[PATCH 2/3] drm/i915/gtt: Wrap page_table with page_directory
Chris Wilson
chris at chris-wilson.co.uk
Sat Jul 6 12:37:21 UTC 2019
The page directory extends the page table with the shadown entries. Make
the page directory struct embed the page table for better code reuse.
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 2 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 57 ++++++++++------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 31 ++++++++----
3 files changed, 49 insertions(+), 41 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
index b33cfc56f623..9163b5238082 100644
--- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c
@@ -1503,7 +1503,7 @@ static int load_pd_dir(struct i915_request *rq, const struct i915_ppgtt *ppgtt)
*cs++ = MI_LOAD_REGISTER_IMM(1);
*cs++ = i915_mmio_reg_offset(RING_PP_DIR_BASE(engine->mmio_base));
- *cs++ = ppgtt->pd->base.ggtt_offset << 10;
+ *cs++ = px_base(ppgtt->pd)->ggtt_offset << 10;
intel_ring_advance(rq, cs);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 570d487b5cc8..8c9bb4a9235f 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -709,17 +709,12 @@ static struct i915_page_directory *__alloc_pd(void)
struct i915_page_directory *pd;
pd = kmalloc(sizeof(*pd), I915_GFP_ALLOW_FAIL);
-
if (unlikely(!pd))
return NULL;
- memset(&pd->base, 0, sizeof(pd->base));
- atomic_set(&pd->used, 0);
+ atomic_set(px_used(pd), 0);
spin_lock_init(&pd->lock);
- /* for safety */
- pd->entry[0] = NULL;
-
return pd;
}
@@ -731,7 +726,7 @@ static struct i915_page_directory *alloc_pd(struct i915_address_space *vm)
if (unlikely(!pd))
return ERR_PTR(-ENOMEM);
- if (unlikely(setup_page_dma(vm, &pd->base))) {
+ if (unlikely(setup_page_dma(vm, px_base(pd)))) {
kfree(pd);
return ERR_PTR(-ENOMEM);
}
@@ -742,7 +737,7 @@ static struct i915_page_directory *alloc_pd(struct i915_address_space *vm)
static void free_pd(struct i915_address_space *vm,
struct i915_page_directory *pd)
{
- cleanup_page_dma(vm, &pd->base);
+ cleanup_page_dma(vm, px_base(pd));
kfree(pd);
}
@@ -770,9 +765,9 @@ __set_pd_entry(struct i915_page_directory * const pd,
struct i915_page_dma * const to,
u64 (*encode)(const dma_addr_t, const enum i915_cache_level))
{
- GEM_BUG_ON(atomic_read(&pd->used) > 512);
+ GEM_BUG_ON(atomic_read(px_used(pd)) > 512);
- atomic_inc(&pd->used);
+ atomic_inc(px_used(pd));
pd->entry[pde] = to;
write_dma_entry(px_base(pd), pde, encode(to->daddr, I915_CACHE_LLC));
}
@@ -783,11 +778,11 @@ __clear_pd_entry(struct i915_page_directory * const pd,
struct i915_page_dma * const to,
u64 (*encode)(const dma_addr_t, const enum i915_cache_level))
{
- GEM_BUG_ON(atomic_read(&pd->used) == 0);
+ GEM_BUG_ON(atomic_read(px_used(pd)) == 0);
write_dma_entry(px_base(pd), pde, encode(to->daddr, I915_CACHE_LLC));
pd->entry[pde] = to;
- atomic_dec(&pd->used);
+ atomic_dec(px_used(pd));
}
#define set_pd_entry(pd, pde, to) \
@@ -799,13 +794,13 @@ __clear_pd_entry(struct i915_page_directory * const pd,
static bool
release_pd_entry(struct i915_page_directory * const pd,
