1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright (C) 2020 Google LLC
4 * Author: Will Deacon <will@kernel.org>
5 */
6
7 #ifndef __ARM64_KVM_PGTABLE_H__
8 #define __ARM64_KVM_PGTABLE_H__
9
10 #include <linux/bits.h>
11 #include <linux/kvm_host.h>
12 #include <linux/types.h>
13
14 #define KVM_PGTABLE_FIRST_LEVEL -1
15 #define KVM_PGTABLE_LAST_LEVEL 3
16
17 /*
18 * The largest supported block sizes for KVM (no 52-bit PA support):
19 * - 4K (level 1): 1GB
20 * - 16K (level 2): 32MB
21 * - 64K (level 2): 512MB
22 */
23 #ifdef CONFIG_ARM64_4K_PAGES
24 #define KVM_PGTABLE_MIN_BLOCK_LEVEL 1
25 #else
26 #define KVM_PGTABLE_MIN_BLOCK_LEVEL 2
27 #endif
28
29 #define kvm_lpa2_is_enabled() system_supports_lpa2()
30
kvm_get_parange_max(void)31 static inline u64 kvm_get_parange_max(void)
32 {
33 if (kvm_lpa2_is_enabled() ||
34 (IS_ENABLED(CONFIG_ARM64_PA_BITS_52) && PAGE_SHIFT == 16))
35 return ID_AA64MMFR0_EL1_PARANGE_52;
36 else
37 return ID_AA64MMFR0_EL1_PARANGE_48;
38 }
39
kvm_get_parange(u64 mmfr0)40 static inline u64 kvm_get_parange(u64 mmfr0)
41 {
42 u64 parange_max = kvm_get_parange_max();
43 u64 parange = cpuid_feature_extract_unsigned_field(mmfr0,
44 ID_AA64MMFR0_EL1_PARANGE_SHIFT);
45 if (parange > parange_max)
46 parange = parange_max;
47
48 return parange;
49 }
50
51 typedef u64 kvm_pte_t;
52
53 #define KVM_PTE_VALID BIT(0)
54
55 #define KVM_PTE_ADDR_MASK GENMASK(47, PAGE_SHIFT)
56 #define KVM_PTE_ADDR_51_48 GENMASK(15, 12)
57 #define KVM_PTE_ADDR_MASK_LPA2 GENMASK(49, PAGE_SHIFT)
58 #define KVM_PTE_ADDR_51_50_LPA2 GENMASK(9, 8)
59
60 #define KVM_PHYS_INVALID (-1ULL)
61
kvm_pte_valid(kvm_pte_t pte)62 static inline bool kvm_pte_valid(kvm_pte_t pte)
63 {
64 return pte & KVM_PTE_VALID;
65 }
66
kvm_pte_to_phys(kvm_pte_t pte)67 static inline u64 kvm_pte_to_phys(kvm_pte_t pte)
68 {
69 u64 pa;
70
71 if (kvm_lpa2_is_enabled()) {
72 pa = pte & KVM_PTE_ADDR_MASK_LPA2;
73 pa |= FIELD_GET(KVM_PTE_ADDR_51_50_LPA2, pte) << 50;
74 } else {
75 pa = pte & KVM_PTE_ADDR_MASK;
76 if (PAGE_SHIFT == 16)
77 pa |= FIELD_GET(KVM_PTE_ADDR_51_48, pte) << 48;
78 }
79
80 return pa;
81 }
82
kvm_phys_to_pte(u64 pa)83 static inline kvm_pte_t kvm_phys_to_pte(u64 pa)
84 {
85 kvm_pte_t pte;
86
87 if (kvm_lpa2_is_enabled()) {
88 pte = pa & KVM_PTE_ADDR_MASK_LPA2;
89 pa &= GENMASK(51, 50);
90 pte |= FIELD_PREP(KVM_PTE_ADDR_51_50_LPA2, pa >> 50);
91 } else {
92 pte = pa & KVM_PTE_ADDR_MASK;
93 if (PAGE_SHIFT == 16) {
94 pa &= GENMASK(51, 48);
95 pte |= FIELD_PREP(KVM_PTE_ADDR_51_48, pa >> 48);
96 }
97 }
98
99 return pte;
100 }
101
kvm_pte_to_pfn(kvm_pte_t pte)102 static inline kvm_pfn_t kvm_pte_to_pfn(kvm_pte_t pte)
103 {
104 return __phys_to_pfn(kvm_pte_to_phys(pte));
105 }
106
kvm_granule_shift(s8 level)107 static inline u64 kvm_granule_shift(s8 level)
108 {
109 /* Assumes KVM_PGTABLE_LAST_LEVEL is 3 */
110 return ARM64_HW_PGTABLE_LEVEL_SHIFT(level);
111 }
112
kvm_granule_size(s8 level)113 static inline u64 kvm_granule_size(s8 level)
114 {
115 return BIT(kvm_granule_shift(level));
116 }
117
kvm_level_supports_block_mapping(s8 level)118 static inline bool kvm_level_supports_block_mapping(s8 level)
119 {
120 return level >= KVM_PGTABLE_MIN_BLOCK_LEVEL;
121 }
122
