1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2003 Peter Wemm. 5 * Copyright (c) 1991 Regents of the University of California. 6 * All rights reserved. 7 * 8 * This code is derived from software contributed to Berkeley by 9 * the Systems Programming Group of the University of Utah Computer 10 * Science Department and William Jolitz of UUNET Technologies Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * Derived from hp300 version by Mike Hibler, this version by William 37 * Jolitz uses a recursive map [a pde points to the page directory] to 38 * map the page tables using the pagetables themselves. This is done to 39 * reduce the impact on kernel virtual memory for lots of sparse address 40 * space, and to reduce the cost of memory to each process. 41 * 42 * from: hp300: @(#)pmap.h 7.2 (Berkeley) 12/16/90 43 * from: @(#)pmap.h 7.4 (Berkeley) 5/12/91 44 * $FreeBSD$ 45 */ 46 47 #ifdef __i386__ 48 #include <i386/pmap.h> 49 #else /* !__i386__ */ 50 51 #ifndef _MACHINE_PMAP_H_ 52 #define _MACHINE_PMAP_H_ 53 54 /* 55 * Page-directory and page-table entries follow this format, with a few 56 * of the fields not present here and there, depending on a lot of things. 57 */ 58 /* ---- Intel Nomenclature ---- */ 59 #define X86_PG_V 0x001 /* P Valid */ 60 #define X86_PG_RW 0x002 /* R/W Read/Write */ 61 #define X86_PG_U 0x004 /* U/S User/Supervisor */ 62 #define X86_PG_NC_PWT 0x008 /* PWT Write through */ 63 #define X86_PG_NC_PCD 0x010 /* PCD Cache disable */ 64 #define X86_PG_A 0x020 /* A Accessed */ 65 #define X86_PG_M 0x040 /* D Dirty */ 66 #define X86_PG_PS 0x080 /* PS Page size (0=4k,1=2M) */ 67 #define X86_PG_PTE_PAT 0x080 /* PAT PAT index */ 68 #define X86_PG_G 0x100 /* G Global */ 69 #define X86_PG_AVAIL1 0x200 /* / Available for system */ 70 #define X86_PG_AVAIL2 0x400 /* < programmers use */ 71 #define X86_PG_AVAIL3 0x800 /* \ */ 72 #define X86_PG_PDE_PAT 0x1000 /* PAT PAT index */ 73 #define X86_PG_PKU(idx) ((pt_entry_t)idx << 59) 74 #define X86_PG_NX (1ul<<63) /* No-execute */ 75 #define X86_PG_AVAIL(x) (1ul << (x)) 76 77 /* Page level cache control fields used to determine the PAT type */ 78 #define X86_PG_PDE_CACHE (X86_PG_PDE_PAT | X86_PG_NC_PWT | X86_PG_NC_PCD) 79 #define X86_PG_PTE_CACHE (X86_PG_PTE_PAT | X86_PG_NC_PWT | X86_PG_NC_PCD) 80 81 /* Protection keys indexes */ 82 #define PMAP_MAX_PKRU_IDX 0xf 83 #define X86_PG_PKU_MASK X86_PG_PKU(PMAP_MAX_PKRU_IDX) 84 85 /* 86 * Intel extended page table (EPT) bit definitions. 87 */ 88 #define EPT_PG_READ 0x001 /* R Read */ 89 #define EPT_PG_WRITE 0x002 /* W Write */ 90 #define EPT_PG_EXECUTE 0x004 /* X Execute */ 91 #define EPT_PG_IGNORE_PAT 0x040 /* IPAT Ignore PAT */ 92 #define EPT_PG_PS 0x080 /* PS Page size */ 93 #define EPT_PG_A 0x100 /* A Accessed */ 94 #define EPT_PG_M 0x200 /* D Dirty */ 95 #define EPT_PG_MEMORY_TYPE(x) ((x) << 3) /* MT Memory Type */ 96 97 /* 98 * Define the PG_xx macros in terms of the