Commit dd19674b authored by Chris Wilson's avatar Chris Wilson

drm/i915: Remove bitmap tracking for used-ptes

We only operate on known extents (both for alloc/clear) and so we can use
both the knowledge of the bind/unbind range along with the knowledge of
the existing pagetable to avoid having to allocate temporary and
auxiliary bitmaps.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99295Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: default avatarMatthew Auld <matthew.auld@intel.com>
Reviewed-by: default avatarMichał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-12-chris@chris-wilson.co.uk
parent 16a011c8
...@@ -457,62 +457,38 @@ static void cleanup_scratch_page(struct i915_address_space *vm) ...@@ -457,62 +457,38 @@ static void cleanup_scratch_page(struct i915_address_space *vm)
static struct i915_page_table *alloc_pt(struct i915_address_space *vm) static struct i915_page_table *alloc_pt(struct i915_address_space *vm)
{ {
struct i915_page_table *pt; struct i915_page_table *pt;
const size_t count = INTEL_GEN(vm->i915) >= 8 ? GEN8_PTES : GEN6_PTES;
int ret = -ENOMEM;
pt = kzalloc(sizeof(*pt), GFP_KERNEL); pt = kmalloc(sizeof(*pt), GFP_KERNEL | __GFP_NOWARN);
if (!pt) if (unlikely(!pt))
return ERR_PTR(-ENOMEM); return ERR_PTR(-ENOMEM);
pt->used_ptes = kcalloc(BITS_TO_LONGS(count), sizeof(*pt->used_ptes), if (unlikely(setup_px(vm, pt))) {
GFP_KERNEL); kfree(pt);
return ERR_PTR(-ENOMEM);
if (!pt->used_ptes) }
goto fail_bitmap;
ret = setup_px(vm, pt);
if (ret)
goto fail_page_m;
pt->used_ptes = 0;
return pt; return pt;
fail_page_m:
kfree(pt->used_ptes);
fail_bitmap:
kfree(pt);
return ERR_PTR(ret);
} }
static void free_pt(struct i915_address_space *vm, struct i915_page_table *pt) static void free_pt(struct i915_address_space *vm, struct i915_page_table *pt)
{ {
cleanup_px(vm, pt); cleanup_px(vm, pt);
kfree(pt->used_ptes);
kfree(pt); kfree(pt);
} }
static void gen8_initialize_pt(struct i915_address_space *vm, static void gen8_initialize_pt(struct i915_address_space *vm,
struct i915_page_table *pt) struct i915_page_table *pt)
{ {
gen8_pte_t scratch_pte; fill_px(vm, pt,
gen8_pte_encode(vm->scratch_page.daddr, I915_CACHE_LLC));
scratch_pte = gen8_pte_encode(vm->scratch_page.daddr,
I915_CACHE_LLC);
fill_px(vm, pt, scratch_pte);
} }
static void gen6_initialize_pt(struct i915_address_space *vm, static void gen6_initialize_pt(struct i915_address_space *vm,
struct i915_page_table *pt) struct i915_page_table *pt)
{ {
gen6_pte_t scratch_pte; fill32_px(vm, pt,
vm->pte_encode(vm->scratch_page.daddr, I915_CACHE_LLC, 0));
WARN_ON(vm->scratch_page.daddr == 0);
scratch_pte = vm->pte_encode(vm->scratch_page.daddr,
I915_CACHE_LLC, 0);
fill32_px(vm, pt, scratch_pte);
} }
static struct i915_page_directory *alloc_pd(struct i915_address_space *vm) static struct i915_page_directory *alloc_pd(struct i915_address_space *vm)
