Merge tag 'drm-for-v4.15' of git://people.freedesktop.org/~airlied/linux

Pull drm updates from Dave Airlie:
 "This is the main drm pull request for v4.15.

  Core:
   - Atomic object lifetime fixes
   - Atomic iterator improvements
   - Sparse/smatch fixes
   - Legacy kms ioctls to be interruptible
   - EDID override improvements
   - fb/gem helper cleanups
   - Simple outreachy patches
   - Documentation improvements
   - Fix dma-buf rcu races
   - DRM mode object leasing for improving VR use cases.
   - vgaarb improvements for non-x86 platforms.

  New driver:
   - tve200: Faraday Technology TVE200 block.

     This "TV Encoder" encodes a ITU-T BT.656 stream and can be found in
     the StorLink SL3516 (later Cortina Systems CS3516) as well as the
     Grain Media GM8180.

  New bridges:
   - SiI9234 support

  New panels:
   - S6E63J0X03, OTM8009A, Seiko 43WVF1G, 7" rpi touch panel, Toshiba
     LT089AC19000, Innolux AT043TN24

  i915:
   - Remove Coffeelake from alpha support
   - Cannonlake workarounds
   - Infoframe refactoring for DisplayPort
   - VBT updates
   - DisplayPort vswing/emph/buffer translation refactoring
   - CCS fixes
   - Restore GPU clock boost on missed vblanks
   - Scatter list updates for userptr allocations
   - Gen9+ transition watermarks
   - Display IPC (Isochronous Priority Control)
   - Private PAT management
   - GVT: improved error handling and pci config sanitizing
   - Execlist refactoring
   - Transparent Huge Page support
   - User defined priorities support
   - HuC/GuC firmware refactoring
   - DP MST fixes
   - eDP power sequencing fixes
   - Use RCU instead of stop_machine
   - PSR state tracking support
   - Eviction fixes
   - BDW DP aux channel timeout fixes
   - LSPCON fixes
   - Cannonlake PLL fixes

  amdgpu:
   - Per VM BO support
   - Powerplay cleanups
   - CI powerplay support
   - PASID mgr for kfd
   - SR-IOV fixes
   - initial GPU reset for vega10
   - Prime mmap support
   - TTM updates
   - Clock query interface for Raven
   - Fence to handle ioctl
   - UVD encode ring support on Polaris
   - Transparent huge page DMA support
   - Compute LRU pipe tweaks
   - BO flag to allow buffers to opt out of implicit sync
   - CTX priority setting API
   - VRAM lost infrastructure plumbing

  qxl:
   - fix flicker since atomic rework

  amdkfd:
   - Further improvements from internal AMD tree
   - Usermode events
   - Drop radeon support

  nouveau:
   - Pascal temperature sensor support
   - Improved BAR2 handling
   - MMU rework to support Pascal MMU

  exynos:
   - Improved HDMI/mixer support
   - HDMI audio interface support

  tegra:
   - Prep work for tegra186
   - Cleanup/fixes

  msm:
   - Preemption support for a5xx
   - Display fixes for 8x96 (snapdragon 820)
   - Async cursor plane fixes
   - FW loading rework
   - GPU debugging improvements

  vc4:
   - Prep for DSI panels
   - fix T-format tiling scanout
   - New madvise ioctl

  Rockchip:
   - LVDS support

  omapdrm:
   - omap4 HDMI CEC support

  etnaviv:
   - GPU performance counters groundwork

  sun4i:
   - refactor driver load + TCON backend
   - HDMI improvements
   - A31 support
   - Misc fixes

  udl:
   - Probe/EDID read fixes.

  tilcdc:
   - Misc fixes.

  pl111:
   - Support more variants

  adv7511:
   - Improve EDID handling.
   - HDMI CEC support

  sii8620:
   - Add remote control support"

* tag 'drm-for-v4.15' of git://people.freedesktop.org/~airlied/linux: (1480 commits)
  drm/rockchip: analogix_dp: Use mutex rather than spinlock
  drm/mode_object: fix documentation for object lookups.
  drm/i915: Reorder context-close to avoid calling i915_vma_close() under RCU
  drm/i915: Move init_clock_gating() back to where it was
  drm/i915: Prune the reservation shared fence array
  drm/i915: Idle the GPU before shinking everything
  drm/i915: Lock llist_del_first() vs llist_del_all()
  drm/i915: Calculate ironlake intermediate watermarks correctly, v2.
  drm/i915: Disable lazy PPGTT page table optimization for vGPU
  drm/i915/execlists: Remove the priority "optimisation"
  drm/i915: Filter out spurious execlists context-switch interrupts
  drm/amdgpu: use irq-safe lock for kiq->ring_lock
  drm/amdgpu: bypass lru touch for KIQ ring submission
  drm/amdgpu: Potential uninitialized variable in amdgpu_vm_update_directories()
  drm/amdgpu: potential uninitialized variable in amdgpu_vce_ring_parse_cs()
  drm/amd/powerplay: initialize a variable before using it
  drm/amd/powerplay: suppress KASAN out of bounds warning in vega10_populate_all_memory_levels
  drm/amd/amdgpu: fix evicted VRAM bo adjudgement condition
  drm/vblank: Tune drm_crtc_accurate_vblank_count() WARN down to a debug
  drm/rockchip: add CONFIG_OF dependency for lvds
  ...

