drm/nouveau/fifo: switch to device pri macros

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

7 files changed, 345 insertions, 289 deletions

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c
index bff5867e24ce..dfc27a0d2ac2 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/g84.c
@@ -80,10 +80,11 @@ static int
 g84_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 			struct nvkm_object *object)
 {
-	struct nvkm_bar *bar = nvkm_bar(parent);
 	struct nv50_fifo *fifo = (void *)parent->engine;
 	struct nv50_fifo_base *base = (void *)parent->parent;
 	struct nv50_fifo_chan *chan = (void *)parent;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	struct nvkm_bar *bar = device->bar;
 	u32 addr, save, engn;
 	bool done;
 
@@ -103,10 +104,10 @@ g84_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 		return -EINVAL;
 	}
 
-	save = nv_mask(fifo, 0x002520, 0x0000003f, 1 << engn);
-	nv_wr32(fifo, 0x0032fc, nv_gpuobj(base)->addr >> 12);
+	save = nvkm_mask(device, 0x002520, 0x0000003f, 1 << engn);
+	nvkm_wr32(device, 0x0032fc, nv_gpuobj(base)->addr >> 12);
 	done = nv_wait_ne(fifo, 0x0032fc, 0xffffffff, 0xffffffff);
-	nv_wr32(fifo, 0x002520, save);
+	nvkm_wr32(device, 0x002520, save);
 	if (!done) {
 		nv_error(fifo, "channel %d [%s] unload timeout\n",
 			 chan->base.chid, nvkm_client_name(chan));
@@ -313,6 +314,7 @@ g84_fifo_chan_init(struct nvkm_object *object)
 	struct nv50_fifo_base *base = (void *)object->parent;
 	struct nv50_fifo_chan *chan = (void *)object;
 	struct nvkm_gpuobj *ramfc = base->ramfc;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 	int ret;
 
@@ -320,7 +322,7 @@ g84_fifo_chan_init(struct nvkm_object *object)
 	if (ret)
 		return ret;
 
-	nv_wr32(fifo, 0x002600 + (chid * 4), 0x80000000 | ramfc->addr >> 8);
+	nvkm_wr32(device, 0x002600 + (chid * 4), 0x80000000 | ramfc->addr >> 8);
 	nv50_fifo_playlist_update(fifo);
 	return 0;
 }
@@ -422,14 +424,16 @@ static void
 g84_fifo_uevent_init(struct nvkm_event *event, int type, int index)
 {
 	struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), uevent);
-	nv_mask(fifo, 0x002140, 0x40000000, 0x40000000);
+	struct nvkm_device *device = fifo->engine.subdev.device;
+	nvkm_mask(device, 0x002140, 0x40000000, 0x40000000);
 }
 
 static void
 g84_fifo_uevent_fini(struct nvkm_event *event, int type, int index)
 {
 	struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), uevent);
-	nv_mask(fifo, 0x002140, 0x40000000, 0x00000000);
+	struct nvkm_device *device = fifo->engine.subdev.device;
+	nvkm_mask(device, 0x002140, 0x40000000, 0x00000000);
 }
 
 static const struct nvkm_event_func
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c
index 0a7971a3317c..f714bda4230f 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gf100.c
@@ -76,7 +76,8 @@ struct gf100_fifo_chan {
 static void
 gf100_fifo_runlist_update(struct gf100_fifo *fifo)
 {
-	struct nvkm_bar *bar = nvkm_bar(fifo);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	struct nvkm_bar *bar = device->bar;
 	struct nvkm_gpuobj *cur;
 	int i, p;
 
@@ -94,11 +95,11 @@ gf100_fifo_runlist_update(struct gf100_fifo *fifo)
 	}
 	bar->flush(bar);
 
-	nv_wr32(fifo, 0x002270, cur->addr >> 12);
-	nv_wr32(fifo, 0x002274, 0x01f00000 | (p >> 3));
+	nvkm_wr32(device, 0x002270, cur->addr >> 12);
+	nvkm_wr32(device, 0x002274, 0x01f00000 | (p >> 3));
 
 	if (wait_event_timeout(fifo->runlist.wait,
-			       !(nv_rd32(fifo, 0x00227c) & 0x00100000),
+			       !(nvkm_rd32(device, 0x00227c) & 0x00100000),
 			       msecs_to_jiffies(2000)) == 0)
 		nv_error(fifo, "runlist update timeout\n");
 	mutex_unlock(&nv_subdev(fifo)->mutex);
@@ -145,10 +146,11 @@ static int
 gf100_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 			  struct nvkm_object *object)
 {
-	struct nvkm_bar *bar = nvkm_bar(parent);
 	struct gf100_fifo *fifo = (void *)parent->engine;
 	struct gf100_fifo_base *base = (void *)parent->parent;
 	struct gf100_fifo_chan *chan = (void *)parent;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	struct nvkm_bar *bar = device->bar;
 	u32 addr;
 
 	switch (nv_engidx(object->engine)) {
@@ -163,7 +165,7 @@ gf100_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 		return -EINVAL;
 	}
 
-	nv_wr32(fifo, 0x002634, chan->base.chid);
+	nvkm_wr32(device, 0x002634, chan->base.chid);
 	if (!nv_wait(fifo, 0x002634, 0xffffffff, chan->base.chid)) {
 		nv_error(fifo, "channel %d [%s] kick timeout\n",
 			 chan->base.chid, nvkm_client_name(chan));
@@ -253,6 +255,7 @@ gf100_fifo_chan_init(struct nvkm_object *object)
 	struct nvkm_gpuobj *base = nv_gpuobj(object->parent);
 	struct gf100_fifo *fifo = (void *)object->engine;
 	struct gf100_fifo_chan *chan = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 	int ret;
 
@@ -260,10 +263,10 @@ gf100_fifo_chan_init(struct nvkm_object *object)
 	if (ret)
 		return ret;
 
-	nv_wr32(fifo, 0x003000 + (chid * 8), 0xc0000000 | base->addr >> 12);
+	nvkm_wr32(device, 0x003000 + (chid * 8), 0xc0000000 | base->addr >> 12);
 
 	if (chan->state == STOPPED && (chan->state = RUNNING) == RUNNING) {
-		nv_wr32(fifo, 0x003004 + (chid * 8), 0x001f0001);
+		nvkm_wr32(device, 0x003004 + (chid * 8), 0x001f0001);
 		gf100_fifo_runlist_update(fifo);
 	}
 
@@ -277,16 +280,17 @@ gf100_fifo_chan_fini(struct nvkm_object *object, bool suspend)
 {
 	struct gf100_fifo *fifo = (void *)object->engine;
 	struct gf100_fifo_chan *chan = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 
 	if (chan->state == RUNNING && (chan->state = STOPPED) == STOPPED) {
-		nv_mask(fifo, 0x003004 + (chid * 8), 0x00000001, 0x00000000);
+		nvkm_mask(device, 0x003004 + (chid * 8), 0x00000001, 0x00000000);
 		gf100_fifo_runlist_update(fifo);
 	}
 
 	gf100_fifo_intr_engine(fifo);
 
-	nv_wr32(fifo, 0x003000 + (chid * 8), 0x00000000);
+	nvkm_wr32(device, 0x003000 + (chid * 8), 0x00000000);
 	return nvkm_fifo_channel_fini(&chan->base, suspend);
 }
 
@@ -408,6 +412,7 @@ static void
 gf100_fifo_recover_work(struct work_struct *work)
 {
 	struct gf100_fifo *fifo = container_of(work, typeof(*fifo), fault);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	struct nvkm_object *engine;
 	unsigned long flags;
 	u32 engn, engm = 0;
@@ -420,7 +425,7 @@ gf100_fifo_recover_work(struct work_struct *work)
 
 	for (todo = mask; engn = __ffs64(todo), todo; todo &= ~(1 << engn))
 		engm |= 1 << gf100_fifo_engidx(fifo, engn);
-	nv_mask(fifo, 0x002630, engm, engm);
+	nvkm_mask(device, 0x002630, engm, engm);
 
 	for (todo = mask; engn = __ffs64(todo), todo; todo &= ~(1 << engn)) {
 		if ((engine = (void *)nvkm_engine(fifo, engn))) {
@@ -430,21 +435,22 @@ gf100_fifo_recover_work(struct work_struct *work)
 	}
 
 	gf100_fifo_runlist_update(fifo);
-	nv_wr32(fifo, 0x00262c, engm);
-	nv_mask(fifo, 0x002630, engm, 0x00000000);
+	nvkm_wr32(device, 0x00262c, engm);
+	nvkm_mask(device, 0x002630, engm, 0x00000000);
 }
 
 static void
 gf100_fifo_recover(struct gf100_fifo *fifo, struct nvkm_engine *engine,
 		   struct gf100_fifo_chan *chan)
 {
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 	unsigned long flags;
 
 	nv_error(fifo, "%s engine fault on channel %d, recovering...\n",
 		       nv_subdev(engine)->name, chid);
 