bits on x86 PTEs. 99 */ 100 #define PG_V X86_PG_V 101 #define PG_RW X86_PG_RW 102 #define PG_U X86_PG_U 103 #define PG_NC_PWT X86_PG_NC_PWT 104 #define PG_NC_PCD X86_PG_NC_PCD 105 #define PG_A X86_PG_A 106 #define PG_M X86_PG_M 107 #define PG_PS X86_PG_PS 108 #define PG_PTE_PAT X86_PG_PTE_PAT 109 #define PG_G X86_PG_G 110 #define PG_AVAIL1 X86_PG_AVAIL1 111 #define PG_AVAIL2 X86_PG_AVAIL2 112 #define PG_AVAIL3 X86_PG_AVAIL3 113 #define PG_PDE_PAT X86_PG_PDE_PAT 114 #define PG_NX X86_PG_NX 115 #define PG_PDE_CACHE X86_PG_PDE_CACHE 116 #define PG_PTE_CACHE X86_PG_PTE_CACHE 117 118 /* Our various interpretations of the above */ 119 #define PG_W X86_PG_AVAIL3 /* "Wired" pseudoflag */ 120 #define PG_MANAGED X86_PG_AVAIL2 121 #define EPT_PG_EMUL_V X86_PG_AVAIL(52) 122 #define EPT_PG_EMUL_RW X86_PG_AVAIL(53) 123 #define PG_PROMOTED X86_PG_AVAIL(54) /* PDE only */ 124 #define PG_FRAME (0x000ffffffffff000ul) 125 #define PG_PS_FRAME (0x000fffffffe00000ul) 126 #define PG_PS_PDP_FRAME (0x000fffffc0000000ul) 127 128 /* 129 * Promotion to a 2MB (PDE) page mapping requires that the corresponding 4KB 130 * (PTE) page mappings have identical settings for the following fields: 131 */ 132 #define PG_PTE_PROMOTE (PG_NX | PG_MANAGED | PG_W | PG_G | PG_PTE_CACHE | \ 133 PG_M | PG_A | PG_U | PG_RW | PG_V | PG_PKU_MASK) 134 135 /* 136 * Page Protection Exception bits 137 */ 138 139 #define PGEX_P 0x01 /* Protection violation vs. not present */ 140 #define PGEX_W 0x02 /* during a Write cycle */ 141 #define PGEX_U 0x04 /* access from User mode (UPL) */ 142 #define PGEX_RSV 0x08 /* reserved PTE field is non-zero */ 143 #define PGEX_I 0x10 /* during an instruction fetch */ 144 #define PGEX_PK 0x20 /* protection key violation */ 145 #define PGEX_SGX 0x8000 /* SGX-related */ 146 147 /* 148 * undef the PG_xx macros that define bits in the regular x86 PTEs that 149 * have a different position in nested PTEs. This is done when compiling 150 * code that needs to be aware of the differences between regular x86 and 151 * nested PTEs. 152 * 153 * The appropriate bitmask will be calculated at runtime based on the pmap 154 * type. 155 */ 156 #ifdef AMD64_NPT_AWARE 157 #undef PG_AVAIL1 /* X86_PG_AVAIL1 aliases with EPT_PG_M */ 158 #undef PG_G 159 #undef PG_A 160 #undef PG_M 161 #undef PG_PDE_PAT 162 #undef PG_PDE_CACHE 163 #undef PG_PTE_PAT 164 #undef PG_PTE_CACHE 165 #undef PG_RW 166 #undef PG_V 167 #endif 168 169 /* 170 * Pte related macros. This is complicated by having to deal with 171 * the sign extension of the 48th bit. 172 */ 173 #define KV4ADDR(l4, l3, l2, l1) ( \ 174 ((unsigned long)-1 << 47) | \ 175 ((unsigned long)(l4) << PML4SHIFT) | \ 176 ((unsigned long)(l3) << PDPSHIFT) | \ 177 ((unsigned long)(l2) << PDRSHIFT) | \ 178 ((unsigned long)(l1) << PAGE_SHIFT)) 179 #define KV5ADDR(l5, l4, l3, l2, l1) ( \ 180 ((unsigned long)-1 << 56) | \ 181 ((unsigned long)(l5) << PML5SHIFT) | \ 182 ((unsigned