Merge branch 'drm-nouveau-next' of git://anongit.freedesktop.org/git/nouveau/linux-2.6 into drm-next

- Page flipping fixes, with support for syncing them to vblank (finally...).
- Misc other general fixes

* 'drm-nouveau-next' of git://anongit.freedesktop.org/git/nouveau/linux-2.6:
  drm/nouveau: do not map evicted vram buffers in nouveau_bo_vma_add
  drm/nvc0-/gr: shift wrapping bug in nvc0_grctx_generate_r406800
  drm/nouveau/pwr: fix missing mutex unlock in a failure path
  drm/nv40/therm: fix slowing down fan when pstate undefined
  drm/nv11-: synchronise flips to vblank, unless async flip requested
  drm/nvc0-: remove nasty fifo swmthd hack for flip completion method
  drm/nv10-: we no longer need to create nvsw object on user channels
  drm/nouveau: always queue flips relative to kernel channel activity
  drm/nouveau: there is no need to reserve/fence the new fb when flipping
  drm/nouveau: when bailing out of a pushbuf ioctl, do not remove previous fence
  drm/nouveau: allow nouveau_fence_ref() to be a noop
  drm/nvc8/mc: msi rearm is via the nvc0 method

drivers/gpu/drm/nouveau/core/engine/device/nvc0.c

@@ -256,7 +256,7 @@ nvc0_identify(struct nouveau_device *device)
 		device->oclass[NVDEV_SUBDEV_THERM  ] = &nva3_therm_oclass;
 		device->oclass[NVDEV_SUBDEV_MXM    ] = &nv50_mxm_oclass;
 		device->oclass[NVDEV_SUBDEV_DEVINIT] = &nvc0_devinit_oclass;
-		device->oclass[NVDEV_SUBDEV_MC     ] =  nvc3_mc_oclass;
+		device->oclass[NVDEV_SUBDEV_MC     ] =  nvc0_mc_oclass;
 		device->oclass[NVDEV_SUBDEV_BUS    ] =  nvc0_bus_oclass;
 		device->oclass[NVDEV_SUBDEV_TIMER  ] = &nv04_timer_oclass;
 		device->oclass[NVDEV_SUBDEV_FB     ] =  nvc0_fb_oclass;

drivers/gpu/drm/nouveau/core/engine/fifo/nvc0.c

@@ -494,13 +494,6 @@ nvc0_fifo_isr_subfifo_intr(struct nvc0_fifo_priv *priv, int unit)
 	u32 mthd = (addr & 0x00003ffc);
 	u32 show = stat;
 
-	if (stat & 0x00200000) {
-		if (mthd == 0x0054) {
-			if (!nvc0_fifo_swmthd(priv, chid, 0x0500, 0x00000000))
-				show &= ~0x00200000;
-		}
-	}
-
 	if (stat & 0x00800000) {
 		if (!nvc0_fifo_swmthd(priv, chid, mthd, data))
 			show &= ~0x00800000;

drivers/gpu/drm/nouveau/core/engine/fifo/nve0.c

@@ -481,13 +481,6 @@ nve0_fifo_isr_subfifo_intr(struct nve0_fifo_priv *priv, int unit)
 	u32 mthd = (addr & 0x00003ffc);
 	u32 show = stat;
 
-	if (stat & 0x00200000) {
-		if (mthd == 0x0054) {
-			if (!nve0_fifo_swmthd(priv, chid, 0x0500, 0x00000000))
-				show &= ~0x00200000;
-		}
-	}
-
 	if (stat & 0x00800000) {
 		if (!nve0_fifo_swmthd(priv, chid, mthd, data))
 			show &= ~0x00800000;

drivers/gpu/drm/nouveau/core/engine/graph/ctxnvc0.c

@@ -1039,7 +1039,7 @@ nvc0_grctx_generate_r406800(struct nvc0_graph_priv *priv)
 			} while (!tpcnr[gpc]);
 			tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--;
 
-			tpc_set |= 1 << ((gpc * 8) + tpc);
+			tpc_set |= 1ULL << ((gpc * 8) + tpc);
 		}
 
 		nv_wr32(priv, 0x406800 + (i * 0x20), lower_32_bits(tpc_set));

drivers/gpu/drm/nouveau/core/subdev/pwr/base.c

@@ -32,6 +32,11 @@ nouveau_pwr_send(struct nouveau_pwr *ppwr, u32 reply[2],
 	struct nouveau_subdev *subdev = nv_subdev(ppwr);
 	u32 addr;
 
