1 /* $NetBSD: pmap.h,v 1.34 2011/02/01 20:19:32 chuck 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 #define ptei(VA) (((VA_SIGN_POS(VA)) & L1_MASK) >> L1_SHIFT) 71 72 /* 73 * pl*_pi: index in the ptp page for a pde mapping a VA. 74 * (pl*_i below is the index in the virtual array of all pdes per level) 75 */ 76 #define pl1_pi(VA) (((VA_SIGN_POS(VA)) & L1_MASK) >> L1_SHIFT) 77 #define pl2_pi(VA) (((VA_SIGN_POS(VA)) & L2_MASK) >> L2_SHIFT) 78 #define pl3_pi(VA) (((VA_SIGN_POS(VA)) & L3_MASK) >> L3_SHIFT) 79 #define pl4_pi(VA) (((VA_SIGN_POS(VA)) & L4_MASK) >> L4_SHIFT) 80 81 /* 82 * pl*_i: generate index into pde/pte arrays in virtual space 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 /* 112 * pmap data structures: see pmap.c for details of locking. 113 */ 114 115 /* 116 * we maintain a list of all non-kernel pmaps 117 */ 118 119 LIST_HEAD(pmap_head, pmap); /* struct pmap_head: head of a pmap list */ 120 121 /* 122 * the pmap structure 123 * 124 * note that the pm_obj contains the simple_lock, the reference count, 125 * page list, and number of PTPs within the pmap. 126 * 127 * pm_lock is the same as the spinlock for vm object 0. Changes to 128 * the other objects may only be made if that lock has been taken 129 * (the other object locks are only used when uvm_pagealloc is called) 130 * 131 * XXX If we ever support processor numbers higher than 31, we'll have 132 * XXX to rethink the CPU mask. 133 */ 134 135 struct pmap { 136 struct uvm_object pm_obj[PTP_LEVELS-1]; /* objects for lvl >= 1) */ 137 #define pm_lock pm_obj[0].vmobjlock 138 LIST_ENTRY(pmap) pm_list; /* list (lck by pm_list lock) */ 139 pd_entry_t *pm_pdir; /* VA of PD (lck by object lock) */ 140 paddr_t pm_pdirpa[PDP_SIZE]; /* PA of PDs (read-only after create) */ 141 struct vm_page *pm_ptphint[PTP_LEVELS-1]; 142 /* pointer to a PTP in our pmap */ 143 struct pmap_statistics pm_stats; /* pmap stats (lck by object lock) */ 144 145 #if !defined(__x86_64__) 146 vaddr_t pm_hiexec; /* highest executable mapping */ 147 #endif /* !defined(__x86_64__) */ 148 int pm_flags; /* see below */ 149 150 union descriptor *pm_ldt; /* user-set LDT */ 151 size_t pm_ldt_len; /* size of LDT in bytes */ 152 int pm_ldt_sel; /* LDT selector */ 153 uint32_t pm_cpus; /* mask of CPUs using pmap */ 154 uint32_t pm_kernel_cpus; /* mask of CPUs using kernel part 155 of pmap */ 156 }; 157 158 /* macro to access pm_pdirpa slots */ 159 #ifdef PAE 160 #define pmap_pdirpa(pmap, index) \ 161 ((pmap)->pm_pdirpa[l2tol3(index)] + l2tol2(index) * sizeof(pd_entry_t)) 162 #else 163 #define pmap_pdirpa(pmap, index) \ 164 ((pmap)->pm_pdirpa[0] + (index) * sizeof(pd_entry_t)) 165 #endif 166 167 /* 168 * MD flags that we use for pmap_enter and pmap_kenter_pa: 169 */ 170 171 /* 172 * global kernel variables 173 */ 174 175 /* 176 * PDPpaddr is the physical address of the kernel's PDP. 177 * - i386 non-PAE and amd64: PDPpaddr corresponds directly to the %cr3 178 * value associated to the kernel process, proc0. 