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 "gt/intel_reset.h"
42 #include "i915_request.h"
43 #include "i915_selftest.h"
44 #include "i915_timeline.h"
45 
46 #define I915_GTT_PAGE_SIZE_4K	BIT_ULL(12)
47 #define I915_GTT_PAGE_SIZE_64K	BIT_ULL(16)
48 #define I915_GTT_PAGE_SIZE_2M	BIT_ULL(21)
49 
50 #define I915_GTT_PAGE_SIZE I915_GTT_PAGE_SIZE_4K
51 #define I915_GTT_MAX_PAGE_SIZE I915_GTT_PAGE_SIZE_2M
52 
53 #define I915_GTT_PAGE_MASK -I915_GTT_PAGE_SIZE
54 
55 #define I915_GTT_MIN_ALIGNMENT I915_GTT_PAGE_SIZE
56 
57 #define I915_FENCE_REG_NONE -1
58 #define I915_MAX_NUM_FENCES 32
59 /* 32 fences + sign bit for FENCE_REG_NONE */
60 #define I915_MAX_NUM_FENCE_BITS 6
61 
62 struct drm_i915_file_private;
63 struct drm_i915_fence_reg;
64 struct i915_vma;
65 
66 typedef u32 gen6_pte_t;
67 typedef u64 gen8_pte_t;
68 typedef u64 gen8_pde_t;
69 typedef u64 gen8_ppgtt_pdpe_t;
70 typedef u64 gen8_ppgtt_pml4e_t;
71 
72 #define ggtt_total_entries(ggtt) ((ggtt)->vm.total >> PAGE_SHIFT)
73 
74 /* gen6-hsw has bit 11-4 for physical addr bit 39-32 */
75 #define GEN6_GTT_ADDR_ENCODE(addr)	((addr) | (((addr) >> 28) & 0xff0))
76 #define GEN6_PTE_ADDR_ENCODE(addr)	GEN6_GTT_ADDR_ENCODE(addr)
77 #define GEN6_PDE_ADDR_ENCODE(addr)	GEN6_GTT_ADDR_ENCODE(addr)
78 #define GEN6_PTE_CACHE_LLC		(2 << 1)
79 #define GEN6_PTE_UNCACHED		(1 << 1)
80 #define GEN6_PTE_VALID			(1 << 0)
81 
82 #define I915_PTES(pte_len)		((unsigned int)(PAGE_SIZE / (pte_len)))
83 #define I915_PTE_MASK(pte_len)		(I915_PTES(pte_len) - 1)
84 #define I915_PDES			512
85 #define I915_PDE_MASK			(I915_PDES - 1)
86 #define NUM_PTE(pde_shift)     (1 << (pde_shift - PAGE_SHIFT))
87 
88 #define GEN6_PTES			I915_PTES(sizeof(gen6_pte_t))
89 #define GEN6_PD_SIZE		        (I915_PDES * PAGE_SIZE)
90 #define GEN6_PD_ALIGN			(PAGE_SIZE * 16)
91 #define GEN6_PDE_SHIFT			22
92 #define GEN6_PDE_VALID			(1 << 0)
93 
94 #define GEN7_PTE_CACHE_L3_LLC		(3 << 1)
95 
96 #define BYT_PTE_SNOOPED_BY_CPU_CACHES	(1 << 2)
97 #define BYT_PTE_WRITEABLE		(1 << 1)
98 
99 /* Cacheability Control is a 4-bit value. The low three bits are stored in bits
100  * 3:1 of the PTE, while the fourth bit is stored in bit 11 of the PTE.
101  */
102 #define HSW_CACHEABILITY_CONTROL(bits)	((((bits) & 0x7) << 1) | \
103 					 (((bits) & 0x8) << (11 - 3)))
104 #define HSW_WB_LLC_AGE3			HSW_CACHEABILITY_CONTROL(0x2)
105 #define HSW_WB_LLC_AGE0			HSW_CACHEABILITY_CONTROL(0x3)
106 #define HSW_WB_ELLC_LLC_AGE3		HSW_CACHEABILITY_CONTROL(0x8)
107 #define HSW_WB_ELLC_LLC_AGE0		HSW_CACHEABILITY_CONTROL(0xb)
108 #define HSW_WT_ELLC_LLC_AGE3		HSW_CACHEABILITY_CONTROL(0x7)
109 #define HSW_WT_ELLC_LLC_AGE0		HSW_CACHEABILITY_CONTROL(0x6)
110 #define HSW_PTE_UNCACHED		(0)
111 #define HSW_GTT_ADDR_ENCODE(addr)	((addr) | (((addr) >> 28) & 0x7f0))
112 #define HSW_PTE_ADDR_ENCODE(addr)	HSW_GTT_ADDR_ENCODE(addr)
113 
114 /* GEN8 32b style address is defined as a 3 level page table:
115  * 31:30 | 29:21 | 20:12 |  11:0
116  * PDPE  |  PDE  |  PTE  | offset
117  * The difference as compared to normal x86 3 level page table is the PDPEs are
118  * programmed via register.
