linux_dsm_epyc7002/drivers/gpu/drm/i915/i915_gem_gtt.h

670 lines
19 KiB
C
Raw Normal View History

/*
* Copyright © 2014 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*
* Please try to maintain the following order within this file unless it makes
* sense to do otherwise. From top to bottom:
* 1. typedefs
* 2. #defines, and macros
* 3. structure definitions
* 4. function prototypes
*
* Within each section, please try to order by generation in ascending order,
* from top to bottom (ie. gen6 on the top, gen8 on the bottom).
*/
#ifndef __I915_GEM_GTT_H__
#define __I915_GEM_GTT_H__
#include <linux/io-mapping.h>
#include <linux/kref.h>
#include <linux/mm.h>
#include <linux/pagevec.h>
#include <linux/workqueue.h>
#include <drm/drm_mm.h>
#include "gt/intel_reset.h"
#include "i915_gem_fence_reg.h"
#include "i915_request.h"
#include "i915_scatterlist.h"
#include "i915_selftest.h"
#include "gt/intel_timeline.h"
#define I915_GTT_PAGE_SIZE_4K BIT_ULL(12)
#define I915_GTT_PAGE_SIZE_64K BIT_ULL(16)
#define I915_GTT_PAGE_SIZE_2M BIT_ULL(21)
#define I915_GTT_PAGE_SIZE I915_GTT_PAGE_SIZE_4K
#define I915_GTT_MAX_PAGE_SIZE I915_GTT_PAGE_SIZE_2M
#define I915_GTT_PAGE_MASK -I915_GTT_PAGE_SIZE
#define I915_GTT_MIN_ALIGNMENT I915_GTT_PAGE_SIZE
#define I915_FENCE_REG_NONE -1
#define I915_MAX_NUM_FENCES 32
/* 32 fences + sign bit for FENCE_REG_NONE */
#define I915_MAX_NUM_FENCE_BITS 6
struct drm_i915_file_private;
struct drm_i915_gem_object;
struct i915_vma;
struct intel_gt;
typedef u32 gen6_pte_t;
typedef u64 gen8_pte_t;
#define ggtt_total_entries(ggtt) ((ggtt)->vm.total >> PAGE_SHIFT)
/* gen6-hsw has bit 11-4 for physical addr bit 39-32 */
#define GEN6_GTT_ADDR_ENCODE(addr) ((addr) | (((addr) >> 28) & 0xff0))
#define GEN6_PTE_ADDR_ENCODE(addr) GEN6_GTT_ADDR_ENCODE(addr)
#define GEN6_PDE_ADDR_ENCODE(addr) GEN6_GTT_ADDR_ENCODE(addr)
#define GEN6_PTE_CACHE_LLC (2 << 1)
#define GEN6_PTE_UNCACHED (1 << 1)
#define GEN6_PTE_VALID (1 << 0)
#define I915_PTES(pte_len) ((unsigned int)(PAGE_SIZE / (pte_len)))
#define I915_PTE_MASK(pte_len) (I915_PTES(pte_len) - 1)
#define I915_PDES 512
#define I915_PDE_MASK (I915_PDES - 1)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
#define NUM_PTE(pde_shift) (1 << (pde_shift - PAGE_SHIFT))
#define GEN6_PTES I915_PTES(sizeof(gen6_pte_t))
#define GEN6_PD_SIZE (I915_PDES * PAGE_SIZE)
#define GEN6_PD_ALIGN (PAGE_SIZE * 16)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
#define GEN6_PDE_SHIFT 22
#define GEN6_PDE_VALID (1 << 0)
#define GEN7_PTE_CACHE_L3_LLC (3 << 1)
#define BYT_PTE_SNOOPED_BY_CPU_CACHES (1 << 2)
#define BYT_PTE_WRITEABLE (1 << 1)
/* Cacheability Control is a 4-bit value. The low three bits are stored in bits
* 3:1 of the PTE, while the fourth bit is stored in bit 11 of the PTE.
*/
#define HSW_CACHEABILITY_CONTROL(bits) ((((bits) & 0x7) << 1) | \
(((bits) & 0x8) << (11 - 3)))
#define HSW_WB_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x2)
#define HSW_WB_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x3)
#define HSW_WB_ELLC_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x8)
#define HSW_WB_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0xb)
#define HSW_WT_ELLC_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x7)
#define HSW_WT_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x6)
#define HSW_PTE_UNCACHED (0)
#define HSW_GTT_ADDR_ENCODE(addr) ((addr) | (((addr) >> 28) & 0x7f0))
#define HSW_PTE_ADDR_ENCODE(addr) HSW_GTT_ADDR_ENCODE(addr)
/*
* GEN8 32b style address is defined as a 3 level page table:
* 31:30 | 29:21 | 20:12 | 11:0
* PDPE | PDE | PTE | offset
* The difference as compared to normal x86 3 level page table is the PDPEs are
* programmed via register.
