xref: /linux/drivers/gpu/drm/i915/i915_gem_gtt.h (revision bd628c1bed7902ec1f24ba0fe70758949146abbe)
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 
37 #include <linux/io-mapping.h>
38 #include <linux/mm.h>
39 #include <linux/pagevec.h>
40 
41 #include "i915_request.h"
42 #include "i915_selftest.h"
43 #include "i915_timeline.h"
44 
45 #define I915_GTT_PAGE_SIZE_4K	BIT_ULL(12)
46 #define I915_GTT_PAGE_SIZE_64K	BIT_ULL(16)
47 #define I915_GTT_PAGE_SIZE_2M	BIT_ULL(21)
48 
49 #define I915_GTT_PAGE_SIZE I915_GTT_PAGE_SIZE_4K
50 #define I915_GTT_MAX_PAGE_SIZE I915_GTT_PAGE_SIZE_2M
51 
52 #define I915_GTT_PAGE_MASK -I915_GTT_PAGE_SIZE
53 
54 #define I915_GTT_MIN_ALIGNMENT I915_GTT_PAGE_SIZE
55 
56 #define I915_FENCE_REG_NONE -1
57 #define I915_MAX_NUM_FENCES 32
58 /* 32 fences + sign bit for FENCE_REG_NONE */
59 #define I915_MAX_NUM_FENCE_BITS 6
60 
61 struct drm_i915_file_private;
62 struct drm_i915_fence_reg;
63 struct i915_vma;
64 
65 typedef u32 gen6_pte_t;
66 typedef u64 gen8_pte_t;
67 typedef u64 gen8_pde_t;
68 typedef u64 gen8_ppgtt_pdpe_t;
69 typedef u64 gen8_ppgtt_pml4e_t;
70 
71 #define ggtt_total_entries(ggtt) ((ggtt)->vm.total >> PAGE_SHIFT)
72 
73 /* gen6-hsw has bit 11-4 for physical addr bit 39-32 */
74 #define GEN6_GTT_ADDR_ENCODE(addr)	((addr) | (((addr) >> 28) & 0xff0))
75 #define GEN6_PTE_ADDR_ENCODE(addr)	GEN6_GTT_ADDR_ENCODE(addr)
76 #define GEN6_PDE_ADDR_ENCODE(addr)	GEN6_GTT_ADDR_ENCODE(addr)
77 #define GEN6_PTE_CACHE_LLC		(2 << 1)
78 #define GEN6_PTE_UNCACHED		(1 << 1)
79 #define GEN6_PTE_VALID			(1 << 0)
80 
81 #define I915_PTES(pte_len)		((unsigned int)(PAGE_SIZE / (pte_len)))
82 #define I915_PTE_MASK(pte_len)		(I915_PTES(pte_len) - 1)
83 #define I915_PDES			512
84 #define I915_PDE_MASK			(I915_PDES - 1)
85 #define NUM_PTE(pde_shift)     (1 << (pde_shift - PAGE_SHIFT))
86 
87 #define GEN6_PTES			I915_PTES(sizeof(gen6_pte_t))
88 #define GEN6_PD_SIZE		        (I915_PDES * PAGE_SIZE)
89 #define GEN6_PD_ALIGN			(PAGE_SIZE * 16)
90 #define GEN6_PDE_SHIFT			22
91 #define GEN6_PDE_VALID			(1 << 0)
92 
93 #define GEN7_PTE_CACHE_L3_LLC		(3 << 1)
94 
95 #define BYT_PTE_SNOOPED_BY_CPU_CACHES	(1 << 2)
96 #define BYT_PTE_WRITEABLE		(1 << 1)
97 
98 /* Cacheability Control is a 4-bit value. The low three bits are stored in bits
99  * 3:1 of the PTE, while the fourth bit is stored in bit 11 of the PTE.