long)(l4) << PML4SHIFT) | \ 183 ((unsigned long)(l3) << PDPSHIFT) | \ 184 ((unsigned long)(l2) << PDRSHIFT) | \ 185 ((unsigned long)(l1) << PAGE_SHIFT)) 186 187 #define UVADDR(l5, l4, l3, l2, l1) ( \ 188 ((unsigned long)(l5) << PML5SHIFT) | \ 189 ((unsigned long)(l4) << PML4SHIFT) | \ 190 ((unsigned long)(l3) << PDPSHIFT) | \ 191 ((unsigned long)(l2) << PDRSHIFT) | \ 192 ((unsigned long)(l1) << PAGE_SHIFT)) 193 194 /* 195 * Number of kernel PML4 slots. Can be anywhere from 1 to 64 or so, 196 * but setting it larger than NDMPML4E makes no sense. 197 * 198 * Each slot provides .5 TB of kernel virtual space. 199 */ 200 #define NKPML4E 4 201 202 /* 203 * Number of PML4 slots for the KASAN shadow map. It requires 1 byte of memory 204 * for every 8 bytes of the kernel address space. 205 */ 206 #define NKASANPML4E ((NKPML4E + 7) / 8) 207 208 /* 209 * Number of PML4 slots for the KMSAN shadow and origin maps. These are 210 * one-to-one with the kernel map. 211 */ 212 #define NKMSANSHADPML4E NKPML4E 213 #define NKMSANORIGPML4E NKPML4E 214 215 /* 216 * We use the same numbering of the page table pages for 5-level and 217 * 4-level paging structures. 218 */ 219 #define NUPML5E (NPML5EPG / 2) /* number of userland PML5 220 pages */ 221 #define NUPML4E (NUPML5E * NPML4EPG) /* number of userland PML4 222 pages */ 223 #define NUPDPE (NUPML4E * NPDPEPG) /* number of userland PDP 224 pages */ 225 #define NUPDE (NUPDPE * NPDEPG) /* number of userland PD 226 entries */ 227 #define NUP4ML4E (NPML4EPG / 2) 228 229 /* 230 * NDMPML4E is the maximum number of PML4 entries that will be 231 * used to implement the direct map. It must be a power of two, 232 * and should generally exceed NKPML4E. The maximum possible 233 * value is 64; using 128 will make the direct map intrude into 234 * the recursive page table map. 235 */ 236 #define NDMPML4E 8 237 238 /* 239 * These values control the layout of virtual memory. The starting address 240 * of the direct map, which is controlled by DMPML4I, must be a multiple of 241 * its size. (See the PHYS_TO_DMAP() and DMAP_TO_PHYS() macros.) 242 * 243 * Note: KPML4I is the index of the (single) level 4 page that maps 244 * the KVA that holds KERNBASE, while KPML4BASE is the index of the 245 * first level 4 page that maps VM_MIN_KERNEL_ADDRESS. If NKPML4E 246 * is 1, these are the same, otherwise KPML4BASE < KPML4I and extra 247 * level 4 PDEs are needed to map from VM_MIN_KERNEL_ADDRESS up to 248 * KERNBASE. 249 * 250 * (KPML4I combines with KPDPI to choose where KERNBASE starts. 251 * Or, in other words, KPML4I provides bits 39..47 of KERNBASE, 252 * and KPDPI provides bits 30..38.) 