*
* GEN8 48b style address is defined as a 4 level page table:
* 47:39 | 38:30 | 29:21 | 20:12 | 11:0
* PML4E | PDPE | PDE | PTE | offset
*/
#define GEN8_3LVL_PDPES 4
#define PPAT_UNCACHED (_PAGE_PWT | _PAGE_PCD)
#define PPAT_CACHED_PDE 0 /* WB LLC */
#define PPAT_CACHED _PAGE_PAT /* WB LLCeLLC */
#define PPAT_DISPLAY_ELLC _PAGE_PCD /* WT eLLC */
#define CHV_PPAT_SNOOP (1<<6)
#define GEN8_PPAT_AGE(x) ((x)<<4)
#define GEN8_PPAT_LLCeLLC (3<<2)
#define GEN8_PPAT_LLCELLC (2<<2)
#define GEN8_PPAT_LLC (1<<2)
#define GEN8_PPAT_WB (3<<0)
#define GEN8_PPAT_WT (2<<0)
#define GEN8_PPAT_WC (1<<0)
#define GEN8_PPAT_UC (0<<0)
#define GEN8_PPAT_ELLC_OVERRIDE (0<<2)
#define GEN8_PPAT(i, x) ((u64)(x) << ((i) * 8))
#define GEN8_PDE_IPS_64K BIT(11)
#define GEN8_PDE_PS_2M BIT(7)
#define for_each_sgt_daddr(__dp, __iter, __sgt) \
__for_each_sgt_daddr(__dp, __iter, __sgt, I915_GTT_PAGE_SIZE)
struct intel_remapped_plane_info {
/* in gtt pages */
unsigned int width, height, stride, offset;
} __packed;
struct intel_remapped_info {
struct intel_remapped_plane_info plane[2];
unsigned int unused_mbz;
} __packed;
drm/i915/skl: Support secondary (rotated) frame buffer mapping 90/270 rotated scanout needs a rotated GTT view of the framebuffer. This is put in a separate VMA with a dedicated ggtt view and wired such that it is created when a framebuffer is pinned to a 90/270 rotated plane. Rotation is only possible with Yb/Yf buffers and error is propagated to user space in case of a mismatch. Special rotated page view is constructed at the VMA creation time by borrowing the DMA addresses from obj->pages. v2: * Do not bother with pages for rotated sg list, just populate the DMA addresses. (Daniel Vetter) * Checkpatch cleanup. v3: * Rebased on top of new plane handling (create rotated mapping when setting the rotation property). * Unpin rotated VMA on unpinning from display plane. * Simplify rotation check using bitwise AND. (Chris Wilson) v4: * Fix unpinning of optional rotated mapping so it is really considered to be optional. v5: * Rebased for fb modifier changes. * Rebased for atomic commit. * Only pin needed view for display. (Ville Syrjälä, Daniel Vetter) v6: * Rebased after preparatory work has been extracted out. (Daniel Vetter) v7: * Slightly simplified tiling geometry calculation. * Moved rotated GGTT view implementation into i915_gem_gtt.c (Daniel Vetter) v8: * Do not use i915_gem_obj_size to get object size since that actually returns the size of an VMA which may not exist. * Rebased for ggtt view changes. v9: * Rebased after code review changes on the preceding patches. * Tidy function definitions. (Joonas Lahtinen) For: VIZ-4726 Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Michel Thierry <michel.thierry@intel.com> (v4) Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-23 18:10:36 +07:00
struct intel_rotation_info {
struct intel_remapped_plane_info plane[2];
} __packed;
struct intel_partial_info {
u64 offset;
unsigned int size;
} __packed;
enum i915_ggtt_view_type {
I915_GGTT_VIEW_NORMAL = 0,
I915_GGTT_VIEW_ROTATED = sizeof(struct intel_rotation_info),
I915_GGTT_VIEW_PARTIAL = sizeof(struct intel_partial_info),
I915_GGTT_VIEW_REMAPPED = sizeof(struct intel_remapped_info),
};
static inline void assert_i915_gem_gtt_types(void)
{
BUILD_BUG_ON(sizeof(struct intel_rotation_info) != 8*sizeof(unsigned int));
BUILD_BUG_ON(sizeof(struct intel_partial_info) != sizeof(u64) + sizeof(unsigned int));
BUILD_BUG_ON(sizeof(struct intel_remapped_info) != 9*sizeof(unsigned int));
/* Check that rotation/remapped shares offsets for simplicity */
BUILD_BUG_ON(offsetof(struct intel_remapped_info, plane[0]) !=
offsetof(struct intel_rotation_info, plane[0]));
BUILD_BUG_ON(offsetofend(struct intel_remapped_info, plane[1]) !=
offsetofend(struct intel_rotation_info, plane[1]));
/* As we encode the size of each branch inside the union into its type,
* we have to be careful that each branch has a unique size.