100  */
101 #define HSW_CACHEABILITY_CONTROL(bits)	((((bits) & 0x7) << 1) | \
102 					 (((bits) & 0x8) << (11 - 3)))
103 #define HSW_WB_LLC_AGE3			HSW_CACHEABILITY_CONTROL(0x2)
104 #define HSW_WB_LLC_AGE0			HSW_CACHEABILITY_CONTROL(0x3)
105 #define HSW_WB_ELLC_LLC_AGE3		HSW_CACHEABILITY_CONTROL(0x8)
106 #define HSW_WB_ELLC_LLC_AGE0		HSW_CACHEABILITY_CONTROL(0xb)
107 #define HSW_WT_ELLC_LLC_AGE3		HSW_CACHEABILITY_CONTROL(0x7)
108 #define HSW_WT_ELLC_LLC_AGE0		HSW_CACHEABILITY_CONTROL(0x6)
109 #define HSW_PTE_UNCACHED		(0)
110 #define HSW_GTT_ADDR_ENCODE(addr)	((addr) | (((addr) >> 28) & 0x7f0))
111 #define HSW_PTE_ADDR_ENCODE(addr)	HSW_GTT_ADDR_ENCODE(addr)
112 
113 /* GEN8 32b style address is defined as a 3 level page table:
114  * 31:30 | 29:21 | 20:12 |  11:0
115  * PDPE  |  PDE  |  PTE  | offset
116  * The difference as compared to normal x86 3 level page table is the PDPEs are
117  * programmed via register.
118  */
119 #define GEN8_3LVL_PDPES			4
120 #define GEN8_PDE_SHIFT			21
121 #define GEN8_PDE_MASK			0x1ff
122 #define GEN8_PTE_SHIFT			12
123 #define GEN8_PTE_MASK			0x1ff
124 #define GEN8_PTES			I915_PTES(sizeof(gen8_pte_t))
125 
126 /* GEN8 48b style address is defined as a 4 level page table:
127  * 47:39 | 38:30 | 29:21 | 20:12 |  11:0
128  * PML4E | PDPE  |  PDE  |  PTE  | offset
129  */
130 #define GEN8_PML4ES_PER_PML4		512
131 #define GEN8_PML4E_SHIFT		39
132 #define GEN8_PML4E_MASK			(GEN8_PML4ES_PER_PML4 - 1)
133 #define GEN8_PDPE_SHIFT			30
134 /* NB: GEN8_PDPE_MASK is untrue for 32b platforms, but it has no impact on 32b page
135  * tables */
136 #define GEN8_PDPE_MASK			0x1ff
137 
138 #define PPAT_UNCACHED			(_PAGE_PWT | _PAGE_PCD)
139 #define PPAT_CACHED_PDE			0 /* WB LLC */
140 #define PPAT_CACHED			_PAGE_PAT /* WB LLCeLLC */
141 #define PPAT_DISPLAY_ELLC		_PAGE_PCD /* WT eLLC */
142 
143 #define CHV_PPAT_SNOOP			(1<<6)
144 #define GEN8_PPAT_AGE(x)		((x)<<4)
145 #define GEN8_PPAT_LLCeLLC		(3<<2)
146 #define GEN8_PPAT_LLCELLC		(2<<2)
147 #define GEN8_PPAT_LLC			(1<<2)
148 #define GEN8_PPAT_WB			(3<<0)
149 #define GEN8_PPAT_WT			(2<<0)
150 #define GEN8_PPAT_WC			(1<<0)
151 #define GEN8_PPAT_UC			(0<<0)
152 #define GEN8_PPAT_ELLC_OVERRIDE		(0<<2)
153 #define GEN8_PPAT(i, x)			((u64)(x) << ((i) * 8))
154 
155 #define GEN8_PPAT_GET_CA(x) ((x) & 3)
156 #define GEN8_PPAT_GET_TC(x) ((x) & (3 << 2))
157 #define GEN8_PPAT_GET_AGE(x) ((x) & (3 << 4))
158 #define CHV_PPAT_GET_SNOOP(x) ((x) & (1 << 6))
159 
160 #define GEN8_PDE_IPS_64K BIT(11)
161 #define GEN8_PDE_PS_2M   BIT(7)
162 
163 struct sg_table;
164 
165 struct intel_rotation_info {
166 	struct intel_rotation_plane_info {
167 		/* tiles */
168 		unsigned int width, height, stride, offset;
169 	} plane[2];
170 } __packed;
171 
172 struct intel_partial_info {
173 	u64 offset;
174 	unsigned int size;
175 } __packed;
176 
177 enum i915_ggtt_view_type {
178 	I915_GGTT_VIEW_NORMAL = 0,
179 	I915_GGTT_VIEW_ROTATED = sizeof(struct intel_rotation_info),
180 	I915_GGTT_VIEW_PARTIAL = sizeof(struct intel_partial_info),
181 };
182 
183 static inline void assert_i915_gem_gtt_types(void)
184 {
185 	BUILD_BUG_ON(sizeof(struct intel_rotation_info) != 8*sizeof(unsigned int));
186 	BUILD_BUG_ON(sizeof(struct intel_partial_info) != sizeof(u64) + sizeof(unsigned int));
187 
188 	/* As we encode the size of each branch inside the union into its type,
189 	 * we have to be careful that each branch has a unique size.
