1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _ASM_X86_PGTABLE_H
3 #define _ASM_X86_PGTABLE_H
5 #include <linux/mem_encrypt.h>
7 #include <asm/pgtable_types.h>
10 * Macro to mark a page protection value as UC-
12 #define pgprot_noncached(prot) \
13 ((boot_cpu_data.x86 > 3) \
14 ? (__pgprot(pgprot_val(prot) | \
15 cachemode2protval(_PAGE_CACHE_MODE_UC_MINUS))) \
19 * Macros to add or remove encryption attribute
21 #define pgprot_encrypted(prot) __pgprot(__sme_set(pgprot_val(prot)))
22 #define pgprot_decrypted(prot) __pgprot(__sme_clr(pgprot_val(prot)))
25 #include <asm/x86_init.h>
27 extern pgd_t early_top_pgt
[PTRS_PER_PGD
];
28 int __init
__early_make_pgtable(unsigned long address
, pmdval_t pmd
);
30 void ptdump_walk_pgd_level(struct seq_file
*m
, pgd_t
*pgd
);
31 void ptdump_walk_pgd_level_checkwx(void);
33 #ifdef CONFIG_DEBUG_WX
34 #define debug_checkwx() ptdump_walk_pgd_level_checkwx()
36 #define debug_checkwx() do { } while (0)
40 * ZERO_PAGE is a global shared page that is always zero: used
41 * for zero-mapped memory areas etc..
43 extern unsigned long empty_zero_page
[PAGE_SIZE
/ sizeof(unsigned long)]
45 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
47 extern spinlock_t pgd_lock
;
48 extern struct list_head pgd_list
;
50 extern struct mm_struct
*pgd_page_get_mm(struct page
*page
);
52 extern pmdval_t early_pmd_flags
;
54 #ifdef CONFIG_PARAVIRT
55 #include <asm/paravirt.h>
56 #else /* !CONFIG_PARAVIRT */
57 #define set_pte(ptep, pte) native_set_pte(ptep, pte)
58 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
60 #define set_pte_atomic(ptep, pte) \
61 native_set_pte_atomic(ptep, pte)
63 #define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
65 #ifndef __PAGETABLE_P4D_FOLDED
66 #define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
67 #define pgd_clear(pgd) native_pgd_clear(pgd)
71 # define set_p4d(p4dp, p4d) native_set_p4d(p4dp, p4d)
74 #ifndef __PAGETABLE_PUD_FOLDED
75 #define p4d_clear(p4d) native_p4d_clear(p4d)
79 # define set_pud(pudp, pud) native_set_pud(pudp, pud)
82 #ifndef __PAGETABLE_PUD_FOLDED
83 #define pud_clear(pud) native_pud_clear(pud)
86 #define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
87 #define pmd_clear(pmd) native_pmd_clear(pmd)
89 #define pgd_val(x) native_pgd_val(x)
90 #define __pgd(x) native_make_pgd(x)
92 #ifndef __PAGETABLE_P4D_FOLDED
93 #define p4d_val(x) native_p4d_val(x)
94 #define __p4d(x) native_make_p4d(x)
97 #ifndef __PAGETABLE_PUD_FOLDED
98 #define pud_val(x) native_pud_val(x)
99 #define __pud(x) native_make_pud(x)
102 #ifndef __PAGETABLE_PMD_FOLDED
103 #define pmd_val(x) native_pmd_val(x)
104 #define __pmd(x) native_make_pmd(x)
107 #define pte_val(x) native_pte_val(x)
108 #define __pte(x) native_make_pte(x)
110 #define arch_end_context_switch(prev) do {} while(0)
112 #endif /* CONFIG_PARAVIRT */
115 * The following only work if pte_present() is true.
116 * Undefined behaviour if not..