*/
switch ((enum i915_ggtt_view_type)0) {
case I915_GGTT_VIEW_NORMAL:
case I915_GGTT_VIEW_PARTIAL:
case I915_GGTT_VIEW_ROTATED:
case I915_GGTT_VIEW_REMAPPED:
/* gcc complains if these are identical cases */
break;
}
}
drm/i915: Infrastructure for supporting different GGTT views per object Things like reliable GGTT mappings and mirrored 2d-on-3d display will need to map objects into the same address space multiple times. Added a GGTT view concept and linked it with the VMA to distinguish between multiple instances per address space. New objects and GEM functions which do not take this new view as a parameter assume the default of zero (I915_GGTT_VIEW_NORMAL) which preserves the previous behaviour. This now means that objects can have multiple VMA entries so the code which assumed there will only be one also had to be modified. Alternative GGTT views are supposed to borrow DMA addresses from obj->pages which is DMA mapped on first VMA instantiation and unmapped on the last one going away. v2: * Removed per view special casing in i915_gem_ggtt_prepare / finish_object in favour of creating and destroying DMA mappings on first VMA instantiation and last VMA destruction. (Daniel Vetter) * Simplified i915_vma_unbind which does not need to count the GGTT views. (Daniel Vetter) * Also moved obj->map_and_fenceable reset under the same check. * Checkpatch cleanups. v3: * Only retire objects once the last VMA is unbound. v4: * Keep scatter-gather table for alternative views persistent for the lifetime of the VMA. * Propagate binding errors to callers and handle appropriately. v5: * Explicitly look for normal GGTT view in i915_gem_obj_bound to align usage in i915_gem_object_ggtt_unpin. (Michel Thierry) * Change to single if statement in i915_gem_obj_to_ggtt. (Michel Thierry) * Removed stray semi-colon in i915_gem_object_set_cache_level. For: VIZ-4544 Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by: Michel Thierry <michel.thierry@intel.com> [danvet: Drop hunk from i915_gem_shrink since it's just prettification but upsets a __must_check warning.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-12-11 00:27:58 +07:00
struct i915_ggtt_view {
enum i915_ggtt_view_type type;
union {
/* Members need to contain no holes/padding */
struct intel_partial_info partial;
struct intel_rotation_info rotated;
struct intel_remapped_info remapped;
};
drm/i915: Infrastructure for supporting different GGTT views per object Things like reliable GGTT mappings and mirrored 2d-on-3d display will need to map objects into the same address space multiple times. Added a GGTT view concept and linked it with the VMA to distinguish between multiple instances per address space. New objects and GEM functions which do not take this new view as a parameter assume the default of zero (I915_GGTT_VIEW_NORMAL) which preserves the previous behaviour. This now means that objects can have multiple VMA entries so the code which assumed there will only be one also had to be modified. Alternative GGTT views are supposed to borrow DMA addresses from obj->pages which is DMA mapped on first VMA instantiation and unmapped on the last one going away. v2: * Removed per view special casing in i915_gem_ggtt_prepare / finish_object in favour of creating and destroying DMA mappings on first VMA instantiation and last VMA destruction. (Daniel Vetter) * Simplified i915_vma_unbind which does not need to count the GGTT views. (Daniel Vetter) * Also moved obj->map_and_fenceable reset under the same check. * Checkpatch cleanups. v3: * Only retire objects once the last VMA is unbound. v4: * Keep scatter-gather table for alternative views persistent for the lifetime of the VMA. * Propagate binding errors to callers and handle appropriately. v5: * Explicitly look for normal GGTT view in i915_gem_obj_bound to align usage in i915_gem_object_ggtt_unpin. (Michel Thierry) * Change to single if statement in i915_gem_obj_to_ggtt. (Michel Thierry) * Removed stray semi-colon in i915_gem_object_set_cache_level. For: VIZ-4544 Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by: Michel Thierry <michel.thierry@intel.com> [danvet: Drop hunk from i915_gem_shrink since it's just prettification but upsets a __must_check warning.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-12-11 00:27:58 +07:00
};
enum i915_cache_level;
drm/i915: Infrastructure for supporting different GGTT views per object Things like reliable GGTT mappings and mirrored 2d-on-3d display will need to map objects into the same address space multiple times. Added a GGTT view concept and linked it with the VMA to distinguish between multiple instances per address space. New objects and GEM functions which do not take this new view as a parameter assume the default of zero (I915_GGTT_VIEW_NORMAL) which preserves the previous behaviour. This now means that objects can have multiple VMA entries so the code which assumed there will only be one also had to be modified. Alternative GGTT views are supposed to borrow DMA addresses from obj->pages which is DMA mapped on first VMA instantiation and unmapped on the last one going away. v2: * Removed per view special casing in i915_gem_ggtt_prepare / finish_object in favour of creating and destroying DMA mappings on first VMA instantiation and last VMA destruction. (Daniel Vetter) * Simplified i915_vma_unbind which does not need to count the GGTT views. (Daniel Vetter) * Also moved obj->map_and_fenceable reset under the same check. * Checkpatch cleanups. v3: * Only retire objects once the last VMA is unbound. v4: * Keep scatter-gather table for alternative views persistent for the lifetime of the VMA. * Propagate binding errors to callers and handle appropriately. v5: * Explicitly look for normal GGTT view in i915_gem_obj_bound to align usage in i915_gem_object_ggtt_unpin. (Michel Thierry) * Change to single if statement in i915_gem_obj_to_ggtt. (Michel Thierry) * Removed stray semi-colon in i915_gem_object_set_cache_level. For: VIZ-4544 Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Reviewed-by: Michel Thierry <michel.thierry@intel.com> [danvet: Drop hunk from i915_gem_shrink since it's just prettification but upsets a __must_check warning.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-12-11 00:27:58 +07:00
struct i915_vma;
struct i915_page_dma {
struct page *page;
union {
dma_addr_t daddr;
/* For gen6/gen7 only. This is the offset in the GGTT
* where the page directory entries for PPGTT begin
*/
u32 ggtt_offset;
};
};
struct i915_page_scratch {
struct i915_page_dma base;
u64 encode;
};
struct i915_page_table {
struct i915_page_dma base;
atomic_t used;
};
struct i915_page_directory {
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_scratch *, &__x->base, \
__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,
enum i915_cache_level cache_level,
u32 flags);
/*
* Unmap an object from an address space. This usually consists of
* setting the valid PTE entries to a reserved scratch page.
