Merge branch 'drm-next' of git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "This is the main pull request for the drm for 4.3.  Nouveau is
  probably the biggest amount of changes in here, since it missed 4.2.
  Highlights below, along with the usual bunch of fixes.

  All stuff outside drm should have applicable acks.

  Highlights:

   - new drivers:
        freescale dcu kms driver

   - core:
        more atomic fixes
        disable some dri1 interfaces on kms drivers
        drop fb panic handling, this was just getting more broken, as more locking was required.
        new core fbdev Kconfig support - instead of each driver enable/disabling it
        struct_mutex cleanups

   - panel:
        more new panels
        cleanup Kconfig

   - i915:
        Skylake support enabled by default
        legacy modesetting using atomic infrastructure
        Skylake fixes
        GEN9 workarounds

   - amdgpu:
        Fiji support
        CGS support for amdgpu
        Initial GPU scheduler - off by default
        Lots of bug fixes and optimisations.

   - radeon:
        DP fixes
        misc fixes

   - amdkfd:
        Add Carrizo support for amdkfd using amdgpu.

   - nouveau:
        long pending cleanup to complete driver,
        fully bisectable which makes it larger,
        perfmon work
        more reclocking improvements
        maxwell displayport fixes

   - vmwgfx:
        new DX device support, supports OpenGL 3.3
        screen targets support

   - mgag200:
        G200eW support
        G200e new revision support

   - msm:
        dragonboard 410c support, msm8x94 support, msm8x74v1 support
        yuv format support
        dma plane support
        mdp5 rotation
        initial hdcp

   - sti:
        atomic support

   - exynos:
        lots of cleanups
        atomic modesetting/pageflipping support
        render node support

   - tegra:
        tegra210 support (dc, dsi, dp/hdmi)
        dpms with atomic modesetting support

   - atmel:
        support for 3 more atmel SoCs
        new input formats, PRIME support.

   - dwhdmi:
        preparing to add audio support

   - rockchip:
        yuv plane support"

* 'drm-next' of git://people.freedesktop.org/~airlied/linux: (1369 commits)
  drm/amdgpu: rename gmc_v8_0_init_compute_vmid
  drm/amdgpu: fix vce3 instance handling
  drm/amdgpu: remove ib test for the second VCE Ring
  drm/amdgpu: properly enable VM fault interrupts
  drm/amdgpu: fix warning in scheduler
  drm/amdgpu: fix buffer placement under memory pressure
  drm/amdgpu/cz: fix cz_dpm_update_low_memory_pstate logic
  drm/amdgpu: fix typo in dce11 watermark setup
  drm/amdgpu: fix typo in dce10 watermark setup
  drm/amdgpu: use top down allocation for non-CPU accessible vram
  drm/amdgpu: be explicit about cpu vram access for driver BOs (v2)
  drm/amdgpu: set MEC doorbell range for Fiji
  drm/amdgpu: implement burst NOP for SDMA
  drm/amdgpu: add insert_nop ring func and default implementation
  drm/amdgpu: add amdgpu_get_sdma_instance helper function
  drm/amdgpu: add AMDGPU_MAX_SDMA_INSTANCES
  drm/amdgpu: add burst_nop flag for sdma
  drm/amdgpu: add count field for the SDMA NOP packet v2
  drm/amdgpu: use PT for VM sync on unmap
  drm/amdgpu: make wait_event uninterruptible in push_job
  ...

1 files changed, 365 insertions, 459 deletions

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c
index d6ace41830e9..4542867fa9e6 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c
@@ -21,13 +21,13 @@
  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
  * DEALINGS IN THE SOFTWARE.
  */
-#include <engine/gr.h>
+#include "nv10.h"
 #include "regs.h"
 
 #include <core/client.h>
-#include <core/device.h>
-#include <core/handle.h>
+#include <core/gpuobj.h>
 #include <engine/fifo.h>
+#include <engine/fifo/chan.h>
 #include <subdev/fb.h>
 
 struct pipe_state {
@@ -386,14 +386,19 @@ static int nv17_gr_ctx_regs[] = {
 	0x00400a04,
 };
 
-struct nv10_gr_priv {
+#define nv10_gr(p) container_of((p), struct nv10_gr, base)
+
+struct nv10_gr {
 	struct nvkm_gr base;
 	struct nv10_gr_chan *chan[32];
 	spinlock_t lock;
 };
 
+#define nv10_gr_chan(p) container_of((p), struct nv10_gr_chan, object)
+
 struct nv10_gr_chan {
-	struct nvkm_object base;
+	struct nvkm_object object;
+	struct nv10_gr *gr;
 	int chid;
 	int nv10[ARRAY_SIZE(nv10_gr_ctx_regs)];
 	int nv17[ARRAY_SIZE(nv17_gr_ctx_regs)];
@@ -402,214 +407,151 @@ struct nv10_gr_chan {
 };
 
 
-static inline struct nv10_gr_priv *
-nv10_gr_priv(struct nv10_gr_chan *chan)
-{
-	return (void *)nv_object(chan)->engine;
-}
-
 /*******************************************************************************
  * Graphics object classes
  ******************************************************************************/
 
-#define PIPE_SAVE(priv, state, addr)					\
+#define PIPE_SAVE(gr, state, addr)					\
 	do {								\
 		int __i;						\
-		nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
 		for (__i = 0; __i < ARRAY_SIZE(state); __i++)		\
-			state[__i] = nv_rd32(priv, NV10_PGRAPH_PIPE_DATA); \
+			state[__i] = nvkm_rd32(device, NV10_PGRAPH_PIPE_DATA); \
 	} while (0)
 
-#define PIPE_RESTORE(priv, state, addr)					\
+#define PIPE_RESTORE(gr, state, addr)					\
 	do {								\
 		int __i;						\
-		nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, addr);		\
 		for (__i = 0; __i < ARRAY_SIZE(state); __i++)		\
-			nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, state[__i]); \
+			nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, state[__i]); \
 	} while (0)
 
