linux/drivers/gpu/drm/i915/i915_vma.h
<<
>>
Prefs
   1/*
   2 * Copyright © 2016 Intel Corporation
   3 *
   4 * Permission is hereby granted, free of charge, to any person obtaining a
   5 * copy of this software and associated documentation files (the "Software"),
   6 * to deal in the Software without restriction, including without limitation
   7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8 * and/or sell copies of the Software, and to permit persons to whom the
   9 * Software is furnished to do so, subject to the following conditions:
  10 *
  11 * The above copyright notice and this permission notice (including the next
  12 * paragraph) shall be included in all copies or substantial portions of the
  13 * Software.
  14 *
  15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21 * IN THE SOFTWARE.
  22 *
  23 */
  24
  25#ifndef __I915_VMA_H__
  26#define __I915_VMA_H__
  27
  28#include <linux/io-mapping.h>
  29#include <linux/rbtree.h>
  30
  31#include <drm/drm_mm.h>
  32
  33#include "i915_gem_gtt.h"
  34#include "i915_gem_fence_reg.h"
  35#include "gem/i915_gem_object.h"
  36
  37#include "i915_active.h"
  38#include "i915_request.h"
  39
  40enum i915_cache_level;
  41
  42/**
  43 * DOC: Virtual Memory Address
  44 *
  45 * A VMA represents a GEM BO that is bound into an address space. Therefore, a
  46 * VMA's presence cannot be guaranteed before binding, or after unbinding the
  47 * object into/from the address space.
  48 *
  49 * To make things as simple as possible (ie. no refcounting), a VMA's lifetime
  50 * will always be <= an objects lifetime. So object refcounting should cover us.
  51 */
  52struct i915_vma {
  53        struct drm_mm_node node;
  54        struct drm_i915_gem_object *obj;
  55        struct i915_address_space *vm;
  56        const struct i915_vma_ops *ops;
  57        struct i915_fence_reg *fence;
  58        struct reservation_object *resv; /** Alias of obj->resv */
  59        struct sg_table *pages;
  60        void __iomem *iomap;
  61        void *private; /* owned by creator */
  62        u64 size;
  63        u64 display_alignment;
  64        struct i915_page_sizes page_sizes;
  65
  66        u32 fence_size;
  67        u32 fence_alignment;
  68
  69        /**
  70         * Count of the number of times this vma has been opened by different
  71         * handles (but same file) for execbuf, i.e. the number of aliases
  72         * that exist in the ctx->handle_vmas LUT for this vma.
  73         */
  74        atomic_t open_count;
  75        unsigned long flags;
  76        /**
  77         * How many users have pinned this object in GTT space.
  78         *
  79         * This is a tightly bound, fairly small number of users, so we
  80         * stuff inside the flags field so that we can both check for overflow
  81         * and detect a no-op i915_vma_pin() in a single check, while also
  82         * pinning the vma.
  83         *
  84         * The worst case display setup would have the same vma pinned for
  85         * use on each plane on each crtc, while also building the next atomic
  86         * state and holding a pin for the length of the cleanup queue. In the
  87         * future, the flip queue may be increased from 1.
  88         * Estimated worst case: 3 [qlen] * 4 [max crtcs] * 7 [max planes] = 84
  89         *
  90         * For GEM, the number of concurrent users for pwrite/pread is
  91         * unbounded. For execbuffer, it is currently one but will in future
  92         * be extended to allow multiple clients to pin vma concurrently.
  93         *
  94         * We also use suballocated pages, with each suballocation claiming
  95         * its own pin on the shared vma. At present, this is limited to
  96         * exclusive cachelines of a single page, so a maximum of 64 possible
  97         * users.
  98         */
  99#define I915_VMA_PIN_MASK 0xff
 100#define I915_VMA_PIN_OVERFLOW   BIT(8)
 101
 102        /** Flags and address space this VMA is bound to */
 103#define I915_VMA_GLOBAL_BIND    BIT(9)
 104#define I915_VMA_LOCAL_BIND     BIT(10)
 105#define I915_VMA_BIND_MASK (I915_VMA_GLOBAL_BIND | I915_VMA_LOCAL_BIND | I915_VMA_PIN_OVERFLOW)
 106
 107#define I915_VMA_GGTT           BIT(11)
 108#define I915_VMA_CAN_FENCE      BIT(12)
 109#define I915_VMA_USERFAULT_BIT  13
 110#define I915_VMA_USERFAULT      BIT(I915_VMA_USERFAULT_BIT)
 111#define I915_VMA_GGTT_WRITE     BIT(14)
 112
 113        struct i915_active active;
 114        struct i915_active_request last_fence;
 115
 116        /**
 117         * Support different GGTT views into the same object.