*/
void (*unbind_vma)(struct i915_vma *vma);
int (*set_pages)(struct i915_vma *vma);
void (*clear_pages)(struct i915_vma *vma);
};
struct pagestash {
spinlock_t lock;
struct pagevec pvec;
};
struct i915_address_space {
struct kref ref;
struct rcu_work rcu;
struct drm_mm mm;
struct intel_gt *gt;
struct drm_i915_private *i915;
struct device *dma;
/* Every address space belongs to a struct file - except for the global
* GTT that is owned by the driver (and so @file is set to NULL). In
* principle, no information should leak from one context to another
* (or between files/processes etc) unless explicitly shared by the
* owner. Tracking the owner is important in order to free up per-file
* objects along with the file, to aide resource tracking, and to
* assign blame.
*/
struct drm_i915_file_private *file;
u64 total; /* size addr space maps (ex. 2GB for ggtt) */
u64 reserved; /* size addr space reserved */
unsigned int bind_async_flags;
drm/i915: Pull i915_vma_pin under the vm->mutex Replace the struct_mutex requirement for pinning the i915_vma with the local vm->mutex instead. Note that the vm->mutex is tainted by the shrinker (we require unbinding from inside fs-reclaim) and so we cannot allocate while holding that mutex. Instead we have to preallocate workers to do allocate and apply the PTE updates after we have we reserved their slot in the drm_mm (using fences to order the PTE writes with the GPU work and with later unbind). In adding the asynchronous vma binding, one subtle requirement is to avoid coupling the binding fence into the backing object->resv. That is the asynchronous binding only applies to the vma timeline itself and not to the pages as that is a more global timeline (the binding of one vma does not need to be ordered with another vma, nor does the implicit GEM fencing depend on a vma, only on writes to the backing store). Keeping the vma binding distinct from the backing store timelines is verified by a number of async gem_exec_fence and gem_exec_schedule tests. The way we do this is quite simple, we keep the fence for the vma binding separate and only wait on it as required, and never add it to the obj->resv itself. Another consequence in reducing the locking around the vma is the destruction of the vma is no longer globally serialised by struct_mutex. A natural solution would be to add a kref to i915_vma, but that requires decoupling the reference cycles, possibly by introducing a new i915_mm_pages object that is own by both obj->mm and vma->pages. However, we have not taken that route due to the overshadowing lmem/ttm discussions, and instead play a series of complicated games with trylocks to (hopefully) ensure that only one destruction path is called! v2: Add some commentary, and some helpers to reduce patch churn. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191004134015.13204-4-chris@chris-wilson.co.uk
2019-10-04 20:39:58 +07:00
/*
* Each active user context has its own address space (in full-ppgtt).
* Since the vm may be shared between multiple contexts, we count how
* many contexts keep us "open". Once open hits zero, we are closed
* and do not allow any new attachments, and proceed to shutdown our
* vma and page directories.
*/
atomic_t open;
struct mutex mutex; /* protects vma and our lists */
#define VM_CLASS_GGTT 0
#define VM_CLASS_PPGTT 1
struct i915_page_scratch scratch[4];
unsigned int scratch_order;
unsigned int top;
/**
drm/i915: Stop tracking MRU activity on VMA Our goal is to remove struct_mutex and replace it with fine grained locking. One of the thorny issues is our eviction logic for reclaiming space for an execbuffer (or GTT mmaping, among a few other examples). While eviction itself is easy to move under a per-VM mutex, performing the activity tracking is less agreeable. One solution is not to do any MRU tracking and do a simple coarse evaluation during eviction of active/inactive, with a loose temporal ordering of last insertion/evaluation. That keeps all the locking constrained to when we are manipulating the VM itself, neatly avoiding the tricky handling of possible recursive locking during execbuf and elsewhere. Note that discarding the MRU (currently implemented as a pair of lists, to avoid scanning the active list for a NONBLOCKING search) is unlikely to impact upon our efficiency to reclaim VM space (where we think a LRU model is best) as our current strategy is to use random idle replacement first before doing a search, and over time the use of softpinned 48b per-ppGTT is growing (thereby eliminating any need to perform any eviction searches, in theory at least) with the remaining users being found on much older devices (gen2-gen6). v2: Changelog and commentary rewritten to elaborate on the duality of a single list being both an inactive and active list. v3: Consolidate bool parameters into a single set of flags; don't comment on the duality of a single variable being a multiplicity of bits. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190128102356.15037-1-chris@chris-wilson.co.uk
2019-01-28 17:23:52 +07:00
* List of vma currently bound.
