1 /* $NetBSD: pmap.h,v 1.52 2012/04/20 22:23:24 rmind Exp $ */ 2 3 /* 4 * Copyright (c) 1997 Charles D. Cranor and Washington University. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 /* 29 * Copyright (c) 2001 Wasabi Systems, Inc. 30 * All rights reserved. 31 * 32 * Written by Frank van der Linden for Wasabi Systems, Inc. 33 * 34 * Redistribution and use in source and binary forms, with or without 35 * modification, are permitted provided that the following conditions 36 * are met: 37 * 1. Redistributions of source code must retain the above copyright 38 * notice, this list of conditions and the following disclaimer. 39 * 2. Redistributions in binary form must reproduce the above copyright 40 * notice, this list of conditions and the following disclaimer in the 41 * documentation and/or other materials provided with the distribution. 42 * 3. All advertising materials mentioning features or use of this software 43 * must display the following acknowledgement: 44 * This product includes software developed for the NetBSD Project by 45 * Wasabi Systems, Inc. 46 * 4. The name of Wasabi Systems, Inc. may not be used to endorse 47 * or promote products derived from this software without specific prior 48 * written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 52 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 53 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC 54 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 55 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 56 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 57 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 58 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 59 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 60 * POSSIBILITY OF SUCH DAMAGE. 61 */ 62 63 /* 64 * pmap.h: see pmap.c for the history of this pmap module. 65 */ 66 67 #ifndef _X86_PMAP_H_ 68 #define _X86_PMAP_H_ 69 70 /* 71 * pl*_pi: index in the ptp page for a pde mapping a VA. 72 * (pl*_i below is the index in the virtual array of all pdes per level) 73 */ 74 #define pl1_pi(VA) (((VA_SIGN_POS(VA)) & L1_MASK) >> L1_SHIFT) 75 #define pl2_pi(VA) (((VA_SIGN_POS(VA)) & L2_MASK) >> L2_SHIFT) 76 #define pl3_pi(VA) (((VA_SIGN_POS(VA)) & L3_MASK) >> L3_SHIFT) 77 #define pl4_pi(VA) (((VA_SIGN_POS(VA)) & L4_MASK) >> L4_SHIFT) 78 79 /* 80 * pl*_i: generate index into pde/pte arrays in virtual space 81 * 82 * pl_i(va, X) == plX_i(va) <= pl_i_roundup(va, X) 83 */ 84 #define pl1_i(VA) (((VA_SIGN_POS(VA)) & L1_FRAME) >> L1_SHIFT) 85 #define pl2_i(VA) (((VA_SIGN_POS(VA)) & L2_FRAME) >> L2_SHIFT) 86 #define pl3_i(VA) (((VA_SIGN_POS(VA)) & L3_FRAME) >> L3_SHIFT) 87 #define pl4_i(VA) (((VA_SIGN_POS(VA)) & L4_FRAME) >> L4_SHIFT) 88 #define pl_i(va, lvl) \ 89 (((VA_SIGN_POS(va)) & ptp_masks[(lvl)-1]) >> ptp_shifts[(lvl)-1]) 90 91 #define pl_i_roundup(va, lvl) pl_i((va)+ ~ptp_masks[(lvl)-1], (lvl)) 92 93 /* 94 * PTP macros: 95 * a PTP's index is the PD index of the PDE that points to it 96 * a PTP's offset is the byte-offset in the PTE space that this PTP is at 97 * a PTP's VA is the first VA mapped by that PTP 98 */ 99 100 #define ptp_va2o(va, lvl) (pl_i(va, (lvl)+1) * PAGE_SIZE) 101 102 /* size of a PDP: usually one page, except for PAE */ 103 #ifdef PAE 104 #define PDP_SIZE 4 105 #else 106 #define PDP_SIZE 1 107 #endif 108 109 110 #if defined(_KERNEL) 111 #include <sys/kcpuset.h> 112 113 /* 114 * pmap data structures: see pmap.c for details of locking. 115 */ 116 117 /* 118 * we maintain a list of all non-kernel pmaps 119 */ 120 121 LIST_HEAD(pmap_head, pmap); /* struct pmap_head: head of a pmap list */ 122 123 /* 124 * linked list of all non-kernel pmaps 125 */ 126 extern struct pmap_head pmaps; 127 extern kmutex_t pmaps_lock; /* protects pmaps */ 128 129 /* 130 * pool_cache(9) that PDPs are allocated from 131 */ 132 extern struct pool_cache pmap_pdp_cache; 133 134 /* 135 * the pmap structure 136 * 137 * note that the pm_obj contains the lock pointer, the reference count, 138 * page list, and number of PTPs within the pmap. 139 * 140 * pm_lock is the same as the lock for vm object 0. Changes to 141 * the other objects may only be made if that lock has been taken 142 * (the other object locks are only used when uvm_pagealloc is called) 143 * 144 * XXX If we ever support processor numbers higher than 31, we'll have 145 * XXX to rethink the CPU mask. 146 */ 147 148 struct pmap { 149 struct uvm_object pm_obj[PTP_LEVELS-1]; /* objects for lvl >= 1) */ 150 #define pm_lock pm_obj[0].vmobjlock 151 kmutex_t pm_obj_lock[PTP_LEVELS-1]; /* locks for pm_objs */ 152 LIST_ENTRY(pmap) pm_list; /* list (lck by pm_list lock) */ 153 pd_entry_t *pm_pdir; /* VA of PD (lck by object lock) */ 154 paddr_t pm_pdirpa[PDP_SIZE]; /* PA of PDs (read-only after create) */ 155 struct vm_page *pm_ptphint[PTP_LEVELS-1]; 156 /* pointer to a PTP in our pmap */ 157 struct pmap_statistics pm_stats; /* pmap stats (lck by object lock) */ 158 159 #if !defined(__x86_64__) 160 vaddr_t pm_hiexec; /* highest executable mapping */ 161 #endif /* !defined(__x86_64__) */ 162 int pm_flags; /* see below */ 163 164 union descriptor *pm_ldt; /* user-set LDT */ 165 size_t pm_ldt_len; /* size of LDT in bytes */ 166 int pm_ldt_sel; /* LDT selector */ 167 kcpuset_t *pm_cpus; /* mask of CPUs using pmap */ 168 kcpuset_t *pm_kernel_cpus; /* mask of CPUs using kernel part 169 of pmap */ 170 kcpuset_t *pm_xen_ptp_cpus; /* mask of CPUs which have this pmap's 171 ptp mapped */ 172 uint64_t pm_ncsw; /* for assertions */ 173 struct vm_page *pm_gc_ptp; /* pages from pmap g/c */ 174 }; 175 176 /* macro to access pm_pdirpa slots */ 177 #ifdef PAE 178 #define pmap_pdirpa(pmap, index) \ 179 ((pmap)->pm_pdirpa[l2tol3(index)] + l2tol2(index) * sizeof(pd_entry_t)) 180 #else 181 #define pmap_pdirpa(pmap, index) \ 182 ((pmap)->pm_pdirpa[0] + (index) * sizeof(pd_entry_t)) 183 #endif 184 185 /* 186 * flag to be used for kernel mappings: PG_u on Xen/amd64, 187 * 0 otherwise. 