-static struct nvkm_oclass
-nv10_gr_sclass[] = {
-	{ 0x0012, &nv04_gr_ofuncs }, /* beta1 */
-	{ 0x0019, &nv04_gr_ofuncs }, /* clip */
-	{ 0x0030, &nv04_gr_ofuncs }, /* null */
-	{ 0x0039, &nv04_gr_ofuncs }, /* m2mf */
-	{ 0x0043, &nv04_gr_ofuncs }, /* rop */
-	{ 0x0044, &nv04_gr_ofuncs }, /* pattern */
-	{ 0x004a, &nv04_gr_ofuncs }, /* gdi */
-	{ 0x0052, &nv04_gr_ofuncs }, /* swzsurf */
-	{ 0x005f, &nv04_gr_ofuncs }, /* blit */
-	{ 0x0062, &nv04_gr_ofuncs }, /* surf2d */
-	{ 0x0072, &nv04_gr_ofuncs }, /* beta4 */
-	{ 0x0089, &nv04_gr_ofuncs }, /* sifm */
-	{ 0x008a, &nv04_gr_ofuncs }, /* ifc */
-	{ 0x009f, &nv04_gr_ofuncs }, /* blit */
-	{ 0x0093, &nv04_gr_ofuncs }, /* surf3d */
-	{ 0x0094, &nv04_gr_ofuncs }, /* ttri */
-	{ 0x0095, &nv04_gr_ofuncs }, /* mtri */
-	{ 0x0056, &nv04_gr_ofuncs }, /* celcius */
-	{},
-};
-
-static struct nvkm_oclass
-nv15_gr_sclass[] = {
-	{ 0x0012, &nv04_gr_ofuncs }, /* beta1 */
-	{ 0x0019, &nv04_gr_ofuncs }, /* clip */
-	{ 0x0030, &nv04_gr_ofuncs }, /* null */
-	{ 0x0039, &nv04_gr_ofuncs }, /* m2mf */
-	{ 0x0043, &nv04_gr_ofuncs }, /* rop */
-	{ 0x0044, &nv04_gr_ofuncs }, /* pattern */
-	{ 0x004a, &nv04_gr_ofuncs }, /* gdi */
-	{ 0x0052, &nv04_gr_ofuncs }, /* swzsurf */
-	{ 0x005f, &nv04_gr_ofuncs }, /* blit */
-	{ 0x0062, &nv04_gr_ofuncs }, /* surf2d */
-	{ 0x0072, &nv04_gr_ofuncs }, /* beta4 */
-	{ 0x0089, &nv04_gr_ofuncs }, /* sifm */
-	{ 0x008a, &nv04_gr_ofuncs }, /* ifc */
-	{ 0x009f, &nv04_gr_ofuncs }, /* blit */
-	{ 0x0093, &nv04_gr_ofuncs }, /* surf3d */
-	{ 0x0094, &nv04_gr_ofuncs }, /* ttri */
-	{ 0x0095, &nv04_gr_ofuncs }, /* mtri */
-	{ 0x0096, &nv04_gr_ofuncs }, /* celcius */
-	{},
-};
-
-static int
-nv17_gr_mthd_lma_window(struct nvkm_object *object, u32 mthd,
-			void *args, u32 size)
+static void
+nv17_gr_mthd_lma_window(struct nv10_gr_chan *chan, u32 mthd, u32 data)
 {
-	struct nv10_gr_chan *chan = (void *)object->parent;
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nvkm_device *device = chan->object.engine->subdev.device;
+	struct nvkm_gr *gr = &chan->gr->base;
 	struct pipe_state *pipe = &chan->pipe_state;
 	u32 pipe_0x0040[1], pipe_0x64c0[8], pipe_0x6a80[3], pipe_0x6ab0[3];
 	u32 xfmode0, xfmode1;
-	u32 data = *(u32 *)args;
 	int i;
 
 	chan->lma_window[(mthd - 0x1638) / 4] = data;
 
 	if (mthd != 0x1644)
-		return 0;
+		return;
 
-	nv04_gr_idle(priv);
+	nv04_gr_idle(gr);
 
-	PIPE_SAVE(priv, pipe_0x0040, 0x0040);
-	PIPE_SAVE(priv, pipe->pipe_0x0200, 0x0200);
+	PIPE_SAVE(device, pipe_0x0040, 0x0040);
+	PIPE_SAVE(device, pipe->pipe_0x0200, 0x0200);
 
-	PIPE_RESTORE(priv, chan->lma_window, 0x6790);
+	PIPE_RESTORE(device, chan->lma_window, 0x6790);
 
-	nv04_gr_idle(priv);
+	nv04_gr_idle(gr);
 
-	xfmode0 = nv_rd32(priv, NV10_PGRAPH_XFMODE0);
-	xfmode1 = nv_rd32(priv, NV10_PGRAPH_XFMODE1);
+	xfmode0 = nvkm_rd32(device, NV10_PGRAPH_XFMODE0);
+	xfmode1 = nvkm_rd32(device, NV10_PGRAPH_XFMODE1);
 
-	PIPE_SAVE(priv, pipe->pipe_0x4400, 0x4400);
-	PIPE_SAVE(priv, pipe_0x64c0, 0x64c0);
-	PIPE_SAVE(priv, pipe_0x6ab0, 0x6ab0);
-	PIPE_SAVE(priv, pipe_0x6a80, 0x6a80);
+	PIPE_SAVE(device, pipe->pipe_0x4400, 0x4400);
+	PIPE_SAVE(device, pipe_0x64c0, 0x64c0);
+	PIPE_SAVE(device, pipe_0x6ab0, 0x6ab0);
+	PIPE_SAVE(device, pipe_0x6a80, 0x6a80);
 
-	nv04_gr_idle(priv);
+	nv04_gr_idle(gr);
 
-	nv_wr32(priv, NV10_PGRAPH_XFMODE0, 0x10000000);
-	nv_wr32(priv, NV10_PGRAPH_XFMODE1, 0x00000000);
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE0, 0x10000000);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE1, 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
 	for (i = 0; i < 4; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
 	for (i = 0; i < 4; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000);
 
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
 	for (i = 0; i < 3; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
 
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
 	for (i = 0; i < 3; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
-
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
-	nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000008);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000);
 
-	PIPE_RESTORE(priv, pipe->pipe_0x0200, 0x0200);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000008);
 
-	nv04_gr_idle(priv);
+	PIPE_RESTORE(device, pipe->pipe_0x0200, 0x0200);
 
-	PIPE_RESTORE(priv, pipe_0x0040, 0x0040);
+	nv04_gr_idle(gr);
 
-	nv_wr32(priv, NV10_PGRAPH_XFMODE0, xfmode0);
-	nv_wr32(priv, NV10_PGRAPH_XFMODE1, xfmode1);
+	PIPE_RESTORE(device, pipe_0x0040, 0x0040);
 
-	PIPE_RESTORE(priv, pipe_0x64c0, 0x64c0);
-	PIPE_RESTORE(priv, pipe_0x6ab0, 0x6ab0);
-	PIPE_RESTORE(priv, pipe_0x6a80, 0x6a80);
-	PIPE_RESTORE(priv, pipe->pipe_0x4400, 0x4400);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE0, xfmode0);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE1, xfmode1);
 