*/
drm/i915: Stop tracking MRU activity on VMA Our goal is to remove struct_mutex and replace it with fine grained locking. One of the thorny issues is our eviction logic for reclaiming space for an execbuffer (or GTT mmaping, among a few other examples). While eviction itself is easy to move under a per-VM mutex, performing the activity tracking is less agreeable. One solution is not to do any MRU tracking and do a simple coarse evaluation during eviction of active/inactive, with a loose temporal ordering of last insertion/evaluation. That keeps all the locking constrained to when we are manipulating the VM itself, neatly avoiding the tricky handling of possible recursive locking during execbuf and elsewhere. Note that discarding the MRU (currently implemented as a pair of lists, to avoid scanning the active list for a NONBLOCKING search) is unlikely to impact upon our efficiency to reclaim VM space (where we think a LRU model is best) as our current strategy is to use random idle replacement first before doing a search, and over time the use of softpinned 48b per-ppGTT is growing (thereby eliminating any need to perform any eviction searches, in theory at least) with the remaining users being found on much older devices (gen2-gen6). v2: Changelog and commentary rewritten to elaborate on the duality of a single list being both an inactive and active list. v3: Consolidate bool parameters into a single set of flags; don't comment on the duality of a single variable being a multiplicity of bits. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190128102356.15037-1-chris@chris-wilson.co.uk
2019-01-28 17:23:52 +07:00
struct list_head bound_list;
struct pagestash free_pages;
/* Global GTT */
bool is_ggtt:1;
/* Some systems require uncached updates of the page directories */
bool pt_kmap_wc:1;
/* Some systems support read-only mappings for GGTT and/or PPGTT */
bool has_read_only:1;
u64 (*pte_encode)(dma_addr_t addr,
enum i915_cache_level level,
u32 flags); /* Create a valid PTE */
#define PTE_READ_ONLY (1<<0)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
int (*allocate_va_range)(struct i915_address_space *vm,
u64 start, u64 length);
void (*clear_range)(struct i915_address_space *vm,
u64 start, u64 length);
void (*insert_page)(struct i915_address_space *vm,
dma_addr_t addr,
u64 offset,
enum i915_cache_level cache_level,
u32 flags);
void (*insert_entries)(struct i915_address_space *vm,
struct i915_vma *vma,
enum i915_cache_level cache_level,
u32 flags);
void (*cleanup)(struct i915_address_space *vm);
struct i915_vma_ops vma_ops;
I915_SELFTEST_DECLARE(struct fault_attr fault_attr);
I915_SELFTEST_DECLARE(bool scrub_64K);
};
#define i915_is_ggtt(vm) ((vm)->is_ggtt)
static inline bool
i915_vm_is_4lvl(const struct i915_address_space *vm)
{
return (vm->total - 1) >> 32;
}
static inline bool
i915_vm_has_scratch_64K(struct i915_address_space *vm)
{
return vm->scratch_order == get_order(I915_GTT_PAGE_SIZE_64K);
}
static inline bool
i915_vm_has_cache_coloring(struct i915_address_space *vm)
{
return i915_is_ggtt(vm) && vm->mm.color_adjust;
}
/* The Graphics Translation Table is the way in which GEN hardware translates a
* Graphics Virtual Address into a Physical Address. In addition to the normal
* collateral associated with any va->pa translations GEN hardware also has a
* portion of the GTT which can be mapped by the CPU and remain both coherent
* and correct (in cases like swizzling). That region is referred to as GMADR in
* the spec.