...@@ -556,11 +532,12 @@ static void free_pd(struct i915_address_space *vm, ...@@ -556,11 +532,12 @@ static void free_pd(struct i915_address_space *vm,
static void gen8_initialize_pd(struct i915_address_space *vm, static void gen8_initialize_pd(struct i915_address_space *vm,
struct i915_page_directory *pd) struct i915_page_directory *pd)
{ {
gen8_pde_t scratch_pde; unsigned int i;
scratch_pde = gen8_pde_encode(px_dma(vm->scratch_pt), I915_CACHE_LLC); fill_px(vm, pd,
gen8_pde_encode(px_dma(vm->scratch_pt), I915_CACHE_LLC));
fill_px(vm, pd, scratch_pde); for (i = 0; i < I915_PDES; i++)
pd->page_table[i] = vm->scratch_pt;
} }
static int __pdp_init(struct drm_i915_private *dev_priv, static int __pdp_init(struct drm_i915_private *dev_priv,
...@@ -744,8 +721,7 @@ static void mark_tlbs_dirty(struct i915_hw_ppgtt *ppgtt) ...@@ -744,8 +721,7 @@ static void mark_tlbs_dirty(struct i915_hw_ppgtt *ppgtt)
*/ */
static bool gen8_ppgtt_clear_pt(struct i915_address_space *vm, static bool gen8_ppgtt_clear_pt(struct i915_address_space *vm,
struct i915_page_table *pt, struct i915_page_table *pt,
uint64_t start, u64 start, u64 length)
uint64_t length)
{ {
unsigned int num_entries = gen8_pte_count(start, length); unsigned int num_entries = gen8_pte_count(start, length);
unsigned int pte = gen8_pte_index(start); unsigned int pte = gen8_pte_index(start);
...@@ -754,16 +730,11 @@ static bool gen8_ppgtt_clear_pt(struct i915_address_space *vm, ...@@ -754,16 +730,11 @@ static bool gen8_ppgtt_clear_pt(struct i915_address_space *vm,
gen8_pte_encode(vm->scratch_page.daddr, I915_CACHE_LLC); gen8_pte_encode(vm->scratch_page.daddr, I915_CACHE_LLC);
gen8_pte_t *vaddr; gen8_pte_t *vaddr;
if (WARN_ON(!px_page(pt))) GEM_BUG_ON(num_entries > pt->used_ptes);
return false;
GEM_BUG_ON(pte_end > GEN8_PTES);
bitmap_clear(pt->used_ptes, pte, num_entries); pt->used_ptes -= num_entries;
if (USES_FULL_PPGTT(vm->i915)) { if (!pt->used_ptes)
if (bitmap_empty(pt->used_ptes, GEN8_PTES)) return true;
return true;
}
vaddr = kmap_atomic_px(pt); vaddr = kmap_atomic_px(pt);
while (pte < pte_end) while (pte < pte_end)
...@@ -773,31 +744,38 @@ static bool gen8_ppgtt_clear_pt(struct i915_address_space *vm, ...@@ -773,31 +744,38 @@ static bool gen8_ppgtt_clear_pt(struct i915_address_space *vm,
return false; return false;
} }
static void gen8_ppgtt_set_pde(struct i915_address_space *vm,
struct i915_page_directory *pd,
struct i915_page_table *pt,
unsigned int pde)
{
gen8_pde_t *vaddr;
pd->page_table[pde] = pt;
vaddr = kmap_atomic_px(pd);
vaddr[pde] = gen8_pde_encode(px_dma(pt), I915_CACHE_LLC);
kunmap_atomic(vaddr);
}
/* Removes entries from a single page dir, releasing it if it's empty. /* Removes entries from a single page dir, releasing it if it's empty.