kvm_supported_block_sizes(void)123 static inline u32 kvm_supported_block_sizes(void)
124 {
125 s8 level = KVM_PGTABLE_MIN_BLOCK_LEVEL;
126 u32 r = 0;
127
128 for (; level <= KVM_PGTABLE_LAST_LEVEL; level++)
129 r |= BIT(kvm_granule_shift(level));
130
131 return r;
132 }
133
kvm_is_block_size_supported(u64 size)134 static inline bool kvm_is_block_size_supported(u64 size)
135 {
136 bool is_power_of_two = IS_ALIGNED(size, size);
137
138 return is_power_of_two && (size & kvm_supported_block_sizes());
139 }
140
141 /**
142 * struct kvm_pgtable_mm_ops - Memory management callbacks.
143 * @zalloc_page: Allocate a single zeroed memory page.
144 * The @arg parameter can be used by the walker
145 * to pass a memcache. The initial refcount of
146 * the page is 1.
147 * @zalloc_pages_exact: Allocate an exact number of zeroed memory pages.
148 * The @size parameter is in bytes, and is rounded
149 * up to the next page boundary. The resulting
150 * allocation is physically contiguous.
151 * @free_pages_exact: Free an exact number of memory pages previously
152 * allocated by zalloc_pages_exact.
153 * @free_unlinked_table: Free an unlinked paging structure by unlinking and
154 * dropping references.
155 * @get_page: Increment the refcount on a page.
156 * @put_page: Decrement the refcount on a page. When the
157 * refcount reaches 0 the page is automatically
158 * freed.
159 * @page_count: Return the refcount of a page.
160 * @phys_to_virt: Convert a physical address into a virtual
161 * address mapped in the current context.
162 * @virt_to_phys: Convert a virtual address mapped in the current
163 * context into a physical address.
164 * @dcache_clean_inval_poc: Clean and invalidate the data cache to the PoC
165 * for the specified memory address range.
166 * @icache_inval_pou: Invalidate the instruction cache to the PoU
167 * for the specified memory address range.
168 */
169 struct kvm_pgtable_mm_ops {
170 void* (*zalloc_page)(void *arg);
171 void* (*zalloc_pages_exact)(size_t size);
172 void (*free_pages_exact)(void *addr, size_t size);
173 void (*free_unlinked_table)(void *addr, s8 level);
174 void (*get_page)(void *addr);
175 void (*put_page)(void *addr);
176 int (*page_count)(void *addr);
177 void* (*phys_to_virt)(phys_addr_t phys);
178 phys_addr_t (*virt_to_phys)(void *addr);
179 void (*dcache_clean_inval_poc)(void *addr, size_t size);
180 void (*icache_inval_pou)(void *addr, size_t size);
181 };
182
183 /**
184 * enum kvm_pgtable_stage2_flags - Stage-2 page-table flags.
185 * @KVM_PGTABLE_S2_NOFWB: Don't enforce Normal-WB even if the CPUs have
186 * ARM64_HAS_STAGE2_FWB.
187 * @KVM_PGTABLE_S2_IDMAP: Only use identity mappings.
188 */
189 enum kvm_pgtable_stage2_flags {
190 KVM_PGTABLE_S2_NOFWB = BIT(0),
191 KVM_PGTABLE_S2_IDMAP = BIT(1),
192 };
193
194 /**
195 * enum kvm_pgtable_prot - Page-table permissions and attributes.
196 * @KVM_PGTABLE_PROT_X: Execute permission.
197 * @KVM_PGTABLE_PROT_W: Write permission.
198 * @KVM_PGTABLE_PROT_R: Read permission.
199 * @KVM_PGTABLE_PROT_DEVICE: Device attributes.
200 * @KVM_PGTABLE_PROT_NORMAL_NC: Normal noncacheable attributes.
201 * @KVM_PGTABLE_PROT_SW0: Software bit 0.
202 * @KVM_PGTABLE_PROT_SW1: Software bit 1.
