Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1#ifndef _ASM_X86_PGTABLE_H
2#define _ASM_X86_PGTABLE_H
3
4#define FIRST_USER_ADDRESS 0
5
6#define _PAGE_BIT_PRESENT 0 /* is present */
7#define _PAGE_BIT_RW 1 /* writeable */
8#define _PAGE_BIT_USER 2 /* userspace addressable */
9#define _PAGE_BIT_PWT 3 /* page write through */
10#define _PAGE_BIT_PCD 4 /* page cache disabled */
11#define _PAGE_BIT_ACCESSED 5 /* was accessed (raised by CPU) */
12#define _PAGE_BIT_DIRTY 6 /* was written to (raised by CPU) */
13#define _PAGE_BIT_FILE 6
14#define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page */
15#define _PAGE_BIT_PAT 7 /* on 4KB pages */
16#define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
17#define _PAGE_BIT_UNUSED1 9 /* available for programmer */
18#define _PAGE_BIT_UNUSED2 10
19#define _PAGE_BIT_UNUSED3 11
20#define _PAGE_BIT_PAT_LARGE 12 /* On 2MB or 1GB pages */
21#define _PAGE_BIT_SPECIAL _PAGE_BIT_UNUSED1
22#define _PAGE_BIT_NX 63 /* No execute: only valid after cpuid check */
23
24#define _PAGE_PRESENT (_AT(pteval_t, 1) << _PAGE_BIT_PRESENT)
25#define _PAGE_RW (_AT(pteval_t, 1) << _PAGE_BIT_RW)
26#define _PAGE_USER (_AT(pteval_t, 1) << _PAGE_BIT_USER)
27#define _PAGE_PWT (_AT(pteval_t, 1) << _PAGE_BIT_PWT)
28#define _PAGE_PCD (_AT(pteval_t, 1) << _PAGE_BIT_PCD)
29#define _PAGE_ACCESSED (_AT(pteval_t, 1) << _PAGE_BIT_ACCESSED)
30#define _PAGE_DIRTY (_AT(pteval_t, 1) << _PAGE_BIT_DIRTY)
31#define _PAGE_PSE (_AT(pteval_t, 1) << _PAGE_BIT_PSE)
32#define _PAGE_GLOBAL (_AT(pteval_t, 1) << _PAGE_BIT_GLOBAL)
33#define _PAGE_UNUSED1 (_AT(pteval_t, 1) << _PAGE_BIT_UNUSED1)
34#define _PAGE_UNUSED2 (_AT(pteval_t, 1) << _PAGE_BIT_UNUSED2)
35#define _PAGE_UNUSED3 (_AT(pteval_t, 1) << _PAGE_BIT_UNUSED3)
36#define _PAGE_PAT (_AT(pteval_t, 1) << _PAGE_BIT_PAT)
37#define _PAGE_PAT_LARGE (_AT(pteval_t, 1) << _PAGE_BIT_PAT_LARGE)
38#define _PAGE_SPECIAL (_AT(pteval_t, 1) << _PAGE_BIT_SPECIAL)
39#define __HAVE_ARCH_PTE_SPECIAL
40
41#if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE)
42#define _PAGE_NX (_AT(pteval_t, 1) << _PAGE_BIT_NX)
43#else
44#define _PAGE_NX (_AT(pteval_t, 0))
45#endif
46
47/* If _PAGE_PRESENT is clear, we use these: */
48#define _PAGE_FILE _PAGE_DIRTY /* nonlinear file mapping,
49 * saved PTE; unset:swap */
50#define _PAGE_PROTNONE _PAGE_PSE /* if the user mapped it with PROT_NONE;
51 pte_present gives true */
52
53#define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | \
54 _PAGE_ACCESSED | _PAGE_DIRTY)
55#define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | \
56 _PAGE_DIRTY)
57
58/* Set of bits not changed in pte_modify */
59#define _PAGE_CHG_MASK (PTE_PFN_MASK | _PAGE_PCD | _PAGE_PWT | \
60 _PAGE_SPECIAL | _PAGE_ACCESSED | _PAGE_DIRTY)
61
62#define _PAGE_CACHE_MASK (_PAGE_PCD | _PAGE_PWT)
63#define _PAGE_CACHE_WB (0)
64#define _PAGE_CACHE_WC (_PAGE_PWT)
65#define _PAGE_CACHE_UC_MINUS (_PAGE_PCD)
66#define _PAGE_CACHE_UC (_PAGE_PCD | _PAGE_PWT)
67
68#define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
69#define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | \
70 _PAGE_ACCESSED | _PAGE_NX)
71
72#define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | \
73 _PAGE_USER | _PAGE_ACCESSED)
