We never used any invalid ptes, those were put in place for
a possibility of doing gpu faults. However our batchbuffers are not
restricted in length, so everything needs to be pointing to something
and thus out-of-bounds is pointing to scratch.
Remove the valid flag as it is always true.
v2: Expand commit msg, patch reorder (Mika)
v3: Rebase
Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Michel Thierry <michel.thierry@xxxxxxxxx>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@xxxxxxxxxxxxxxx>
Reviewed-by: Mika Kuoppala <mika.kuoppala@xxxxxxxxx>
Signed-off-by: Michał Winiarski <michal.winiarski@xxxxxxxxx>
---
drivers/gpu/drm/i915/i915_gem.c | 6 +-
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 3 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 98 ++++++++++++------------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 5 +-
drivers/gpu/drm/i915/i915_gpu_error.c | 2 +-
5 files changed, 46 insertions(+), 68 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index fdd496e..ec2335c 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -919,8 +919,7 @@ out_unpin:
if (node.allocated) {
wmb();
ggtt->base.clear_range(&ggtt->base,
- node.start, node.size,
- true);
+ node.start, node.size);
i915_gem_object_unpin_pages(obj);
remove_mappable_node(&node);
} else {
@@ -1228,8 +1227,7 @@ out_unpin:
if (node.allocated) {
wmb();
ggtt->base.clear_range(&ggtt->base,
- node.start, node.size,
- true);
+ node.start, node.size);
i915_gem_object_unpin_pages(obj);
remove_mappable_node(&node);
} else {
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 72c7c18..6835074 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -370,8 +370,7 @@ static void reloc_cache_fini(struct reloc_cache *cache)
ggtt->base.clear_range(&ggtt->base,
cache->node.start,
- cache->node.size,
- true);
+ cache->node.size);
drm_mm_remove_node(&cache->node);
} else {
i915_vma_unpin((struct i915_vma *)cache->node.mm);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 2d846aa..48ec9c5 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -191,15 +191,13 @@ static void ppgtt_unbind_vma(struct i915_vma *vma)
{
vma->vm->clear_range(vma->vm,
vma->node.start,
- vma->size,
- true);
+ vma->size);
}
static gen8_pte_t gen8_pte_encode(dma_addr_t addr,
- enum i915_cache_level level,
- bool valid)
+ enum i915_cache_level level)
{
- gen8_pte_t pte = valid ? _PAGE_PRESENT | _PAGE_RW : 0;
+ gen8_pte_t pte = _PAGE_PRESENT | _PAGE_RW;
pte |= addr;
switch (level) {
@@ -234,9 +232,9 @@ static gen8_pde_t gen8_pde_encode(const dma_addr_t addr,
static gen6_pte_t snb_pte_encode(dma_addr_t addr,
enum i915_cache_level level,
- bool valid, u32 unused)
+ u32 unused)
{
- gen6_pte_t pte = valid ? GEN6_PTE_VALID : 0;
+ gen6_pte_t pte = GEN6_PTE_VALID;
pte |= GEN6_PTE_ADDR_ENCODE(addr);
switch (level) {
@@ -256,9 +254,9 @@ static gen6_pte_t snb_pte_encode(dma_addr_t addr,
static gen6_pte_t ivb_pte_encode(dma_addr_t addr,
enum i915_cache_level level,
- bool valid, u32 unused)
+ u32 unused)
{
- gen6_pte_t pte = valid ? GEN6_PTE_VALID : 0;
+ gen6_pte_t pte = GEN6_PTE_VALID;
pte |= GEN6_PTE_ADDR_ENCODE(addr);
switch (level) {
@@ -280,9 +278,9 @@ static gen6_pte_t ivb_pte_encode(dma_addr_t addr,
static gen6_pte_t byt_pte_encode(dma_addr_t addr,
enum i915_cache_level level,
- bool valid, u32 flags)
+ u32 flags)
{