 118         * This means there can be multiple VMA mappings per object and per VM.
 119         * i915_ggtt_view_type is used to distinguish between those entries.
 120         * The default one of zero (I915_GGTT_VIEW_NORMAL) is default and also
 121         * assumed in GEM functions which take no ggtt view parameter.
 122         */
 123        struct i915_ggtt_view ggtt_view;
 124
 125        /** This object's place on the active/inactive lists */
 126        struct list_head vm_link;
 127
 128        struct list_head obj_link; /* Link in the object's VMA list */
 129        struct rb_node obj_node;
 130        struct hlist_node obj_hash;
 131
 132        /** This vma's place in the execbuf reservation list */
 133        struct list_head exec_link;
 134        struct list_head reloc_link;
 135
 136        /** This vma's place in the eviction list */
 137        struct list_head evict_link;
 138
 139        struct list_head closed_link;
 140
 141        /**
 142         * Used for performing relocations during execbuffer insertion.
 143         */
 144        unsigned int *exec_flags;
 145        struct hlist_node exec_node;
 146        u32 exec_handle;
 147};
 148
 149struct i915_vma *
 150i915_vma_instance(struct drm_i915_gem_object *obj,
 151                  struct i915_address_space *vm,
 152                  const struct i915_ggtt_view *view);
 153
 154void i915_vma_unpin_and_release(struct i915_vma **p_vma, unsigned int flags);
 155#define I915_VMA_RELEASE_MAP BIT(0)
 156
 157static inline bool i915_vma_is_active(const struct i915_vma *vma)
 158{
 159        return !i915_active_is_idle(&vma->active);
 160}
 161
 162int __must_check i915_vma_move_to_active(struct i915_vma *vma,
 163                                         struct i915_request *rq,
 164                                         unsigned int flags);
 165
 166static inline bool i915_vma_is_ggtt(const struct i915_vma *vma)
 167{
 168        return vma->flags & I915_VMA_GGTT;
 169}
 170
 171static inline bool i915_vma_has_ggtt_write(const struct i915_vma *vma)
 172{
 173        return vma->flags & I915_VMA_GGTT_WRITE;
 174}
 175
 176static inline void i915_vma_set_ggtt_write(struct i915_vma *vma)
 177{
 178        GEM_BUG_ON(!i915_vma_is_ggtt(vma));
 179        vma->flags |= I915_VMA_GGTT_WRITE;
 180}
 181
 182static inline void i915_vma_unset_ggtt_write(struct i915_vma *vma)
 183{
 184        vma->flags &= ~I915_VMA_GGTT_WRITE;
 185}
 186
 187void i915_vma_flush_writes(struct i915_vma *vma);
 188
 189static inline bool i915_vma_is_map_and_fenceable(const struct i915_vma *vma)
 190{
 191        return vma->flags & I915_VMA_CAN_FENCE;
 192}
 193
 194static inline bool i915_vma_set_userfault(struct i915_vma *vma)
 195{
 196        GEM_BUG_ON(!i915_vma_is_map_and_fenceable(vma));
 197        return __test_and_set_bit(I915_VMA_USERFAULT_BIT, &vma->flags);
 198}
 199
 200static inline void i915_vma_unset_userfault(struct i915_vma *vma)
 201{
 202        return __clear_bit(I915_VMA_USERFAULT_BIT, &vma->flags);
 203}
 204
 205static inline bool i915_vma_has_userfault(const struct i915_vma *vma)
 206{
 207        return test_bit(I915_VMA_USERFAULT_BIT, &vma->flags);
 208}
 209
 210static inline bool i915_vma_is_closed(const struct i915_vma *vma)
 211{
 212        return !list_empty(&vma->closed_link);
 213}
 214
 215static inline u32 i915_ggtt_offset(const struct i915_vma *vma)
 216{
 217        GEM_BUG_ON(!i915_vma_is_ggtt(vma));
 218        GEM_BUG_ON(!vma->node.allocated);
 219        GEM_BUG_ON(upper_32_bits(vma->node.start));
 220        GEM_BUG_ON(upper_32_bits(vma->node.start + vma->node.size - 1));
 221        return lower_32_bits(vma->node.start);
 222}
 223
 224static inline u32 i915_ggtt_pin_bias(struct i915_vma *vma)
 225{
 226        return i915_vm_to_ggtt(vma->vm)->pin_bias;
 227}
 228
 229static inline struct i915_vma *i915_vma_get(struct i915_vma *vma)
 230{
 231        i915_gem_object_get(vma->obj);
 232        return vma;
 233}
 234
 235static inline void i915_vma_put(struct i915_vma *vma)
 236{
 237        i915_gem_object_put(vma->obj);
 238}
 239
 240static __always_inline ptrdiff_t ptrdiff(const void *a, const void *b)
 241{
 242        return a - b;
 243}
 244
 245static inline long
 246i915_vma_compare(struct i915_vma *vma,
 247                 struct i915_address_space *vm,
 248                 const struct i915_ggtt_view *view)
 249{
 250        ptrdiff_t cmp;
 251
 252        GEM_BUG_ON(view && !i915_is_ggtt(vm));
 253
 254        cmp = ptrdiff(vma->vm, vm);
 255        if (cmp)
 256                return cmp;
 257
 258        BUILD_BUG_ON(I915_GGTT_VIEW_NORMAL != 0);
 259        cmp = vma->ggtt_view.type;
 260        if (!view)
 261                return cmp;
 262
 263        cmp -= view->type;
 264        if (cmp)
 265                return cmp;
 266
 267        assert_i915_gem_gtt_types();
 268
 269        /* ggtt_view.type also encodes its size so that we both distinguish
 270         * different views using it as a "type" and also use a compact (no
 271         * accessing of uninitialised padding bytes) memcmp without storing
 272         * an extra parameter or adding more code.
