drm/nouveau: Implement the pageflip ioctl.

nv0x-nv4x should be mostly fine, nv50 doesn't work yet.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

14 files changed, 265 insertions, 9 deletions

diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index c09928322eb9..cdc8f544d47f 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -942,6 +942,18 @@ nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
 	return ttm_bo_validate(bo, &nvbo->placement, false, true, false);
 }
 
+void
+nouveau_bo_fence(struct nouveau_bo *nvbo, struct nouveau_fence *fence)
+{
+	spin_lock(&nvbo->bo.bdev->fence_lock);
+	__nouveau_fence_unref(&nvbo->bo.sync_obj);
+
+	if (likely(fence))
+		nvbo->bo.sync_obj = nouveau_fence_ref(fence);
+
+	spin_unlock(&nvbo->bo.bdev->fence_lock);
+}
+
 struct ttm_bo_driver nouveau_bo_driver = {
 	.create_ttm_backend_entry = nouveau_bo_create_ttm_backend_entry,
 	.invalidate_caches = nouveau_bo_invalidate_caches,
diff --git a/drivers/gpu/drm/nouveau/nouveau_channel.c b/drivers/gpu/drm/nouveau/nouveau_channel.c
index c9cdbd786dae..11b2370e16da 100644
--- a/drivers/gpu/drm/nouveau/nouveau_channel.c
+++ b/drivers/gpu/drm/nouveau/nouveau_channel.c
@@ -148,6 +148,7 @@ nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret,
 
 	NV_DEBUG(dev, "initialising channel %d\n", chan->id);
 	INIT_LIST_HEAD(&chan->nvsw.vbl_wait);
+	INIT_LIST_HEAD(&chan->nvsw.flip);
 	INIT_LIST_HEAD(&chan->fence.pending);
 
 	/* Allocate DMA push buffer */
diff --git a/drivers/gpu/drm/nouveau/nouveau_display.c b/drivers/gpu/drm/nouveau/nouveau_display.c
index f8987bcb7f51..505c6bfb4d75 100644
--- a/drivers/gpu/drm/nouveau/nouveau_display.c
+++ b/drivers/gpu/drm/nouveau/nouveau_display.c
@@ -30,6 +30,8 @@
 #include "nouveau_fb.h"
 #include "nouveau_fbcon.h"
 #include "nouveau_hw.h"
+#include "nouveau_crtc.h"
+#include "nouveau_dma.h"
 