- gen6_pte_t pte = valid ? GEN6_PTE_VALID : 0;
+ gen6_pte_t pte = GEN6_PTE_VALID;
pte |= GEN6_PTE_ADDR_ENCODE(addr);
if (!(flags & PTE_READ_ONLY))
@@ -296,9 +294,9 @@ static gen6_pte_t byt_pte_encode(dma_addr_t addr,
static gen6_pte_t hsw_pte_encode(dma_addr_t addr,
enum i915_cache_level level,
- bool valid, u32 unused)
+ u32 unused)
{
- gen6_pte_t pte = valid ? GEN6_PTE_VALID : 0;
+ gen6_pte_t pte = GEN6_PTE_VALID;
pte |= HSW_PTE_ADDR_ENCODE(addr);
if (level != I915_CACHE_NONE)
@@ -309,9 +307,9 @@ static gen6_pte_t hsw_pte_encode(dma_addr_t addr,
static gen6_pte_t iris_pte_encode(dma_addr_t addr,
enum i915_cache_level level,
- bool valid, u32 unused)
+ u32 unused)
{
- gen6_pte_t pte = valid ? GEN6_PTE_VALID : 0;
+ gen6_pte_t pte = GEN6_PTE_VALID;
pte |= HSW_PTE_ADDR_ENCODE(addr);
switch (level) {
@@ -472,7 +470,7 @@ static void gen8_initialize_pt(struct i915_address_space *vm,
gen8_pte_t scratch_pte;
scratch_pte = gen8_pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC, true);
+ I915_CACHE_LLC);
fill_px(vm->dev, pt, scratch_pte);
}
@@ -485,7 +483,7 @@ static void gen6_initialize_pt(struct i915_address_space *vm,
WARN_ON(vm->scratch_page.daddr == 0);
scratch_pte = vm->pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC, true, 0);
+ I915_CACHE_LLC, 0);
fill32_px(vm->dev, pt, scratch_pte);
}
@@ -763,13 +761,11 @@ static void gen8_ppgtt_clear_pte_range(struct i915_address_space *vm,
}
static void gen8_ppgtt_clear_range(struct i915_address_space *vm,
- uint64_t start,
- uint64_t length,
- bool use_scratch)
+ uint64_t start, uint64_t length)
{
struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm);
gen8_pte_t scratch_pte = gen8_pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC, use_scratch);
+ I915_CACHE_LLC);
if (!USES_FULL_48BIT_PPGTT(vm->dev)) {
gen8_ppgtt_clear_pte_range(vm, &ppgtt->pdp, start, length,
@@ -809,7 +805,7 @@ gen8_ppgtt_insert_pte_entries(struct i915_address_space *vm,
pt_vaddr[pte] =
gen8_pte_encode(sg_page_iter_dma_address(sg_iter),
- cache_level, true);
+ cache_level);
if (++pte == GEN8_PTES) {
kunmap_px(ppgtt, pt_vaddr);
pt_vaddr = NULL;
@@ -1452,7 +1448,7 @@ static void gen8_dump_ppgtt(struct i915_hw_ppgtt *ppgtt, struct seq_file *m)
uint64_t start = ppgtt->base.start;
uint64_t length = ppgtt->base.total;
gen8_pte_t scratch_pte = gen8_pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC, true);
+ I915_CACHE_LLC);
if (!USES_FULL_48BIT_PPGTT(vm->dev)) {
gen8_dump_pdp(&ppgtt->pdp, start, length, scratch_pte, m);
@@ -1569,7 +1565,7 @@ static void gen6_dump_ppgtt(struct i915_hw_ppgtt *ppgtt, struct seq_file *m)
uint32_t start = ppgtt->base.start, length = ppgtt->base.total;
scratch_pte = vm->pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC, true, 0);
+ I915_CACHE_LLC, 0);
gen6_for_each_pde(unused, &ppgtt->pd, start, length, pde) {
u32 expected;
@@ -1782,8 +1778,7 @@ static void gen6_ppgtt_enable(struct drm_device *dev)
/* PPGTT support for Sandybdrige/Gen6 and later */
static void gen6_ppgtt_clear_range(struct i915_address_space *vm,
uint64_t start,
- uint64_t length,
- bool use_scratch)
+ uint64_t length)
{
struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm);
gen6_pte_t *pt_vaddr, scratch_pte;
@@ -1794,7 +1789,7 @@ static void gen6_ppgtt_clear_range(struct i915_address_space *vm,
unsigned last_pte, i;
