1 /*- 2 * Copyright (c) 1991 Regents of the University of California. 3 * All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * the Systems Programming Group of the University of Utah Computer 7 * Science Department and William Jolitz of UUNET Technologies Inc. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * Derived from hp300 version by Mike Hibler, this version by William 34 * Jolitz uses a recursive map [a pde points to the page directory] to 35 * map the page tables using the pagetables themselves. This is done to 36 * reduce the impact on kernel virtual memory for lots of sparse address 37 * space, and to reduce the cost of memory to each process. 38 * 39 * from: hp300: @(#)pmap.h 7.2 (Berkeley) 12/16/90 40 * from: @(#)pmap.h 7.4 (Berkeley) 5/12/91 41 * $FreeBSD$ 42 */ 43 44 #ifndef _MACHINE_PMAP_H_ 45 #define _MACHINE_PMAP_H_ 46 47 /* 48 * Page-directory and page-table entries follow this format, with a few 49 * of the fields not present here and there, depending on a lot of things. 50 */ 51 /* ---- Intel Nomenclature ---- */ 52 #define PG_V 0x001 /* P Valid */ 53 #define PG_RW 0x002 /* R/W Read/Write */ 54 #define PG_U 0x004 /* U/S User/Supervisor */ 55 #define PG_NC_PWT 0x008 /* PWT Write through */ 56 #define PG_NC_PCD 0x010 /* PCD Cache disable */ 57 #define PG_A 0x020 /* A Accessed */ 58 #define PG_M 0x040 /* D Dirty */ 59 #define PG_PS 0x080 /* PS Page size (0=4k,1=4M) */ 60 #define PG_PTE_PAT 0x080 /* PAT PAT index */ 61 #define PG_G 0x100 /* G Global */ 62 #define PG_AVAIL1 0x200 /* / Available for system */ 63 #define PG_AVAIL2 0x400 /* < programmers use */ 64 #define PG_AVAIL3 0x800 /* \ */ 65 #define PG_PDE_PAT 0x1000 /* PAT PAT index */ 66 #ifdef PAE 67 #define PG_NX (1ull<<63) /* No-execute */ 68 #endif 69 70 71 /* Our various interpretations of the above */ 72 #define PG_W PG_AVAIL1 /* "Wired" pseudoflag */ 73 #define PG_MANAGED PG_AVAIL2 74 #ifdef PAE 75 #define PG_FRAME (0x000ffffffffff000ull) 76 #define PG_PS_FRAME (0x000fffffffe00000ull) 77 #else 78 #define PG_FRAME (~PAGE_MASK) 79 #define PG_PS_FRAME (0xffc00000) 80 #endif 81 #define PG_PROT (PG_RW|PG_U) /* all protection bits . */ 82 #define PG_N (PG_NC_PWT|PG_NC_PCD) /* Non-cacheable */ 83 84 /* Page level cache control fields used to determine the PAT type */ 85 #define PG_PDE_CACHE (PG_PDE_PAT | PG_NC_PWT | PG_NC_PCD) 86 #define PG_PTE_CACHE (PG_PTE_PAT | PG_NC_PWT | PG_NC_PCD) 87 88 /* 89 * Promotion to a 2 or 4MB (PDE) page mapping requires that the corresponding 90 * 4KB (PTE) page mappings have identical settings for the following fields: 91 */ 92 #define PG_PTE_PROMOTE (PG_MANAGED | PG_W | PG_G | PG_PTE_PAT | \ 93 PG_M | PG_A | PG_NC_PCD | PG_NC_PWT | PG_U | PG_RW | PG_V) 94 95 /* 96 * Page Protection Exception bits 97 */ 98 99 #define PGEX_P 0x01 /* Protection violation vs. not present */ 100 #define PGEX_W 0x02 /* during a Write cycle */ 101 #define PGEX_U 0x04 /* access from User mode (UPL) */ 102 #define PGEX_RSV 0x08 /* reserved PTE field is non-zero */ 103 #define PGEX_I 0x10 /* during an instruction fetch */ 104 105 /* 106 * Size of Kernel address space. This is the number of page table pages 107 * (4MB each) to use for the kernel. 