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0);
-	nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+	PIPE_RESTORE(device, pipe_0x64c0, 0x64c0);
+	PIPE_RESTORE(device, pipe_0x6ab0, 0x6ab0);
+	PIPE_RESTORE(device, pipe_0x6a80, 0x6a80);
+	PIPE_RESTORE(device, pipe->pipe_0x4400, 0x4400);
 
-	nv04_gr_idle(priv);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000);
 
-	return 0;
+	nv04_gr_idle(gr);
 }
 
-static int
-nv17_gr_mthd_lma_enable(struct nvkm_object *object, u32 mthd,
-			void *args, u32 size)
+static void
+nv17_gr_mthd_lma_enable(struct nv10_gr_chan *chan, u32 mthd, u32 data)
 {
-	struct nv10_gr_chan *chan = (void *)object->parent;
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nvkm_device *device = chan->object.engine->subdev.device;
+	struct nvkm_gr *gr = &chan->gr->base;
 
-	nv04_gr_idle(priv);
+	nv04_gr_idle(gr);
 
-	nv_mask(priv, NV10_PGRAPH_DEBUG_4, 0x00000100, 0x00000100);
-	nv_mask(priv, 0x4006b0, 0x08000000, 0x08000000);
-	return 0;
+	nvkm_mask(device, NV10_PGRAPH_DEBUG_4, 0x00000100, 0x00000100);
+	nvkm_mask(device, 0x4006b0, 0x08000000, 0x08000000);
 }
 
-static struct nvkm_omthds
-nv17_celcius_omthds[] = {
-	{ 0x1638, 0x1638, nv17_gr_mthd_lma_window },
-	{ 0x163c, 0x163c, nv17_gr_mthd_lma_window },
-	{ 0x1640, 0x1640, nv17_gr_mthd_lma_window },
-	{ 0x1644, 0x1644, nv17_gr_mthd_lma_window },
-	{ 0x1658, 0x1658, nv17_gr_mthd_lma_enable },
-	{}
-};
+static bool
+nv17_gr_mthd_celcius(struct nv10_gr_chan *chan, u32 mthd, u32 data)
+{
+	void (*func)(struct nv10_gr_chan *, u32, u32);
+	switch (mthd) {
+	case 0x1638 ... 0x1644:
+		     func = nv17_gr_mthd_lma_window; break;
+	case 0x1658: func = nv17_gr_mthd_lma_enable; break;
+	default:
+		return false;
+	}
+	func(chan, mthd, data);
+	return true;
+}
 
-static struct nvkm_oclass
-nv17_gr_sclass[] = {
-	{ 0x0012, &nv04_gr_ofuncs }, /* beta1 */
-	{ 0x0019, &nv04_gr_ofuncs }, /* clip */
-	{ 0x0030, &nv04_gr_ofuncs }, /* null */
-	{ 0x0039, &nv04_gr_ofuncs }, /* m2mf */
-	{ 0x0043, &nv04_gr_ofuncs }, /* rop */
-	{ 0x0044, &nv04_gr_ofuncs }, /* pattern */
-	{ 0x004a, &nv04_gr_ofuncs }, /* gdi */
-	{ 0x0052, &nv04_gr_ofuncs }, /* swzsurf */
-	{ 0x005f, &nv04_gr_ofuncs }, /* blit */
-	{ 0x0062, &nv04_gr_ofuncs }, /* surf2d */
-	{ 0x0072, &nv04_gr_ofuncs }, /* beta4 */
-	{ 0x0089, &nv04_gr_ofuncs }, /* sifm */
-	{ 0x008a, &nv04_gr_ofuncs }, /* ifc */
-	{ 0x009f, &nv04_gr_ofuncs }, /* blit */
-	{ 0x0093, &nv04_gr_ofuncs }, /* surf3d */
-	{ 0x0094, &nv04_gr_ofuncs }, /* ttri */
-	{ 0x0095, &nv04_gr_ofuncs }, /* mtri */
-	{ 0x0099, &nv04_gr_ofuncs, nv17_celcius_omthds },
-	{},
-};
+static bool
+nv10_gr_mthd(struct nv10_gr_chan *chan, u8 class, u32 mthd, u32 data)
+{
+	bool (*func)(struct nv10_gr_chan *, u32, u32);
+	switch (class) {
+	case 0x99: func = nv17_gr_mthd_celcius; break;
+	default:
+		return false;
+	}
+	return func(chan, mthd, data);
+}
 
 /*******************************************************************************
  * PGRAPH context
  ******************************************************************************/
 
 static struct nv10_gr_chan *
-nv10_gr_channel(struct nv10_gr_priv *priv)
+nv10_gr_channel(struct nv10_gr *gr)
 {
+	struct nvkm_device *device = gr->base.engine.subdev.device;
 	struct nv10_gr_chan *chan = NULL;
-	if (nv_rd32(priv, 0x400144) & 0x00010000) {
-		int chid = nv_rd32(priv, 0x400148) >> 24;
-		if (chid < ARRAY_SIZE(priv->chan))
-			chan = priv->chan[chid];
+	if (nvkm_rd32(device, 0x400144) & 0x00010000) {
+		int chid = nvkm_rd32(device, 0x400148) >> 24;
+		if (chid < ARRAY_SIZE(gr->chan))
+			chan = gr->chan[chid];
 	}
 	return chan;
 }
@@ -617,75 +559,78 @@ nv10_gr_channel(struct nv10_gr_priv *priv)
 static void
 nv10_gr_save_pipe(struct nv10_gr_chan *chan)
 {
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nv10_gr *gr = chan->gr;
 	struct pipe_state *pipe = &chan->pipe_state;
-
-	PIPE_SAVE(priv, pipe->pipe_0x4400, 0x4400);
-	PIPE_SAVE(priv, pipe->pipe_0x0200, 0x0200);
-	PIPE_SAVE(priv, pipe->pipe_0x6400, 0x6400);
-	PIPE_SAVE(priv, pipe->pipe_0x6800, 0x6800);
-	PIPE_SAVE(priv, pipe->pipe_0x6c00, 0x6c00);
-	PIPE_SAVE(priv, pipe->pipe_0x7000, 0x7000);
-	PIPE_SAVE(priv, pipe->pipe_0x7400, 0x7400);
-	PIPE_SAVE(priv, pipe->pipe_0x7800, 0x7800);
-	PIPE_SAVE(priv, pipe->pipe_0x0040, 0x0040);
-	PIPE_SAVE(priv, pipe->pipe_0x0000, 0x0000);
+	struct nvkm_device *device = gr->base.engine.subdev.device;
+
+	PIPE_SAVE(gr, pipe->pipe_0x4400, 0x4400);
+	PIPE_SAVE(gr, pipe->pipe_0x0200, 0x0200);
+	PIPE_SAVE(gr, pipe->pipe_0x6400, 0x6400);
+	PIPE_SAVE(gr, pipe->pipe_0x6800, 0x6800);
+	PIPE_SAVE(gr, pipe->pipe_0x6c00, 0x6c00);
+	PIPE_SAVE(gr, pipe->pipe_0x7000, 0x7000);
+	PIPE_SAVE(gr, pipe->pipe_0x7400, 0x7400);
+	PIPE_SAVE(gr, pipe->pipe_0x7800, 0x7800);
+	PIPE_SAVE(gr, pipe->pipe_0x0040, 0x0040);
+	PIPE_SAVE(gr, pipe->pipe_0x0000, 0x0000);
 }
 