 static void
 nouveau_user_framebuffer_destroy(struct drm_framebuffer *drm_fb)
@@ -131,3 +133,181 @@ nouveau_vblank_disable(struct drm_device *dev, int crtc)
 	else
 		NVWriteCRTC(dev, crtc, NV_PCRTC_INTR_EN_0, 0);
 }
+
+static int
+nouveau_page_flip_reserve(struct nouveau_bo *old_bo,
+			  struct nouveau_bo *new_bo)
+{
+	int ret;
+
+	ret = nouveau_bo_pin(new_bo, TTM_PL_FLAG_VRAM);
+	if (ret)
+		return ret;
+
+	ret = ttm_bo_reserve(&new_bo->bo, false, false, false, 0);
+	if (ret)
+		goto fail;
+
+	ret = ttm_bo_reserve(&old_bo->bo, false, false, false, 0);
+	if (ret)
+		goto fail_unreserve;
+
+	return 0;
+
+fail_unreserve:
+	ttm_bo_unreserve(&new_bo->bo);
+fail:
+	nouveau_bo_unpin(new_bo);
+	return ret;
+}
+
+static void
+nouveau_page_flip_unreserve(struct nouveau_bo *old_bo,
+			    struct nouveau_bo *new_bo,
+			    struct nouveau_fence *fence)
+{
+	nouveau_bo_fence(new_bo, fence);
+	ttm_bo_unreserve(&new_bo->bo);
+
+	nouveau_bo_fence(old_bo, fence);
+	ttm_bo_unreserve(&old_bo->bo);
+
+	nouveau_bo_unpin(old_bo);
+}
+
+static int
+nouveau_page_flip_emit(struct nouveau_channel *chan,
+		       struct nouveau_bo *old_bo,
+		       struct nouveau_bo *new_bo,
+		       struct nouveau_page_flip_state *s,
+		       struct nouveau_fence **pfence)
+{
+	struct drm_device *dev = chan->dev;
+	unsigned long flags;
+	int ret;
+
+	/* Queue it to the pending list */
+	spin_lock_irqsave(&dev->event_lock, flags);
+	list_add_tail(&s->head, &chan->nvsw.flip);
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+
+	/* Synchronize with the old framebuffer */
+	ret = nouveau_fence_sync(old_bo->bo.sync_obj, chan);
+	if (ret)
+		goto fail;
+
+	/* Emit the pageflip */
+	ret = RING_SPACE(chan, 2);
+	if (ret)
+		goto fail;
+
+	BEGIN_RING(chan, NvSubSw, NV_SW_PAGE_FLIP, 1);
+	OUT_RING(chan, 0);
+	FIRE_RING(chan);
+
+	ret = nouveau_fence_new(chan, pfence, true);
+	if (ret)
+		goto fail;
+
+	return 0;
+fail:
+	spin_lock_irqsave(&dev->event_lock, flags);
+	list_del(&s->head);
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+	return ret;
+}
+
+int
+nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
+		       struct drm_pending_vblank_event *event)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	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;
+	struct nouveau_fence *fence;
+	int ret;
+
+	if (dev_priv->engine.graph.accel_blocked)
+		return -ENODEV;
+
+	s = kzalloc(sizeof(*s), GFP_KERNEL);
+	if (!s)
+		return -ENOMEM;
+
+	/* Don't let the buffers go away while we flip */
+	ret = nouveau_page_flip_reserve(old_bo, new_bo);
+	if (ret)
+		goto fail_free;
+
+	/* Initialize a page flip struct */
+	*s = (struct nouveau_page_flip_state)
+		{ { }, s->event, nouveau_crtc(crtc)->index,
+		  fb->bits_per_pixel, fb->pitch, crtc->x, crtc->y,
+		  new_bo->bo.offset };
+
+	/* Choose the channel the flip will be handled in */
+	chan = nouveau_fence_channel(new_bo->bo.sync_obj);
+	if (!chan)
+		chan = nouveau_channel_get_unlocked(dev_priv->channel);
+	mutex_lock(&chan->mutex);
+
+	/* Emit a page flip */
+	ret = nouveau_page_flip_emit(chan, old_bo, new_bo, s, &fence);
+	nouveau_channel_put(&chan);
+	if (ret)
+		goto fail_unreserve;
+
+	/* Update the crtc struct and cleanup */
+	crtc->fb = fb;
+
+	nouveau_page_flip_unreserve(old_bo, new_bo, fence);
+	nouveau_fence_unref(&fence);
+	return 0;
+
+fail_unreserve:
+	nouveau_page_flip_unreserve(old_bo, new_bo, NULL);
+fail_free:
+	kfree(s);
+	return ret;
+}
+
+int
+nouveau_finish_page_flip(struct nouveau_channel *chan,
+			 struct nouveau_page_flip_state *ps)
+{
+	struct drm_device *dev = chan->dev;
+	struct nouveau_page_flip_state *s;
+	unsigned long flags;
+
+	spin_lock_irqsave(&dev->event_lock, flags);
+
+	if (list_empty(&chan->nvsw.flip)) {
+		NV_ERROR(dev, "Unexpected pageflip in channel %d.\n", chan->id);
+		spin_unlock_irqrestore(&dev->event_lock, flags);
+		return -EINVAL;
+	}
+
+	s = list_first_entry(&chan->nvsw.flip,
+			     struct nouveau_page_flip_state, head);
+	if (s->event) {
+		struct drm_pending_vblank_event *e = s->event;
+		struct timeval now;
+
+		do_gettimeofday(&now);
+		e->event.sequence = 0;
+		e->event.tv_sec = now.tv_sec;
+		e->event.tv_usec = now.tv_usec;
+		list_add_tail(&e->base.link, &e->base.file_priv->event_list);
+		wake_up_interruptible(&e->base.file_priv->event_wait);
+	}
+
+	list_del(&s->head);
+	*ps = *s;
+	kfree(s);
+
+	spin_unlock_irqrestore(&dev->event_lock, flags);
+	return 0;
+}
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
index 7cf034fd5cd0..2bb1f1572a55 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -166,6 +166,13 @@ struct nouveau_gpuobj {
 	void *priv;
 };
 
