Warning: This file is not a C or C++ file. It does not have highlighting.
| 1 | /* SPDX-License-Identifier: GPL-2.0-only */ |
|---|---|
| 2 | /* |
| 3 | * |
| 4 | * Copyright SUSE Linux Products GmbH 2010 |
| 5 | * |
| 6 | * Authors: Alexander Graf <agraf@suse.de> |
| 7 | */ |
| 8 | |
| 9 | #ifndef __ASM_KVM_BOOK3S_64_H__ |
| 10 | #define __ASM_KVM_BOOK3S_64_H__ |
| 11 | |
| 12 | #include <linux/string.h> |
| 13 | #include <asm/bitops.h> |
| 14 | #include <asm/book3s/64/mmu-hash.h> |
| 15 | #include <asm/cpu_has_feature.h> |
| 16 | #include <asm/ppc-opcode.h> |
| 17 | #include <asm/pte-walk.h> |
| 18 | |
| 19 | /* |
| 20 | * Structure for a nested guest, that is, for a guest that is managed by |
| 21 | * one of our guests. |
| 22 | */ |
| 23 | struct kvm_nested_guest { |
| 24 | struct kvm *l1_host; /* L1 VM that owns this nested guest */ |
| 25 | int l1_lpid; /* lpid L1 guest thinks this guest is */ |
| 26 | int shadow_lpid; /* real lpid of this nested guest */ |
| 27 | pgd_t *shadow_pgtable; /* our page table for this guest */ |
| 28 | u64 l1_gr_to_hr; /* L1's addr of part'n-scoped table */ |
| 29 | u64 process_table; /* process table entry for this guest */ |
| 30 | long refcnt; /* number of pointers to this struct */ |
| 31 | struct mutex tlb_lock; /* serialize page faults and tlbies */ |
| 32 | struct kvm_nested_guest *next; |
| 33 | cpumask_t need_tlb_flush; |
| 34 | short prev_cpu[NR_CPUS]; |
| 35 | u8 radix; /* is this nested guest radix */ |
| 36 | }; |
| 37 | |
| 38 | /* |
| 39 | * We define a nested rmap entry as a single 64-bit quantity |
| 40 | * 0xFFF0000000000000 12-bit lpid field |
| 41 | * 0x000FFFFFFFFFF000 40-bit guest 4k page frame number |
| 42 | * 0x0000000000000001 1-bit single entry flag |
| 43 | */ |
| 44 | #define RMAP_NESTED_LPID_MASK 0xFFF0000000000000UL |
| 45 | #define RMAP_NESTED_LPID_SHIFT (52) |
| 46 | #define RMAP_NESTED_GPA_MASK 0x000FFFFFFFFFF000UL |
| 47 | #define RMAP_NESTED_IS_SINGLE_ENTRY 0x0000000000000001UL |
| 48 | |
| 49 | /* Structure for a nested guest rmap entry */ |
| 50 | struct rmap_nested { |
| 51 | struct llist_node list; |
| 52 | u64 rmap; |
| 53 | }; |
| 54 | |
| 55 | /* |
| 56 | * for_each_nest_rmap_safe - iterate over the list of nested rmap entries |
| 57 | * safe against removal of the list entry or NULL list |
| 58 | * @pos: a (struct rmap_nested *) to use as a loop cursor |
| 59 | * @node: pointer to the first entry |
| 60 | * NOTE: this can be NULL |
| 61 | * @rmapp: an (unsigned long *) in which to return the rmap entries on each |
| 62 | * iteration |
| 63 | * NOTE: this must point to already allocated memory |
| 64 | * |
| 65 | * The nested_rmap is a llist of (struct rmap_nested) entries pointed to by the |
| 66 | * rmap entry in the memslot. The list is always terminated by a "single entry" |
| 67 | * stored in the list element of the final entry of the llist. If there is ONLY |
| 68 | * a single entry then this is itself in the rmap entry of the memslot, not a |
| 69 | * llist head pointer. |
| 70 | * |
| 71 | * Note that the iterator below assumes that a nested rmap entry is always |
| 72 | * non-zero. This is true for our usage because the LPID field is always |