-	nv_mask(fifo, 0x003004 + (chid * 0x08), 0x00000001, 0x00000000);
+	nvkm_mask(device, 0x003004 + (chid * 0x08), 0x00000001, 0x00000000);
 	chan->state = KILLED;
 
 	spin_lock_irqsave(&fifo->base.lock, flags);
@@ -488,12 +494,13 @@ gf100_fifo_sched_reason[] = {
 static void
 gf100_fifo_intr_sched_ctxsw(struct gf100_fifo *fifo)
 {
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	struct nvkm_engine *engine;
 	struct gf100_fifo_chan *chan;
 	u32 engn;
 
 	for (engn = 0; engn < 6; engn++) {
-		u32 stat = nv_rd32(fifo, 0x002640 + (engn * 0x04));
+		u32 stat = nvkm_rd32(device, 0x002640 + (engn * 0x04));
 		u32 busy = (stat & 0x80000000);
 		u32 save = (stat & 0x00100000); /* maybe? */
 		u32 unk0 = (stat & 0x00040000);
@@ -514,7 +521,8 @@ gf100_fifo_intr_sched_ctxsw(struct gf100_fifo *fifo)
 static void
 gf100_fifo_intr_sched(struct gf100_fifo *fifo)
 {
-	u32 intr = nv_rd32(fifo, 0x00254c);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 intr = nvkm_rd32(device, 0x00254c);
 	u32 code = intr & 0x000000ff;
 	const struct nvkm_enum *en;
 	char enunk[6] = "";
@@ -596,10 +604,11 @@ gf100_fifo_fault_gpcclient[] = {
 static void
 gf100_fifo_intr_fault(struct gf100_fifo *fifo, int unit)
 {
-	u32 inst = nv_rd32(fifo, 0x002800 + (unit * 0x10));
-	u32 valo = nv_rd32(fifo, 0x002804 + (unit * 0x10));
-	u32 vahi = nv_rd32(fifo, 0x002808 + (unit * 0x10));
-	u32 stat = nv_rd32(fifo, 0x00280c + (unit * 0x10));
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 inst = nvkm_rd32(device, 0x002800 + (unit * 0x10));
+	u32 valo = nvkm_rd32(device, 0x002804 + (unit * 0x10));
+	u32 vahi = nvkm_rd32(device, 0x002808 + (unit * 0x10));
+	u32 stat = nvkm_rd32(device, 0x00280c + (unit * 0x10));
 	u32 gpc    = (stat & 0x1f000000) >> 24;
 	u32 client = (stat & 0x00001f00) >> 8;
 	u32 write  = (stat & 0x00000080);
@@ -621,13 +630,13 @@ gf100_fifo_intr_fault(struct gf100_fifo *fifo, int unit)
 	if (eu) {
 		switch (eu->data2) {
 		case NVDEV_SUBDEV_BAR:
-			nv_mask(fifo, 0x001704, 0x00000000, 0x00000000);
+			nvkm_mask(device, 0x001704, 0x00000000, 0x00000000);
 			break;
 		case NVDEV_SUBDEV_INSTMEM:
-			nv_mask(fifo, 0x001714, 0x00000000, 0x00000000);
+			nvkm_mask(device, 0x001714, 0x00000000, 0x00000000);
 			break;
 		case NVDEV_ENGINE_IFB:
-			nv_mask(fifo, 0x001718, 0x00000000, 0x00000000);
+			nvkm_mask(device, 0x001718, 0x00000000, 0x00000000);
 			break;
 		default:
 			engine = nvkm_engine(fifo, eu->data2);
@@ -680,10 +689,11 @@ gf100_fifo_pbdma_intr[] = {
 static void
 gf100_fifo_intr_pbdma(struct gf100_fifo *fifo, int unit)
 {
-	u32 stat = nv_rd32(fifo, 0x040108 + (unit * 0x2000));
-	u32 addr = nv_rd32(fifo, 0x0400c0 + (unit * 0x2000));
-	u32 data = nv_rd32(fifo, 0x0400c4 + (unit * 0x2000));
-	u32 chid = nv_rd32(fifo, 0x040120 + (unit * 0x2000)) & 0x7f;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 stat = nvkm_rd32(device, 0x040108 + (unit * 0x2000));
+	u32 addr = nvkm_rd32(device, 0x0400c0 + (unit * 0x2000));
+	u32 data = nvkm_rd32(device, 0x0400c4 + (unit * 0x2000));
+	u32 chid = nvkm_rd32(device, 0x040120 + (unit * 0x2000)) & 0x7f;
 	u32 subc = (addr & 0x00070000) >> 16;
 	u32 mthd = (addr & 0x00003ffc);
 	u32 show = stat;
@@ -704,35 +714,37 @@ gf100_fifo_intr_pbdma(struct gf100_fifo *fifo, int unit)
 			 subc, mthd, data);
 	}
 
-	nv_wr32(fifo, 0x0400c0 + (unit * 0x2000), 0x80600008);
-	nv_wr32(fifo, 0x040108 + (unit * 0x2000), stat);
+	nvkm_wr32(device, 0x0400c0 + (unit * 0x2000), 0x80600008);
+	nvkm_wr32(device, 0x040108 + (unit * 0x2000), stat);
 }
 
 static void
 gf100_fifo_intr_runlist(struct gf100_fifo *fifo)
 {
-	u32 intr = nv_rd32(fifo, 0x002a00);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 intr = nvkm_rd32(device, 0x002a00);
 
 	if (intr & 0x10000000) {
 		wake_up(&fifo->runlist.wait);
-		nv_wr32(fifo, 0x002a00, 0x10000000);
+		nvkm_wr32(device, 0x002a00, 0x10000000);
 		intr &= ~0x10000000;
 	}
 
 	if (intr) {
 		nv_error(fifo, "RUNLIST 0x%08x\n", intr);
-		nv_wr32(fifo, 0x002a00, intr);
+		nvkm_wr32(device, 0x002a00, intr);
 	}
 }
 
 static void
 gf100_fifo_intr_engine_unit(struct gf100_fifo *fifo, int engn)
 {
-	u32 intr = nv_rd32(fifo, 0x0025a8 + (engn * 0x04));
-	u32 inte = nv_rd32(fifo, 0x002628);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 intr = nvkm_rd32(device, 0x0025a8 + (engn * 0x04));
+	u32 inte = nvkm_rd32(device, 0x002628);
 	u32 unkn;
 
-	nv_wr32(fifo, 0x0025a8 + (engn * 0x04), intr);
+	nvkm_wr32(device, 0x0025a8 + (engn * 0x04), intr);
 
 	for (unkn = 0; unkn < 8; unkn++) {
 		u32 ints = (intr >> (unkn * 0x04)) & inte;
@@ -742,7 +754,7 @@ gf100_fifo_intr_engine_unit(struct gf100_fifo *fifo, int engn)
 		}
 		if (ints) {
 			nv_error(fifo, "ENGINE %d %d %01x", engn, unkn, ints);
-			nv_mask(fifo, 0x002628, ints, 0);
+			nvkm_mask(device, 0x002628, ints, 0);
 		}
 	}
 }
@@ -750,7 +762,8 @@ gf100_fifo_intr_engine_unit(struct gf100_fifo *fifo, int engn)
 static void
 gf100_fifo_intr_engine(struct gf100_fifo *fifo)
 {
-	u32 mask = nv_rd32(fifo, 0x0025a4);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 mask = nvkm_rd32(device, 0x0025a4);
 	while (mask) {
 		u32 unit = __ffs(mask);
 		gf100_fifo_intr_engine_unit(fifo, unit);
@@ -762,53 +775,54 @@ static void
 gf100_fifo_intr(struct nvkm_subdev *subdev)
 {
 	struct gf100_fifo *fifo = (void *)subdev;
-	u32 mask = nv_rd32(fifo, 0x002140);
-	u32 stat = nv_rd32(fifo, 0x002100) & mask;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 mask = nvkm_rd32(device, 0x002140);
+	u32 stat = nvkm_rd32(device, 0x002100) & mask;
 
 	if (stat & 0x00000001) {
-		u32 intr = nv_rd32(fifo, 0x00252c);
+		u32 intr = nvkm_rd32(device, 0x00252c);
 		nv_warn(fifo, "INTR 0x00000001: 0x%08x\n", intr);
-		nv_wr32(fifo, 0x002100, 0x00000001);
+		nvkm_wr32(device, 0x002100, 0x00000001);
 		stat &= ~0x00000001;
 	}
 
 	if (stat & 0x00000100) {
 		gf100_fifo_intr_sched(fifo);
-		nv_wr32(fifo, 0x002100, 0x00000100);
+		nvkm_wr32(device, 0x002100, 0x00000100);
 		stat &= ~0x00000100;
 	}
 
 	if (stat & 0x00010000) {
-		u32 intr = nv_rd32(fifo, 0x00256c);
+		u32 intr = nvkm_rd32(device, 0x00256c);
 		nv_warn(fifo, "INTR 0x00010000: 0x%08x\n", intr);
-		nv_wr32(fifo, 0x002100, 0x00010000);
+		nvkm_wr32(device, 0x002100, 0x00010000);
 		stat &= ~0x00010000;
 	}
 
 	if (stat & 0x01000000) {
-		u32 intr = nv_rd32(fifo, 0x00258c);
+		u32 intr = nvkm_rd32(device, 0x00258c);
 		nv_warn(fifo, "INTR 0x01000000: 0x%08x\n", intr);
-		nv_wr32(fifo, 0x002100, 0x01000000);
+		nvkm_wr32(device, 0x002100, 0x01000000);
 		stat &= ~0x01000000;
 	}
 
 	if (stat & 0x10000000) {
-		u32 mask = nv_rd32(fifo, 0x00259c);
+		u32 mask = nvkm_rd32(device, 0x00259c);
 		while (mask) {
 			u32 unit = __ffs(mask);
 			gf100_fifo_intr_fault(fifo, unit);
-			nv_wr32(fifo, 0x00259c, (1 << unit));
+			nvkm_wr32(device, 0x00259c, (1 << unit));
 			mask &= ~(1 << unit);
 		}
 		stat &= ~0x10000000;
 	}
 
 	if (stat & 0x20000000) {
-		u32 mask = nv_rd32(fifo, 0x0025a0);
+		u32 mask = nvkm_rd32(device, 0x0025a0);
 		while (mask) {
 			u32 unit = __ffs(mask);
 			gf100_fifo_intr_pbdma(fifo, unit);
-			nv_wr32(fifo, 0x0025a0, (1 << unit));
+			nvkm_wr32(device, 0x0025a0, (1 << unit));
 			mask &= ~(1 << unit);
 		}
 		stat &= ~0x20000000;
@@ -826,8 +840,8 @@ gf100_fifo_intr(struct nvkm_subdev *subdev)
 
 	if (stat) {
 		nv_error(fifo, "INTR 0x%08x\n", stat);
-		nv_mask(fifo, 0x002140, stat, 0x00000000);
-		nv_wr32(fifo, 0x002100, stat);
+		nvkm_mask(device, 0x002140, stat, 0x00000000);
+		nvkm_wr32(device, 0x002100, stat);
 	}
 }
 