+	/* wait for a free slot in the fifo */
+	addr  = nv_rd32(ppwr, 0x10a4a0);
+	if (!nv_wait_ne(ppwr, 0x10a4b0, 0xffffffff, addr ^ 8))
+		return -EBUSY;
+
 	/* we currently only support a single process at a time waiting
 	 * on a synchronous reply, take the PPWR mutex and tell the
 	 * receive handler what we're waiting for
@@ -42,11 +47,6 @@ nouveau_pwr_send(struct nouveau_pwr *ppwr, u32 reply[2],
 		ppwr->recv.process = process;
 	}
 
-	/* wait for a free slot in the fifo */
-	addr  = nv_rd32(ppwr, 0x10a4a0);
-	if (!nv_wait_ne(ppwr, 0x10a4b0, 0xffffffff, addr ^ 8))
-		return -EBUSY;
-
 	/* acquire data segment access */
 	do {
 		nv_wr32(ppwr, 0x10a580, 0x00000001);

drivers/gpu/drm/nouveau/core/subdev/therm/base.c

@@ -117,7 +117,8 @@ nouveau_therm_update(struct nouveau_therm *therm, int mode)
 		    priv->fan->bios.linear_max_temp) {
 			duty = nouveau_therm_update_linear(therm);
 		} else {
-			duty = priv->cstate;
+			if (priv->cstate)
+				duty = priv->cstate;
 			poll = false;
 		}
 		immd = false;

drivers/gpu/drm/nouveau/nouveau_abi16.c

@@ -298,7 +298,7 @@ nouveau_abi16_ioctl_channel_alloc(ABI16_IOCTL_ARGS)
 	else
 		init->pushbuf_domains = NOUVEAU_GEM_DOMAIN_GART;
 
-	if (device->card_type < NV_C0) {
+	if (device->card_type < NV_10) {
 		init->subchan[0].handle = 0x00000000;
 		init->subchan[0].grclass = 0x0000;
 		init->subchan[1].handle = NvSw;

drivers/gpu/drm/nouveau/nouveau_bo.c

@@ -98,12 +98,7 @@ nv10_bo_put_tile_region(struct drm_device *dev, struct nouveau_drm_tile *tile,
 
 	if (tile) {
 		spin_lock(&drm->tile.lock);
-		if (fence) {
-			/* Mark it as pending. */
-			tile->fence = fence;
-			nouveau_fence_ref(fence);
-		}
-
+		tile->fence = nouveau_fence_ref(fence);
 		tile->used = false;
 		spin_unlock(&drm->tile.lock);
 	}
@@ -1462,14 +1457,12 @@ nouveau_ttm_tt_unpopulate(struct ttm_tt *ttm)
 void
 nouveau_bo_fence(struct nouveau_bo *nvbo, struct nouveau_fence *fence)
 {
+	struct nouveau_fence *new_fence = nouveau_fence_ref(fence);
 	struct nouveau_fence *old_fence = NULL;
 
-	if (likely(fence))
-		nouveau_fence_ref(fence);
-
 	spin_lock(&nvbo->bo.bdev->fence_lock);
 	old_fence = nvbo->bo.sync_obj;
-	nvbo->bo.sync_obj = fence;
+	nvbo->bo.sync_obj = new_fence;
 	spin_unlock(&nvbo->bo.bdev->fence_lock);
 
 	nouveau_fence_unref(&old_fence);
@@ -1552,7 +1545,8 @@ nouveau_bo_vma_add(struct nouveau_bo *nvbo, struct nouveau_vm *vm,
 
 	if (nvbo->bo.mem.mem_type == TTM_PL_VRAM)
 		nouveau_vm_map(vma, nvbo->bo.mem.mm_node);
-	else if (nvbo->bo.mem.mem_type == TTM_PL_TT) {
+	else if (nvbo->bo.mem.mem_type == TTM_PL_TT &&
+		 nvbo->page_shift == vma->vm->vmm->spg_shift) {
 		if (node->sg)
 			nouveau_vm_map_sg_table(vma, 0, size, node);
 		else

drivers/gpu/drm/nouveau/nouveau_chan.c

@@ -346,22 +346,17 @@ nouveau_channel_init(struct nouveau_channel *chan, u32 vram, u32 gart)
 	for (i = 0; i < NOUVEAU_DMA_SKIPS; i++)
 		OUT_RING(chan, 0x00000000);
 