| 73 | * non-zero (zero is reserved for the host). |
| 74 | * |
| 75 | * This should be used to iterate over the list of rmap_nested entries with |
| 76 | * processing done on the u64 rmap value given by each iteration. This is safe |
| 77 | * against removal of list entries and it is always safe to call free on (pos). |
| 78 | * |
| 79 | * e.g. |
| 80 | * struct rmap_nested *cursor; |
| 81 | * struct llist_node *first; |
| 82 | * unsigned long rmap; |
| 83 | * for_each_nest_rmap_safe(cursor, first, &rmap) { |
| 84 | * do_something(rmap); |
| 85 | * free(cursor); |
| 86 | * } |
| 87 | */ |
| 88 | #define for_each_nest_rmap_safe(pos, node, rmapp) \ |
| 89 | for ((pos) = llist_entry((node), typeof(*(pos)), list); \ |
| 90 | (node) && \ |
| 91 | (*(rmapp) = ((RMAP_NESTED_IS_SINGLE_ENTRY & ((u64) (node))) ? \ |
| 92 | ((u64) (node)) : ((pos)->rmap))) && \ |
| 93 | (((node) = ((RMAP_NESTED_IS_SINGLE_ENTRY & ((u64) (node))) ? \ |
| 94 | ((struct llist_node *) ((pos) = NULL)) : \ |
| 95 | (pos)->list.next)), true); \ |
| 96 | (pos) = llist_entry((node), typeof(*(pos)), list)) |
| 97 | |
| 98 | struct kvm_nested_guest *kvmhv_get_nested(struct kvm *kvm, int l1_lpid, |
| 99 | bool create); |
| 100 | void kvmhv_put_nested(struct kvm_nested_guest *gp); |
| 101 | int kvmhv_nested_next_lpid(struct kvm *kvm, int lpid); |
| 102 | |
| 103 | /* Encoding of first parameter for H_TLB_INVALIDATE */ |
| 104 | #define H_TLBIE_P1_ENC(ric, prs, r) (___PPC_RIC(ric) | ___PPC_PRS(prs) | \ |
| 105 | ___PPC_R(r)) |
| 106 | |
| 107 | /* Power architecture requires HPT is at least 256kiB, at most 64TiB */ |
| 108 | #define PPC_MIN_HPT_ORDER 18 |
| 109 | #define PPC_MAX_HPT_ORDER 46 |
| 110 | |
| 111 | #ifdef CONFIG_KVM_BOOK3S_PR_POSSIBLE |
| 112 | static inline struct kvmppc_book3s_shadow_vcpu *svcpu_get(struct kvm_vcpu *vcpu) |
| 113 | { |
| 114 | preempt_disable(); |
| 115 | return &get_paca()->shadow_vcpu; |
| 116 | } |
| 117 | |
| 118 | static inline void svcpu_put(struct kvmppc_book3s_shadow_vcpu *svcpu) |
| 119 | { |
| 120 | preempt_enable(); |
| 121 | } |
| 122 | #endif |
| 123 | |
| 124 | #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE |
| 125 | |
| 126 | static inline bool kvm_is_radix(struct kvm *kvm) |
| 127 | { |
| 128 | return kvm->arch.radix; |
| 129 | } |
| 130 | |
| 131 | static inline bool kvmhv_vcpu_is_radix(struct kvm_vcpu *vcpu) |
| 132 | { |
| 133 | bool radix; |
| 134 | |
| 135 | if (vcpu->arch.nested) |
| 136 | radix = vcpu->arch.nested->radix; |
| 137 | else |
| 138 | radix = kvm_is_radix(vcpu->kvm); |
| 139 | |
| 140 | return radix; |
| 141 | } |
| 142 | |
| 143 | unsigned long kvmppc_msr_hard_disable_set_facilities(struct kvm_vcpu *vcpu, unsigned long msr); |
| 144 | |
| 145 | int kvmhv_vcpu_entry_p9(struct kvm_vcpu *vcpu, u64 time_limit, unsigned long lpcr, u64 *tb); |
| 146 | |
| 147 | #define KVM_DEFAULT_HPT_ORDER 24 /* 16MB HPT by default */ |
| 148 | #endif |
| 149 | |
| 150 | /* |
| 151 | * Invalid HDSISR value which is used to indicate when HW has not set the reg. |
| 152 | * Used to work around an errata. |
| 153 | */ |
| 154 | #define HDSISR_CANARY 0x7fff |
| 155 | |
| 156 | /* |