253 */ 254 #define PML4PML4I (NPML4EPG / 2) /* Index of recursive pml4 mapping */ 255 #define PML5PML5I (NPML5EPG / 2) /* Index of recursive pml5 mapping */ 256 257 #define KPML4BASE (NPML4EPG-NKPML4E) /* KVM at highest addresses */ 258 #define DMPML4I rounddown(KPML4BASE-NDMPML4E, NDMPML4E) /* Below KVM */ 259 260 #define KPML4I (NPML4EPG-1) 261 #define KPDPI (NPDPEPG-2) /* kernbase at -2GB */ 262 263 #define KASANPML4I (DMPML4I - NKASANPML4E) /* Below the direct map */ 264 265 #define KMSANSHADPML4I (KPML4BASE - NKMSANSHADPML4E) 266 #define KMSANORIGPML4I (DMPML4I - NKMSANORIGPML4E) 267 268 /* Large map: index of the first and max last pml4 entry */ 269 #define LMSPML4I (PML4PML4I + 1) 270 #define LMEPML4I (KASANPML4I - 1) 271 272 /* 273 * XXX doesn't really belong here I guess... 274 */ 275 #define ISA_HOLE_START 0xa0000 276 #define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START) 277 278 #define PMAP_PCID_NONE 0xffffffff 279 #define PMAP_PCID_KERN 0 280 #define PMAP_PCID_OVERMAX 0x1000 281 #define PMAP_PCID_OVERMAX_KERN 0x800 282 #define PMAP_PCID_USER_PT 0x800 283 284 #define PMAP_NO_CR3 0xffffffffffffffff 285 #define PMAP_UCR3_NOMASK 0xffffffffffffffff 286 287 #ifndef LOCORE 288 289 #include <sys/queue.h> 290 #include <sys/_cpuset.h> 291 #include <sys/_lock.h> 292 #include <sys/_mutex.h> 293 #include <sys/_pctrie.h> 294 #include <sys/_pv_entry.h> 295 #include <sys/_rangeset.h> 296 #include <sys/_smr.h> 297 298 #include <vm/_vm_radix.h> 299 300 typedef u_int64_t pd_entry_t; 301 typedef u_int64_t pt_entry_t; 302 typedef u_int64_t pdp_entry_t; 303 typedef u_int64_t pml4_entry_t; 304 typedef u_int64_t pml5_entry_t; 305 306 /* 307 * Address of current address space page table maps and directories. 308 */ 309 #ifdef _KERNEL 310 #define addr_P4Tmap (KV4ADDR(PML4PML4I, 0, 0, 0)) 311 #define addr_P4Dmap (KV4ADDR(PML4PML4I, PML4PML4I, 0, 0)) 312 #define addr_P4DPmap (KV4ADDR(PML4PML4I, PML4PML4I, PML4PML4I, 0)) 313 #define addr_P4ML4map (KV4ADDR(PML4PML4I, PML4PML4I, PML4PML4I, PML4PML4I)) 314 #define addr_P4ML4pml4e (addr_PML4map + (PML4PML4I * sizeof(pml4_entry_t))) 315 #define P4Tmap ((pt_entry_t *)(addr_P4Tmap)) 316 #define P4Dmap ((pd_entry_t *)(addr_P4Dmap)) 317 318 #define addr_P5Tmap (KV5ADDR(PML5PML5I, 0, 0, 0, 0)) 319 #define addr_P5Dmap (KV5ADDR(PML5PML5I, PML5PML5I, 0, 0, 0)) 320 #define addr_P5DPmap (KV5ADDR(PML5PML5I, PML5PML5I, PML5PML5I, 0, 0)) 321 #define addr_P5ML4map (KV5ADDR(PML5PML5I, PML5PML5I, PML5PML5I, PML5PML5I, 0)) 322 #define addr_P5ML5map \ 323 (KVADDR(PML5PML5I, PML5PML5I, PML5PML5I, PML5PML5I, PML5PML5I)) 324 #define addr_P5ML5pml5e (addr_P5ML5map + (PML5PML5I * sizeof(pml5_entry_t))) 325 #define P5Tmap ((pt_entry_t *)(addr_P5Tmap)) 326 #define P5Dmap ((pd_entry_t *)(addr_P5Dmap)) 327 328 extern int nkpt; /* Initial number of kernel page tables */ 329 extern u_int64_t KPML4phys; /* physical address of kernel level 4 */ 330 extern u_int64_t KPML5phys; /* physical address of kernel level 5 */ 331 332 /* 333 * virtual address to page table entry and 334 * to physical address. 