256 pages == 1 Gigabyte. 108 * This **MUST** be a multiple of 4 (eg: 252, 256, 260, etc). 109 * For PAE, the page table page unit size is 2MB. This means that 512 pages 110 * is 1 Gigabyte. Double everything. It must be a multiple of 8 for PAE. 111 */ 112 #ifndef KVA_PAGES 113 #ifdef PAE 114 #define KVA_PAGES 512 115 #else 116 #define KVA_PAGES 256 117 #endif 118 #endif 119 120 /* 121 * Pte related macros 122 */ 123 #define VADDR(pdi, pti) ((vm_offset_t)(((pdi)<<PDRSHIFT)|((pti)<<PAGE_SHIFT))) 124 125 /* Initial number of kernel page tables. */ 126 #ifndef NKPT 127 #ifdef PAE 128 /* 152 page tables needed to map 16G (76B "struct vm_page", 2M page tables). */ 129 #define NKPT 240 130 #else 131 /* 18 page tables needed to map 4G (72B "struct vm_page", 4M page tables). */ 132 #define NKPT 30 133 #endif 134 #endif 135 136 #ifndef NKPDE 137 #define NKPDE (KVA_PAGES) /* number of page tables/pde's */ 138 #endif 139 140 /* 141 * The *PTDI values control the layout of virtual memory 142 * 143 * XXX This works for now, but I am not real happy with it, I'll fix it 144 * right after I fix locore.s and the magic 28K hole 145 */ 146 #define KPTDI (NPDEPTD-NKPDE) /* start of kernel virtual pde's */ 147 #define PTDPTDI (KPTDI-NPGPTD) /* ptd entry that points to ptd! */ 148 149 /* 150 * XXX doesn't really belong here I guess... 151 */ 152 #define ISA_HOLE_START 0xa0000 153 #define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START) 154 155 #ifndef LOCORE 156 157 #include <sys/queue.h> 158 #include <sys/_cpuset.h> 159 #include <sys/_lock.h> 160 #include <sys/_mutex.h> 161 162 #ifdef PAE 163 164 typedef uint64_t pdpt_entry_t; 165 typedef uint64_t pd_entry_t; 166 typedef uint64_t pt_entry_t; 167 168 #define PTESHIFT (3) 169 #define PDESHIFT (3) 170 171 #else 172 173 typedef uint32_t pd_entry_t; 174 typedef uint32_t pt_entry_t; 175 176 #define PTESHIFT (2) 177 #define PDESHIFT (2) 178 179 #endif 180 181 /* 182 * Address of current address space page table maps and directories. 183 */ 184 #ifdef _KERNEL 185 extern pt_entry_t PTmap[]; 186 extern pd_entry_t PTD[]; 187 extern pd_entry_t PTDpde[]; 188 189 #ifdef PAE 190 extern pdpt_entry_t *IdlePDPT; 191 #endif 192 extern pd_entry_t *IdlePTD; /* physical address of "Idle" state directory */ 193 194 /* 195 * Translate a virtual address to the kernel virtual address of its page table 196 * entry (PTE). This can be used recursively. If the address of a PTE as 197 * previously returned by this macro is itself given as the argument, then the 198 * address of the page directory entry (PDE) that maps the PTE will be 199 * returned. 200 * 201 * This macro may be used before pmap_bootstrap() is called. 202 */ 203 #define vtopte(va) (PTmap + i386_btop(va)) 204 205 /* 206 * Translate a virtual address to its physical address. 207 * 208 * This macro may be used before pmap_bootstrap() is called. 