 273         *
 274         * To ensure that the memcmp is valid for all branches of the union,
 275         * even though the code looks like it is just comparing one branch,
 276         * we assert above that all branches have the same address, and that
 277         * each branch has a unique type/size.
 278         */
 279        BUILD_BUG_ON(I915_GGTT_VIEW_NORMAL >= I915_GGTT_VIEW_PARTIAL);
 280        BUILD_BUG_ON(I915_GGTT_VIEW_PARTIAL >= I915_GGTT_VIEW_ROTATED);
 281        BUILD_BUG_ON(I915_GGTT_VIEW_ROTATED >= I915_GGTT_VIEW_REMAPPED);
 282        BUILD_BUG_ON(offsetof(typeof(*view), rotated) !=
 283                     offsetof(typeof(*view), partial));
 284        BUILD_BUG_ON(offsetof(typeof(*view), rotated) !=
 285                     offsetof(typeof(*view), remapped));
 286        return memcmp(&vma->ggtt_view.partial, &view->partial, view->type);
 287}
 288
 289int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
 290                  u32 flags);
 291bool i915_gem_valid_gtt_space(struct i915_vma *vma, unsigned long cache_level);
 292bool i915_vma_misplaced(const struct i915_vma *vma,
 293                        u64 size, u64 alignment, u64 flags);
 294void __i915_vma_set_map_and_fenceable(struct i915_vma *vma);
 295void i915_vma_revoke_mmap(struct i915_vma *vma);
 296int __must_check i915_vma_unbind(struct i915_vma *vma);
 297void i915_vma_unlink_ctx(struct i915_vma *vma);
 298void i915_vma_close(struct i915_vma *vma);
 299void i915_vma_reopen(struct i915_vma *vma);
 300void i915_vma_destroy(struct i915_vma *vma);
 301
 302#define assert_vma_held(vma) reservation_object_assert_held((vma)->resv)
 303
 304static inline void i915_vma_lock(struct i915_vma *vma)
 305{
 306        reservation_object_lock(vma->resv, NULL);
 307}
 308
 309static inline void i915_vma_unlock(struct i915_vma *vma)
 310{
 311        reservation_object_unlock(vma->resv);
 312}
 313
 314int __i915_vma_do_pin(struct i915_vma *vma,
 315                      u64 size, u64 alignment, u64 flags);
 316static inline int __must_check
 317i915_vma_pin(struct i915_vma *vma, u64 size, u64 alignment, u64 flags)
 318{
 319        BUILD_BUG_ON(PIN_MBZ != I915_VMA_PIN_OVERFLOW);
 320        BUILD_BUG_ON(PIN_GLOBAL != I915_VMA_GLOBAL_BIND);
 321        BUILD_BUG_ON(PIN_USER != I915_VMA_LOCAL_BIND);
 322
 323        /* Pin early to prevent the shrinker/eviction logic from destroying
 324         * our vma as we insert and bind.