 static void
 nv10_gr_load_pipe(struct nv10_gr_chan *chan)
 {
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nv10_gr *gr = chan->gr;
 	struct pipe_state *pipe = &chan->pipe_state;
+	struct nvkm_device *device = gr->base.engine.subdev.device;
 	u32 xfmode0, xfmode1;
 	int i;
 
-	nv04_gr_idle(priv);
+	nv04_gr_idle(&gr->base);
 	/* XXX check haiku comments */
-	xfmode0 = nv_rd32(priv, NV10_PGRAPH_XFMODE0);
-	xfmode1 = nv_rd32(priv, NV10_PGRAPH_XFMODE1);
-	nv_wr32(priv, NV10_PGRAPH_XFMODE0, 0x10000000);
-	nv_wr32(priv, NV10_PGRAPH_XFMODE1, 0x00000000);
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
+	xfmode0 = nvkm_rd32(device, NV10_PGRAPH_XFMODE0);
+	xfmode1 = nvkm_rd32(device, NV10_PGRAPH_XFMODE1);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE0, 0x10000000);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE1, 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0);
 	for (i = 0; i < 4; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
 	for (i = 0; i < 4; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000);
 
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0);
 	for (i = 0; i < 3; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000);
 
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80);
 	for (i = 0; i < 3; i++)
-		nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000);
+		nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000);
 
-	nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
-	nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000008);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040);
+	nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000008);
 
 
-	PIPE_RESTORE(priv, pipe->pipe_0x0200, 0x0200);
-	nv04_gr_idle(priv);
+	PIPE_RESTORE(gr, pipe->pipe_0x0200, 0x0200);
+	nv04_gr_idle(&gr->base);
 
 	/* restore XFMODE */
-	nv_wr32(priv, NV10_PGRAPH_XFMODE0, xfmode0);
-	nv_wr32(priv, NV10_PGRAPH_XFMODE1, xfmode1);
-	PIPE_RESTORE(priv, pipe->pipe_0x6400, 0x6400);
-	PIPE_RESTORE(priv, pipe->pipe_0x6800, 0x6800);
-	PIPE_RESTORE(priv, pipe->pipe_0x6c00, 0x6c00);
-	PIPE_RESTORE(priv, pipe->pipe_0x7000, 0x7000);
-	PIPE_RESTORE(priv, pipe->pipe_0x7400, 0x7400);
-	PIPE_RESTORE(priv, pipe->pipe_0x7800, 0x7800);
-	PIPE_RESTORE(priv, pipe->pipe_0x4400, 0x4400);
-	PIPE_RESTORE(priv, pipe->pipe_0x0000, 0x0000);
-	PIPE_RESTORE(priv, pipe->pipe_0x0040, 0x0040);
-	nv04_gr_idle(priv);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE0, xfmode0);
+	nvkm_wr32(device, NV10_PGRAPH_XFMODE1, xfmode1);
+	PIPE_RESTORE(gr, pipe->pipe_0x6400, 0x6400);
+	PIPE_RESTORE(gr, pipe->pipe_0x6800, 0x6800);
+	PIPE_RESTORE(gr, pipe->pipe_0x6c00, 0x6c00);
+	PIPE_RESTORE(gr, pipe->pipe_0x7000, 0x7000);
+	PIPE_RESTORE(gr, pipe->pipe_0x7400, 0x7400);
+	PIPE_RESTORE(gr, pipe->pipe_0x7800, 0x7800);
+	PIPE_RESTORE(gr, pipe->pipe_0x4400, 0x4400);
+	PIPE_RESTORE(gr, pipe->pipe_0x0000, 0x0000);
+	PIPE_RESTORE(gr, pipe->pipe_0x0040, 0x0040);
+	nv04_gr_idle(&gr->base);
 }
 
 static void
 nv10_gr_create_pipe(struct nv10_gr_chan *chan)
 {
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nv10_gr *gr = chan->gr;
+	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
 	struct pipe_state *pipe_state = &chan->pipe_state;
 	u32 *pipe_state_addr;
 	int i;
@@ -698,7 +643,7 @@ nv10_gr_create_pipe(struct nv10_gr_chan *chan)
 		u32 *__end_addr = pipe_state->pipe_##addr + \
 				ARRAY_SIZE(pipe_state->pipe_##addr); \
 		if (pipe_state_addr != __end_addr) \
-			nv_error(priv, "incomplete pipe init for 0x%x :  %p/%p\n", \
+			nvkm_error(subdev, "incomplete pipe init for 0x%x :  %p/%p\n", \
 				addr, pipe_state_addr, __end_addr); \
 	} while (0)
 #define NV_WRITE_PIPE_INIT(value) *(pipe_state_addr++) = value
@@ -838,33 +783,36 @@ nv10_gr_create_pipe(struct nv10_gr_chan *chan)
 }
 
 static int
-nv10_gr_ctx_regs_find_offset(struct nv10_gr_priv *priv, int reg)
+nv10_gr_ctx_regs_find_offset(struct nv10_gr *gr, int reg)
 {
+	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
 	int i;
 	for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++) {
 		if (nv10_gr_ctx_regs[i] == reg)
 			return i;
 	}
-	nv_error(priv, "unknown offset nv10_ctx_regs %d\n", reg);
+	nvkm_error(subdev, "unknown offset nv10_ctx_regs %d\n", reg);
 	return -1;
 }
 
 static int
-nv17_gr_ctx_regs_find_offset(struct nv10_gr_priv *priv, int reg)
+nv17_gr_ctx_regs_find_offset(struct nv10_gr *gr, int reg)
 {
+	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
 	int i;
 	for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++) {
 		if (nv17_gr_ctx_regs[i] == reg)
 			return i;
 	}
-	nv_error(priv, "unknown offset nv17_ctx_regs %d\n", reg);
+	nvkm_error(subdev, "unknown offset nv17_ctx_regs %d\n", reg);
 	return -1;
 }
 
 static void
 nv10_gr_load_dma_vtxbuf(struct nv10_gr_chan *chan, int chid, u32 inst)
 {
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nv10_gr *gr = chan->gr;
+	struct nvkm_device *device = gr->base.engine.subdev.device;
 	u32 st2, st2_dl, st2_dh, fifo_ptr, fifo[0x60/4];
 	u32 ctx_user, ctx_switch[5];
 	int i, subchan = -1;
@@ -876,7 +824,7 @@ nv10_gr_load_dma_vtxbuf(struct nv10_gr_chan *chan, int chid, u32 inst)
 