@@ -835,14 +849,16 @@ static void
 gf100_fifo_uevent_init(struct nvkm_event *event, int type, int index)
 {
 	struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), uevent);
-	nv_mask(fifo, 0x002140, 0x80000000, 0x80000000);
+	struct nvkm_device *device = fifo->engine.subdev.device;
+	nvkm_mask(device, 0x002140, 0x80000000, 0x80000000);
 }
 
 static void
 gf100_fifo_uevent_fini(struct nvkm_event *event, int type, int index)
 {
 	struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), uevent);
-	nv_mask(fifo, 0x002140, 0x80000000, 0x00000000);
+	struct nvkm_device *device = fifo->engine.subdev.device;
+	nvkm_mask(device, 0x002140, 0x80000000, 0x00000000);
 }
 
 static const struct nvkm_event_func
@@ -917,41 +933,42 @@ static int
 gf100_fifo_init(struct nvkm_object *object)
 {
 	struct gf100_fifo *fifo = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	int ret, i;
 
 	ret = nvkm_fifo_init(&fifo->base);
 	if (ret)
 		return ret;
 
-	nv_wr32(fifo, 0x000204, 0xffffffff);
-	nv_wr32(fifo, 0x002204, 0xffffffff);
+	nvkm_wr32(device, 0x000204, 0xffffffff);
+	nvkm_wr32(device, 0x002204, 0xffffffff);
 
-	fifo->spoon_nr = hweight32(nv_rd32(fifo, 0x002204));
+	fifo->spoon_nr = hweight32(nvkm_rd32(device, 0x002204));
 	nv_debug(fifo, "%d PBDMA unit(s)\n", fifo->spoon_nr);
 
 	/* assign engines to PBDMAs */
 	if (fifo->spoon_nr >= 3) {
-		nv_wr32(fifo, 0x002208, ~(1 << 0)); /* PGRAPH */
-		nv_wr32(fifo, 0x00220c, ~(1 << 1)); /* PVP */
-		nv_wr32(fifo, 0x002210, ~(1 << 1)); /* PMSPP */
-		nv_wr32(fifo, 0x002214, ~(1 << 1)); /* PMSVLD */
-		nv_wr32(fifo, 0x002218, ~(1 << 2)); /* PCE0 */
-		nv_wr32(fifo, 0x00221c, ~(1 << 1)); /* PCE1 */
+		nvkm_wr32(device, 0x002208, ~(1 << 0)); /* PGRAPH */
+		nvkm_wr32(device, 0x00220c, ~(1 << 1)); /* PVP */
+		nvkm_wr32(device, 0x002210, ~(1 << 1)); /* PMSPP */
+		nvkm_wr32(device, 0x002214, ~(1 << 1)); /* PMSVLD */
+		nvkm_wr32(device, 0x002218, ~(1 << 2)); /* PCE0 */
+		nvkm_wr32(device, 0x00221c, ~(1 << 1)); /* PCE1 */
 	}
 
 	/* PBDMA[n] */
 	for (i = 0; i < fifo->spoon_nr; i++) {
-		nv_mask(fifo, 0x04013c + (i * 0x2000), 0x10000100, 0x00000000);
-		nv_wr32(fifo, 0x040108 + (i * 0x2000), 0xffffffff); /* INTR */
-		nv_wr32(fifo, 0x04010c + (i * 0x2000), 0xfffffeff); /* INTREN */
+		nvkm_mask(device, 0x04013c + (i * 0x2000), 0x10000100, 0x00000000);
+		nvkm_wr32(device, 0x040108 + (i * 0x2000), 0xffffffff); /* INTR */
+		nvkm_wr32(device, 0x04010c + (i * 0x2000), 0xfffffeff); /* INTREN */
 	}
 
-	nv_mask(fifo, 0x002200, 0x00000001, 0x00000001);
-	nv_wr32(fifo, 0x002254, 0x10000000 | fifo->user.bar.offset >> 12);
+	nvkm_mask(device, 0x002200, 0x00000001, 0x00000001);
+	nvkm_wr32(device, 0x002254, 0x10000000 | fifo->user.bar.offset >> 12);
 
-	nv_wr32(fifo, 0x002100, 0xffffffff);
-	nv_wr32(fifo, 0x002140, 0x7fffffff);
-	nv_wr32(fifo, 0x002628, 0x00000001); /* ENGINE_INTR_EN */
+	nvkm_wr32(device, 0x002100, 0xffffffff);
+	nvkm_wr32(device, 0x002140, 0x7fffffff);
+	nvkm_wr32(device, 0x002628, 0x00000001); /* ENGINE_INTR_EN */
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c
index ed8d3820a044..cec1281dbd12 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/gk104.c
@@ -95,8 +95,9 @@ struct gk104_fifo_chan {
 static void
 gk104_fifo_runlist_update(struct gk104_fifo *fifo, u32 engine)
 {
-	struct nvkm_bar *bar = nvkm_bar(fifo);
 	struct gk104_fifo_engn *engn = &fifo->engine[engine];
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	struct nvkm_bar *bar = device->bar;
 	struct nvkm_gpuobj *cur;
 	int i, p;
 
@@ -114,10 +115,10 @@ gk104_fifo_runlist_update(struct gk104_fifo *fifo, u32 engine)
 	}
 	bar->flush(bar);
 
-	nv_wr32(fifo, 0x002270, cur->addr >> 12);
-	nv_wr32(fifo, 0x002274, (engine << 20) | (p >> 3));
+	nvkm_wr32(device, 0x002270, cur->addr >> 12);
+	nvkm_wr32(device, 0x002274, (engine << 20) | (p >> 3));
 
-	if (wait_event_timeout(engn->wait, !(nv_rd32(fifo, 0x002284 +
+	if (wait_event_timeout(engn->wait, !(nvkm_rd32(device, 0x002284 +
 			       (engine * 0x08)) & 0x00100000),
 				msecs_to_jiffies(2000)) == 0)
 		nv_error(fifo, "runlist %d update timeout\n", engine);
@@ -170,8 +171,9 @@ gk104_fifo_chan_kick(struct gk104_fifo_chan *chan)
 {
 	struct nvkm_object *obj = (void *)chan;
 	struct gk104_fifo *fifo = (void *)obj->engine;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 
-	nv_wr32(fifo, 0x002634, chan->base.chid);
+	nvkm_wr32(device, 0x002634, chan->base.chid);
 	if (!nv_wait(fifo, 0x002634, 0x100000, 0x000000)) {
 		nv_error(fifo, "channel %d [%s] kick timeout\n",
 			 chan->base.chid, nvkm_client_name(chan));
@@ -300,6 +302,7 @@ gk104_fifo_chan_init(struct nvkm_object *object)
 	struct nvkm_gpuobj *base = nv_gpuobj(object->parent);
 	struct gk104_fifo *fifo = (void *)object->engine;
 	struct gk104_fifo_chan *chan = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 	int ret;
 
@@ -307,13 +310,13 @@ gk104_fifo_chan_init(struct nvkm_object *object)
 	if (ret)
 		return ret;
 
-	nv_mask(fifo, 0x800004 + (chid * 8), 0x000f0000, chan->engine << 16);
-	nv_wr32(fifo, 0x800000 + (chid * 8), 0x80000000 | base->addr >> 12);
+	nvkm_mask(device, 0x800004 + (chid * 8), 0x000f0000, chan->engine << 16);
+	nvkm_wr32(device, 0x800000 + (chid * 8), 0x80000000 | base->addr >> 12);
 
 	if (chan->state == STOPPED && (chan->state = RUNNING) == RUNNING) {
-		nv_mask(fifo, 0x800004 + (chid * 8), 0x00000400, 0x00000400);
+		nvkm_mask(device, 0x800004 + (chid * 8), 0x00000400, 0x00000400);
 		gk104_fifo_runlist_update(fifo, chan->engine);
-		nv_mask(fifo, 0x800004 + (chid * 8), 0x00000400, 0x00000400);
+		nvkm_mask(device, 0x800004 + (chid * 8), 0x00000400, 0x00000400);
 	}
 
 	return 0;
@@ -324,14 +327,15 @@ gk104_fifo_chan_fini(struct nvkm_object *object, bool suspend)
 {
 	struct gk104_fifo *fifo = (void *)object->engine;
 	struct gk104_fifo_chan *chan = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 
 	if (chan->state == RUNNING && (chan->state = STOPPED) == STOPPED) {
-		nv_mask(fifo, 0x800004 + (chid * 8), 0x00000800, 0x00000800);
+		nvkm_mask(device, 0x800004 + (chid * 8), 0x00000800, 0x00000800);
 		gk104_fifo_runlist_update(fifo, chan->engine);
 	}
 
-	nv_wr32(fifo, 0x800000 + (chid * 8), 0x00000000);
+	nvkm_wr32(device, 0x800000 + (chid * 8), 0x00000000);
 	return nvkm_fifo_channel_fini(&chan->base, suspend);
 }
 
@@ -445,6 +449,7 @@ static void
 gk104_fifo_recover_work(struct work_struct *work)
 {
 	struct gk104_fifo *fifo = container_of(work, typeof(*fifo), fault);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	struct nvkm_object *engine;
 	unsigned long flags;
 	u32 engn, engm = 0;
@@ -457,7 +462,7 @@ gk104_fifo_recover_work(struct work_struct *work)
 
 	for (todo = mask; engn = __ffs64(todo), todo; todo &= ~(1 << engn))
 		engm |= 1 << gk104_fifo_engidx(fifo, engn);
-	nv_mask(fifo, 0x002630, engm, engm);
+	nvkm_mask(device, 0x002630, engm, engm);
 
 	for (todo = mask; engn = __ffs64(todo), todo; todo &= ~(1 << engn)) {
 		if ((engine = (void *)nvkm_engine(fifo, engn))) {
@@ -467,21 +472,22 @@ gk104_fifo_recover_work(struct work_struct *work)
 		gk104_fifo_runlist_update(fifo, gk104_fifo_engidx(fifo, engn));
 	}
 
-	nv_wr32(fifo, 0x00262c, engm);
-	nv_mask(fifo, 0x002630, engm, 0x00000000);
+	nvkm_wr32(device, 0x00262c, engm);
+	nvkm_mask(device, 0x002630, engm, 0x00000000);
 }
 
 static void
 gk104_fifo_recover(struct gk104_fifo *fifo, struct nvkm_engine *engine,
 		  struct gk104_fifo_chan *chan)
 {
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 	unsigned long flags;
 
 	nv_error(fifo, "%s engine fault on channel %d, recovering...\n",
 		       nv_subdev(engine)->name, chid);
 