*/
struct i915_ggtt {
struct i915_address_space vm;
struct io_mapping iomap; /* Mapping to our CPU mappable region */
struct resource gmadr; /* GMADR resource */
resource_size_t mappable_end; /* End offset that we can CPU map */
/** "Graphics Stolen Memory" holds the global PTEs */
void __iomem *gsm;
void (*invalidate)(struct i915_ggtt *ggtt);
/** PPGTT used for aliasing the PPGTT with the GTT */
struct i915_ppgtt *alias;
bool do_idle_maps;
int mtrr;
/** Bit 6 swizzling required for X tiling */
u32 bit_6_swizzle_x;
/** Bit 6 swizzling required for Y tiling */
u32 bit_6_swizzle_y;
drm/i915/guc: Move the pin bias value from GuC to GGTT Removing the pin bias from GuC allows us to not check for GuC every time we pin a context, which fixes the assertion error on unresolved GuC platform default in mock contexts selftest. It also seems that we were using uninitialized WOPCM variables when setting the GuC pin bias. The pin bias has to be set after the WOPCM, but before the call to i915_gem_contexts_init where the first contexts are pinned. v2: This also makes it so that there's no need to set GuC variables from within the WOPCM init function or to move the WOPCM init, while keeping the correct initialization order. Also for mock tests the pin bias is left at 0 and we make sure that the pin bias with GuC will not be smaller than without GuC. v3: Avoid unused i915 in intel_guc_ggtt_offset if debug is disabled. v4: Squash with WOPCM init reordering. Moved the i915_ggtt_pin_bias helper to this patch, and made some functions use it instead of directly dereferencing i915->ggtt. v5: Since we now don't use wopcm.guc.base for the pin bias there's no need to validate it. It also has already been verified in WOPCM init. v6: Deleted the now unnecessarily introduced includes from previous versions. Dropped naming changes from dev_priv to i915 for better patch readability. v7: Changed some comments to make more sense in the context they're in. v8: Moved and renamed the function which now returns the wopcm.guc.size to intel_guc.c:intel_guc_reserved_gtt_size to avoid any possible confusion with the pin_bias in ggtt, which should be used for pinning. Fixed patch not applying or the most recent upstream. Fixes: f7dc0157e4b5 ("drm/i915/uc: Fetch GuC/HuC firmwares from guc/huc specific init") Testcase: igt/drv_selftest/mock_contexts #GuC Signed-off-by: Jakub Bartmiński <jakub.bartminski@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Michał Winiarski <michal.winiarski@intel.com> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180727141148.30874-3-jakub.bartminski@intel.com
2018-07-27 21:11:45 +07:00
u32 pin_bias;
unsigned int num_fences;
struct i915_fence_reg fence_regs[I915_MAX_NUM_FENCES];
struct list_head fence_list;
/** List of all objects in gtt_space, currently mmaped by userspace.
* All objects within this list must also be on bound_list.
*/
struct list_head userfault_list;
/* Manual runtime pm autosuspend delay for user GGTT mmaps */
struct intel_wakeref_auto userfault_wakeref;
struct drm_mm_node error_capture;
struct drm_mm_node uc_fw;
};
struct i915_ppgtt {
struct i915_address_space vm;
struct i915_page_directory *pd;
};
struct gen6_ppgtt {
struct i915_ppgtt base;
struct mutex flush;
struct i915_vma *vma;
gen6_pte_t __iomem *pd_addr;
atomic_t pin_count;
struct mutex pin_mutex;
bool scan_for_unused_pt;
};
#define __to_gen6_ppgtt(base) container_of(base, struct gen6_ppgtt, base)
static inline struct gen6_ppgtt *to_gen6_ppgtt(struct i915_ppgtt *base)
{
BUILD_BUG_ON(offsetof(struct gen6_ppgtt, base));
return __to_gen6_ppgtt(base);
}
/*
* gen6_for_each_pde() iterates over every pde from start until start+length.
* If start and start+length are not perfectly divisible, the macro will round
* down and up as needed. Start=0 and length=2G effectively iterates over
* every PDE in the system. The macro modifies ALL its parameters except 'pd',
* so each of the other parameters should preferably be a simple variable, or
* at most an lvalue with no side-effects!
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
*/
#define gen6_for_each_pde(pt, pd, start, length, iter) \
for (iter = gen6_pde_index(start); \
length > 0 && iter < I915_PDES && \
(pt = i915_pt_entry(pd, iter), true); \
({ u32 temp = ALIGN(start+1, 1 << GEN6_PDE_SHIFT); \
temp = min(temp - start, length); \
start += temp, length -= temp; }), ++iter)
#define gen6_for_all_pdes(pt, pd, iter) \
for (iter = 0; \
iter < I915_PDES && \
(pt = i915_pt_entry(pd, iter), true); \
++iter)
static inline u32 i915_pte_index(u64 address, unsigned int pde_shift)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
{
const u32 mask = NUM_PTE(pde_shift) - 1;
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
return (address >> PAGE_SHIFT) & mask;
}
/* Helper to counts the number of PTEs within the given length. This count
* does not cross a page table boundary, so the max value would be
* GEN6_PTES for GEN6, and GEN8_PTES for GEN8.