* Caller can use the return value to update higher-level entries * Caller can use the return value to update higher-level entries
*/ */
static bool gen8_ppgtt_clear_pd(struct i915_address_space *vm, static bool gen8_ppgtt_clear_pd(struct i915_address_space *vm,
struct i915_page_directory *pd, struct i915_page_directory *pd,
uint64_t start, u64 start, u64 length)
uint64_t length)
{ {
struct i915_page_table *pt; struct i915_page_table *pt;
uint64_t pde; u32 pde;
gen8_pde_t *pde_vaddr;
gen8_pde_t scratch_pde = gen8_pde_encode(px_dma(vm->scratch_pt),
I915_CACHE_LLC);
gen8_for_each_pde(pt, pd, start, length, pde) { gen8_for_each_pde(pt, pd, start, length, pde) {
if (WARN_ON(!pd->page_table[pde])) if (!gen8_ppgtt_clear_pt(vm, pt, start, length))
break; continue;
if (gen8_ppgtt_clear_pt(vm, pt, start, length)) { gen8_ppgtt_set_pde(vm, pd, vm->scratch_pt, pde);
__clear_bit(pde, pd->used_pdes); __clear_bit(pde, pd->used_pdes);
pde_vaddr = kmap_atomic_px(pd);
pde_vaddr[pde] = scratch_pde; free_pt(vm, pt);
kunmap_atomic(pde_vaddr);
free_pt(vm, pt);
}
} }
if (bitmap_empty(pd->used_pdes, I915_PDES)) if (bitmap_empty(pd->used_pdes, I915_PDES))
...@@ -1124,8 +1102,6 @@ static void gen8_ppgtt_cleanup(struct i915_address_space *vm) ...@@ -1124,8 +1102,6 @@ static void gen8_ppgtt_cleanup(struct i915_address_space *vm)
* @pd: Page directory for this address range. * @pd: Page directory for this address range.
* @start: Starting virtual address to begin allocations. * @start: Starting virtual address to begin allocations.
* @length: Size of the allocations. * @length: Size of the allocations.
* @new_pts: Bitmap set by function with new allocations. Likely used by the
* caller to free on error.
* *
* Allocate the required number of page tables. Extremely similar to * Allocate the required number of page tables. Extremely similar to
* gen8_ppgtt_alloc_page_directories(). The main difference is here we are limited by * gen8_ppgtt_alloc_page_directories(). The main difference is here we are limited by
...@@ -1138,37 +1114,30 @@ static void gen8_ppgtt_cleanup(struct i915_address_space *vm) ...@@ -1138,37 +1114,30 @@ static void gen8_ppgtt_cleanup(struct i915_address_space *vm)
*/ */
static int gen8_ppgtt_alloc_pagetabs(struct i915_address_space *vm, static int gen8_ppgtt_alloc_pagetabs(struct i915_address_space *vm,
struct i915_page_directory *pd, struct i915_page_directory *pd,
uint64_t start, u64 start, u64 length)
uint64_t length,
unsigned long *new_pts)
{ {
struct i915_page_table *pt; struct i915_page_table *pt;
u64 from = start;
uint32_t pde; uint32_t pde;
gen8_for_each_pde(pt, pd, start, length, pde) { gen8_for_each_pde(pt, pd, start, length, pde) {
/* Don't reallocate page tables */ /* Don't reallocate page tables */
if (test_bit(pde, pd->used_pdes)) { if (!test_bit(pde, pd->used_pdes)) {
/* Scratch is never allocated this way */ pt = alloc_pt(vm);
WARN_ON(pt == vm->scratch_pt); if (IS_ERR(pt))
continue; goto unwind;
}
pt = alloc_pt(vm);
if (IS_ERR(pt))
goto unwind_out;
gen8_initialize_pt(vm, pt); gen8_initialize_pt(vm, pt);
pd->page_table[pde] = pt; pd->page_table[pde] = pt;
__set_bit(pde, new_pts); }
pt->used_ptes += gen8_pte_count(start, length);