-	nv_mask(fifo, 0x800004 + (chid * 0x08), 0x00000800, 0x00000800);
+	nvkm_mask(device, 0x800004 + (chid * 0x08), 0x00000800, 0x00000800);
 	chan->state = KILLED;
 
 	spin_lock_irqsave(&fifo->base.lock, flags);
@@ -530,7 +536,8 @@ gk104_fifo_bind_reason[] = {
 static void
 gk104_fifo_intr_bind(struct gk104_fifo *fifo)
 {
-	u32 intr = nv_rd32(fifo, 0x00252c);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 intr = nvkm_rd32(device, 0x00252c);
 	u32 code = intr & 0x000000ff;
 	const struct nvkm_enum *en;
 	char enunk[6] = "";
@@ -551,12 +558,13 @@ gk104_fifo_sched_reason[] = {
 static void
 gk104_fifo_intr_sched_ctxsw(struct gk104_fifo *fifo)
 {
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	struct nvkm_engine *engine;
 	struct gk104_fifo_chan *chan;
 	u32 engn;
 
 	for (engn = 0; engn < ARRAY_SIZE(fifo_engine); engn++) {
-		u32 stat = nv_rd32(fifo, 0x002640 + (engn * 0x04));
+		u32 stat = nvkm_rd32(device, 0x002640 + (engn * 0x04));
 		u32 busy = (stat & 0x80000000);
 		u32 next = (stat & 0x07ff0000) >> 16;
 		u32 chsw = (stat & 0x00008000);
@@ -579,7 +587,8 @@ gk104_fifo_intr_sched_ctxsw(struct gk104_fifo *fifo)
 static void
 gk104_fifo_intr_sched(struct gk104_fifo *fifo)
 {
-	u32 intr = nv_rd32(fifo, 0x00254c);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 intr = nvkm_rd32(device, 0x00254c);
 	u32 code = intr & 0x000000ff;
 	const struct nvkm_enum *en;
 	char enunk[6] = "";
@@ -602,15 +611,17 @@ gk104_fifo_intr_sched(struct gk104_fifo *fifo)
 static void
 gk104_fifo_intr_chsw(struct gk104_fifo *fifo)
 {
-	u32 stat = nv_rd32(fifo, 0x00256c);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 stat = nvkm_rd32(device, 0x00256c);
 	nv_error(fifo, "CHSW_ERROR 0x%08x\n", stat);
-	nv_wr32(fifo, 0x00256c, stat);
+	nvkm_wr32(device, 0x00256c, stat);
 }
 
 static void
 gk104_fifo_intr_dropped_fault(struct gk104_fifo *fifo)
 {
-	u32 stat = nv_rd32(fifo, 0x00259c);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 stat = nvkm_rd32(device, 0x00259c);
 	nv_error(fifo, "DROPPED_MMU_FAULT 0x%08x\n", stat);
 }
 
@@ -722,10 +733,11 @@ gk104_fifo_fault_gpcclient[] = {
 static void
 gk104_fifo_intr_fault(struct gk104_fifo *fifo, int unit)
 {
-	u32 inst = nv_rd32(fifo, 0x002800 + (unit * 0x10));
-	u32 valo = nv_rd32(fifo, 0x002804 + (unit * 0x10));
-	u32 vahi = nv_rd32(fifo, 0x002808 + (unit * 0x10));
-	u32 stat = nv_rd32(fifo, 0x00280c + (unit * 0x10));
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 inst = nvkm_rd32(device, 0x002800 + (unit * 0x10));
+	u32 valo = nvkm_rd32(device, 0x002804 + (unit * 0x10));
+	u32 vahi = nvkm_rd32(device, 0x002808 + (unit * 0x10));
+	u32 stat = nvkm_rd32(device, 0x00280c + (unit * 0x10));
 	u32 gpc    = (stat & 0x1f000000) >> 24;
 	u32 client = (stat & 0x00001f00) >> 8;
 	u32 write  = (stat & 0x00000080);
@@ -747,13 +759,13 @@ gk104_fifo_intr_fault(struct gk104_fifo *fifo, int unit)
 	if (eu) {
 		switch (eu->data2) {
 		case NVDEV_SUBDEV_BAR:
-			nv_mask(fifo, 0x001704, 0x00000000, 0x00000000);
+			nvkm_mask(device, 0x001704, 0x00000000, 0x00000000);
 			break;
 		case NVDEV_SUBDEV_INSTMEM:
-			nv_mask(fifo, 0x001714, 0x00000000, 0x00000000);
+			nvkm_mask(device, 0x001714, 0x00000000, 0x00000000);
 			break;
 		case NVDEV_ENGINE_IFB:
-			nv_mask(fifo, 0x001718, 0x00000000, 0x00000000);
+			nvkm_mask(device, 0x001718, 0x00000000, 0x00000000);
 			break;
 		default:
 			engine = nvkm_engine(fifo, eu->data2);
@@ -833,11 +845,12 @@ static const struct nvkm_bitfield gk104_fifo_pbdma_intr_0[] = {
 static void
 gk104_fifo_intr_pbdma_0(struct gk104_fifo *fifo, int unit)
 {
-	u32 mask = nv_rd32(fifo, 0x04010c + (unit * 0x2000));
-	u32 stat = nv_rd32(fifo, 0x040108 + (unit * 0x2000)) & mask;
-	u32 addr = nv_rd32(fifo, 0x0400c0 + (unit * 0x2000));
-	u32 data = nv_rd32(fifo, 0x0400c4 + (unit * 0x2000));
-	u32 chid = nv_rd32(fifo, 0x040120 + (unit * 0x2000)) & 0xfff;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 mask = nvkm_rd32(device, 0x04010c + (unit * 0x2000));
+	u32 stat = nvkm_rd32(device, 0x040108 + (unit * 0x2000)) & mask;
+	u32 addr = nvkm_rd32(device, 0x0400c0 + (unit * 0x2000));
+	u32 data = nvkm_rd32(device, 0x0400c4 + (unit * 0x2000));
+	u32 chid = nvkm_rd32(device, 0x040120 + (unit * 0x2000)) & 0xfff;
 	u32 subc = (addr & 0x00070000) >> 16;
 	u32 mthd = (addr & 0x00003ffc);
 	u32 show = stat;
@@ -845,7 +858,7 @@ gk104_fifo_intr_pbdma_0(struct gk104_fifo *fifo, int unit)
 	if (stat & 0x00800000) {
 		if (!gk104_fifo_swmthd(fifo, chid, mthd, data))
 			show &= ~0x00800000;
-		nv_wr32(fifo, 0x0400c0 + (unit * 0x2000), 0x80600008);
+		nvkm_wr32(device, 0x0400c0 + (unit * 0x2000), 0x80600008);
 	}
 
 	if (show) {
@@ -859,7 +872,7 @@ gk104_fifo_intr_pbdma_0(struct gk104_fifo *fifo, int unit)
 			 subc, mthd, data);
 	}
 
-	nv_wr32(fifo, 0x040108 + (unit * 0x2000), stat);
+	nvkm_wr32(device, 0x040108 + (unit * 0x2000), stat);
 }
 
 static const struct nvkm_bitfield gk104_fifo_pbdma_intr_1[] = {
@@ -874,30 +887,32 @@ static const struct nvkm_bitfield gk104_fifo_pbdma_intr_1[] = {
 static void
 gk104_fifo_intr_pbdma_1(struct gk104_fifo *fifo, int unit)
 {
-	u32 mask = nv_rd32(fifo, 0x04014c + (unit * 0x2000));
-	u32 stat = nv_rd32(fifo, 0x040148 + (unit * 0x2000)) & mask;
-	u32 chid = nv_rd32(fifo, 0x040120 + (unit * 0x2000)) & 0xfff;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 mask = nvkm_rd32(device, 0x04014c + (unit * 0x2000));
+	u32 stat = nvkm_rd32(device, 0x040148 + (unit * 0x2000)) & mask;
+	u32 chid = nvkm_rd32(device, 0x040120 + (unit * 0x2000)) & 0xfff;
 
 	if (stat) {
 		nv_error(fifo, "PBDMA%d:", unit);
 		nvkm_bitfield_print(gk104_fifo_pbdma_intr_1, stat);
 		pr_cont("\n");
 		nv_error(fifo, "PBDMA%d: ch %d %08x %08x\n", unit, chid,
-			 nv_rd32(fifo, 0x040150 + (unit * 0x2000)),
-			 nv_rd32(fifo, 0x040154 + (unit * 0x2000)));
+			 nvkm_rd32(device, 0x040150 + (unit * 0x2000)),
+			 nvkm_rd32(device, 0x040154 + (unit * 0x2000)));
 	}
 