| 157 | * We use a lock bit in HPTE dword 0 to synchronize updates and |
| 158 | * accesses to each HPTE, and another bit to indicate non-present |
| 159 | * HPTEs. |
| 160 | */ |
| 161 | #define HPTE_V_HVLOCK 0x40UL |
| 162 | #define HPTE_V_ABSENT 0x20UL |
| 163 | |
| 164 | /* |
| 165 | * We use this bit in the guest_rpte field of the revmap entry |
| 166 | * to indicate a modified HPTE. |
| 167 | */ |
| 168 | #define HPTE_GR_MODIFIED (1ul << 62) |
| 169 | |
| 170 | /* These bits are reserved in the guest view of the HPTE */ |
| 171 | #define HPTE_GR_RESERVED HPTE_GR_MODIFIED |
| 172 | |
| 173 | static inline long try_lock_hpte(__be64 *hpte, unsigned long bits) |
| 174 | { |
| 175 | unsigned long tmp, old; |
| 176 | __be64 be_lockbit, be_bits; |
| 177 | |
| 178 | /* |
| 179 | * We load/store in native endian, but the HTAB is in big endian. If |
| 180 | * we byte swap all data we apply on the PTE we're implicitly correct |
| 181 | * again. |
| 182 | */ |
| 183 | be_lockbit = cpu_to_be64(HPTE_V_HVLOCK); |
| 184 | be_bits = cpu_to_be64(bits); |
| 185 | |
| 186 | asm volatile(" ldarx %0,0,%2\n" |
| 187 | " and. %1,%0,%3\n" |
| 188 | " bne 2f\n" |
| 189 | " or %0,%0,%4\n" |
| 190 | " stdcx. %0,0,%2\n" |
| 191 | " beq+ 2f\n" |
| 192 | " mr %1,%3\n" |
| 193 | "2: isync" |
| 194 | : "=&r" (tmp), "=&r" (old) |
| 195 | : "r" (hpte), "r" (be_bits), "r" (be_lockbit) |
| 196 | : "cc", "memory"); |
| 197 | return old == 0; |
| 198 | } |
| 199 | |
| 200 | static inline void unlock_hpte(__be64 *hpte, unsigned long hpte_v) |
| 201 | { |
| 202 | hpte_v &= ~HPTE_V_HVLOCK; |
| 203 | asm volatile(PPC_RELEASE_BARRIER "" : : : "memory"); |
| 204 | hpte[0] = cpu_to_be64(hpte_v); |
| 205 | } |
| 206 | |
| 207 | /* Without barrier */ |
| 208 | static inline void __unlock_hpte(__be64 *hpte, unsigned long hpte_v) |
| 209 | { |
| 210 | hpte_v &= ~HPTE_V_HVLOCK; |
| 211 | hpte[0] = cpu_to_be64(hpte_v); |
| 212 | } |
| 213 | |
| 214 | /* |
| 215 | * These functions encode knowledge of the POWER7/8/9 hardware |
| 216 | * interpretations of the HPTE LP (large page size) field. |
| 217 | */ |
| 218 | static inline int kvmppc_hpte_page_shifts(unsigned long h, unsigned long l) |
| 219 | { |
| 220 | unsigned int lphi; |
| 221 | |
| 222 | if (!(h & HPTE_V_LARGE)) |
| 223 | return 12; /* 4kB */ |
| 224 | lphi = (l >> 16) & 0xf; |
| 225 | switch ((l >> 12) & 0xf) { |
| 226 | case 0: |
| 227 | return !lphi ? 24 : 0; /* 16MB */ |
| 228 | break; |
| 229 | case 1: |
| 230 | return 16; /* 64kB */ |
| 231 | break; |
| 232 | case 3: |
| 233 | return !lphi ? 34 : 0; /* 16GB */ |
| 234 | break; |
| 235 | case 7: |
| 236 | return (16 << 8) + 12; /* 64kB in 4kB */ |
| 237 | break; |
| 238 | case 8: |
| 239 | if (!lphi) |
| 240 | return (24 << 8) + 16; /* 16MB in 64kkB */ |
| 241 | if (lphi == 3) |
| 242 | return (24 << 8) + 12; /* 16MB in 4kB */ |
| 243 | break; |
| 244 | } |
| 245 | return 0; |
| 246 | } |
| 247 | |
| 248 | static inline int kvmppc_hpte_base_page_shift(unsigned long h, unsigned long l) |
| 249 | { |
| 250 | return kvmppc_hpte_page_shifts(h, l) & 0xff; |
| 251 | } |
| 252 | |
| 253 | static inline int kvmppc_hpte_actual_page_shift(unsigned long h, unsigned long l) |