119  */
120 #define GEN8_3LVL_PDPES			4
121 #define GEN8_PDE_SHIFT			21
122 #define GEN8_PDE_MASK			0x1ff
123 #define GEN8_PTE_SHIFT			12
124 #define GEN8_PTE_MASK			0x1ff
125 #define GEN8_PTES			I915_PTES(sizeof(gen8_pte_t))
126 
127 /* GEN8 48b style address is defined as a 4 level page table:
128  * 47:39 | 38:30 | 29:21 | 20:12 |  11:0
129  * PML4E | PDPE  |  PDE  |  PTE  | offset
130  */
131 #define GEN8_PML4ES_PER_PML4		512
132 #define GEN8_PML4E_SHIFT		39
133 #define GEN8_PML4E_MASK			(GEN8_PML4ES_PER_PML4 - 1)
134 #define GEN8_PDPE_SHIFT			30
135 /* NB: GEN8_PDPE_MASK is untrue for 32b platforms, but it has no impact on 32b page
136  * tables */
137 #define GEN8_PDPE_MASK			0x1ff
138 
139 #define PPAT_UNCACHED			(_PAGE_PWT | _PAGE_PCD)
140 #define PPAT_CACHED_PDE			0 /* WB LLC */
141 #define PPAT_CACHED			_PAGE_PAT /* WB LLCeLLC */
142 #define PPAT_DISPLAY_ELLC		_PAGE_PCD /* WT eLLC */
143 
144 #define CHV_PPAT_SNOOP			(1<<6)
145 #define GEN8_PPAT_AGE(x)		((x)<<4)
146 #define GEN8_PPAT_LLCeLLC		(3<<2)
147 #define GEN8_PPAT_LLCELLC		(2<<2)
148 #define GEN8_PPAT_LLC			(1<<2)
149 #define GEN8_PPAT_WB			(3<<0)
150 #define GEN8_PPAT_WT			(2<<0)
151 #define GEN8_PPAT_WC			(1<<0)
152 #define GEN8_PPAT_UC			(0<<0)
153 #define GEN8_PPAT_ELLC_OVERRIDE		(0<<2)
154 #define GEN8_PPAT(i, x)			((u64)(x) << ((i) * 8))
155 
156 #define GEN8_PPAT_GET_CA(x) ((x) & 3)
157 #define GEN8_PPAT_GET_TC(x) ((x) & (3 << 2))
158 #define GEN8_PPAT_GET_AGE(x) ((x) & (3 << 4))
159 #define CHV_PPAT_GET_SNOOP(x) ((x) & (1 << 6))
160 
161 #define GEN8_PDE_IPS_64K BIT(11)
162 #define GEN8_PDE_PS_2M   BIT(7)
163 
164 struct sg_table;
165 
166 struct intel_remapped_plane_info {
167 	/* in gtt pages */
168 	unsigned int width, height, stride, offset;
169 } __packed;
170 
171 struct intel_remapped_info {
172 	struct intel_remapped_plane_info plane[2];
173 	unsigned int unused_mbz;
174 } __packed;
175 
176 struct intel_rotation_info {
177 	struct intel_remapped_plane_info plane[2];
178 } __packed;
179 
180 struct intel_partial_info {
181 	u64 offset;
182 	unsigned int size;
183 } __packed;
184 
185 enum i915_ggtt_view_type {
186 	I915_GGTT_VIEW_NORMAL = 0,
187 	I915_GGTT_VIEW_ROTATED = sizeof(struct intel_rotation_info),
188 	I915_GGTT_VIEW_PARTIAL = sizeof(struct intel_partial_info),
189 	I915_GGTT_VIEW_REMAPPED = sizeof(struct intel_remapped_info),
190 };
191 
192 static inline void assert_i915_gem_gtt_types(void)
193 {
194 	BUILD_BUG_ON(sizeof(struct intel_rotation_info) != 8*sizeof(unsigned int));
195 	BUILD_BUG_ON(sizeof(struct intel_partial_info) != sizeof(u64) + sizeof(unsigned int));
196 	BUILD_BUG_ON(sizeof(struct intel_remapped_info) != 9*sizeof(unsigned int));
197 
198 	/* Check that rotation/remapped shares offsets for simplicity */
199 	BUILD_BUG_ON(offsetof(struct intel_remapped_info, plane[0]) !=
200 		     offsetof(struct intel_rotation_info, plane[0]));
201 	BUILD_BUG_ON(offsetofend(struct intel_remapped_info, plane[1]) !=
202 		     offsetofend(struct intel_rotation_info, plane[1]));
203 
204 	/* As we encode the size of each branch inside the union into its type,
205 	 * we have to be careful that each branch has a unique size.