 325         */
 326        if (likely(((++vma->flags ^ flags) & I915_VMA_BIND_MASK) == 0)) {
 327                GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
 328                GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags));
 329                return 0;
 330        }
 331
 332        return __i915_vma_do_pin(vma, size, alignment, flags);
 333}
 334
 335static inline int i915_vma_pin_count(const struct i915_vma *vma)
 336{
 337        return vma->flags & I915_VMA_PIN_MASK;
 338}
 339
 340static inline bool i915_vma_is_pinned(const struct i915_vma *vma)
 341{
 342        return i915_vma_pin_count(vma);
 343}
 344
 345static inline void __i915_vma_pin(struct i915_vma *vma)
 346{
 347        vma->flags++;
 348        GEM_BUG_ON(vma->flags & I915_VMA_PIN_OVERFLOW);
 349}
 350
 351static inline void __i915_vma_unpin(struct i915_vma *vma)
 352{
 353        vma->flags--;
 354}
 355
 356static inline void i915_vma_unpin(struct i915_vma *vma)
 357{
 358        GEM_BUG_ON(!i915_vma_is_pinned(vma));
 359        GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
 360        __i915_vma_unpin(vma);
 361}
 362
 363static inline bool i915_vma_is_bound(const struct i915_vma *vma,
 364                                     unsigned int where)
 365{
 366        return vma->flags & where;
 367}
 368
 369/**
 370 * i915_vma_pin_iomap - calls ioremap_wc to map the GGTT VMA via the aperture
 371 * @vma: VMA to iomap
 372 *
 373 * The passed in VMA has to be pinned in the global GTT mappable region.
 374 * An extra pinning of the VMA is acquired for the return iomapping,
 375 * the caller must call i915_vma_unpin_iomap to relinquish the pinning
 376 * after the iomapping is no longer required.
 377 *
 378 * Callers must hold the struct_mutex.
 379 *
 380 * Returns a valid iomapped pointer or ERR_PTR.
 381 */
 382void __iomem *i915_vma_pin_iomap(struct i915_vma *vma);
 383#define IO_ERR_PTR(x) ((void __iomem *)ERR_PTR(x))
 384
 385/**
 386 * i915_vma_unpin_iomap - unpins the mapping returned from i915_vma_iomap
 387 * @vma: VMA to unpin
 388 *
 389 * Unpins the previously iomapped VMA from i915_vma_pin_iomap().
 390 *
 391 * Callers must hold the struct_mutex. This function is only valid to be
 392 * called on a VMA previously iomapped by the caller with i915_vma_pin_iomap().
 393 */
 394void i915_vma_unpin_iomap(struct i915_vma *vma);
 395
 396static inline struct page *i915_vma_first_page(struct i915_vma *vma)
 397{
 398        GEM_BUG_ON(!vma->pages);
 399        return sg_page(vma->pages->sgl);
 400}
 401
 402/**
 403 * i915_vma_pin_fence - pin fencing state
 404 * @vma: vma to pin fencing for
 405 *
 406 * This pins the fencing state (whether tiled or untiled) to make sure the
 407 * vma (and its object) is ready to be used as a scanout target. Fencing
 408 * status must be synchronize first by calling i915_vma_get_fence():
 409 *
 410 * The resulting fence pin reference must be released again with
 411 * i915_vma_unpin_fence().
 412 *
 413 * Returns:
 414 *
 415 * True if the vma has a fence, false otherwise.
 416 */
 417int i915_vma_pin_fence(struct i915_vma *vma);
 418int __must_check i915_vma_put_fence(struct i915_vma *vma);
 419
 420static inline void __i915_vma_unpin_fence(struct i915_vma *vma)
 421{
 422        GEM_BUG_ON(vma->fence->pin_count <= 0);
 423        vma->fence->pin_count--;
 424}
 425
 426/**
 427 * i915_vma_unpin_fence - unpin fencing state
 428 * @vma: vma to unpin fencing for
 429 *
 430 * This releases the fence pin reference acquired through
 431 * i915_vma_pin_fence. It will handle both objects with and without an
 432 * attached fence correctly, callers do not need to distinguish this.
 433 */
 434static inline void
 435i915_vma_unpin_fence(struct i915_vma *vma)
 436{
 437        /* lockdep_assert_held(&vma->vm->i915->drm.struct_mutex); */
 438        if (vma->fence)
 439                __i915_vma_unpin_fence(vma);
 440}
 441
 442void i915_vma_parked(struct drm_i915_private *i915);
 443
 444#define for_each_until(cond) if (cond) break; else
 445
 446/**
 447 * for_each_ggtt_vma - Iterate over the GGTT VMA belonging to an object.
 448 * @V: the #i915_vma iterator
 449 * @OBJ: the #drm_i915_gem_object
 450 *
 451 * GGTT VMA are placed at the being of the object's vma_list, see
 452 * vma_create(), so we can stop our walk as soon as we see a ppgtt VMA,
 453 * or the list is empty ofc.
 454 */
 455#define for_each_ggtt_vma(V, OBJ) \
 456        list_for_each_entry(V, &(OBJ)->vma.list, obj_link)              \
 457                for_each_until(!i915_vma_is_ggtt(V))
 458
 459struct i915_vma *i915_vma_alloc(void);
 460void i915_vma_free(struct i915_vma *vma);
 461
 462#endif
 463