188 */ 189 #if defined(XEN) && defined(__x86_64__) 190 #define PG_k PG_u 191 #else 192 #define PG_k 0 193 #endif 194 195 /* 196 * MD flags that we use for pmap_enter and pmap_kenter_pa: 197 */ 198 199 /* 200 * global kernel variables 201 */ 202 203 /* 204 * PDPpaddr is the physical address of the kernel's PDP. 205 * - i386 non-PAE and amd64: PDPpaddr corresponds directly to the %cr3 206 * value associated to the kernel process, proc0. 207 * - i386 PAE: it still represents the PA of the kernel's PDP (L2). Due to 208 * the L3 PD, it cannot be considered as the equivalent of a %cr3 any more. 209 * - Xen: it corresponds to the PFN of the kernel's PDP. 210 */ 211 extern u_long PDPpaddr; 212 213 extern int pmap_pg_g; /* do we support PG_G? */ 214 extern long nkptp[PTP_LEVELS]; 215 216 /* 217 * macros 218 */ 219 220 #define pmap_resident_count(pmap) ((pmap)->pm_stats.resident_count) 221 #define pmap_wired_count(pmap) ((pmap)->pm_stats.wired_count) 222 223 #define pmap_clear_modify(pg) pmap_clear_attrs(pg, PG_M) 224 #define pmap_clear_reference(pg) pmap_clear_attrs(pg, PG_U) 225 #define pmap_copy(DP,SP,D,L,S) 226 #define pmap_is_modified(pg) pmap_test_attrs(pg, PG_M) 227 #define pmap_is_referenced(pg) pmap_test_attrs(pg, PG_U) 228 #define pmap_move(DP,SP,D,L,S) 229 #define pmap_phys_address(ppn) (x86_ptob(ppn) & ~X86_MMAP_FLAG_MASK) 230 #define pmap_mmap_flags(ppn) x86_mmap_flags(ppn) 231 #define pmap_valid_entry(E) ((E) & PG_V) /* is PDE or PTE valid? */ 232 233 #if defined(__x86_64__) || defined(PAE) 234 #define X86_MMAP_FLAG_SHIFT (64 - PGSHIFT) 235 #else 236 #define X86_MMAP_FLAG_SHIFT (32 - PGSHIFT) 237 #endif 238 239 #define X86_MMAP_FLAG_MASK 0xf 240 #define X86_MMAP_FLAG_PREFETCH 0x1 241 242 /* 243 * prototypes 244 */ 245 246 void pmap_activate(struct lwp *); 247 void pmap_bootstrap(vaddr_t); 248 bool pmap_clear_attrs(struct vm_page *, unsigned); 249 void pmap_deactivate(struct lwp *); 250 void pmap_page_remove (struct vm_page *); 251 void pmap_remove(struct pmap *, vaddr_t, vaddr_t); 252 bool pmap_test_attrs(struct vm_page *, unsigned); 253 void pmap_write_protect(struct pmap *, vaddr_t, vaddr_t, vm_prot_t); 254 void pmap_load(void); 255 paddr_t pmap_init_tmp_pgtbl(paddr_t); 256 void pmap_remove_all(struct pmap *); 257 void pmap_ldt_sync(struct pmap *); 258 259 void pmap_emap_enter(vaddr_t, paddr_t, vm_prot_t); 260 void pmap_emap_remove(vaddr_t, vsize_t); 261 void pmap_emap_sync(bool); 262 263 void pmap_map_ptes(struct pmap *, struct pmap **, pd_entry_t **, 264 pd_entry_t * const **); 265 void pmap_unmap_ptes(struct pmap *, struct pmap *); 266 267 int pmap_pdes_invalid(vaddr_t, pd_entry_t * const *, pd_entry_t *); 268 269 u_int x86_mmap_flags(paddr_t); 270 271 bool pmap_is_curpmap(struct pmap *); 272 273 vaddr_t