203 * @KVM_PGTABLE_PROT_SW2: Software bit 2.
204 * @KVM_PGTABLE_PROT_SW3: Software bit 3.
205 */
206 enum kvm_pgtable_prot {
207 KVM_PGTABLE_PROT_X = BIT(0),
208 KVM_PGTABLE_PROT_W = BIT(1),
209 KVM_PGTABLE_PROT_R = BIT(2),
210
211 KVM_PGTABLE_PROT_DEVICE = BIT(3),
212 KVM_PGTABLE_PROT_NORMAL_NC = BIT(4),
213
214 KVM_PGTABLE_PROT_SW0 = BIT(55),
215 KVM_PGTABLE_PROT_SW1 = BIT(56),
216 KVM_PGTABLE_PROT_SW2 = BIT(57),
217 KVM_PGTABLE_PROT_SW3 = BIT(58),
218 };
219
220 #define KVM_PGTABLE_PROT_RW (KVM_PGTABLE_PROT_R | KVM_PGTABLE_PROT_W)
221 #define KVM_PGTABLE_PROT_RWX (KVM_PGTABLE_PROT_RW | KVM_PGTABLE_PROT_X)
222
223 #define PKVM_HOST_MEM_PROT KVM_PGTABLE_PROT_RWX
224 #define PKVM_HOST_MMIO_PROT KVM_PGTABLE_PROT_RW
225
226 #define PAGE_HYP KVM_PGTABLE_PROT_RW
227 #define PAGE_HYP_EXEC (KVM_PGTABLE_PROT_R | KVM_PGTABLE_PROT_X)
228 #define PAGE_HYP_RO (KVM_PGTABLE_PROT_R)
229 #define PAGE_HYP_DEVICE (PAGE_HYP | KVM_PGTABLE_PROT_DEVICE)
230
231 typedef bool (*kvm_pgtable_force_pte_cb_t)(u64 addr, u64 end,
232 enum kvm_pgtable_prot prot);
233
234 /**
235 * enum kvm_pgtable_walk_flags - Flags to control a depth-first page-table walk.
236 * @KVM_PGTABLE_WALK_LEAF: Visit leaf entries, including invalid
237 * entries.
238 * @KVM_PGTABLE_WALK_TABLE_PRE: Visit table entries before their
239 * children.
240 * @KVM_PGTABLE_WALK_TABLE_POST: Visit table entries after their
241 * children.
242 * @KVM_PGTABLE_WALK_SHARED: Indicates the page-tables may be shared
243 * with other software walkers.
244 * @KVM_PGTABLE_WALK_HANDLE_FAULT: Indicates the page-table walk was
245 * invoked from a fault handler.
246 * @KVM_PGTABLE_WALK_SKIP_BBM_TLBI: Visit and update table entries
247 * without Break-before-make's
248 * TLB invalidation.
249 * @KVM_PGTABLE_WALK_SKIP_CMO: Visit and update table entries
250 * without Cache maintenance
251 * operations required.
252 */
253 enum kvm_pgtable_walk_flags {
254 KVM_PGTABLE_WALK_LEAF = BIT(0),
255 KVM_PGTABLE_WALK_TABLE_PRE = BIT(1),
256 KVM_PGTABLE_WALK_TABLE_POST = BIT(2),
257 KVM_PGTABLE_WALK_SHARED = BIT(3),
258 KVM_PGTABLE_WALK_HANDLE_FAULT = BIT(4),
259 KVM_PGTABLE_WALK_SKIP_BBM_TLBI = BIT(5),
260 KVM_PGTABLE_WALK_SKIP_CMO = BIT(6),
261 };
262
263 struct kvm_pgtable_visit_ctx {
264 kvm_pte_t *ptep;
265 kvm_pte_t old;
266 void *arg;
267 struct kvm_pgtable_mm_ops *mm_ops;
268 u64 start;
269 u64 addr;
270 u64 end;
271 s8 level;
272 enum kvm_pgtable_walk_flags flags;
273 };
274
275 typedef int (*kvm_pgtable_visitor_fn_t)(const struct kvm_pgtable_visit_ctx *ctx,
276 enum kvm_pgtable_walk_flags visit);
277
kvm_pgtable_walk_shared(const struct kvm_pgtable_visit_ctx * ctx)278 static inline bool kvm_pgtable_walk_shared(const struct kvm_pgtable_visit_ctx *ctx)
279 {
280 return ctx->flags & KVM_PGTABLE_WALK_SHARED;
281 }
282
283 /**
284 * struct kvm_pgtable_walker - Hook into a page-table walk.
