i915_gem_gtt.h 10 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284
  1. /*
  2. * Copyright © 2014 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. * Please try to maintain the following order within this file unless it makes
  24. * sense to do otherwise. From top to bottom:
  25. * 1. typedefs
  26. * 2. #defines, and macros
  27. * 3. structure definitions
  28. * 4. function prototypes
  29. *
  30. * Within each section, please try to order by generation in ascending order,
  31. * from top to bottom (ie. gen6 on the top, gen8 on the bottom).
  32. */
  33. #ifndef __I915_GEM_GTT_H__
  34. #define __I915_GEM_GTT_H__
  35. typedef uint32_t gen6_gtt_pte_t;
  36. typedef uint64_t gen8_gtt_pte_t;
  37. typedef gen8_gtt_pte_t gen8_ppgtt_pde_t;
  38. #define gtt_total_entries(gtt) ((gtt).base.total >> PAGE_SHIFT)
  39. #define I915_PPGTT_PT_ENTRIES (PAGE_SIZE / sizeof(gen6_gtt_pte_t))
  40. /* gen6-hsw has bit 11-4 for physical addr bit 39-32 */
  41. #define GEN6_GTT_ADDR_ENCODE(addr) ((addr) | (((addr) >> 28) & 0xff0))
  42. #define GEN6_PTE_ADDR_ENCODE(addr) GEN6_GTT_ADDR_ENCODE(addr)
  43. #define GEN6_PDE_ADDR_ENCODE(addr) GEN6_GTT_ADDR_ENCODE(addr)
  44. #define GEN6_PTE_CACHE_LLC (2 << 1)
  45. #define GEN6_PTE_UNCACHED (1 << 1)
  46. #define GEN6_PTE_VALID (1 << 0)
  47. #define GEN6_PPGTT_PD_ENTRIES 512
  48. #define GEN6_PD_SIZE (GEN6_PPGTT_PD_ENTRIES * PAGE_SIZE)
  49. #define GEN6_PD_ALIGN (PAGE_SIZE * 16)
  50. #define GEN6_PDE_VALID (1 << 0)
  51. #define GEN7_PTE_CACHE_L3_LLC (3 << 1)
  52. #define BYT_PTE_SNOOPED_BY_CPU_CACHES (1 << 2)
  53. #define BYT_PTE_WRITEABLE (1 << 1)
  54. /* Cacheability Control is a 4-bit value. The low three bits are stored in bits
  55. * 3:1 of the PTE, while the fourth bit is stored in bit 11 of the PTE.
  56. */
  57. #define HSW_CACHEABILITY_CONTROL(bits) ((((bits) & 0x7) << 1) | \
  58. (((bits) & 0x8) << (11 - 3)))
  59. #define HSW_WB_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x2)
  60. #define HSW_WB_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x3)
  61. #define HSW_WB_ELLC_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x8)
  62. #define HSW_WB_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0xb)
  63. #define HSW_WT_ELLC_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x7)
  64. #define HSW_WT_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x6)
  65. #define HSW_PTE_UNCACHED (0)
  66. #define HSW_GTT_ADDR_ENCODE(addr) ((addr) | (((addr) >> 28) & 0x7f0))
  67. #define HSW_PTE_ADDR_ENCODE(addr) HSW_GTT_ADDR_ENCODE(addr)
  68. /* GEN8 legacy style address is defined as a 3 level page table:
  69. * 31:30 | 29:21 | 20:12 | 11:0
  70. * PDPE | PDE | PTE | offset
  71. * The difference as compared to normal x86 3 level page table is the PDPEs are
  72. * programmed via register.