23 files changed, 146 insertions, 129 deletions

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c
index bc77eea351a5..881015080d83 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c
@@ -986,14 +986,14 @@ gf100_grctx_pack_tpc[] = {
  ******************************************************************************/
 
 int
-gf100_grctx_mmio_data(struct gf100_grctx *info, u32 size, u32 align, u32 access)
+gf100_grctx_mmio_data(struct gf100_grctx *info, u32 size, u32 align, bool priv)
 {
 	if (info->data) {
 		info->buffer[info->buffer_nr] = round_up(info->addr, align);
 		info->addr = info->buffer[info->buffer_nr] + size;
 		info->data->size = size;
 		info->data->align = align;
-		info->data->access = access;
+		info->data->priv = priv;
 		info->data++;
 		return info->buffer_nr++;
 	}
@@ -1028,9 +1028,8 @@ void
 gf100_grctx_generate_bundle(struct gf100_grctx *info)
 {
 	const struct gf100_grctx_func *grctx = info->gr->func->grctx;
-	const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS;
 	const int s = 8;
-	const int b = mmio_vram(info, grctx->bundle_size, (1 << s), access);
+	const int b = mmio_vram(info, grctx->bundle_size, (1 << s), true);
 	mmio_refn(info, 0x408004, 0x00000000, s, b);
 	mmio_wr32(info, 0x408008, 0x80000000 | (grctx->bundle_size >> s));
 	mmio_refn(info, 0x418808, 0x00000000, s, b);
@@ -1041,9 +1040,8 @@ void
 gf100_grctx_generate_pagepool(struct gf100_grctx *info)
 {
 	const struct gf100_grctx_func *grctx = info->gr->func->grctx;
-	const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS;
 	const int s = 8;
-	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), access);
+	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), true);
 	mmio_refn(info, 0x40800c, 0x00000000, s, b);
 	mmio_wr32(info, 0x408010, 0x80000000);
 	mmio_refn(info, 0x419004, 0x00000000, s, b);
@@ -1057,9 +1055,8 @@ gf100_grctx_generate_attrib(struct gf100_grctx *info)
 	const struct gf100_grctx_func *grctx = gr->func->grctx;
 	const u32 attrib = grctx->attrib_nr;
 	const u32   size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max);
-	const u32 access = NV_MEM_ACCESS_RW;
 	const int s = 12;
-	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access);
+	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), false);
 	int gpc, tpc;
 	u32 bo = 0;
 
@@ -1267,85 +1264,87 @@ gf100_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info)
 	nvkm_mc_unk260(device, 1);
 }
 
+#define CB_RESERVED 0x80000
+
 int
 gf100_grctx_generate(struct gf100_gr *gr)
 {
 	const struct gf100_grctx_func *grctx = gr->func->grctx;
 	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
 	struct nvkm_device *device = subdev->device;
-	struct nvkm_memory *chan;
+	struct nvkm_memory *inst = NULL;
+	struct nvkm_memory *data = NULL;
+	struct nvkm_vmm *vmm = NULL;
+	struct nvkm_vma *ctx = NULL;
 	struct gf100_grctx info;
 	int ret, i;
 	u64 addr;
 
-	/* allocate memory to for a "channel", which we'll use to generate
-	 * the default context values
+	/* Allocate memory to for a "channel", which we'll use to generate
+	 * the default context values.
 	 */
-	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x80000 + gr->size,
-			      0x1000, true, &chan);
-	if (ret) {
-		nvkm_error(subdev, "failed to allocate chan memory, %d\n", ret);
-		return ret;
-	}
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST,
+			      0x1000, 0x1000, true, &inst);
+	if (ret)
+		goto done;
 
-	addr = nvkm_memory_addr(chan);
+	ret = nvkm_vmm_new(device, 0, 0, NULL, 0, NULL, "grctx", &vmm);
+	if (ret)
+		goto done;
 
-	/* PGD pointer */
-	nvkm_kmap(chan);
-	nvkm_wo32(chan, 0x0200, lower_32_bits(addr + 0x1000));
-	nvkm_wo32(chan, 0x0204, upper_32_bits(addr + 0x1000));
-	nvkm_wo32(chan, 0x0208, 0xffffffff);
-	nvkm_wo32(chan, 0x020c, 0x000000ff);
+	vmm->debug = subdev->debug;
 
