Commit Graph

22218 Commits

Author SHA1 Message Date
Ben Skeggs
6cf813fb26 drm/nouveau/device: prepare for new-style subdevs
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:35 +10:00
Ben Skeggs
7974dd1bdb drm/nouveau/device: separate construction of pci/tegra devices
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:34 +10:00
Ben Skeggs
168c2e213d drm/nouveau/engine: implement support for new-style nvkm_engine
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:34 +10:00
Ben Skeggs
f0290215c4 drm/nouveau/subdev: implement support for new-style nvkm_subdev
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:34 +10:00
Ben Skeggs
cbea21e2ab drm/nouveau/object: implement support for new-style nvkm_object
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:34 +10:00
Ben Skeggs
89c651e221 drm/nouveau/engine: rename some functions to avoid upcoming conflicts
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:34 +10:00
Ben Skeggs
3a8c3400f3 drm/nouveau/subdev: rename some functions to avoid upcoming conflicts
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:33 +10:00
Ben Skeggs
aa35888ff0 drm/nouveau/object: rename some functions to avoid upcoming conflicts
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:33 +10:00
Ben Skeggs
d61f4c178c drm/nouveau/nvif: device time mthd
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:33 +10:00
Ben Skeggs
159045cdc4 drm/nouveau/nvif: replace pushbuf with vm in fermi/kepler gpfifo class args
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:33 +10:00
Ben Skeggs
22827fa42b drm/nouveau/nvif: return chipset/board names from device info method
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:33 +10:00
Ben Skeggs
d56241794c drm/nouveau/nvif: add supported engines query to kepler gpfifo class
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:33 +10:00
Ben Skeggs
41a634064d drm/nouveau/nvif: return min/max versions for supported object classes
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:32 +10:00
Ben Skeggs
f58ddf9581 drm/nouveau/nvif: assign internal class identifiers to sw classes
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:32 +10:00
Ben Skeggs
315a8b2edf drm/nouveau/nvif: use negative oclass identifier for internal classes
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:32 +10:00
Ben Skeggs
5c15bf709e drm/nouveau/nvif: return success from sclass even for objects without children
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:32 +10:00
Ben Skeggs
99d4d36ad6 drm/nouveau/nvif: extend nop ioctl to return nvif version identifier
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:32 +10:00
Ben Skeggs
bf81df9be2 drm/nouveau/nvif: replace path-based object identification
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:32 +10:00
Ben Skeggs
a01ca78c8f 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>
2015-08-28 12:40:32 +10:00
Ben Skeggs
4e7e62d607 drm/nouveau/client: store default device by handle, not reference
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:31 +10:00
Ben Skeggs
102b49da15 drm/nouveau/parent: remove some (now) unnecessary special-case handling
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:31 +10:00
Ben Skeggs
493f189dc0 drm/nouveau/object: store object type data outside of handle
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:31 +10:00
Ben Skeggs
a1e8873622 drm/nouveau/device: decouple from engine machinery
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:31 +10:00
Ben Skeggs
a1bfb29a33 drm/nouveau/device: split user device implementation out on its own
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:31 +10:00
Ben Skeggs
0ac9d210a1 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>
2015-08-28 12:40:31 +10:00
Ben Skeggs
e781dc8f6c drm/nouveau/device: tidy ctor/dtor interfaces
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:30 +10:00
Ben Skeggs
76ecea5b4b drm/nouveau/client: tidy ctor/dtor interfaces
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:30 +10:00
Ben Skeggs
47b2505efb drm/nouveau/platform: remove subclassing of nvkm_device
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:30 +10:00
Ben Skeggs
f2c906fc0c drm/nouveau/disp: transition outp/conn away from being based on nvkm_object
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:30 +10:00
Ben Skeggs
2aa5eac516 drm/nouveau/i2c: transition pad/ports away from being based on nvkm_object
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:29 +10:00
Ben Skeggs
d36a99d2da drm/nouveau/fb: transition nvkm_ram away from being based on nvkm_object
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:29 +10:00
Ben Skeggs
a8dae9fe0e drm/nouveau/vga: require nvkm_device pointer in accessor functions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:29 +10:00
Ben Skeggs
95cf469cc3 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>
2015-08-28 12:40:29 +10:00
Ben Skeggs
cfdc4c44b4 drm/nouveau/core: wrap direct use of object accessor functions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:29 +10:00
Ben Skeggs
572fb13db2 drm/nouveau/mpeg: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:28 +10:00
Ben Skeggs
142ea05f49 drm/nouveau/gr: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:28 +10:00
Ben Skeggs
5444e770e3 drm/nouveau/fifo: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:28 +10:00
Ben Skeggs
3f532ef1e2 drm/nouveau/dma: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:28 +10:00
Ben Skeggs
6d06fd68cb drm/nouveau/cipher: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:28 +10:00
Ben Skeggs
edb1dc51de drm/nouveau/xtensa: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:28 +10:00
Ben Skeggs
9cc264a3aa drm/nouveau/falcon: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:28 +10:00
Ben Skeggs
cd821077aa drm/nouveau/mmu: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:27 +10:00
Ben Skeggs
1918707556 drm/nouveau/bar: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:27 +10:00
Ben Skeggs
45ea503161 drm/nouveau/core: switch to gpuobj accessor macros
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:27 +10:00
Ben Skeggs
384af9acea 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>
2015-08-28 12:40:27 +10:00
Ben Skeggs
9ccdc76006 drm/nouveau/xtensa: remove object accessor functions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:27 +10:00
Ben Skeggs
63902181a7 drm/nouveau/falcon: remove object accessor functions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:27 +10:00
Ben Skeggs
a47474261e drm/nouveau/imem: remove object accessor functions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:26 +10:00
Ben Skeggs
7e24c1145a drm/nouveau/fuse: remove object accessor functions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:26 +10:00
Ben Skeggs
7f5f518fd7 drm/nouveau/bios: remove object accessor functions
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2015-08-28 12:40:26 +10:00