  73. */
  74. #define GEN8_PDPE_SHIFT 30
  75. #define GEN8_PDPE_MASK 0x3
  76. #define GEN8_PDE_SHIFT 21
  77. #define GEN8_PDE_MASK 0x1ff
  78. #define GEN8_PTE_SHIFT 12
  79. #define GEN8_PTE_MASK 0x1ff
  80. #define GEN8_LEGACY_PDPS 4
  81. #define GEN8_PTES_PER_PAGE (PAGE_SIZE / sizeof(gen8_gtt_pte_t))
  82. #define GEN8_PDES_PER_PAGE (PAGE_SIZE / sizeof(gen8_ppgtt_pde_t))
  83. #define PPAT_UNCACHED_INDEX (_PAGE_PWT | _PAGE_PCD)
  84. #define PPAT_CACHED_PDE_INDEX 0 /* WB LLC */
  85. #define PPAT_CACHED_INDEX _PAGE_PAT /* WB LLCeLLC */
  86. #define PPAT_DISPLAY_ELLC_INDEX _PAGE_PCD /* WT eLLC */
  87. #define CHV_PPAT_SNOOP (1<<6)
  88. #define GEN8_PPAT_AGE(x) (x<<4)
  89. #define GEN8_PPAT_LLCeLLC (3<<2)
  90. #define GEN8_PPAT_LLCELLC (2<<2)
  91. #define GEN8_PPAT_LLC (1<<2)
  92. #define GEN8_PPAT_WB (3<<0)
  93. #define GEN8_PPAT_WT (2<<0)
  94. #define GEN8_PPAT_WC (1<<0)
  95. #define GEN8_PPAT_UC (0<<0)
  96. #define GEN8_PPAT_ELLC_OVERRIDE (0<<2)
  97. #define GEN8_PPAT(i, x) ((uint64_t) (x) << ((i) * 8))
  98. enum i915_cache_level;
  99. /**
  100. * A VMA represents a GEM BO that is bound into an address space. Therefore, a
  101. * VMA's presence cannot be guaranteed before binding, or after unbinding the
  102. * object into/from the address space.
  103. *
  104. * To make things as simple as possible (ie. no refcounting), a VMA's lifetime
  105. * will always be <= an objects lifetime. So object refcounting should cover us.
  106. */
  107. struct i915_vma {
  108. struct drm_mm_node node;
  109. struct drm_i915_gem_object *obj;
  110. struct i915_address_space *vm;
  111. /** This object's place on the active/inactive lists */
  112. struct list_head mm_list;
  113. struct list_head vma_link; /* Link in the object's VMA list */
  114. /** This vma's place in the batchbuffer or on the eviction list */
  115. struct list_head exec_list;
  116. /**
  117. * Used for performing relocations during execbuffer insertion.
  118. */
  119. struct hlist_node exec_node;
  120. unsigned long exec_handle;
  121. struct drm_i915_gem_exec_object2 *exec_entry;
  122. /**
  123. * How many users have pinned this object in GTT space. The following
  124. * users can each hold at most one reference: pwrite/pread, pin_ioctl
  125. * (via user_pin_count), execbuffer (objects are not allowed multiple
  126. * times for the same batchbuffer), and the framebuffer code. When
  127. * switching/pageflipping, the framebuffer code has at most two buffers
  128. * pinned per crtc.
  129. *
  130. * In the worst case this is 1 + 1 + 1 + 2*2 = 7. That would fit into 3
  131. * bits with absolutely no headroom. So use 4 bits. */
  132. unsigned int pin_count:4;
  133. #define DRM_I915_GEM_OBJECT_MAX_PIN_COUNT 0xf
  134. /** Unmap an object from an address space. This usually consists of
  135. * setting the valid PTE entries to a reserved scratch page. */
  136. void (*unbind_vma)(struct i915_vma *vma);
  137. /* Map an object into an address space with the given cache flags. */
  138. #define GLOBAL_BIND (1<<0)
  139. void (*bind_vma)(struct i915_vma *vma,
  140. enum i915_cache_level cache_level,
  141. u32 flags);
  142. };
  143. struct i915_address_space {
  144. struct drm_mm mm;
  145. struct drm_device *dev;
  146. struct list_head global_link;
  147. unsigned long start; /* Start offset always 0 for dri2 */
  148. size_t total; /* size addr space maps (ex. 2GB for ggtt) */
  149. struct {
  150. dma_addr_t addr;
  151. struct page *page;
  152. } scratch;
  153. /**
  154. * List of objects currently involved in rendering.
  155. *
  156. * Includes buffers having the contents of their GPU caches
  157. * flushed, not necessarily primitives. last_rendering_seqno
  158. * represents when the rendering involved will be completed.
  159. *
  160. * A reference is held on the buffer while on this list.
  161. */
  162. struct list_head active_list;
  163. /**
  164. * LRU list of objects which are not in the ringbuffer and
  165. * are ready to unbind, but are still in the GTT.
  166. *
  167. * last_rendering_seqno is 0 while an object is in this list.
  168. *
  169. * A reference is not held on the buffer while on this list,
  170. * as merely being GTT-bound shouldn't prevent its being
  171. * freed, and we'll pull it off the list in the free path.
  172. */
  173. struct list_head inactive_list;
  174. /* FIXME: Need a more generic return type */
  175. gen6_gtt_pte_t (*pte_encode)(dma_addr_t addr,
  176. enum i915_cache_level level,
  177. bool valid); /* Create a valid PTE */
  178. void (*clear_range)(struct i915_address_space *vm,
  179. uint64_t start,
  180. uint64_t length,
  181. bool use_scratch);
  182. void (*insert_entries)(struct i915_address_space *vm,
  183. struct sg_table *st,
  184. uint64_t start,
  185. enum i915_cache_level cache_level);
  186. void (*cleanup)(struct i915_address_space *vm);
  187. };
  188. /* The Graphics Translation Table is the way in which GEN hardware translates a
  189. * Graphics Virtual Address into a Physical Address. In addition to the normal
  190. * collateral associated with any va->pa translations GEN hardware also has a
  191. * portion of the GTT which can be mapped by the CPU and remain both coherent
  192. * and correct (in cases like swizzling). That region is referred to as GMADR in
  193. * the spec.
  194. */
  195. struct i915_gtt {
  196. struct i915_address_space base;
  197. size_t stolen_size; /* Total size of stolen memory */
  198. unsigned long mappable_end; /* End offset that we can CPU map */
  199. struct io_mapping *mappable; /* Mapping to our CPU mappable region */
  200. phys_addr_t mappable_base; /* PA of our GMADR */
  201. /** "Graphics Stolen Memory" holds the global PTEs */
  202. void __iomem *gsm;
  203. bool do_idle_maps;
  204. int mtrr;
  205. /* global gtt ops */
  206. int (*gtt_probe)(struct drm_device *dev, size_t *gtt_total,
  207. size_t *stolen, phys_addr_t *mappable_base,
  208. unsigned long *mappable_end);
  209. };
  210. struct i915_hw_ppgtt {
  211. struct i915_address_space base;
  212. struct kref ref;
  213. struct drm_mm_node node;
  214. unsigned num_pd_entries;
  215. unsigned num_pd_pages; /* gen8+ */
  216. union {
  217. struct page **pt_pages;
  218. struct page **gen8_pt_pages[GEN8_LEGACY_PDPS];
  219. };
  220. struct page *pd_pages;
  221. union {
  222. uint32_t pd_offset;
  223. dma_addr_t pd_dma_addr[GEN8_LEGACY_PDPS];
  224. };
  225. union {
  226. dma_addr_t *pt_dma_addr;
  227. dma_addr_t *gen8_pt_dma_addr[4];
  228. };
  229. struct intel_context *ctx;
  230. int (*enable)(struct i915_hw_ppgtt *ppgtt);
  231. int (*switch_mm)(struct i915_hw_ppgtt *ppgtt,
  232. struct intel_engine_cs *ring,
  233. bool synchronous);
  234. void (*debug_dump)(struct i915_hw_ppgtt *ppgtt, struct seq_file *m);
  235. };
  236. int i915_gem_gtt_init(struct drm_device *dev);
  237. void i915_gem_init_global_gtt(struct drm_device *dev);
  238. void i915_gem_setup_global_gtt(struct drm_device *dev, unsigned long start,
  239. unsigned long mappable_end, unsigned long end);
  240. bool intel_enable_ppgtt(struct drm_device *dev, bool full);
  241. int i915_gem_init_ppgtt(struct drm_device *dev, struct i915_hw_ppgtt *ppgtt);
  242. void i915_check_and_clear_faults(struct drm_device *dev);
  243. void i915_gem_suspend_gtt_mappings(struct drm_device *dev);
  244. void i915_gem_restore_gtt_mappings(struct drm_device *dev);
  245. int __must_check i915_gem_gtt_prepare_object(struct drm_i915_gem_object *obj);
  246. void i915_gem_gtt_finish_object(struct drm_i915_gem_object *obj);
  247. #endif