| 254 | { |
| 255 | int tmp = kvmppc_hpte_page_shifts(h, l); |
| 256 | |
| 257 | if (tmp >= 0x100) |
| 258 | tmp >>= 8; |
| 259 | return tmp; |
| 260 | } |
| 261 | |
| 262 | static inline unsigned long kvmppc_actual_pgsz(unsigned long v, unsigned long r) |
| 263 | { |
| 264 | int shift = kvmppc_hpte_actual_page_shift(v, r); |
| 265 | |
| 266 | if (shift) |
| 267 | return 1ul << shift; |
| 268 | return 0; |
| 269 | } |
| 270 | |
| 271 | static inline int kvmppc_pgsize_lp_encoding(int base_shift, int actual_shift) |
| 272 | { |
| 273 | switch (base_shift) { |
| 274 | case 12: |
| 275 | switch (actual_shift) { |
| 276 | case 12: |
| 277 | return 0; |
| 278 | case 16: |
| 279 | return 7; |
| 280 | case 24: |
| 281 | return 0x38; |
| 282 | } |
| 283 | break; |
| 284 | case 16: |
| 285 | switch (actual_shift) { |
| 286 | case 16: |
| 287 | return 1; |
| 288 | case 24: |
| 289 | return 8; |
| 290 | } |
| 291 | break; |
| 292 | case 24: |
| 293 | return 0; |
| 294 | } |
| 295 | return -1; |
| 296 | } |
| 297 | |
| 298 | static inline unsigned long compute_tlbie_rb(unsigned long v, unsigned long r, |
| 299 | unsigned long pte_index) |
| 300 | { |
| 301 | int a_pgshift, b_pgshift; |
| 302 | unsigned long rb = 0, va_low, sllp; |
| 303 | |
| 304 | b_pgshift = a_pgshift = kvmppc_hpte_page_shifts(v, r); |
| 305 | if (a_pgshift >= 0x100) { |
| 306 | b_pgshift &= 0xff; |
| 307 | a_pgshift >>= 8; |
| 308 | } |
| 309 | |
| 310 | /* |
| 311 | * Ignore the top 14 bits of va |
| 312 | * v have top two bits covering segment size, hence move |
| 313 | * by 16 bits, Also clear the lower HPTE_V_AVPN_SHIFT (7) bits. |
| 314 | * AVA field in v also have the lower 23 bits ignored. |
| 315 | * For base page size 4K we need 14 .. 65 bits (so need to |
| 316 | * collect extra 11 bits) |
| 317 | * For others we need 14..14+i |
| 318 | */ |
| 319 | /* This covers 14..54 bits of va*/ |
| 320 | rb = (v & ~0x7fUL) << 16; /* AVA field */ |
| 321 | |
| 322 | /* |
| 323 | * AVA in v had cleared lower 23 bits. We need to derive |
| 324 | * that from pteg index |
| 325 | */ |
| 326 | va_low = pte_index >> 3; |
| 327 | if (v & HPTE_V_SECONDARY) |
| 328 | va_low = ~va_low; |
| 329 | /* |
| 330 | * get the vpn bits from va_low using reverse of hashing. |
| 331 | * In v we have va with 23 bits dropped and then left shifted |
| 332 | * HPTE_V_AVPN_SHIFT (7) bits. Now to find vsid we need |
| 333 | * right shift it with (SID_SHIFT - (23 - 7)) |
| 334 | */ |
| 335 | if (!(v & HPTE_V_1TB_SEG)) |
| 336 | va_low ^= v >> (SID_SHIFT - 16); |
| 337 | else |
| 338 | va_low ^= v >> (SID_SHIFT_1T - 16); |
| 339 | va_low &= 0x7ff; |
| 340 | |
| 341 | if (b_pgshift <= 12) { |
| 342 | if (a_pgshift > 12) { |
| 343 | sllp = (a_pgshift == 16) ? 5 : 4; |
| 344 | rb |= sllp << 5; /* AP field */ |
| 345 | } |
| 346 | rb |= (va_low & 0x7ff) << 12; /* remaining 11 bits of AVA */ |
| 347 | } else { |
| 348 | int aval_shift; |
| 349 | /* |
| 350 | * remaining bits of AVA/LP fields |
| 351 | * Also contain the rr bits of LP |
| 352 | */ |
| 353 | rb |= (va_low << b_pgshift) & 0x7ff000; |
| 354 | /* |
| 355 | * Now clear not needed LP bits based on actual psize |