scratch_pte = vm->pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC, true, 0);
+ I915_CACHE_LLC, 0);
while (num_entries) {
last_pte = first_pte + num_entries;
@@ -1832,7 +1827,7 @@ static void gen6_ppgtt_insert_entries(struct i915_address_space *vm,
pt_vaddr = kmap_px(ppgtt->pd.page_table[act_pt]);
pt_vaddr[act_pte] =
- vm->pte_encode(addr, cache_level, true, flags);
+ vm->pte_encode(addr, cache_level, flags);
if (++act_pte == GEN6_PTES) {
kunmap_px(ppgtt, pt_vaddr);
@@ -2286,8 +2281,7 @@ void i915_gem_suspend_gtt_mappings(struct drm_device *dev)
i915_check_and_clear_faults(dev_priv);
- ggtt->base.clear_range(&ggtt->base, ggtt->base.start, ggtt->base.total,
- true);
+ ggtt->base.clear_range(&ggtt->base, ggtt->base.start, ggtt->base.total);
i915_ggtt_flush(dev_priv);
}
@@ -2321,7 +2315,7 @@ static void gen8_ggtt_insert_page(struct i915_address_space *vm,
rpm_atomic_seq = assert_rpm_atomic_begin(dev_priv);
- gen8_set_pte(pte, gen8_pte_encode(addr, level, true));
+ gen8_set_pte(pte, gen8_pte_encode(addr, level));
I915_WRITE(GFX_FLSH_CNTL_GEN6, GFX_FLSH_CNTL_EN);
POSTING_READ(GFX_FLSH_CNTL_GEN6);
@@ -2348,7 +2342,7 @@ static void gen8_ggtt_insert_entries(struct i915_address_space *vm,
gtt_entries = (gen8_pte_t __iomem *)ggtt->gsm + (start >> PAGE_SHIFT);
for_each_sgt_dma(addr, sgt_iter, st) {
- gtt_entry = gen8_pte_encode(addr, level, true);
+ gtt_entry = gen8_pte_encode(addr, level);
gen8_set_pte(>t_entries[i++], gtt_entry);
}
@@ -2412,7 +2406,7 @@ static void gen6_ggtt_insert_page(struct i915_address_space *vm,
rpm_atomic_seq = assert_rpm_atomic_begin(dev_priv);
- iowrite32(vm->pte_encode(addr, level, true, flags), pte);
+ iowrite32(vm->pte_encode(addr, level, flags), pte);
I915_WRITE(GFX_FLSH_CNTL_GEN6, GFX_FLSH_CNTL_EN);
POSTING_READ(GFX_FLSH_CNTL_GEN6);
@@ -2445,7 +2439,7 @@ static void gen6_ggtt_insert_entries(struct i915_address_space *vm,
gtt_entries = (gen6_pte_t __iomem *)ggtt->gsm + (start >> PAGE_SHIFT);
for_each_sgt_dma(addr, sgt_iter, st) {
- gtt_entry = vm->pte_encode(addr, level, true, flags);
+ gtt_entry = vm->pte_encode(addr, level, flags);
iowrite32(gtt_entry, >t_entries[i++]);
}
@@ -2469,16 +2463,12 @@ static void gen6_ggtt_insert_entries(struct i915_address_space *vm,
}
static void nop_clear_range(struct i915_address_space *vm,
- uint64_t start,
- uint64_t length,
- bool use_scratch)
+ uint64_t start, uint64_t length)
{
}
static void gen8_ggtt_clear_range(struct i915_address_space *vm,
- uint64_t start,
- uint64_t length,
- bool use_scratch)
+ uint64_t start, uint64_t length)
{
struct drm_i915_private *dev_priv = to_i915(vm->dev);
struct i915_ggtt *ggtt = i915_vm_to_ggtt(vm);
@@ -2498,8 +2488,7 @@ static void gen8_ggtt_clear_range(struct i915_address_space *vm,
num_entries = max_entries;
scratch_pte = gen8_pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC,
- use_scratch);
+ I915_CACHE_LLC);
for (i = 0; i < num_entries; i++)
gen8_set_pte(>t_base[i], scratch_pte);
readl(gtt_base);
@@ -2509,8 +2498,7 @@ static void gen8_ggtt_clear_range(struct i915_address_space *vm,
static void gen6_ggtt_clear_range(struct i915_address_space *vm,
uint64_t start,
- uint64_t length,
- bool use_scratch)
+ uint64_t length)
{
struct drm_i915_private *dev_priv = to_i915(vm->dev);
struct i915_ggtt *ggtt = i915_vm_to_ggtt(vm);
@@ -2530,7 +2518,7 @@ static void gen6_ggtt_clear_range(struct i915_address_space *vm,