reserve_dumppages(vaddr_t); /* XXX: not a pmap fn */ 274 275 typedef enum tlbwhy { 276 TLBSHOOT_APTE, 277 TLBSHOOT_KENTER, 278 TLBSHOOT_KREMOVE, 279 TLBSHOOT_FREE_PTP1, 280 TLBSHOOT_FREE_PTP2, 281 TLBSHOOT_REMOVE_PTE, 282 TLBSHOOT_REMOVE_PTES, 283 TLBSHOOT_SYNC_PV1, 284 TLBSHOOT_SYNC_PV2, 285 TLBSHOOT_WRITE_PROTECT, 286 TLBSHOOT_ENTER, 287 TLBSHOOT_UPDATE, 288 TLBSHOOT_BUS_DMA, 289 TLBSHOOT_BUS_SPACE, 290 TLBSHOOT__MAX, 291 } tlbwhy_t; 292 293 void pmap_tlb_init(void); 294 void pmap_tlb_cpu_init(struct cpu_info *); 295 void pmap_tlb_shootdown(pmap_t, vaddr_t, pt_entry_t, tlbwhy_t); 296 void pmap_tlb_shootnow(void); 297 void pmap_tlb_intr(void); 298 299 #define __HAVE_PMAP_EMAP 300 301 #define PMAP_GROWKERNEL /* turn on pmap_growkernel interface */ 302 #define PMAP_FORK /* turn on pmap_fork interface */ 303 304 /* 305 * Do idle page zero'ing uncached to avoid polluting the cache. 306 */ 307 bool pmap_pageidlezero(paddr_t); 308 #define PMAP_PAGEIDLEZERO(pa) pmap_pageidlezero((pa)) 309 310 /* 311 * inline functions 312 */ 313 314 __inline static bool __unused 315 pmap_pdes_valid(vaddr_t va, pd_entry_t * const *pdes, pd_entry_t *lastpde) 316 { 317 return pmap_pdes_invalid(va, pdes, lastpde) == 0; 318 } 319 320 /* 321 * pmap_update_pg: flush one page from the TLB (or flush the whole thing 322 * if hardware doesn't support one-page flushing) 323 */ 324 325 __inline static void __unused 326 pmap_update_pg(vaddr_t va) 327 { 328 invlpg(va); 329 } 330 331 /* 332 * pmap_update_2pg: flush two pages from the TLB 333 */ 334 335 __inline static void __unused 336 pmap_update_2pg(vaddr_t va, vaddr_t vb) 337 { 338 invlpg(va); 339 invlpg(vb); 340 } 341 342 /* 343 * pmap_page_protect: change the protection of all recorded mappings 344 * of a managed page 345 * 346 * => this function is a frontend for pmap_page_remove/pmap_clear_attrs 347 * => we only have to worry about making the page more protected. 348 * unprotecting a page is done on-demand at fault time. 349 */ 350 351 __inline static void __unused 352 pmap_page_protect(struct vm_page *pg, vm_prot_t prot) 353 { 354 if ((prot & VM_PROT_WRITE) == 0) { 355 if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) { 356 (void) pmap_clear_attrs(pg, PG_RW); 357 } else { 358 pmap_page_remove(pg); 359 } 360 } 361 } 362 363 /* 364 * pmap_protect: change the protection of pages in a pmap 365 * 366 * => this function is a frontend for pmap_remove/pmap_write_protect 367 * => we only have to worry about making the page more protected. 