-	/* PGT[0] pointer */
-	nvkm_wo32(chan, 0x1000, 0x00000000);
-	nvkm_wo32(chan, 0x1004, 0x00000001 | (addr + 0x2000) >> 8);
+	ret = nvkm_vmm_join(vmm, inst);
+	if (ret)
+		goto done;
 
-	/* identity-map the whole "channel" into its own vm */
-	for (i = 0; i < nvkm_memory_size(chan) / 4096; i++) {
-		u64 addr = ((nvkm_memory_addr(chan) + (i * 4096)) >> 8) | 1;
-		nvkm_wo32(chan, 0x2000 + (i * 8), lower_32_bits(addr));
-		nvkm_wo32(chan, 0x2004 + (i * 8), upper_32_bits(addr));
-	}
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST,
+			      CB_RESERVED + gr->size, 0, true, &data);
+	if (ret)
+		goto done;
 
-	/* context pointer (virt) */
-	nvkm_wo32(chan, 0x0210, 0x00080004);
-	nvkm_wo32(chan, 0x0214, 0x00000000);
-	nvkm_done(chan);
+	ret = nvkm_vmm_get(vmm, 0, nvkm_memory_size(data), &ctx);
+	if (ret)
+		goto done;
 
-	nvkm_wr32(device, 0x100cb8, (addr + 0x1000) >> 8);
-	nvkm_wr32(device, 0x100cbc, 0x80000001);
-	nvkm_msec(device, 2000,
-		if (nvkm_rd32(device, 0x100c80) & 0x00008000)
-			break;
-	);
+	ret = nvkm_memory_map(data, 0, vmm, ctx, NULL, 0);
+	if (ret)
+		goto done;
+
+
+	/* Setup context pointer. */
+	nvkm_kmap(inst);
+	nvkm_wo32(inst, 0x0210, lower_32_bits(ctx->addr + CB_RESERVED) | 4);
+	nvkm_wo32(inst, 0x0214, upper_32_bits(ctx->addr + CB_RESERVED));
+	nvkm_done(inst);
 
-	/* setup default state for mmio list construction */
+	/* Setup default state for mmio list construction. */
 	info.gr = gr;
 	info.data = gr->mmio_data;
 	info.mmio = gr->mmio_list;
-	info.addr = 0x2000 + (i * 8);
+	info.addr = ctx->addr;
 	info.buffer_nr = 0;
 
-	/* make channel current */
+	/* Make channel current. */
+	addr = nvkm_memory_addr(inst) >> 12;
 	if (gr->firmware) {
 		nvkm_wr32(device, 0x409840, 0x00000030);
-		nvkm_wr32(device, 0x409500, 0x80000000 | addr >> 12);
+		nvkm_wr32(device, 0x409500, 0x80000000 | addr);
 		nvkm_wr32(device, 0x409504, 0x00000003);
 		nvkm_msec(device, 2000,
 			if (nvkm_rd32(device, 0x409800) & 0x00000010)
 				break;
 		);
 
-		nvkm_kmap(chan);
-		nvkm_wo32(chan, 0x8001c, 1);
-		nvkm_wo32(chan, 0x80020, 0);
-		nvkm_wo32(chan, 0x80028, 0);
-		nvkm_wo32(chan, 0x8002c, 0);
-		nvkm_done(chan);
+		nvkm_kmap(data);
+		nvkm_wo32(data, 0x1c, 1);
+		nvkm_wo32(data, 0x20, 0);
+		nvkm_wo32(data, 0x28, 0);
+		nvkm_wo32(data, 0x2c, 0);
+		nvkm_done(data);
 	} else {
 		nvkm_wr32(device, 0x409840, 0x80000000);
-		nvkm_wr32(device, 0x409500, 0x80000000 | addr >> 12);
+		nvkm_wr32(device, 0x409500, 0x80000000 | addr);
 		nvkm_wr32(device, 0x409504, 0x00000001);
 		nvkm_msec(device, 2000,
 			if (nvkm_rd32(device, 0x409800) & 0x80000000)
@@ -1355,8 +1354,8 @@ gf100_grctx_generate(struct gf100_gr *gr)
 
 	grctx->main(gr, &info);
 
-	/* trigger a context unload by unsetting the "next channel valid" bit
-	 * and faking a context switch interrupt
+	/* Trigger a context unload by unsetting the "next channel valid" bit
+	 * and faking a context switch interrupt.
 	 */
 	nvkm_mask(device, 0x409b04, 0x80000000, 0x00000000);
 	nvkm_wr32(device, 0x409000, 0x00000100);
@@ -1370,17 +1369,21 @@ gf100_grctx_generate(struct gf100_gr *gr)
 
 	gr->data = kmalloc(gr->size, GFP_KERNEL);
 	if (gr->data) {
-		nvkm_kmap(chan);
+		nvkm_kmap(data);
 		for (i = 0; i < gr->size; i += 4)
-			gr->data[i / 4] = nvkm_ro32(chan, 0x80000 + i);
-		nvkm_done(chan);
+			gr->data[i / 4] = nvkm_ro32(data, CB_RESERVED + i);
+		nvkm_done(data);
 		ret = 0;
 	} else {
 		ret = -ENOMEM;
 	}
 