285 * @cb: Callback function to invoke during the walk.
286 * @arg: Argument passed to the callback function.
287 * @flags: Bitwise-OR of flags to identify the entry types on which to
288 * invoke the callback function.
289 */
290 struct kvm_pgtable_walker {
291 const kvm_pgtable_visitor_fn_t cb;
292 void * const arg;
293 const enum kvm_pgtable_walk_flags flags;
294 };
295
296 /*
297 * RCU cannot be used in a non-kernel context such as the hyp. As such, page
298 * table walkers used in hyp do not call into RCU and instead use other
299 * synchronization mechanisms (such as a spinlock).
300 */
301 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__)
302
303 typedef kvm_pte_t *kvm_pteref_t;
304
kvm_dereference_pteref(struct kvm_pgtable_walker * walker,kvm_pteref_t pteref)305 static inline kvm_pte_t *kvm_dereference_pteref(struct kvm_pgtable_walker *walker,
306 kvm_pteref_t pteref)
307 {
308 return pteref;
309 }
310
kvm_pgtable_walk_begin(struct kvm_pgtable_walker * walker)311 static inline int kvm_pgtable_walk_begin(struct kvm_pgtable_walker *walker)
312 {
313 /*
314 * Due to the lack of RCU (or a similar protection scheme), only
315 * non-shared table walkers are allowed in the hypervisor.
316 */
317 if (walker->flags & KVM_PGTABLE_WALK_SHARED)
318 return -EPERM;
319
320 return 0;
321 }
322
kvm_pgtable_walk_end(struct kvm_pgtable_walker * walker)323 static inline void kvm_pgtable_walk_end(struct kvm_pgtable_walker *walker) {}
324
kvm_pgtable_walk_lock_held(void)325 static inline bool kvm_pgtable_walk_lock_held(void)
326 {
327 return true;
328 }
329
330 #else
331
332 typedef kvm_pte_t __rcu *kvm_pteref_t;
333
kvm_dereference_pteref(struct kvm_pgtable_walker * walker,kvm_pteref_t pteref)334 static inline kvm_pte_t *kvm_dereference_pteref(struct kvm_pgtable_walker *walker,
335 kvm_pteref_t pteref)
336 {
337 return rcu_dereference_check(pteref, !(walker->flags & KVM_PGTABLE_WALK_SHARED));
338 }
339
kvm_pgtable_walk_begin(struct kvm_pgtable_walker * walker)340 static inline int kvm_pgtable_walk_begin(struct kvm_pgtable_walker *walker)
341 {
342 if (walker->flags & KVM_PGTABLE_WALK_SHARED)
343 rcu_read_lock();
344
345 return 0;
346 }
347
kvm_pgtable_walk_end(struct kvm_pgtable_walker * walker)348 static inline void kvm_pgtable_walk_end(struct kvm_pgtable_walker *walker)
349 {
350 if (walker->flags & KVM_PGTABLE_WALK_SHARED)
351 rcu_read_unlock();
352 }
353
kvm_pgtable_walk_lock_held(void)354 static inline bool kvm_pgtable_walk_lock_held(void)
355 {
356 return rcu_read_lock_held();
357 }
358
359 #endif
360
361 /**
362 * struct kvm_pgtable - KVM page-table.
363 * @ia_bits: Maximum input address size, in bits.
364 * @start_level: Level at which the page-table walk starts.
365 * @pgd: Pointer to the first top-level entry of the page-table.
366 * @mm_ops: Memory management callbacks.
367 * @mmu: Stage-2 KVM MMU struct. Unused for stage-1 page-tables.
368 * @flags: Stage-2 page-table flags.
369 * @force_pte_cb: Function that returns true if page level mappings must
370 * be used instead of block mappings.
371 */
372 struct kvm_pgtable {
373 u32 ia_bits;
374 s8 start_level;
375 kvm_pteref_t pgd;
376 struct kvm_pgtable_mm_ops *mm_ops;
377
378 /* Stage-2 only */
379 struct kvm_s2_mmu *mmu;
380 enum kvm_pgtable_stage2_flags flags;
381 kvm_pgtable_force_pte_cb_t force_pte_cb;
382 };
383
384 /**
385 * kvm_pgtable_hyp_init() - Initialise a hypervisor stage-1 page-table.
386 * @pgt: Uninitialised page-table structure to initialise.
387 * @va_bits: Maximum virtual address bits.
388 * @mm_ops: Memory management callbacks.
389 *
390 * Return: 0 on success, negative error code on failure.