const unsigned short pde,
- atomic_t *counter,
+ struct i915_page_table * const pt,
struct i915_page_dma * const scratch)
{
bool free = false;
spin_lock(&pd->lock);
- if (atomic_dec_and_test(counter)) {
+ if (atomic_dec_and_test(&pt->used)) {
clear_pd_entry(pd, pde, scratch);
free = true;
}
@@ -856,7 +851,7 @@ static void gen8_ppgtt_clear_pd(struct i915_address_space *vm,
atomic_inc(&pt->used);
gen8_ppgtt_clear_pt(vm, pt, start, length);
- if (release_pd_entry(pd, pde, &pt->used, &vm->scratch_pt))
+ if (release_pd_entry(pd, pde, pt, &vm->scratch_pt))
free_pt(vm, pt);
}
}
@@ -874,9 +869,9 @@ static void gen8_ppgtt_clear_pdp(struct i915_address_space *vm,
gen8_for_each_pdpe(pd, pdp, start, length, pdpe) {
GEM_BUG_ON(px_base(pd) == &vm->scratch_pd);
- atomic_inc(&pd->used);
+ atomic_inc(px_used(pd));
gen8_ppgtt_clear_pd(vm, pd, start, length);
- if (release_pd_entry(pdp, pdpe, &pd->used, &vm->scratch_pd))
+ if (release_pd_entry(pdp, pdpe, &pd->pt, &vm->scratch_pd))
free_pd(vm, pd);
}
}
@@ -904,9 +899,9 @@ static void gen8_ppgtt_clear_4lvl(struct i915_address_space *vm,
gen8_for_each_pml4e(pdp, pml4, start, length, pml4e) {
GEM_BUG_ON(px_base(pdp) == &vm->scratch_pdp);
- atomic_inc(&pdp->used);
+ atomic_inc(px_used(pdp));
gen8_ppgtt_clear_pdp(vm, pdp, start, length);
- if (release_pd_entry(pml4, pml4e, &pdp->used, &vm->scratch_pdp))
+ if (release_pd_entry(pml4, pml4e, &pdp->pt, &vm->scratch_pdp))
free_pd(vm, pdp);
}
}
@@ -1241,9 +1236,9 @@ static int gen8_ppgtt_notify_vgt(struct i915_ppgtt *ppgtt, bool create)
int i;
if (create)
- atomic_inc(&ppgtt->pd->used); /* never remove */
+ atomic_inc(px_used(ppgtt->pd)); /* never remove */
else
- atomic_dec(&ppgtt->pd->used);
+ atomic_dec(px_used(ppgtt->pd));
if (i915_vm_is_4lvl(vm)) {
const u64 daddr = px_dma(ppgtt->pd);
@@ -1416,7 +1411,7 @@ static int gen8_ppgtt_alloc_pdp(struct i915_address_space *vm,
pd = pdp->entry[pdpe];
}
}
- atomic_inc(&pd->used);
+ atomic_inc(px_used(pd));
spin_unlock(&pdp->lock);
ret = gen8_ppgtt_alloc_pd(vm, pd, start, length);
@@ -1424,13 +1419,13 @@ static int gen8_ppgtt_alloc_pdp(struct i915_address_space *vm,
goto unwind_pd;
spin_lock(&pdp->lock);
- atomic_dec(&pd->used);
+ atomic_dec(px_used(pd));
}
spin_unlock(&pdp->lock);
goto out;
unwind_pd:
- if (release_pd_entry(pdp, pdpe, &pd->used, &vm->scratch_pd))
+ if (release_pd_entry(pdp, pdpe, &pd->pt, &vm->scratch_pd))
free_pd(vm, pd);
unwind:
gen8_ppgtt_clear_pdp(vm, pdp, from, start - from);
@@ -1480,7 +1475,7 @@ static int gen8_ppgtt_alloc_4lvl(struct i915_address_space *vm,
pdp = pml4->entry[pml4e];
}
}
- atomic_inc(&pdp->used);
+ atomic_inc(px_used(pdp));
spin_unlock(&pml4->lock);
ret = gen8_ppgtt_alloc_pdp(vm, pdp, start, length);
@@ -1488,13 +1483,13 @@ static int gen8_ppgtt_alloc_4lvl(struct i915_address_space *vm,
goto unwind_pdp;
spin_lock(&pml4->lock);
- atomic_dec(&pdp->used);
+ atomic_dec(px_used(pdp));
}
spin_unlock(&pml4->lock);
goto out;
unwind_pdp:
- if (release_pd_entry(pml4, pml4e, &pdp->used, &vm->scratch_pdp))
