drm/nouveau/fb: remove dependence on namedb/engctx lookup

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

drm/nouveau/sw: turn flip completion into an event

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

drm/nouveau/sw/nv04: replace direct context access with GetRef method

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

drm/nouveau/ramht: remove dependence on namedb

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

drm/nouveau/gpuobj: separate allocation from nvkm_object

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

drm/nouveau/gr: directly use instmem where currently possible

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

drm/nouveau/fifo: directly use instmem for runlists and polling areas

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

drm/nouveau/xtensa: directly use instmem for ucode image

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

drm/nouveau/falcon: directly use instmem for ucode image

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

drm/nouveau/imem/nv04-nv40: directly use instmem for vbios/ramfc/ramro

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

drm/nouveau/bar/gf100: directly use instmem for channel descriptors

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

drm/nouveau/mmu: directly use instmem for page tables

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

drm/nouveau/imem: improve management of instance memory

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

drm/nouveau/mmu: protect each vm with its own mutex

An upcoming commit requires being able to modify the PRAMIN BAR page
tables while already holding the MMU subdev mutex.

To solve this issue, each VM has been given its own mutex.  As a nice
side-effect, this also allows separate VMs to be updated concurrently.

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

drm/nouveau/core: prepare for new-style objects

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

drm/nouveau/device: prepare for new-style subdevs

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

drm/nouveau/device: separate construction of pci/tegra devices

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

drm/nouveau/engine: implement support for new-style nvkm_engine

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

drm/nouveau/subdev: implement support for new-style nvkm_subdev

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

drm/nouveau/object: implement support for new-style nvkm_object

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

drm/nouveau/engine: rename some functions to avoid upcoming conflicts

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

drm/nouveau/subdev: rename some functions to avoid upcoming conflicts

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

drm/nouveau/object: rename some functions to avoid upcoming conflicts

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

drm/nouveau/nvif: device time mthd

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

drm/nouveau/nvif: replace pushbuf with vm in fermi/kepler gpfifo class args

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

drm/nouveau/nvif: return chipset/board names from device info method

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

drm/nouveau/nvif: add supported engines query to kepler gpfifo class

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

drm/nouveau/nvif: return min/max versions for supported object classes

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

drm/nouveau/nvif: assign internal class identifiers to sw classes

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

drm/nouveau/nvif: use negative oclass identifier for internal classes

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

drm/nouveau/nvif: return success from sclass even for objects without children

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

drm/nouveau/nvif: extend nop ioctl to return nvif version identifier

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

drm/nouveau/nvif: replace path-based object identification

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

drm/nouveau/nvif: simplify and tidy library interfaces

A variety of tweaks to the NVIF library interfaces, mostly ripping out
things that turned out to be not so useful.

- Removed refcounting from nvif_object, callers are expected to not be
  stupid instead.
- nvif_client is directly reachable from anything derived from nvif_object,
  removing the need for heuristics to locate it
- _new() versions of interfaces, that allocate memory for the object
  they construct, have been removed.  The vast majority of callers used
  the embedded _init() interfaces.
- No longer storing constructor arguments (and the data returned from
  nvkm) inside nvif_object, it's more or less unused and just wastes
  memory.

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

drm/nouveau/client: store default device by handle, not reference

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

drm/nouveau/parent: remove some (now) unnecessary special-case handling

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

drm/nouveau/object: store object type data outside of handle

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

drm/nouveau/device: decouple from engine machinery

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

drm/nouveau/device: split user device implementation out on its own

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

drm/nouveau/device: simplify subdev construction

Replaces the piece-by-piece (in response to NV_DEVICE ctor args) device
contruction with a once-off all-or-nothing approach, eliminating some
tricky refcounting issues.  The partial device init capability was only
required by some tools, and has been moved to probe time instead.

Temporarily removes a workaround for some boards where we need to fiddle
with AGP registers before executing the DEVINIT scripts.  A later commit
in this series reinstates it.

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

drm/nouveau/bin: punt client/device argument handling into a common helper

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

drm/nouveau/device: tidy ctor/dtor interfaces

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

drm/nouveau/client: tidy ctor/dtor interfaces

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

drm/nouveau/platform: remove subclassing of nvkm_device

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

drm/nouveau/disp: transition outp/conn away from being based on nvkm_object

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

drm/nouveau/i2c: transition pad/ports away from being based on nvkm_object

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

drm/nouveau/fb: transition nvkm_ram away from being based on nvkm_object

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

drm/nouveau/vga: require nvkm_device pointer in accessor functions

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

drm/nouveau/imem: use object accessor function for suspend/resume

Very much not ideal, but this will be improved later.