190 	 */
191 	switch ((enum i915_ggtt_view_type)0) {
192 	case I915_GGTT_VIEW_NORMAL:
193 	case I915_GGTT_VIEW_PARTIAL:
194 	case I915_GGTT_VIEW_ROTATED:
195 		/* gcc complains if these are identical cases */
196 		break;
197 	}
198 }
199 
200 struct i915_ggtt_view {
201 	enum i915_ggtt_view_type type;
202 	union {
203 		/* Members need to contain no holes/padding */
204 		struct intel_partial_info partial;
205 		struct intel_rotation_info rotated;
206 	};
207 };
208 
209 enum i915_cache_level;
210 
211 struct i915_vma;
212 
213 struct i915_page_dma {
214 	struct page *page;
215 	int order;
216 	union {
217 		dma_addr_t daddr;
218 
219 		/* For gen6/gen7 only. This is the offset in the GGTT
220 		 * where the page directory entries for PPGTT begin
221 		 */
222 		u32 ggtt_offset;
223 	};
224 };
225 
226 #define px_base(px) (&(px)->base)
227 #define px_dma(px) (px_base(px)->daddr)
228 
229 struct i915_page_table {
230 	struct i915_page_dma base;
231 	unsigned int used_ptes;
232 };
233 
234 struct i915_page_directory {
235 	struct i915_page_dma base;
236 
237 	struct i915_page_table *page_table[I915_PDES]; /* PDEs */
238 	unsigned int used_pdes;
239 };
240 
241 struct i915_page_directory_pointer {
242 	struct i915_page_dma base;
243 	struct i915_page_directory **page_directory;
244 	unsigned int used_pdpes;
245 };
246 
247 struct i915_pml4 {
248 	struct i915_page_dma base;
249 	struct i915_page_directory_pointer *pdps[GEN8_PML4ES_PER_PML4];
250 };
251 
252 struct i915_vma_ops {
253 	/* Map an object into an address space with the given cache flags. */
254 	int (*bind_vma)(struct i915_vma *vma,
255 			enum i915_cache_level cache_level,
256 			u32 flags);
257 	/*
258 	 * Unmap an object from an address space. This usually consists of
259 	 * setting the valid PTE entries to a reserved scratch page.
260 	 */
261 	void (*unbind_vma)(struct i915_vma *vma);
262 
263 	int (*set_pages)(struct i915_vma *vma);
264 	void (*clear_pages)(struct i915_vma *vma);
265 };
266 
267 struct pagestash {
268 	spinlock_t lock;
269 	struct pagevec pvec;
270 };
271 
272 struct i915_address_space {
273 	struct drm_mm mm;
274 	struct drm_i915_private *i915;
275 	struct device *dma;
276 	/* Every address space belongs to a struct file - except for the global
277 	 * GTT that is owned by the driver (and so @file is set to NULL). In
278 	 * principle, no information should leak from one context to another
279 	 * (or between files/processes etc) unless explicitly shared by the
280 	 * owner. Tracking the owner is important in order to free up per-file
281 	 * objects along with the file, to aide resource tracking, and to
282 	 * assign blame.