335 * Note: these work recursively, thus vtopte of a pte will give 336 * the corresponding pde that in turn maps it. 337 */ 338 pt_entry_t *vtopte(vm_offset_t); 339 #define vtophys(va) pmap_kextract(((vm_offset_t) (va))) 340 341 #define pte_load_store(ptep, pte) atomic_swap_long(ptep, pte) 342 #define pte_load_clear(ptep) atomic_swap_long(ptep, 0) 343 #define pte_store(ptep, pte) do { \ 344 *(u_long *)(ptep) = (u_long)(pte); \ 345 } while (0) 346 #define pte_clear(ptep) pte_store(ptep, 0) 347 348 #define pde_store(pdep, pde) pte_store(pdep, pde) 349 350 extern pt_entry_t pg_nx; 351 352 #endif /* _KERNEL */ 353 354 /* 355 * Pmap stuff 356 */ 357 358 /* 359 * Locks 360 * (p) PV list lock 361 */ 362 struct md_page { 363 TAILQ_HEAD(, pv_entry) pv_list; /* (p) */ 364 int pv_gen; /* (p) */ 365 int pat_mode; 366 }; 367 368 enum pmap_type { 369 PT_X86, /* regular x86 page tables */ 370 PT_EPT, /* Intel's nested page tables */ 371 PT_RVI, /* AMD's nested page tables */ 372 }; 373 374 struct pmap_pcids { 375 uint32_t pm_pcid; 376 uint32_t pm_gen; 377 }; 378 379 /* 380 * The kernel virtual address (KVA) of the level 4 page table page is always 381 * within the direct map (DMAP) region. 382 */ 383 struct pmap { 384 struct mtx pm_mtx; 385 pml4_entry_t *pm_pmltop; /* KVA of top level page table */ 386 pml4_entry_t *pm_pmltopu; /* KVA of user top page table */ 387 uint64_t pm_cr3; 388 uint64_t pm_ucr3; 389 TAILQ_HEAD(,pv_chunk) pm_pvchunk; /* list of mappings in pmap */ 390 cpuset_t pm_active; /* active on cpus */ 391 enum pmap_type pm_type; /* regular or nested tables */ 392 struct pmap_statistics pm_stats; /* pmap statistics */ 393 struct vm_radix pm_root; /* spare page table pages */ 394 long pm_eptgen; /* EPT pmap generation id */ 395 smr_t pm_eptsmr; 396 int pm_flags; 397 struct pmap_pcids pm_pcids[MAXCPU]; 398 struct rangeset pm_pkru; 399 }; 400 401 /* flags */ 402 #define PMAP_NESTED_IPIMASK 0xff 403 #define PMAP_PDE_SUPERPAGE (1 << 8) /* supports 2MB superpages */ 404 #define PMAP_EMULATE_AD_BITS (1 << 9) /* needs A/D bits emulation */ 405 #define PMAP_SUPPORTS_EXEC_ONLY (1 << 10) /* execute only mappings ok */ 406 407 typedef struct pmap *pmap_t; 408 409 #ifdef _KERNEL 410 extern struct pmap kernel_pmap_store; 411 #define kernel_pmap (&kernel_pmap_store) 412 413 #define PMAP_LOCK(pmap) mtx_lock(&(pmap)->pm_mtx) 414 #define PMAP_LOCK_ASSERT(pmap, type) \ 415 mtx_assert(&(pmap)->pm_mtx, (type)) 416 #define PMAP_LOCK_DESTROY(pmap) mtx_destroy(&(pmap)->pm_mtx) 417 #define PMAP_LOCK_INIT(pmap) mtx_init(&(pmap)->pm_mtx, "pmap", \ 418 NULL, MTX_DEF | MTX_DUPOK) 419 #define PMAP_LOCKED(pmap) mtx_owned(&(pmap)->pm_mtx) 420 #define PMAP_MTX(pmap) (&(pmap)->pm_mtx) 421 #define PMAP_TRYLOCK(pmap) mtx_trylock(&(pmap)->pm_mtx) 422 #define PMAP_UNLOCK(pmap) mtx_unlock(&(pmap)->pm_mtx) 423 424 int pmap_pinit_type(pmap_t pmap, enum pmap_type pm_type, int flags); 425 int pmap_emulate_accessed_dirty(pmap_t pmap, vm_offset_t va, int ftype); 426 427 extern caddr_t CADDR1; 428 extern pt_entry_t *CMAP1; 429 extern vm_offset_t