*/
static inline u32 i915_pte_count(u64 addr, u64 length, unsigned int pde_shift)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
{
const u64 mask = ~((1ULL << pde_shift) - 1);
u64 end;
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
GEM_BUG_ON(length == 0);
GEM_BUG_ON(offset_in_page(addr | length));
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
end = addr + length;
if ((addr & mask) != (end & mask))
return NUM_PTE(pde_shift) - i915_pte_index(addr, pde_shift);
return i915_pte_index(end, pde_shift) - i915_pte_index(addr, pde_shift);
}
static inline u32 i915_pde_index(u64 addr, u32 shift)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
{
return (addr >> shift) & I915_PDE_MASK;
}
static inline u32 gen6_pte_index(u32 addr)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
{
return i915_pte_index(addr, GEN6_PDE_SHIFT);
}
static inline u32 gen6_pte_count(u32 addr, u32 length)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
{
return i915_pte_count(addr, length, GEN6_PDE_SHIFT);
}
static inline u32 gen6_pde_index(u32 addr)
drm/i915: Track GEN6 page table usage Instead of implementing the full tracking + dynamic allocation, this patch does a bit less than half of the work, by tracking and warning on unexpected conditions. The tracking itself follows which PTEs within a page table are currently being used for objects. The next patch will modify this to actually allocate the page tables only when necessary. With the current patch there isn't much in the way of making a gen agnostic range allocation function. However, in the next patch we'll add more specificity which makes having separate functions a bit easier to manage. One important change introduced here is that DMA mappings are created/destroyed at the same page directories/tables are allocated/deallocated. Notice that aliasing PPGTT is not managed here. The patch which actually begins dynamic allocation/teardown explains the reasoning for this. v2: s/pdp.page_directory/pdp.page_directories Make a scratch page allocation helper v3: Rebase and expand commit message. v4: Allocate required pagetables only when it is needed, _bind_to_vm instead of bind_vma (Daniel). v5: Rebased to remove the unnecessary noise in the diff, also: - PDE mask is GEN agnostic, renamed GEN6_PDE_MASK to I915_PDE_MASK. - Removed unnecessary checks in gen6_alloc_va_range. - Changed map/unmap_px_single macros to use dma functions directly and be part of a static inline function instead. - Moved drm_device plumbing through page tables operation to its own patch. - Moved allocate/teardown_va_range calls until they are fully implemented (in subsequent patch). - Merged pt and scratch_pt unmap_and_free path. - Moved scratch page allocator helper to the patch that will use it. v6: Reduce complexity by not tearing down pagetables dynamically, the same can be achieved while freeing empty vms. (Daniel) v7: s/i915_dma_map_px_single/i915_dma_map_single s/gen6_write_pdes/gen6_write_pde Prevent a NULL case when only GGTT is available. (Mika) v8: Rebased after s/page_tables/page_table/. v9: Reworked i915_pte_index and i915_pte_count. Also exercise bitmap allocation here (gen6_alloc_va_range) and fix incorrect write_page_range in i915_gem_restore_gtt_mappings (Mika). Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com> Signed-off-by: Ben Widawsky <ben@bwidawsk.net> Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+) Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-03-16 23:00:56 +07:00
{
return i915_pde_index(addr, GEN6_PDE_SHIFT);
}
static inline struct i915_page_table *
i915_pt_entry(const struct i915_page_directory * const pd,
const unsigned short n)
{
return pd->entry[n];
}
static inline struct i915_page_directory *
i915_pd_entry(const struct i915_page_directory * const pdp,
const unsigned short n)
{
return pdp->entry[n];
}
static inline dma_addr_t
i915_page_dir_dma_addr(const struct i915_ppgtt *ppgtt, const unsigned int n)
{
struct i915_page_dma *pt = ppgtt->pd->entry[n];
return px_dma(pt ?: px_base(&ppgtt->vm.scratch[ppgtt->vm.top]));
}
static inline struct i915_ggtt *
i915_vm_to_ggtt(struct i915_address_space *vm)
{
BUILD_BUG_ON(offsetof(struct i915_ggtt, vm));
GEM_BUG_ON(!i915_is_ggtt(vm));
return container_of(vm, struct i915_ggtt, vm);
}
static inline struct i915_ppgtt *
i915_vm_to_ppgtt(struct i915_address_space *vm)
{
BUILD_BUG_ON(offsetof(struct i915_ppgtt, vm));
GEM_BUG_ON(i915_is_ggtt(vm));
return container_of(vm, struct i915_ppgtt, vm);
}
int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv);
int i915_ggtt_init_hw(struct drm_i915_private *dev_priv);
int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv);
void i915_ggtt_enable_guc(struct i915_ggtt *ggtt);
void i915_ggtt_disable_guc(struct i915_ggtt *ggtt);
int i915_init_ggtt(struct drm_i915_private *dev_priv);
void i915_ggtt_driver_release(struct drm_i915_private *dev_priv);
static inline bool i915_ggtt_has_aperture(const struct i915_ggtt *ggtt)
{
return ggtt->mappable_end > 0;
}
int i915_ppgtt_init_hw(struct intel_gt *gt);
struct i915_ppgtt *i915_ppgtt_create(struct drm_i915_private *dev_priv);
static inline struct i915_address_space *
i915_vm_get(struct i915_address_space *vm)
{
kref_get(&vm->ref);
return vm;
}
void i915_vm_release(struct kref *kref);
static inline void i915_vm_put(struct i915_address_space *vm)
{
kref_put(&vm->ref, i915_vm_release);
}