 done:
-	nvkm_memory_del(&chan);
+	nvkm_vmm_put(vmm, &ctx);
+	nvkm_memory_unref(&data);
+	nvkm_vmm_part(vmm, inst);
+	nvkm_vmm_unref(&vmm);
+	nvkm_memory_unref(&inst);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h
index 2812ca511c9c..5199e5aa0cb7 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h
@@ -12,7 +12,7 @@ struct gf100_grctx {
 	u64 addr;
 };
 
-int  gf100_grctx_mmio_data(struct gf100_grctx *, u32 size, u32 align, u32 access);
+int  gf100_grctx_mmio_data(struct gf100_grctx *, u32 size, u32 align, bool priv);
 void gf100_grctx_mmio_item(struct gf100_grctx *, u32 addr, u32 data, int s, int);
 
 #define mmio_vram(a,b,c,d) gf100_grctx_mmio_data((a), (b), (c), (d))
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c
index 505cdcbfc085..82f71b10c06e 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c
@@ -735,9 +735,8 @@ gf108_grctx_generate_attrib(struct gf100_grctx *info)
 	const u32  alpha = grctx->alpha_nr;
 	const u32   beta = grctx->attrib_nr;
 	const u32   size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max);
-	const u32 access = NV_MEM_ACCESS_RW;
 	const int s = 12;
-	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access);
+	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), false);
 	const int timeslice_mode = 1;
 	const int max_batches = 0xffff;
 	u32 bo = 0;
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c
index 74a64e3fd59a..19301d88577d 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c
@@ -187,9 +187,8 @@ gf117_grctx_generate_attrib(struct gf100_grctx *info)
 	const u32  alpha = grctx->alpha_nr;
 	const u32   beta = grctx->attrib_nr;
 	const u32   size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max);
-	const u32 access = NV_MEM_ACCESS_RW;
 	const int s = 12;
-	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access);
+	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), false);
 	const int timeslice_mode = 1;
 	const int max_batches = 0xffff;
 	u32 bo = 0;
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c
index c46b3fdf7203..825c8fd500bc 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c
@@ -847,9 +847,8 @@ gk104_grctx_generate_bundle(struct gf100_grctx *info)
 	const u32 state_limit = min(grctx->bundle_min_gpm_fifo_depth,
 				    grctx->bundle_size / 0x20);
 	const u32 token_limit = grctx->bundle_token_limit;
-	const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS;
 	const int s = 8;
-	const int b = mmio_vram(info, grctx->bundle_size, (1 << s), access);
+	const int b = mmio_vram(info, grctx->bundle_size, (1 << s), true);
 	mmio_refn(info, 0x408004, 0x00000000, s, b);
 	mmio_wr32(info, 0x408008, 0x80000000 | (grctx->bundle_size >> s));
 	mmio_refn(info, 0x418808, 0x00000000, s, b);
@@ -861,9 +860,8 @@ void
 gk104_grctx_generate_pagepool(struct gf100_grctx *info)
 {
 	const struct gf100_grctx_func *grctx = info->gr->func->grctx;
-	const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS;
 	const int s = 8;
-	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), access);
+	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), true);
 	mmio_refn(info, 0x40800c, 0x00000000, s, b);
 	mmio_wr32(info, 0x408010, 0x80000000);
 	mmio_refn(info, 0x419004, 0x00000000, s, b);
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c
index 4c4b5ab6e46d..9b43d4ce3eaa 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c
@@ -867,9 +867,8 @@ gm107_grctx_generate_bundle(struct gf100_grctx *info)
 	const u32 state_limit = min(grctx->bundle_min_gpm_fifo_depth,
 				    grctx->bundle_size / 0x20);
 	const u32 token_limit = grctx->bundle_token_limit;
-	const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS;
 	const int s = 8;
-	const int b = mmio_vram(info, grctx->bundle_size, (1 << s), access);
+	const int b = mmio_vram(info, grctx->bundle_size, (1 << s), true);
 	mmio_refn(info, 0x408004, 0x00000000, s, b);
 	mmio_wr32(info, 0x408008, 0x80000000 | (grctx->bundle_size >> s));
 	mmio_refn(info, 0x418e24, 0x00000000, s, b);
@@ -881,9 +880,8 @@ void
 gm107_grctx_generate_pagepool(struct gf100_grctx *info)
 {
 	const struct gf100_grctx_func *grctx = info->gr->func->grctx;
-	const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS;
 	const int s = 8;
-	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), access);
+	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), true);
 	mmio_refn(info, 0x40800c, 0x00000000, s, b);
 	mmio_wr32(info, 0x408010, 0x80000000);
 	mmio_refn(info, 0x419004, 0x00000000, s, b);
@@ -900,9 +898,8 @@ gm107_grctx_generate_attrib(struct gf100_grctx *info)
 	const u32  alpha = grctx->alpha_nr;
 	const u32 attrib = grctx->attrib_nr;
 	const u32   size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max);
-	const u32 access = NV_MEM_ACCESS_RW;
 	const int s = 12;
-	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access);
+	const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), false);
 	const int max_batches = 0xffff;
 	u32 bo = 0;
 	u32 ao = bo + grctx->attrib_nr_max * gr->tpc_total;
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c
index 7833bc777a29..88ea322d956c 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp100.c
@@ -33,9 +33,8 @@ void
 gp100_grctx_generate_pagepool(struct gf100_grctx *info)
 {
 	const struct gf100_grctx_func *grctx = info->gr->func->grctx;
-	const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS;
 	const int s = 8;
-	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), access);
+	const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), true);
 	mmio_refn(info, 0x40800c, 0x00000000, s, b);
 	mmio_wr32(info, 0x408010, 0x80000000);
 	mmio_refn(info, 0x419004, 0x00000000, s, b);
@@ -51,9 +50,8 @@ gp100_grctx_generate_attrib(struct gf100_grctx *info)
 	const u32 attrib = grctx->attrib_nr;
 	const u32 pertpc = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max);
 	const u32   size = roundup(gr->tpc_total * pertpc, 0x80);
-	const u32 access = NV_MEM_ACCESS_RW;
 	const int s = 12;
-	const int b = mmio_vram(info, size, (1 << s), access);
+	const int b = mmio_vram(info, size, (1 << s), false);
 	const int max_batches = 0xffff;
 	u32 ao = 0;
 	u32 bo = ao + grctx->alpha_nr_max * gr->tpc_total;
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c
index 80b7ab0bee3a..7a66b4c2eb18 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgp102.c
@@ -38,9 +38,8 @@ gp102_grctx_generate_attrib(struct gf100_grctx *info)
 	const u32 attrib = grctx->attrib_nr;
 	const u32 pertpc = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max);
 	const u32   size = roundup(gr->tpc_total * pertpc, 0x80);
-	const u32 access = NV_MEM_ACCESS_RW;
 	const int s = 12;
-	const int b = mmio_vram(info, size, (1 << s), access);
+	const int b = mmio_vram(info, size, (1 << s), false);
 	const int max_batches = 0xffff;
 	u32 ao = 0;
 	u32 bo = ao + grctx->alpha_nr_max * gr->tpc_total;
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c
index 99689f4de502..2f8dc107047d 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c
@@ -37,6 +37,7 @@
 