209 */ 210 #define vtophys(va) pmap_kextract((vm_offset_t)(va)) 211 212 #if defined(XEN) 213 #include <sys/param.h> 214 #include <machine/xen/xen-os.h> 215 #include <machine/xen/xenvar.h> 216 #include <machine/xen/xenpmap.h> 217 218 extern pt_entry_t pg_nx; 219 220 #define PG_KERNEL (PG_V | PG_A | PG_RW | PG_M) 221 222 #define MACH_TO_VM_PAGE(ma) PHYS_TO_VM_PAGE(xpmap_mtop((ma))) 223 #define VM_PAGE_TO_MACH(m) xpmap_ptom(VM_PAGE_TO_PHYS((m))) 224 225 #define VTOM(va) xpmap_ptom(VTOP(va)) 226 227 static __inline vm_paddr_t 228 pmap_kextract_ma(vm_offset_t va) 229 { 230 vm_paddr_t ma; 231 if ((ma = PTD[va >> PDRSHIFT]) & PG_PS) { 232 ma = (ma & ~(NBPDR - 1)) | (va & (NBPDR - 1)); 233 } else { 234 ma = (*vtopte(va) & PG_FRAME) | (va & PAGE_MASK); 235 } 236 return ma; 237 } 238 239 static __inline vm_paddr_t 240 pmap_kextract(vm_offset_t va) 241 { 242 return xpmap_mtop(pmap_kextract_ma(va)); 243 } 244 #define vtomach(va) pmap_kextract_ma(((vm_offset_t) (va))) 245 246 vm_paddr_t pmap_extract_ma(struct pmap *pmap, vm_offset_t va); 247 248 void pmap_kenter_ma(vm_offset_t va, vm_paddr_t pa); 249 void pmap_map_readonly(struct pmap *pmap, vm_offset_t va, int len); 250 void pmap_map_readwrite(struct pmap *pmap, vm_offset_t va, int len); 251 252 static __inline pt_entry_t 253 pte_load_store(pt_entry_t *ptep, pt_entry_t v) 254 { 255 pt_entry_t r; 256 257 r = *ptep; 258 PT_SET_VA(ptep, v, TRUE); 259 return (r); 260 } 261 262 static __inline pt_entry_t 263 pte_load_store_ma(pt_entry_t *ptep, pt_entry_t v) 264 { 265 pt_entry_t r; 266 267 r = *ptep; 268 PT_SET_VA_MA(ptep, v, TRUE); 269 return (r); 270 } 271 272 #define pte_load_clear(ptep) pte_load_store((ptep), (pt_entry_t)0ULL) 273 274 #define pte_store(ptep, pte) pte_load_store((ptep), (pt_entry_t)pte) 275 #define pte_store_ma(ptep, pte) pte_load_store_ma((ptep), (pt_entry_t)pte) 276 #define pde_store_ma(ptep, pte) pte_load_store_ma((ptep), (pt_entry_t)pte) 277 278 #elif !defined(XEN) 279 280 /* 281 * KPTmap is a linear mapping of the kernel page table. It differs from the 282 * recursive mapping in two ways: (1) it only provides access to kernel page 283 * table pages, and not user page table pages, and (2) it provides access to 284 * a kernel page table page after the corresponding virtual addresses have 285 * been promoted to a 2/4MB page mapping. 286 * 287 * KPTmap is first initialized by locore to support just NPKT page table 288 * pages. Later, it is reinitialized by pmap_bootstrap() to allow for 289 * expansion of the kernel page table. 290 */ 291 extern pt_entry_t *KPTmap; 292 293 /* 294 * Extract from the kernel page table the physical address that is mapped by 295 * the given virtual address "va". 296 * 297 * This function may be used before pmap_bootstrap() is called. 298 */ 299 static __inline vm_paddr_t 300 pmap_kextract(vm_offset_t va) 301 { 302 vm_paddr_t pa; 303 304 if ((pa = PTD[va >> PDRSHIFT]) & PG_PS) { 305 pa = (pa & PG_PS_FRAME) | (va & PDRMASK); 306 } else { 307 /* 308 * Beware of a concurrent promotion that changes the PDE at 309 * this point! For example, vtopte() must not be used to 310 * access the PTE because it would use the new PDE. It is, 311 * however, safe to use the old PDE because the page table 312 * page is preserved by the promotion. 