trace_i915_page_table_entry_alloc(vm, pde, start, GEN8_PDE_SHIFT); trace_i915_page_table_entry_alloc(vm, pde, start, GEN8_PDE_SHIFT);
} }
return 0; return 0;
unwind_out: unwind:
for_each_set_bit(pde, new_pts, I915_PDES) gen8_ppgtt_clear_pd(vm, pd, from, start - from);
free_pt(vm, pd->page_table[pde]);
return -ENOMEM; return -ENOMEM;
} }
...@@ -1285,9 +1254,8 @@ gen8_ppgtt_alloc_page_dirpointers(struct i915_address_space *vm, ...@@ -1285,9 +1254,8 @@ gen8_ppgtt_alloc_page_dirpointers(struct i915_address_space *vm,
} }
static void static void
free_gen8_temp_bitmaps(unsigned long *new_pds, unsigned long *new_pts) free_gen8_temp_bitmaps(unsigned long *new_pds)
{ {
kfree(new_pts);
kfree(new_pds); kfree(new_pds);
} }
...@@ -1296,29 +1264,16 @@ free_gen8_temp_bitmaps(unsigned long *new_pds, unsigned long *new_pts) ...@@ -1296,29 +1264,16 @@ free_gen8_temp_bitmaps(unsigned long *new_pds, unsigned long *new_pts)
*/ */
static static
int __must_check alloc_gen8_temp_bitmaps(unsigned long **new_pds, int __must_check alloc_gen8_temp_bitmaps(unsigned long **new_pds,
unsigned long **new_pts,
uint32_t pdpes) uint32_t pdpes)
{ {
unsigned long *pds; unsigned long *pds;
unsigned long *pts;
pds = kcalloc(BITS_TO_LONGS(pdpes), sizeof(unsigned long), GFP_TEMPORARY); pds = kcalloc(BITS_TO_LONGS(pdpes), sizeof(unsigned long), GFP_TEMPORARY);
if (!pds) if (!pds)
return -ENOMEM; return -ENOMEM;
pts = kcalloc(pdpes, BITS_TO_LONGS(I915_PDES) * sizeof(unsigned long),
GFP_TEMPORARY);
if (!pts)
goto err_out;
*new_pds = pds; *new_pds = pds;
*new_pts = pts;
return 0; return 0;
err_out:
free_gen8_temp_bitmaps(pds, pts);
return -ENOMEM;
} }
static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm, static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm,
...@@ -1327,7 +1282,7 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm, ...@@ -1327,7 +1282,7 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm,
uint64_t length) uint64_t length)
{ {
struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm);
unsigned long *new_page_dirs, *new_page_tables; unsigned long *new_page_dirs;
struct i915_page_directory *pd; struct i915_page_directory *pd;
const uint64_t orig_start = start; const uint64_t orig_start = start;
const uint64_t orig_length = length; const uint64_t orig_length = length;
...@@ -1335,7 +1290,7 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm, ...@@ -1335,7 +1290,7 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm,
uint32_t pdpes = I915_PDPES_PER_PDP(dev_priv); uint32_t pdpes = I915_PDPES_PER_PDP(dev_priv);
int ret; int ret;
ret = alloc_gen8_temp_bitmaps(&new_page_dirs, &new_page_tables, pdpes); ret = alloc_gen8_temp_bitmaps(&new_page_dirs, pdpes);
if (ret) if (ret)
return ret; return ret;
...@@ -1343,14 +1298,13 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm, ...@@ -1343,14 +1298,13 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm,
ret = gen8_ppgtt_alloc_page_directories(vm, pdp, start, length, ret = gen8_ppgtt_alloc_page_directories(vm, pdp, start, length,
new_page_dirs); new_page_dirs);
if (ret) { if (ret) {
free_gen8_temp_bitmaps(new_page_dirs, new_page_tables); free_gen8_temp_bitmaps(new_page_dirs);
return ret; return ret;
} }
/* For every page directory referenced, allocate page tables */ /* For every page directory referenced, allocate page tables */