virtual_avail; 430 extern vm_offset_t virtual_end; 431 extern vm_paddr_t dmaplimit; 432 extern int pmap_pcid_enabled; 433 extern int invpcid_works; 434 extern int pmap_pcid_invlpg_workaround; 435 extern int pmap_pcid_invlpg_workaround_uena; 436 437 #define pmap_page_get_memattr(m) ((vm_memattr_t)(m)->md.pat_mode) 438 #define pmap_page_is_write_mapped(m) (((m)->a.flags & PGA_WRITEABLE) != 0) 439 #define pmap_unmapbios(va, sz) pmap_unmapdev((va), (sz)) 440 441 #define pmap_vm_page_alloc_check(m) \ 442 KASSERT(m->phys_addr < kernphys || \ 443 m->phys_addr >= kernphys + (vm_offset_t)&_end - KERNSTART, \ 444 ("allocating kernel page %p pa %#lx kernphys %#lx end %p", \ 445 m, m->phys_addr, kernphys, &_end)); 446 447 struct thread; 448 449 void pmap_activate_boot(pmap_t pmap); 450 void pmap_activate_sw(struct thread *); 451 void pmap_allow_2m_x_ept_recalculate(void); 452 void pmap_bootstrap(vm_paddr_t *); 453 int pmap_cache_bits(pmap_t pmap, int mode, boolean_t is_pde); 454 int pmap_change_attr(vm_offset_t, vm_size_t, int); 455 int pmap_change_prot(vm_offset_t, vm_size_t, vm_prot_t); 456 void pmap_demote_DMAP(vm_paddr_t base, vm_size_t len, boolean_t invalidate); 457 void pmap_flush_cache_range(vm_offset_t, vm_offset_t); 458 void pmap_flush_cache_phys_range(vm_paddr_t, vm_paddr_t, vm_memattr_t); 459 void pmap_init_pat(void); 460 void pmap_kenter(vm_offset_t va, vm_paddr_t pa); 461 void *pmap_kenter_temporary(vm_paddr_t pa, int i); 462 vm_paddr_t pmap_kextract(vm_offset_t); 463 void pmap_kremove(vm_offset_t); 464 int pmap_large_map(vm_paddr_t, vm_size_t, void **, vm_memattr_t); 465 void pmap_large_map_wb(void *sva, vm_size_t len); 466 void pmap_large_unmap(void *sva, vm_size_t len); 467 void *pmap_mapbios(vm_paddr_t, vm_size_t); 468 void *pmap_mapdev(vm_paddr_t, vm_size_t); 469 void *pmap_mapdev_attr(vm_paddr_t, vm_size_t, int); 470 void *pmap_mapdev_pciecfg(vm_paddr_t pa, vm_size_t size); 471 bool pmap_not_in_di(void); 472 boolean_t pmap_page_is_mapped(vm_page_t m); 473 void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma); 474 void pmap_page_set_memattr_noflush(vm_page_t m, vm_memattr_t ma); 475 void pmap_pinit_pml4(vm_page_t); 476 void pmap_pinit_pml5(vm_page_t); 477 bool pmap_ps_enabled(pmap_t pmap); 478 void pmap_unmapdev(void *, vm_size_t); 479 void pmap_invalidate_page(pmap_t, vm_offset_t); 480 void pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t); 481 void pmap_invalidate_all(pmap_t); 482 void pmap_invalidate_cache(void); 483 void pmap_invalidate_cache_pages(vm_page_t *pages, int count); 484 void pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva); 485 void pmap_force_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva); 486 void pmap_get_mapping(pmap_t pmap, vm_offset_t va, uint64_t *ptr, int *num); 487 boolean_t pmap_map_io_transient(vm_page_t *, vm_offset_t *, int, boolean_t); 488 void pmap_unmap_io_transient(vm_page_t *, vm_offset_t *, int, boolean_t); 