+ if (release_pd_entry(pml4, pml4e, &pdp->pt, &vm->scratch_pdp))
free_pd(vm, pdp);
unwind:
gen8_ppgtt_clear_4lvl(vm, from, start - from);
@@ -1526,7 +1521,7 @@ static int gen8_preallocate_top_level_pdp(struct i915_ppgtt *ppgtt)
unwind:
gen8_ppgtt_clear_pdp(vm, pdp, from, start - from);
- atomic_set(&pdp->used, 0);
+ atomic_set(px_used(pdp), 0);
return -ENOMEM;
}
@@ -1553,7 +1548,7 @@ static void init_pd_n(struct i915_address_space *vm,
const unsigned int entries)
{
const u64 daddr = gen8_pde_encode(to->daddr, I915_CACHE_LLC);
- u64 * const vaddr = kmap_atomic(pd->base.page);
+ u64 * const vaddr = kmap_atomic_px(pd);
memset64(vaddr, daddr, entries);
kunmap_atomic(vaddr);
@@ -1581,7 +1576,7 @@ gen8_alloc_top_pd(struct i915_address_space *vm)
pd->entry[GEN8_3LVL_PDPES] = NULL;
- if (unlikely(setup_page_dma(vm, &pd->base))) {
+ if (unlikely(setup_page_dma(vm, px_base(pd)))) {
kfree(pd);
return ERR_PTR(-ENOMEM);
}
@@ -1956,7 +1951,7 @@ static int pd_vma_bind(struct i915_vma *vma,
struct i915_page_table *pt;
unsigned int pde;
- ppgtt->base.pd->base.ggtt_offset = ggtt_offset * sizeof(gen6_pte_t);
+ px_base(ppgtt->base.pd)->ggtt_offset = ggtt_offset * sizeof(gen6_pte_t);
ppgtt->pd_addr = (gen6_pte_t __iomem *)ggtt->gsm + ggtt_offset;
gen6_for_all_pdes(pt, ppgtt->base.pd, pde)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 860850411a1b..7019bb2b1941 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -240,21 +240,37 @@ struct i915_page_dma {
};
};
-#define px_base(px) (&(px)->base)
-#define px_dma(px) (px_base(px)->daddr)
-
struct i915_page_table {
struct i915_page_dma base;
atomic_t used;
};
struct i915_page_directory {
- struct i915_page_dma base;
- atomic_t used;
+ struct i915_page_table pt;
spinlock_t lock;
void *entry[512];
};
+#define __px_choose_expr(x, type, expr, other) \
+ __builtin_choose_expr( \
+ __builtin_types_compatible_p(typeof(x), type) || \
+ __builtin_types_compatible_p(typeof(x), const type), \
+ ({ type __x = (type)(x); expr; }), \
+ other)
+
+#define px_base(px) \
+ __px_choose_expr(px, struct i915_page_dma *, __x, \
+ __px_choose_expr(px, struct i915_page_table *, &__x->base, \
+ __px_choose_expr(px, struct i915_page_directory *, &__x->pt.base, \
+ (void)0)))
+#define px_dma(px) (px_base(px)->daddr)
+
+#define px_pt(px) \
+ __px_choose_expr(px, struct i915_page_table *, __x, \
+ __px_choose_expr(px, struct i915_page_directory *, &__x->pt, \
+ (void)0))
+#define px_used(px) (&px_pt(px)->used)
+
struct i915_vma_ops {
/* Map an object into an address space with the given cache flags. */
int (*bind_vma)(struct i915_vma *vma,
@@ -589,10 +605,7 @@ static inline u64 gen8_pte_count(u64 address, u64 length)
static inline dma_addr_t
i915_page_dir_dma_addr(const struct i915_ppgtt *ppgtt, const unsigned int n)
{
- struct i915_page_directory *pd;
-
- pd = i915_pdp_entry(ppgtt->pd, n);
- return px_dma(pd);
+ return px_dma(i915_pdp_entry(ppgtt->pd, n));
}
static inline struct i915_ggtt *
--
2.20.1
More information about the Intel-gfx-trybot
mailing list