gen8_for_each_pdpe(pd, pdp, start, length, pdpe) { gen8_for_each_pdpe(pd, pdp, start, length, pdpe) {
ret = gen8_ppgtt_alloc_pagetabs(vm, pd, start, length, ret = gen8_ppgtt_alloc_pagetabs(vm, pd, start, length);
new_page_tables + pdpe * BITS_TO_LONGS(I915_PDES));
if (ret) if (ret)
goto err_out; goto err_out;
} }
...@@ -1376,11 +1330,6 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm, ...@@ -1376,11 +1330,6 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm,
WARN_ON(!pd_len); WARN_ON(!pd_len);
WARN_ON(!gen8_pte_count(pd_start, pd_len)); WARN_ON(!gen8_pte_count(pd_start, pd_len));
/* Set our used ptes within the page table */
bitmap_set(pt->used_ptes,
gen8_pte_index(pd_start),
gen8_pte_count(pd_start, pd_len));
/* Our pde is now pointing to the pagetable, pt */ /* Our pde is now pointing to the pagetable, pt */
__set_bit(pde, pd->used_pdes); __set_bit(pde, pd->used_pdes);
...@@ -1389,8 +1338,7 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm, ...@@ -1389,8 +1338,7 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm,
I915_CACHE_LLC); I915_CACHE_LLC);
trace_i915_page_table_entry_map(&ppgtt->base, pde, pt, trace_i915_page_table_entry_map(&ppgtt->base, pde, pt,
gen8_pte_index(start), gen8_pte_index(start),
gen8_pte_count(start, length), gen8_pte_count(start, length));
GEN8_PTES);
/* NB: We haven't yet mapped ptes to pages. At this /* NB: We haven't yet mapped ptes to pages. At this
* point we're still relying on insert_entries() */ * point we're still relying on insert_entries() */
...@@ -1401,23 +1349,15 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm, ...@@ -1401,23 +1349,15 @@ static int gen8_alloc_va_range_3lvl(struct i915_address_space *vm,
gen8_setup_pdpe(ppgtt, pdp, pd, pdpe); gen8_setup_pdpe(ppgtt, pdp, pd, pdpe);
} }
free_gen8_temp_bitmaps(new_page_dirs, new_page_tables); free_gen8_temp_bitmaps(new_page_dirs);
mark_tlbs_dirty(ppgtt); mark_tlbs_dirty(ppgtt);
return 0; return 0;
err_out: err_out:
while (pdpe--) {
unsigned long temp;
for_each_set_bit(temp, new_page_tables + pdpe *
BITS_TO_LONGS(I915_PDES), I915_PDES)
free_pt(vm, pdp->page_directory[pdpe]->page_table[temp]);
}
for_each_set_bit(pdpe, new_page_dirs, pdpes) for_each_set_bit(pdpe, new_page_dirs, pdpes)
free_pd(vm, pdp->page_directory[pdpe]); free_pd(vm, pdp->page_directory[pdpe]);
free_gen8_temp_bitmaps(new_page_dirs, new_page_tables); free_gen8_temp_bitmaps(new_page_dirs);
mark_tlbs_dirty(ppgtt); mark_tlbs_dirty(ppgtt);
return ret; return ret;
} }
...@@ -1559,14 +1499,14 @@ static void gen8_dump_ppgtt(struct i915_hw_ppgtt *ppgtt, struct seq_file *m) ...@@ -1559,14 +1499,14 @@ static void gen8_dump_ppgtt(struct i915_hw_ppgtt *ppgtt, struct seq_file *m)
static int gen8_preallocate_top_level_pdps(struct i915_hw_ppgtt *ppgtt) static int gen8_preallocate_top_level_pdps(struct i915_hw_ppgtt *ppgtt)
{ {
unsigned long *new_page_dirs, *new_page_tables; unsigned long *new_page_dirs;
uint32_t pdpes = I915_PDPES_PER_PDP(to_i915(ppgtt->base.dev)); uint32_t pdpes = I915_PDPES_PER_PDP(to_i915(ppgtt->base.dev));
int ret; int ret;
/* We allocate temp bitmap for page tables for no gain /* We allocate temp bitmap for page tables for no gain