206 	 */
207 	switch ((enum i915_ggtt_view_type)0) {
208 	case I915_GGTT_VIEW_NORMAL:
209 	case I915_GGTT_VIEW_PARTIAL:
210 	case I915_GGTT_VIEW_ROTATED:
211 	case I915_GGTT_VIEW_REMAPPED:
212 		/* gcc complains if these are identical cases */
213 		break;
214 	}
215 }
216 
217 struct i915_ggtt_view {
218 	enum i915_ggtt_view_type type;
219 	union {
220 		/* Members need to contain no holes/padding */
221 		struct intel_partial_info partial;
222 		struct intel_rotation_info rotated;
223 		struct intel_remapped_info remapped;
224 	};
225 };
226 
227 enum i915_cache_level;
228 
229 struct i915_vma;
230 
231 struct i915_page_dma {
232 	struct page *page;
233 	union {
234 		dma_addr_t daddr;
235 
236 		/* For gen6/gen7 only. This is the offset in the GGTT
237 		 * where the page directory entries for PPGTT begin
238 		 */
239 		u32 ggtt_offset;
240 	};
241 };
242 
243 #define px_base(px) (&(px)->base)
244 #define px_dma(px) (px_base(px)->daddr)
245 
246 struct i915_page_table {
247 	struct i915_page_dma base;
248 	unsigned int used_ptes;
249 };
250 
251 struct i915_page_directory {
252 	struct i915_page_dma base;
253 
254 	struct i915_page_table *page_table[I915_PDES]; /* PDEs */
255 	unsigned int used_pdes;
256 };
257 
258 struct i915_page_directory_pointer {
259 	struct i915_page_dma base;
260 	struct i915_page_directory **page_directory;
261 	unsigned int used_pdpes;
262 };
263 
264 struct i915_pml4 {
265 	struct i915_page_dma base;
266 	struct i915_page_directory_pointer *pdps[GEN8_PML4ES_PER_PML4];
267 };
268 
269 struct i915_vma_ops {
270 	/* Map an object into an address space with the given cache flags. */
271 	int (*bind_vma)(struct i915_vma *vma,
272 			enum i915_cache_level cache_level,
273 			u32 flags);
274 	/*
275 	 * Unmap an object from an address space. This usually consists of
276 	 * setting the valid PTE entries to a reserved scratch page.
277 	 */
278 	void (*unbind_vma)(struct i915_vma *vma);
279 
280 	int (*set_pages)(struct i915_vma *vma);
281 	void (*clear_pages)(struct i915_vma *vma);
282 };
283 
284 struct pagestash {
285 	spinlock_t lock;
286 	struct pagevec pvec;
287 };
288 
289 struct i915_address_space {
290 	struct drm_mm mm;
291 	struct drm_i915_private *i915;
292 	struct device *dma;
293 	/* Every address space belongs to a struct file - except for the global
294 	 * GTT that is owned by the driver (and so @file is set to NULL). In
295 	 * principle, no information should leak from one context to another
296 	 * (or between files/processes etc) unless explicitly shared by the
297 	 * owner. Tracking the owner is important in order to free up per-file
298 	 * objects along with the file, to aide resource tracking, and to
299 	 * assign blame.