 	/* Look for a celsius object */
 	for (i = 0; i < 8; i++) {
-		int class = nv_rd32(priv, NV10_PGRAPH_CTX_CACHE(i, 0)) & 0xfff;
+		int class = nvkm_rd32(device, NV10_PGRAPH_CTX_CACHE(i, 0)) & 0xfff;
 
 		if (class == 0x56 || class == 0x96 || class == 0x99) {
 			subchan = i;
@@ -888,159 +836,183 @@ nv10_gr_load_dma_vtxbuf(struct nv10_gr_chan *chan, int chid, u32 inst)
 		return;
 
 	/* Save the current ctx object */
-	ctx_user = nv_rd32(priv, NV10_PGRAPH_CTX_USER);
+	ctx_user = nvkm_rd32(device, NV10_PGRAPH_CTX_USER);
 	for (i = 0; i < 5; i++)
-		ctx_switch[i] = nv_rd32(priv, NV10_PGRAPH_CTX_SWITCH(i));
+		ctx_switch[i] = nvkm_rd32(device, NV10_PGRAPH_CTX_SWITCH(i));
 
 	/* Save the FIFO state */
-	st2 = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2);
-	st2_dl = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2_DL);
-	st2_dh = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2_DH);
-	fifo_ptr = nv_rd32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR);
+	st2 = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2);
+	st2_dl = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2_DL);
+	st2_dh = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2_DH);
+	fifo_ptr = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR);
 
 	for (i = 0; i < ARRAY_SIZE(fifo); i++)
-		fifo[i] = nv_rd32(priv, 0x4007a0 + 4 * i);
+		fifo[i] = nvkm_rd32(device, 0x4007a0 + 4 * i);
 
 	/* Switch to the celsius subchannel */
 	for (i = 0; i < 5; i++)
-		nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(i),
-			nv_rd32(priv, NV10_PGRAPH_CTX_CACHE(subchan, i)));
-	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xe000, subchan << 13);
+		nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(i),
+			nvkm_rd32(device, NV10_PGRAPH_CTX_CACHE(subchan, i)));
+	nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xe000, subchan << 13);
 
 	/* Inject NV10TCL_DMA_VTXBUF */
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR, 0);
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2,
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR, 0);
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2,
 		0x2c000000 | chid << 20 | subchan << 16 | 0x18c);
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DL, inst);
-	nv_mask(priv, NV10_PGRAPH_CTX_CONTROL, 0, 0x10000);
-	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001);
-	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DL, inst);
+	nvkm_mask(device, NV10_PGRAPH_CTX_CONTROL, 0, 0x10000);
+	nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001);
+	nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000);
 
 	/* Restore the FIFO state */
 	for (i = 0; i < ARRAY_SIZE(fifo); i++)
-		nv_wr32(priv, 0x4007a0 + 4 * i, fifo[i]);
+		nvkm_wr32(device, 0x4007a0 + 4 * i, fifo[i]);
 
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR, fifo_ptr);
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2, st2);
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DL, st2_dl);
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DH, st2_dh);
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR, fifo_ptr);
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, st2);
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DL, st2_dl);
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DH, st2_dh);
 
 	/* Restore the current ctx object */
 	for (i = 0; i < 5; i++)
-		nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(i), ctx_switch[i]);
-	nv_wr32(priv, NV10_PGRAPH_CTX_USER, ctx_user);
+		nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(i), ctx_switch[i]);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_USER, ctx_user);
 }
 
 static int
 nv10_gr_load_context(struct nv10_gr_chan *chan, int chid)
 {
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nv10_gr *gr = chan->gr;
+	struct nvkm_device *device = gr->base.engine.subdev.device;
 	u32 inst;
 	int i;
 
 	for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++)
-		nv_wr32(priv, nv10_gr_ctx_regs[i], chan->nv10[i]);
+		nvkm_wr32(device, nv10_gr_ctx_regs[i], chan->nv10[i]);
 
-	if (nv_device(priv)->card_type >= NV_11 &&
-	    nv_device(priv)->chipset >= 0x17) {
+	if (device->card_type >= NV_11 && device->chipset >= 0x17) {
 		for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++)
-			nv_wr32(priv, nv17_gr_ctx_regs[i], chan->nv17[i]);
+			nvkm_wr32(device, nv17_gr_ctx_regs[i], chan->nv17[i]);
 	}
 
 	nv10_gr_load_pipe(chan);
 
-	inst = nv_rd32(priv, NV10_PGRAPH_GLOBALSTATE1) & 0xffff;
+	inst = nvkm_rd32(device, NV10_PGRAPH_GLOBALSTATE1) & 0xffff;
 	nv10_gr_load_dma_vtxbuf(chan, chid, inst);
 
-	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
-	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, chid << 24);
-	nv_mask(priv, NV10_PGRAPH_FFINTFC_ST2, 0x30000000, 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
+	nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, chid << 24);
+	nvkm_mask(device, NV10_PGRAPH_FFINTFC_ST2, 0x30000000, 0x00000000);
 	return 0;
 }
 
 static int
 nv10_gr_unload_context(struct nv10_gr_chan *chan)
 {
-	struct nv10_gr_priv *priv = nv10_gr_priv(chan);
+	struct nv10_gr *gr = chan->gr;
+	struct nvkm_device *device = gr->base.engine.subdev.device;
 	int i;
 
 	for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++)
-		chan->nv10[i] = nv_rd32(priv, nv10_gr_ctx_regs[i]);
+		chan->nv10[i] = nvkm_rd32(device, nv10_gr_ctx_regs[i]);
 
-	if (nv_device(priv)->card_type >= NV_11 &&
-	    nv_device(priv)->chipset >= 0x17) {
+	if (device->card_type >= NV_11 && device->chipset >= 0x17) {
 		for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++)
-			chan->nv17[i] = nv_rd32(priv, nv17_gr_ctx_regs[i]);
+			chan->nv17[i] = nvkm_rd32(device, nv17_gr_ctx_regs[i]);
 	}
 
 	nv10_gr_save_pipe(chan);
 