-	nv_wr32(fifo, 0x040148 + (unit * 0x2000), stat);
+	nvkm_wr32(device, 0x040148 + (unit * 0x2000), stat);
 }
 
 static void
 gk104_fifo_intr_runlist(struct gk104_fifo *fifo)
 {
-	u32 mask = nv_rd32(fifo, 0x002a00);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 mask = nvkm_rd32(device, 0x002a00);
 	while (mask) {
 		u32 engn = __ffs(mask);
 		wake_up(&fifo->engine[engn].wait);
-		nv_wr32(fifo, 0x002a00, 1 << engn);
+		nvkm_wr32(device, 0x002a00, 1 << engn);
 		mask &= ~(1 << engn);
 	}
 }
@@ -912,69 +927,70 @@ static void
 gk104_fifo_intr(struct nvkm_subdev *subdev)
 {
 	struct gk104_fifo *fifo = (void *)subdev;
-	u32 mask = nv_rd32(fifo, 0x002140);
-	u32 stat = nv_rd32(fifo, 0x002100) & mask;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	u32 mask = nvkm_rd32(device, 0x002140);
+	u32 stat = nvkm_rd32(device, 0x002100) & mask;
 
 	if (stat & 0x00000001) {
 		gk104_fifo_intr_bind(fifo);
-		nv_wr32(fifo, 0x002100, 0x00000001);
+		nvkm_wr32(device, 0x002100, 0x00000001);
 		stat &= ~0x00000001;
 	}
 
 	if (stat & 0x00000010) {
 		nv_error(fifo, "PIO_ERROR\n");
-		nv_wr32(fifo, 0x002100, 0x00000010);
+		nvkm_wr32(device, 0x002100, 0x00000010);
 		stat &= ~0x00000010;
 	}
 
 	if (stat & 0x00000100) {
 		gk104_fifo_intr_sched(fifo);
-		nv_wr32(fifo, 0x002100, 0x00000100);
+		nvkm_wr32(device, 0x002100, 0x00000100);
 		stat &= ~0x00000100;
 	}
 
 	if (stat & 0x00010000) {
 		gk104_fifo_intr_chsw(fifo);
-		nv_wr32(fifo, 0x002100, 0x00010000);
+		nvkm_wr32(device, 0x002100, 0x00010000);
 		stat &= ~0x00010000;
 	}
 
 	if (stat & 0x00800000) {
 		nv_error(fifo, "FB_FLUSH_TIMEOUT\n");
-		nv_wr32(fifo, 0x002100, 0x00800000);
+		nvkm_wr32(device, 0x002100, 0x00800000);
 		stat &= ~0x00800000;
 	}
 
 	if (stat & 0x01000000) {
 		nv_error(fifo, "LB_ERROR\n");
-		nv_wr32(fifo, 0x002100, 0x01000000);
+		nvkm_wr32(device, 0x002100, 0x01000000);
 		stat &= ~0x01000000;
 	}
 
 	if (stat & 0x08000000) {
 		gk104_fifo_intr_dropped_fault(fifo);
-		nv_wr32(fifo, 0x002100, 0x08000000);
+		nvkm_wr32(device, 0x002100, 0x08000000);
 		stat &= ~0x08000000;
 	}
 
 	if (stat & 0x10000000) {
-		u32 mask = nv_rd32(fifo, 0x00259c);
+		u32 mask = nvkm_rd32(device, 0x00259c);
 		while (mask) {
 			u32 unit = __ffs(mask);
 			gk104_fifo_intr_fault(fifo, unit);
-			nv_wr32(fifo, 0x00259c, (1 << unit));
+			nvkm_wr32(device, 0x00259c, (1 << unit));
 			mask &= ~(1 << unit);
 		}
 		stat &= ~0x10000000;
 	}
 
 	if (stat & 0x20000000) {
-		u32 mask = nv_rd32(fifo, 0x0025a0);
+		u32 mask = nvkm_rd32(device, 0x0025a0);
 		while (mask) {
 			u32 unit = __ffs(mask);
 			gk104_fifo_intr_pbdma_0(fifo, unit);
 			gk104_fifo_intr_pbdma_1(fifo, unit);
-			nv_wr32(fifo, 0x0025a0, (1 << unit));
+			nvkm_wr32(device, 0x0025a0, (1 << unit));
 			mask &= ~(1 << unit);
 		}
 		stat &= ~0x20000000;
@@ -986,15 +1002,15 @@ gk104_fifo_intr(struct nvkm_subdev *subdev)
 	}
 
 	if (stat & 0x80000000) {
-		nv_wr32(fifo, 0x002100, 0x80000000);
+		nvkm_wr32(device, 0x002100, 0x80000000);
 		gk104_fifo_intr_engine(fifo);
 		stat &= ~0x80000000;
 	}
 
 	if (stat) {
 		nv_error(fifo, "INTR 0x%08x\n", stat);
-		nv_mask(fifo, 0x002140, stat, 0x00000000);
-		nv_wr32(fifo, 0x002100, stat);
+		nvkm_mask(device, 0x002140, stat, 0x00000000);
+		nvkm_wr32(device, 0x002100, stat);
 	}
 }
 
@@ -1002,14 +1018,16 @@ static void
 gk104_fifo_uevent_init(struct nvkm_event *event, int type, int index)
 {
 	struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), uevent);
-	nv_mask(fifo, 0x002140, 0x80000000, 0x80000000);
+	struct nvkm_device *device = fifo->engine.subdev.device;
+	nvkm_mask(device, 0x002140, 0x80000000, 0x80000000);
 }
 
 static void
 gk104_fifo_uevent_fini(struct nvkm_event *event, int type, int index)
 {
 	struct nvkm_fifo *fifo = container_of(event, typeof(*fifo), uevent);
-	nv_mask(fifo, 0x002140, 0x80000000, 0x00000000);
+	struct nvkm_device *device = fifo->engine.subdev.device;
+	nvkm_mask(device, 0x002140, 0x80000000, 0x00000000);
 }
 
 static const struct nvkm_event_func
@@ -1023,6 +1041,7 @@ int
 gk104_fifo_fini(struct nvkm_object *object, bool suspend)
 {
 	struct gk104_fifo *fifo = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	int ret;
 
 	ret = nvkm_fifo_fini(&fifo->base, suspend);
@@ -1030,7 +1049,7 @@ gk104_fifo_fini(struct nvkm_object *object, bool suspend)
 		return ret;
 
 	/* allow mmu fault interrupts, even when we're not using fifo */
-	nv_mask(fifo, 0x002140, 0x10000000, 0x10000000);
+	nvkm_mask(device, 0x002140, 0x10000000, 0x10000000);
 	return 0;
 }
 
@@ -1038,6 +1057,7 @@ int
 gk104_fifo_init(struct nvkm_object *object)
 {
 	struct gk104_fifo *fifo = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	int ret, i;
 
 	ret = nvkm_fifo_init(&fifo->base);
@@ -1045,27 +1065,27 @@ gk104_fifo_init(struct nvkm_object *object)
 		return ret;
 
 	/* enable all available PBDMA units */
-	nv_wr32(fifo, 0x000204, 0xffffffff);
-	fifo->spoon_nr = hweight32(nv_rd32(fifo, 0x000204));
+	nvkm_wr32(device, 0x000204, 0xffffffff);
+	fifo->spoon_nr = hweight32(nvkm_rd32(device, 0x000204));
 	nv_debug(fifo, "%d PBDMA unit(s)\n", fifo->spoon_nr);
 
 	/* PBDMA[n] */
 	for (i = 0; i < fifo->spoon_nr; i++) {
-		nv_mask(fifo, 0x04013c + (i * 0x2000), 0x10000100, 0x00000000);
-		nv_wr32(fifo, 0x040108 + (i * 0x2000), 0xffffffff); /* INTR */
-		nv_wr32(fifo, 0x04010c + (i * 0x2000), 0xfffffeff); /* INTREN */
+		nvkm_mask(device, 0x04013c + (i * 0x2000), 0x10000100, 0x00000000);
+		nvkm_wr32(device, 0x040108 + (i * 0x2000), 0xffffffff); /* INTR */
+		nvkm_wr32(device, 0x04010c + (i * 0x2000), 0xfffffeff); /* INTREN */
 	}
 
 	/* PBDMA[n].HCE */
 	for (i = 0; i < fifo->spoon_nr; i++) {
-		nv_wr32(fifo, 0x040148 + (i * 0x2000), 0xffffffff); /* INTR */
-		nv_wr32(fifo, 0x04014c + (i * 0x2000), 0xffffffff); /* INTREN */
+		nvkm_wr32(device, 0x040148 + (i * 0x2000), 0xffffffff); /* INTR */
+		nvkm_wr32(device, 0x04014c + (i * 0x2000), 0xffffffff); /* INTREN */
 	}
 
-	nv_wr32(fifo, 0x002254, 0x10000000 | fifo->user.bar.offset >> 12);
+	nvkm_wr32(device, 0x002254, 0x10000000 | fifo->user.bar.offset >> 12);
 
-	nv_wr32(fifo, 0x002100, 0xffffffff);
-	nv_wr32(fifo, 0x002140, 0x7fffffff);
+	nvkm_wr32(device, 0x002100, 0xffffffff);
+	nvkm_wr32(device, 0x002140, 0x7fffffff);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c
index 91a2080bdaf9..6339b473feb0 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv04.c
@@ -174,6 +174,7 @@ nv04_fifo_chan_init(struct nvkm_object *object)
 {
 	struct nv04_fifo *fifo = (void *)object->engine;
 	struct nv04_fifo_chan *chan = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 mask = 1 << chan->base.chid;
 	unsigned long flags;
 	int ret;
@@ -183,7 +184,7 @@ nv04_fifo_chan_init(struct nvkm_object *object)
 		return ret;
 
 	spin_lock_irqsave(&fifo->base.lock, flags);
-	nv_mask(fifo, NV04_PFIFO_MODE, mask, mask);
+	nvkm_mask(device, NV04_PFIFO_MODE, mask, mask);
 	spin_unlock_irqrestore(&fifo->base.lock, flags);
 	return 0;
 }
@@ -194,6 +195,7 @@ nv04_fifo_chan_fini(struct nvkm_object *object, bool suspend)
 	struct nv04_fifo *fifo = (void *)object->engine;
 	struct nv04_fifo_chan *chan = (void *)object;
 	struct nvkm_gpuobj *fctx = fifo->ramfc;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	struct ramfc_desc *c;
 	unsigned long flags;
 	u32 data = chan->ramfc;
@@ -201,39 +203,39 @@ nv04_fifo_chan_fini(struct nvkm_object *object, bool suspend)
 
 	/* prevent fifo context switches */
 	spin_lock_irqsave(&fifo->base.lock, flags);
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 0);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 0);
 
 	/* if this channel is active, replace it with a null context */
-	chid = nv_rd32(fifo, NV03_PFIFO_CACHE1_PUSH1) & fifo->base.max;
+	chid = nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH1) & fifo->base.max;
 	if (chid == chan->base.chid) {
-		nv_mask(fifo, NV04_PFIFO_CACHE1_DMA_PUSH, 0x00000001, 0);
-		nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH0, 0);
-		nv_mask(fifo, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0);
+		nvkm_mask(device, NV04_PFIFO_CACHE1_DMA_PUSH, 0x00000001, 0);
+		nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 0);
+		nvkm_mask(device, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0);
 
 		c = fifo->ramfc_desc;
 		do {
 			u32 rm = ((1ULL << c->bits) - 1) << c->regs;
 			u32 cm = ((1ULL << c->bits) - 1) << c->ctxs;
-			u32 rv = (nv_rd32(fifo, c->regp) &  rm) >> c->regs;
+			u32 rv = (nvkm_rd32(device, c->regp) &  rm) >> c->regs;
 			u32 cv = (nv_ro32(fctx, c->ctxp + data) & ~cm);
 			nv_wo32(fctx, c->ctxp + data, cv | (rv << c->ctxs));
 		} while ((++c)->bits);
 
 		c = fifo->ramfc_desc;
 		do {
-			nv_wr32(fifo, c->regp, 0x00000000);
+			nvkm_wr32(device, c->regp, 0x00000000);
 		} while ((++c)->bits);
 
