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 #if defined(PAE) || defined(PAE_TABLES) 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 #if defined(PAE) || defined(PAE_TABLES) 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 #if defined(PAE) || defined(PAE_TABLES) 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 /* 126 * The initial number of kernel page table pages that are constructed 127 * by locore must be sufficient to map vm_page_array. That number can 128 * be calculated as follows: 129 * max_phys / PAGE_SIZE * sizeof(struct vm_page) / NBPDR 130 * PAE: max_phys 16G, sizeof(vm_page) 76, NBPDR 2M, 152 page table pages. 131 * PAE_TABLES: max_phys 4G, sizeof(vm_page) 68, NBPDR 2M, 36 page table pages. 132 * Non-PAE: max_phys 4G, sizeof(vm_page) 68, NBPDR 4M, 18 page table pages. 133 */ 134 #ifndef NKPT 135 #if defined(PAE) 136 #define NKPT 240 137 #elif defined(PAE_TABLES) 138 #define NKPT 60 139 #else 140 #define NKPT 30 141 #endif 142 #endif 143 144 #ifndef NKPDE 145 #define NKPDE (KVA_PAGES) /* number of page tables/pde's */ 146 #endif 147 148 /* 149 * The *PTDI values control the layout of virtual memory 150 * 151 * XXX This works for now, but I am not real happy with it, I'll fix it 152 * right after I fix locore.s and the magic 28K hole 153 */ 154 #define KPTDI (NPDEPTD-NKPDE) /* start of kernel virtual pde's */ 155 #define PTDPTDI (KPTDI-NPGPTD) /* ptd entry that points to ptd! */ 156 157 /* 158 * XXX doesn't really belong here I guess... 159 */ 160 #define ISA_HOLE_START 0xa0000 161 #define ISA_HOLE_LENGTH (0x100000-ISA_HOLE_START) 162 163 #ifndef LOCORE 164 165 #include <sys/queue.h> 166 #include <sys/_cpuset.h> 167 #include <sys/_lock.h> 168 #include <sys/_mutex.h> 169 170 #include <vm/_vm_radix.h> 171 172 #if defined(PAE) || defined(PAE_TABLES) 173 174 typedef uint64_t pdpt_entry_t; 175 typedef uint64_t pd_entry_t; 176 typedef uint64_t pt_entry_t; 177 178 #define PTESHIFT (3) 179 #define PDESHIFT (3) 180 181 #else 182 183 typedef uint32_t pd_entry_t; 184 typedef uint32_t pt_entry_t; 185 186 #define PTESHIFT (2) 187 #define PDESHIFT (2) 188 189 #endif 190 191 /* 192 * Address of current address space page table maps and directories. 193 */ 194 #ifdef _KERNEL 195 extern pt_entry_t PTmap[]; 196 extern pd_entry_t PTD[]; 197 extern pd_entry_t PTDpde[]; 198 199 #if defined(PAE) || defined(PAE_TABLES) 200 extern pdpt_entry_t *IdlePDPT; 201 #endif 202 extern pd_entry_t *IdlePTD; /* physical address of "Idle" state directory */ 203 204 /* 205 * Translate a virtual address to the kernel virtual address of its page table 206 * entry (PTE). This can be used recursively. If the address of a PTE as 207 * previously returned by this macro is itself given as the argument, then the 208 * address of the page directory entry (PDE) that maps the PTE will be 209 * returned. 210 * 211 * This macro may be used before pmap_bootstrap() is called. 212 */ 213 #define vtopte(va) (PTmap + i386_btop(va)) 214 215 /* 216 * Translate a virtual address to its physical address. 217 * 218 * This macro may be used before pmap_bootstrap() is called. 