-	/* allocate software object class (used for fences on <= nv05, and
-	 * to signal flip completion), bind it to a subchannel.
-	 */
-	if ((device->card_type < NV_E0) || gart /* nve0: want_nvsw */) {
+	/* allocate software object class (used for fences on <= nv05) */
+	if (device->card_type < NV_10) {
 		ret = nouveau_object_new(nv_object(client), chan->handle,
-					 NvSw, nouveau_abi16_swclass(chan->drm),
-					 NULL, 0, &object);
+					 NvSw, 0x006e, NULL, 0, &object);
 		if (ret)
 			return ret;
 
 		swch = (void *)object->parent;
 		swch->flip = nouveau_flip_complete;
 		swch->flip_data = chan;
-	}
 
-	if (device->card_type < NV_C0) {
 		ret = RING_SPACE(chan, 2);
 		if (ret)
 			return ret;

drivers/gpu/drm/nouveau/nouveau_display.c

@@ -26,7 +26,6 @@
 
 #include <drm/drmP.h>
 #include <drm/drm_crtc_helper.h>
-#include <drm/ttm/ttm_execbuf_util.h>
 
 #include "nouveau_fbcon.h"
 #include "dispnv04/hw.h"
@@ -399,6 +398,11 @@ nouveau_display_create(struct drm_device *dev)
 	dev->mode_config.preferred_depth = 24;
 	dev->mode_config.prefer_shadow = 1;
 
+	if (nv_device(drm->device)->chipset < 0x11)
+		dev->mode_config.async_page_flip = false;
+	else
+		dev->mode_config.async_page_flip = true;
+
 	drm_kms_helper_poll_init(dev);
 	drm_kms_helper_poll_disable(dev);
 
@@ -555,19 +559,15 @@ nouveau_page_flip_emit(struct nouveau_channel *chan,
 		goto fail;
 
 	/* Emit the pageflip */
-	ret = RING_SPACE(chan, 3);
+	ret = RING_SPACE(chan, 2);
 	if (ret)
 		goto fail;
 
-	if (nv_device(drm->device)->card_type < NV_C0) {
+	if (nv_device(drm->device)->card_type < NV_C0)
 		BEGIN_NV04(chan, NvSubSw, NV_SW_PAGE_FLIP, 1);
-		OUT_RING  (chan, 0x00000000);
-		OUT_RING  (chan, 0x00000000);
-	} else {
-		BEGIN_NVC0(chan, 0, NV10_SUBCHAN_REF_CNT, 1);
-		OUT_RING  (chan, 0);
-		BEGIN_IMC0(chan, 0, NVSW_SUBCHAN_PAGE_FLIP, 0x0000);
-	}
+	else
+		BEGIN_NVC0(chan, FermiSw, NV_SW_PAGE_FLIP, 1);
+	OUT_RING  (chan, 0x00000000);
 	FIRE_RING (chan);
 
 	ret = nouveau_fence_new(chan, false, pfence);
@@ -584,22 +584,16 @@ fail:
 
 int
 nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
-		       struct drm_pending_vblank_event *event,
-		       uint32_t page_flip_flags)
+		       struct drm_pending_vblank_event *event, u32 flags)
 {
+	const int swap_interval = (flags & DRM_MODE_PAGE_FLIP_ASYNC) ? 0 : 1;
 	struct drm_device *dev = crtc->dev;
 	struct nouveau_drm *drm = nouveau_drm(dev);
 	struct nouveau_bo *old_bo = nouveau_framebuffer(crtc->fb)->nvbo;
 	struct nouveau_bo *new_bo = nouveau_framebuffer(fb)->nvbo;
 	struct nouveau_page_flip_state *s;
-	struct nouveau_channel *chan = NULL;
+	struct nouveau_channel *chan = drm->channel;
 	struct nouveau_fence *fence;
-	struct ttm_validate_buffer resv[2] = {
-		{ .bo = &old_bo->bo },
-		{ .bo = &new_bo->bo },
-	};
-	struct ww_acquire_ctx ticket;
-	LIST_HEAD(res);
 	int ret;
 
 	if (!drm->channel)
@@ -609,26 +603,22 @@ nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 	if (!s)
 		return -ENOMEM;
 
-	/* Choose the channel the flip will be handled in */
-	spin_lock(&old_bo->bo.bdev->fence_lock);
-	fence = new_bo->bo.sync_obj;
-	if (fence)
-		chan = fence->channel;
-	if (!chan)
-		chan = drm->channel;
-	spin_unlock(&old_bo->bo.bdev->fence_lock);
+	/* synchronise rendering channel with the kernel's channel */
+	spin_lock(&new_bo->bo.bdev->fence_lock);
+	fence = nouveau_fence_ref(new_bo->bo.sync_obj);
+	spin_unlock(&new_bo->bo.bdev->fence_lock);
+	ret = nouveau_fence_sync(fence, chan);
+	if (ret)
+		return ret;
 
 	if (new_bo != old_bo) {
 		ret = nouveau_bo_pin(new_bo, TTM_PL_FLAG_VRAM);
 		if (ret)
 			goto fail_free;
-
-		list_add(&resv[1].head, &res);
 	}
-	list_add(&resv[0].head, &res);
 
 	mutex_lock(&chan->cli->mutex);
-	ret = ttm_eu_reserve_buffers(&ticket, &res);
+	ret = ttm_bo_reserve(&old_bo->bo, true, false, false, NULL);
 	if (ret)
 		goto fail_unpin;
 