179 * - i386 PAE: it still represents the PA of the kernel's PDP (L2). Due to 180 * the L3 PD, it cannot be considered as the equivalent of a %cr3 any more. 181 * - Xen: it corresponds to the PFN of the kernel's PDP. 182 */ 183 extern u_long PDPpaddr; 184 185 extern int pmap_pg_g; /* do we support PG_G? */ 186 extern long nkptp[PTP_LEVELS]; 187 188 /* 189 * macros 190 */ 191 192 #define pmap_resident_count(pmap) ((pmap)->pm_stats.resident_count) 193 #define pmap_wired_count(pmap) ((pmap)->pm_stats.wired_count) 194 195 #define pmap_clear_modify(pg) pmap_clear_attrs(pg, PG_M) 196 #define pmap_clear_reference(pg) pmap_clear_attrs(pg, PG_U) 197 #define pmap_copy(DP,SP,D,L,S) 198 #define pmap_is_modified(pg) pmap_test_attrs(pg, PG_M) 199 #define pmap_is_referenced(pg) pmap_test_attrs(pg, PG_U) 200 #define pmap_move(DP,SP,D,L,S) 201 #define pmap_phys_address(ppn) x86_ptob(ppn) 202 #define pmap_valid_entry(E) ((E) & PG_V) /* is PDE or PTE valid? */ 203 204 205 /* 206 * prototypes 207 */ 208 209 void pmap_activate(struct lwp *); 210 void pmap_bootstrap(vaddr_t); 211 bool pmap_clear_attrs(struct vm_page *, unsigned); 212 void pmap_deactivate(struct lwp *); 213 void pmap_page_remove (struct vm_page *); 214 void pmap_remove(struct pmap *, vaddr_t, vaddr_t); 215 bool pmap_test_attrs(struct vm_page *, unsigned); 216 void pmap_write_protect(struct pmap *, vaddr_t, vaddr_t, vm_prot_t); 217 void pmap_load(void); 218 paddr_t pmap_init_tmp_pgtbl(paddr_t); 219 void pmap_remove_all(struct pmap *); 220 void pmap_ldt_sync(struct pmap *); 221 222 void pmap_emap_enter(vaddr_t, paddr_t, vm_prot_t); 223 void pmap_emap_remove(vaddr_t, vsize_t); 224 void pmap_emap_sync(bool); 225 226 void pmap_map_ptes(struct pmap *, struct pmap **, pd_entry_t **, 227 pd_entry_t * const **); 228 void pmap_unmap_ptes(struct pmap *, struct pmap *); 229 230 int pmap_pdes_invalid(vaddr_t, pd_entry_t * const *, pd_entry_t *); 231 232 vaddr_t reserve_dumppages(vaddr_t); /* XXX: not a pmap fn */ 233 234 void pmap_tlb_shootdown(pmap_t, vaddr_t, vaddr_t, pt_entry_t); 235 void pmap_tlb_shootwait(void); 236 237 #define __HAVE_PMAP_EMAP 238 239 #define PMAP_GROWKERNEL /* turn on pmap_growkernel interface */ 240 #define PMAP_FORK /* turn on pmap_fork interface */ 241 242 /* 243 * Do idle page zero'ing uncached to avoid polluting the cache. 244 */ 245 bool pmap_pageidlezero(paddr_t); 246 #define PMAP_PAGEIDLEZERO(pa) pmap_pageidlezero((pa)) 247 248 /* 249 * inline functions 250 */ 251 252 __inline static bool __unused 253 pmap_pdes_valid(vaddr_t va, pd_entry_t * const *pdes, pd_entry_t *lastpde) 254 { 255 return pmap_pdes_invalid(va, pdes, lastpde) == 0; 256 } 257 258 /* 259 * pmap_update_pg: flush one page from the TLB (or flush the whole thing 260 * if hardware doesn't support one-page flushing) 261 */ 262 263 __inline static void __unused 264 pmap_update_pg(vaddr_t va) 265 { 266 invlpg(va); 267 } 268 269 /* 270 * pmap_update_2pg: flush two pages from the TLB 271 */ 272 273 __inline static void __unused 274 pmap_update_2pg(vaddr_t va, vaddr_t vb) 275 { 276 invlpg(va); 277 invlpg(vb); 278 } 279 280 /* 281 * pmap_page_protect: change the protection of all recorded mappings 282 * of a managed page 283 * 284 * => this function is a frontend for pmap_page_remove/pmap_clear_attrs 285 * => we only have to worry about making the page more protected. 286 * unprotecting a page is done on-demand at fault time. 287 */ 288 289 __inline static void __unused 290 pmap_page_protect(struct vm_page *pg, vm_prot_t prot) 291 { 292 if ((prot & VM_PROT_WRITE) == 0) { 293 if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) { 294 (void) pmap_clear_attrs(pg, PG_RW); 295 } else { 296 pmap_page_remove(pg); 297 } 298 } 299 } 300 301 /* 302 * pmap_protect: change the protection of pages in a pmap 303 * 304 * => this function is a frontend for pmap_remove/pmap_write_protect 305 * => we only have to worry about making the page more protected. 