219 */ 220 #define vtophys(va) pmap_kextract((vm_offset_t)(va)) 221 222 /* 223 * KPTmap is a linear mapping of the kernel page table. It differs from the 224 * recursive mapping in two ways: (1) it only provides access to kernel page 225 * table pages, and not user page table pages, and (2) it provides access to 226 * a kernel page table page after the corresponding virtual addresses have 227 * been promoted to a 2/4MB page mapping. 228 * 229 * KPTmap is first initialized by locore to support just NPKT page table 230 * pages. Later, it is reinitialized by pmap_bootstrap() to allow for 231 * expansion of the kernel page table. 232 */ 233 extern pt_entry_t *KPTmap; 234 235 /* 236 * Extract from the kernel page table the physical address that is mapped by 237 * the given virtual address "va". 238 * 239 * This function may be used before pmap_bootstrap() is called. 240 */ 241 static __inline vm_paddr_t 242 pmap_kextract(vm_offset_t va) 243 { 244 vm_paddr_t pa; 245 246 if ((pa = PTD[va >> PDRSHIFT]) & PG_PS) { 247 pa = (pa & PG_PS_FRAME) | (va & PDRMASK); 248 } else { 249 /* 250 * Beware of a concurrent promotion that changes the PDE at 251 * this point! For example, vtopte() must not be used to 252 * access the PTE because it would use the new PDE. It is, 253 * however, safe to use the old PDE because the page table 254 * page is preserved by the promotion. 255 */ 256 pa = KPTmap[i386_btop(va)]; 257 pa = (pa & PG_FRAME) | (va & PAGE_MASK); 258 } 259 return (pa); 260 } 261 262 #if (defined(PAE) || defined(PAE_TABLES)) 263 264 #define pde_cmpset(pdep, old, new) atomic_cmpset_64_i586(pdep, old, new) 265 #define pte_load_store(ptep, pte) atomic_swap_64_i586(ptep, pte) 266 #define pte_load_clear(ptep) atomic_swap_64_i586(ptep, 0) 267 #define pte_store(ptep, pte) atomic_store_rel_64_i586(ptep, pte) 268 269 extern pt_entry_t pg_nx; 270 271 #else /* !(PAE || PAE_TABLES) */ 272 273 #define pde_cmpset(pdep, old, new) atomic_cmpset_int(pdep, old, new) 274 #define pte_load_store(ptep, pte) atomic_swap_int(ptep, pte) 275 #define pte_load_clear(ptep) atomic_swap_int(ptep, 0) 276 #define pte_store(ptep, pte) do { \ 277 *(u_int *)(ptep) = (u_int)(pte); \ 278 } while (0) 279 280 #endif /* !(PAE || PAE_TABLES) */ 281 282 #define pte_clear(ptep) pte_store(ptep, 0) 283 284 #define pde_store(pdep, pde) pte_store(pdep, pde) 285 286 #endif /* _KERNEL */ 287 288 /* 289 * Pmap stuff 290 */ 291 struct pv_entry; 292 struct pv_chunk; 293 294 struct md_page { 295 TAILQ_HEAD(,pv_entry) pv_list; 296 int pat_mode; 297 }; 298 299 struct pmap { 300 struct mtx pm_mtx; 301 pd_entry_t *pm_pdir; /* KVA of page directory */ 302 TAILQ_HEAD(,pv_chunk) pm_pvchunk; /* list of mappings in pmap */ 303 cpuset_t pm_active; /* active on cpus */ 304 struct pmap_statistics pm_stats; /* pmap statistics */ 305 LIST_ENTRY(pmap) pm_list; /* List of all pmaps */ 306 #if defined(PAE) || defined(PAE_TABLES) 307 pdpt_entry_t *pm_pdpt; /* KVA of page directory pointer 308 table */ 309 #endif 310 struct vm_radix pm_root; /* spare page table pages */ 311 }; 312 313 typedef struct pmap *pmap_t; 314 315 #ifdef _KERNEL 316 extern struct pmap kernel_pmap_store; 317 #define kernel_pmap (&kernel_pmap_store) 318 319 #define PMAP_LOCK(pmap) mtx_lock(&(pmap)->pm_mtx) 320 #define PMAP_LOCK_ASSERT(pmap, type) \ 321 mtx_assert(&(pmap)->pm_mtx, (type)) 322 #define PMAP_LOCK_DESTROY(pmap) mtx_destroy(&(pmap)->pm_mtx) 323 #define PMAP_LOCK_INIT(pmap) mtx_init(&(pmap)->pm_mtx, "pmap", \ 324 NULL, MTX_DEF | MTX_DUPOK) 325 #define PMAP_LOCKED(pmap) mtx_owned(&(pmap)->pm_mtx) 326 #define PMAP_MTX(pmap) (&(pmap)->pm_mtx) 327 #define PMAP_TRYLOCK(pmap) mtx_trylock(&(pmap)->pm_mtx) 328 #define PMAP_UNLOCK(pmap) mtx_unlock(&(pmap)->pm_mtx) 329 #endif 330 331 /* 332 * For each vm_page_t, there is a list of all currently valid virtual 333 * mappings of that page. An entry is a pv_entry_t, the list is pv_list. 334 */ 335 typedef struct pv_entry { 336 vm_offset_t pv_va; /* virtual address for mapping */ 337 TAILQ_ENTRY(pv_entry) pv_next; 338 } *pv_entry_t; 339 340 /* 341 * pv_entries are allocated in chunks per-process. This avoids the 342 * need to track per-pmap assignments. 343 */ 344 #define _NPCM 11 345 #define _NPCPV 336 346 struct pv_chunk { 347 pmap_t pc_pmap; 348 TAILQ_ENTRY(pv_chunk) pc_list; 349 uint32_t pc_map[_NPCM]; /* bitmap; 1 = free */ 350 TAILQ_ENTRY(pv_chunk) pc_lru; 351 struct pv_entry pc_pventry[_NPCPV]; 352 }; 353 354 #ifdef _KERNEL 355 356 extern caddr_t CADDR3; 357 extern pt_entry_t *CMAP3; 358 extern vm_paddr_t phys_avail[]; 359 extern vm_paddr_t dump_avail[]; 360 extern int pseflag; 361 extern int pgeflag; 362 extern char *ptvmmap; /* poor name! */ 363 extern vm_offset_t virtual_avail; 364 extern vm_offset_t virtual_end; 365 366 #define pmap_page_get_memattr(m) ((vm_memattr_t)(m)->md.pat_mode) 367 #define pmap_page_is_write_mapped(m) (((m)->aflags & PGA_WRITEABLE) != 0) 368 #define pmap_unmapbios(va, sz) pmap_unmapdev((va), (sz)) 369 370 /* 371 * Only the following functions or macros may be used before pmap_bootstrap() 372 * is called: pmap_kenter(), pmap_kextract(), pmap_kremove(), vtophys(), and 373 * vtopte(). 374 */ 375 void pmap_bootstrap(vm_paddr_t); 376 int pmap_cache_bits(int mode, boolean_t is_pde); 377 int pmap_change_attr(vm_offset_t, vm_size_t, int); 378 void pmap_init_pat(void); 379 void pmap_kenter(vm_offset_t va, vm_paddr_t pa); 380 void *pmap_kenter_temporary(vm_paddr_t pa, int i); 381 void pmap_kremove(vm_offset_t); 382 void *pmap_mapbios(vm_paddr_t, vm_size_t); 383 void *pmap_mapdev(vm_paddr_t, vm_size_t); 384 void *pmap_mapdev_attr(vm_paddr_t, vm_size_t, int); 385 boolean_t pmap_page_is_mapped(vm_page_t m); 386 void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma); 387 void pmap_unmapdev(vm_offset_t, vm_size_t); 388 pt_entry_t *pmap_pte(pmap_t, vm_offset_t) __pure2; 389 void pmap_invalidate_page(pmap_t, vm_offset_t); 390 void pmap_invalidate_range(pmap_t, vm_offset_t, vm_offset_t); 391 void pmap_invalidate_all(pmap_t); 392 void pmap_invalidate_cache(void); 393 void pmap_invalidate_cache_pages(vm_page_t *pages, int count); 394 void pmap_invalidate_cache_range(vm_offset_t sva, vm_offset_t eva, 395 boolean_t force); 396 397 void invltlb_glob(void); 398 399 #endif /* _KERNEL */ 400 401 #endif /* !LOCORE */ 402 403 #endif /* !_MACHINE_PMAP_H_ */ 404