drm/nouveau: convert to dev_pm_ops

This is a precursor to dynamic power management support for nouveau,

we need to use pm ops for that, so first convert the driver to using pm ops
interfaces.

Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: Add interface to detect optimus and v1 support

This is required to decide if we can auto-powerdown and how to implement it.

Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/disp: add support for 10bpc over DisplayPort

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: rename nvd0_display to nv50_display to reflect reality since merge

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50-nvc0: switch to common disp impl, removing previous version

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: implement support for colour vibrance control

Ported from original nv50 commit by Christoph, with added support for the
setting on later chipsets.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: remove fetching of capabilities report

Unused, and caused a race with evo_sync().  Will revisit using it properly
later on.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: enable support for older display classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: support creation of fb dma objects on older chipsets

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: implement support for older DISP_SYNC classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: implement sor support for older display classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: implement dac support for older display classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: start implementing support for older display classes

Currently unused, but checkpointing the merged head handling routines.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/dp: remove last bits of VBIOS parsing from DRM code

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/dp: move core link training calls to common code

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: remove (now obsolete) BIT U table parsing from DRM code

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: move remaining interrupt handling into core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: calculate U script id in supervisor interrupt

This is like we do on nv50:nvd9 already.  There's been no problems seen
yet with using this *seemingly* scratch register to store the value, but
we won't be able to do this anymore once nv50's code is merged.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv84/disp: move hdmi control into core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nva3/disp: move hda codec handling to core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: move dp link training helpers into core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: call into core for dac load detection

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: call into core to handle dac/sor power state changes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: allocate display from driver core

EVO channels still handled "manually", this won't be ported here, and
will instead be held off until nv50_display/nvd0_display are merged.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/bios: implement some dcb output entry parsing/matching functions

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: handle DP transfer unit setup from second supervisor interrupt

This is what we've done forever in nv50_display.c, and also allows the
last direct MMIO accesses to be removed from nvd0_display.c.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: move HDMI control to core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: move HDA codec setup to core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: call into core to handle dac power state changes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: call into core to handle sor power state changes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: move link training helpers into core as display methods

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/core: allow representing method ranges in nouveau_omthds

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/core: expose full method calling capabilities with nv_exec

nv_call() just allows mthd+u32 submission, nv_exec() exposes the
mthd+data+size interface which will be used in future commits.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/bios: implement "full" BIT 'd' table (and subtable) parsing in core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: move remaining interrupt handling to core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/core: add some missing subdev/engine disable flags

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/bios: implement BIT 'U' table (and subtable) parsing in core

This will, in the near future, replace what's currently in the DRM
nouveau_bios.c code.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: allocate display and evo channels from driver core

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: run mode_config destructor before destroying internal display state

Later changes will depend on being able to pull down CRTCs etc with the
master display state still intact.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/drm/nvd0/disp: display->disp

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/disp: introduce a nvd0_head as a subclass of nouveau_crtc

This will be used instead of storing a heap of per-head data (such as evo
channels) in nvd0_display in some other way.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0-nve0/disp: initial implementation of evo channel classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: allow PCI_US pushbuf binding

drm/nv50/disp: initial implementation of the various channel object classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: enable interrupts and setup memory area

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: copy caps to evo

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/disp: create skeleton display/channel object classes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/dmaobj: initial bind() method implementation

Currently unused, and rudimentary.  Lots to figure out here still, but
this is sufficient for what disp will need.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvc0/dmaobj: implement initial bind() method

Currently unused.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50/dmaobj: extend class to allow gpu-specific attributes to be defined

disp is going to need to be able to create more specific dma objects
than was previously possible.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvd0/dmaobj: duplicate fermi class, will diverge real soon now

The hardware dmaobj format completely changed in GF119, so these will
need a separate implementation.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nvc0/dmaobj: stub bind function for now so we can call unconditionally

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/dmaobj: move parent class check to bind() method

Otherwise when nvc0- gains a bind() method (disp needs it), the fifo
engine will attempt to create a dma object for the push buffer, which
is unnecessary on fermi.

The only sane place to put these checks is in the bind method itself,
and have it unconditionally called from wherever it might be needed.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/dmaobj: merge everything except ctor and bind together

Simplifies things a little, and currently no reason to need
chipset-specific dmaobj constructors.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: pass address to object accessor functions as u64

Will be required by future work.  Make the API change now to catch any
(but hopefully none) unexpected fallout.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/gpio: place upper limit on using old-school tvdac bios data

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: remove newline-only NV_DEBUG calls

This used to output the function name, now doesn't, so just looks
stupid.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50: decode PGRAPH status registers on TLB flush fail

SIgned-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/clk: fix crystal frequency retrieval on nv25

Signed-off-by: Viktor Novotný <noviktor@seznam.cz>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: remove unused variable from nouveau_bios_shadow_of

drivers/gpu/drm/nouveau/core/subdev/bios/base.c: warning: unused variable 'i'

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv40/fb: use an actual compressed zeta format

SPLIT is apparently just that, and only splits Z and S components.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv30/fb: enable z compression

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv40/fb: enable z compression

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv20/fb: fixup compression tag allocation size

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv30-nv40/graph: poke zcomp regs from tile_prog hook

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv30-nv40/fb: call zcomp setup hook from tiling setup

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fb: create tag heap from common code for all relevant chipsets

A nv2x bug wrt hardcoded tag counts is now also fixed as a side-effect.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv30/fb: start bashing zcomp registers with 'disabled' (for now)

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: replace some open-coded mm_initialised checks

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fb: read TILE_BASE after writing it to avoid a hardware race

Apparently needed for turbocache nv4x chips at least, we'll just do it
everywhere...

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv40/fb: start bashing zcomp registers on relevant chipsets

Always bashing "disabled" for now, actual compressing coming up...

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv30-nv40/fb: make use of bankoff for zeta buffers, where supported

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv20/fb: fix zcomp register calculation on big-endian systems

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv40/fb: split implementation into nv40/nv41/nv44/nv46/nv47/nv49/nv4e pieces

Wow, this is a nice complicated mess of build-your-own-mc blocks...

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv30/fb: split implementation into nv34(nv10)/nv30/nv35 pieces

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv20/fb: split implementation into nv20/nv25 pieces

There's more stuff that can be shared in the constructor, will be merged
together again later.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/fb: split tile and compression region handling

This is in preparation for extending the support to the remaining
chipsets, to allow for sharing more functions.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/bios: attempt to fetch entire acpi rom image in one shot

v2: fdo#55948 - the _ROM method silently truncates size to 4KiB, perform
    a checksum test and fall back to slow _ROM access on failure.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/doc: add new dp helpers into drm DocBook

I didn't bother with documenting the really trivial new "extract
something from dpcd" helpers, but the i2c over aux ch is now
documented a bit.

v2: Clarify the comment for i2c_dp_aux_add_bus a bit.

v3: Fix more spelling fail spotted by Laurent Pinchart.

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>

drm/doc: integrate fb helper reference into docs

Again only minimal changes to make kerneldoc no longer shout. Plus a
little introduction in the form of a inline DOC: section to quickly
explain what this is all about.

v2: Fixup spelling fail.

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>

drm/doc: integrate crtc helper api into docbook

- Add the missing doc for drm_helper_move_panel_connectors_to_head.
- Fixup any outdated stuff in existing sections. I've only looked at
  those kerneldoc headers that actually resulted in a complaint from
  the kerneldoc parser tool.

v2:
- Actually include the docbook snippet in the right patch.
- Fix spelling fail.

v3: It's now called drm_crtc_helper_set_mode, spotted by Laurent
Pinchart.

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>

drm/doc: Helpers are not a Midlayer!

I'm devoting all my wrath to that fight, so don't misname it ;-)

v2: Make it clear that this section talks about kms helpers.

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>

drm/edid: tune down debug message in parse_hdmi_vsdb

Those tend to be totally not interesting for end-users, and for
debugging we tend to dump the entire noise anyway by enabling all
debug messages.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=57388
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm: only build ati_pcigart if PCI enabled

Prevent ati_pcigart.c being built unless PCI is enabled. The exported
functions in this file are only used by drivers which depend on PCI
(namely r128 and radeon), and it tries to use PCI specific functions
(pci_unmap_page, pci_map_page, and pci_dma_mapping_error) that cause
compiler errors when PCI is disabled.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Tighten the security around buffer maps

Make sure that other DRM clients can't map the contents of
non-shareable buffer objects.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm: tegra: Add Tegra30 support

Add support for host1x, the display controllers and HDMI on the Tegra30
SoC.

Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de>
Tested-by: Stephen Warren <swarren@nvidia.com>
Tested-by: Mark Zhang <markz@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm: tegra: Use framebuffer pitch as line stride

Instead of using the stride derived from the display mode, use the pitch
associated with the currently active framebuffer. This fixes a bug where
the LCD display content would be skewed when enabling HDMI with a video
mode different from that of the LCD.

Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Mark Zhang <markz@nvidia.com>
Tested-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/ttm: Optimize vm locking using kref_get_unless_zero v3

Removes the need for a write lock each time we call ttm_bo_unref().

v2: Remove an unused variable.
v3: Really remove the unused variable.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/ttm: Use the hashtab _rcu interface for ttm_objects

Also move a kref_init() out of spinlocked region

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/ttm: Fix locking in an error path

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Free user-space fence objects correctly

They need to be freed after an rcu grace period.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

kref: Add kref_get_unless_zero documentation

Document how kref_get_unless_zero should be used and how it helps
solve a typical kref / locking problem.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm: Add a hash-tab rcu-safe API

While hashtab should now be RCU-safe, Add a drm_ht_xxx_api for consumers
to use to make it obvious what locking mechanism is used.

Document the way the rcu-safe interface should be used.

Don't use rcu-safe list traversal in modify operations where we should use
a spinlock / mutex anyway.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Add and make use of a header for surface size calculation.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Break out surface and context management to separate files

Add a resource-private header for common resource definitions

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Refactor resource management

Refactor resource management to make it easy to hook up resources
that are backed up by buffers. In particular, resources and their
backing buffers can be evicted and rebound, if supported by the device.
To avoid query deadlocks, the query code is also modified somewhat.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Make vmw_dmabuf_unreference handle NULL objects

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Refactor module load to not require fifo unless fbdev is loaded

This also fixes a bug where the fence manager was left without irq
enabled when waiting for fences, causing various errors at module
load time

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Make screen object code not require fifo at init time

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/vmwgfx: Make overlay code not require fifo at init time

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dmitry Torokhov <dtor@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>