+struct nouveau_page_flip_state {
+	struct list_head head;
+	struct drm_pending_vblank_event *event;
+	int crtc, bpp, pitch, x, y;
+	uint64_t offset;
+};
+
 struct nouveau_channel {
 	struct drm_device *dev;
 	int id;
@@ -253,6 +260,7 @@ struct nouveau_channel {
 		uint32_t vblsem_offset;
 		uint32_t vblsem_rval;
 		struct list_head vbl_wait;
+		struct list_head flip;
 	} nvsw;
 
 	struct {
@@ -1076,6 +1084,8 @@ extern void nv04_graph_destroy_context(struct nouveau_channel *);
 extern int  nv04_graph_load_context(struct nouveau_channel *);
 extern int  nv04_graph_unload_context(struct drm_device *);
 extern void nv04_graph_context_switch(struct drm_device *);
+extern int  nv04_graph_mthd_page_flip(struct nouveau_channel *chan,
+				      u32 class, u32 mthd, u32 data);
 
 /* nv10_graph.c */
 extern int  nv10_graph_init(struct drm_device *);
@@ -1249,6 +1259,7 @@ extern u16 nouveau_bo_rd16(struct nouveau_bo *nvbo, unsigned index);
 extern void nouveau_bo_wr16(struct nouveau_bo *nvbo, unsigned index, u16 val);
 extern u32 nouveau_bo_rd32(struct nouveau_bo *nvbo, unsigned index);
 extern void nouveau_bo_wr32(struct nouveau_bo *nvbo, unsigned index, u32 val);
+extern void nouveau_bo_fence(struct nouveau_bo *, struct nouveau_fence *);
 
 /* nouveau_fence.c */
 struct nouveau_fence;
@@ -1315,6 +1326,10 @@ extern int nouveau_gem_ioctl_info(struct drm_device *, void *,
 /* nouveau_display.c */
 int nouveau_vblank_enable(struct drm_device *dev, int crtc);
 void nouveau_vblank_disable(struct drm_device *dev, int crtc);
+int nouveau_crtc_page_flip(struct drm_crtc *crtc, struct drm_framebuffer *fb,
+			   struct drm_pending_vblank_event *event);
+int nouveau_finish_page_flip(struct nouveau_channel *,
+			     struct nouveau_page_flip_state *);
 
 /* nv10_gpio.c */
 int nv10_gpio_get(struct drm_device *dev, enum dcb_gpio_tag tag);
@@ -1514,5 +1529,6 @@ nv_match_device(struct drm_device *dev, unsigned device,
 #define NV_SW_VBLSEM_OFFSET                                          0x00000400
 #define NV_SW_VBLSEM_RELEASE_VALUE                                   0x00000404
 #define NV_SW_VBLSEM_RELEASE                                         0x00000408
+#define NV_SW_PAGE_FLIP                                              0x00000500
 
 #endif /* __NOUVEAU_DRV_H__ */
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index de8535b58710..9886b644f27d 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -231,15 +231,8 @@ 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);
-		if (likely(fence)) {
-			struct nouveau_fence *prev_fence;
-
-			spin_lock(&nvbo->bo.bdev->fence_lock);
-			prev_fence = nvbo->bo.sync_obj;
-			nvbo->bo.sync_obj = nouveau_fence_ref(fence);
-			spin_unlock(&nvbo->bo.bdev->fence_lock);
-			nouveau_fence_unref(&prev_fence);
-		}
+
+		nouveau_bo_fence(nvbo, fence);
 
 		if (unlikely(nvbo->validate_mapped)) {
 			ttm_bo_kunmap(&nvbo->kmap);
diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c
index d72aa8d19a19..ec9d193b6f61 100644
--- a/drivers/gpu/drm/nouveau/nouveau_state.c
+++ b/drivers/gpu/drm/nouveau/nouveau_state.c
@@ -1090,6 +1090,9 @@ int nouveau_ioctl_getparam(struct drm_device *dev, void *data,
 	case NOUVEAU_GETPARAM_HAS_BO_USAGE:
 		getparam->value = 1;
 		break;
+	case NOUVEAU_GETPARAM_HAS_PAGEFLIP:
+		getparam->value = (dev_priv->card_type < NV_50);
+		break;
 	case NOUVEAU_GETPARAM_GRAPH_UNITS:
 		/* NV40 and NV50 versions are quite different, but register
 		 * address is the same. User is supposed to know the card
diff --git a/drivers/gpu/drm/nouveau/nv04_crtc.c b/drivers/gpu/drm/nouveau/nv04_crtc.c
index b61a7ff8b7b7..e4aea721971f 100644
--- a/drivers/gpu/drm/nouveau/nv04_crtc.c
+++ b/drivers/gpu/drm/nouveau/nv04_crtc.c
@@ -989,6 +989,7 @@ static const struct drm_crtc_funcs nv04_crtc_funcs = {
 	.cursor_move = nv04_crtc_cursor_move,
 	.gamma_set = nv_crtc_gamma_set,
 	.set_config = drm_crtc_helper_set_config,
+	.page_flip = nouveau_crtc_page_flip,
 	.destroy = nv_crtc_destroy,
 };
 