368 * unprotecting a page is done on-demand at fault time. 369 */ 370 371 __inline static void __unused 372 pmap_protect(struct pmap *pmap, vaddr_t sva, vaddr_t eva, vm_prot_t prot) 373 { 374 if ((prot & VM_PROT_WRITE) == 0) { 375 if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) { 376 pmap_write_protect(pmap, sva, eva, prot); 377 } else { 378 pmap_remove(pmap, sva, eva); 379 } 380 } 381 } 382 383 /* 384 * various address inlines 385 * 386 * vtopte: return a pointer to the PTE mapping a VA, works only for 387 * user and PT addresses 388 * 389 * kvtopte: return a pointer to the PTE mapping a kernel VA 390 */ 391 392 #include <lib/libkern/libkern.h> 393 394 static __inline pt_entry_t * __unused 395 vtopte(vaddr_t va) 396 { 397 398 KASSERT(va < VM_MIN_KERNEL_ADDRESS); 399 400 return (PTE_BASE + pl1_i(va)); 401 } 402 403 static __inline pt_entry_t * __unused 404 kvtopte(vaddr_t va) 405 { 406 pd_entry_t *pde; 407 408 KASSERT(va >= VM_MIN_KERNEL_ADDRESS); 409 410 pde = L2_BASE + pl2_i(va); 411 if (*pde & PG_PS) 412 return ((pt_entry_t *)pde); 413 414 return (PTE_BASE + pl1_i(va)); 415 } 416 417 paddr_t vtophys(vaddr_t); 418 vaddr_t pmap_map(vaddr_t, paddr_t, paddr_t, vm_prot_t); 419 void pmap_cpu_init_late(struct cpu_info *); 420 bool sse2_idlezero_page(void *); 421 422 #ifdef XEN 423 #include <sys/bitops.h> 424 425 #define XPTE_MASK L1_FRAME 426 /* Selects the index of a PTE in (A)PTE_BASE */ 427 #define XPTE_SHIFT (L1_SHIFT - ilog2(sizeof(pt_entry_t))) 428 429 /* PTE access inline fuctions */ 430 431 /* 432 * Get the machine address of the pointed pte 433 * We use hardware MMU to get value so works only for levels 1-3 434 */ 435 436 static __inline paddr_t 437 xpmap_ptetomach(pt_entry_t *pte) 438 { 439 pt_entry_t *up_pte; 440 vaddr_t va = (vaddr_t) pte; 441 442 va = ((va & XPTE_MASK) >> XPTE_SHIFT) | (vaddr_t) PTE_BASE; 443 up_pte = (pt_entry_t *) va; 444 445 return (paddr_t) (((*up_pte) & PG_FRAME) + (((vaddr_t) pte) & (~PG_FRAME & ~VA_SIGN_MASK))); 446 } 447 448 /* Xen helpers to change bits of a pte */ 449 #define XPMAP_UPDATE_DIRECT 1 /* Update direct map entry flags too */ 450 451 paddr_t vtomach(vaddr_t); 452 #define vtomfn(va) (vtomach(va) >> PAGE_SHIFT) 453 #endif /* XEN */ 454 455 /* pmap functions with machine addresses */ 456 void pmap_kenter_ma(vaddr_t, paddr_t, vm_prot_t, u_int); 457 int pmap_enter_ma(struct pmap *, vaddr_t, paddr_t, paddr_t, 458 vm_prot_t, u_int, int); 459 bool pmap_extract_ma(pmap_t, vaddr_t, paddr_t *); 460 461 /* 462 * Hooks for the pool allocator. 463 */ 464 #define POOL_VTOPHYS(va) vtophys((vaddr_t) (va)) 465 466 #ifdef __HAVE_DIRECT_MAP 467 468 #define L4_SLOT_DIRECT 509 469 #define PDIR_SLOT_DIRECT L4_SLOT_DIRECT 470 471 #define PMAP_DIRECT_BASE (VA_SIGN_NEG((L4_SLOT_DIRECT * NBPD_L4))) 472 #define PMAP_DIRECT_END (VA_SIGN_NEG(((L4_SLOT_DIRECT + 1) * NBPD_L4))) 473 474 #define PMAP_DIRECT_MAP(pa) ((vaddr_t)PMAP_DIRECT_BASE + (pa)) 475 #define PMAP_DIRECT_UNMAP(va) ((paddr_t)(va) - PMAP_DIRECT_BASE) 476 477 /* 478 * Alternate mapping hooks for pool pages. 479 */ 480 #define PMAP_MAP_POOLPAGE(pa) PMAP_DIRECT_MAP((pa)) 481 #define PMAP_UNMAP_POOLPAGE(va) PMAP_DIRECT_UNMAP((va)) 482 483 void pagezero(vaddr_t); 484 485 #endif /* __HAVE_DIRECT_MAP */ 486 487 #endif /* _KERNEL */ 488 489 #endif /* _X86_PMAP_H_ */ 490