From: Ben Skeggs <bskeggs@redhat.com>
Date: Thu, 10 Feb 2011 03:41:01 +0000 (+1000)
Subject: drm/nv50-nvc0: restrict memtype to those specified at creation time
X-Git-Tag: v2.6.39-rc1~458^2~20^2~7
X-Git-Url: https://git.openpandora.org/cgi-bin/gitweb.cgi?p=pandora-kernel.git;a=commitdiff_plain;h=db5c8e299a30db48a3a60dadc676cf05d19d268d

drm/nv50-nvc0: restrict memtype to those specified at creation time

Upcoming patches are going to enable full support for buffers that keep
a constant GPU virtual address whenever they're validated for use by
the GPU.

In order for this to work properly while keeping support for large pages,
we need to know if it's ever going to be possible for a buffer to end
up in GART, and if so, disable large pages for the buffer's VMA.

This is a new restriction that's not present in earlier kernel's, but
should not break userspace as the current code never attempts to validate
buffers into a memtype other than it was created with.

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

diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 931dade12edd..cfdecd31f802 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -54,8 +54,8 @@ nouveau_bo_del_ttm(struct ttm_buffer_object *bo)
 }
 
 static void
-nouveau_bo_fixup_align(struct nouveau_bo *nvbo, int *align, int *size,
-		       int *page_shift)
+nouveau_bo_fixup_align(struct nouveau_bo *nvbo, u32 flags,
+		       int *align, int *size, int *page_shift)
 {
 	struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
 
@@ -80,7 +80,7 @@ nouveau_bo_fixup_align(struct nouveau_bo *nvbo, int *align, int *size,
 		}
 	} else {
 		if (likely(dev_priv->chan_vm)) {
-			if (*size > 256 * 1024)
+			if (!(flags & TTM_PL_FLAG_TT) &&  *size > 256 * 1024)
 				*page_shift = dev_priv->chan_vm->lpg_shift;
 			else
 				*page_shift = dev_priv->chan_vm->spg_shift;
@@ -113,7 +113,7 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
 	nvbo->tile_flags = tile_flags;
 	nvbo->bo.bdev = &dev_priv->ttm.bdev;
 
-	nouveau_bo_fixup_align(nvbo, &align, &size, &page_shift);
+	nouveau_bo_fixup_align(nvbo, flags, &align, &size, &page_shift);
 	align >>= PAGE_SHIFT;
 
 	if (dev_priv->chan_vm) {
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
index fce748f0f925..d409c0dc425b 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -90,6 +90,7 @@ struct nouveau_tile_reg {
 struct nouveau_bo {
 	struct ttm_buffer_object bo;
 	struct ttm_placement placement;
+	u32 valid_domains;
 	u32 placements[3];
 	u32 busy_placements[3];
 	struct ttm_bo_kmap_obj kmap;
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index cd4ed9e86704..3ce58d2222cb 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -64,6 +64,7 @@ nouveau_gem_new(struct drm_device *dev, struct nouveau_channel *chan,
 		int size, int align, uint32_t domain, uint32_t tile_mode,
 		uint32_t tile_flags, struct nouveau_bo **pnvbo)
 {
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_bo *nvbo;
 	u32 flags = 0;
 	int ret;
@@ -81,6 +82,15 @@ nouveau_gem_new(struct drm_device *dev, struct nouveau_channel *chan,
 		return ret;
 	nvbo = *pnvbo;
 
+	/* we restrict allowed domains on nv50+ to only the types
+	 * that were requested at creation time.  not possibly on
+	 * earlier chips without busting the ABI.
+	 */
+	nvbo->valid_domains = NOUVEAU_GEM_DOMAIN_VRAM |
+			      NOUVEAU_GEM_DOMAIN_GART;
+	if (dev_priv->card_type >= NV_50)
+		nvbo->valid_domains &= domain;
+
 	nvbo->gem = drm_gem_object_alloc(dev, nvbo->bo.mem.size);
 	if (!nvbo->gem) {
 		nouveau_bo_ref(NULL, pnvbo);
@@ -159,7 +169,7 @@ nouveau_gem_set_domain(struct drm_gem_object *gem, uint32_t read_domains,
 {
 	struct nouveau_bo *nvbo = gem->driver_private;
 	struct ttm_buffer_object *bo = &nvbo->bo;
-	uint32_t domains = valid_domains &
+	uint32_t domains = valid_domains & nvbo->valid_domains &
 		(write_domains ? write_domains : read_domains);
 	uint32_t pref_flags = 0, valid_flags = 0;