diff --git a/drivers/gpu/drm/nouveau/nv04_graph.c b/drivers/gpu/drm/nouveau/nv04_graph.c
index 81aba097ef95..239519aefce6 100644
--- a/drivers/gpu/drm/nouveau/nv04_graph.c
+++ b/drivers/gpu/drm/nouveau/nv04_graph.c
@@ -26,6 +26,7 @@
 #include "drm.h"
 #include "nouveau_drm.h"
 #include "nouveau_drv.h"
+#include "nouveau_hw.h"
 
 static int nv04_graph_register(struct drm_device *dev);
 
@@ -553,6 +554,20 @@ nv04_graph_mthd_set_ref(struct nouveau_channel *chan,
 	return 0;
 }
 
+int
+nv04_graph_mthd_page_flip(struct nouveau_channel *chan,
+			  u32 class, u32 mthd, u32 data)
+{
+	struct drm_device *dev = chan->dev;
+	struct nouveau_page_flip_state s;
+
+	if (!nouveau_finish_page_flip(chan, &s))
+		nv_set_crtc_base(dev, s.crtc,
+				 s.offset + s.y * s.pitch + s.x * s.bpp / 8);
+
+	return 0;
+}
+
 /*
  * Software methods, why they are needed, and how they all work:
  *
@@ -1204,6 +1219,7 @@ nv04_graph_register(struct drm_device *dev)
 	/* nvsw */
 	NVOBJ_CLASS(dev, 0x506e, SW);
 	NVOBJ_MTHD (dev, 0x506e, 0x0150, nv04_graph_mthd_set_ref);
+	NVOBJ_MTHD (dev, 0x506e, 0x0500, nv04_graph_mthd_page_flip);
 
 	dev_priv->engine.graph.registered = true;
 	return 0;
diff --git a/drivers/gpu/drm/nouveau/nv10_graph.c b/drivers/gpu/drm/nouveau/nv10_graph.c
index a571bfd4471b..3fbb49dfc09c 100644
--- a/drivers/gpu/drm/nouveau/nv10_graph.c
+++ b/drivers/gpu/drm/nouveau/nv10_graph.c
@@ -1113,6 +1113,10 @@ nv10_graph_register(struct drm_device *dev)
 		NVOBJ_MTHD (dev, 0x0099, 0x1658, nv17_graph_mthd_lma_enable);
 	}
 
+	/* nvsw */
+	NVOBJ_CLASS(dev, 0x506e, SW);
+	NVOBJ_MTHD (dev, 0x506e, 0x0500, nv04_graph_mthd_page_flip);
+
 	dev_priv->engine.graph.registered = true;
 	return 0;
 }
diff --git a/drivers/gpu/drm/nouveau/nv20_graph.c b/drivers/gpu/drm/nouveau/nv20_graph.c
index 7720bccb3c98..51b9dd12949d 100644
--- a/drivers/gpu/drm/nouveau/nv20_graph.c
+++ b/drivers/gpu/drm/nouveau/nv20_graph.c
@@ -801,6 +801,10 @@ nv20_graph_register(struct drm_device *dev)
 	else
 		NVOBJ_CLASS(dev, 0x0597, GR);
 
+	/* nvsw */
+	NVOBJ_CLASS(dev, 0x506e, SW);
+	NVOBJ_MTHD (dev, 0x506e, 0x0500, nv04_graph_mthd_page_flip);
+
 	dev_priv->engine.graph.registered = true;
 	return 0;
 }
@@ -841,6 +845,10 @@ nv30_graph_register(struct drm_device *dev)
 	if (0x000001e0 & (1 << (dev_priv->chipset & 0x0f)))
 		NVOBJ_CLASS(dev, 0x0497, GR);
 