@@ -640,12 +630,29 @@ nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 
 	/* Emit a page flip */
 	if (nv_device(drm->device)->card_type >= NV_50) {
-		ret = nv50_display_flip_next(crtc, fb, chan, 0);
+		ret = nv50_display_flip_next(crtc, fb, chan, swap_interval);
 		if (ret)
 			goto fail_unreserve;
 	} else {
 		struct nv04_display *dispnv04 = nv04_display(dev);
-		nouveau_bo_ref(new_bo, &dispnv04->image[nouveau_crtc(crtc)->index]);
+		int head = nouveau_crtc(crtc)->index;
+
+		if (swap_interval) {
+			ret = RING_SPACE(chan, 8);
+			if (ret)
+				goto fail_unreserve;
+
+			BEGIN_NV04(chan, NvSubImageBlit, 0x012c, 1);
+			OUT_RING  (chan, 0);
+			BEGIN_NV04(chan, NvSubImageBlit, 0x0134, 1);
+			OUT_RING  (chan, head);
+			BEGIN_NV04(chan, NvSubImageBlit, 0x0100, 1);
+			OUT_RING  (chan, 0);
+			BEGIN_NV04(chan, NvSubImageBlit, 0x0130, 1);
+			OUT_RING  (chan, 0);
+		}
+
+		nouveau_bo_ref(new_bo, &dispnv04->image[head]);
 	}
 
 	ret = nouveau_page_flip_emit(chan, old_bo, new_bo, s, &fence);
@@ -656,14 +663,15 @@ nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 	/* Update the crtc struct and cleanup */
 	crtc->fb = fb;
 
-	ttm_eu_fence_buffer_objects(&ticket, &res, fence);
+	nouveau_bo_fence(old_bo, fence);
+	ttm_bo_unreserve(&old_bo->bo);
 	if (old_bo != new_bo)
 		nouveau_bo_unpin(old_bo);
 	nouveau_fence_unref(&fence);
 	return 0;
 
 fail_unreserve:
-	ttm_eu_backoff_reservation(&ticket, &res);
+	ttm_bo_unreserve(&old_bo->bo);
 fail_unpin:
 	mutex_unlock(&chan->cli->mutex);
 	if (old_bo != new_bo)

drivers/gpu/drm/nouveau/nouveau_dma.h

@@ -51,9 +51,11 @@ enum {
 	NvSubCtxSurf2D  = 0,
 	NvSubSw		= 1,
 	NvSubImageBlit  = 2,
-	NvSub2D		= 3,
 	NvSubGdiRect    = 3,
-	NvSubCopy	= 4,
+
+	NvSub2D		= 3, /* DO NOT CHANGE - hardcoded for kepler gr fifo */
+	NvSubCopy	= 4, /* DO NOT CHANGE - hardcoded for kepler gr fifo */
+	FermiSw		= 5, /* DO NOT CHANGE (well.. 6/7 will work...) */
 };
 
 /* Object handles. */
@@ -194,7 +196,6 @@ WIND_RING(struct nouveau_channel *chan)
 #define NV84_SUBCHAN_UEVENT                                          0x00000020
 #define NV84_SUBCHAN_WRCACHE_FLUSH                                   0x00000024
 #define NV10_SUBCHAN_REF_CNT                                         0x00000050
-#define NVSW_SUBCHAN_PAGE_FLIP                                       0x00000054
 #define NV11_SUBCHAN_DMA_SEMAPHORE                                   0x00000060
 #define NV11_SUBCHAN_SEMAPHORE_OFFSET                                0x00000064
 #define NV11_SUBCHAN_SEMAPHORE_ACQUIRE                               0x00000068

drivers/gpu/drm/nouveau/nouveau_drm.c

@@ -37,6 +37,7 @@
 #include <engine/device.h>
 #include <engine/disp.h>
 #include <engine/fifo.h>
+#include <engine/software.h>
 