391 */
392 int kvm_pgtable_hyp_init(struct kvm_pgtable *pgt, u32 va_bits,
393 struct kvm_pgtable_mm_ops *mm_ops);
394
395 /**
396 * kvm_pgtable_hyp_destroy() - Destroy an unused hypervisor stage-1 page-table.
397 * @pgt: Page-table structure initialised by kvm_pgtable_hyp_init().
398 *
399 * The page-table is assumed to be unreachable by any hardware walkers prior
400 * to freeing and therefore no TLB invalidation is performed.
401 */
402 void kvm_pgtable_hyp_destroy(struct kvm_pgtable *pgt);
403
404 /**
405 * kvm_pgtable_hyp_map() - Install a mapping in a hypervisor stage-1 page-table.
406 * @pgt: Page-table structure initialised by kvm_pgtable_hyp_init().
407 * @addr: Virtual address at which to place the mapping.
408 * @size: Size of the mapping.
409 * @phys: Physical address of the memory to map.
410 * @prot: Permissions and attributes for the mapping.
411 *
412 * The offset of @addr within a page is ignored, @size is rounded-up to
413 * the next page boundary and @phys is rounded-down to the previous page
414 * boundary.
415 *
416 * If device attributes are not explicitly requested in @prot, then the
417 * mapping will be normal, cacheable. Attempts to install a new mapping
418 * for a virtual address that is already mapped will be rejected with an
419 * error and a WARN().
420 *
421 * Return: 0 on success, negative error code on failure.
422 */
423 int kvm_pgtable_hyp_map(struct kvm_pgtable *pgt, u64 addr, u64 size, u64 phys,
424 enum kvm_pgtable_prot prot);
425
426 /**
427 * kvm_pgtable_hyp_unmap() - Remove a mapping from a hypervisor stage-1 page-table.
428 * @pgt: Page-table structure initialised by kvm_pgtable_hyp_init().
429 * @addr: Virtual address from which to remove the mapping.
430 * @size: Size of the mapping.
431 *
432 * The offset of @addr within a page is ignored, @size is rounded-up to
433 * the next page boundary and @phys is rounded-down to the previous page
434 * boundary.
435 *
436 * TLB invalidation is performed for each page-table entry cleared during the
437 * unmapping operation and the reference count for the page-table page
438 * containing the cleared entry is decremented, with unreferenced pages being
439 * freed. The unmapping operation will stop early if it encounters either an
440 * invalid page-table entry or a valid block mapping which maps beyond the range
441 * being unmapped.
442 *
443 * Return: Number of bytes unmapped, which may be 0.
444 */
445 u64 kvm_pgtable_hyp_unmap(struct kvm_pgtable *pgt, u64 addr, u64 size);
446
447 /**
448 * kvm_get_vtcr() - Helper to construct VTCR_EL2
449 * @mmfr0: Sanitized value of SYS_ID_AA64MMFR0_EL1 register.
450 * @mmfr1: Sanitized value of SYS_ID_AA64MMFR1_EL1 register.
451 * @phys_shfit: Value to set in VTCR_EL2.T0SZ.
452 *
453 * The VTCR value is common across all the physical CPUs on the system.
454 * We use system wide sanitised values to fill in different fields,
455 * except for Hardware Management of Access Flags. HA Flag is set
456 * unconditionally on all CPUs, as it is safe to run with or without
457 * the feature and the bit is RES0 on CPUs that don't support it.
458 *
459 * Return: VTCR_EL2 value
460 */
461 u64 kvm_get_vtcr(u64 mmfr0, u64 mmfr1, u32 phys_shift);
462
463 /**
464 * kvm_pgtable_stage2_pgd_size() - Helper to compute size of a stage-2 PGD
465 * @vtcr: Content of the VTCR register.
466 *
467 * Return: the size (in bytes) of the stage-2 PGD
468 */
469 size_t kvm_pgtable_stage2_pgd_size(u64 vtcr);
470
471 /**
472 * __kvm_pgtable_stage2_init() - Initialise a guest stage-2 page-table.
473 * @pgt: Uninitialised page-table structure to initialise.
474 * @mmu: S2 MMU context for this S2 translation
475 * @mm_ops: Memory management callbacks.
476 * @flags: Stage-2 configuration flags.
477 * @force_pte_cb: Function that returns true if page level mappings must
478 * be used instead of block mappings.
479 *
480 * Return: 0 on success, negative error code on failure.