300 	 */
301 	struct drm_i915_file_private *file;
302 	u64 total;		/* size addr space maps (ex. 2GB for ggtt) */
303 	u64 reserved;		/* size addr space reserved */
304 
305 	bool closed;
306 
307 	struct mutex mutex; /* protects vma and our lists */
308 #define VM_CLASS_GGTT 0
309 #define VM_CLASS_PPGTT 1
310 
311 	u64 scratch_pte;
312 	int scratch_order;
313 	struct i915_page_dma scratch_page;
314 	struct i915_page_table *scratch_pt;
315 	struct i915_page_directory *scratch_pd;
316 	struct i915_page_directory_pointer *scratch_pdp; /* GEN8+ & 48b PPGTT */
317 
318 	/**
319 	 * List of vma currently bound.
320 	 */
321 	struct list_head bound_list;
322 
323 	/**
324 	 * List of vma that are not unbound.
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_4lvl(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_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 	struct drm_mm_node uc_fw;
405 };
406 
407 struct i915_hw_ppgtt {
408 	struct i915_address_space vm;
409 	struct kref ref;
410 
411 	intel_engine_mask_t pd_dirty_engines;
412 	union {
413 		struct i915_pml4 pml4;		/* GEN8+ & 48b PPGTT */
414 		struct i915_page_directory_pointer pdp;	/* GEN8+ */
415 		struct i915_page_directory pd;		/* GEN6-7 */
416 	};
417 };
418 
419 struct gen6_hw_ppgtt {
420 	struct i915_hw_ppgtt base;
421 
422 	struct i915_vma *vma;
423 	gen6_pte_t __iomem *pd_addr;
424 
425 	unsigned int pin_count;
426 	bool scan_for_unused_pt;
427 
428 	struct gen6_ppgtt_cleanup_work *work;
429 };
430 
431 #define __to_gen6_ppgtt(base) container_of(base, struct gen6_hw_ppgtt, base)
432 
433 static inline struct gen6_hw_ppgtt *to_gen6_ppgtt(struct i915_hw_ppgtt *base)
434 {
435 	BUILD_BUG_ON(offsetof(struct gen6_hw_ppgtt, base));
436 	return __to_gen6_ppgtt(base);
437 }
438 
439 /*
440  * gen6_for_each_pde() iterates over every pde from start until start+length.
441  * If start and start+length are not perfectly divisible, the macro will round
442  * down and up as needed. Start=0 and length=2G effectively iterates over
443  * every PDE in the system. The macro modifies ALL its parameters except 'pd',
444  * so each of the other parameters should preferably be a simple variable, or
445  * at most an lvalue with no side-effects!
446  */
447 #define gen6_for_each_pde(pt, pd, start, length, iter)			\
448 	for (iter = gen6_pde_index(start);				\
449 	     length > 0 && iter < I915_PDES &&				\
450 		(pt = (pd)->page_table[iter], true);			\
451 	     ({ u32 temp = ALIGN(start+1, 1 << GEN6_PDE_SHIFT);		\
452 		    temp = min(temp - start, length);			\
453 		    start += temp, length -= temp; }), ++iter)
454 
455 #define gen6_for_all_pdes(pt, pd, iter)					\
456 	for (iter = 0;							\
457 	     iter < I915_PDES &&					\
458 		(pt = (pd)->page_table[iter], true);			\
459 	     ++iter)
460 
461 static inline u32 i915_pte_index(u64 address, unsigned int pde_shift)
462 {
463 	const u32 mask = NUM_PTE(pde_shift) - 1;
464 
465 	return (address >> PAGE_SHIFT) & mask;
466 }
467 
468 /* Helper to counts the number of PTEs within the given length. This count
469  * does not cross a page table boundary, so the max value would be
470  * GEN6_PTES for GEN6, and GEN8_PTES for GEN8.