 #include <subdev/vm.h>
 
@@ -191,6 +192,32 @@ nouveau_accel_init(struct nouveau_drm *drm)
 		return;
 	}
 
+	ret = nouveau_object_new(nv_object(drm), NVDRM_CHAN, NVDRM_NVSW,
+				 nouveau_abi16_swclass(drm), NULL, 0, &object);
+	if (ret == 0) {
+		struct nouveau_software_chan *swch = (void *)object->parent;
+		ret = RING_SPACE(drm->channel, 2);
+		if (ret == 0) {
+			if (device->card_type < NV_C0) {
+				BEGIN_NV04(drm->channel, NvSubSw, 0, 1);
+				OUT_RING  (drm->channel, NVDRM_NVSW);
+			} else
+			if (device->card_type < NV_E0) {
+				BEGIN_NVC0(drm->channel, FermiSw, 0, 1);
+				OUT_RING  (drm->channel, 0x001f0000);
+			}
+		}
+		swch = (void *)object->parent;
+		swch->flip = nouveau_flip_complete;
+		swch->flip_data = drm->channel;
+	}
+
+	if (ret) {
+		NV_ERROR(drm, "failed to allocate software object, %d\n", ret);
+		nouveau_accel_fini(drm);
+		return;
+	}
+
 	if (device->card_type < NV_C0) {
 		ret = nouveau_gpuobj_new(drm->device, NULL, 32, 0, 0,
 					&drm->notify);

drivers/gpu/drm/nouveau/nouveau_drm.h

@@ -56,6 +56,7 @@ enum nouveau_drm_handle {
 	NVDRM_CONTROL = 0xdddddddc,
 	NVDRM_PUSH    = 0xbbbb0000, /* |= client chid */
 	NVDRM_CHAN    = 0xcccc0000, /* |= client chid */
+	NVDRM_NVSW    = 0x55550000,
 };
 
 struct nouveau_cli {

drivers/gpu/drm/nouveau/nouveau_fence.c

@@ -306,7 +306,8 @@ nouveau_fence_unref(struct nouveau_fence **pfence)
 struct nouveau_fence *
 nouveau_fence_ref(struct nouveau_fence *fence)
 {
-	kref_get(&fence->kref);
+	if (fence)
+		kref_get(&fence->kref);
 	return fence;
 }
 

drivers/gpu/drm/nouveau/nouveau_gem.c

@@ -106,8 +106,7 @@ nouveau_gem_object_unmap(struct nouveau_bo *nvbo, struct nouveau_vma *vma)
 
 	if (mapped) {
 		spin_lock(&nvbo->bo.bdev->fence_lock);
-		if (nvbo->bo.sync_obj)
-			fence = nouveau_fence_ref(nvbo->bo.sync_obj);
+		fence = nouveau_fence_ref(nvbo->bo.sync_obj);
 		spin_unlock(&nvbo->bo.bdev->fence_lock);
 	}
 
@@ -309,7 +308,8 @@ validate_fini_list(struct list_head *list, struct nouveau_fence *fence,
 	list_for_each_safe(entry, tmp, list) {
 		nvbo = list_entry(entry, struct nouveau_bo, entry);
 
-		nouveau_bo_fence(nvbo, fence);
+		if (likely(fence))
+			nouveau_bo_fence(nvbo, fence);
 
 		if (unlikely(nvbo->validate_mapped)) {
 			ttm_bo_kunmap(&nvbo->kmap);
@@ -438,8 +438,7 @@ validate_sync(struct nouveau_channel *chan, struct nouveau_bo *nvbo)
 	int ret = 0;
 
 	spin_lock(&nvbo->bo.bdev->fence_lock);
-	if (nvbo->bo.sync_obj)
-		fence = nouveau_fence_ref(nvbo->bo.sync_obj);
+	fence = nouveau_fence_ref(nvbo->bo.sync_obj);
 	spin_unlock(&nvbo->bo.bdev->fence_lock);
 
 	if (fence) {

drivers/gpu/drm/nouveau/nv04_fbcon.c

@@ -255,6 +255,12 @@ nv04_fbcon_accel_init(struct fb_info *info)
 	OUT_RING(chan, NvCtxSurf2D);
 	BEGIN_NV04(chan, NvSubImageBlit, 0x02fc, 1);
 	OUT_RING(chan, 3);
+	if (device->chipset >= 0x11 /*XXX: oclass == 0x009f*/) {
+		BEGIN_NV04(chan, NvSubImageBlit, 0x0120, 3);
+		OUT_RING(chan, 0);
+		OUT_RING(chan, 1);
+		OUT_RING(chan, 2);
+	}
 
 	BEGIN_NV04(chan, NvSubGdiRect, 0x0000, 1);
 	OUT_RING(chan, NvGdiRect);