283 	 */
284 	struct drm_i915_file_private *file;
285 	u64 total;		/* size addr space maps (ex. 2GB for ggtt) */
286 	u64 reserved;		/* size addr space reserved */
287 
288 	bool closed;
289 
290 	struct mutex mutex; /* protects vma and our lists */
291 
292 	u64 scratch_pte;
293 	struct i915_page_dma scratch_page;
294 	struct i915_page_table *scratch_pt;
295 	struct i915_page_directory *scratch_pd;
296 	struct i915_page_directory_pointer *scratch_pdp; /* GEN8+ & 48b PPGTT */
297 
298 	/**
299 	 * List of objects currently involved in rendering.
300 	 *
301 	 * Includes buffers having the contents of their GPU caches
302 	 * flushed, not necessarily primitives. last_read_req
303 	 * represents when the rendering involved will be completed.
304 	 *
305 	 * A reference is held on the buffer while on this list.
306 	 */
307 	struct list_head active_list;
308 
309 	/**
310 	 * LRU list of objects which are not in the ringbuffer and
311 	 * are ready to unbind, but are still in the GTT.
312 	 *
313 	 * last_read_req is NULL while an object is in this list.
314 	 *
315 	 * A reference is not held on the buffer while on this list,
316 	 * as merely being GTT-bound shouldn't prevent its being
317 	 * freed, and we'll pull it off the list in the free path.
318 	 */
319 	struct list_head inactive_list;
320 
321 	/**
322 	 * List of vma that have been unbound.
323 	 *
324 	 * A reference is not held on the buffer while on this list.
325 	 */
326 	struct list_head unbound_list;
327 
328 	struct pagestash free_pages;
329 
330 	/* Global GTT */
331 	bool is_ggtt:1;
332 
333 	/* Some systems require uncached updates of the page directories */
334 	bool pt_kmap_wc:1;
335 
336 	/* Some systems support read-only mappings for GGTT and/or PPGTT */
337 	bool has_read_only:1;
338 
339 	u64 (*pte_encode)(dma_addr_t addr,
340 			  enum i915_cache_level level,
341 			  u32 flags); /* Create a valid PTE */
342 #define PTE_READ_ONLY	(1<<0)
343 
344 	int (*allocate_va_range)(struct i915_address_space *vm,
345 				 u64 start, u64 length);
346 	void (*clear_range)(struct i915_address_space *vm,
347 			    u64 start, u64 length);
348 	void (*insert_page)(struct i915_address_space *vm,
349 			    dma_addr_t addr,
350 			    u64 offset,
351 			    enum i915_cache_level cache_level,
352 			    u32 flags);
353 	void (*insert_entries)(struct i915_address_space *vm,
354 			       struct i915_vma *vma,
355 			       enum i915_cache_level cache_level,
356 			       u32 flags);
357 	void (*cleanup)(struct i915_address_space *vm);
358 
359 	struct i915_vma_ops vma_ops;
360 
361 	I915_SELFTEST_DECLARE(struct fault_attr fault_attr);
362 	I915_SELFTEST_DECLARE(bool scrub_64K);
363 };
364 
365 #define i915_is_ggtt(vm) ((vm)->is_ggtt)
366 
367 static inline bool
368 i915_vm_is_48bit(const struct i915_address_space *vm)
369 {
370 	return (vm->total - 1) >> 32;
371 }
372 
373 static inline bool
374 i915_vm_has_scratch_64K(struct i915_address_space *vm)
375 {
376 	return vm->scratch_page.order == get_order(I915_GTT_PAGE_SIZE_64K);
377 }
378 
379 /* The Graphics Translation Table is the way in which GEN hardware translates a
380  * Graphics Virtual Address into a Physical Address. In addition to the normal
381  * collateral associated with any va->pa translations GEN hardware also has a
382  * portion of the GTT which can be mapped by the CPU and remain both coherent
383  * and correct (in cases like swizzling). That region is referred to as GMADR in
384  * the spec.