+	/* nvsw */
+	NVOBJ_CLASS(dev, 0x506e, SW);
+	NVOBJ_MTHD (dev, 0x506e, 0x0500, nv04_graph_mthd_page_flip);
+
 	dev_priv->engine.graph.registered = true;
 	return 0;
 }
diff --git a/drivers/gpu/drm/nouveau/nv40_graph.c b/drivers/gpu/drm/nouveau/nv40_graph.c
index b9361e28687c..159bdcd757d4 100644
--- a/drivers/gpu/drm/nouveau/nv40_graph.c
+++ b/drivers/gpu/drm/nouveau/nv40_graph.c
@@ -446,6 +446,10 @@ nv40_graph_register(struct drm_device *dev)
 	else
 		NVOBJ_CLASS(dev, 0x4097, GR);
 
+	/* nvsw */
+	NVOBJ_CLASS(dev, 0x506e, SW);
+	NVOBJ_MTHD (dev, 0x506e, 0x0500, nv04_graph_mthd_page_flip);
+
 	dev_priv->engine.graph.registered = true;
 	return 0;
 }
diff --git a/drivers/gpu/drm/nouveau/nv50_crtc.c b/drivers/gpu/drm/nouveau/nv50_crtc.c
index 56476d0c6de8..1225ea0a841b 100644
--- a/drivers/gpu/drm/nouveau/nv50_crtc.c
+++ b/drivers/gpu/drm/nouveau/nv50_crtc.c
@@ -437,6 +437,7 @@ static const struct drm_crtc_funcs nv50_crtc_funcs = {
 	.cursor_move = nv50_crtc_cursor_move,
 	.gamma_set = nv50_crtc_gamma_set,
 	.set_config = drm_crtc_helper_set_config,
+	.page_flip = nouveau_crtc_page_flip,
 	.destroy = nv50_crtc_destroy,
 };
 
diff --git a/drivers/gpu/drm/nouveau/nv50_graph.c b/drivers/gpu/drm/nouveau/nv50_graph.c
index ac7f62d524bb..6d81f4dab37d 100644
--- a/drivers/gpu/drm/nouveau/nv50_graph.c
+++ b/drivers/gpu/drm/nouveau/nv50_graph.c
@@ -29,6 +29,8 @@
 #include "nouveau_drv.h"
 #include "nouveau_ramht.h"
 #include "nouveau_grctx.h"
+#include "nouveau_dma.h"
+#include "nv50_evo.h"
 
 static int nv50_graph_register(struct drm_device *);
 
@@ -390,6 +392,19 @@ nv50_graph_nvsw_vblsem_release(struct nouveau_channel *chan,
 }
 
 static int
+nv50_graph_nvsw_mthd_page_flip(struct nouveau_channel *chan,
+			       u32 class, u32 mthd, u32 data)
+{
+	struct nouveau_page_flip_state s;
+
+	if (!nouveau_finish_page_flip(chan, &s)) {
+		/* XXX - Do something here */
+	}
+
+	return 0;
+}
+
+static int
 nv50_graph_register(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
@@ -402,6 +417,7 @@ nv50_graph_register(struct drm_device *dev)
 	NVOBJ_MTHD (dev, 0x506e, 0x0400, nv50_graph_nvsw_vblsem_offset);
 	NVOBJ_MTHD (dev, 0x506e, 0x0404, nv50_graph_nvsw_vblsem_release_val);
 	NVOBJ_MTHD (dev, 0x506e, 0x0408, nv50_graph_nvsw_vblsem_release);
+	NVOBJ_MTHD (dev, 0x506e, 0x0500, nv50_graph_nvsw_mthd_page_flip);
 
 	NVOBJ_CLASS(dev, 0x0030, GR); /* null */
 	NVOBJ_CLASS(dev, 0x5039, GR); /* m2mf */
diff --git a/include/drm/nouveau_drm.h b/include/drm/nouveau_drm.h
index 60a7b3e9c2e9..39aa0fd6a529 100644
--- a/include/drm/nouveau_drm.h
+++ b/include/drm/nouveau_drm.h
@@ -81,6 +81,7 @@ struct drm_nouveau_gpuobj_free {
 #define NOUVEAU_GETPARAM_GRAPH_UNITS     13
 #define NOUVEAU_GETPARAM_PTIMER_TIME     14
 #define NOUVEAU_GETPARAM_HAS_BO_USAGE    15
+#define NOUVEAU_GETPARAM_HAS_PAGEFLIP    16
 struct drm_nouveau_getparam {
 	uint64_t param;
 	uint64_t value;