-	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000000);
-	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000000);
+	nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000);
 	return 0;
 }
 
 static void
-nv10_gr_context_switch(struct nv10_gr_priv *priv)
+nv10_gr_context_switch(struct nv10_gr *gr)
 {
+	struct nvkm_device *device = gr->base.engine.subdev.device;
 	struct nv10_gr_chan *prev = NULL;
 	struct nv10_gr_chan *next = NULL;
-	unsigned long flags;
 	int chid;
 
-	spin_lock_irqsave(&priv->lock, flags);
-	nv04_gr_idle(priv);
+	nv04_gr_idle(&gr->base);
 
 	/* If previous context is valid, we need to save it */
-	prev = nv10_gr_channel(priv);
+	prev = nv10_gr_channel(gr);
 	if (prev)
 		nv10_gr_unload_context(prev);
 
 	/* load context for next channel */
-	chid = (nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f;
-	next = priv->chan[chid];
+	chid = (nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f;
+	next = gr->chan[chid];
 	if (next)
 		nv10_gr_load_context(next, chid);
+}
+
+static int
+nv10_gr_chan_fini(struct nvkm_object *object, bool suspend)
+{
+	struct nv10_gr_chan *chan = nv10_gr_chan(object);
+	struct nv10_gr *gr = chan->gr;
+	struct nvkm_device *device = gr->base.engine.subdev.device;
+	unsigned long flags;
 
-	spin_unlock_irqrestore(&priv->lock, flags);
+	spin_lock_irqsave(&gr->lock, flags);
+	nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000);
+	if (nv10_gr_channel(gr) == chan)
+		nv10_gr_unload_context(chan);
+	nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001);
+	spin_unlock_irqrestore(&gr->lock, flags);
+	return 0;
+}
+
+static void *
+nv10_gr_chan_dtor(struct nvkm_object *object)
+{
+	struct nv10_gr_chan *chan = nv10_gr_chan(object);
+	struct nv10_gr *gr = chan->gr;
+	unsigned long flags;
+
+	spin_lock_irqsave(&gr->lock, flags);
+	gr->chan[chan->chid] = NULL;
+	spin_unlock_irqrestore(&gr->lock, flags);
+	return chan;
 }
 
+static const struct nvkm_object_func
+nv10_gr_chan = {
+	.dtor = nv10_gr_chan_dtor,
+	.fini = nv10_gr_chan_fini,
+};
+
 #define NV_WRITE_CTX(reg, val) do { \
-	int offset = nv10_gr_ctx_regs_find_offset(priv, reg); \
+	int offset = nv10_gr_ctx_regs_find_offset(gr, reg); \
 	if (offset > 0) \
 		chan->nv10[offset] = val; \
 	} while (0)
 
 #define NV17_WRITE_CTX(reg, val) do { \
-	int offset = nv17_gr_ctx_regs_find_offset(priv, reg); \
+	int offset = nv17_gr_ctx_regs_find_offset(gr, reg); \
 	if (offset > 0) \
 		chan->nv17[offset] = val; \
 	} while (0)
 
-static int
-nv10_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine,
-		     struct nvkm_oclass *oclass, void *data, u32 size,
-		     struct nvkm_object **pobject)
+int
+nv10_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch,
+		 const struct nvkm_oclass *oclass, struct nvkm_object **pobject)
 {
-	struct nvkm_fifo_chan *fifo = (void *)parent;
-	struct nv10_gr_priv *priv = (void *)engine;
+	struct nv10_gr *gr = nv10_gr(base);
 	struct nv10_gr_chan *chan;
+	struct nvkm_device *device = gr->base.engine.subdev.device;
 	unsigned long flags;
-	int ret;
-
-	ret = nvkm_object_create(parent, engine, oclass, 0, &chan);
-	*pobject = nv_object(chan);
-	if (ret)
-		return ret;
-
-	spin_lock_irqsave(&priv->lock, flags);
-	if (priv->chan[fifo->chid]) {
-		*pobject = nv_object(priv->chan[fifo->chid]);
-		atomic_inc(&(*pobject)->refcount);
-		spin_unlock_irqrestore(&priv->lock, flags);
-		nvkm_object_destroy(&chan->base);
-		return 1;
-	}
+
+	if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL)))
+		return -ENOMEM;
+	nvkm_object_ctor(&nv10_gr_chan, oclass, &chan->object);
+	chan->gr = gr;
+	chan->chid = fifoch->chid;
+	*pobject = &chan->object;
 
 	NV_WRITE_CTX(0x00400e88, 0x08000000);
 	NV_WRITE_CTX(0x00400e9c, 0x4b7fffff);
@@ -1049,12 +1021,11 @@ nv10_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine,
 	NV_WRITE_CTX(0x00400e14, 0x00001000);
 	NV_WRITE_CTX(0x00400e30, 0x00080008);
 	NV_WRITE_CTX(0x00400e34, 0x00080008);
-	if (nv_device(priv)->card_type >= NV_11 &&
-	    nv_device(priv)->chipset >= 0x17) {
+	if (device->card_type >= NV_11 && device->chipset >= 0x17) {
 		/* is it really needed ??? */
 		NV17_WRITE_CTX(NV10_PGRAPH_DEBUG_4,
-					nv_rd32(priv, NV10_PGRAPH_DEBUG_4));
-		NV17_WRITE_CTX(0x004006b0, nv_rd32(priv, 0x004006b0));
+			       nvkm_rd32(device, NV10_PGRAPH_DEBUG_4));
+		NV17_WRITE_CTX(0x004006b0, nvkm_rd32(device, 0x004006b0));
 		NV17_WRITE_CTX(0x00400eac, 0x0fff0000);
 		NV17_WRITE_CTX(0x00400eb0, 0x0fff0000);
 		NV17_WRITE_CTX(0x00400ec0, 0x00000080);
@@ -1064,74 +1035,32 @@ nv10_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine,
 
 	nv10_gr_create_pipe(chan);
 
-	priv->chan[fifo->chid] = chan;
-	chan->chid = fifo->chid;
-	spin_unlock_irqrestore(&priv->lock, flags);
+	spin_lock_irqsave(&gr->lock, flags);
+	gr->chan[chan->chid] = chan;
+	spin_unlock_irqrestore(&gr->lock, flags);
 	return 0;
 }
 