481 */
482 int __kvm_pgtable_stage2_init(struct kvm_pgtable *pgt, struct kvm_s2_mmu *mmu,
483 struct kvm_pgtable_mm_ops *mm_ops,
484 enum kvm_pgtable_stage2_flags flags,
485 kvm_pgtable_force_pte_cb_t force_pte_cb);
486
487 #define kvm_pgtable_stage2_init(pgt, mmu, mm_ops) \
488 __kvm_pgtable_stage2_init(pgt, mmu, mm_ops, 0, NULL)
489
490 /**
491 * kvm_pgtable_stage2_destroy() - Destroy an unused guest stage-2 page-table.
492 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
493 *
494 * The page-table is assumed to be unreachable by any hardware walkers prior
495 * to freeing and therefore no TLB invalidation is performed.
496 */
497 void kvm_pgtable_stage2_destroy(struct kvm_pgtable *pgt);
498
499 /**
500 * kvm_pgtable_stage2_free_unlinked() - Free an unlinked stage-2 paging structure.
501 * @mm_ops: Memory management callbacks.
502 * @pgtable: Unlinked stage-2 paging structure to be freed.
503 * @level: Level of the stage-2 paging structure to be freed.
504 *
505 * The page-table is assumed to be unreachable by any hardware walkers prior to
506 * freeing and therefore no TLB invalidation is performed.
507 */
508 void kvm_pgtable_stage2_free_unlinked(struct kvm_pgtable_mm_ops *mm_ops, void *pgtable, s8 level);
509
510 /**
511 * kvm_pgtable_stage2_create_unlinked() - Create an unlinked stage-2 paging structure.
512 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
513 * @phys: Physical address of the memory to map.
514 * @level: Starting level of the stage-2 paging structure to be created.
515 * @prot: Permissions and attributes for the mapping.
516 * @mc: Cache of pre-allocated and zeroed memory from which to allocate
517 * page-table pages.
518 * @force_pte: Force mappings to PAGE_SIZE granularity.
519 *
520 * Returns an unlinked page-table tree. This new page-table tree is
521 * not reachable (i.e., it is unlinked) from the root pgd and it's
522 * therefore unreachableby the hardware page-table walker. No TLB
523 * invalidation or CMOs are performed.
524 *
525 * If device attributes are not explicitly requested in @prot, then the
526 * mapping will be normal, cacheable.
527 *
528 * Return: The fully populated (unlinked) stage-2 paging structure, or
529 * an ERR_PTR(error) on failure.
530 */
531 kvm_pte_t *kvm_pgtable_stage2_create_unlinked(struct kvm_pgtable *pgt,
532 u64 phys, s8 level,
533 enum kvm_pgtable_prot prot,
534 void *mc, bool force_pte);
535
536 /**
537 * kvm_pgtable_stage2_map() - Install a mapping in a guest stage-2 page-table.
538 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
539 * @addr: Intermediate physical address at which to place the mapping.
540 * @size: Size of the mapping.
541 * @phys: Physical address of the memory to map.
542 * @prot: Permissions and attributes for the mapping.
543 * @mc: Cache of pre-allocated and zeroed memory from which to allocate
544 * page-table pages.
545 * @flags: Flags to control the page-table walk (ex. a shared walk)
546 *
547 * The offset of @addr within a page is ignored, @size is rounded-up to
548 * the next page boundary and @phys is rounded-down to the previous page
549 * boundary.
550 *
551 * If device attributes are not explicitly requested in @prot, then the
552 * mapping will be normal, cacheable.
553 *
554 * Note that the update of a valid leaf PTE in this function will be aborted,
555 * if it's trying to recreate the exact same mapping or only change the access
556 * permissions. Instead, the vCPU will exit one more time from guest if still
557 * needed and then go through the path of relaxing permissions.
558 *
559 * Note that this function will both coalesce existing table entries and split
560 * existing block mappings, relying on page-faults to fault back areas outside
561 * of the new mapping lazily.
562 *
563 * Return: 0 on success, negative error code on failure.
564 */
565 int kvm_pgtable_stage2_map(struct kvm_pgtable *pgt, u64 addr, u64 size,
566 u64 phys, enum kvm_pgtable_prot prot,
567 void *mc, enum kvm_pgtable_walk_flags flags);
568
569 /**
570 * kvm_pgtable_stage2_set_owner() - Unmap and annotate pages in the IPA space to
571 * track ownership.
572 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
573 * @addr: Base intermediate physical address to annotate.
574 * @size: Size of the annotated range.
575 * @mc: Cache of pre-allocated and zeroed memory from which to allocate
576 * page-table pages.
577 * @owner_id: Unique identifier for the owner of the page.