-		nv_wr32(fifo, NV03_PFIFO_CACHE1_GET, 0);
-		nv_wr32(fifo, NV03_PFIFO_CACHE1_PUT, 0);
-		nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
-		nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH0, 1);
-		nv_wr32(fifo, NV04_PFIFO_CACHE1_PULL0, 1);
+		nvkm_wr32(device, NV03_PFIFO_CACHE1_GET, 0);
+		nvkm_wr32(device, NV03_PFIFO_CACHE1_PUT, 0);
+		nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
+		nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1);
+		nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1);
 	}
 
 	/* restore normal operation, after disabling dma mode */
-	nv_mask(fifo, NV04_PFIFO_MODE, 1 << chan->base.chid, 0);
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 1);
+	nvkm_mask(device, NV04_PFIFO_MODE, 1 << chan->base.chid, 0);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 1);
 	spin_unlock_irqrestore(&fifo->base.lock, flags);
 
 	return nvkm_fifo_channel_fini(&chan->base, suspend);
@@ -301,13 +303,14 @@ nv04_fifo_pause(struct nvkm_fifo *obj, unsigned long *pflags)
 __acquires(fifo->base.lock)
 {
 	struct nv04_fifo *fifo = container_of(obj, typeof(*fifo), base);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	unsigned long flags;
 
 	spin_lock_irqsave(&fifo->base.lock, flags);
 	*pflags = flags;
 
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 0x00000000);
-	nv_mask(fifo, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000000);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 0x00000000);
+	nvkm_mask(device, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000000);
 
 	/* in some cases the puller may be left in an inconsistent state
 	 * if you try to stop it while it's busy translating handles.
@@ -322,11 +325,11 @@ __acquires(fifo->base.lock)
 			   NV04_PFIFO_CACHE1_PULL0_HASH_BUSY, 0x00000000))
 		nv_warn(fifo, "timeout idling puller\n");
 
-	if (nv_rd32(fifo, NV04_PFIFO_CACHE1_PULL0) &
+	if (nvkm_rd32(device, NV04_PFIFO_CACHE1_PULL0) &
 			  NV04_PFIFO_CACHE1_PULL0_HASH_FAILED)
-		nv_wr32(fifo, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_CACHE_ERROR);
+		nvkm_wr32(device, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_CACHE_ERROR);
 
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_HASH, 0x00000000);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_HASH, 0x00000000);
 }
 
 void
@@ -334,10 +337,11 @@ nv04_fifo_start(struct nvkm_fifo *obj, unsigned long *pflags)
 __releases(fifo->base.lock)
 {
 	struct nv04_fifo *fifo = container_of(obj, typeof(*fifo), base);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	unsigned long flags = *pflags;
 
-	nv_mask(fifo, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000001);
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 0x00000001);
+	nvkm_mask(device, NV04_PFIFO_CACHE1_PULL0, 0x00000001, 0x00000001);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 0x00000001);
 
 	spin_unlock_irqrestore(&fifo->base.lock, flags);
 }
@@ -355,6 +359,7 @@ nv_dma_state_err(u32 state)
 static bool
 nv04_fifo_swmthd(struct nv04_fifo *fifo, u32 chid, u32 addr, u32 data)
 {
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	struct nv04_fifo_chan *chan = NULL;
 	struct nvkm_handle *bind;
 	const int subc = (addr >> 13) & 0x7;
@@ -380,13 +385,13 @@ nv04_fifo_swmthd(struct nv04_fifo *fifo, u32 chid, u32 addr, u32 data)
 			chan->subc[subc] = data;
 			handled = true;
 
-			nv_mask(fifo, NV04_PFIFO_CACHE1_ENGINE, engine, 0);
+			nvkm_mask(device, NV04_PFIFO_CACHE1_ENGINE, engine, 0);
 		}
 
 		nvkm_namedb_put(bind);
 		break;
 	default:
-		engine = nv_rd32(fifo, NV04_PFIFO_CACHE1_ENGINE);
+		engine = nvkm_rd32(device, NV04_PFIFO_CACHE1_ENGINE);
 		if (unlikely(((engine >> (subc * 4)) & 0xf) != 0))
 			break;
 
@@ -419,11 +424,11 @@ nv04_fifo_cache_error(struct nvkm_device *device,
 	ptr = (get & 0x7ff) >> 2;
 
 	if (device->card_type < NV_40) {
-		mthd = nv_rd32(fifo, NV04_PFIFO_CACHE1_METHOD(ptr));
-		data = nv_rd32(fifo, NV04_PFIFO_CACHE1_DATA(ptr));
+		mthd = nvkm_rd32(device, NV04_PFIFO_CACHE1_METHOD(ptr));
+		data = nvkm_rd32(device, NV04_PFIFO_CACHE1_DATA(ptr));
 	} else {
-		mthd = nv_rd32(fifo, NV40_PFIFO_CACHE1_METHOD(ptr));
-		data = nv_rd32(fifo, NV40_PFIFO_CACHE1_DATA(ptr));
+		mthd = nvkm_rd32(device, NV40_PFIFO_CACHE1_METHOD(ptr));
+		data = nvkm_rd32(device, NV40_PFIFO_CACHE1_DATA(ptr));
 	}
 
 	if (!nv04_fifo_swmthd(fifo, chid, mthd, data)) {
@@ -435,19 +440,19 @@ nv04_fifo_cache_error(struct nvkm_device *device,
 			 data);
 	}
 
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_DMA_PUSH, 0);
-	nv_wr32(fifo, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_CACHE_ERROR);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_DMA_PUSH, 0);
+	nvkm_wr32(device, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_CACHE_ERROR);
 
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH0,
-		nv_rd32(fifo, NV03_PFIFO_CACHE1_PUSH0) & ~1);
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_GET, get + 4);
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH0,
-		nv_rd32(fifo, NV03_PFIFO_CACHE1_PUSH0) | 1);
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_HASH, 0);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0,
+		nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH0) & ~1);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_GET, get + 4);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0,
+		nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH0) | 1);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_HASH, 0);
 
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_DMA_PUSH,
-		nv_rd32(fifo, NV04_PFIFO_CACHE1_DMA_PUSH) | 1);
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_PULL0, 1);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_DMA_PUSH,
+		nvkm_rd32(device, NV04_PFIFO_CACHE1_DMA_PUSH) | 1);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1);
 }
 
 static void
@@ -455,18 +460,18 @@ nv04_fifo_dma_pusher(struct nvkm_device *device,
 		     struct nv04_fifo *fifo, u32 chid)
 {
 	const char *client_name;
-	u32 dma_get = nv_rd32(fifo, 0x003244);
-	u32 dma_put = nv_rd32(fifo, 0x003240);
-	u32 push = nv_rd32(fifo, 0x003220);
-	u32 state = nv_rd32(fifo, 0x003228);
+	u32 dma_get = nvkm_rd32(device, 0x003244);
+	u32 dma_put = nvkm_rd32(device, 0x003240);
+	u32 push = nvkm_rd32(device, 0x003220);
+	u32 state = nvkm_rd32(device, 0x003228);
 
 	client_name = nvkm_client_name_for_fifo_chid(&fifo->base, chid);
 
 	if (device->card_type == NV_50) {
-		u32 ho_get = nv_rd32(fifo, 0x003328);
-		u32 ho_put = nv_rd32(fifo, 0x003320);
-		u32 ib_get = nv_rd32(fifo, 0x003334);
-		u32 ib_put = nv_rd32(fifo, 0x003330);
+		u32 ho_get = nvkm_rd32(device, 0x003328);
+		u32 ho_put = nvkm_rd32(device, 0x003320);
+		u32 ib_get = nvkm_rd32(device, 0x003334);
+		u32 ib_put = nvkm_rd32(device, 0x003330);
 
 		nv_error(fifo,
 			 "DMA_PUSHER - ch %d [%s] get 0x%02x%08x put 0x%02x%08x ib_get 0x%08x ib_put 0x%08x state 0x%08x (err: %s) push 0x%08x\n",
@@ -474,13 +479,13 @@ nv04_fifo_dma_pusher(struct nvkm_device *device,
 			 ib_get, ib_put, state, nv_dma_state_err(state), push);
 
 		/* METHOD_COUNT, in DMA_STATE on earlier chipsets */
-		nv_wr32(fifo, 0x003364, 0x00000000);
+		nvkm_wr32(device, 0x003364, 0x00000000);
 		if (dma_get != dma_put || ho_get != ho_put) {
-			nv_wr32(fifo, 0x003244, dma_put);
-			nv_wr32(fifo, 0x003328, ho_put);
+			nvkm_wr32(device, 0x003244, dma_put);
+			nvkm_wr32(device, 0x003328, ho_put);
 		} else
 		if (ib_get != ib_put)
-			nv_wr32(fifo, 0x003334, ib_put);
+			nvkm_wr32(device, 0x003334, ib_put);
 	} else {
 		nv_error(fifo,
 			 "DMA_PUSHER - ch %d [%s] get 0x%08x put 0x%08x state 0x%08x (err: %s) push 0x%08x\n",
@@ -488,12 +493,12 @@ nv04_fifo_dma_pusher(struct nvkm_device *device,
 			 nv_dma_state_err(state), push);
 
 		if (dma_get != dma_put)
-			nv_wr32(fifo, 0x003244, dma_put);
+			nvkm_wr32(device, 0x003244, dma_put);
 	}
 