 #include <nvif/class.h>
 #include <nvif/cl9097.h>
+#include <nvif/if900d.h>
 #include <nvif/unpack.h>
 
 /*******************************************************************************
@@ -327,13 +328,13 @@ gf100_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent,
 
 	if (!gr->firmware) {
 		nvkm_wo32(*pgpuobj, 0x00, chan->mmio_nr / 2);
-		nvkm_wo32(*pgpuobj, 0x04, chan->mmio_vma.offset >> 8);
+		nvkm_wo32(*pgpuobj, 0x04, chan->mmio_vma->addr >> 8);
 	} else {
 		nvkm_wo32(*pgpuobj, 0xf4, 0);
 		nvkm_wo32(*pgpuobj, 0xf8, 0);
 		nvkm_wo32(*pgpuobj, 0x10, chan->mmio_nr / 2);
-		nvkm_wo32(*pgpuobj, 0x14, lower_32_bits(chan->mmio_vma.offset));
-		nvkm_wo32(*pgpuobj, 0x18, upper_32_bits(chan->mmio_vma.offset));
+		nvkm_wo32(*pgpuobj, 0x14, lower_32_bits(chan->mmio_vma->addr));
+		nvkm_wo32(*pgpuobj, 0x18, upper_32_bits(chan->mmio_vma->addr));
 		nvkm_wo32(*pgpuobj, 0x1c, 1);
 		nvkm_wo32(*pgpuobj, 0x20, 0);
 		nvkm_wo32(*pgpuobj, 0x28, 0);
@@ -350,18 +351,13 @@ gf100_gr_chan_dtor(struct nvkm_object *object)
 	int i;
 
 	for (i = 0; i < ARRAY_SIZE(chan->data); i++) {
-		if (chan->data[i].vma.node) {
-			nvkm_vm_unmap(&chan->data[i].vma);
-			nvkm_vm_put(&chan->data[i].vma);
-		}
-		nvkm_memory_del(&chan->data[i].mem);
+		nvkm_vmm_put(chan->vmm, &chan->data[i].vma);
+		nvkm_memory_unref(&chan->data[i].mem);
 	}
 
-	if (chan->mmio_vma.node) {
-		nvkm_vm_unmap(&chan->mmio_vma);
-		nvkm_vm_put(&chan->mmio_vma);
-	}
-	nvkm_memory_del(&chan->mmio);
+	nvkm_vmm_put(chan->vmm, &chan->mmio_vma);
+	nvkm_memory_unref(&chan->mmio);
+	nvkm_vmm_unref(&chan->vmm);
 	return chan;
 }
 
@@ -380,6 +376,7 @@ gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch,
 	struct gf100_gr_data *data = gr->mmio_data;
 	struct gf100_gr_mmio *mmio = gr->mmio_list;
 	struct gf100_gr_chan *chan;
+	struct gf100_vmm_map_v0 args = { .priv = 1 };
 	struct nvkm_device *device = gr->base.engine.subdev.device;
 	int ret, i;
 
@@ -387,6 +384,7 @@ gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch,
 		return -ENOMEM;
 	nvkm_object_ctor(&gf100_gr_chan, oclass, &chan->object);
 	chan->gr = gr;
+	chan->vmm = nvkm_vmm_ref(fifoch->vmm);
 	*pobject = &chan->object;
 
 	/* allocate memory for a "mmio list" buffer that's used by the HUB
@@ -398,12 +396,14 @@ gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch,
 	if (ret)
 		return ret;
 
-	ret = nvkm_vm_get(fifoch->vm, 0x1000, 12, NV_MEM_ACCESS_RW |
-			  NV_MEM_ACCESS_SYS, &chan->mmio_vma);
+	ret = nvkm_vmm_get(fifoch->vmm, 12, 0x1000, &chan->mmio_vma);
 	if (ret)
 		return ret;
 
-	nvkm_memory_map(chan->mmio, &chan->mmio_vma, 0);
+	ret = nvkm_memory_map(chan->mmio, 0, fifoch->vmm,
+			      chan->mmio_vma, &args, sizeof(args));
+	if (ret)
+		return ret;
 
 	/* allocate buffers referenced by mmio list */
 	for (i = 0; data->size && i < ARRAY_SIZE(gr->mmio_data); i++) {
@@ -413,13 +413,19 @@ gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch,
 		if (ret)
 			return ret;
 
-		ret = nvkm_vm_get(fifoch->vm,
-				  nvkm_memory_size(chan->data[i].mem), 12,
-				  data->access, &chan->data[i].vma);
+		ret = nvkm_vmm_get(fifoch->vmm, 12,
+				   nvkm_memory_size(chan->data[i].mem),
+				   &chan->data[i].vma);
+		if (ret)
+			return ret;
+
+		args.priv = data->priv;
+
+		ret = nvkm_memory_map(chan->data[i].mem, 0, chan->vmm,
+				      chan->data[i].vma, &args, sizeof(args));
 		if (ret)
 			return ret;
 
-		nvkm_memory_map(chan->data[i].mem, &chan->data[i].vma, 0);
 		data++;
 	}
 
@@ -430,7 +436,7 @@ gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch,
 		u32 data = mmio->data;
 
 		if (mmio->buffer >= 0) {
-			u64 info = chan->data[mmio->buffer].vma.offset;
+			u64 info = chan->data[mmio->buffer].vma->addr;
 			data |= info >> mmio->shift;
 		}
 
@@ -1855,8 +1861,12 @@ gf100_gr_ctor_fw(struct gf100_gr *gr, const char *fwname,
 	int ret;
 
 	ret = nvkm_firmware_get(device, fwname, &fw);
-	if (ret)
-		return gf100_gr_ctor_fw_legacy(gr, fwname, fuc, ret);
+	if (ret) {
+		ret = gf100_gr_ctor_fw_legacy(gr, fwname, fuc, ret);
+		if (ret)
+			return -ENODEV;
+		return 0;
+	}
 
 	fuc->size = fw->size;
 	fuc->data = kmemdup(fw->data, fuc->size, GFP_KERNEL);
@@ -1903,25 +1913,33 @@ gf100_gr_new_(const struct gf100_gr_func *func, struct nvkm_device *device,
 	return 0;
 }
 
+void
+gf100_gr_init_gpc_mmu(struct gf100_gr *gr)
+{
+	struct nvkm_device *device = gr->base.engine.subdev.device;
+	struct nvkm_fb *fb = device->fb;
+
+	nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0x00000001);
+	nvkm_wr32(device, 0x4188a4, 0x00000000);
+	nvkm_wr32(device, 0x418888, 0x00000000);
+	nvkm_wr32(device, 0x41888c, 0x00000000);
+	nvkm_wr32(device, 0x418890, 0x00000000);
+	nvkm_wr32(device, 0x418894, 0x00000000);
+	nvkm_wr32(device, 0x4188b4, nvkm_memory_addr(fb->mmu_wr) >> 8);
+	nvkm_wr32(device, 0x4188b8, nvkm_memory_addr(fb->mmu_rd) >> 8);
+}
+
 int
 gf100_gr_init(struct gf100_gr *gr)
 {
 	struct nvkm_device *device = gr->base.engine.subdev.device;
-	struct nvkm_fb *fb = device->fb;
 	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total);
 	u32 data[TPC_MAX / 8] = {};
 	u8  tpcnr[GPC_MAX];
 	int gpc, tpc, rop;
 	int i;
 
-	nvkm_wr32(device, GPC_BCAST(0x0880), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x08a4), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x0888), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x088c), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(fb->mmu_wr) >> 8);
-	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(fb->mmu_rd) >> 8);
+	gr->func->init_gpc_mmu(gr);
 
 	gf100_gr_mmio(gr, gr->func->mmio);
 
@@ -2036,6 +2054,7 @@ gf100_gr_gpccs_ucode = {
 static const struct gf100_gr_func
 gf100_gr = {
 	.init = gf100_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.mmio = gf100_gr_pack_mmio,
 	.fecs.ucode = &gf100_gr_fecs_ucode,
 	.gpccs.ucode = &gf100_gr_gpccs_ucode,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h
index a36e45a4a635..d7c2adb9b543 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h
@@ -45,7 +45,7 @@
 struct gf100_gr_data {
 	u32 size;
 	u32 align;
-	u32 access;
+	bool priv;
 };
 
 struct gf100_gr_mmio {
@@ -156,18 +156,20 @@ int gp100_gr_init(struct gf100_gr *);
 void gp100_gr_init_rop_active_fbps(struct gf100_gr *);
 
 #define gf100_gr_chan(p) container_of((p), struct gf100_gr_chan, object)
+#include <core/object.h>
 
 struct gf100_gr_chan {
 	struct nvkm_object object;
 	struct gf100_gr *gr;
+	struct nvkm_vmm *vmm;
 
 	struct nvkm_memory *mmio;
-	struct nvkm_vma mmio_vma;
+	struct nvkm_vma *mmio_vma;
 	int mmio_nr;
 
 	struct {
 		struct nvkm_memory *mem;
-		struct nvkm_vma vma;
+		struct nvkm_vma *vma;
 	} data[4];
 };
 
@@ -253,6 +255,7 @@ extern const struct gf100_gr_init gf100_gr_init_mpc_0[];
 extern const struct gf100_gr_init gf100_gr_init_be_0[];
 extern const struct gf100_gr_init gf100_gr_init_fe_1[];
 extern const struct gf100_gr_init gf100_gr_init_pe_1[];
+void gf100_gr_init_gpc_mmu(struct gf100_gr *);
 
 extern const struct gf100_gr_init gf104_gr_init_ds_0[];
 extern const struct gf100_gr_init gf104_gr_init_tex_0[];
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c
index d736dcd55ea2..ec0f11983b23 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c
@@ -115,6 +115,7 @@ gf104_gr_pack_mmio[] = {
 static const struct gf100_gr_func
 gf104_gr = {
 	.init = gf100_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.mmio = gf104_gr_pack_mmio,
 	.fecs.ucode = &gf100_gr_fecs_ucode,
 	.gpccs.ucode = &gf100_gr_gpccs_ucode,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c
index 2f0d24498427..cc152eb74123 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c
@@ -106,6 +106,7 @@ gf108_gr_pack_mmio[] = {
 static const struct gf100_gr_func
 gf108_gr = {
 	.init = gf100_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.mmio = gf108_gr_pack_mmio,
 	.fecs.ucode = &gf100_gr_fecs_ucode,
 	.gpccs.ucode = &gf100_gr_gpccs_ucode,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c
index d1d942eb86af..10d2d73ca8c3 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c
@@ -87,6 +87,7 @@ gf110_gr_pack_mmio[] = {
 static const struct gf100_gr_func
 gf110_gr = {
 	.init = gf100_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.mmio = gf110_gr_pack_mmio,
 	.fecs.ucode = &gf100_gr_fecs_ucode,
 	.gpccs.ucode = &gf100_gr_gpccs_ucode,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c
index 0124e468086e..ac09a07c4150 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c
@@ -123,6 +123,7 @@ gf117_gr_gpccs_ucode = {
 static const struct gf100_gr_func
 gf117_gr = {
 	.init = gf100_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.mmio = gf117_gr_pack_mmio,
 	.fecs.ucode = &gf117_gr_fecs_ucode,
 	.gpccs.ucode = &gf117_gr_gpccs_ucode,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c
index 8d8e4cafe28f..7f449ec6f760 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c
@@ -178,6 +178,7 @@ gf119_gr_pack_mmio[] = {
 static const struct gf100_gr_func
 gf119_gr = {
 	.init = gf100_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.mmio = gf119_gr_pack_mmio,
 	.fecs.ucode = &gf100_gr_fecs_ucode,
 	.gpccs.ucode = &gf100_gr_gpccs_ucode,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c
index ec22da6c99fc..5e82f94c2245 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c
@@ -24,8 +24,6 @@
 #include "gf100.h"
 #include "ctxgf100.h"
 