* but as this is for init only, lets keep the things simple * but as this is for init only, lets keep the things simple
*/ */
ret = alloc_gen8_temp_bitmaps(&new_page_dirs, &new_page_tables, pdpes); ret = alloc_gen8_temp_bitmaps(&new_page_dirs, pdpes);
if (ret) if (ret)
return ret; return ret;
...@@ -1579,7 +1519,7 @@ static int gen8_preallocate_top_level_pdps(struct i915_hw_ppgtt *ppgtt) ...@@ -1579,7 +1519,7 @@ static int gen8_preallocate_top_level_pdps(struct i915_hw_ppgtt *ppgtt)
if (!ret) if (!ret)
*ppgtt->pdp.used_pdpes = *new_page_dirs; *ppgtt->pdp.used_pdpes = *new_page_dirs;
free_gen8_temp_bitmaps(new_page_dirs, new_page_tables); free_gen8_temp_bitmaps(new_page_dirs);
return ret; return ret;
} }
...@@ -1728,16 +1668,15 @@ static void gen6_write_page_range(struct i915_hw_ppgtt *ppgtt, ...@@ -1728,16 +1668,15 @@ static void gen6_write_page_range(struct i915_hw_ppgtt *ppgtt,
gen6_for_each_pde(pt, &ppgtt->pd, start, length, pde) gen6_for_each_pde(pt, &ppgtt->pd, start, length, pde)
gen6_write_pde(ppgtt, pde, pt); gen6_write_pde(ppgtt, pde, pt);
wmb();
mark_tlbs_dirty(ppgtt); mark_tlbs_dirty(ppgtt);
wmb();
} }
static uint32_t get_pd_offset(struct i915_hw_ppgtt *ppgtt) static inline uint32_t get_pd_offset(struct i915_hw_ppgtt *ppgtt)
{ {
BUG_ON(ppgtt->pd.base.ggtt_offset & 0x3f); GEM_BUG_ON(ppgtt->pd.base.ggtt_offset & 0x3f);
return ppgtt->pd.base.ggtt_offset << 10;
return (ppgtt->pd.base.ggtt_offset / 64) << 16;
} }
static int hsw_mm_switch(struct i915_hw_ppgtt *ppgtt, static int hsw_mm_switch(struct i915_hw_ppgtt *ppgtt,
...@@ -1869,35 +1808,36 @@ static void gen6_ppgtt_enable(struct drm_i915_private *dev_priv) ...@@ -1869,35 +1808,36 @@ static void gen6_ppgtt_enable(struct drm_i915_private *dev_priv)
/* PPGTT support for Sandybdrige/Gen6 and later */ /* PPGTT support for Sandybdrige/Gen6 and later */
static void gen6_ppgtt_clear_range(struct i915_address_space *vm, static void gen6_ppgtt_clear_range(struct i915_address_space *vm,
uint64_t start, u64 start, u64 length)
uint64_t length)
{ {
struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm);
gen6_pte_t *pt_vaddr, scratch_pte; unsigned int first_entry = start >> PAGE_SHIFT;
unsigned first_entry = start >> PAGE_SHIFT; unsigned int pde = first_entry / GEN6_PTES;
unsigned num_entries = length >> PAGE_SHIFT; unsigned int pte = first_entry % GEN6_PTES;
unsigned act_pt = first_entry / GEN6_PTES; unsigned int num_entries = length >> PAGE_SHIFT;
unsigned first_pte = first_entry % GEN6_PTES; gen6_pte_t scratch_pte =
unsigned last_pte, i; vm->pte_encode(vm->scratch_page.daddr, I915_CACHE_LLC, 0);
scratch_pte = vm->pte_encode(vm->scratch_page.daddr,
I915_CACHE_LLC, 0);
while (num_entries) { while (num_entries) {
last_pte = first_pte + num_entries; struct i915_page_table *pt = ppgtt->pd.page_table[pde++];
if (last_pte > GEN6_PTES) unsigned int end = min(pte + num_entries, GEN6_PTES);
last_pte = GEN6_PTES; gen6_pte_t *vaddr;
pt_vaddr = kmap_atomic_px(ppgtt->pd.page_table[act_pt]); num_entries -= end - pte;
for (i = first_pte; i < last_pte; i++) /* Note that the hw doesn't support removing PDE on the fly
pt_vaddr[i] = scratch_pte; * (they are cached inside the context with no means to
* invalidate the cache), so we can only reset the PTE
* entries back to scratch.