385  */
386 struct i915_ggtt {
387 	struct i915_address_space vm;
388 
389 	struct io_mapping iomap;	/* Mapping to our CPU mappable region */
390 	struct resource gmadr;          /* GMADR resource */
391 	resource_size_t mappable_end;	/* End offset that we can CPU map */
392 
393 	/** "Graphics Stolen Memory" holds the global PTEs */
394 	void __iomem *gsm;
395 	void (*invalidate)(struct drm_i915_private *dev_priv);
396 
397 	bool do_idle_maps;
398 
399 	int mtrr;
400 
401 	u32 pin_bias;
402 
403 	struct drm_mm_node error_capture;
404 };
405 
406 struct i915_hw_ppgtt {
407 	struct i915_address_space vm;
408 	struct kref ref;
409 
410 	unsigned long pd_dirty_rings;
411 	union {
412 		struct i915_pml4 pml4;		/* GEN8+ & 48b PPGTT */
413 		struct i915_page_directory_pointer pdp;	/* GEN8+ */
414 		struct i915_page_directory pd;		/* GEN6-7 */
415 	};
416 
417 	void (*debug_dump)(struct i915_hw_ppgtt *ppgtt, struct seq_file *m);
418 };
419 
420 struct gen6_hw_ppgtt {
421 	struct i915_hw_ppgtt base;
422 
423 	struct i915_vma *vma;
424 	gen6_pte_t __iomem *pd_addr;
425 
426 	unsigned int pin_count;
427 	bool scan_for_unused_pt;
428 };
429 
430 #define __to_gen6_ppgtt(base) container_of(base, struct gen6_hw_ppgtt, base)
431 
432 static inline struct gen6_hw_ppgtt *to_gen6_ppgtt(struct i915_hw_ppgtt *base)
433 {
434 	BUILD_BUG_ON(offsetof(struct gen6_hw_ppgtt, base));
435 	return __to_gen6_ppgtt(base);
436 }
437 
438 /*
439  * gen6_for_each_pde() iterates over every pde from start until start+length.
440  * If start and start+length are not perfectly divisible, the macro will round
441  * down and up as needed. Start=0 and length=2G effectively iterates over
442  * every PDE in the system. The macro modifies ALL its parameters except 'pd',
443  * so each of the other parameters should preferably be a simple variable, or
444  * at most an lvalue with no side-effects!
445  */
446 #define gen6_for_each_pde(pt, pd, start, length, iter)			\
447 	for (iter = gen6_pde_index(start);				\
448 	     length > 0 && iter < I915_PDES &&				\
449 		(pt = (pd)->page_table[iter], true);			\
450 	     ({ u32 temp = ALIGN(start+1, 1 << GEN6_PDE_SHIFT);		\
451 		    temp = min(temp - start, length);			\
452 		    start += temp, length -= temp; }), ++iter)
453 
454 #define gen6_for_all_pdes(pt, pd, iter)					\
455 	for (iter = 0;							\
456 	     iter < I915_PDES &&					\
457 		(pt = (pd)->page_table[iter], true);			\
458 	     ++iter)
459 
460 static inline u32 i915_pte_index(u64 address, unsigned int pde_shift)
461 {
462 	const u32 mask = NUM_PTE(pde_shift) - 1;
463 
464 	return (address >> PAGE_SHIFT) & mask;
465 }
466 
467 /* Helper to counts the number of PTEs within the given length. This count
468  * does not cross a page table boundary, so the max value would be
469  * GEN6_PTES for GEN6, and GEN8_PTES for GEN8.
470 */
471 static inline u32 i915_pte_count(u64 addr, u64 length, unsigned int pde_shift)
472 {
473 	const u64 mask = ~((1ULL << pde_shift) - 1);
474 	u64 end;
475 
476 	GEM_BUG_ON(length == 0);
477 	GEM_BUG_ON(offset_in_page(addr | length));
478 
479 	end = addr + length;
480 
481 	if ((addr & mask) != (end & mask))
482 		return NUM_PTE(pde_shift) - i915_pte_index(addr, pde_shift);
483 
484 	return i915_pte_index(end, pde_shift) - i915_pte_index(addr, pde_shift);
485 }
486 
487 static inline u32 i915_pde_index(u64 addr, u32 shift)
488 {
489 	return (addr >> shift) & I915_PDE_MASK;
490 }
491 
492 static inline u32 gen6_pte_index(u32 addr)
493 {
494 	return i915_pte_index(addr, GEN6_PDE_SHIFT);
495 }
496 
497 static inline u32 gen6_pte_count(u32 addr, u32 length)
498 {
499 	return i915_pte_count(addr, length, GEN6_PDE_SHIFT);
500 }
501 
502 static inline u32 gen6_pde_index(u32 addr)
503 {
504 	return i915_pde_index(addr, GEN6_PDE_SHIFT);
505 }
506 
507 static inline unsigned int
508 i915_pdpes_per_pdp(const struct i915_address_space *vm)
509 {
510 	if (i915_vm_is_48bit(vm))
511 		return GEN8_PML4ES_PER_PML4;
512 
513 	return GEN8_3LVL_PDPES;
514 }
515 
516 /* Equivalent to the gen6 version, For each pde iterates over every pde
517  * between from start until start + length. On gen8+ it simply iterates
518  * over every page directory entry in a page directory.