-static void
-nv10_gr_context_dtor(struct nvkm_object *object)
-{
-	struct nv10_gr_priv *priv = (void *)object->engine;
-	struct nv10_gr_chan *chan = (void *)object;
-	unsigned long flags;
-
-	spin_lock_irqsave(&priv->lock, flags);
-	priv->chan[chan->chid] = NULL;
-	spin_unlock_irqrestore(&priv->lock, flags);
-
-	nvkm_object_destroy(&chan->base);
-}
-
-static int
-nv10_gr_context_fini(struct nvkm_object *object, bool suspend)
-{
-	struct nv10_gr_priv *priv = (void *)object->engine;
-	struct nv10_gr_chan *chan = (void *)object;
-	unsigned long flags;
-
-	spin_lock_irqsave(&priv->lock, flags);
-	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000);
-	if (nv10_gr_channel(priv) == chan)
-		nv10_gr_unload_context(chan);
-	nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001);
-	spin_unlock_irqrestore(&priv->lock, flags);
-
-	return nvkm_object_fini(&chan->base, suspend);
-}
-
-static struct nvkm_oclass
-nv10_gr_cclass = {
-	.handle = NV_ENGCTX(GR, 0x10),
-	.ofuncs = &(struct nvkm_ofuncs) {
-		.ctor = nv10_gr_context_ctor,
-		.dtor = nv10_gr_context_dtor,
-		.init = nvkm_object_init,
-		.fini = nv10_gr_context_fini,
-	},
-};
-
 /*******************************************************************************
  * PGRAPH engine/subdev functions
  ******************************************************************************/
 
-static void
-nv10_gr_tile_prog(struct nvkm_engine *engine, int i)
+void
+nv10_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile)
 {
-	struct nvkm_fb_tile *tile = &nvkm_fb(engine)->tile.region[i];
-	struct nvkm_fifo *pfifo = nvkm_fifo(engine);
-	struct nv10_gr_priv *priv = (void *)engine;
+	struct nv10_gr *gr = nv10_gr(base);
+	struct nvkm_device *device = gr->base.engine.subdev.device;
+	struct nvkm_fifo *fifo = device->fifo;
 	unsigned long flags;
 
-	pfifo->pause(pfifo, &flags);
-	nv04_gr_idle(priv);
+	nvkm_fifo_pause(fifo, &flags);
+	nv04_gr_idle(&gr->base);
 
-	nv_wr32(priv, NV10_PGRAPH_TLIMIT(i), tile->limit);
-	nv_wr32(priv, NV10_PGRAPH_TSIZE(i), tile->pitch);
-	nv_wr32(priv, NV10_PGRAPH_TILE(i), tile->addr);
+	nvkm_wr32(device, NV10_PGRAPH_TLIMIT(i), tile->limit);
+	nvkm_wr32(device, NV10_PGRAPH_TSIZE(i), tile->pitch);
+	nvkm_wr32(device, NV10_PGRAPH_TILE(i), tile->addr);
 
-	pfifo->start(pfifo, &flags);
+	nvkm_fifo_start(fifo, &flags);
 }
 
 const struct nvkm_bitfield nv10_gr_intr_name[] = {
@@ -1148,168 +1077,145 @@ const struct nvkm_bitfield nv10_gr_nstatus[] = {
 	{}
 };
 
-static void
-nv10_gr_intr(struct nvkm_subdev *subdev)
+void
+nv10_gr_intr(struct nvkm_gr *base)
 {
-	struct nv10_gr_priv *priv = (void *)subdev;
-	struct nv10_gr_chan *chan = NULL;
-	struct nvkm_namedb *namedb = NULL;
-	struct nvkm_handle *handle = NULL;
-	u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR);
-	u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE);
-	u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS);
-	u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR);
+	struct nv10_gr *gr = nv10_gr(base);
+	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
+	struct nvkm_device *device = subdev->device;
+	u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR);
+	u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE);
+	u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS);
+	u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR);
 	u32 chid = (addr & 0x01f00000) >> 20;
 	u32 subc = (addr & 0x00070000) >> 16;
 	u32 mthd = (addr & 0x00001ffc);
-	u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA);
-	u32 class = nv_rd32(priv, 0x400160 + subc * 4) & 0xfff;
+	u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA);
+	u32 class = nvkm_rd32(device, 0x400160 + subc * 4) & 0xfff;
 	u32 show = stat;
+	char msg[128], src[128], sta[128];
+	struct nv10_gr_chan *chan;
 	unsigned long flags;
 
-	spin_lock_irqsave(&priv->lock, flags);
-	chan = priv->chan[chid];
-	if (chan)
-		namedb = (void *)nv_pclass(nv_object(chan), NV_NAMEDB_CLASS);
-	spin_unlock_irqrestore(&priv->lock, flags);
+	spin_lock_irqsave(&gr->lock, flags);
+	chan = gr->chan[chid];
 
 	if (stat & NV_PGRAPH_INTR_ERROR) {
 		if (chan && (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD)) {
-			handle = nvkm_namedb_get_class(namedb, class);
-			if (handle && !nv_call(handle->object, mthd, data))
+			if (!nv10_gr_mthd(chan, class, mthd, data))
 				show &= ~NV_PGRAPH_INTR_ERROR;
 		}
 	}
 
 	if (stat & NV_PGRAPH_INTR_CONTEXT_SWITCH) {
-		nv_wr32(priv, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH);
+		nvkm_wr32(device, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH);
 		stat &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
 		show &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH;
-		nv10_gr_context_switch(priv);
+		nv10_gr_context_switch(gr);
 	}
 
-	nv_wr32(priv, NV03_PGRAPH_INTR, stat);
-	nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001);
+	nvkm_wr32(device, NV03_PGRAPH_INTR, stat);
+	nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001);
 
 	if (show) {
-		nv_error(priv, "%s", "");
-		nvkm_bitfield_print(nv10_gr_intr_name, show);
-		pr_cont(" nsource:");
-		nvkm_bitfield_print(nv04_gr_nsource, nsource);
-		pr_cont(" nstatus:");
-		nvkm_bitfield_print(nv10_gr_nstatus, nstatus);
-		pr_cont("\n");
-		nv_error(priv,
-			 "ch %d [%s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n",
-			 chid, nvkm_client_name(chan), subc, class, mthd,
-			 data);
+		nvkm_snprintbf(msg, sizeof(msg), nv10_gr_intr_name, show);
+		nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource);
+		nvkm_snprintbf(sta, sizeof(sta), nv10_gr_nstatus, nstatus);
+		nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] "
+				   "nstatus %08x [%s] ch %d [%s] subc %d "
+				   "class %04x mthd %04x data %08x\n",
+			   show, msg, nsource, src, nstatus, sta, chid,
+			   chan ? chan->object.client->name : "unknown",
+			   subc, class, mthd, data);
 	}
 