-#include <subdev/fb.h>
-
 #include <nvif/class.h>
 
 /*******************************************************************************
@@ -207,21 +205,13 @@ int
 gk104_gr_init(struct gf100_gr *gr)
 {
 	struct nvkm_device *device = gr->base.engine.subdev.device;
-	struct nvkm_fb *fb = device->fb;
 	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total);
 	u32 data[TPC_MAX / 8] = {};
 	u8  tpcnr[GPC_MAX];
 	int gpc, tpc, rop;
 	int i;
 
-	nvkm_wr32(device, GPC_BCAST(0x0880), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x08a4), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x0888), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x088c), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(fb->mmu_wr) >> 8);
-	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(fb->mmu_rd) >> 8);
+	gr->func->init_gpc_mmu(gr);
 
 	gf100_gr_mmio(gr, gr->func->mmio);
 
@@ -339,6 +329,7 @@ gk104_gr_gpccs_ucode = {
 static const struct gf100_gr_func
 gk104_gr = {
 	.init = gk104_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.init_rop_active_fbps = gk104_gr_init_rop_active_fbps,
 	.init_ppc_exceptions = gk104_gr_init_ppc_exceptions,
 	.mmio = gk104_gr_pack_mmio,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c
index f31b171a4102..a38e19b61c1d 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c
@@ -183,6 +183,7 @@ gk110_gr_gpccs_ucode = {
 static const struct gf100_gr_func
 gk110_gr = {
 	.init = gk104_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.init_rop_active_fbps = gk104_gr_init_rop_active_fbps,
 	.init_ppc_exceptions = gk104_gr_init_ppc_exceptions,
 	.mmio = gk110_gr_pack_mmio,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c
index d76dd178007f..1912c0bfd7ee 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c
@@ -103,6 +103,7 @@ gk110b_gr_pack_mmio[] = {
 static const struct gf100_gr_func
 gk110b_gr = {
 	.init = gk104_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.init_rop_active_fbps = gk104_gr_init_rop_active_fbps,
 	.init_ppc_exceptions = gk104_gr_init_ppc_exceptions,
 	.mmio = gk110b_gr_pack_mmio,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c
index 14bbe6ed02a9..1fc258163f25 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c
@@ -162,6 +162,7 @@ gk208_gr_gpccs_ucode = {
 static const struct gf100_gr_func
 gk208_gr = {
 	.init = gk104_gr_init,
+	.init_gpc_mmu = gf100_gr_init_gpc_mmu,
 	.init_rop_active_fbps = gk104_gr_init_rop_active_fbps,
 	.init_ppc_exceptions = gk104_gr_init_ppc_exceptions,
 	.mmio = gk208_gr_pack_mmio,
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c
index df2cd864147c..111c8bb4497b 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c
@@ -60,7 +60,7 @@ void *
 nv20_gr_chan_dtor(struct nvkm_object *object)
 {
 	struct nv20_gr_chan *chan = nv20_gr_chan(object);
-	nvkm_memory_del(&chan->inst);
+	nvkm_memory_unref(&chan->inst);
 	return chan;
 }
 
@@ -324,7 +324,7 @@ void *
 nv20_gr_dtor(struct nvkm_gr *base)
 {
 	struct nv20_gr *gr = nv20_gr(base);
-	nvkm_memory_del(&gr->ctxtab);
+	nvkm_memory_unref(&gr->ctxtab);
 	return gr;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h
index ad7e53bb7c23..979dc5f7b32e 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h
@@ -20,6 +20,7 @@ void nv20_gr_tile(struct nvkm_gr *, int, struct nvkm_fb_tile *);
 int nv30_gr_init(struct nvkm_gr *);
 
 #define nv20_gr_chan(p) container_of((p), struct nv20_gr_chan, object)
+#include <core/object.h>
 
 struct nv20_gr_chan {
 	struct nvkm_object object;
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h
index 89b773233ac5..731400937edd 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h
@@ -17,6 +17,7 @@ void nv40_gr_intr(struct nvkm_gr *);
 u64 nv40_gr_units(struct nvkm_gr *);
 
 #define nv40_gr_chan(p) container_of((p), struct nv40_gr_chan, object)
+#include <core/object.h>
 
 struct nv40_gr_chan {
 	struct nvkm_object object;
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h
index 567fa4f3e518..5b9d99bee207 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h
@@ -20,6 +20,7 @@ u64 nv50_gr_units(struct nvkm_gr *);
 int g84_gr_tlb_flush(struct nvkm_gr *);
 
 #define nv50_gr_chan(p) container_of((p), struct nv50_gr_chan, object)
+#include <core/object.h>
 
 struct nv50_gr_chan {
 	struct nvkm_object object;