519  */
520 #define gen8_for_each_pde(pt, pd, start, length, iter)			\
521 	for (iter = gen8_pde_index(start);				\
522 	     length > 0 && iter < I915_PDES &&				\
523 		(pt = (pd)->page_table[iter], true);			\
524 	     ({ u64 temp = ALIGN(start+1, 1 << GEN8_PDE_SHIFT);		\
525 		    temp = min(temp - start, length);			\
526 		    start += temp, length -= temp; }), ++iter)
527 
528 #define gen8_for_each_pdpe(pd, pdp, start, length, iter)		\
529 	for (iter = gen8_pdpe_index(start);				\
530 	     length > 0 && iter < i915_pdpes_per_pdp(vm) &&		\
531 		(pd = (pdp)->page_directory[iter], true);		\
532 	     ({ u64 temp = ALIGN(start+1, 1 << GEN8_PDPE_SHIFT);	\
533 		    temp = min(temp - start, length);			\
534 		    start += temp, length -= temp; }), ++iter)
535 
536 #define gen8_for_each_pml4e(pdp, pml4, start, length, iter)		\
537 	for (iter = gen8_pml4e_index(start);				\
538 	     length > 0 && iter < GEN8_PML4ES_PER_PML4 &&		\
539 		(pdp = (pml4)->pdps[iter], true);			\
540 	     ({ u64 temp = ALIGN(start+1, 1ULL << GEN8_PML4E_SHIFT);	\
541 		    temp = min(temp - start, length);			\
542 		    start += temp, length -= temp; }), ++iter)
543 
544 static inline u32 gen8_pte_index(u64 address)
545 {
546 	return i915_pte_index(address, GEN8_PDE_SHIFT);
547 }
548 
549 static inline u32 gen8_pde_index(u64 address)
550 {
551 	return i915_pde_index(address, GEN8_PDE_SHIFT);
552 }
553 
554 static inline u32 gen8_pdpe_index(u64 address)
555 {
556 	return (address >> GEN8_PDPE_SHIFT) & GEN8_PDPE_MASK;
557 }
558 
559 static inline u32 gen8_pml4e_index(u64 address)
560 {
561 	return (address >> GEN8_PML4E_SHIFT) & GEN8_PML4E_MASK;
562 }
563 
564 static inline u64 gen8_pte_count(u64 address, u64 length)
565 {
566 	return i915_pte_count(address, length, GEN8_PDE_SHIFT);
567 }
568 
569 static inline dma_addr_t
570 i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n)
571 {
572 	return px_dma(ppgtt->pdp.page_directory[n]);
573 }
574 
575 static inline struct i915_ggtt *
576 i915_vm_to_ggtt(struct i915_address_space *vm)
577 {
578 	GEM_BUG_ON(!i915_is_ggtt(vm));
579 	return container_of(vm, struct i915_ggtt, vm);
580 }
581 
582 #define INTEL_MAX_PPAT_ENTRIES 8
583 #define INTEL_PPAT_PERFECT_MATCH (~0U)
584 
585 struct intel_ppat;
586 
587 struct intel_ppat_entry {
588 	struct intel_ppat *ppat;
589 	struct kref ref;
590 	u8 value;
591 };
592 
593 struct intel_ppat {
594 	struct intel_ppat_entry entries[INTEL_MAX_PPAT_ENTRIES];
595 	DECLARE_BITMAP(used, INTEL_MAX_PPAT_ENTRIES);
596 	DECLARE_BITMAP(dirty, INTEL_MAX_PPAT_ENTRIES);
597 	unsigned int max_entries;
598 	u8 clear_value;
599 	/*
600 	 * Return a score to show how two PPAT values match,
601 	 * a INTEL_PPAT_PERFECT_MATCH indicates a perfect match
602 	 */
603 	unsigned int (*match)(u8 src, u8 dst);
604 	void (*update_hw)(struct drm_i915_private *i915);
605 
606 	struct drm_i915_private *i915;
607 };
608 
