drm/nv41/vm: fix and enable use of "real" pciegart

Hopefully fixed the tlb flush timeout issue.  Was able to observe this
condition occur occasionally, and it appears the binary driver doesn't
wait on the old condition either..

Should give 39-bit DMA addressing on the relevant chipsets.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drivers/gpu/drm/nouveau/core/subdev/device/nv40.c

@@ -78,7 +78,7 @@ nv40_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_SUBDEV_TIMER  ] = &nv04_timer_oclass;
 		device->oclass[NVDEV_SUBDEV_FB     ] = &nv40_fb_oclass;
 		device->oclass[NVDEV_SUBDEV_INSTMEM] = &nv40_instmem_oclass;
-		device->oclass[NVDEV_SUBDEV_VM     ] = &nv04_vmmgr_oclass;
+		device->oclass[NVDEV_SUBDEV_VM     ] = &nv41_vmmgr_oclass;
 		device->oclass[NVDEV_ENGINE_DMAOBJ ] = &nv04_dmaeng_oclass;
 		device->oclass[NVDEV_ENGINE_FIFO   ] = &nv40_fifo_oclass;
 		device->oclass[NVDEV_ENGINE_SW     ] = &nv10_software_oclass;
@@ -98,7 +98,7 @@ nv40_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_SUBDEV_TIMER  ] = &nv04_timer_oclass;
 		device->oclass[NVDEV_SUBDEV_FB     ] = &nv40_fb_oclass;
 		device->oclass[NVDEV_SUBDEV_INSTMEM] = &nv40_instmem_oclass;
-		device->oclass[NVDEV_SUBDEV_VM     ] = &nv04_vmmgr_oclass;
+		device->oclass[NVDEV_SUBDEV_VM     ] = &nv41_vmmgr_oclass;
 		device->oclass[NVDEV_ENGINE_DMAOBJ ] = &nv04_dmaeng_oclass;
 		device->oclass[NVDEV_ENGINE_FIFO   ] = &nv40_fifo_oclass;
 		device->oclass[NVDEV_ENGINE_SW     ] = &nv10_software_oclass;
@@ -118,7 +118,7 @@ nv40_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_SUBDEV_TIMER  ] = &nv04_timer_oclass;
 		device->oclass[NVDEV_SUBDEV_FB     ] = &nv40_fb_oclass;
 		device->oclass[NVDEV_SUBDEV_INSTMEM] = &nv40_instmem_oclass;
-		device->oclass[NVDEV_SUBDEV_VM     ] = &nv04_vmmgr_oclass;
+		device->oclass[NVDEV_SUBDEV_VM     ] = &nv41_vmmgr_oclass;
 		device->oclass[NVDEV_ENGINE_DMAOBJ ] = &nv04_dmaeng_oclass;
 		device->oclass[NVDEV_ENGINE_FIFO   ] = &nv40_fifo_oclass;
 		device->oclass[NVDEV_ENGINE_SW     ] = &nv10_software_oclass;
@@ -158,7 +158,7 @@ nv40_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_SUBDEV_TIMER  ] = &nv04_timer_oclass;
 		device->oclass[NVDEV_SUBDEV_FB     ] = &nv40_fb_oclass;
 		device->oclass[NVDEV_SUBDEV_INSTMEM] = &nv40_instmem_oclass;
-		device->oclass[NVDEV_SUBDEV_VM     ] = &nv04_vmmgr_oclass;
+		device->oclass[NVDEV_SUBDEV_VM     ] = &nv41_vmmgr_oclass;
 		device->oclass[NVDEV_ENGINE_DMAOBJ ] = &nv04_dmaeng_oclass;
 		device->oclass[NVDEV_ENGINE_FIFO   ] = &nv40_fifo_oclass;
 		device->oclass[NVDEV_ENGINE_SW     ] = &nv10_software_oclass;
@@ -178,7 +178,7 @@ nv40_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_SUBDEV_TIMER  ] = &nv04_timer_oclass;
 		device->oclass[NVDEV_SUBDEV_FB     ] = &nv40_fb_oclass;
 		device->oclass[NVDEV_SUBDEV_INSTMEM] = &nv40_instmem_oclass;
-		device->oclass[NVDEV_SUBDEV_VM     ] = &nv04_vmmgr_oclass;
+		device->oclass[NVDEV_SUBDEV_VM     ] = &nv41_vmmgr_oclass;
 		device->oclass[NVDEV_ENGINE_DMAOBJ ] = &nv04_dmaeng_oclass;
 		device->oclass[NVDEV_ENGINE_FIFO   ] = &nv40_fifo_oclass;
 		device->oclass[NVDEV_ENGINE_SW     ] = &nv10_software_oclass;
@@ -198,7 +198,7 @@ nv40_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_SUBDEV_TIMER  ] = &nv04_timer_oclass;
 		device->oclass[NVDEV_SUBDEV_FB     ] = &nv40_fb_oclass;
 		device->oclass[NVDEV_SUBDEV_INSTMEM] = &nv40_instmem_oclass;
-		device->oclass[NVDEV_SUBDEV_VM     ] = &nv04_vmmgr_oclass;
+		device->oclass[NVDEV_SUBDEV_VM     ] = &nv41_vmmgr_oclass;
 		device->oclass[NVDEV_ENGINE_DMAOBJ ] = &nv04_dmaeng_oclass;
 		device->oclass[NVDEV_ENGINE_FIFO   ] = &nv40_fifo_oclass;
 		device->oclass[NVDEV_ENGINE_SW     ] = &nv10_software_oclass;

drivers/gpu/drm/nouveau/core/subdev/vm/nv41.c

@@ -23,6 +23,7 @@
  */
 
 #include <core/gpuobj.h>
+#include <core/option.h>
 
 #include <subdev/timer.h>
 #include <subdev/vm.h>
@@ -70,7 +71,7 @@ nv41_vm_flush(struct nouveau_vm *vm)
 
 	mutex_lock(&nv_subdev(priv)->mutex);
 	nv_wr32(priv, 0x100810, 0x00000022);
-	if (!nv_wait(priv, 0x100810, 0x00000100, 0x00000100)) {
+	if (!nv_wait(priv, 0x100810, 0x00000020, 0x00000020)) {
 		nv_warn(priv, "flush timeout, 0x%08x\n",
 			nv_rd32(priv, 0x100810));
 	}
@@ -87,9 +88,15 @@ nv41_vmmgr_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
 		struct nouveau_oclass *oclass, void *data, u32 size,
 		struct nouveau_object **pobject)
 {
+	struct nouveau_device *device = nv_device(parent);
 	struct nv04_vmmgr_priv *priv;
 	int ret;
 
+	if (!nouveau_boolopt(device->cfgopt, "NvPCIE", true)) {
+		return nouveau_object_ctor(parent, engine, &nv04_vmmgr_oclass,
+					   data, size, pobject);
+	}
+
 	ret = nouveau_vmmgr_create(parent, engine, oclass, "PCIEGART",
 				   "pciegart", &priv);
 	*pobject = nv_object(priv);