host1x: stop casting VMA offsets to 32bit
David Herrmann [Tue, 13 Aug 2013 12:19:58 +0000 (14:19 +0200)]
drm/host1x: stop casting VMA offsets to 32bit

VMA offsets are 64bit so do not cast them to "unsigned int". Also remove
the (now useless) offset-retrieval helper. The VMA manager provides simple
enough helpers.

Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: "Terje Bergström" <tbergstrom@nvidia.com>
Cc: Arto Merilainen <amerilainen@nvidia.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: use ida to allocate connector ids
Ilia Mirkin [Thu, 8 Aug 2013 02:34:48 +0000 (22:34 -0400)]
drm: use ida to allocate connector ids

This makes it so that reloading a module does not cause all the
connector ids to change, which are user-visible and sometimes used
for configuration.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/omap: kill omap_gem_helpers.c
Rob Clark [Wed, 7 Aug 2013 17:41:27 +0000 (13:41 -0400)]
drm/omap: kill omap_gem_helpers.c

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/udl: use gem get/put page helpers
Rob Clark [Wed, 7 Aug 2013 17:41:26 +0000 (13:41 -0400)]
drm/udl: use gem get/put page helpers

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gma500: use gem get/put page helpers
Rob Clark [Wed, 7 Aug 2013 17:41:25 +0000 (13:41 -0400)]
drm/gma500: use gem get/put page helpers

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: add shmem get/put page helpers
Rob Clark [Wed, 7 Aug 2013 17:41:24 +0000 (13:41 -0400)]
drm/gem: add shmem get/put page helpers

Basically just extracting some code duplicated in gma500, omapdrm, udl,
and upcoming msm driver.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gem: add drm_gem_create_mmap_offset_size()
Rob Clark [Wed, 7 Aug 2013 17:41:23 +0000 (13:41 -0400)]
drm/gem: add drm_gem_create_mmap_offset_size()

Variant of drm_gem_create_mmap_offset() which doesn't make the
assumption that virtual size and physical size (obj->size) are the same.
This is needed in omapdrm to deal with tiled buffers.  And lets us get
rid of a duplicated and slightly modified version of
drm_gem_create_mmap_offset() in omapdrm.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/omap: use flip-work helper
Rob Clark [Wed, 7 Aug 2013 17:41:21 +0000 (13:41 -0400)]
drm/omap: use flip-work helper

And simplify how we hold a ref+pin to what is being scanned out by using
fb refcnt'ing.  The previous logic pre-dated fb refcnt, and as a result
was less straightforward than it could have been.  By holding a ref to
the fb, we don't have to care about how many plane's there are and
holding a ref to each color plane's bo.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/tilcdc: use flip-work helper
Rob Clark [Wed, 7 Aug 2013 17:41:20 +0000 (13:41 -0400)]
drm/tilcdc: use flip-work helper

Signed-off-by: Rob Clark <robdclark@gmail.com>
Tested-by: Darren Etheridge <detheridge@ti.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: add flip-work helper
Rob Clark [Wed, 7 Aug 2013 18:41:54 +0000 (14:41 -0400)]
drm: add flip-work helper

A small helper to queue up work to do, from workqueue context, after a
flip.  Typically useful to defer unreffing buffers that may be read by
the display controller until vblank.

v1: original
v2: wire up docbook + couple docbook fixes

Signed-off-by: Rob Clark <robdclark@gmail.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: Remove drm_mode_validate_clocks
Stéphane Marchesin [Tue, 13 Aug 2013 18:55:12 +0000 (11:55 -0700)]
drm: Remove drm_mode_validate_clocks

This function is unused.

Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: no-op out GET_STATS ioctl
Daniel Vetter [Thu, 8 Aug 2013 13:41:32 +0000 (15:41 +0200)]
drm: no-op out GET_STATS ioctl

Again only used by a tests in libdrm and by dristat. Nowadays we have
much better tracing tools to get detailed insights into what a drm
driver is doing. And for a simple "does it work" kind of question that
these stats could answer we have plenty of dmesg debug log spew.

So I don't see any use for this stat gathering complexity at all.

To be able to gradually drop things start with ripping out the
interfaces to it, here the ioctl.

To prevent dristat from eating its own stack garbage we can't use the
drm_noop ioctl though, since we need to clear the return data with a
memset.

Cc: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: hollow-out GET_CLIENT ioctl
Daniel Vetter [Thu, 8 Aug 2013 13:41:31 +0000 (15:41 +0200)]
drm: hollow-out GET_CLIENT ioctl

We not only have debugfs files to do pretty much the equivalent of
lsof, we also have an ioctl. Not that compared to lsof this dumps a
wee bit more information, but we can still get at that from debugfs
easily.

I've dug around in mesa, libdrm and ddx histories and the only users
seem to be drm/tests/dristat.c and drm/tests/getclients.c. The later
is a testcase for the ioctl itself since up to

commit b018fcdaa5e8b4eabb8cffda687d00004a3c4785
Author: Eric Anholt <eric@anholt.net>
Date:   Thu Nov 22 18:46:54 2007 +1000

    drm: Make DRM_IOCTL_GET_CLIENT return EINVAL when it can't find client #idx

there was actually no way at all for userspace to enumerate all
clients since the kernel just wouldn't tell it when to stop. Which
completely broke it's only user, dristat -c.

So obviously that ioctl wasn't much use for debugging. Hence I don't
see any point in keeping support for a tool which was pretty obviously
never really used, and while we have good replacements in the form of
equivalent debugfs files.

Still, to keep dristat -c from looping forever again stop it early by
returning an unconditional -EINVAL. Also add a comment in the code
about why.

v2: Slightly less hollowed-out implementation. libva uses GET_CLIENTS
to figure out whether the fd it has is already authenticated or not.
So we need to keep that part of things working. Simplest way is to
just return one entry to keep va_drm_is_authenticated in
libva/va/drm/va_drm_auth.c working.

This is exercised by igt/drm_get_client_auth which contains a
copypasta of the libva auth check code.

Cc: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/memory: don't export agp helpers
Daniel Vetter [Thu, 8 Aug 2013 13:41:30 +0000 (15:41 +0200)]
drm/memory: don't export agp helpers

They're only used by the agpgart support code in drm_agpgart.c,
not by any drivers.

I think long-term we should create a drm_internal.h include file with
all the various functions only used by the drm core and not exported
to drivers, and remove them from drmP.h. Oh, and someone should kill
that upper-case P sometimes ;-) But that's all stuff for future patch
bombs.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: remove a bunch of unused #defines from drmP.h
Daniel Vetter [Thu, 8 Aug 2013 13:41:26 +0000 (15:41 +0200)]
drm: remove a bunch of unused #defines from drmP.h

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: rip out a few unused DRIVER flags
Daniel Vetter [Thu, 8 Aug 2013 13:41:25 +0000 (15:41 +0200)]
drm: rip out a few unused DRIVER flags

The gma500 driver somehow set the DRIVER_IRQ_VBL flag, but since
there's no code at all to check for this we can kill it. The other two
are completely unused.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: rip out DRIVER_FB_DMA and related code
Daniel Vetter [Thu, 8 Aug 2013 13:41:24 +0000 (15:41 +0200)]
drm: rip out DRIVER_FB_DMA and related code

No driver ever sets that flag, so good riddance!

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: remove FASYNC support
Daniel Vetter [Thu, 8 Aug 2013 13:41:23 +0000 (15:41 +0200)]
drm: remove FASYNC support

So I've stumbled over drm_fasync and wondered what it does. Digging
that up is quite a story.

First I've had to read up on what this does and ended up being rather
bewildered why peopled loved signals so much back in the days that
they've created SIGIO just for that ...

Then I wondered how this ever works, and what that strange "No-op."
comment right above it should mean. After all calling the core fasync
helper is pretty obviously not a noop. After reading through the
kernels FASYNC implementation I've noticed that signals are only sent
out to the processes attached with FASYNC by calling kill_fasync.

No merged drm driver has ever done that.

After more digging I've found out that the only driver that ever used
this is the so called GAMMA driver. I've frankly never heard of such a
gpu brand ever before. Now FASYNC seems to not have been the only bad
thing with that driver, since Dave Airlie removed it from the drm
driver with prejudice:

commit 1430163b4bbf7b00367ea1066c1c5fe85dbeefed
Author: Dave Airlie <airlied@linux.ie>
Date:   Sun Aug 29 12:04:35 2004 +0000

    Drop GAMMA DRM from a great height ...

Long story short, the drm fasync support seems to be doing absolutely
nothing. And the only user of it was never merged into the upstream
kernel. And we don't need any fops->fasync callback since the fcntl
implementation in the kernel already implements the noop case
correctly.

So stop this particular cargo-cult and rip it all out.

v2: Kill drm_fasync assignments in rcar (newly added) and imx drivers
(somehow I've missed that one in staging). Also drop the reference in
the drm DocBook. ARM compile-fail reported by Rob Clark.

v3: Move the removal of dev->buf_asnyc assignment in drm_setup to this
patch here.

v4: Actually git add ... tsk.

Cc: Dave Airlie <airlied@linux.ie>
Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Rob Clark <robdclark@gmail.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/vmwgfx: remove redundant clearing of driver->dma_quiescent
Daniel Vetter [Thu, 8 Aug 2013 13:41:22 +0000 (15:41 +0200)]
drm/vmwgfx: remove redundant clearing of driver->dma_quiescent

It's kzalloced ...

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: mark context support as a legacy subsystem
Daniel Vetter [Thu, 8 Aug 2013 13:41:21 +0000 (15:41 +0200)]
drm: mark context support as a legacy subsystem

So after a lot of digging around in git histories it looks like this
has only ever be used by dri1 render clients. Hence we can fully
disable the entire thing for modesetting drivers and so greatly reduce
the attack surface for potential exploits (or at least tools like
trinity ...).

Also add the drm_legacy prefix for functions which are called from
common code. To further reduce the impact on common code also extract
all the ctx release handling into a function (instead of only
releasing individual handles) and make ctxbitmap_cleanup return void -
it can never fail.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: disallow legacy dma ioctls for modesetting drivers
Daniel Vetter [Thu, 8 Aug 2013 13:41:20 +0000 (15:41 +0200)]
drm: disallow legacy dma ioctls for modesetting drivers

Now only legacy ums drivers have the DRIVER_HAVE_DMA driver feature
flag set, so strictly speaking the modesetting check is redundant. But
adding it has the upside that it makes it very clear that the dma
support is legacy stuff.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: mark dma setup/teardown as legacy systems
Daniel Vetter [Thu, 8 Aug 2013 13:41:19 +0000 (15:41 +0200)]
drm: mark dma setup/teardown as legacy systems

And hide the checks a bit better. This was already disallowed for
modesetting drivers, so no functinal change here.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: disallow legacy sg ioctls for modesetting drivers
Daniel Vetter [Thu, 8 Aug 2013 13:41:18 +0000 (15:41 +0200)]
drm: disallow legacy sg ioctls for modesetting drivers

Only the radeon/r128/ati ums drivers use this. Furthermore the cleanup
was already only done for UMS drivers. Also a quick check of the ATI
ddx git history shows that only the UMS code ever used this facility.

So we can safely disallow these pair of ioctls for modesetting
drivers.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: hide legacy sg cleanup better from common code
Daniel Vetter [Thu, 8 Aug 2013 13:41:17 +0000 (15:41 +0200)]
drm: hide legacy sg cleanup better from common code

I've decided that some clear markers for what's legacy dri1/non-gem
code is useful. I've opted to use the drm_legacy prefix and then hide
all the checks in that function for better readability in the common
code.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: kill dev->driver->set_version
Daniel Vetter [Thu, 8 Aug 2013 13:41:16 +0000 (15:41 +0200)]
drm: kill dev->driver->set_version

Totally unused, so just rip it out. Anyway, we want drivers to be
fully backwards compatible, allowing them to change behaviour is just
a recipe for them to break badly.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon: kill firstopen callback for kms driver
Daniel Vetter [Thu, 8 Aug 2013 13:41:12 +0000 (15:41 +0200)]
drm/radeon: kill firstopen callback for kms driver

Again, it does nothing.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/omap: kill firstopen callback
Daniel Vetter [Thu, 8 Aug 2013 13:41:11 +0000 (15:41 +0200)]
drm/omap: kill firstopen callback

KMS drivers really shouldn't need to do anything on firstopen, so kill
empty callbacks.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ttm: kill unused functions
Maarten Lankhorst [Thu, 15 Aug 2013 14:03:17 +0000 (16:03 +0200)]
drm/ttm: kill unused functions

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/radeon: remove stale gem->driver_private access
David Herrmann [Wed, 14 Aug 2013 13:07:18 +0000 (15:07 +0200)]
drm/radeon: remove stale gem->driver_private access

This field is never read. No need to set it in radeon. Besides, DRM gem
core clears it during setup, anyway.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/qxl: remove unused object_pin/unpin() helpers
David Herrmann [Wed, 14 Aug 2013 13:07:17 +0000 (15:07 +0200)]
drm/qxl: remove unused object_pin/unpin() helpers

These two helpers are unused. Remove them. They rely on
gem_obj->driver_private, which is set to NULL during setup. As this field
isn't used by the driver, anymore, we can remove this assignment as well.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/cirrus: remove unused driver_private access
David Herrmann [Wed, 14 Aug 2013 13:07:16 +0000 (15:07 +0200)]
drm/cirrus: remove unused driver_private access

gem_bo->driver_private is never read by cirrus nor DRM core. No need to
set it. Besides, drm core clears it during setup, anyway.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/mgag200: remove unused driver_private access
David Herrmann [Wed, 14 Aug 2013 13:07:15 +0000 (15:07 +0200)]
drm/mgag200: remove unused driver_private access

gem_bo->driver_private is never read by mgag200 nor DRM core. No need to
set it. Besides, drm core clears it during setup, anyway.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/ast: remove unused driver_private access
David Herrmann [Wed, 14 Aug 2013 13:07:14 +0000 (15:07 +0200)]
drm/ast: remove unused driver_private access

gem_bo->driver_private is never read by ast nor DRM core. No need to set
it. Besides, drm core clears it during setup, anyway.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge remote-tracking branch 'pfdo/drm-rcar-for-v3.12' into drm-next
Dave Airlie [Sun, 18 Aug 2013 23:24:13 +0000 (09:24 +1000)]
Merge remote-tracking branch 'pfdo/drm-rcar-for-v3.12' into drm-next

Merge the rcar stable branch that is being shared with the arm-soc tree.

Signed-off-by: Dave Airlie <airlied@redhat.com>
* pfdo/drm-rcar-for-v3.12: (220 commits)
  drm/rcar-du: Add FBDEV emulation support
  drm/rcar-du: Add internal LVDS encoder support
  drm/rcar-du: Configure RGB output routing to DPAD0
  drm/rcar-du: Rework output routing support
  drm/rcar-du: Add support for DEFR8 register
  drm/rcar-du: Add support for multiple groups
  drm/rcar-du: Fix buffer pitch alignment for R8A7790 DU
  drm/rcar-du: Add support for the R8A7790 DU
  drm/rcar-du: Move output routing configuration to group
  drm/rcar-du: Remove register definitions for the second channel
  drm/rcar-du: Use dynamic number of CRTCs instead of CRTCs array size
  drm/rcar-du: Introduce CRTCs groups
  drm/rcar-du: Rename rcar_du_plane_(init|register) to rcar_du_planes_*
  drm/rcar-du: Create rcar_du_planes structure
  drm/rcar-du: Rename platform data fields to match what they describe
  drm/rcar-du: Merge LVDS and VGA encoder code
  drm/rcar-du: Split VGA encoder and connector
  drm/rcar-du: Split LVDS encoder and connector
  drm/rcar-du: Clarify comment regarding plane Y source coordinate
  drm/rcar-du: Support per-CRTC clock and IRQ
  ...

Conflicts:
drivers/gpu/drm/i915/i915_dma.c
drivers/gpu/drm/i915/intel_pm.c
drivers/gpu/drm/qxl/qxl_release.c

11 years agodrm/tilcdc fixup mode to workaround sync for tda998x
Darren Etheridge [Wed, 14 Aug 2013 19:43:33 +0000 (21:43 +0200)]
drm/tilcdc fixup mode to workaround sync for tda998x

Add a fixup function that will flip the hsync priority and
add a hskew value that is used to shift the tda998x to the
right by a variable number of pixels depending on the mode.
This works around an issue with the sync timings that tilcdc
is outputing.

Signed-off-by: Darren Etheridge <detheridge@ti.com>
Tested-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/i2c: tda998x: prepare for broken sync workaround
Sebastian Hesselbarth [Wed, 14 Aug 2013 19:43:32 +0000 (21:43 +0200)]
drm/i2c: tda998x: prepare for broken sync workaround

Some LCD controller cannot provide valid VESA style sync, i.e. coincident
HS/VS edges. First, this patch adds hskew passed from the adjusted_mode to
reference pixel calculation to allow those controllers to add an offset
relative to the expected reference pixel.

Signed-off-by: Darren Etheridge <detheridge@ti.com>
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/i2c: tda998x: fix sync generation and calculation
Sebastian Hesselbarth [Wed, 14 Aug 2013 19:43:31 +0000 (21:43 +0200)]
drm/i2c: tda998x: fix sync generation and calculation

This fixes the wrong sync generation and sync calculation of TDA998x
for HS/VS-based sync detection.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Darren Etheridge <detheridge@ti.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/i2c: tda998x: add video and audio input configuration
Russell King [Wed, 14 Aug 2013 19:43:30 +0000 (21:43 +0200)]
drm/i2c: tda998x: add video and audio input configuration

This patch adds tda998x specific parameters to allow it to be configured
for different boards using it. Also, this implements rudimentary audio
support for S/PDIF attached controllers.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Darren Etheridge <detheridge@ti.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/i2c: tda998x: prepare for video input configuration
Russell King [Wed, 14 Aug 2013 19:43:29 +0000 (21:43 +0200)]
drm/i2c: tda998x: prepare for video input configuration

The video-input-port (VIP) is highly configurable. This prepares
current driver to allow to configure VIP configuration, as some
boards connect lcd controller and TDA998x "pin-swapped" and depend
on VIP to swap the pins by register configuration.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Tested-by: Darren Etheridge <detheridge@ti.com>
Tested-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/i2c: tda998x: fix npix/nline programming
Russell King [Wed, 14 Aug 2013 19:43:28 +0000 (21:43 +0200)]
drm/i2c: tda998x: fix npix/nline programming

The npix/nline registers are supposed to be programmed with the total
number of pixels/lines, not the displayed pixels/lines, and not minus
one either.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Tested-by: Darren Etheridge <detheridge@ti.com>
Tested-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/i2c: tda998x: ensure VIP output mux is properly set
Russell King [Wed, 14 Aug 2013 19:43:27 +0000 (21:43 +0200)]
drm/i2c: tda998x: ensure VIP output mux is properly set

When switching between various drivers for this device, it's possible
that some critical registers are left containing values which affect
the device operation.  One such case encountered is the VIP output
mux register.  This defaults to 0x24 on powerup, but other drivers may
set this to 0x12.  This results in incorrect colours.

Fix this by ensuring that the register is always set to the power on
default setting.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Tested-by: Darren Etheridge <detheridge@ti.com>
Tested-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/i2c: tda998x: fix EDID reading on TDA19988 devices
Russell King [Wed, 14 Aug 2013 19:43:26 +0000 (21:43 +0200)]
drm/i2c: tda998x: fix EDID reading on TDA19988 devices

TDA19988 devices need their RAM enabled in order to read EDID
information.  Add support for this.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Rob Clark <robdclark@gmail.com>
Tested-by: Darren Etheridge <detheridge@ti.com>
Tested-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Russell King <rmk_kernel@arm.linux.org.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm: DRM should depend on HAS_DMA
Geert Uytterhoeven [Sun, 18 Aug 2013 10:54:25 +0000 (12:54 +0200)]
drm: DRM should depend on HAS_DMA

If NO_DMA=y:

drivers/built-in.o: In function `__drm_pci_free':
drivers/gpu/drm/drm_pci.c:112: undefined reference to `dma_free_coherent'
drivers/built-in.o: In function `drm_pci_alloc':
drivers/gpu/drm/drm_pci.c:72: undefined reference to `dma_alloc_coherent'
drivers/built-in.o: In function `drm_gem_unmap_dma_buf':
drivers/gpu/drm/drm_prime.c:87: undefined reference to `dma_unmap_sg'
drivers/built-in.o: In function `drm_gem_map_dma_buf':
drivers/gpu/drm/drm_prime.c:78: undefined reference to `dma_map_sg'

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agodrm/gma500/cdv: Add and hook up chip op for disabling sr
Patrik Jakobsson [Wed, 14 Aug 2013 22:54:44 +0000 (00:54 +0200)]
drm/gma500/cdv: Add and hook up chip op for disabling sr

Add a callback hook to the chip ops struct to allow chips to have their
specific self-refresh function. Currently only used by cdv.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
11 years agodrm/gma500/cdv: Add and hook up chip op for watermarks
Patrik Jakobsson [Wed, 14 Aug 2013 17:14:17 +0000 (19:14 +0200)]
drm/gma500/cdv: Add and hook up chip op for watermarks

Add a callback hook to the chip ops struct to allow chips to have their
specific fifo watermark update function. Currently only cdv actually
tries to set wms based on crtc configuration but if/when the other chips
needs it we can attach a callback for them as well.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
11 years agoMerge branch 'drm/next/du' of git://linuxtv.org/pinchartl/fbdev into drm-rcar-stable
Dave Airlie [Fri, 9 Aug 2013 21:16:50 +0000 (07:16 +1000)]
Merge branch 'drm/next/du' of git://linuxtv.org/pinchartl/fbdev into drm-rcar-stable

Create topic branch for rcar for shmobile tree to pull as well, arm-soc should
probably merge after drm merges if possible.

Signed-off-by: Dave Airlie <airlied@redhat.com>
* 'drm/next/du' of git://linuxtv.org/pinchartl/fbdev: (23 commits)
  drm/rcar-du: Add FBDEV emulation support
  drm/rcar-du: Add internal LVDS encoder support
  drm/rcar-du: Configure RGB output routing to DPAD0
  drm/rcar-du: Rework output routing support
  drm/rcar-du: Add support for DEFR8 register
  drm/rcar-du: Add support for multiple groups
  drm/rcar-du: Fix buffer pitch alignment for R8A7790 DU
  drm/rcar-du: Add support for the R8A7790 DU
  drm/rcar-du: Move output routing configuration to group
  drm/rcar-du: Remove register definitions for the second channel
  drm/rcar-du: Use dynamic number of CRTCs instead of CRTCs array size
  drm/rcar-du: Introduce CRTCs groups
  drm/rcar-du: Rename rcar_du_plane_(init|register) to rcar_du_planes_*
  drm/rcar-du: Create rcar_du_planes structure
  drm/rcar-du: Rename platform data fields to match what they describe
  drm/rcar-du: Merge LVDS and VGA encoder code
  drm/rcar-du: Split VGA encoder and connector
  drm/rcar-du: Split LVDS encoder and connector
  drm/rcar-du: Clarify comment regarding plane Y source coordinate
  drm/rcar-du: Support per-CRTC clock and IRQ
  ...

11 years agodrm/rcar-du: Add FBDEV emulation support
Laurent Pinchart [Thu, 14 Mar 2013 21:45:22 +0000 (22:45 +0100)]
drm/rcar-du: Add FBDEV emulation support

Use the FB CMA helpers to implement FBDEV emulation support. The VGA
connector status must be reported as connector_status_connected instead
of connector_status_unknown to be usable by the emulation layer.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Add internal LVDS encoder support
Laurent Pinchart [Mon, 17 Jun 2013 11:48:27 +0000 (13:48 +0200)]
drm/rcar-du: Add internal LVDS encoder support

The R8A7790 includes two internal LVDS encoders. Support them in the DU
driver.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Configure RGB output routing to DPAD0
Laurent Pinchart [Mon, 17 Jun 2013 01:20:08 +0000 (03:20 +0200)]
drm/rcar-du: Configure RGB output routing to DPAD0

The R8A7790 DU variant has a single RGB output called DPAD0 that can be
fed with the output of DU0, DU1 or DU2. Making the routing configurable.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Rework output routing support
Laurent Pinchart [Mon, 17 Jun 2013 01:13:11 +0000 (03:13 +0200)]
drm/rcar-du: Rework output routing support

Split the output routing specification between SoC-internal data,
specified in the rcar_du_device_info structure, and board data, passed
through platform data.