*/
kunmap_atomic(pt_vaddr); vaddr = kmap_atomic_px(pt);
do {
vaddr[pte++] = scratch_pte;
} while (pte < end);
kunmap_atomic(vaddr);
num_entries -= last_pte - first_pte; pte = 0;
first_pte = 0;
act_pt++;
} }
} }
...@@ -1941,89 +1881,37 @@ static void gen6_ppgtt_insert_entries(struct i915_address_space *vm, ...@@ -1941,89 +1881,37 @@ static void gen6_ppgtt_insert_entries(struct i915_address_space *vm,
} }
static int gen6_alloc_va_range(struct i915_address_space *vm, static int gen6_alloc_va_range(struct i915_address_space *vm,
uint64_t start_in, uint64_t length_in) u64 start, u64 length)
{ {
DECLARE_BITMAP(new_page_tables, I915_PDES);
struct drm_i915_private *dev_priv = vm->i915;
struct i915_ggtt *ggtt = &dev_priv->ggtt;
struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm);
struct i915_page_table *pt; struct i915_page_table *pt;
uint32_t start, length, start_save, length_save; u64 from = start;
uint32_t pde; unsigned int pde;
int ret; bool flush = false;
start = start_save = start_in;
length = length_save = length_in;
bitmap_zero(new_page_tables, I915_PDES);
/* The allocation is done in two stages so that we can bail out with
* minimal amount of pain. The first stage finds new page tables that
* need allocation. The second stage marks use ptes within the page
* tables.
*/
gen6_for_each_pde(pt, &ppgtt->pd, start, length, pde) { gen6_for_each_pde(pt, &ppgtt->pd, start, length, pde) {
if (pt != vm->scratch_pt) { if (pt == vm->scratch_pt) {
WARN_ON(bitmap_empty(pt->used_ptes, GEN6_PTES)); pt = alloc_pt(vm);
continue; if (IS_ERR(pt))
} goto unwind_out;
/* We've already allocated a page table */
WARN_ON(!bitmap_empty(pt->used_ptes, GEN6_PTES));
pt = alloc_pt(vm); gen6_initialize_pt(vm, pt);
if (IS_ERR(pt)) { ppgtt->pd.page_table[pde] = pt;
ret = PTR_ERR(pt); gen6_write_pde(ppgtt, pde, pt);
goto unwind_out; flush = true;
} }
gen6_initialize_pt(vm, pt);
ppgtt->pd.page_table[pde] = pt;
__set_bit(pde, new_page_tables);
trace_i915_page_table_entry_alloc(vm, pde, start, GEN6_PDE_SHIFT);
} }
start = start_save; if (flush) {
length = length_save; mark_tlbs_dirty(ppgtt);
wmb();
gen6_for_each_pde(pt, &ppgtt->pd, start, length, pde) {
DECLARE_BITMAP(tmp_bitmap, GEN6_PTES);
bitmap_zero(tmp_bitmap, GEN6_PTES);
bitmap_set(tmp_bitmap, gen6_pte_index(start),
gen6_pte_count(start, length));
if (__test_and_clear_bit(pde, new_page_tables))
gen6_write_pde(ppgtt, pde, pt);
trace_i915_page_table_entry_map(vm, pde, pt,
gen6_pte_index(start),
gen6_pte_count(start, length),
GEN6_PTES);
bitmap_or(pt->used_ptes, tmp_bitmap, pt->used_ptes,
GEN6_PTES);
} }
WARN_ON(!bitmap_empty(new_page_tables, I915_PDES));
/* Make sure write is complete before other code can use this page
* table. Also require for WC mapped PTEs */
readl(ggtt->gsm);
mark_tlbs_dirty(ppgtt);
return 0; return 0;
unwind_out: unwind_out:
for_each_set_bit(pde, new_page_tables, I915_PDES) { gen6_ppgtt_clear_range(vm, from, start);
struct i915_page_table *pt = ppgtt->pd.page_table[pde]; return -ENOMEM;
ppgtt->pd.page_table[pde] = vm->scratch_pt;
free_pt(vm, pt);
}
mark_tlbs_dirty(ppgtt);
return ret;
} }
static int gen6_init_scratch(struct i915_address_space *vm) static int gen6_init_scratch(struct i915_address_space *vm)
......