-	nvkm_namedb_put(handle);
+	spin_unlock_irqrestore(&gr->lock, flags);
 }
 
-static int
-nv10_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine,
-	     struct nvkm_oclass *oclass, void *data, u32 size,
-	     struct nvkm_object **pobject)
+int
+nv10_gr_init(struct nvkm_gr *base)
 {
-	struct nv10_gr_priv *priv;
-	int ret;
-
-	ret = nvkm_gr_create(parent, engine, oclass, true, &priv);
-	*pobject = nv_object(priv);
-	if (ret)
-		return ret;
-
-	nv_subdev(priv)->unit = 0x00001000;
-	nv_subdev(priv)->intr = nv10_gr_intr;
-	nv_engine(priv)->cclass = &nv10_gr_cclass;
-
-	if (nv_device(priv)->chipset <= 0x10)
-		nv_engine(priv)->sclass = nv10_gr_sclass;
-	else
-	if (nv_device(priv)->chipset <  0x17 ||
-	    nv_device(priv)->card_type < NV_11)
-		nv_engine(priv)->sclass = nv15_gr_sclass;
-	else
-		nv_engine(priv)->sclass = nv17_gr_sclass;
-
-	nv_engine(priv)->tile_prog = nv10_gr_tile_prog;
-	spin_lock_init(&priv->lock);
-	return 0;
-}
-
-static void
-nv10_gr_dtor(struct nvkm_object *object)
-{
-	struct nv10_gr_priv *priv = (void *)object;
-	nvkm_gr_destroy(&priv->base);
-}
-
-static int
-nv10_gr_init(struct nvkm_object *object)
-{
-	struct nvkm_engine *engine = nv_engine(object);
-	struct nvkm_fb *pfb = nvkm_fb(object);
-	struct nv10_gr_priv *priv = (void *)engine;
-	int ret, i;
-
-	ret = nvkm_gr_init(&priv->base);
-	if (ret)
-		return ret;
-
-	nv_wr32(priv, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
-	nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
-
-	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
-	nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000);
-	nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x00118700);
-	/* nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x24E00810); */ /* 0x25f92ad9 */
-	nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x25f92ad9);
-	nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0x55DE0830 | (1 << 29) | (1 << 31));
-
-	if (nv_device(priv)->card_type >= NV_11 &&
-	    nv_device(priv)->chipset >= 0x17) {
-		nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x1f000000);
-		nv_wr32(priv, 0x400a10, 0x03ff3fb6);
-		nv_wr32(priv, 0x400838, 0x002f8684);
-		nv_wr32(priv, 0x40083c, 0x00115f3f);
-		nv_wr32(priv, 0x4006b0, 0x40000020);
+	struct nv10_gr *gr = nv10_gr(base);
+	struct nvkm_device *device = gr->base.engine.subdev.device;
+
+	nvkm_wr32(device, NV03_PGRAPH_INTR   , 0xFFFFFFFF);
+	nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
+
+	nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
+	nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000);
+	nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x00118700);
+	/* nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x24E00810); */ /* 0x25f92ad9 */
+	nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x25f92ad9);
+	nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0x55DE0830 | (1 << 29) | (1 << 31));
+
+	if (device->card_type >= NV_11 && device->chipset >= 0x17) {
+		nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x1f000000);
+		nvkm_wr32(device, 0x400a10, 0x03ff3fb6);
+		nvkm_wr32(device, 0x400838, 0x002f8684);
+		nvkm_wr32(device, 0x40083c, 0x00115f3f);
+		nvkm_wr32(device, 0x4006b0, 0x40000020);
 	} else {
-		nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00000000);
+		nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00000000);
 	}
 
-	/* Turn all the tiling regions off. */
-	for (i = 0; i < pfb->tile.regions; i++)
-		engine->tile_prog(engine, i);
-
-	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(0), 0x00000000);
-	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(1), 0x00000000);
-	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(2), 0x00000000);
-	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(3), 0x00000000);
-	nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(4), 0x00000000);
-	nv_wr32(priv, NV10_PGRAPH_STATE, 0xFFFFFFFF);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(0), 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(1), 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(2), 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(3), 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(4), 0x00000000);
+	nvkm_wr32(device, NV10_PGRAPH_STATE, 0xFFFFFFFF);
 
-	nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000);
-	nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
-	nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2, 0x08000000);
+	nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000);
+	nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
+	nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, 0x08000000);
 	return 0;
 }
 
-static int
-nv10_gr_fini(struct nvkm_object *object, bool suspend)
+int
+nv10_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device,
+	     int index, struct nvkm_gr **pgr)
 {
-	struct nv10_gr_priv *priv = (void *)object;
-	return nvkm_gr_fini(&priv->base, suspend);
+	struct nv10_gr *gr;
+
+	if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL)))
+		return -ENOMEM;
+	spin_lock_init(&gr->lock);
+	*pgr = &gr->base;
+
+	return nvkm_gr_ctor(func, device, index, 0x00001000, true, &gr->base);
 }
 
-struct nvkm_oclass
-nv10_gr_oclass = {
-	.handle = NV_ENGINE(GR, 0x10),
-	.ofuncs = &(struct nvkm_ofuncs) {
-		.ctor = nv10_gr_ctor,
-		.dtor = nv10_gr_dtor,
-		.init = nv10_gr_init,
-		.fini = nv10_gr_fini,
-	},
+static const struct nvkm_gr_func
+nv10_gr = {
+	.init = nv10_gr_init,
+	.intr = nv10_gr_intr,
+	.tile = nv10_gr_tile,
+	.chan_new = nv10_gr_chan_new,
+	.sclass = {
+		{ -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */
+		{ -1, -1, 0x0019, &nv04_gr_object }, /* clip */
+		{ -1, -1, 0x0030, &nv04_gr_object }, /* null */
+		{ -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */
+		{ -1, -1, 0x0043, &nv04_gr_object }, /* rop */
+		{ -1, -1, 0x0044, &nv04_gr_object }, /* pattern */
+		{ -1, -1, 0x004a, &nv04_gr_object }, /* gdi */
+		{ -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */
+		{ -1, -1, 0x005f, &nv04_gr_object }, /* blit */
+		{ -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */
+		{ -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */
+		{ -1, -1, 0x0089, &nv04_gr_object }, /* sifm */
+		{ -1, -1, 0x008a, &nv04_gr_object }, /* ifc */
+		{ -1, -1, 0x009f, &nv04_gr_object }, /* blit */
+		{ -1, -1, 0x0093, &nv04_gr_object }, /* surf3d */
+		{ -1, -1, 0x0094, &nv04_gr_object }, /* ttri */
+		{ -1, -1, 0x0095, &nv04_gr_object }, /* mtri */
+		{ -1, -1, 0x0056, &nv04_gr_object }, /* celcius */
+		{}
+	}
 };
+
+int
+nv10_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr)
+{
+	return nv10_gr_new_(&nv10_gr, device, index, pgr);
+}