| 356 | */ |
| 357 | rb &= ~((1ul << a_pgshift) - 1); |
| 358 | /* |
| 359 | * AVAL field 58..77 - base_page_shift bits of va |
| 360 | * we have space for 58..64 bits, Missing bits should |
| 361 | * be zero filled. +1 is to take care of L bit shift |
| 362 | */ |
| 363 | aval_shift = 64 - (77 - b_pgshift) + 1; |
| 364 | rb |= ((va_low << aval_shift) & 0xfe); |
| 365 | |
| 366 | rb |= 1; /* L field */ |
| 367 | rb |= r & 0xff000 & ((1ul << a_pgshift) - 1); /* LP field */ |
| 368 | } |
| 369 | /* |
| 370 | * This sets both bits of the B field in the PTE. 0b1x values are |
| 371 | * reserved, but those will have been filtered by kvmppc_do_h_enter. |
| 372 | */ |
| 373 | rb |= (v >> HPTE_V_SSIZE_SHIFT) << 8; /* B field */ |
| 374 | return rb; |
| 375 | } |
| 376 | |
| 377 | static inline unsigned long hpte_rpn(unsigned long ptel, unsigned long psize) |
| 378 | { |
| 379 | return ((ptel & HPTE_R_RPN) & ~(psize - 1)) >> PAGE_SHIFT; |
| 380 | } |
| 381 | |
| 382 | static inline int hpte_is_writable(unsigned long ptel) |
| 383 | { |
| 384 | unsigned long pp = ptel & (HPTE_R_PP0 | HPTE_R_PP); |
| 385 | |
| 386 | return pp != PP_RXRX && pp != PP_RXXX; |
| 387 | } |
| 388 | |
| 389 | static inline unsigned long hpte_make_readonly(unsigned long ptel) |
| 390 | { |
| 391 | if ((ptel & HPTE_R_PP0) || (ptel & HPTE_R_PP) == PP_RWXX) |
| 392 | ptel = (ptel & ~HPTE_R_PP) | PP_RXXX; |
| 393 | else |
| 394 | ptel |= PP_RXRX; |
| 395 | return ptel; |
| 396 | } |
| 397 | |
| 398 | static inline bool hpte_cache_flags_ok(unsigned long hptel, bool is_ci) |
| 399 | { |
| 400 | unsigned int wimg = hptel & HPTE_R_WIMG; |
| 401 | |
| 402 | /* Handle SAO */ |
| 403 | if (wimg == (HPTE_R_W | HPTE_R_I | HPTE_R_M) && |
| 404 | cpu_has_feature(CPU_FTR_ARCH_206)) |
| 405 | wimg = HPTE_R_M; |
| 406 | |
| 407 | if (!is_ci) |
| 408 | return wimg == HPTE_R_M; |
| 409 | /* |
| 410 | * if host is mapped cache inhibited, make sure hptel also have |
| 411 | * cache inhibited. |
| 412 | */ |
| 413 | if (wimg & HPTE_R_W) /* FIXME!! is this ok for all guest. ? */ |
| 414 | return false; |
| 415 | return !!(wimg & HPTE_R_I); |
| 416 | } |
| 417 | |
| 418 | /* |
| 419 | * If it's present and writable, atomically set dirty and referenced bits and |
| 420 | * return the PTE, otherwise return 0. |
| 421 | */ |
| 422 | static inline pte_t kvmppc_read_update_linux_pte(pte_t *ptep, int writing) |
| 423 | { |
| 424 | pte_t old_pte, new_pte = __pte(0); |
| 425 | |
| 426 | while (1) { |
| 427 | /* |
| 428 | * Make sure we don't reload from ptep |
| 429 | */ |
| 430 | old_pte = READ_ONCE(*ptep); |
| 431 | /* |
| 432 | * wait until H_PAGE_BUSY is clear then set it atomically |
| 433 | */ |
| 434 | if (unlikely(pte_val(old_pte) & H_PAGE_BUSY)) { |
| 435 | cpu_relax(); |
| 436 | continue; |
| 437 | } |
| 438 | /* If pte is not present return None */ |
| 439 | if (unlikely(!pte_present(old_pte))) |
| 440 | return __pte(0); |
| 441 | |
| 442 | new_pte = pte_mkyoung(old_pte); |
| 443 | if (writing && pte_write(old_pte)) |
| 444 | new_pte = pte_mkdirty(new_pte); |
| 445 | |
| 446 | if (pte_xchg(ptep, old_pte, new_pte)) |
| 447 | break; |
| 448 | } |