-	nv_wr32(fifo, 0x003228, 0x00000000);
-	nv_wr32(fifo, 0x003220, 0x00000001);
-	nv_wr32(fifo, 0x002100, NV_PFIFO_INTR_DMA_PUSHER);
+	nvkm_wr32(device, 0x003228, 0x00000000);
+	nvkm_wr32(device, 0x003220, 0x00000001);
+	nvkm_wr32(device, 0x002100, NV_PFIFO_INTR_DMA_PUSHER);
 }
 
 void
@@ -501,15 +506,15 @@ nv04_fifo_intr(struct nvkm_subdev *subdev)
 {
 	struct nvkm_device *device = nv_device(subdev);
 	struct nv04_fifo *fifo = (void *)subdev;
-	u32 mask = nv_rd32(fifo, NV03_PFIFO_INTR_EN_0);
-	u32 stat = nv_rd32(fifo, NV03_PFIFO_INTR_0) & mask;
+	u32 mask = nvkm_rd32(device, NV03_PFIFO_INTR_EN_0);
+	u32 stat = nvkm_rd32(device, NV03_PFIFO_INTR_0) & mask;
 	u32 reassign, chid, get, sem;
 
-	reassign = nv_rd32(fifo, NV03_PFIFO_CACHES) & 1;
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 0);
+	reassign = nvkm_rd32(device, NV03_PFIFO_CACHES) & 1;
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 0);
 
-	chid = nv_rd32(fifo, NV03_PFIFO_CACHE1_PUSH1) & fifo->base.max;
-	get  = nv_rd32(fifo, NV03_PFIFO_CACHE1_GET);
+	chid = nvkm_rd32(device, NV03_PFIFO_CACHE1_PUSH1) & fifo->base.max;
+	get  = nvkm_rd32(device, NV03_PFIFO_CACHE1_GET);
 
 	if (stat & NV_PFIFO_INTR_CACHE_ERROR) {
 		nv04_fifo_cache_error(device, fifo, chid, get);
@@ -523,23 +528,23 @@ nv04_fifo_intr(struct nvkm_subdev *subdev)
 
 	if (stat & NV_PFIFO_INTR_SEMAPHORE) {
 		stat &= ~NV_PFIFO_INTR_SEMAPHORE;
-		nv_wr32(fifo, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_SEMAPHORE);
+		nvkm_wr32(device, NV03_PFIFO_INTR_0, NV_PFIFO_INTR_SEMAPHORE);
 
-		sem = nv_rd32(fifo, NV10_PFIFO_CACHE1_SEMAPHORE);
-		nv_wr32(fifo, NV10_PFIFO_CACHE1_SEMAPHORE, sem | 0x1);
+		sem = nvkm_rd32(device, NV10_PFIFO_CACHE1_SEMAPHORE);
+		nvkm_wr32(device, NV10_PFIFO_CACHE1_SEMAPHORE, sem | 0x1);
 
-		nv_wr32(fifo, NV03_PFIFO_CACHE1_GET, get + 4);
-		nv_wr32(fifo, NV04_PFIFO_CACHE1_PULL0, 1);
+		nvkm_wr32(device, NV03_PFIFO_CACHE1_GET, get + 4);
+		nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1);
 	}
 
 	if (device->card_type == NV_50) {
 		if (stat & 0x00000010) {
 			stat &= ~0x00000010;
-			nv_wr32(fifo, 0x002100, 0x00000010);
+			nvkm_wr32(device, 0x002100, 0x00000010);
 		}
 
 		if (stat & 0x40000000) {
-			nv_wr32(fifo, 0x002100, 0x40000000);
+			nvkm_wr32(device, 0x002100, 0x40000000);
 			nvkm_fifo_uevent(&fifo->base);
 			stat &= ~0x40000000;
 		}
@@ -547,11 +552,11 @@ nv04_fifo_intr(struct nvkm_subdev *subdev)
 
 	if (stat) {
 		nv_warn(fifo, "unknown intr 0x%08x\n", stat);
-		nv_mask(fifo, NV03_PFIFO_INTR_EN_0, stat, 0x00000000);
-		nv_wr32(fifo, NV03_PFIFO_INTR_0, stat);
+		nvkm_mask(device, NV03_PFIFO_INTR_EN_0, stat, 0x00000000);
+		nvkm_wr32(device, NV03_PFIFO_INTR_0, stat);
 	}
 
-	nv_wr32(fifo, NV03_PFIFO_CACHES, reassign);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, reassign);
 }
 
 static int
@@ -596,29 +601,30 @@ int
 nv04_fifo_init(struct nvkm_object *object)
 {
 	struct nv04_fifo *fifo = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	int ret;
 
 	ret = nvkm_fifo_init(&fifo->base);
 	if (ret)
 		return ret;
 
-	nv_wr32(fifo, NV04_PFIFO_DELAY_0, 0x000000ff);
-	nv_wr32(fifo, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff);
+	nvkm_wr32(device, NV04_PFIFO_DELAY_0, 0x000000ff);
+	nvkm_wr32(device, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff);
 
-	nv_wr32(fifo, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+	nvkm_wr32(device, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
 				       ((fifo->ramht->bits - 9) << 16) |
 				        (fifo->ramht->gpuobj.addr >> 8));
-	nv_wr32(fifo, NV03_PFIFO_RAMRO, fifo->ramro->addr >> 8);
-	nv_wr32(fifo, NV03_PFIFO_RAMFC, fifo->ramfc->addr >> 8);
+	nvkm_wr32(device, NV03_PFIFO_RAMRO, fifo->ramro->addr >> 8);
+	nvkm_wr32(device, NV03_PFIFO_RAMFC, fifo->ramfc->addr >> 8);
 
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
 
-	nv_wr32(fifo, NV03_PFIFO_INTR_0, 0xffffffff);
-	nv_wr32(fifo, NV03_PFIFO_INTR_EN_0, 0xffffffff);
+	nvkm_wr32(device, NV03_PFIFO_INTR_0, 0xffffffff);
+	nvkm_wr32(device, NV03_PFIFO_INTR_EN_0, 0xffffffff);
 
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH0, 1);
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_PULL0, 1);
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 1);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 1);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c
index 6f8787fbacc0..6a57df6f78c4 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv17.c
@@ -177,29 +177,30 @@ static int
 nv17_fifo_init(struct nvkm_object *object)
 {
 	struct nv04_fifo *fifo = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	int ret;
 
 	ret = nvkm_fifo_init(&fifo->base);
 	if (ret)
 		return ret;
 
-	nv_wr32(fifo, NV04_PFIFO_DELAY_0, 0x000000ff);
-	nv_wr32(fifo, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff);
+	nvkm_wr32(device, NV04_PFIFO_DELAY_0, 0x000000ff);
+	nvkm_wr32(device, NV04_PFIFO_DMA_TIMESLICE, 0x0101ffff);
 
-	nv_wr32(fifo, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+	nvkm_wr32(device, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
 				       ((fifo->ramht->bits - 9) << 16) |
 				        (fifo->ramht->gpuobj.addr >> 8));
-	nv_wr32(fifo, NV03_PFIFO_RAMRO, fifo->ramro->addr >> 8);
-	nv_wr32(fifo, NV03_PFIFO_RAMFC, fifo->ramfc->addr >> 8 | 0x00010000);
+	nvkm_wr32(device, NV03_PFIFO_RAMRO, fifo->ramro->addr >> 8);
+	nvkm_wr32(device, NV03_PFIFO_RAMFC, fifo->ramfc->addr >> 8 | 0x00010000);
 
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
 
-	nv_wr32(fifo, NV03_PFIFO_INTR_0, 0xffffffff);
-	nv_wr32(fifo, NV03_PFIFO_INTR_EN_0, 0xffffffff);
+	nvkm_wr32(device, NV03_PFIFO_INTR_0, 0xffffffff);
+	nvkm_wr32(device, NV03_PFIFO_INTR_EN_0, 0xffffffff);
 
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH0, 1);
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_PULL0, 1);
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 1);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 1);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c
index 4c1ed3f29e6b..7f9eb1018814 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv40.c
@@ -105,6 +105,7 @@ nv40_fifo_context_attach(struct nvkm_object *parent, struct nvkm_object *engctx)
 {
 	struct nv04_fifo *fifo = (void *)parent->engine;
 	struct nv04_fifo_chan *chan = (void *)parent;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	unsigned long flags;
 	u32 reg, ctx;
 
@@ -125,13 +126,13 @@ nv40_fifo_context_attach(struct nvkm_object *parent, struct nvkm_object *engctx)
 
 	spin_lock_irqsave(&fifo->base.lock, flags);
 	nv_engctx(engctx)->addr = nv_gpuobj(engctx)->addr >> 4;
-	nv_mask(fifo, 0x002500, 0x00000001, 0x00000000);
+	nvkm_mask(device, 0x002500, 0x00000001, 0x00000000);
 
-	if ((nv_rd32(fifo, 0x003204) & fifo->base.max) == chan->base.chid)
-		nv_wr32(fifo, reg, nv_engctx(engctx)->addr);
+	if ((nvkm_rd32(device, 0x003204) & fifo->base.max) == chan->base.chid)
+		nvkm_wr32(device, reg, nv_engctx(engctx)->addr);
 	nv_wo32(fifo->ramfc, chan->ramfc + ctx, nv_engctx(engctx)->addr);
 
-	nv_mask(fifo, 0x002500, 0x00000001, 0x00000001);
+	nvkm_mask(device, 0x002500, 0x00000001, 0x00000001);
 	spin_unlock_irqrestore(&fifo->base.lock, flags);
 	return 0;
 }
@@ -142,6 +143,7 @@ nv40_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 {
 	struct nv04_fifo *fifo = (void *)parent->engine;
 	struct nv04_fifo_chan *chan = (void *)parent;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	unsigned long flags;
 	u32 reg, ctx;
 
@@ -161,13 +163,13 @@ nv40_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 	}
 
 	spin_lock_irqsave(&fifo->base.lock, flags);
-	nv_mask(fifo, 0x002500, 0x00000001, 0x00000000);
+	nvkm_mask(device, 0x002500, 0x00000001, 0x00000000);
 
-	if ((nv_rd32(fifo, 0x003204) & fifo->base.max) == chan->base.chid)
-		nv_wr32(fifo, reg, 0x00000000);
+	if ((nvkm_rd32(device, 0x003204) & fifo->base.max) == chan->base.chid)
+		nvkm_wr32(device, reg, 0x00000000);
 	nv_wo32(fifo->ramfc, chan->ramfc + ctx, 0x00000000);
 