313 */ 314 pa = KPTmap[i386_btop(va)]; 315 pa = (pa & PG_FRAME) | (va & PAGE_MASK); 316 } 317 return (pa); 318 } 319 #endif 320 321 #if !defined(XEN) 322 #define PT_UPDATES_FLUSH() 323 #endif 324 325 #if defined(PAE) && !defined(XEN) 326 327 #define pde_cmpset(pdep, old, new) \ 328 atomic_cmpset_64((pdep), (old), (new)) 329 330 static __inline pt_entry_t 331 pte_load(pt_entry_t *ptep) 332 { 333 pt_entry_t r; 334 335 __asm __volatile( 336 "lock; cmpxchg8b %1" 337 : "=A" (r) 338 : "m" (*ptep), "a" (0), "d" (0), "b" (0), "c" (0)); 339 return (r); 340 } 341 342 static __inline pt_entry_t 343 pte_load_store(pt_entry_t *ptep, pt_entry_t v) 344 { 345 pt_entry_t r; 346 347 r = *ptep; 348 __asm __volatile( 349 "1:\n" 350 "\tlock; cmpxchg8b %1\n" 351 "\tjnz 1b" 352 : "+A" (r) 353 : "m" (*ptep), "b" ((uint32_t)v), "c" ((uint32_t)(v >> 32))); 354 return (r); 355 } 356 357 /* XXXRU move to atomic.h? */ 358 static __inline int 359 atomic_cmpset_64(volatile uint64_t *dst, uint64_t exp, uint64_t src) 360 { 361 int64_t res = exp; 362 363 __asm __volatile ( 364 " lock ; " 365 " cmpxchg8b %2 ; " 366 " setz %%al ; " 367 " movzbl %%al,%0 ; " 368 "# atomic_cmpset_64" 369 : "+A" (res), /* 0 (result) */ 370 "=m" (*dst) /* 1 */ 371 : "m" (*dst), /* 2 */ 372 "b" ((uint32_t)src), 373 "c" ((uint32_t)(src >> 32))); 374 375 return (res); 376 } 377 378 #define pte_load_clear(ptep) pte_load_store((ptep), (pt_entry_t)0ULL) 379 380 #define pte_store(ptep, pte) pte_load_store((ptep), (pt_entry_t)pte) 381 382 extern pt_entry_t pg_nx; 383 384 #elif !defined(PAE) && !defined (XEN) 385 386 #define pde_cmpset(pdep, old, new) \ 387 atomic_cmpset_int((pdep), (old), (new)) 388 389 static __inline pt_entry_t 390 pte_load(pt_entry_t *ptep) 391 { 392 pt_entry_t r; 393 394 r = *ptep; 395 return (r); 396 } 397 398 static __inline pt_entry_t 399 pte_load_store(pt_entry_t *ptep, pt_entry_t pte) 400 { 401 __asm volatile("xchgl %0, %1" : "+m" (*ptep), "+r" (pte)); 402 return (pte); 403 } 404 405 #define pte_load_clear(pte) atomic_readandclear_int(pte) 406 407 static __inline void 408 pte_store(pt_entry_t *ptep, pt_entry_t pte) 409 { 410 411 *ptep = pte; 412 } 413 414 #endif /* PAE */ 415 416 #define pte_clear(ptep) pte_store((ptep), (pt_entry_t)0ULL) 417 418 #define pde_store(pdep, pde) pte_store((pdep), (pde)) 419 420 #endif /* _KERNEL */ 421 422 /* 423 * Pmap stuff 424 */ 425 struct pv_entry; 426 struct pv_chunk; 427 428 struct md_page { 429 TAILQ_HEAD(,pv_entry) pv_list; 430 int pat_mode; 431 }; 432 433 struct pmap { 434 struct mtx pm_mtx; 435 pd_entry_t *pm_pdir; /* KVA of page directory */ 436 TAILQ_HEAD(,pv_chunk) pm_pvchunk; /* list of mappings in pmap */ 437 cpuset_t pm_active; /* active on cpus */ 438 struct pmap_statistics pm_stats; /* pmap statistics */ 439 LIST_ENTRY(pmap) pm_list; /* List of all pmaps */ 440 #ifdef PAE 441 pdpt_entry_t *pm_pdpt; /* KVA of page director pointer 442 table */ 443 #endif 444 vm_page_t pm_root; /* spare page table pages */ 445 }; 446 447 typedef struct pmap *pmap_t; 448 449 #ifdef _KERNEL 450 extern struct pmap kernel_pmap_store; 451 #define kernel_pmap (&kernel_pmap_store) 452 453 #define PMAP_LOCK(pmap) mtx_lock(&(pmap)->pm_mtx) 454 #define PMAP_LOCK_ASSERT(pmap, type) \ 455 mtx_assert(&(pmap)->pm_mtx, (type)) 456 #define PMAP_LOCK_DESTROY(pmap) mtx_destroy(&(pmap)->pm_mtx) 457 #define PMAP_LOCK_INIT(pmap) mtx_init(&(pmap)->pm_mtx, "pmap", \ 458 NULL, MTX_DEF | MTX_DUPOK) 459 #define PMAP_LOCKED(pmap) mtx_owned(&(pmap)->pm_mtx) 460 #define PMAP_MTX(pmap) (&(pmap)->pm_mtx) 461 #define PMAP_TRYLOCK(pmap) mtx_trylock(&(pmap)->pm_mtx) 462 #define PMAP_UNLOCK(pmap) mtx_unlock(&(pmap)->pm_mtx) 463 #endif 464 465 /* 466 * For each vm_page_t, there is a list of all currently valid virtual 467 * mappings of that page. An entry is a pv_entry_t, the list is pv_list. 468 */ 469 typedef struct pv_entry { 470 vm_offset_t pv_va; /* virtual address for mapping */ 471 TAILQ_ENTRY(pv_entry) pv_list; 472 } *pv_entry_t; 473 474 /* 475 * pv_entries are allocated in chunks per-process. This avoids the 476 * need to track per-pmap assignments. 477 */ 478 #define _NPCM 11 479 #define _NPCPV 336 480 struct pv_chunk { 481 pmap_t pc_pmap; 482 TAILQ_ENTRY(pv_chunk) pc_list; 483 uint32_t pc_map[_NPCM]; /* bitmap; 1 = free */ 484 uint32_t pc_spare[2]; 485 struct pv_entry pc_pventry[_NPCPV]; 486 }; 487 488 #ifdef _KERNEL 489 490 extern caddr_t CADDR1; 491 extern pt_entry_t *CMAP1; 492 extern vm_paddr_t phys_avail[]; 493 extern vm_paddr_t dump_avail[]; 494 extern int pseflag; 495 extern int pgeflag; 496 extern char *ptvmmap; /* poor name! */ 497 extern vm_offset_t virtual_avail; 498 extern vm_offset_t virtual_end; 499 500 #define pmap_page_get_memattr(m) ((vm_memattr_t)(m)->md.pat_mode) 501 #define pmap_unmapbios(va, sz) pmap_unmapdev((va), (sz)) 502 503 /* 504 * Only the following functions or macros may be used before pmap_bootstrap() 505 * is called: pmap_kenter(), pmap_kextract(), pmap_kremove(), vtophys(), and 506 * vtopte(). 507 */ 508 void pmap_bootstrap(vm_paddr_t); 509 int pmap_cache_bits(int mode, boolean_t is_pde); 510 int pmap_change_attr(vm_offset_t, vm_size_t, int); 511 void pmap_init_pat(void); 512 void pmap_kenter(vm_offset_t va, vm_paddr_t pa); 513 void *pmap_kenter_temporary(vm_paddr_t pa, int i); 514 void pmap_kremove(vm_offset_t); 515 void *pmap_mapbios(vm_paddr_t, vm_size_t); 516 void *pmap_mapdev(vm_paddr_t, vm_size_t); 517 void *pmap_mapdev_attr(vm_paddr_t, vm_size_t, int); 518 boolean_t pmap_page_is_mapped(vm_page_t m); 519 void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma); 520 void pmap_unmapdev(vm_offset_t, vm_size_t); 521 pt_entry_t *pmap_pte(pmap_t, vm_offset_t) __pure2; 522 void pmap_invalidate_page(pmap_t, vm_offset_t); 523 void pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t); 524 void pmap_invalidate_all(pmap_t); 525 void pmap_invalidate_cache(void); 526 void pmap_invalidate_cache_pages(vm_page_t *pages, int count); 527 void pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva); 528 529 #endif /* _KERNEL */ 530 531 #endif /* !LOCORE */ 532 533 #endif /* !_MACHINE_PMAP_H_ */ 534