578 *
579 * By default, all page-tables are owned by identifier 0. This function can be
580 * used to mark portions of the IPA space as owned by other entities. When a
581 * stage 2 is used with identity-mappings, these annotations allow to use the
582 * page-table data structure as a simple rmap.
583 *
584 * Return: 0 on success, negative error code on failure.
585 */
586 int kvm_pgtable_stage2_set_owner(struct kvm_pgtable *pgt, u64 addr, u64 size,
587 void *mc, u8 owner_id);
588
589 /**
590 * kvm_pgtable_stage2_unmap() - Remove a mapping from a guest stage-2 page-table.
591 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
592 * @addr: Intermediate physical address from which to remove the mapping.
593 * @size: Size of the mapping.
594 *
595 * The offset of @addr within a page is ignored and @size is rounded-up to
596 * the next page boundary.
597 *
598 * TLB invalidation is performed for each page-table entry cleared during the
599 * unmapping operation and the reference count for the page-table page
600 * containing the cleared entry is decremented, with unreferenced pages being
601 * freed. Unmapping a cacheable page will ensure that it is clean to the PoC if
602 * FWB is not supported by the CPU.
603 *
604 * Return: 0 on success, negative error code on failure.
605 */
606 int kvm_pgtable_stage2_unmap(struct kvm_pgtable *pgt, u64 addr, u64 size);
607
608 /**
609 * kvm_pgtable_stage2_wrprotect() - Write-protect guest stage-2 address range
610 * without TLB invalidation.
611 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
612 * @addr: Intermediate physical address from which to write-protect,
613 * @size: Size of the range.
614 *
615 * The offset of @addr within a page is ignored and @size is rounded-up to
616 * the next page boundary.
617 *
618 * Note that it is the caller's responsibility to invalidate the TLB after
619 * calling this function to ensure that the updated permissions are visible
620 * to the CPUs.
621 *
622 * Return: 0 on success, negative error code on failure.
623 */
624 int kvm_pgtable_stage2_wrprotect(struct kvm_pgtable *pgt, u64 addr, u64 size);
625
626 /**
627 * kvm_pgtable_stage2_mkyoung() - Set the access flag in a page-table entry.
628 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
629 * @addr: Intermediate physical address to identify the page-table entry.
630 *
631 * The offset of @addr within a page is ignored.
632 *
633 * If there is a valid, leaf page-table entry used to translate @addr, then
634 * set the access flag in that entry.
635 *
636 * Return: The old page-table entry prior to setting the flag, 0 on failure.
637 */
638 kvm_pte_t kvm_pgtable_stage2_mkyoung(struct kvm_pgtable *pgt, u64 addr);
639
640 /**
641 * kvm_pgtable_stage2_test_clear_young() - Test and optionally clear the access
642 * flag in a page-table entry.
643 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
644 * @addr: Intermediate physical address to identify the page-table entry.
645 * @size: Size of the address range to visit.
646 * @mkold: True if the access flag should be cleared.
647 *
648 * The offset of @addr within a page is ignored.
649 *
650 * Tests and conditionally clears the access flag for every valid, leaf
651 * page-table entry used to translate the range [@addr, @addr + @size).
652 *
653 * Note that it is the caller's responsibility to invalidate the TLB after
654 * calling this function to ensure that the updated permissions are visible
655 * to the CPUs.
656 *
657 * Return: True if any of the visited PTEs had the access flag set.
658 */
659 bool kvm_pgtable_stage2_test_clear_young(struct kvm_pgtable *pgt, u64 addr,
660 u64 size, bool mkold);
661
662 /**
663 * kvm_pgtable_stage2_relax_perms() - Relax the permissions enforced by a
664 * page-table entry.
665 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
666 * @addr: Intermediate physical address to identify the page-table entry.
667 * @prot: Additional permissions to grant for the mapping.
668 *
669 * The offset of @addr within a page is ignored.
670 *
671 * If there is a valid, leaf page-table entry used to translate @addr, then
672 * relax the permissions in that entry according to the read, write and
673 * execute permissions specified by @prot. No permissions are removed, and
674 * TLB invalidation is performed after updating the entry. Software bits cannot
675 * be set or cleared using kvm_pgtable_stage2_relax_perms().
676 *
677 * Return: 0 on success, negative error code on failure.
678 */
679 int kvm_pgtable_stage2_relax_perms(struct kvm_pgtable *pgt, u64 addr,
680 enum kvm_pgtable_prot prot);
681
682 /**
683 * kvm_pgtable_stage2_flush_range() - Clean and invalidate data cache to Point
684 * of Coherency for guest stage-2 address
685 * range.
686 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init*().
687 * @addr: Intermediate physical address from which to flush.