118 static inline int pte_dirty(pte_t pte
)
120 return pte_flags(pte
) & _PAGE_DIRTY
;
124 static inline u32
read_pkru(void)
126 if (boot_cpu_has(X86_FEATURE_OSPKE
))
127 return __read_pkru();
131 static inline void write_pkru(u32 pkru
)
133 if (boot_cpu_has(X86_FEATURE_OSPKE
))
137 static inline int pte_young(pte_t pte
)
139 return pte_flags(pte
) & _PAGE_ACCESSED
;
142 static inline int pmd_dirty(pmd_t pmd
)
144 return pmd_flags(pmd
) & _PAGE_DIRTY
;
147 static inline int pmd_young(pmd_t pmd
)
149 return pmd_flags(pmd
) & _PAGE_ACCESSED
;
152 static inline int pud_dirty(pud_t pud
)
154 return pud_flags(pud
) & _PAGE_DIRTY
;
157 static inline int pud_young(pud_t pud
)
159 return pud_flags(pud
) & _PAGE_ACCESSED
;
162 static inline int pte_write(pte_t pte
)
164 return pte_flags(pte
) & _PAGE_RW
;
167 static inline int pte_huge(pte_t pte
)
169 return pte_flags(pte
) & _PAGE_PSE
;
172 static inline int pte_global(pte_t pte
)
174 return pte_flags(pte
) & _PAGE_GLOBAL
;
177 static inline int pte_exec(pte_t pte
)
179 return !(pte_flags(pte
) & _PAGE_NX
);
182 static inline int pte_special(pte_t pte
)
184 return pte_flags(pte
) & _PAGE_SPECIAL
;
187 static inline unsigned long pte_pfn(pte_t pte
)
189 return (pte_val(pte
) & PTE_PFN_MASK
) >> PAGE_SHIFT
;
192 static inline unsigned long pmd_pfn(pmd_t pmd
)
194 return (pmd_val(pmd
) & pmd_pfn_mask(pmd
)) >> PAGE_SHIFT
;
197 static inline unsigned long pud_pfn(pud_t pud
)
199 return (pud_val(pud
) & pud_pfn_mask(pud
)) >> PAGE_SHIFT
;
202 static inline unsigned long p4d_pfn(p4d_t p4d
)
204 return (p4d_val(p4d
) & p4d_pfn_mask(p4d
)) >> PAGE_SHIFT
;
207 static inline unsigned long pgd_pfn(pgd_t pgd
)
209 return (pgd_val(pgd
) & PTE_PFN_MASK
) >> PAGE_SHIFT
;
212 static inline int p4d_large(p4d_t p4d
)
214 /* No 512 GiB pages yet */
218 #define pte_page(pte) pfn_to_page(pte_pfn(pte))
220 static inline int pmd_large(pmd_t pte
)
222 return pmd_flags(pte
) & _PAGE_PSE
;
225 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
226 static inline int pmd_trans_huge(pmd_t pmd
)
228 return (pmd_val(pmd
) & (_PAGE_PSE
|_PAGE_DEVMAP
)) == _PAGE_PSE
;
231 #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
232 static inline int pud_trans_huge(pud_t pud
)
234 return (pud_val(pud
) & (_PAGE_PSE
|_PAGE_DEVMAP
)) == _PAGE_PSE
;
238 #define has_transparent_hugepage has_transparent_hugepage
239 static inline int has_transparent_hugepage(void)
241 return boot_cpu_has(X86_FEATURE_PSE
);
244 #ifdef __HAVE_ARCH_PTE_DEVMAP
245 static inline int pmd_devmap(pmd_t pmd
)
247 return !!(pmd_val(pmd
) & _PAGE_DEVMAP
);
250 #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
251 static inline int pud_devmap(pud_t pud
)
253 return !!(pud_val(pud
) & _PAGE_DEVMAP
);
256 static inline int pud_devmap(pud_t pud
)
262 static inline int pgd_devmap(pgd_t pgd
)
267 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
269 static inline pte_t
pte_set_flags(pte_t pte
, pteval_t set
)
271 pteval_t v
= native_pte_val(pte
);
273 return native_make_pte(v
| set
);
276 static inline pte_t
pte_clear_flags(pte_t pte
, pteval_t clear
)
278 pteval_t v
= native_pte_val(pte
);
280 return native_make_pte(v
& ~clear
);
283 static inline pte_t
pte_mkclean(pte_t pte
)
285 return pte_clear_flags(pte
, _PAGE_DIRTY
);
288 static inline pte_t
pte_mkold(pte_t pte
)
290 return pte_clear_flags(pte
, _PAGE_ACCESSED
);
293 static inline pte_t
pte_wrprotect(pte_t pte
)
295 return pte_clear_flags(pte
, _PAGE_RW
);
298 static inline pte_t
pte_mkexec(pte_t pte
)
300 return pte_clear_flags(pte
, _PAGE_NX
);
303 static inline pte_t
pte_mkdirty(pte_t pte
)
305 return pte_set_flags(pte
, _PAGE_DIRTY
| _PAGE_SOFT_DIRTY
);
308 static inline pte_t
pte_mkyoung(pte_t pte
)
310 return pte_set_flags(pte
, _PAGE_ACCESSED
);
313 static inline pte_t
pte_mkwrite(pte_t pte
)
315 return pte_set_flags(pte
, _PAGE_RW
);
318 static inline pte_t
pte_mkhuge(pte_t pte
)
320 return pte_set_flags(pte
, _PAGE_PSE
);
323 static inline pte_t
pte_clrhuge(pte_t pte
)
325 return pte_clear_flags(pte
, _PAGE_PSE
);
328 static inline pte_t
pte_mkglobal(pte_t pte
)
330 return pte_set_flags(pte
, _PAGE_GLOBAL
);
333 static inline pte_t
pte_clrglobal(pte_t pte
)
335 return pte_clear_flags(pte
, _PAGE_GLOBAL
);
338 static inline pte_t
pte_mkspecial(pte_t pte
)
340 return pte_set_flags(pte
, _PAGE_SPECIAL
);
343 static inline pte_t
pte_mkdevmap(pte_t pte
)
345 return pte_set_flags(pte
, _PAGE_SPECIAL
|_PAGE_DEVMAP
);
348 static inline pmd_t
pmd_set_flags(pmd_t pmd
, pmdval_t set
)
350 pmdval_t v
= native_pmd_val(pmd
);
352 return __pmd(v
| set
);
355 static inline pmd_t
pmd_clear_flags(pmd_t pmd
, pmdval_t clear
)
357 pmdval_t v
= native_pmd_val(pmd
);
359 return __pmd(v
& ~clear
);
362 static inline pmd_t
pmd_mkold(pmd_t pmd
)
364 return pmd_clear_flags(pmd
, _PAGE_ACCESSED
);
367 static inline pmd_t
pmd_mkclean(pmd_t pmd
)
369 return pmd_clear_flags(pmd
, _PAGE_DIRTY
);
372 static inline pmd_t
pmd_wrprotect(pmd_t pmd
)
374 return pmd_clear_flags(pmd
, _PAGE_RW
);
377 static inline pmd_t
pmd_mkdirty(pmd_t pmd
)
379 return pmd_set_flags(pmd
, _PAGE_DIRTY
| _PAGE_SOFT_DIRTY
);
382 static inline pmd_t
pmd_mkdevmap(pmd_t pmd
)
384 return pmd_set_flags(pmd
, _PAGE_DEVMAP
);
387 static inline pmd_t
pmd_mkhuge(pmd_t pmd
)
389 return pmd_set_flags(pmd
, _PAGE_PSE
);
392 static inline pmd_t
pmd_mkyoung(pmd_t pmd
)
394 return pmd_set_flags(pmd
, _PAGE_ACCESSED
);
397 static inline pmd_t
pmd_mkwrite(pmd_t pmd
)
399 return pmd_set_flags(pmd
, _PAGE_RW
);
402 static inline pmd_t
pmd_mknotpresent(pmd_t pmd
)
404 return pmd_clear_flags(pmd
, _PAGE_PRESENT
| _PAGE_PROTNONE
);
407 static inline pud_t
pud_set_flags(pud_t pud
, pudval_t set
)
409 pudval_t v
= native_pud_val(pud
);
411 return __pud(v
| set
);
414 static inline pud_t
pud_clear_flags(pud_t pud
, pudval_t clear
)
416 pudval_t v
= native_pud_val(pud
);
418 return __pud(v
& ~clear
);
421 static inline pud_t
pud_mkold(pud_t pud
)
423 return pud_clear_flags(pud
, _PAGE_ACCESSED
);
426 static inline pud_t
pud_mkclean(pud_t pud
)
428 return pud_clear_flags(pud
, _PAGE_DIRTY
);
431 static inline pud_t
pud_wrprotect(pud_t pud
)
433 return pud_clear_flags(pud
, _PAGE_RW
);
436 static inline pud_t
pud_mkdirty(pud_t pud
)
438 return pud_set_flags(pud
, _PAGE_DIRTY
| _PAGE_SOFT_DIRTY
);
441 static inline pud_t
pud_mkdevmap(pud_t pud
)
443 return pud_set_flags(pud
, _PAGE_DEVMAP
);
446 static inline pud_t
pud_mkhuge(pud_t pud
)
448 return pud_set_flags(pud
, _PAGE_PSE
);
451 static inline pud_t
pud_mkyoung(pud_t pud
)
453 return pud_set_flags(pud
, _PAGE_ACCESSED
);
456 static inline pud_t
pud_mkwrite(pud_t pud
)
458 return pud_set_flags(pud
, _PAGE_RW
);
461 static inline pud_t
pud_mknotpresent(pud_t pud
)
463 return pud_clear_flags(pud
, _PAGE_PRESENT
| _PAGE_PROTNONE
);
466 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
467 static inline int pte_soft_dirty(pte_t pte
)
469 return pte_flags(pte
) & _PAGE_SOFT_DIRTY
;
472 static inline int pmd_soft_dirty(pmd_t pmd
)
474 return pmd_flags(pmd
) & _PAGE_SOFT_DIRTY
;
477 static inline int pud_soft_dirty(pud_t pud
)
479 return pud_flags(pud
) & _PAGE_SOFT_DIRTY
;
482 static inline pte_t
pte_mksoft_dirty(pte_t pte
)
484 return pte_set_flags(pte
, _PAGE_SOFT_DIRTY
);
487 static inline pmd_t
pmd_mksoft_dirty(pmd_t pmd
)
489 return pmd_set_flags(pmd
, _PAGE_SOFT_DIRTY
);
492 static inline pud_t
pud_mksoft_dirty(pud_t pud
)
494 return pud_set_flags(pud
, _PAGE_SOFT_DIRTY
);
497 static inline pte_t
pte_clear_soft_dirty(pte_t pte
)
499 return pte_clear_flags(pte
, _PAGE_SOFT_DIRTY
);
502 static inline pmd_t
pmd_clear_soft_dirty(pmd_t pmd
)
504 return pmd_clear_flags(pmd
, _PAGE_SOFT_DIRTY
);
507 static inline pud_t
pud_clear_soft_dirty(pud_t pud
)
509 return pud_clear_flags(pud
, _PAGE_SOFT_DIRTY
);
512 #endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */
515 * Mask out unsupported bits in a present pgprot. Non-present pgprots
516 * can use those bits for other purposes, so leave them be.
518 static inline pgprotval_t
massage_pgprot(pgprot_t pgprot
)
520 pgprotval_t protval
= pgprot_val(pgprot
);
522 if (protval
& _PAGE_PRESENT
)
523 protval
&= __supported_pte_mask
;
528 static inline pte_t
pfn_pte(unsigned long page_nr
, pgprot_t pgprot
)
530 return __pte(((phys_addr_t
)page_nr
<< PAGE_SHIFT
) |
531 massage_pgprot(pgprot
));
534 static inline pmd_t
pfn_pmd(unsigned long page_nr
, pgprot_t pgprot
)
536 return __pmd(((phys_addr_t
)page_nr
<< PAGE_SHIFT
) |
537 massage_pgprot(pgprot
));
540 static inline pud_t
pfn_pud(unsigned long page_nr
, pgprot_t pgprot
)
542 return __pud(((phys_addr_t
)page_nr
<< PAGE_SHIFT
) |
543 massage_pgprot(pgprot
));
546 static inline pte_t
pte_modify(pte_t pte
, pgprot_t newprot
)
548 pteval_t val
= pte_val(pte
);
551 * Chop off the NX bit (if present), and add the NX portion of
552 * the newprot (if present):
554 val
&= _PAGE_CHG_MASK
;
555 val
|= massage_pgprot(newprot
) & ~_PAGE_CHG_MASK
;
560 static inline pmd_t
pmd_modify(pmd_t pmd
, pgprot_t newprot
)
562 pmdval_t val
= pmd_val(pmd
);
564 val
&= _HPAGE_CHG_MASK
;
565 val
|= massage_pgprot(newprot
) & ~_HPAGE_CHG_MASK
;
570 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
571 #define pgprot_modify pgprot_modify
572 static inline pgprot_t
pgprot_modify(pgprot_t oldprot
, pgprot_t newprot
)
574 pgprotval_t preservebits
= pgprot_val(oldprot
) & _PAGE_CHG_MASK
;
575 pgprotval_t addbits
= pgprot_val(newprot
);
576 return __pgprot(preservebits
| addbits
);
579 #define pte_pgprot(x) __pgprot(pte_flags(x))
580 #define pmd_pgprot(x) __pgprot(pmd_flags(x))
581 #define pud_pgprot(x) __pgprot(pud_flags(x))
582 #define p4d_pgprot(x) __pgprot(p4d_flags(x))
584 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
586 static inline int is_new_memtype_allowed(u64 paddr
, unsigned long size
,
587 enum page_cache_mode pcm
,
588 enum page_cache_mode new_pcm
)
591 * PAT type is always WB for untracked ranges, so no need to check.
593 if (x86_platform
.is_untracked_pat_range(paddr
, paddr
+ size
))
597 * Certain new memtypes are not allowed with certain
599 * - request is uncached, return cannot be write-back
600 * - request is write-combine, return cannot be write-back
601 * - request is write-through, return cannot be write-back
602 * - request is write-through, return cannot be write-combine
604 if ((pcm
== _PAGE_CACHE_MODE_UC_MINUS
&&
605 new_pcm
== _PAGE_CACHE_MODE_WB
) ||
606 (pcm
== _PAGE_CACHE_MODE_WC
&&
607 new_pcm
== _PAGE_CACHE_MODE_WB
) ||
608 (pcm
== _PAGE_CACHE_MODE_WT
&&
609 new_pcm
== _PAGE_CACHE_MODE_WB
) ||
610 (pcm
== _PAGE_CACHE_MODE_WT
&&
611 new_pcm
== _PAGE_CACHE_MODE_WC
)) {
618 pmd_t
*populate_extra_pmd(unsigned long vaddr
);
619 pte_t
*populate_extra_pte(unsigned long vaddr
);
620 #endif /* __ASSEMBLY__ */
623 # include <asm/pgtable_32.h>
625 # include <asm/pgtable_64.h>
629 #include <linux/mm_types.h>
630 #include <linux/mmdebug.h>
631 #include <linux/log2.h>
632 #include <asm/fixmap.h>
634 static inline int pte_none(pte_t pte
)
636 return !(pte
.pte
& ~(_PAGE_KNL_ERRATUM_MASK
));
639 #define __HAVE_ARCH_PTE_SAME
640 static inline int pte_same(pte_t a
, pte_t b
)
642 return a
.pte
== b
.pte
;
645 static inline int pte_present(pte_t a
)
647 return pte_flags(a
) & (_PAGE_PRESENT
| _PAGE_PROTNONE
);
650 #ifdef __HAVE_ARCH_PTE_DEVMAP
651 static inline int pte_devmap(pte_t a
)
653 return (pte_flags(a
) & _PAGE_DEVMAP
) == _PAGE_DEVMAP
;
657 #define pte_accessible pte_accessible
658 static inline bool pte_accessible(struct mm_struct
*mm
, pte_t a
)
660 if (pte_flags(a
) & _PAGE_PRESENT
)
663 if ((pte_flags(a
) & _PAGE_PROTNONE
) &&
664 mm_tlb_flush_pending(mm
))
670 static inline int pmd_present(pmd_t pmd
)
673 * Checking for _PAGE_PSE is needed too because
674 * split_huge_page will temporarily clear the present bit (but
675 * the _PAGE_PSE flag will remain set at all times while the
676 * _PAGE_PRESENT bit is clear).
678 return pmd_flags(pmd
) & (_PAGE_PRESENT
| _PAGE_PROTNONE
| _PAGE_PSE
);
681 #ifdef CONFIG_NUMA_BALANCING
683 * These work without NUMA balancing but the kernel does not care. See the
684 * comment in include/asm-generic/pgtable.h
686 static inline int pte_protnone(pte_t pte
)
688 return (pte_flags(pte
) & (_PAGE_PROTNONE
| _PAGE_PRESENT
))
692 static inline int pmd_protnone(pmd_t pmd
)
694 return (pmd_flags(pmd
) & (_PAGE_PROTNONE
| _PAGE_PRESENT
))
697 #endif /* CONFIG_NUMA_BALANCING */
699 static inline int pmd_none(pmd_t pmd
)
701 /* Only check low word on 32-bit platforms, since it might be
702 out of sync with upper half. */
703 unsigned long val
= native_pmd_val(pmd
);
704 return (val
& ~_PAGE_KNL_ERRATUM_MASK
) == 0;
707 static inline unsigned long pmd_page_vaddr(pmd_t pmd
)
709 return (unsigned long)__va(pmd_val(pmd
) & pmd_pfn_mask(pmd
));
713 * Currently stuck as a macro due to indirect forward reference to
714 * linux/mmzone.h's __section_mem_map_addr() definition:
716 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
719 * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
721 * this macro returns the index of the entry in the pmd page which would
722 * control the given virtual address
724 static inline unsigned long pmd_index(unsigned long address
)
726 return (address
>> PMD_SHIFT
) & (PTRS_PER_PMD
- 1);
730 * Conversion functions: convert a page and protection to a page entry,
731 * and a page entry and page directory to the page they refer to.
733 * (Currently stuck as a macro because of indirect forward reference
734 * to linux/mm.h:page_to_nid())
736 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
739 * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
741 * this function returns the index of the entry in the pte page which would
742 * control the given virtual address
744 static inline unsigned long pte_index(unsigned long address
)
746 return (address
>> PAGE_SHIFT
) & (PTRS_PER_PTE
- 1);
749 static inline pte_t
*pte_offset_kernel(pmd_t
*pmd
, unsigned long address
)
751 return (pte_t
*)pmd_page_vaddr(*pmd
) + pte_index(address
);
754 static inline int pmd_bad(pmd_t pmd
)
756 return (pmd_flags(pmd
) & ~_PAGE_USER
) != _KERNPG_TABLE
;
759 static inline unsigned long pages_to_mb(unsigned long npg
)
761 return npg
>> (20 - PAGE_SHIFT
);
764 #if CONFIG_PGTABLE_LEVELS > 2
765 static inline int pud_none(pud_t pud
)
767 return (native_pud_val(pud
) & ~(_PAGE_KNL_ERRATUM_MASK
)) == 0;
770 static inline int pud_present(pud_t pud
)
772 return pud_flags(pud
) & _PAGE_PRESENT
;
775 static inline unsigned long pud_page_vaddr(pud_t pud
)
777 return (unsigned long)__va(pud_val(pud
) & pud_pfn_mask(pud
));
781 * Currently stuck as a macro due to indirect forward reference to
782 * linux/mmzone.h's __section_mem_map_addr() definition:
784 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
786 /* Find an entry in the second-level page table.. */
787 static inline pmd_t
*pmd_offset(pud_t
*pud
, unsigned long address
)
789 return (pmd_t
*)pud_page_vaddr(*pud
) + pmd_index(address
);
792 static inline int pud_large(pud_t pud
)
794 return (pud_val(pud
) & (_PAGE_PSE
| _PAGE_PRESENT
)) ==
795 (_PAGE_PSE
| _PAGE_PRESENT
);
798 static inline int pud_bad(pud_t pud
)
800 return (pud_flags(pud
) & ~(_KERNPG_TABLE
| _PAGE_USER
)) != 0;
803 static inline int pud_large(pud_t pud
)
807 #endif /* CONFIG_PGTABLE_LEVELS > 2 */
809 static inline unsigned long pud_index(unsigned long address
)
811 return (address
>> PUD_SHIFT
) & (PTRS_PER_PUD
- 1);
814 #if CONFIG_PGTABLE_LEVELS > 3
815 static inline int p4d_none(p4d_t p4d
)
817 return (native_p4d_val(p4d
) & ~(_PAGE_KNL_ERRATUM_MASK
)) == 0;
820 static inline int p4d_present(p4d_t p4d
)
822 return p4d_flags(p4d
) & _PAGE_PRESENT
;
825 static inline unsigned long p4d_page_vaddr(p4d_t p4d
)
827 return (unsigned long)__va(p4d_val(p4d
) & p4d_pfn_mask(p4d
));
831 * Currently stuck as a macro due to indirect forward reference to
832 * linux/mmzone.h's __section_mem_map_addr() definition:
834 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
836 /* Find an entry in the third-level page table.. */
837 static inline pud_t
*pud_offset(p4d_t
*p4d
, unsigned long address
)
839 return (pud_t
*)p4d_page_vaddr(*p4d
) + pud_index(address
);
842 static inline int p4d_bad(p4d_t p4d
)
844 return (p4d_flags(p4d
) & ~(_KERNPG_TABLE
| _PAGE_USER
)) != 0;
846 #endif /* CONFIG_PGTABLE_LEVELS > 3 */
848 static inline unsigned long p4d_index(unsigned long address
)
850 return (address
>> P4D_SHIFT
) & (PTRS_PER_P4D
- 1);
853 #if CONFIG_PGTABLE_LEVELS > 4
854 static inline int pgd_present(pgd_t pgd
)
856 return pgd_flags(pgd
) & _PAGE_PRESENT
;
859 static inline unsigned long pgd_page_vaddr(pgd_t pgd
)
861 return (unsigned long)__va((unsigned long)pgd_val(pgd
) & PTE_PFN_MASK
);
865 * Currently stuck as a macro due to indirect forward reference to
866 * linux/mmzone.h's __section_mem_map_addr() definition:
868 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
870 /* to find an entry in a page-table-directory. */
871 static inline p4d_t
*p4d_offset(pgd_t
*pgd
, unsigned long address
)
873 return (p4d_t
*)pgd_page_vaddr(*pgd
) + p4d_index(address
);
876 static inline int pgd_bad(pgd_t pgd
)
878 return (pgd_flags(pgd
) & ~_PAGE_USER
) != _KERNPG_TABLE
;
881 static inline int pgd_none(pgd_t pgd
)
884 * There is no need to do a workaround for the KNL stray
885 * A/D bit erratum here. PGDs only point to page tables
886 * except on 32-bit non-PAE which is not supported on
889 return !native_pgd_val(pgd
);
891 #endif /* CONFIG_PGTABLE_LEVELS > 4 */
893 #endif /* __ASSEMBLY__ */
896 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
898 * this macro returns the index of the entry in the pgd page which would
899 * control the given virtual address
901 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
904 * pgd_offset() returns a (pgd_t *)
905 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
907 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
909 * a shortcut which implies the use of the kernel's pgd, instead
912 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
915 #define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
916 #define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
920 extern int direct_gbpages
;
921 void init_mem_mapping(void);
922 void early_alloc_pgt_buf(void);
923 extern void memblock_find_dma_reserve(void);
926 /* Realmode trampoline initialization. */
927 extern pgd_t trampoline_pgd_entry
;
928 static inline void __meminit
init_trampoline_default(void)
930 /* Default trampoline pgd value */
931 trampoline_pgd_entry
= init_top_pgt
[pgd_index(__PAGE_OFFSET
)];
933 # ifdef CONFIG_RANDOMIZE_MEMORY
934 void __meminit
init_trampoline(void);
936 # define init_trampoline init_trampoline_default
939 static inline void init_trampoline(void) { }
942 /* local pte updates need not use xchg for locking */
943 static inline pte_t
native_local_ptep_get_and_clear(pte_t
*ptep
)
947 /* Pure native function needs no input for mm, addr */
948 native_pte_clear(NULL
, 0, ptep
);
952 static inline pmd_t
native_local_pmdp_get_and_clear(pmd_t
*pmdp
)
956 native_pmd_clear(pmdp
);
960 static inline pud_t
native_local_pudp_get_and_clear(pud_t
*pudp
)
964 native_pud_clear(pudp
);
968 static inline void native_set_pte_at(struct mm_struct
*mm
, unsigned long addr
,
969 pte_t
*ptep
, pte_t pte
)
971 native_set_pte(ptep
, pte
);
974 static inline void set_pmd_at(struct mm_struct
*mm
, unsigned long addr
,
975 pmd_t
*pmdp
, pmd_t pmd
)
977 native_set_pmd(pmdp
, pmd
);
980 static inline void set_pud_at(struct mm_struct
*mm
, unsigned long addr
,
981 pud_t
*pudp
, pud_t pud
)
983 native_set_pud(pudp
, pud
);
987 * We only update the dirty/accessed state if we set
988 * the dirty bit by hand in the kernel, since the hardware
989 * will do the accessed bit for us, and we don't want to
990 * race with other CPU's that might be updating the dirty
991 * bit at the same time.
993 struct vm_area_struct
;
995 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
996 extern int ptep_set_access_flags(struct vm_area_struct
*vma
,
997 unsigned long address
, pte_t
*ptep
,
998 pte_t entry
, int dirty
);
1000 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1001 extern int ptep_test_and_clear_young(struct vm_area_struct
*vma
,
1002 unsigned long addr
, pte_t
*ptep
);
1004 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1005 extern int ptep_clear_flush_young(struct vm_area_struct
*vma
,
1006 unsigned long address
, pte_t
*ptep
);
1008 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1009 static inline pte_t
ptep_get_and_clear(struct mm_struct
*mm
, unsigned long addr
,
1012 pte_t pte
= native_ptep_get_and_clear(ptep
);
1016 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1017 static inline pte_t
ptep_get_and_clear_full(struct mm_struct
*mm
,
1018 unsigned long addr
, pte_t
*ptep
,
1024 * Full address destruction in progress; paravirt does not
1025 * care about updates and native needs no locking
1027 pte
= native_local_ptep_get_and_clear(ptep
);
1029 pte
= ptep_get_and_clear(mm
, addr
, ptep
);
1034 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1035 static inline void ptep_set_wrprotect(struct mm_struct
*mm
,
1036 unsigned long addr
, pte_t
*ptep
)
1038 clear_bit(_PAGE_BIT_RW
, (unsigned long *)&ptep
->pte
);
1041 #define flush_tlb_fix_spurious_fault(vma, address) do { } while (0)
1043 #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
1045 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1046 extern int pmdp_set_access_flags(struct vm_area_struct
*vma
,
1047 unsigned long address
, pmd_t
*pmdp
,
1048 pmd_t entry
, int dirty
);
1049 extern int pudp_set_access_flags(struct vm_area_struct
*vma
,
1050 unsigned long address
, pud_t
*pudp
,
1051 pud_t entry
, int dirty
);
1053 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1054 extern int pmdp_test_and_clear_young(struct vm_area_struct
*vma
,
1055 unsigned long addr
, pmd_t
*pmdp
);
1056 extern int pudp_test_and_clear_young(struct vm_area_struct
*vma
,
1057 unsigned long addr
, pud_t
*pudp
);
1059 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1060 extern int pmdp_clear_flush_young(struct vm_area_struct
*vma
,
1061 unsigned long address
, pmd_t
*pmdp
);
1064 #define __HAVE_ARCH_PMD_WRITE
1065 static inline int pmd_write(pmd_t pmd
)
1067 return pmd_flags(pmd
) & _PAGE_RW
;
1070 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1071 static inline pmd_t
pmdp_huge_get_and_clear(struct mm_struct
*mm
, unsigned long addr
,
1074 return native_pmdp_get_and_clear(pmdp
);
1077 #define __HAVE_ARCH_PUDP_HUGE_GET_AND_CLEAR
1078 static inline pud_t
pudp_huge_get_and_clear(struct mm_struct
*mm
,
1079 unsigned long addr
, pud_t
*pudp
)
1081 return native_pudp_get_and_clear(pudp
);
1084 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1085 static inline void pmdp_set_wrprotect(struct mm_struct
*mm
,
1086 unsigned long addr
, pmd_t
*pmdp
)
1088 clear_bit(_PAGE_BIT_RW
, (unsigned long *)pmdp
);
1092 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
1094 * dst - pointer to pgd range anwhere on a pgd page
1096 * count - the number of pgds to copy.
1098 * dst and src can be on the same page, but the range must not overlap,
1099 * and must not cross a page boundary.
1101 static inline void clone_pgd_range(pgd_t
*dst
, pgd_t
*src
, int count
)
1103 memcpy(dst
, src
, count
* sizeof(pgd_t
));
1106 #define PTE_SHIFT ilog2(PTRS_PER_PTE)
1107 static inline int page_level_shift(enum pg_level level
)
1109 return (PAGE_SHIFT
- PTE_SHIFT
) + level
* PTE_SHIFT
;
1111 static inline unsigned long page_level_size(enum pg_level level
)
1113 return 1UL << page_level_shift(level
);
1115 static inline unsigned long page_level_mask(enum pg_level level
)
1117 return ~(page_level_size(level
) - 1);
1121 * The x86 doesn't have any external MMU info: the kernel page
1122 * tables contain all the necessary information.
1124 static inline void update_mmu_cache(struct vm_area_struct
*vma
,
1125 unsigned long addr
, pte_t
*ptep
)
1128 static inline void update_mmu_cache_pmd(struct vm_area_struct
*vma
,
1129 unsigned long addr
, pmd_t
*pmd
)
1132 static inline void update_mmu_cache_pud(struct vm_area_struct
*vma
,
1133 unsigned long addr
, pud_t
*pud
)
1137 #ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
1138 static inline pte_t
pte_swp_mksoft_dirty(pte_t pte
)
1140 return pte_set_flags(pte
, _PAGE_SWP_SOFT_DIRTY
);
1143 static inline int pte_swp_soft_dirty(pte_t pte
)
1145 return pte_flags(pte
) & _PAGE_SWP_SOFT_DIRTY
;
1148 static inline pte_t
pte_swp_clear_soft_dirty(pte_t pte
)
1150 return pte_clear_flags(pte
, _PAGE_SWP_SOFT_DIRTY
);
1153 #ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION
1154 static inline pmd_t
pmd_swp_mksoft_dirty(pmd_t pmd
)
1156 return pmd_set_flags(pmd
, _PAGE_SWP_SOFT_DIRTY
);
1159 static inline int pmd_swp_soft_dirty(pmd_t pmd
)
1161 return pmd_flags(pmd
) & _PAGE_SWP_SOFT_DIRTY
;
1164 static inline pmd_t
pmd_swp_clear_soft_dirty(pmd_t pmd
)
1166 return pmd_clear_flags(pmd
, _PAGE_SWP_SOFT_DIRTY
);
1171 #define PKRU_AD_BIT 0x1
1172 #define PKRU_WD_BIT 0x2
1173 #define PKRU_BITS_PER_PKEY 2
1175 static inline bool __pkru_allows_read(u32 pkru
, u16 pkey
)
1177 int pkru_pkey_bits
= pkey
* PKRU_BITS_PER_PKEY
;
1178 return !(pkru
& (PKRU_AD_BIT
<< pkru_pkey_bits
));
1181 static inline bool __pkru_allows_write(u32 pkru
, u16 pkey
)
1183 int pkru_pkey_bits
= pkey
* PKRU_BITS_PER_PKEY
;
1185 * Access-disable disables writes too so we need to check
1188 return !(pkru
& ((PKRU_AD_BIT
|PKRU_WD_BIT
) << pkru_pkey_bits
));
1191 static inline u16
pte_flags_pkey(unsigned long pte_flags
)
1193 #ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
1194 /* ifdef to avoid doing 59-bit shift on 32-bit values */
1195 return (pte_flags
& _PAGE_PKEY_MASK
) >> _PAGE_BIT_PKEY_BIT0
;
1201 static inline bool __pkru_allows_pkey(u16 pkey
, bool write
)
1203 u32 pkru
= read_pkru();
1205 if (!__pkru_allows_read(pkru
, pkey
))
1207 if (write
&& !__pkru_allows_write(pkru
, pkey
))
1214 * 'pteval' can come from a PTE, PMD or PUD. We only check
1215 * _PAGE_PRESENT, _PAGE_USER, and _PAGE_RW in here which are the
1216 * same value on all 3 types.
1218 static inline bool __pte_access_permitted(unsigned long pteval
, bool write
)
1220 unsigned long need_pte_bits
= _PAGE_PRESENT
|_PAGE_USER
;
1223 need_pte_bits
|= _PAGE_RW
;
1225 if ((pteval
& need_pte_bits
) != need_pte_bits
)
1228 return __pkru_allows_pkey(pte_flags_pkey(pteval
), write
);
1231 #define pte_access_permitted pte_access_permitted
1232 static inline bool pte_access_permitted(pte_t pte
, bool write
)
1234 return __pte_access_permitted(pte_val(pte
), write
);
1237 #define pmd_access_permitted pmd_access_permitted
1238 static inline bool pmd_access_permitted(pmd_t pmd
, bool write
)
1240 return __pte_access_permitted(pmd_val(pmd
), write
);
1243 #define pud_access_permitted pud_access_permitted
1244 static inline bool pud_access_permitted(pud_t pud
, bool write
)
1246 return __pte_access_permitted(pud_val(pud
), write
);
1249 #include <asm-generic/pgtable.h>
1250 #endif /* __ASSEMBLY__ */
1252 #endif /* _ASM_X86_PGTABLE_H */