| 449 | return new_pte; |
| 450 | } |
| 451 | |
| 452 | static inline bool hpte_read_permission(unsigned long pp, unsigned long key) |
| 453 | { |
| 454 | if (key) |
| 455 | return PP_RWRX <= pp && pp <= PP_RXRX; |
| 456 | return true; |
| 457 | } |
| 458 | |
| 459 | static inline bool hpte_write_permission(unsigned long pp, unsigned long key) |
| 460 | { |
| 461 | if (key) |
| 462 | return pp == PP_RWRW; |
| 463 | return pp <= PP_RWRW; |
| 464 | } |
| 465 | |
| 466 | static inline int hpte_get_skey_perm(unsigned long hpte_r, unsigned long amr) |
| 467 | { |
| 468 | unsigned long skey; |
| 469 | |
| 470 | skey = ((hpte_r & HPTE_R_KEY_HI) >> 57) | |
| 471 | ((hpte_r & HPTE_R_KEY_LO) >> 9); |
| 472 | return (amr >> (62 - 2 * skey)) & 3; |
| 473 | } |
| 474 | |
| 475 | static inline void lock_rmap(unsigned long *rmap) |
| 476 | { |
| 477 | do { |
| 478 | while (test_bit(KVMPPC_RMAP_LOCK_BIT, rmap)) |
| 479 | cpu_relax(); |
| 480 | } while (test_and_set_bit_lock(KVMPPC_RMAP_LOCK_BIT, rmap)); |
| 481 | } |
| 482 | |
| 483 | static inline void unlock_rmap(unsigned long *rmap) |
| 484 | { |
| 485 | __clear_bit_unlock(KVMPPC_RMAP_LOCK_BIT, rmap); |
| 486 | } |
| 487 | |
| 488 | static inline bool slot_is_aligned(struct kvm_memory_slot *memslot, |
| 489 | unsigned long pagesize) |
| 490 | { |
| 491 | unsigned long mask = (pagesize >> PAGE_SHIFT) - 1; |
| 492 | |
| 493 | if (pagesize <= PAGE_SIZE) |
| 494 | return true; |
| 495 | return !(memslot->base_gfn & mask) && !(memslot->npages & mask); |
| 496 | } |
| 497 | |
| 498 | /* |
| 499 | * This works for 4k, 64k and 16M pages on POWER7, |
| 500 | * and 4k and 16M pages on PPC970. |
| 501 | */ |
| 502 | static inline unsigned long slb_pgsize_encoding(unsigned long psize) |
| 503 | { |
| 504 | unsigned long senc = 0; |
| 505 | |
| 506 | if (psize > 0x1000) { |
| 507 | senc = SLB_VSID_L; |
| 508 | if (psize == 0x10000) |
| 509 | senc |= SLB_VSID_LP_01; |
| 510 | } |
| 511 | return senc; |
| 512 | } |
| 513 | |
| 514 | static inline int is_vrma_hpte(unsigned long hpte_v) |
| 515 | { |
| 516 | return (hpte_v & ~0xffffffUL) == |
| 517 | (HPTE_V_1TB_SEG | (VRMA_VSID << (40 - 16))); |
| 518 | } |
| 519 | |
| 520 | #ifdef CONFIG_KVM_BOOK3S_HV_POSSIBLE |
| 521 | /* |
| 522 | * Note modification of an HPTE; set the HPTE modified bit |
| 523 | * if anyone is interested. |
| 524 | */ |
| 525 | static inline void note_hpte_modification(struct kvm *kvm, |
| 526 | struct revmap_entry *rev) |
| 527 | { |
| 528 | if (atomic_read(&kvm->arch.hpte_mod_interest)) |
| 529 | rev->guest_rpte |= HPTE_GR_MODIFIED; |
| 530 | } |
| 531 | |
| 532 | /* |
| 533 | * Like kvm_memslots(), but for use in real mode when we can't do |
| 534 | * any RCU stuff (since the secondary threads are offline from the |
| 535 | * kernel's point of view), and we can't print anything. |
| 536 | * Thus we use rcu_dereference_raw() rather than rcu_dereference_check(). |
| 537 | */ |
| 538 | static inline struct kvm_memslots *kvm_memslots_raw(struct kvm *kvm) |
| 539 | { |
| 540 | return rcu_dereference_raw_check(kvm->memslots[0]); |
| 541 | } |
| 542 | |
| 543 | extern void kvmppc_mmu_debugfs_init(struct kvm *kvm); |