...@@ -69,7 +69,7 @@ typedef uint64_t gen8_ppgtt_pml4e_t; ...@@ -69,7 +69,7 @@ typedef uint64_t gen8_ppgtt_pml4e_t;
#define GEN6_PTE_UNCACHED (1 << 1) #define GEN6_PTE_UNCACHED (1 << 1)
#define GEN6_PTE_VALID (1 << 0) #define GEN6_PTE_VALID (1 << 0)
#define I915_PTES(pte_len) (PAGE_SIZE / (pte_len)) #define I915_PTES(pte_len) ((unsigned int)(PAGE_SIZE / (pte_len)))
#define I915_PTE_MASK(pte_len) (I915_PTES(pte_len) - 1) #define I915_PTE_MASK(pte_len) (I915_PTES(pte_len) - 1)
#define I915_PDES 512 #define I915_PDES 512
#define I915_PDE_MASK (I915_PDES - 1) #define I915_PDE_MASK (I915_PDES - 1)
...@@ -220,8 +220,7 @@ struct i915_page_dma { ...@@ -220,8 +220,7 @@ struct i915_page_dma {
struct i915_page_table { struct i915_page_table {
struct i915_page_dma base; struct i915_page_dma base;
unsigned int used_ptes;
unsigned long *used_ptes;
}; };
struct i915_page_directory { struct i915_page_directory {
......
...@@ -245,15 +245,14 @@ DEFINE_EVENT_PRINT(i915_px_entry, i915_page_directory_pointer_entry_alloc, ...@@ -245,15 +245,14 @@ DEFINE_EVENT_PRINT(i915_px_entry, i915_page_directory_pointer_entry_alloc,
DECLARE_EVENT_CLASS(i915_page_table_entry_update, DECLARE_EVENT_CLASS(i915_page_table_entry_update,
TP_PROTO(struct i915_address_space *vm, u32 pde, TP_PROTO(struct i915_address_space *vm, u32 pde,
struct i915_page_table *pt, u32 first, u32 count, u32 bits), struct i915_page_table *pt, u32 first, u32 count),
TP_ARGS(vm, pde, pt, first, count, bits), TP_ARGS(vm, pde, pt, first, count),
TP_STRUCT__entry( TP_STRUCT__entry(
__field(struct i915_address_space *, vm) __field(struct i915_address_space *, vm)
__field(u32, pde) __field(u32, pde)
__field(u32, first) __field(u32, first)
__field(u32, last) __field(u32, last)
__dynamic_array(char, cur_ptes, TRACE_PT_SIZE(bits))
), ),
TP_fast_assign( TP_fast_assign(
...@@ -261,22 +260,16 @@ DECLARE_EVENT_CLASS(i915_page_table_entry_update, ...@@ -261,22 +260,16 @@ DECLARE_EVENT_CLASS(i915_page_table_entry_update,
__entry->pde = pde; __entry->pde = pde;
__entry->first = first; __entry->first = first;
__entry->last = first + count - 1; __entry->last = first + count - 1;
scnprintf(__get_str(cur_ptes),
TRACE_PT_SIZE(bits),
"%*pb",
bits,
pt->used_ptes);
), ),
TP_printk("vm=%p, pde=%d, updating %u:%u\t%s", TP_printk("vm=%p, pde=%d, updating %u:%u",
__entry->vm, __entry->pde, __entry->last, __entry->first, __entry->vm, __entry->pde, __entry->last, __entry->first)
__get_str(cur_ptes))
); );
DEFINE_EVENT(i915_page_table_entry_update, i915_page_table_entry_map, DEFINE_EVENT(i915_page_table_entry_update, i915_page_table_entry_map,
TP_PROTO(struct i915_address_space *vm, u32 pde, TP_PROTO(struct i915_address_space *vm, u32 pde,
struct i915_page_table *pt, u32 first, u32 count, u32 bits), struct i915_page_table *pt, u32 first, u32 count),
TP_ARGS(vm, pde, pt, first, count, bits) TP_ARGS(vm, pde, pt, first, count)
); );
TRACE_EVENT(i915_gem_object_change_domain, TRACE_EVENT(i915_gem_object_change_domain,
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment