Commit | Line | Data |
---|---|---|
0260c420 BW |
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 | ||
34 | #ifndef __I915_GEM_GTT_H__ | |
35 | #define __I915_GEM_GTT_H__ | |
36 | ||
8ef8561f CW |
37 | #include <linux/io-mapping.h> |
38 | ||
4d884705 DV |
39 | struct drm_i915_file_private; |
40 | ||
07749ef3 MT |
41 | typedef uint32_t gen6_pte_t; |
42 | typedef uint64_t gen8_pte_t; | |
43 | typedef uint64_t gen8_pde_t; | |
762d9936 MT |
44 | typedef uint64_t gen8_ppgtt_pdpe_t; |
45 | typedef uint64_t gen8_ppgtt_pml4e_t; | |
0260c420 | 46 | |
72e96d64 | 47 | #define ggtt_total_entries(ggtt) ((ggtt)->base.total >> PAGE_SHIFT) |
0260c420 | 48 | |
0260c420 BW |
49 | /* gen6-hsw has bit 11-4 for physical addr bit 39-32 */ |
50 | #define GEN6_GTT_ADDR_ENCODE(addr) ((addr) | (((addr) >> 28) & 0xff0)) | |
51 | #define GEN6_PTE_ADDR_ENCODE(addr) GEN6_GTT_ADDR_ENCODE(addr) | |
52 | #define GEN6_PDE_ADDR_ENCODE(addr) GEN6_GTT_ADDR_ENCODE(addr) | |
53 | #define GEN6_PTE_CACHE_LLC (2 << 1) | |
54 | #define GEN6_PTE_UNCACHED (1 << 1) | |
55 | #define GEN6_PTE_VALID (1 << 0) | |
56 | ||
07749ef3 MT |
57 | #define I915_PTES(pte_len) (PAGE_SIZE / (pte_len)) |
58 | #define I915_PTE_MASK(pte_len) (I915_PTES(pte_len) - 1) | |
59 | #define I915_PDES 512 | |
60 | #define I915_PDE_MASK (I915_PDES - 1) | |
678d96fb | 61 | #define NUM_PTE(pde_shift) (1 << (pde_shift - PAGE_SHIFT)) |
07749ef3 MT |
62 | |
63 | #define GEN6_PTES I915_PTES(sizeof(gen6_pte_t)) | |
64 | #define GEN6_PD_SIZE (I915_PDES * PAGE_SIZE) | |
0260c420 | 65 | #define GEN6_PD_ALIGN (PAGE_SIZE * 16) |
678d96fb | 66 | #define GEN6_PDE_SHIFT 22 |
0260c420 BW |
67 | #define GEN6_PDE_VALID (1 << 0) |
68 | ||
69 | #define GEN7_PTE_CACHE_L3_LLC (3 << 1) | |
70 | ||
71 | #define BYT_PTE_SNOOPED_BY_CPU_CACHES (1 << 2) | |
72 | #define BYT_PTE_WRITEABLE (1 << 1) | |
73 | ||
74 | /* Cacheability Control is a 4-bit value. The low three bits are stored in bits | |
75 | * 3:1 of the PTE, while the fourth bit is stored in bit 11 of the PTE. | |
76 | */ | |
77 | #define HSW_CACHEABILITY_CONTROL(bits) ((((bits) & 0x7) << 1) | \ | |
78 | (((bits) & 0x8) << (11 - 3))) | |
79 | #define HSW_WB_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x2) | |
80 | #define HSW_WB_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x3) | |
81 | #define HSW_WB_ELLC_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x8) | |
82 | #define HSW_WB_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0xb) | |
83 | #define HSW_WT_ELLC_LLC_AGE3 HSW_CACHEABILITY_CONTROL(0x7) | |
84 | #define HSW_WT_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x6) | |
85 | #define HSW_PTE_UNCACHED (0) | |
86 | #define HSW_GTT_ADDR_ENCODE(addr) ((addr) | (((addr) >> 28) & 0x7f0)) | |
87 | #define HSW_PTE_ADDR_ENCODE(addr) HSW_GTT_ADDR_ENCODE(addr) | |
88 | ||
89 | /* GEN8 legacy style address is defined as a 3 level page table: | |
90 | * 31:30 | 29:21 | 20:12 | 11:0 | |
91 | * PDPE | PDE | PTE | offset | |
92 | * The difference as compared to normal x86 3 level page table is the PDPEs are | |
93 | * programmed via register. | |
81ba8aef MT |
94 | * |
95 | * GEN8 48b legacy style address is defined as a 4 level page table: | |
96 | * 47:39 | 38:30 | 29:21 | 20:12 | 11:0 | |
97 | * PML4E | PDPE | PDE | PTE | offset | |
0260c420 | 98 | */ |
81ba8aef MT |
99 | #define GEN8_PML4ES_PER_PML4 512 |
100 | #define GEN8_PML4E_SHIFT 39 | |
762d9936 | 101 | #define GEN8_PML4E_MASK (GEN8_PML4ES_PER_PML4 - 1) |
0260c420 | 102 | #define GEN8_PDPE_SHIFT 30 |
81ba8aef MT |
103 | /* NB: GEN8_PDPE_MASK is untrue for 32b platforms, but it has no impact on 32b page |
104 | * tables */ | |
105 | #define GEN8_PDPE_MASK 0x1ff | |
0260c420 BW |
106 | #define GEN8_PDE_SHIFT 21 |
107 | #define GEN8_PDE_MASK 0x1ff | |
108 | #define GEN8_PTE_SHIFT 12 | |
109 | #define GEN8_PTE_MASK 0x1ff | |
76643600 | 110 | #define GEN8_LEGACY_PDPES 4 |
07749ef3 | 111 | #define GEN8_PTES I915_PTES(sizeof(gen8_pte_t)) |
0260c420 | 112 | |
81ba8aef MT |
113 | #define I915_PDPES_PER_PDP(dev) (USES_FULL_48BIT_PPGTT(dev) ?\ |
114 | GEN8_PML4ES_PER_PML4 : GEN8_LEGACY_PDPES) | |
6ac18502 | 115 | |
0260c420 BW |
116 | #define PPAT_UNCACHED_INDEX (_PAGE_PWT | _PAGE_PCD) |
117 | #define PPAT_CACHED_PDE_INDEX 0 /* WB LLC */ | |
118 | #define PPAT_CACHED_INDEX _PAGE_PAT /* WB LLCeLLC */ | |
119 | #define PPAT_DISPLAY_ELLC_INDEX _PAGE_PCD /* WT eLLC */ | |
120 | ||
ee0ce478 | 121 | #define CHV_PPAT_SNOOP (1<<6) |
0260c420 BW |
122 | #define GEN8_PPAT_AGE(x) (x<<4) |
123 | #define GEN8_PPAT_LLCeLLC (3<<2) | |
124 | #define GEN8_PPAT_LLCELLC (2<<2) | |
125 | #define GEN8_PPAT_LLC (1<<2) | |
126 | #define GEN8_PPAT_WB (3<<0) | |
127 | #define GEN8_PPAT_WT (2<<0) | |
128 | #define GEN8_PPAT_WC (1<<0) | |
129 | #define GEN8_PPAT_UC (0<<0) | |
130 | #define GEN8_PPAT_ELLC_OVERRIDE (0<<2) | |
131 | #define GEN8_PPAT(i, x) ((uint64_t) (x) << ((i) * 8)) | |
132 | ||
fe14d5f4 TU |
133 | enum i915_ggtt_view_type { |
134 | I915_GGTT_VIEW_NORMAL = 0, | |
8bd7ef16 JL |
135 | I915_GGTT_VIEW_ROTATED, |
136 | I915_GGTT_VIEW_PARTIAL, | |
50470bb0 TU |
137 | }; |
138 | ||
139 | struct intel_rotation_info { | |
89e3e142 | 140 | unsigned int uv_offset; |
50470bb0 | 141 | uint32_t pixel_format; |
dedf278c | 142 | unsigned int uv_start_page; |
1663b9d6 VS |
143 | struct { |
144 | /* tiles */ | |
145 | unsigned int width, height; | |
146 | } plane[2]; | |
fe14d5f4 TU |
147 | }; |
148 | ||
149 | struct i915_ggtt_view { | |
150 | enum i915_ggtt_view_type type; | |
151 | ||
8bd7ef16 JL |
152 | union { |
153 | struct { | |
088e0df4 | 154 | u64 offset; |
8bd7ef16 JL |
155 | unsigned int size; |
156 | } partial; | |
7723f47d | 157 | struct intel_rotation_info rotated; |
8bd7ef16 JL |
158 | } params; |
159 | ||
fe14d5f4 TU |
160 | struct sg_table *pages; |
161 | }; | |
162 | ||
163 | extern const struct i915_ggtt_view i915_ggtt_view_normal; | |
9abc4648 | 164 | extern const struct i915_ggtt_view i915_ggtt_view_rotated; |
fe14d5f4 | 165 | |
0260c420 | 166 | enum i915_cache_level; |
fe14d5f4 | 167 | |
0260c420 BW |
168 | /** |
169 | * A VMA represents a GEM BO that is bound into an address space. Therefore, a | |
170 | * VMA's presence cannot be guaranteed before binding, or after unbinding the | |
171 | * object into/from the address space. | |
172 | * | |
173 | * To make things as simple as possible (ie. no refcounting), a VMA's lifetime | |
174 | * will always be <= an objects lifetime. So object refcounting should cover us. | |
175 | */ | |
176 | struct i915_vma { | |
177 | struct drm_mm_node node; | |
178 | struct drm_i915_gem_object *obj; | |
179 | struct i915_address_space *vm; | |
8ef8561f | 180 | void __iomem *iomap; |
0260c420 | 181 | |
aff43766 TU |
182 | /** Flags and address space this VMA is bound to */ |
183 | #define GLOBAL_BIND (1<<0) | |
184 | #define LOCAL_BIND (1<<1) | |
aff43766 | 185 | unsigned int bound : 4; |
596c5923 | 186 | bool is_ggtt : 1; |
aff43766 | 187 | |
fe14d5f4 TU |
188 | /** |
189 | * Support different GGTT views into the same object. | |
190 | * This means there can be multiple VMA mappings per object and per VM. | |
191 | * i915_ggtt_view_type is used to distinguish between those entries. | |
192 | * The default one of zero (I915_GGTT_VIEW_NORMAL) is default and also | |
193 | * assumed in GEM functions which take no ggtt view parameter. | |
194 | */ | |
195 | struct i915_ggtt_view ggtt_view; | |
196 | ||
0260c420 | 197 | /** This object's place on the active/inactive lists */ |
1c7f4bca | 198 | struct list_head vm_link; |
0260c420 | 199 | |
1c7f4bca | 200 | struct list_head obj_link; /* Link in the object's VMA list */ |
0260c420 BW |
201 | |
202 | /** This vma's place in the batchbuffer or on the eviction list */ | |
203 | struct list_head exec_list; | |
204 | ||
205 | /** | |
206 | * Used for performing relocations during execbuffer insertion. | |
207 | */ | |
208 | struct hlist_node exec_node; | |
209 | unsigned long exec_handle; | |
210 | struct drm_i915_gem_exec_object2 *exec_entry; | |
211 | ||
212 | /** | |
213 | * How many users have pinned this object in GTT space. The following | |
4feb7659 DV |
214 | * users can each hold at most one reference: pwrite/pread, execbuffer |
215 | * (objects are not allowed multiple times for the same batchbuffer), | |
216 | * and the framebuffer code. When switching/pageflipping, the | |
217 | * framebuffer code has at most two buffers pinned per crtc. | |
0260c420 BW |
218 | * |
219 | * In the worst case this is 1 + 1 + 1 + 2*2 = 7. That would fit into 3 | |
220 | * bits with absolutely no headroom. So use 4 bits. */ | |
221 | unsigned int pin_count:4; | |
222 | #define DRM_I915_GEM_OBJECT_MAX_PIN_COUNT 0xf | |
0260c420 BW |
223 | }; |
224 | ||
44159ddb | 225 | struct i915_page_dma { |
d7b3de91 | 226 | struct page *page; |
44159ddb MK |
227 | union { |
228 | dma_addr_t daddr; | |
229 | ||
230 | /* For gen6/gen7 only. This is the offset in the GGTT | |
231 | * where the page directory entries for PPGTT begin | |
232 | */ | |
233 | uint32_t ggtt_offset; | |
234 | }; | |
235 | }; | |
236 | ||
567047be MK |
237 | #define px_base(px) (&(px)->base) |
238 | #define px_page(px) (px_base(px)->page) | |
239 | #define px_dma(px) (px_base(px)->daddr) | |
240 | ||
c114f76a MK |
241 | struct i915_page_scratch { |
242 | struct i915_page_dma base; | |
243 | }; | |
244 | ||
44159ddb MK |
245 | struct i915_page_table { |
246 | struct i915_page_dma base; | |
678d96fb BW |
247 | |
248 | unsigned long *used_ptes; | |
d7b3de91 BW |
249 | }; |
250 | ||
ec565b3c | 251 | struct i915_page_directory { |
44159ddb | 252 | struct i915_page_dma base; |
7324cc04 | 253 | |
33c8819f | 254 | unsigned long *used_pdes; |
ec565b3c | 255 | struct i915_page_table *page_table[I915_PDES]; /* PDEs */ |
d7b3de91 BW |
256 | }; |
257 | ||
ec565b3c | 258 | struct i915_page_directory_pointer { |
6ac18502 MT |
259 | struct i915_page_dma base; |
260 | ||
261 | unsigned long *used_pdpes; | |
262 | struct i915_page_directory **page_directory; | |
d7b3de91 BW |
263 | }; |
264 | ||
81ba8aef MT |
265 | struct i915_pml4 { |
266 | struct i915_page_dma base; | |
267 | ||
268 | DECLARE_BITMAP(used_pml4es, GEN8_PML4ES_PER_PML4); | |
269 | struct i915_page_directory_pointer *pdps[GEN8_PML4ES_PER_PML4]; | |
270 | }; | |
271 | ||
0260c420 BW |
272 | struct i915_address_space { |
273 | struct drm_mm mm; | |
274 | struct drm_device *dev; | |
275 | struct list_head global_link; | |
c44ef60e MK |
276 | u64 start; /* Start offset always 0 for dri2 */ |
277 | u64 total; /* size addr space maps (ex. 2GB for ggtt) */ | |
0260c420 | 278 | |
596c5923 CW |
279 | bool is_ggtt; |
280 | ||
c114f76a | 281 | struct i915_page_scratch *scratch_page; |
79ab9370 MK |
282 | struct i915_page_table *scratch_pt; |
283 | struct i915_page_directory *scratch_pd; | |
69ab76fd | 284 | struct i915_page_directory_pointer *scratch_pdp; /* GEN8+ & 48b PPGTT */ |
0260c420 BW |
285 | |
286 | /** | |
287 | * List of objects currently involved in rendering. | |
288 | * | |
289 | * Includes buffers having the contents of their GPU caches | |
97b2a6a1 | 290 | * flushed, not necessarily primitives. last_read_req |
0260c420 BW |
291 | * represents when the rendering involved will be completed. |
292 | * | |
293 | * A reference is held on the buffer while on this list. | |
294 | */ | |
295 | struct list_head active_list; | |
296 | ||
297 | /** | |
298 | * LRU list of objects which are not in the ringbuffer and | |
299 | * are ready to unbind, but are still in the GTT. | |
300 | * | |
97b2a6a1 | 301 | * last_read_req is NULL while an object is in this list. |
0260c420 BW |
302 | * |
303 | * A reference is not held on the buffer while on this list, | |
304 | * as merely being GTT-bound shouldn't prevent its being | |
305 | * freed, and we'll pull it off the list in the free path. | |
306 | */ | |
307 | struct list_head inactive_list; | |
308 | ||
309 | /* FIXME: Need a more generic return type */ | |
07749ef3 MT |
310 | gen6_pte_t (*pte_encode)(dma_addr_t addr, |
311 | enum i915_cache_level level, | |
312 | bool valid, u32 flags); /* Create a valid PTE */ | |
f329f5f6 DV |
313 | /* flags for pte_encode */ |
314 | #define PTE_READ_ONLY (1<<0) | |
678d96fb BW |
315 | int (*allocate_va_range)(struct i915_address_space *vm, |
316 | uint64_t start, | |
317 | uint64_t length); | |
0260c420 BW |
318 | void (*clear_range)(struct i915_address_space *vm, |
319 | uint64_t start, | |
320 | uint64_t length, | |
321 | bool use_scratch); | |
d6473f56 CW |
322 | void (*insert_page)(struct i915_address_space *vm, |
323 | dma_addr_t addr, | |
324 | uint64_t offset, | |
325 | enum i915_cache_level cache_level, | |
326 | u32 flags); | |
0260c420 BW |
327 | void (*insert_entries)(struct i915_address_space *vm, |
328 | struct sg_table *st, | |
329 | uint64_t start, | |
24f3a8cf | 330 | enum i915_cache_level cache_level, u32 flags); |
0260c420 | 331 | void (*cleanup)(struct i915_address_space *vm); |
777dc5bb DV |
332 | /** Unmap an object from an address space. This usually consists of |
333 | * setting the valid PTE entries to a reserved scratch page. */ | |
334 | void (*unbind_vma)(struct i915_vma *vma); | |
335 | /* Map an object into an address space with the given cache flags. */ | |
70b9f6f8 DV |
336 | int (*bind_vma)(struct i915_vma *vma, |
337 | enum i915_cache_level cache_level, | |
338 | u32 flags); | |
0260c420 BW |
339 | }; |
340 | ||
596c5923 CW |
341 | #define i915_is_ggtt(V) ((V)->is_ggtt) |
342 | ||
0260c420 BW |
343 | /* The Graphics Translation Table is the way in which GEN hardware translates a |
344 | * Graphics Virtual Address into a Physical Address. In addition to the normal | |
345 | * collateral associated with any va->pa translations GEN hardware also has a | |
346 | * portion of the GTT which can be mapped by the CPU and remain both coherent | |
347 | * and correct (in cases like swizzling). That region is referred to as GMADR in | |
348 | * the spec. | |
349 | */ | |
62106b4f | 350 | struct i915_ggtt { |
0260c420 | 351 | struct i915_address_space base; |
0260c420 | 352 | |
c44ef60e | 353 | size_t stolen_size; /* Total size of stolen memory */ |
a9da512b | 354 | size_t stolen_usable_size; /* Total size minus BIOS reserved */ |
274008e8 SAK |
355 | size_t stolen_reserved_base; |
356 | size_t stolen_reserved_size; | |
d507d735 | 357 | size_t size; /* Total size of Global GTT */ |
c44ef60e | 358 | u64 mappable_end; /* End offset that we can CPU map */ |
0260c420 BW |
359 | struct io_mapping *mappable; /* Mapping to our CPU mappable region */ |
360 | phys_addr_t mappable_base; /* PA of our GMADR */ | |
361 | ||
362 | /** "Graphics Stolen Memory" holds the global PTEs */ | |
363 | void __iomem *gsm; | |
364 | ||
365 | bool do_idle_maps; | |
366 | ||
367 | int mtrr; | |
368 | ||
d507d735 | 369 | int (*probe)(struct i915_ggtt *ggtt); |
0260c420 BW |
370 | }; |
371 | ||
372 | struct i915_hw_ppgtt { | |
373 | struct i915_address_space base; | |
374 | struct kref ref; | |
375 | struct drm_mm_node node; | |
563222a7 | 376 | unsigned long pd_dirty_rings; |
d7b3de91 | 377 | union { |
81ba8aef MT |
378 | struct i915_pml4 pml4; /* GEN8+ & 48b PPGTT */ |
379 | struct i915_page_directory_pointer pdp; /* GEN8+ */ | |
380 | struct i915_page_directory pd; /* GEN6-7 */ | |
d7b3de91 | 381 | }; |
0260c420 | 382 | |
4d884705 | 383 | struct drm_i915_file_private *file_priv; |
0260c420 | 384 | |
678d96fb BW |
385 | gen6_pte_t __iomem *pd_addr; |
386 | ||
0260c420 BW |
387 | int (*enable)(struct i915_hw_ppgtt *ppgtt); |
388 | int (*switch_mm)(struct i915_hw_ppgtt *ppgtt, | |
e85b26dc | 389 | struct drm_i915_gem_request *req); |
0260c420 BW |
390 | void (*debug_dump)(struct i915_hw_ppgtt *ppgtt, struct seq_file *m); |
391 | }; | |
392 | ||
731f74c5 DG |
393 | /* |
394 | * gen6_for_each_pde() iterates over every pde from start until start+length. | |
395 | * If start and start+length are not perfectly divisible, the macro will round | |
396 | * down and up as needed. Start=0 and length=2G effectively iterates over | |
397 | * every PDE in the system. The macro modifies ALL its parameters except 'pd', | |
398 | * so each of the other parameters should preferably be a simple variable, or | |
399 | * at most an lvalue with no side-effects! | |
678d96fb | 400 | */ |
731f74c5 DG |
401 | #define gen6_for_each_pde(pt, pd, start, length, iter) \ |
402 | for (iter = gen6_pde_index(start); \ | |
403 | length > 0 && iter < I915_PDES && \ | |
404 | (pt = (pd)->page_table[iter], true); \ | |
405 | ({ u32 temp = ALIGN(start+1, 1 << GEN6_PDE_SHIFT); \ | |
406 | temp = min(temp - start, length); \ | |
407 | start += temp, length -= temp; }), ++iter) | |
408 | ||
409 | #define gen6_for_all_pdes(pt, pd, iter) \ | |
410 | for (iter = 0; \ | |
411 | iter < I915_PDES && \ | |
412 | (pt = (pd)->page_table[iter], true); \ | |
413 | ++iter) | |
09942c65 | 414 | |
678d96fb BW |
415 | static inline uint32_t i915_pte_index(uint64_t address, uint32_t pde_shift) |
416 | { | |
417 | const uint32_t mask = NUM_PTE(pde_shift) - 1; | |
418 | ||
419 | return (address >> PAGE_SHIFT) & mask; | |
420 | } | |
421 | ||
422 | /* Helper to counts the number of PTEs within the given length. This count | |
423 | * does not cross a page table boundary, so the max value would be | |
424 | * GEN6_PTES for GEN6, and GEN8_PTES for GEN8. | |
425 | */ | |
426 | static inline uint32_t i915_pte_count(uint64_t addr, size_t length, | |
427 | uint32_t pde_shift) | |
428 | { | |
69603dbb | 429 | const uint64_t mask = ~((1ULL << pde_shift) - 1); |
678d96fb BW |
430 | uint64_t end; |
431 | ||
432 | WARN_ON(length == 0); | |
433 | WARN_ON(offset_in_page(addr|length)); | |
434 | ||
435 | end = addr + length; | |
436 | ||
437 | if ((addr & mask) != (end & mask)) | |
438 | return NUM_PTE(pde_shift) - i915_pte_index(addr, pde_shift); | |
439 | ||
440 | return i915_pte_index(end, pde_shift) - i915_pte_index(addr, pde_shift); | |
441 | } | |
442 | ||
443 | static inline uint32_t i915_pde_index(uint64_t addr, uint32_t shift) | |
444 | { | |
445 | return (addr >> shift) & I915_PDE_MASK; | |
446 | } | |
447 | ||
448 | static inline uint32_t gen6_pte_index(uint32_t addr) | |
449 | { | |
450 | return i915_pte_index(addr, GEN6_PDE_SHIFT); | |
451 | } | |
452 | ||
453 | static inline size_t gen6_pte_count(uint32_t addr, uint32_t length) | |
454 | { | |
455 | return i915_pte_count(addr, length, GEN6_PDE_SHIFT); | |
456 | } | |
457 | ||
458 | static inline uint32_t gen6_pde_index(uint32_t addr) | |
459 | { | |
460 | return i915_pde_index(addr, GEN6_PDE_SHIFT); | |