num_entries = max_entries;
scratch_pte = vm->pte_encode(vm->scratch_page.daddr,
- I915_CACHE_LLC, use_scratch, 0);
+ I915_CACHE_LLC, 0);
for (i = 0; i < num_entries; i++)
iowrite32(scratch_pte, >t_base[i]);
@@ -2577,8 +2565,7 @@ static void i915_ggtt_insert_entries(struct i915_address_space *vm,
static void i915_ggtt_clear_range(struct i915_address_space *vm,
uint64_t start,
- uint64_t length,
- bool unused)
+ uint64_t length)
{
struct drm_i915_private *dev_priv = to_i915(vm->dev);
unsigned first_entry = start >> PAGE_SHIFT;
@@ -2662,13 +2649,11 @@ static void ggtt_unbind_vma(struct i915_vma *vma)
if (vma->flags & I915_VMA_GLOBAL_BIND)
vma->vm->clear_range(vma->vm,
- vma->node.start, size,
- true);
+ vma->node.start, size);
if (vma->flags & I915_VMA_LOCAL_BIND && appgtt)
appgtt->base.clear_range(&appgtt->base,
- vma->node.start, size,
- true);
+ vma->node.start, size);
}
void i915_gem_gtt_finish_object(struct drm_i915_gem_object *obj)
@@ -2739,13 +2724,12 @@ int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
DRM_DEBUG_KMS("clearing unused GTT space: [%lx, %lx]\n",
hole_start, hole_end);
ggtt->base.clear_range(&ggtt->base, hole_start,
- hole_end - hole_start, true);
+ hole_end - hole_start);
}
/* And finally clear the reserved guard page */
ggtt->base.clear_range(&ggtt->base,
- ggtt->base.total - PAGE_SIZE, PAGE_SIZE,
- true);
+ ggtt->base.total - PAGE_SIZE, PAGE_SIZE);
if (USES_PPGTT(dev_priv) && !USES_FULL_PPGTT(dev_priv)) {
ppgtt = kzalloc(sizeof(*ppgtt), GFP_KERNEL);
@@ -2767,8 +2751,7 @@ int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
ppgtt->base.clear_range(&ppgtt->base,
ppgtt->base.start,
- ppgtt->base.total,
- true);
+ ppgtt->base.total);
dev_priv->mm.aliasing_ppgtt = ppgtt;
WARN_ON(ggtt->base.bind_vma != ggtt_bind_vma);
@@ -3254,8 +3237,7 @@ void i915_gem_restore_gtt_mappings(struct drm_device *dev)
i915_check_and_clear_faults(dev_priv);
/* First fill our portion of the GTT with scratch pages */
- ggtt->base.clear_range(&ggtt->base, ggtt->base.start, ggtt->base.total,
- true);
+ ggtt->base.clear_range(&ggtt->base, ggtt->base.start, ggtt->base.total);
ggtt->base.closed = true; /* skip rewriting PTE on VMA unbind */
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index bd93fb8..c241d81 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -395,7 +395,7 @@ struct i915_address_space {
/* FIXME: Need a more generic return type */
gen6_pte_t (*pte_encode)(dma_addr_t addr,
enum i915_cache_level level,
- bool valid, u32 flags); /* Create a valid PTE */
+ u32 flags); /* Create a valid PTE */
/* flags for pte_encode */
#define PTE_READ_ONLY (1<<0)
int (*allocate_va_range)(struct i915_address_space *vm,
@@ -403,8 +403,7 @@ struct i915_address_space {
uint64_t length);
void (*clear_range)(struct i915_address_space *vm,
uint64_t start,
- uint64_t length,
- bool use_scratch);
+ uint64_t length);
void (*insert_page)(struct i915_address_space *vm,
dma_addr_t addr,
uint64_t offset,
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 78cc13b..f2afcac 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -838,7 +838,7 @@ unwind:
out:
compress_fini(&zstream, dst);
- ggtt->base.clear_range(&ggtt->base, slot, PAGE_SIZE, true);
+ ggtt->base.clear_range(&ggtt->base, slot, PAGE_SIZE);
return dst;
}
--
2.7.4
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