471 */
472 static inline u32 i915_pte_count(u64 addr, u64 length, unsigned int pde_shift)
473 {
474 	const u64 mask = ~((1ULL << pde_shift) - 1);
475 	u64 end;
476 
477 	GEM_BUG_ON(length == 0);
478 	GEM_BUG_ON(offset_in_page(addr | length));
479 
480 	end = addr + length;
481 
482 	if ((addr & mask) != (end & mask))
483 		return NUM_PTE(pde_shift) - i915_pte_index(addr, pde_shift);
484 
485 	return i915_pte_index(end, pde_shift) - i915_pte_index(addr, pde_shift);
486 }
487 
488 static inline u32 i915_pde_index(u64 addr, u32 shift)
489 {
490 	return (addr >> shift) & I915_PDE_MASK;
491 }
492 
493 static inline u32 gen6_pte_index(u32 addr)
494 {
495 	return i915_pte_index(addr, GEN6_PDE_SHIFT);
496 }
497 
498 static inline u32 gen6_pte_count(u32 addr, u32 length)
499 {
500 	return i915_pte_count(addr, length, GEN6_PDE_SHIFT);
501 }
502 
503 static inline u32 gen6_pde_index(u32 addr)
504 {
505 	return i915_pde_index(addr, GEN6_PDE_SHIFT);
506 }
507 
508 static inline unsigned int
509 i915_pdpes_per_pdp(const struct i915_address_space *vm)
510 {
511 	if (i915_vm_is_4lvl(vm))
512 		return GEN8_PML4ES_PER_PML4;
513 
514 	return GEN8_3LVL_PDPES;
515 }
516 
517 /* Equivalent to the gen6 version, For each pde iterates over every pde
518  * between from start until start + length. On gen8+ it simply iterates
519  * over every page directory entry in a page directory.
520  */
521 #define gen8_for_each_pde(pt, pd, start, length, iter)			\
522 	for (iter = gen8_pde_index(start);				\
523 	     length > 0 && iter < I915_PDES &&				\
524 		(pt = (pd)->page_table[iter], true);			\
525 	     ({ u64 temp = ALIGN(start+1, 1 << GEN8_PDE_SHIFT);		\
526 		    temp = min(temp - start, length);			\
527 		    start += temp, length -= temp; }), ++iter)
528 
529 #define gen8_for_each_pdpe(pd, pdp, start, length, iter)		\
530 	for (iter = gen8_pdpe_index(start);				\
531 	     length > 0 && iter < i915_pdpes_per_pdp(vm) &&		\
532 		(pd = (pdp)->page_directory[iter], true);		\
533 	     ({ u64 temp = ALIGN(start+1, 1 << GEN8_PDPE_SHIFT);	\
534 		    temp = min(temp - start, length);			\
535 		    start += temp, length -= temp; }), ++iter)
536 
537 #define gen8_for_each_pml4e(pdp, pml4, start, length, iter)		\
538 	for (iter = gen8_pml4e_index(start);				\
539 	     length > 0 && iter < GEN8_PML4ES_PER_PML4 &&		\
540 		(pdp = (pml4)->pdps[iter], true);			\
541 	     ({ u64 temp = ALIGN(start+1, 1ULL << GEN8_PML4E_SHIFT);	\
542 		    temp = min(temp - start, length);			\
543 		    start += temp, length -= temp; }), ++iter)
544 
545 static inline u32 gen8_pte_index(u64 address)
546 {
547 	return i915_pte_index(address, GEN8_PDE_SHIFT);
548 }
549 
550 static inline u32 gen8_pde_index(u64 address)
551 {
552 	return i915_pde_index(address, GEN8_PDE_SHIFT);
553 }
554 
555 static inline u32 gen8_pdpe_index(u64 address)
556 {
557 	return (address >> GEN8_PDPE_SHIFT) & GEN8_PDPE_MASK;
558 }
559 
560 static inline u32 gen8_pml4e_index(u64 address)
561 {
562 	return (address >> GEN8_PML4E_SHIFT) & GEN8_PML4E_MASK;
563 }
564 
565 static inline u64 gen8_pte_count(u64 address, u64 length)
566 {
567 	return i915_pte_count(address, length, GEN8_PDE_SHIFT);
568 }
569 
570 static inline dma_addr_t
571 i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n)
572 {
573 	return px_dma(ppgtt->pdp.page_directory[n]);
574 }
575 
576 static inline struct i915_ggtt *
577 i915_vm_to_ggtt(struct i915_address_space *vm)
578 {
579 	GEM_BUG_ON(!i915_is_ggtt(vm));
580 	return container_of(vm, struct i915_ggtt, vm);
581 }
582 
583 #define INTEL_MAX_PPAT_ENTRIES 8
584 #define INTEL_PPAT_PERFECT_MATCH (~0U)