| 544 | extern void kvmhv_radix_debugfs_init(struct kvm *kvm); |
| 545 | |
| 546 | extern void kvmhv_rm_send_ipi(int cpu); |
| 547 | |
| 548 | static inline unsigned long kvmppc_hpt_npte(struct kvm_hpt_info *hpt) |
| 549 | { |
| 550 | /* HPTEs are 2**4 bytes long */ |
| 551 | return 1UL << (hpt->order - 4); |
| 552 | } |
| 553 | |
| 554 | static inline unsigned long kvmppc_hpt_mask(struct kvm_hpt_info *hpt) |
| 555 | { |
| 556 | /* 128 (2**7) bytes in each HPTEG */ |
| 557 | return (1UL << (hpt->order - 7)) - 1; |
| 558 | } |
| 559 | |
| 560 | /* Set bits in a dirty bitmap, which is in LE format */ |
| 561 | static inline void set_dirty_bits(unsigned long *map, unsigned long i, |
| 562 | unsigned long npages) |
| 563 | { |
| 564 | |
| 565 | if (npages >= 8) |
| 566 | memset((char *)map + i / 8, 0xff, npages / 8); |
| 567 | else |
| 568 | for (; npages; ++i, --npages) |
| 569 | __set_bit_le(i, map); |
| 570 | } |
| 571 | |
| 572 | static inline void set_dirty_bits_atomic(unsigned long *map, unsigned long i, |
| 573 | unsigned long npages) |
| 574 | { |
| 575 | if (npages >= 8) |
| 576 | memset((char *)map + i / 8, 0xff, npages / 8); |
| 577 | else |
| 578 | for (; npages; ++i, --npages) |
| 579 | set_bit_le(i, map); |
| 580 | } |
| 581 | |
| 582 | static inline u64 sanitize_msr(u64 msr) |
| 583 | { |
| 584 | msr &= ~MSR_HV; |
| 585 | msr |= MSR_ME; |
| 586 | return msr; |
| 587 | } |
| 588 | |
| 589 | #ifdef CONFIG_PPC_TRANSACTIONAL_MEM |
| 590 | static inline void copy_from_checkpoint(struct kvm_vcpu *vcpu) |
| 591 | { |
| 592 | vcpu->arch.regs.ccr = vcpu->arch.cr_tm; |
| 593 | vcpu->arch.regs.xer = vcpu->arch.xer_tm; |
| 594 | vcpu->arch.regs.link = vcpu->arch.lr_tm; |
| 595 | vcpu->arch.regs.ctr = vcpu->arch.ctr_tm; |
| 596 | vcpu->arch.amr = vcpu->arch.amr_tm; |
| 597 | vcpu->arch.ppr = vcpu->arch.ppr_tm; |
| 598 | vcpu->arch.dscr = vcpu->arch.dscr_tm; |
| 599 | vcpu->arch.tar = vcpu->arch.tar_tm; |
| 600 | memcpy(vcpu->arch.regs.gpr, vcpu->arch.gpr_tm, |
| 601 | sizeof(vcpu->arch.regs.gpr)); |
| 602 | vcpu->arch.fp = vcpu->arch.fp_tm; |
| 603 | vcpu->arch.vr = vcpu->arch.vr_tm; |
| 604 | vcpu->arch.vrsave = vcpu->arch.vrsave_tm; |
| 605 | } |
| 606 | |
| 607 | static inline void copy_to_checkpoint(struct kvm_vcpu *vcpu) |
| 608 | { |
| 609 | vcpu->arch.cr_tm = vcpu->arch.regs.ccr; |
| 610 | vcpu->arch.xer_tm = vcpu->arch.regs.xer; |
| 611 | vcpu->arch.lr_tm = vcpu->arch.regs.link; |
| 612 | vcpu->arch.ctr_tm = vcpu->arch.regs.ctr; |
| 613 | vcpu->arch.amr_tm = vcpu->arch.amr; |
| 614 | vcpu->arch.ppr_tm = vcpu->arch.ppr; |
| 615 | vcpu->arch.dscr_tm = vcpu->arch.dscr; |
| 616 | vcpu->arch.tar_tm = vcpu->arch.tar; |
| 617 | memcpy(vcpu->arch.gpr_tm, vcpu->arch.regs.gpr, |
| 618 | sizeof(vcpu->arch.regs.gpr)); |
| 619 | vcpu->arch.fp_tm = vcpu->arch.fp; |
| 620 | vcpu->arch.vr_tm = vcpu->arch.vr; |
| 621 | vcpu->arch.vrsave_tm = vcpu->arch.vrsave; |
| 622 | } |
| 623 | #endif /* CONFIG_PPC_TRANSACTIONAL_MEM */ |
| 624 | |
| 625 | extern int kvmppc_create_pte(struct kvm *kvm, pgd_t *pgtable, pte_t pte, |
| 626 | unsigned long gpa, unsigned int level, |
| 627 | unsigned long mmu_seq, u64 lpid, |