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

drm/nouveau/core: wrap direct use of object accessor functions

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

drm/nouveau/mpeg: switch to gpuobj accessor macros

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

drm/nouveau/gr: switch to gpuobj accessor macros

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

drm/nouveau/fifo: switch to gpuobj accessor macros

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

drm/nouveau/dma: switch to gpuobj accessor macros

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

drm/nouveau/cipher: switch to gpuobj accessor macros

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

drm/nouveau/xtensa: switch to gpuobj accessor macros

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

drm/nouveau/falcon: switch to gpuobj accessor macros

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

drm/nouveau/mmu: switch to gpuobj accessor macros

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

drm/nouveau/bar: switch to gpuobj accessor macros

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

drm/nouveau/core: switch to gpuobj accessor macros

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

drm/nouveau/gpuobj: type-safe accessor macros

These require an explicit struct nvkm_gpuobj pointer, unlike the previous
macros which take a void *, and work with any nvkm_object.

New semantics require acquiring/releasing a gpuobj before accessing them,
which will be made use of in later patches to greatly reduce the overhead
of accesses, particularly when a direct mmio mapping of the object is not
available (suspend/resume, out of ioremap() space, and on GK20A).

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

drm/nouveau/xtensa: remove object accessor functions

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

drm/nouveau/falcon: remove object accessor functions

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

drm/nouveau/imem: remove object accessor functions

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

drm/nouveau/fuse: remove object accessor functions

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

drm/nouveau/bios: remove object accessor functions

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

drm/nouveau/core: remove last printks

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

drm/nouveau: use dev_* for logging

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

drm/nouveau/sec: switch to subdev printk macros

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

drm/nouveau/pm: switch to subdev printk macros

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

drm/nouveau/mpeg: switch to subdev printk macros

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

drm/nouveau/gr: switch to subdev printk macros

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

drm/nouveau/fifo: switch to subdev printk macros

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

drm/nouveau/disp: switch to subdev printk macros

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

drm/nouveau/device: switch to dev_printk macros

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

drm/nouveau/cipher: switch to subdev printk macros

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

drm/nouveau/ce: switch to subdev printk macros

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

drm/nouveau/xtensa: switch to subdev printk macros

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

drm/nouveau/falcon: switch to subdev printk macros

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

drm/nouveau/volt: switch to subdev printk macros

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

drm/nouveau/tmr: switch to subdev printk macros

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

drm/nouveau/therm: switch to subdev printk macros

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

drm/nouveau/pmu: switch to subdev printk macros

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

drm/nouveau/mxm: switch to subdev printk macros

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

drm/nouveau/mmu: switch to subdev printk macros

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

drm/nouveau/mc: switch to subdev printk macros

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

drm/nouveau/ltc: switch to subdev printk macros

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

drm/nouveau/imem: switch to subdev printk macros

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

drm/nouveau/ibus: switch to subdev printk macros

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

drm/nouveau/i2c: switch to subdev printk macros

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

drm/nouveau/fb: switch to subdev printk macros

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

drm/nouveau/devinit: switch to subdev printk macros

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

drm/nouveau/clk: switch to subdev printk macros

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

drm/nouveau/bus: switch to subdev printk macros

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

drm/nouveau/bios: switch to subdev printk macros

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

drm/nouveau/bar: switch to subdev printk macros

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

drm/nouveau/core: type-safe printk macros

These require an explicit pointers to nvkm_object/nvkm_subdev/nvkm_device,
depending on which macros are used.  This is unlike the previous macros
which take a void *, and work for anything derived from nvkm_object (by
way of some awful heuristics).

The output will be a bit confused until everything has been transitioned,
as the logging format used is a more standard style that previously.

In addition, usage of pr_cont(), which doesn't work correctly with the
dev_*() printk functions (and was potentially racy to begin with), will
be replaced.

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

drm/nouveau: switch to new-style timer macros

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

drm/nouveau/mpeg: switch to new-style timer macros

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

drm/nouveau/gr: switch to new-style timer macros

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