585 
586 struct intel_ppat;
587 
588 struct intel_ppat_entry {
589 	struct intel_ppat *ppat;
590 	struct kref ref;
591 	u8 value;
592 };
593 
594 struct intel_ppat {
595 	struct intel_ppat_entry entries[INTEL_MAX_PPAT_ENTRIES];
596 	DECLARE_BITMAP(used, INTEL_MAX_PPAT_ENTRIES);
597 	DECLARE_BITMAP(dirty, INTEL_MAX_PPAT_ENTRIES);
598 	unsigned int max_entries;
599 	u8 clear_value;
600 	/*
601 	 * Return a score to show how two PPAT values match,
602 	 * a INTEL_PPAT_PERFECT_MATCH indicates a perfect match
603 	 */
604 	unsigned int (*match)(u8 src, u8 dst);
605 	void (*update_hw)(struct drm_i915_private *i915);
606 
607 	struct drm_i915_private *i915;
608 };
609 
610 const struct intel_ppat_entry *
611 intel_ppat_get(struct drm_i915_private *i915, u8 value);
612 void intel_ppat_put(const struct intel_ppat_entry *entry);
613 
614 int i915_gem_init_aliasing_ppgtt(struct drm_i915_private *i915);
615 void i915_gem_fini_aliasing_ppgtt(struct drm_i915_private *i915);
616 
617 int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv);
618 int i915_ggtt_init_hw(struct drm_i915_private *dev_priv);
619 int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv);
620 void i915_ggtt_enable_guc(struct drm_i915_private *i915);
621 void i915_ggtt_disable_guc(struct drm_i915_private *i915);
622 int i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
623 void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv);
624 
625 int i915_ppgtt_init_hw(struct drm_i915_private *dev_priv);
626 
627 struct i915_hw_ppgtt *i915_ppgtt_create(struct drm_i915_private *dev_priv);
628 void i915_ppgtt_release(struct kref *kref);
629 
630 static inline struct i915_hw_ppgtt *i915_ppgtt_get(struct i915_hw_ppgtt *ppgtt)
631 {
632 	kref_get(&ppgtt->ref);
633 	return ppgtt;
634 }
635 
636 static inline void i915_ppgtt_put(struct i915_hw_ppgtt *ppgtt)
637 {
638 	if (ppgtt)
639 		kref_put(&ppgtt->ref, i915_ppgtt_release);
640 }
641 
642 int gen6_ppgtt_pin(struct i915_hw_ppgtt *base);
643 void gen6_ppgtt_unpin(struct i915_hw_ppgtt *base);
644 void gen6_ppgtt_unpin_all(struct i915_hw_ppgtt *base);
645 
646 void i915_check_and_clear_faults(struct drm_i915_private *dev_priv);
647 void i915_gem_suspend_gtt_mappings(struct drm_i915_private *dev_priv);
648 void i915_gem_restore_gtt_mappings(struct drm_i915_private *dev_priv);
649 
650 int __must_check i915_gem_gtt_prepare_pages(struct drm_i915_gem_object *obj,
651 					    struct sg_table *pages);
652 void i915_gem_gtt_finish_pages(struct drm_i915_gem_object *obj,
653 			       struct sg_table *pages);
654 
655 int i915_gem_gtt_reserve(struct i915_address_space *vm,
656 			 struct drm_mm_node *node,
657 			 u64 size, u64 offset, unsigned long color,
658 			 unsigned int flags);
659 
660 int i915_gem_gtt_insert(struct i915_address_space *vm,
661 			struct drm_mm_node *node,
662 			u64 size, u64 alignment, unsigned long color,
663 			u64 start, u64 end, unsigned int flags);
664 
665 /* Flags used by pin/bind&friends. */
666 #define PIN_NONBLOCK		BIT_ULL(0)
667 #define PIN_NONFAULT		BIT_ULL(1)
668 #define PIN_NOEVICT		BIT_ULL(2)
669 #define PIN_MAPPABLE		BIT_ULL(3)
670 #define PIN_ZONE_4G		BIT_ULL(4)
671 #define PIN_HIGH		BIT_ULL(5)
672 #define PIN_OFFSET_BIAS		BIT_ULL(6)
673 #define PIN_OFFSET_FIXED	BIT_ULL(7)
674 
675 #define PIN_MBZ			BIT_ULL(8) /* I915_VMA_PIN_OVERFLOW */
676 #define PIN_GLOBAL		BIT_ULL(9) /* I915_VMA_GLOBAL_BIND */
677 #define PIN_USER		BIT_ULL(10) /* I915_VMA_LOCAL_BIND */
678 #define PIN_UPDATE		BIT_ULL(11)
679 
680 #define PIN_OFFSET_MASK		(-I915_GTT_PAGE_SIZE)
681 
682 #endif
683