| 628 | unsigned long *rmapp, struct rmap_nested **n_rmap); |
| 629 | extern void kvmhv_insert_nest_rmap(struct kvm *kvm, unsigned long *rmapp, |
| 630 | struct rmap_nested **n_rmap); |
| 631 | extern void kvmhv_update_nest_rmap_rc_list(struct kvm *kvm, unsigned long *rmapp, |
| 632 | unsigned long clr, unsigned long set, |
| 633 | unsigned long hpa, unsigned long nbytes); |
| 634 | extern void kvmhv_remove_nest_rmap_range(struct kvm *kvm, |
| 635 | const struct kvm_memory_slot *memslot, |
| 636 | unsigned long gpa, unsigned long hpa, |
| 637 | unsigned long nbytes); |
| 638 | |
| 639 | static inline pte_t * |
| 640 | find_kvm_secondary_pte_unlocked(struct kvm *kvm, unsigned long ea, |
| 641 | unsigned *hshift) |
| 642 | { |
| 643 | pte_t *pte; |
| 644 | |
| 645 | pte = __find_linux_pte(kvm->arch.pgtable, ea, NULL, hshift); |
| 646 | return pte; |
| 647 | } |
| 648 | |
| 649 | static inline pte_t *find_kvm_secondary_pte(struct kvm *kvm, unsigned long ea, |
| 650 | unsigned *hshift) |
| 651 | { |
| 652 | pte_t *pte; |
| 653 | |
| 654 | VM_WARN(!spin_is_locked(&kvm->mmu_lock), |
| 655 | "%s called with kvm mmu_lock not held \n", __func__); |
| 656 | pte = __find_linux_pte(kvm->arch.pgtable, ea, NULL, hshift); |
| 657 | |
| 658 | return pte; |
| 659 | } |
| 660 | |
| 661 | static inline pte_t *find_kvm_host_pte(struct kvm *kvm, unsigned long mmu_seq, |
| 662 | unsigned long ea, unsigned *hshift) |
| 663 | { |
| 664 | pte_t *pte; |
| 665 | |
| 666 | VM_WARN(!spin_is_locked(&kvm->mmu_lock), |
| 667 | "%s called with kvm mmu_lock not held \n", __func__); |
| 668 | |
| 669 | if (mmu_invalidate_retry(kvm, mmu_seq)) |
| 670 | return NULL; |
| 671 | |
| 672 | pte = __find_linux_pte(kvm->mm->pgd, ea, NULL, hshift); |
| 673 | |
| 674 | return pte; |
| 675 | } |
| 676 | |
| 677 | extern pte_t *find_kvm_nested_guest_pte(struct kvm *kvm, unsigned long lpid, |
| 678 | unsigned long ea, unsigned *hshift); |
| 679 | |
| 680 | int kvmhv_nestedv2_vcpu_create(struct kvm_vcpu *vcpu, struct kvmhv_nestedv2_io *io); |
| 681 | void kvmhv_nestedv2_vcpu_free(struct kvm_vcpu *vcpu, struct kvmhv_nestedv2_io *io); |
| 682 | int kvmhv_nestedv2_flush_vcpu(struct kvm_vcpu *vcpu, u64 time_limit); |
| 683 | int kvmhv_nestedv2_set_ptbl_entry(unsigned long lpid, u64 dw0, u64 dw1); |
| 684 | int kvmhv_nestedv2_parse_output(struct kvm_vcpu *vcpu); |
| 685 | int kvmhv_nestedv2_set_vpa(struct kvm_vcpu *vcpu, unsigned long vpa); |
| 686 | |
| 687 | int kvmhv_counters_tracepoint_regfunc(void); |
| 688 | void kvmhv_counters_tracepoint_unregfunc(void); |
| 689 | int kvmhv_get_l2_counters_status(void); |
| 690 | void kvmhv_set_l2_counters_status(int cpu, bool status); |
| 691 | u64 kvmhv_get_l1_to_l2_cs_time(void); |
| 692 | u64 kvmhv_get_l2_to_l1_cs_time(void); |
| 693 | u64 kvmhv_get_l2_runtime_agg(void); |
| 694 | u64 kvmhv_get_l1_to_l2_cs_time_vcpu(void); |
| 695 | u64 kvmhv_get_l2_to_l1_cs_time_vcpu(void); |
| 696 | u64 kvmhv_get_l2_runtime_agg_vcpu(void); |
| 697 | |
| 698 | #endif /* CONFIG_KVM_BOOK3S_HV_POSSIBLE */ |
| 699 | |
| 700 | #endif /* __ASM_KVM_BOOK3S_64_H__ */ |
| 701 |
Warning: This file is not a C or C++ file. It does not have highlighting.