461 | } | |
462 | ||
9271d959 MT |
463 | /* Equivalent to the gen6 version, For each pde iterates over every pde |
464 | * between from start until start + length. On gen8+ it simply iterates | |
465 | * over every page directory entry in a page directory. | |
466 | */ | |
e8ebd8e2 DG |
467 | #define gen8_for_each_pde(pt, pd, start, length, iter) \ |
468 | for (iter = gen8_pde_index(start); \ | |
469 | length > 0 && iter < I915_PDES && \ | |
470 | (pt = (pd)->page_table[iter], true); \ | |
471 | ({ u64 temp = ALIGN(start+1, 1 << GEN8_PDE_SHIFT); \ | |
472 | temp = min(temp - start, length); \ | |
473 | start += temp, length -= temp; }), ++iter) | |
474 | ||
475 | #define gen8_for_each_pdpe(pd, pdp, start, length, iter) \ | |
476 | for (iter = gen8_pdpe_index(start); \ | |
477 | length > 0 && iter < I915_PDPES_PER_PDP(dev) && \ | |
478 | (pd = (pdp)->page_directory[iter], true); \ | |
479 | ({ u64 temp = ALIGN(start+1, 1 << GEN8_PDPE_SHIFT); \ | |
480 | temp = min(temp - start, length); \ | |
481 | start += temp, length -= temp; }), ++iter) | |
482 | ||
483 | #define gen8_for_each_pml4e(pdp, pml4, start, length, iter) \ | |
484 | for (iter = gen8_pml4e_index(start); \ | |
485 | length > 0 && iter < GEN8_PML4ES_PER_PML4 && \ | |
486 | (pdp = (pml4)->pdps[iter], true); \ | |
487 | ({ u64 temp = ALIGN(start+1, 1ULL << GEN8_PML4E_SHIFT); \ | |
488 | temp = min(temp - start, length); \ | |
489 | start += temp, length -= temp; }), ++iter) | |
762d9936 | 490 | |
9271d959 MT |
491 | static inline uint32_t gen8_pte_index(uint64_t address) |
492 | { | |
493 | return i915_pte_index(address, GEN8_PDE_SHIFT); | |
494 | } | |
495 | ||
496 | static inline uint32_t gen8_pde_index(uint64_t address) | |
497 | { | |
498 | return i915_pde_index(address, GEN8_PDE_SHIFT); | |
499 | } | |
500 | ||
501 | static inline uint32_t gen8_pdpe_index(uint64_t address) | |
502 | { | |
503 | return (address >> GEN8_PDPE_SHIFT) & GEN8_PDPE_MASK; | |
504 | } | |
505 | ||
506 | static inline uint32_t gen8_pml4e_index(uint64_t address) | |
507 | { | |
762d9936 | 508 | return (address >> GEN8_PML4E_SHIFT) & GEN8_PML4E_MASK; |
9271d959 MT |
509 | } |
510 | ||
33c8819f MT |
511 | static inline size_t gen8_pte_count(uint64_t address, uint64_t length) |
512 | { | |
513 | return i915_pte_count(address, length, GEN8_PDE_SHIFT); | |
514 | } | |
515 | ||
d852c7bf MK |
516 | static inline dma_addr_t |
517 | i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n) | |
518 | { | |
519 | return test_bit(n, ppgtt->pdp.used_pdpes) ? | |
567047be | 520 | px_dma(ppgtt->pdp.page_directory[n]) : |
79ab9370 | 521 | px_dma(ppgtt->base.scratch_pd); |
d852c7bf MK |
522 | } |
523 | ||
0088e522 | 524 | int i915_ggtt_probe_hw(struct drm_device *dev); |
d85489d3 | 525 | int i915_ggtt_init_hw(struct drm_device *dev); |
ac840ae5 | 526 | int i915_ggtt_enable_hw(struct drm_device *dev); |
d85489d3 JL |
527 | void i915_gem_init_ggtt(struct drm_device *dev); |
528 | void i915_ggtt_cleanup_hw(struct drm_device *dev); | |
ee960be7 | 529 | |
82460d97 | 530 | int i915_ppgtt_init_hw(struct drm_device *dev); |
ee960be7 | 531 | void i915_ppgtt_release(struct kref *kref); |
4d884705 DV |
532 | struct i915_hw_ppgtt *i915_ppgtt_create(struct drm_device *dev, |
533 | struct drm_i915_file_private *fpriv); | |
ee960be7 DV |
534 | static inline void i915_ppgtt_get(struct i915_hw_ppgtt *ppgtt) |
535 | { | |
536 | if (ppgtt) | |
537 | kref_get(&ppgtt->ref); | |
538 | } | |
539 | static inline void i915_ppgtt_put(struct i915_hw_ppgtt *ppgtt) | |
540 | { | |
541 | if (ppgtt) | |
542 | kref_put(&ppgtt->ref, i915_ppgtt_release); | |
543 | } | |
0260c420 | 544 | |
dc97997a | 545 | void i915_check_and_clear_faults(struct drm_i915_private *dev_priv); |
0260c420 BW |
546 | void i915_gem_suspend_gtt_mappings(struct drm_device *dev); |
547 | void i915_gem_restore_gtt_mappings(struct drm_device *dev); | |
548 | ||
549 | int __must_check i915_gem_gtt_prepare_object(struct drm_i915_gem_object *obj); | |
550 | void i915_gem_gtt_finish_object(struct drm_i915_gem_object *obj); | |
551 | ||
9abc4648 JL |
552 | static inline bool |
553 | i915_ggtt_view_equal(const struct i915_ggtt_view *a, | |
554 | const struct i915_ggtt_view *b) | |
555 | { | |
556 | if (WARN_ON(!a || !b)) | |
557 | return false; | |
558 | ||
8bd7ef16 JL |
559 | if (a->type != b->type) |
560 | return false; | |
ce7f1728 | 561 | if (a->type != I915_GGTT_VIEW_NORMAL) |
8bd7ef16 JL |
562 | return !memcmp(&a->params, &b->params, sizeof(a->params)); |
563 | return true; | |
9abc4648 JL |
564 | } |
565 | ||
91e6711e JL |
566 | size_t |
567 | i915_ggtt_view_size(struct drm_i915_gem_object *obj, | |
568 | const struct i915_ggtt_view *view); | |
569 | ||
8ef8561f CW |
570 | /** |
571 | * i915_vma_pin_iomap - calls ioremap_wc to map the GGTT VMA via the aperture | |
572 | * @vma: VMA to iomap | |
573 | * | |
574 | * The passed in VMA has to be pinned in the global GTT mappable region. | |
575 | * An extra pinning of the VMA is acquired for the return iomapping, | |
576 | * the caller must call i915_vma_unpin_iomap to relinquish the pinning | |
577 | * after the iomapping is no longer required. | |
578 | * | |
579 | * Callers must hold the struct_mutex. | |
580 | * | |
581 | * Returns a valid iomapped pointer or ERR_PTR. | |
582 | */ | |
583 | void __iomem *i915_vma_pin_iomap(struct i915_vma *vma); | |
406ea8d2 | 584 | #define IO_ERR_PTR(x) ((void __iomem *)ERR_PTR(x)) |
8ef8561f CW |
585 | |
586 | /** | |
587 | * i915_vma_unpin_iomap - unpins the mapping returned from i915_vma_iomap | |
588 | * @vma: VMA to unpin | |
589 | * | |
590 | * Unpins the previously iomapped VMA from i915_vma_pin_iomap(). | |
591 | * | |
592 | * Callers must hold the struct_mutex. This function is only valid to be | |
593 | * called on a VMA previously iomapped by the caller with i915_vma_pin_iomap(). | |
594 | */ | |
595 | static inline void i915_vma_unpin_iomap(struct i915_vma *vma) | |
596 | { | |
597 | lockdep_assert_held(&vma->vm->dev->struct_mutex); | |
598 | GEM_BUG_ON(vma->pin_count == 0); | |
599 | GEM_BUG_ON(vma->iomap == NULL); | |
600 | vma->pin_count--; | |
601 | } | |
602 | ||
0260c420 | 603 | #endif |