489 void pmap_map_delete(pmap_t, vm_offset_t, vm_offset_t); 490 void pmap_pti_add_kva(vm_offset_t sva, vm_offset_t eva, bool exec); 491 void pmap_pti_remove_kva(vm_offset_t sva, vm_offset_t eva); 492 void pmap_pti_pcid_invalidate(uint64_t ucr3, uint64_t kcr3); 493 void pmap_pti_pcid_invlpg(uint64_t ucr3, uint64_t kcr3, vm_offset_t va); 494 void pmap_pti_pcid_invlrng(uint64_t ucr3, uint64_t kcr3, vm_offset_t sva, 495 vm_offset_t eva); 496 int pmap_pkru_clear(pmap_t pmap, vm_offset_t sva, vm_offset_t eva); 497 int pmap_pkru_set(pmap_t pmap, vm_offset_t sva, vm_offset_t eva, 498 u_int keyidx, int flags); 499 void pmap_thread_init_invl_gen(struct thread *td); 500 int pmap_vmspace_copy(pmap_t dst_pmap, pmap_t src_pmap); 501 void pmap_page_array_startup(long count); 502 vm_page_t pmap_page_alloc_below_4g(bool zeroed); 503 504 #if defined(KASAN) || defined(KMSAN) 505 void pmap_san_bootstrap(void); 506 void pmap_san_enter(vm_offset_t); 507 #endif 508 509 /* 510 * Returns a pointer to a set of CPUs on which the pmap is currently active. 511 * Note that the set can be modified without any mutual exclusion, so a copy 512 * must be made if a stable value is required. 513 */ 514 static __inline volatile cpuset_t * 515 pmap_invalidate_cpu_mask(pmap_t pmap) 516 { 517 return (&pmap->pm_active); 518 } 519 520 #if defined(_SYS_PCPU_H_) && defined(_MACHINE_CPUFUNC_H_) 521 /* 522 * It seems that AlderLake+ small cores have some microarchitectural 523 * bug, which results in the INVLPG instruction failing to flush all 524 * global TLB entries when PCID is enabled. Work around it for now, 525 * by doing global invalidation on small cores instead of INVLPG. 526 */ 527 static __inline void 528 pmap_invlpg(pmap_t pmap, vm_offset_t va) 529 { 530 if (pmap == kernel_pmap && PCPU_GET(pcid_invlpg_workaround)) { 531 struct invpcid_descr d = { 0 }; 532 533 invpcid(&d, INVPCID_CTXGLOB); 534 } else { 535 invlpg(va); 536 } 537 } 538 #endif /* sys/pcpu.h && machine/cpufunc.h */ 539 540 #endif /* _KERNEL */ 541 542 /* Return various clipped indexes for a given VA */ 543 static __inline vm_pindex_t 544 pmap_pte_index(vm_offset_t va) 545 { 546 547 return ((va >> PAGE_SHIFT) & ((1ul << NPTEPGSHIFT) - 1)); 548 } 549 550 static __inline vm_pindex_t 551 pmap_pde_index(vm_offset_t va) 552 { 553 554 return ((va >> PDRSHIFT) & ((1ul << NPDEPGSHIFT) - 1)); 555 } 556 557 static __inline vm_pindex_t 558 pmap_pdpe_index(vm_offset_t va) 559 { 560 561 return ((va >> PDPSHIFT) & ((1ul << NPDPEPGSHIFT) - 1)); 562 } 563 564 static __inline vm_pindex_t 565 pmap_pml4e_index(vm_offset_t va) 566 { 567 568 return ((va >> PML4SHIFT) & ((1ul << NPML4EPGSHIFT) - 1)); 569 } 570 571 static __inline vm_pindex_t 572 pmap_pml5e_index(vm_offset_t va) 573 { 574 575 return ((va >> PML5SHIFT) & ((1ul << NPML5EPGSHIFT) - 1)); 576 } 577 578 #endif /* !LOCORE */ 579 580 #endif /* !_MACHINE_PMAP_H_ */ 581 582 #endif /* __i386__ */ 583