drm/i915: Pull i915_vma_pin under the vm->mutex Replace the struct_mutex requirement for pinning the i915_vma with the local vm->mutex instead. Note that the vm->mutex is tainted by the shrinker (we require unbinding from inside fs-reclaim) and so we cannot allocate while holding that mutex. Instead we have to preallocate workers to do allocate and apply the PTE updates after we have we reserved their slot in the drm_mm (using fences to order the PTE writes with the GPU work and with later unbind). In adding the asynchronous vma binding, one subtle requirement is to avoid coupling the binding fence into the backing object->resv. That is the asynchronous binding only applies to the vma timeline itself and not to the pages as that is a more global timeline (the binding of one vma does not need to be ordered with another vma, nor does the implicit GEM fencing depend on a vma, only on writes to the backing store). Keeping the vma binding distinct from the backing store timelines is verified by a number of async gem_exec_fence and gem_exec_schedule tests. The way we do this is quite simple, we keep the fence for the vma binding separate and only wait on it as required, and never add it to the obj->resv itself. Another consequence in reducing the locking around the vma is the destruction of the vma is no longer globally serialised by struct_mutex. A natural solution would be to add a kref to i915_vma, but that requires decoupling the reference cycles, possibly by introducing a new i915_mm_pages object that is own by both obj->mm and vma->pages. However, we have not taken that route due to the overshadowing lmem/ttm discussions, and instead play a series of complicated games with trylocks to (hopefully) ensure that only one destruction path is called! v2: Add some commentary, and some helpers to reduce patch churn. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191004134015.13204-4-chris@chris-wilson.co.uk
2019-10-04 20:39:58 +07:00
static inline struct i915_address_space *
i915_vm_open(struct i915_address_space *vm)
{
GEM_BUG_ON(!atomic_read(&vm->open));
atomic_inc(&vm->open);
return i915_vm_get(vm);
}
static inline bool
i915_vm_tryopen(struct i915_address_space *vm)
{
if (atomic_add_unless(&vm->open, 1, 0))
return i915_vm_get(vm);
return false;
}
void __i915_vm_close(struct i915_address_space *vm);
static inline void
i915_vm_close(struct i915_address_space *vm)
{
GEM_BUG_ON(!atomic_read(&vm->open));
if (atomic_dec_and_test(&vm->open))
__i915_vm_close(vm);
i915_vm_put(vm);
}
int gen6_ppgtt_pin(struct i915_ppgtt *base);
void gen6_ppgtt_unpin(struct i915_ppgtt *base);
void gen6_ppgtt_unpin_all(struct i915_ppgtt *base);
void i915_gem_suspend_gtt_mappings(struct drm_i915_private *dev_priv);
void i915_gem_restore_gtt_mappings(struct drm_i915_private *dev_priv);
int __must_check i915_gem_gtt_prepare_pages(struct drm_i915_gem_object *obj,
struct sg_table *pages);
void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj,
struct sg_table *pages);
int i915_gem_gtt_reserve(struct i915_address_space *vm,
struct drm_mm_node *node,
u64 size, u64 offset, unsigned long color,
unsigned int flags);
int i915_gem_gtt_insert(struct i915_address_space *vm,
struct drm_mm_node *node,
u64 size, u64 alignment, unsigned long color,
u64 start, u64 end, unsigned int flags);
/* Flags used by pin/bind&friends. */
#define PIN_NOEVICT BIT_ULL(0)
#define PIN_NOSEARCH BIT_ULL(1)
#define PIN_NONBLOCK BIT_ULL(2)
#define PIN_MAPPABLE BIT_ULL(3)
#define PIN_ZONE_4G BIT_ULL(4)
#define PIN_HIGH BIT_ULL(5)
#define PIN_OFFSET_BIAS BIT_ULL(6)
#define PIN_OFFSET_FIXED BIT_ULL(7)
drm/i915: Pull i915_vma_pin under the vm->mutex Replace the struct_mutex requirement for pinning the i915_vma with the local vm->mutex instead. Note that the vm->mutex is tainted by the shrinker (we require unbinding from inside fs-reclaim) and so we cannot allocate while holding that mutex. Instead we have to preallocate workers to do allocate and apply the PTE updates after we have we reserved their slot in the drm_mm (using fences to order the PTE writes with the GPU work and with later unbind). In adding the asynchronous vma binding, one subtle requirement is to avoid coupling the binding fence into the backing object->resv. That is the asynchronous binding only applies to the vma timeline itself and not to the pages as that is a more global timeline (the binding of one vma does not need to be ordered with another vma, nor does the implicit GEM fencing depend on a vma, only on writes to the backing store). Keeping the vma binding distinct from the backing store timelines is verified by a number of async gem_exec_fence and gem_exec_schedule tests. The way we do this is quite simple, we keep the fence for the vma binding separate and only wait on it as required, and never add it to the obj->resv itself. Another consequence in reducing the locking around the vma is the destruction of the vma is no longer globally serialised by struct_mutex. A natural solution would be to add a kref to i915_vma, but that requires decoupling the reference cycles, possibly by introducing a new i915_mm_pages object that is own by both obj->mm and vma->pages. However, we have not taken that route due to the overshadowing lmem/ttm discussions, and instead play a series of complicated games with trylocks to (hopefully) ensure that only one destruction path is called! v2: Add some commentary, and some helpers to reduce patch churn. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20191004134015.13204-4-chris@chris-wilson.co.uk
2019-10-04 20:39:58 +07:00
#define PIN_UPDATE BIT_ULL(9)
#define PIN_GLOBAL BIT_ULL(10) /* I915_VMA_GLOBAL_BIND */
#define PIN_USER BIT_ULL(11) /* I915_VMA_LOCAL_BIND */
#define PIN_OFFSET_MASK (-I915_GTT_PAGE_SIZE)
#endif