-	nv_mask(fifo, 0x002500, 0x00000001, 0x00000001);
+	nvkm_mask(device, 0x002500, 0x00000001, 0x00000001);
 	spin_unlock_irqrestore(&fifo->base.lock, flags);
 	return 0;
 }
@@ -295,51 +297,52 @@ static int
 nv40_fifo_init(struct nvkm_object *object)
 {
 	struct nv04_fifo *fifo = (void *)object;
-	struct nvkm_fb *fb = nvkm_fb(object);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	struct nvkm_fb *fb = device->fb;
 	int ret;
 
 	ret = nvkm_fifo_init(&fifo->base);
 	if (ret)
 		return ret;
 
-	nv_wr32(fifo, 0x002040, 0x000000ff);
-	nv_wr32(fifo, 0x002044, 0x2101ffff);
-	nv_wr32(fifo, 0x002058, 0x00000001);
+	nvkm_wr32(device, 0x002040, 0x000000ff);
+	nvkm_wr32(device, 0x002044, 0x2101ffff);
+	nvkm_wr32(device, 0x002058, 0x00000001);
 
-	nv_wr32(fifo, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
+	nvkm_wr32(device, NV03_PFIFO_RAMHT, (0x03 << 24) /* search 128 */ |
 				       ((fifo->ramht->bits - 9) << 16) |
 				        (fifo->ramht->gpuobj.addr >> 8));
-	nv_wr32(fifo, NV03_PFIFO_RAMRO, fifo->ramro->addr >> 8);
+	nvkm_wr32(device, NV03_PFIFO_RAMRO, fifo->ramro->addr >> 8);
 
 	switch (nv_device(fifo)->chipset) {
 	case 0x47:
 	case 0x49:
 	case 0x4b:
-		nv_wr32(fifo, 0x002230, 0x00000001);
+		nvkm_wr32(device, 0x002230, 0x00000001);
 	case 0x40:
 	case 0x41:
 	case 0x42:
 	case 0x43:
 	case 0x45:
 	case 0x48:
-		nv_wr32(fifo, 0x002220, 0x00030002);
+		nvkm_wr32(device, 0x002220, 0x00030002);
 		break;
 	default:
-		nv_wr32(fifo, 0x002230, 0x00000000);
-		nv_wr32(fifo, 0x002220, ((fb->ram->size - 512 * 1024 +
+		nvkm_wr32(device, 0x002230, 0x00000000);
+		nvkm_wr32(device, 0x002220, ((fb->ram->size - 512 * 1024 +
 					 fifo->ramfc->addr) >> 16) |
 					0x00030000);
 		break;
 	}
 
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH1, fifo->base.max);
 
-	nv_wr32(fifo, NV03_PFIFO_INTR_0, 0xffffffff);
-	nv_wr32(fifo, NV03_PFIFO_INTR_EN_0, 0xffffffff);
+	nvkm_wr32(device, NV03_PFIFO_INTR_0, 0xffffffff);
+	nvkm_wr32(device, NV03_PFIFO_INTR_EN_0, 0xffffffff);
 
-	nv_wr32(fifo, NV03_PFIFO_CACHE1_PUSH0, 1);
-	nv_wr32(fifo, NV04_PFIFO_CACHE1_PULL0, 1);
-	nv_wr32(fifo, NV03_PFIFO_CACHES, 1);
+	nvkm_wr32(device, NV03_PFIFO_CACHE1_PUSH0, 1);
+	nvkm_wr32(device, NV04_PFIFO_CACHE1_PULL0, 1);
+	nvkm_wr32(device, NV03_PFIFO_CACHES, 1);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c
index a36f7efc4658..3a868f9c0a39 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/nv50.c
@@ -41,7 +41,8 @@
 static void
 nv50_fifo_playlist_update_locked(struct nv50_fifo *fifo)
 {
-	struct nvkm_bar *bar = nvkm_bar(fifo);
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	struct nvkm_bar *bar = device->bar;
 	struct nvkm_gpuobj *cur;
 	int i, p;
 
@@ -49,15 +50,15 @@ nv50_fifo_playlist_update_locked(struct nv50_fifo *fifo)
 	fifo->cur_playlist = !fifo->cur_playlist;
 
 	for (i = fifo->base.min, p = 0; i < fifo->base.max; i++) {
-		if (nv_rd32(fifo, 0x002600 + (i * 4)) & 0x80000000)
+		if (nvkm_rd32(device, 0x002600 + (i * 4)) & 0x80000000)
 			nv_wo32(cur, p++ * 4, i);
 	}
 
 	bar->flush(bar);
 
-	nv_wr32(fifo, 0x0032f4, cur->addr >> 12);
-	nv_wr32(fifo, 0x0032ec, p);
-	nv_wr32(fifo, 0x002500, 0x00000101);
+	nvkm_wr32(device, 0x0032f4, cur->addr >> 12);
+	nvkm_wr32(device, 0x0032ec, p);
+	nvkm_wr32(device, 0x002500, 0x00000101);
 }
 
 void
@@ -102,10 +103,11 @@ static int
 nv50_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 			 struct nvkm_object *object)
 {
-	struct nvkm_bar *bar = nvkm_bar(parent);
 	struct nv50_fifo *fifo = (void *)parent->engine;
 	struct nv50_fifo_base *base = (void *)parent->parent;
 	struct nv50_fifo_chan *chan = (void *)parent;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
+	struct nvkm_bar *bar = device->bar;
 	u32 addr, me;
 	int ret = 0;
 
@@ -129,17 +131,17 @@ nv50_fifo_context_detach(struct nvkm_object *parent, bool suspend,
 	 * there's also a "ignore these engines" bitmask reg we can use
 	 * if we hit the issue there..
 	 */
-	me = nv_mask(fifo, 0x00b860, 0x00000001, 0x00000001);
+	me = nvkm_mask(device, 0x00b860, 0x00000001, 0x00000001);
 
 	/* do the kickoff... */
-	nv_wr32(fifo, 0x0032fc, nv_gpuobj(base)->addr >> 12);
+	nvkm_wr32(device, 0x0032fc, nv_gpuobj(base)->addr >> 12);
 	if (!nv_wait_ne(fifo, 0x0032fc, 0xffffffff, 0xffffffff)) {
 		nv_error(fifo, "channel %d [%s] unload timeout\n",
 			 chan->base.chid, nvkm_client_name(chan));
 		if (suspend)
 			ret = -EBUSY;
 	}
-	nv_wr32(fifo, 0x00b860, me);
+	nvkm_wr32(device, 0x00b860, me);
 
 	if (ret == 0) {
 		nv_wo32(base->eng, addr + 0x00, 0x00000000);
@@ -324,6 +326,7 @@ nv50_fifo_chan_init(struct nvkm_object *object)
 	struct nv50_fifo_base *base = (void *)object->parent;
 	struct nv50_fifo_chan *chan = (void *)object;
 	struct nvkm_gpuobj *ramfc = base->ramfc;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 	int ret;
 
@@ -331,7 +334,7 @@ nv50_fifo_chan_init(struct nvkm_object *object)
 	if (ret)
 		return ret;
 
-	nv_wr32(fifo, 0x002600 + (chid * 4), 0x80000000 | ramfc->addr >> 12);
+	nvkm_wr32(device, 0x002600 + (chid * 4), 0x80000000 | ramfc->addr >> 12);
 	nv50_fifo_playlist_update(fifo);
 	return 0;
 }
@@ -341,12 +344,13 @@ nv50_fifo_chan_fini(struct nvkm_object *object, bool suspend)
 {
 	struct nv50_fifo *fifo = (void *)object->engine;
 	struct nv50_fifo_chan *chan = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	u32 chid = chan->base.chid;
 
 	/* remove channel from playlist, fifo will unload context */
-	nv_mask(fifo, 0x002600 + (chid * 4), 0x80000000, 0x00000000);
+	nvkm_mask(device, 0x002600 + (chid * 4), 0x80000000, 0x00000000);
 	nv50_fifo_playlist_update(fifo);
-	nv_wr32(fifo, 0x002600 + (chid * 4), 0x00000000);
+	nvkm_wr32(device, 0x002600 + (chid * 4), 0x00000000);
 
 	return nvkm_fifo_channel_fini(&chan->base, suspend);
 }
@@ -498,27 +502,28 @@ int
 nv50_fifo_init(struct nvkm_object *object)
 {
 	struct nv50_fifo *fifo = (void *)object;
+	struct nvkm_device *device = fifo->base.engine.subdev.device;
 	int ret, i;
 
 	ret = nvkm_fifo_init(&fifo->base);
 	if (ret)
 		return ret;
 
-	nv_mask(fifo, 0x000200, 0x00000100, 0x00000000);
-	nv_mask(fifo, 0x000200, 0x00000100, 0x00000100);
-	nv_wr32(fifo, 0x00250c, 0x6f3cfc34);
-	nv_wr32(fifo, 0x002044, 0x01003fff);
+	nvkm_mask(device, 0x000200, 0x00000100, 0x00000000);
+	nvkm_mask(device, 0x000200, 0x00000100, 0x00000100);
+	nvkm_wr32(device, 0x00250c, 0x6f3cfc34);
+	nvkm_wr32(device, 0x002044, 0x01003fff);
 
-	nv_wr32(fifo, 0x002100, 0xffffffff);
-	nv_wr32(fifo, 0x002140, 0xbfffffff);
+	nvkm_wr32(device, 0x002100, 0xffffffff);
+	nvkm_wr32(device, 0x002140, 0xbfffffff);
 
 	for (i = 0; i < 128; i++)
-		nv_wr32(fifo, 0x002600 + (i * 4), 0x00000000);
+		nvkm_wr32(device, 0x002600 + (i * 4), 0x00000000);
 	nv50_fifo_playlist_update_locked(fifo);
 
-	nv_wr32(fifo, 0x003200, 0x00000001);
-	nv_wr32(fifo, 0x003250, 0x00000001);
-	nv_wr32(fifo, 0x002500, 0x00000001);
+	nvkm_wr32(device, 0x003200, 0x00000001);
+	nvkm_wr32(device, 0x003250, 0x00000001);
+	nvkm_wr32(device, 0x002500, 0x00000001);
 	return 0;
 }