609 const struct intel_ppat_entry *
610 intel_ppat_get(struct drm_i915_private *i915, u8 value);
611 void intel_ppat_put(const struct intel_ppat_entry *entry);
612 
613 int i915_gem_init_aliasing_ppgtt(struct drm_i915_private *i915);
614 void i915_gem_fini_aliasing_ppgtt(struct drm_i915_private *i915);
615 
616 int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv);
617 int i915_ggtt_init_hw(struct drm_i915_private *dev_priv);
618 int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv);
619 void i915_ggtt_enable_guc(struct drm_i915_private *i915);
620 void i915_ggtt_disable_guc(struct drm_i915_private *i915);
621 int i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
622 void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv);
623 
624 int i915_ppgtt_init_hw(struct drm_i915_private *dev_priv);
625 void i915_ppgtt_release(struct kref *kref);
626 struct i915_hw_ppgtt *i915_ppgtt_create(struct drm_i915_private *dev_priv,
627 					struct drm_i915_file_private *fpriv);
628 void i915_ppgtt_close(struct i915_address_space *vm);
629 static inline void i915_ppgtt_get(struct i915_hw_ppgtt *ppgtt)
630 {
631 	if (ppgtt)
632 		kref_get(&ppgtt->ref);
633 }
634 static inline void i915_ppgtt_put(struct i915_hw_ppgtt *ppgtt)
635 {
636 	if (ppgtt)
637 		kref_put(&ppgtt->ref, i915_ppgtt_release);
638 }
639 
640 int gen6_ppgtt_pin(struct i915_hw_ppgtt *base);
641 void gen6_ppgtt_unpin(struct i915_hw_ppgtt *base);
642 
643 void i915_check_and_clear_faults(struct drm_i915_private *dev_priv);
644 void i915_gem_suspend_gtt_mappings(struct drm_i915_private *dev_priv);
645 void i915_gem_restore_gtt_mappings(struct drm_i915_private *dev_priv);
646 
647 int __must_check i915_gem_gtt_prepare_pages(struct drm_i915_gem_object *obj,
648 					    struct sg_table *pages);
649 void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj,
650 			       struct sg_table *pages);
651 
652 int i915_gem_gtt_reserve(struct i915_address_space *vm,
653 			 struct drm_mm_node *node,
654 			 u64 size, u64 offset, unsigned long color,
655 			 unsigned int flags);
656 
657 int i915_gem_gtt_insert(struct i915_address_space *vm,
658 			struct drm_mm_node *node,
659 			u64 size, u64 alignment, unsigned long color,
660 			u64 start, u64 end, unsigned int flags);
661 
662 /* Flags used by pin/bind&friends. */
663 #define PIN_NONBLOCK		BIT_ULL(0)
664 #define PIN_MAPPABLE		BIT_ULL(1)
665 #define PIN_ZONE_4G		BIT_ULL(2)
666 #define PIN_NONFAULT		BIT_ULL(3)
667 #define PIN_NOEVICT		BIT_ULL(4)
668 
669 #define PIN_MBZ			BIT_ULL(5) /* I915_VMA_PIN_OVERFLOW */
670 #define PIN_GLOBAL		BIT_ULL(6) /* I915_VMA_GLOBAL_BIND */
671 #define PIN_USER		BIT_ULL(7) /* I915_VMA_LOCAL_BIND */
672 #define PIN_UPDATE		BIT_ULL(8)
673 
674 #define PIN_HIGH		BIT_ULL(9)
675 #define PIN_OFFSET_BIAS		BIT_ULL(10)
676 #define PIN_OFFSET_FIXED	BIT_ULL(11)
677 #define PIN_OFFSET_MASK		(-I915_GTT_PAGE_SIZE)
678 
679 #endif
680