1#ifndef _ASM_X86_PGTABLE_H 2#define _ASM_X86_PGTABLE_H 3 4#include <asm/page.h> 5#include <asm/e820.h> 6 7#include <asm/pgtable_types.h> 8 9/* 10 * Macro to mark a page protection value as UC- 11 */ 12#define pgprot_noncached(prot) \ 13 ((boot_cpu_data.x86 > 3) \ 14 ? (__pgprot(pgprot_val(prot) | _PAGE_CACHE_UC_MINUS)) \ 15 : (prot)) 16 17#ifndef __ASSEMBLY__ 18 19#include <asm/x86_init.h> 20 21/* 22 * ZERO_PAGE is a global shared page that is always zero: used 23 * for zero-mapped memory areas etc.. 24 */ 25extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)]; 26#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page)) 27 28extern spinlock_t pgd_lock; 29extern struct list_head pgd_list; 30 31#ifdef CONFIG_PARAVIRT 32#include <asm/paravirt.h> 33#else /* !CONFIG_PARAVIRT */ 34#define set_pte(ptep, pte) native_set_pte(ptep, pte) 35#define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte) 36 37#define set_pte_atomic(ptep, pte) \ 38 native_set_pte_atomic(ptep, pte) 39 40#define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd) 41 42#ifndef __PAGETABLE_PUD_FOLDED 43#define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd) 44#define pgd_clear(pgd) native_pgd_clear(pgd) 45#endif 46 47#ifndef set_pud 48# define set_pud(pudp, pud) native_set_pud(pudp, pud) 49#endif 50 51#ifndef __PAGETABLE_PMD_FOLDED 52#define pud_clear(pud) native_pud_clear(pud) 53#endif 54 55#define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep) 56#define pmd_clear(pmd) native_pmd_clear(pmd) 57 58#define pte_update(mm, addr, ptep) do { } while (0) 59#define pte_update_defer(mm, addr, ptep) do { } while (0) 60 61#define pgd_val(x) native_pgd_val(x) 62#define __pgd(x) native_make_pgd(x) 63 64#ifndef __PAGETABLE_PUD_FOLDED 65#define pud_val(x) native_pud_val(x) 66#define __pud(x) native_make_pud(x) 67#endif 68 69#ifndef __PAGETABLE_PMD_FOLDED 70#define pmd_val(x) native_pmd_val(x) 71#define __pmd(x) native_make_pmd(x) 72#endif 73 74#define pte_val(x) native_pte_val(x) 75#define __pte(x) native_make_pte(x) 76 77#define arch_end_context_switch(prev) do {} while(0) 78 79#endif /* CONFIG_PARAVIRT */ 80 81/* 82 * The following only work if pte_present() is true. 83 * Undefined behaviour if not.. 84 */ 85static inline int pte_dirty(pte_t pte) 86{ 87 return pte_flags(pte) & _PAGE_DIRTY; 88} 89 90static inline int pte_young(pte_t pte) 91{ 92 return pte_flags(pte) & _PAGE_ACCESSED; 93} 94 95static inline int pte_write(pte_t pte) 96{ 97 return pte_flags(pte) & _PAGE_RW; 98} 99 100static inline int pte_file(pte_t pte) 101{ 102 return pte_flags(pte) & _PAGE_FILE; 103} 104 105static inline int pte_huge(pte_t pte) 106{ 107 return pte_flags(pte) & _PAGE_PSE; 108} 109 110static inline int pte_global(pte_t pte) 111{ 112 return pte_flags(pte) & _PAGE_GLOBAL; 113} 114 115static inline int pte_exec(pte_t pte) 116{ 117 return !(pte_flags(pte) & _PAGE_NX); 118} 119 120static inline int pte_special(pte_t pte) 121{ 122 return pte_flags(pte) & _PAGE_SPECIAL; 123} 124 125static inline unsigned long pte_pfn(pte_t pte) 126{ 127 return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT; 128} 129 130static inline unsigned long pmd_pfn(pmd_t pmd) 131{ 132 return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT; 133} 134 135#define pte_page(pte) pfn_to_page(pte_pfn(pte)) 136 137static inline int pmd_large(pmd_t pte) 138{ 139 return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) == 140 (_PAGE_PSE | _PAGE_PRESENT); 141} 142 143static inline pte_t pte_set_flags(pte_t pte, pteval_t set) 144{ 145 pteval_t v = native_pte_val(pte); 146 147 return native_make_pte(v | set); 148} 149 150static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear) 151{ 152 pteval_t v = native_pte_val(pte); 153 154 return native_make_pte(v & ~clear); 155} 156 157static inline pte_t pte_mkclean(pte_t pte) 158{ 159 return pte_clear_flags(pte, _PAGE_DIRTY); 160} 161 162static inline pte_t pte_mkold(pte_t pte) 163{ 164 return pte_clear_flags(pte, _PAGE_ACCESSED); 165} 166 167static inline pte_t pte_wrprotect(pte_t pte) 168{ 169 return pte_clear_flags(pte, _PAGE_RW); 170} 171 172static inline pte_t pte_mkexec(pte_t pte) 173{ 174 return pte_clear_flags(pte, _PAGE_NX); 175} 176 177static inline pte_t pte_mkdirty(pte_t pte) 178{ 179 return pte_set_flags(pte, _PAGE_DIRTY); 180} 181 182static inline pte_t pte_mkyoung(pte_t pte) 183{ 184 return pte_set_flags(pte, _PAGE_ACCESSED); 185} 186 187static inline pte_t pte_mkwrite(pte_t pte) 188{ 189 return pte_set_flags(pte, _PAGE_RW); 190} 191 192static inline pte_t pte_mkhuge(pte_t pte) 193{ 194 return pte_set_flags(pte, _PAGE_PSE); 195} 196 197static inline pte_t pte_clrhuge(pte_t pte) 198{ 199 return pte_clear_flags(pte, _PAGE_PSE); 200} 201 202static inline pte_t pte_mkglobal(pte_t pte) 203{ 204 return pte_set_flags(pte, _PAGE_GLOBAL); 205} 206 207static inline pte_t pte_clrglobal(pte_t pte) 208{ 209 return pte_clear_flags(pte, _PAGE_GLOBAL); 210} 211 212static inline pte_t pte_mkspecial(pte_t pte) 213{ 214 return pte_set_flags(pte, _PAGE_SPECIAL); 215} 216 217/* 218 * Mask out unsupported bits in a present pgprot. Non-present pgprots 219 * can use those bits for other purposes, so leave them be. 220 */ 221static inline pgprotval_t massage_pgprot(pgprot_t pgprot) 222{ 223 pgprotval_t protval = pgprot_val(pgprot); 224 225 if (protval & _PAGE_PRESENT) 226 protval &= __supported_pte_mask; 227 228 return protval; 229} 230 231static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot) 232{ 233 return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) | 234 massage_pgprot(pgprot)); 235} 236 237static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot) 238{ 239 return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) | 240 massage_pgprot(pgprot)); 241} 242 243static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) 244{ 245 pteval_t val = pte_val(pte); 246 247 /* 248 * Chop off the NX bit (if present), and add the NX portion of 249 * the newprot (if present): 250 */ 251 val &= _PAGE_CHG_MASK; 252 val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK; 253 254 return __pte(val); 255} 256 257/* mprotect needs to preserve PAT bits when updating vm_page_prot */ 258#define pgprot_modify pgprot_modify 259static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot) 260{ 261 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK; 262 pgprotval_t addbits = pgprot_val(newprot); 263 return __pgprot(preservebits | addbits); 264} 265 266#define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK) 267 268#define canon_pgprot(p) __pgprot(massage_pgprot(p)) 269 270static inline int is_new_memtype_allowed(u64 paddr, unsigned long size, 271 unsigned long flags, 272 unsigned long new_flags) 273{ 274 /* 275 * PAT type is always WB for untracked ranges, so no need to check. 276 */ 277 if (x86_platform.is_untracked_pat_range(paddr, paddr + size)) 278 return 1; 279 280 /* 281 * Certain new memtypes are not allowed with certain 282 * requested memtype: 283 * - request is uncached, return cannot be write-back 284 * - request is write-combine, return cannot be write-back 285 */ 286 if ((flags == _PAGE_CACHE_UC_MINUS && 287 new_flags == _PAGE_CACHE_WB) || 288 (flags == _PAGE_CACHE_WC && 289 new_flags == _PAGE_CACHE_WB)) { 290 return 0; 291 } 292 293 return 1; 294} 295 296pmd_t *populate_extra_pmd(unsigned long vaddr); 297pte_t *populate_extra_pte(unsigned long vaddr); 298#endif /* __ASSEMBLY__ */ 299 300#ifdef CONFIG_X86_32 301# include "pgtable_32.h" 302#else 303# include "pgtable_64.h" 304#endif 305 306#ifndef __ASSEMBLY__ 307#include <linux/mm_types.h> 308 309static inline int pte_none(pte_t pte) 310{ 311 return !pte.pte; 312} 313 314#define __HAVE_ARCH_PTE_SAME 315static inline int pte_same(pte_t a, pte_t b) 316{ 317 return a.pte == b.pte; 318} 319 320static inline int pte_present(pte_t a) 321{ 322 return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE); 323} 324 325static inline int pte_hidden(pte_t pte) 326{ 327 return pte_flags(pte) & _PAGE_HIDDEN; 328} 329 330static inline int pmd_present(pmd_t pmd) 331{ 332 return pmd_flags(pmd) & _PAGE_PRESENT; 333} 334 335static inline int pmd_none(pmd_t pmd) 336{ 337 /* Only check low word on 32-bit platforms, since it might be 338 out of sync with upper half. */ 339 return (unsigned long)native_pmd_val(pmd) == 0; 340} 341 342static inline unsigned long pmd_page_vaddr(pmd_t pmd) 343{ 344 return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK); 345} 346 347/* 348 * Currently stuck as a macro due to indirect forward reference to 349 * linux/mmzone.h's __section_mem_map_addr() definition: 350 */ 351#define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT) 352 353/* 354 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD] 355 * 356 * this macro returns the index of the entry in the pmd page which would 357 * control the given virtual address 358 */ 359static inline unsigned long pmd_index(unsigned long address) 360{ 361 return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1); 362} 363 364/* 365 * Conversion functions: convert a page and protection to a page entry, 366 * and a page entry and page directory to the page they refer to. 367 * 368 * (Currently stuck as a macro because of indirect forward reference 369 * to linux/mm.h:page_to_nid()) 370 */ 371#define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot)) 372 373/* 374 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE] 375 * 376 * this function returns the index of the entry in the pte page which would 377 * control the given virtual address 378 */ 379static inline unsigned long pte_index(unsigned long address) 380{ 381 return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1); 382} 383 384static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address) 385{ 386 return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address); 387} 388 389static inline int pmd_bad(pmd_t pmd) 390{ 391 return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE; 392} 393 394static inline unsigned long pages_to_mb(unsigned long npg) 395{ 396 return npg >> (20 - PAGE_SHIFT); 397} 398 399#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \ 400 remap_pfn_range(vma, vaddr, pfn, size, prot) 401 402#if PAGETABLE_LEVELS > 2 403static inline int pud_none(pud_t pud) 404{ 405 return native_pud_val(pud) == 0; 406} 407 408static inline int pud_present(pud_t pud) 409{ 410 return pud_flags(pud) & _PAGE_PRESENT; 411} 412 413static inline unsigned long pud_page_vaddr(pud_t pud) 414{ 415 return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK); 416} 417 418/* 419 * Currently stuck as a macro due to indirect forward reference to 420 * linux/mmzone.h's __section_mem_map_addr() definition: 421 */ 422#define pud_page(pud) pfn_to_page(pud_val(pud) >> PAGE_SHIFT) 423 424/* Find an entry in the second-level page table.. */ 425static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address) 426{ 427 return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address); 428} 429 430static inline int pud_large(pud_t pud) 431{ 432 return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) == 433 (_PAGE_PSE | _PAGE_PRESENT); 434} 435 436static inline int pud_bad(pud_t pud) 437{ 438 return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0; 439} 440#else 441static inline int pud_large(pud_t pud) 442{ 443 return 0; 444} 445#endif /* PAGETABLE_LEVELS > 2 */ 446 447#if PAGETABLE_LEVELS > 3 448static inline int pgd_present(pgd_t pgd) 449{ 450 return pgd_flags(pgd) & _PAGE_PRESENT; 451} 452 453static inline unsigned long pgd_page_vaddr(pgd_t pgd) 454{ 455 return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK); 456} 457 458/* 459 * Currently stuck as a macro due to indirect forward reference to 460 * linux/mmzone.h's __section_mem_map_addr() definition: 461 */ 462#define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT) 463 464/* to find an entry in a page-table-directory. */ 465static inline unsigned long pud_index(unsigned long address) 466{ 467 return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1); 468} 469 470static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address) 471{ 472 return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address); 473} 474 475static inline int pgd_bad(pgd_t pgd) 476{ 477 return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE; 478} 479 480static inline int pgd_none(pgd_t pgd) 481{ 482 return !native_pgd_val(pgd); 483} 484#endif /* PAGETABLE_LEVELS > 3 */ 485 486#endif /* __ASSEMBLY__ */ 487 488/* 489 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD] 490 * 491 * this macro returns the index of the entry in the pgd page which would 492 * control the given virtual address 493 */ 494#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1)) 495 496/* 497 * pgd_offset() returns a (pgd_t *) 498 * pgd_index() is used get the offset into the pgd page's array of pgd_t's; 499 */ 500#define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address))) 501/* 502 * a shortcut which implies the use of the kernel's pgd, instead 503 * of a process's 504 */ 505#define pgd_offset_k(address) pgd_offset(&init_mm, (address)) 506 507 508#define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET) 509#define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY) 510 511#ifndef __ASSEMBLY__ 512 513extern int direct_gbpages; 514 515/* local pte updates need not use xchg for locking */ 516static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep) 517{ 518 pte_t res = *ptep; 519 520 /* Pure native function needs no input for mm, addr */ 521 native_pte_clear(NULL, 0, ptep); 522 return res; 523} 524 525static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr, 526 pte_t *ptep , pte_t pte) 527{ 528 native_set_pte(ptep, pte); 529} 530 531#ifndef CONFIG_PARAVIRT 532/* 533 * Rules for using pte_update - it must be called after any PTE update which 534 * has not been done using the set_pte / clear_pte interfaces. It is used by 535 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE 536 * updates should either be sets, clears, or set_pte_atomic for P->P 537 * transitions, which means this hook should only be called for user PTEs. 538 * This hook implies a P->P protection or access change has taken place, which 539 * requires a subsequent TLB flush. The notification can optionally be delayed 540 * until the TLB flush event by using the pte_update_defer form of the 541 * interface, but care must be taken to assure that the flush happens while 542 * still holding the same page table lock so that the shadow and primary pages 543 * do not become out of sync on SMP. 544 */ 545#define pte_update(mm, addr, ptep) do { } while (0) 546#define pte_update_defer(mm, addr, ptep) do { } while (0) 547#endif 548 549/* 550 * We only update the dirty/accessed state if we set 551 * the dirty bit by hand in the kernel, since the hardware 552 * will do the accessed bit for us, and we don't want to 553 * race with other CPU's that might be updating the dirty 554 * bit at the same time. 555 */ 556struct vm_area_struct; 557 558#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS 559extern int ptep_set_access_flags(struct vm_area_struct *vma, 560 unsigned long address, pte_t *ptep, 561 pte_t entry, int dirty); 562 563#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG 564extern int ptep_test_and_clear_young(struct vm_area_struct *vma, 565 unsigned long addr, pte_t *ptep); 566 567#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH 568extern int ptep_clear_flush_young(struct vm_area_struct *vma, 569 unsigned long address, pte_t *ptep); 570 571#define __HAVE_ARCH_PTEP_GET_AND_CLEAR 572static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, 573 pte_t *ptep) 574{ 575 pte_t pte = native_ptep_get_and_clear(ptep); 576 pte_update(mm, addr, ptep); 577 return pte; 578} 579 580#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL 581static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, 582 unsigned long addr, pte_t *ptep, 583 int full) 584{ 585 pte_t pte; 586 if (full) { 587 /* 588 * Full address destruction in progress; paravirt does not 589 * care about updates and native needs no locking 590 */ 591 pte = native_local_ptep_get_and_clear(ptep); 592 } else { 593 pte = ptep_get_and_clear(mm, addr, ptep); 594 } 595 return pte; 596} 597 598#define __HAVE_ARCH_PTEP_SET_WRPROTECT 599static inline void ptep_set_wrprotect(struct mm_struct *mm, 600 unsigned long addr, pte_t *ptep) 601{ 602 clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte); 603 pte_update(mm, addr, ptep); 604} 605 606/* 607 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count); 608 * 609 * dst - pointer to pgd range anwhere on a pgd page 610 * src - "" 611 * count - the number of pgds to copy. 612 * 613 * dst and src can be on the same page, but the range must not overlap, 614 * and must not cross a page boundary. 615 */ 616static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count) 617{ 618 memcpy(dst, src, count * sizeof(pgd_t)); 619} 620 621 622#include <asm-generic/pgtable.h> 623#endif /* __ASSEMBLY__ */ 624 625#endif /* _ASM_X86_PGTABLE_H */ 626