306 * unprotecting a page is done on-demand at fault time. 307 */ 308 309 __inline static void __unused 310 pmap_protect(struct pmap *pmap, vaddr_t sva, vaddr_t eva, vm_prot_t prot) 311 { 312 if ((prot & VM_PROT_WRITE) == 0) { 313 if (prot & (VM_PROT_READ|VM_PROT_EXECUTE)) { 314 pmap_write_protect(pmap, sva, eva, prot); 315 } else { 316 pmap_remove(pmap, sva, eva); 317 } 318 } 319 } 320 321 /* 322 * various address inlines 323 * 324 * vtopte: return a pointer to the PTE mapping a VA, works only for 325 * user and PT addresses 326 * 327 * kvtopte: return a pointer to the PTE mapping a kernel VA 328 */ 329 330 #include <lib/libkern/libkern.h> 331 332 static __inline pt_entry_t * __unused 333 vtopte(vaddr_t va) 334 { 335 336 KASSERT(va < VM_MIN_KERNEL_ADDRESS); 337 338 return (PTE_BASE + pl1_i(va)); 339 } 340 341 static __inline pt_entry_t * __unused 342 kvtopte(vaddr_t va) 343 { 344 pd_entry_t *pde; 345 346 KASSERT(va >= VM_MIN_KERNEL_ADDRESS); 347 348 pde = L2_BASE + pl2_i(va); 349 if (*pde & PG_PS) 350 return ((pt_entry_t *)pde); 351 352 return (PTE_BASE + pl1_i(va)); 353 } 354 355 paddr_t vtophys(vaddr_t); 356 vaddr_t pmap_map(vaddr_t, paddr_t, paddr_t, vm_prot_t); 357 void pmap_cpu_init_early(struct cpu_info *); 358 void pmap_cpu_init_late(struct cpu_info *); 359 bool sse2_idlezero_page(void *); 360 361 362 #ifdef XEN 363 364 #define XPTE_MASK L1_FRAME 365 /* XPTE_SHIFT = L1_SHIFT - log2(sizeof(pt_entry_t)) */ 366 #if defined(__x86_64__) || defined(PAE) 367 #define XPTE_SHIFT 9 368 #else 369 #define XPTE_SHIFT 10 370 #endif 371 372 /* PTE access inline fuctions */ 373 374 /* 375 * Get the machine address of the pointed pte 376 * We use hardware MMU to get value so works only for levels 1-3 377 */ 378 379 static __inline paddr_t 380 xpmap_ptetomach(pt_entry_t *pte) 381 { 382 pt_entry_t *up_pte; 383 vaddr_t va = (vaddr_t) pte; 384 385 va = ((va & XPTE_MASK) >> XPTE_SHIFT) | (vaddr_t) PTE_BASE; 386 up_pte = (pt_entry_t *) va; 387 388 return (paddr_t) (((*up_pte) & PG_FRAME) + (((vaddr_t) pte) & (~PG_FRAME & ~VA_SIGN_MASK))); 389 } 390 391 /* 392 * xpmap_update() 393 * Update an active pt entry with Xen 394 * Equivalent to *pte = npte 395 */ 396 397 static __inline void 398 xpmap_update (pt_entry_t *pte, pt_entry_t npte) 399 { 400 int s = splvm(); 401 402 xpq_queue_pte_update(xpmap_ptetomach(pte), npte); 403 xpq_flush_queue(); 404 splx(s); 405 } 406 407 408 /* Xen helpers to change bits of a pte */ 409 #define XPMAP_UPDATE_DIRECT 1 /* Update direct map entry flags too */ 410 411 paddr_t vtomach(vaddr_t); 412 #define vtomfn(va) (vtomach(va) >> PAGE_SHIFT) 413 414 #endif /* XEN */ 415 416 /* pmap functions with machine addresses */ 417 void pmap_kenter_ma(vaddr_t, paddr_t, vm_prot_t, u_int); 418 int pmap_enter_ma(struct pmap *, vaddr_t, paddr_t, paddr_t, 419 vm_prot_t, u_int, int); 420 bool pmap_extract_ma(pmap_t, vaddr_t, paddr_t *); 421 422 /* 423 * Hooks for the pool allocator. 424 */ 425 #define POOL_VTOPHYS(va) vtophys((vaddr_t) (va)) 426 427 /* 428 * TLB shootdown mailbox. 429 */ 430 431 struct pmap_mbox { 432 volatile void *mb_pointer; 433 volatile uintptr_t mb_addr1; 434 volatile uintptr_t mb_addr2; 435 volatile uintptr_t mb_head; 436 volatile uintptr_t mb_tail; 437 volatile uintptr_t mb_global; 438 }; 439 440 #endif /* _KERNEL */ 441 442 #endif /* _X86_PMAP_H_ */ 443