The DU has 5 possible outputs (DPAD0/1, LVDS0/1, TCON). SoC-internal
output routing data specify which output are valid, which CRTCs can be
connected to the valid outputs, and the type of in-SoC encoder for the
output.

Platform data then specifies external encoders and the output they are
connected to.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Add support for DEFR8 register
Laurent Pinchart [Sat, 15 Jun 2013 00:40:57 +0000 (02:40 +0200)]
drm/rcar-du: Add support for DEFR8 register

The R8A7790 DU has a new extended function control register. Support it.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Add support for multiple groups
Laurent Pinchart [Sun, 16 Jun 2013 22:29:25 +0000 (00:29 +0200)]
drm/rcar-du: Add support for multiple groups

The R8A7790 DU has 3 CRTCs, split in two groups. Support them.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Fix buffer pitch alignment for R8A7790 DU
Laurent Pinchart [Fri, 14 Jun 2013 18:52:52 +0000 (20:52 +0200)]
drm/rcar-du: Fix buffer pitch alignment for R8A7790 DU

The R8A7790 DU seems to require a 128 bytes pitch alignment, even though
the documentation only mentions a 16 pixels alignement as for the
R8A7779 DU. Make this configurable through a device flag.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Add support for the R8A7790 DU
Laurent Pinchart [Fri, 14 Jun 2013 12:16:35 +0000 (14:16 +0200)]
drm/rcar-du: Add support for the R8A7790 DU

The DU revision in the R8A7790 SoC uses one IRQ and clock per CRTC. Add
a corresponding entry in the module platform ID table.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Move output routing configuration to group
Laurent Pinchart [Sun, 16 Jun 2013 22:11:05 +0000 (00:11 +0200)]
drm/rcar-du: Move output routing configuration to group

Output routing is configured in group registers, move the corresponding
code from rcar_du_crtc.c to rcar_du_group.c.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Remove register definitions for the second channel
Laurent Pinchart [Sun, 16 Jun 2013 21:48:10 +0000 (23:48 +0200)]
drm/rcar-du: Remove register definitions for the second channel

Channels are accessed through a global channel memory offset, there's no
need to define register addresses for the second channel.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Use dynamic number of CRTCs instead of CRTCs array size
Laurent Pinchart [Sun, 16 Jun 2013 20:22:23 +0000 (22:22 +0200)]
drm/rcar-du: Use dynamic number of CRTCs instead of CRTCs array size

The rcar_du_device structure contains a field that stores the number of
CRTCs, use it instead of the CRTCs array size. This prepares the driver
to support a variable number of CRTCs.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Introduce CRTCs groups
Laurent Pinchart [Sun, 16 Jun 2013 19:01:02 +0000 (21:01 +0200)]
drm/rcar-du: Introduce CRTCs groups

The R8A7779 DU is split in per-CRTC resources (scan-out engine, blending
unit, timings generator, ...) and device-global resources (start/stop
control, planes, ...) shared between the two CRTCs.

The R8A7790 introduced a third CRTC with its own set of global resources
This would be modeled as two separate DU device instances if it wasn't
for a handful or resources that are shared between the three CRTCs
(mostly related to input and output routing). For this reason the
R8A7790 DU must be modeled as a single device with three CRTCs, two sets
of "semi-global" resources, and a few device-global resources.

Introduce a new rcar_du_group driver-specific object, without any real
counterpart in the DU documentation, that models those semi-global
resources.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Rename rcar_du_plane_(init|register) to rcar_du_planes_*
Laurent Pinchart [Sun, 16 Jun 2013 17:18:31 +0000 (19:18 +0200)]
drm/rcar-du: Rename rcar_du_plane_(init|register) to rcar_du_planes_*

The functions initialize or register all planes, rename them
accordingly.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
11 years agodrm/rcar-du: Create rcar_du_planes structure
Laurent Pinchart [Sun, 16 Jun 2013 19:02:49 +0000 (21:02 +0200)]
drm/rcar-du: Create rcar_du_planes structure

Move the plane-related fields of struct rcar_du_device to their own
structure.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>