688 * @size: Size of the range.
689 *
690 * The offset of @addr within a page is ignored and @size is rounded-up to
691 * the next page boundary.
692 *
693 * Return: 0 on success, negative error code on failure.
694 */
695 int kvm_pgtable_stage2_flush(struct kvm_pgtable *pgt, u64 addr, u64 size);
696
697 /**
698 * kvm_pgtable_stage2_split() - Split a range of huge pages into leaf PTEs pointing
699 * to PAGE_SIZE guest pages.
700 * @pgt: Page-table structure initialised by kvm_pgtable_stage2_init().
701 * @addr: Intermediate physical address from which to split.
702 * @size: Size of the range.
703 * @mc: Cache of pre-allocated and zeroed memory from which to allocate
704 * page-table pages.
705 *
706 * The function tries to split any level 1 or 2 entry that overlaps
707 * with the input range (given by @addr and @size).
708 *
709 * Return: 0 on success, negative error code on failure. Note that
710 * kvm_pgtable_stage2_split() is best effort: it tries to break as many
711 * blocks in the input range as allowed by @mc_capacity.
712 */
713 int kvm_pgtable_stage2_split(struct kvm_pgtable *pgt, u64 addr, u64 size,
714 struct kvm_mmu_memory_cache *mc);
715
716 /**
717 * kvm_pgtable_walk() - Walk a page-table.
718 * @pgt: Page-table structure initialised by kvm_pgtable_*_init().
719 * @addr: Input address for the start of the walk.
720 * @size: Size of the range to walk.
721 * @walker: Walker callback description.
722 *
723 * The offset of @addr within a page is ignored and @size is rounded-up to
724 * the next page boundary.
725 *
726 * The walker will walk the page-table entries corresponding to the input
727 * address range specified, visiting entries according to the walker flags.
728 * Invalid entries are treated as leaf entries. The visited page table entry is
729 * reloaded after invoking the walker callback, allowing the walker to descend
730 * into a newly installed table.
731 *
732 * Returning a negative error code from the walker callback function will
733 * terminate the walk immediately with the same error code.
734 *
735 * Return: 0 on success, negative error code on failure.
736 */
737 int kvm_pgtable_walk(struct kvm_pgtable *pgt, u64 addr, u64 size,
738 struct kvm_pgtable_walker *walker);
739
740 /**
741 * kvm_pgtable_get_leaf() - Walk a page-table and retrieve the leaf entry
742 * with its level.
743 * @pgt: Page-table structure initialised by kvm_pgtable_*_init()
744 * or a similar initialiser.
745 * @addr: Input address for the start of the walk.
746 * @ptep: Pointer to storage for the retrieved PTE.
747 * @level: Pointer to storage for the level of the retrieved PTE.
748 *
749 * The offset of @addr within a page is ignored.
750 *
751 * The walker will walk the page-table entries corresponding to the input
752 * address specified, retrieving the leaf corresponding to this address.
753 * Invalid entries are treated as leaf entries.
754 *
755 * Return: 0 on success, negative error code on failure.
756 */
757 int kvm_pgtable_get_leaf(struct kvm_pgtable *pgt, u64 addr,
758 kvm_pte_t *ptep, s8 *level);
759
760 /**
761 * kvm_pgtable_stage2_pte_prot() - Retrieve the protection attributes of a
762 * stage-2 Page-Table Entry.
763 * @pte: Page-table entry
764 *
765 * Return: protection attributes of the page-table entry in the enum
766 * kvm_pgtable_prot format.
767 */
768 enum kvm_pgtable_prot kvm_pgtable_stage2_pte_prot(kvm_pte_t pte);
769
770 /**
771 * kvm_pgtable_hyp_pte_prot() - Retrieve the protection attributes of a stage-1
772 * Page-Table Entry.
773 * @pte: Page-table entry
774 *
775 * Return: protection attributes of the page-table entry in the enum
776 * kvm_pgtable_prot format.
777 */
778 enum kvm_pgtable_prot kvm_pgtable_hyp_pte_prot(kvm_pte_t pte);
779
780 /**
781 * kvm_tlb_flush_vmid_range() - Invalidate/flush a range of TLB entries
782 *
783 * @mmu: Stage-2 KVM MMU struct
784 * @addr: The base Intermediate physical address from which to invalidate
785 * @size: Size of the range from the base to invalidate
786 */
787 void kvm_tlb_flush_vmid_range(struct kvm_s2_mmu *mmu,
788 phys_addr_t addr, size_t size);
789 #endif /* __ARM64_KVM_PGTABLE_H__ */
790