74#define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | \
75 _PAGE_ACCESSED | _PAGE_NX)
76#define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | \
77 _PAGE_ACCESSED)
78#define PAGE_COPY PAGE_COPY_NOEXEC
79#define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER | \
80 _PAGE_ACCESSED | _PAGE_NX)
81#define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | \
82 _PAGE_ACCESSED)
83
84#define __PAGE_KERNEL_EXEC \
85 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_GLOBAL)
86#define __PAGE_KERNEL (__PAGE_KERNEL_EXEC | _PAGE_NX)
87
88#define __PAGE_KERNEL_RO (__PAGE_KERNEL & ~_PAGE_RW)
89#define __PAGE_KERNEL_RX (__PAGE_KERNEL_EXEC & ~_PAGE_RW)
90#define __PAGE_KERNEL_EXEC_NOCACHE (__PAGE_KERNEL_EXEC | _PAGE_PCD | _PAGE_PWT)
91#define __PAGE_KERNEL_WC (__PAGE_KERNEL | _PAGE_CACHE_WC)
92#define __PAGE_KERNEL_NOCACHE (__PAGE_KERNEL | _PAGE_PCD | _PAGE_PWT)
93#define __PAGE_KERNEL_UC_MINUS (__PAGE_KERNEL | _PAGE_PCD)
94#define __PAGE_KERNEL_VSYSCALL (__PAGE_KERNEL_RX | _PAGE_USER)
95#define __PAGE_KERNEL_VSYSCALL_NOCACHE (__PAGE_KERNEL_VSYSCALL | _PAGE_PCD | _PAGE_PWT)
96#define __PAGE_KERNEL_LARGE (__PAGE_KERNEL | _PAGE_PSE)
97#define __PAGE_KERNEL_LARGE_NOCACHE (__PAGE_KERNEL | _PAGE_CACHE_UC | _PAGE_PSE)
98#define __PAGE_KERNEL_LARGE_EXEC (__PAGE_KERNEL_EXEC | _PAGE_PSE)
99
100#define PAGE_KERNEL __pgprot(__PAGE_KERNEL)
101#define PAGE_KERNEL_RO __pgprot(__PAGE_KERNEL_RO)
102#define PAGE_KERNEL_EXEC __pgprot(__PAGE_KERNEL_EXEC)
103#define PAGE_KERNEL_RX __pgprot(__PAGE_KERNEL_RX)
104#define PAGE_KERNEL_WC __pgprot(__PAGE_KERNEL_WC)
105#define PAGE_KERNEL_NOCACHE __pgprot(__PAGE_KERNEL_NOCACHE)
106#define PAGE_KERNEL_UC_MINUS __pgprot(__PAGE_KERNEL_UC_MINUS)
107#define PAGE_KERNEL_EXEC_NOCACHE __pgprot(__PAGE_KERNEL_EXEC_NOCACHE)
108#define PAGE_KERNEL_LARGE __pgprot(__PAGE_KERNEL_LARGE)
109#define PAGE_KERNEL_LARGE_NOCACHE __pgprot(__PAGE_KERNEL_LARGE_NOCACHE)
110#define PAGE_KERNEL_LARGE_EXEC __pgprot(__PAGE_KERNEL_LARGE_EXEC)
111#define PAGE_KERNEL_VSYSCALL __pgprot(__PAGE_KERNEL_VSYSCALL)
112#define PAGE_KERNEL_VSYSCALL_NOCACHE __pgprot(__PAGE_KERNEL_VSYSCALL_NOCACHE)
113
114/* xwr */
115#define __P000 PAGE_NONE
116#define __P001 PAGE_READONLY
117#define __P010 PAGE_COPY
118#define __P011 PAGE_COPY
119#define __P100 PAGE_READONLY_EXEC
120#define __P101 PAGE_READONLY_EXEC
121#define __P110 PAGE_COPY_EXEC
122#define __P111 PAGE_COPY_EXEC
123
124#define __S000 PAGE_NONE
125#define __S001 PAGE_READONLY
126#define __S010 PAGE_SHARED
127#define __S011 PAGE_SHARED
128#define __S100 PAGE_READONLY_EXEC
129#define __S101 PAGE_READONLY_EXEC
130#define __S110 PAGE_SHARED_EXEC
131#define __S111 PAGE_SHARED_EXEC
132
133#ifndef __ASSEMBLY__
134
135/*
136 * ZERO_PAGE is a global shared page that is always zero: used
137 * for zero-mapped memory areas etc..
138 */
139extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
140#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
141
142extern spinlock_t pgd_lock;
143extern struct list_head pgd_list;
144
145/*
146 * The following only work if pte_present() is true.
147 * Undefined behaviour if not..
148 */
149static inline int pte_dirty(pte_t pte)
150{
151 return pte_flags(pte) & _PAGE_DIRTY;
152}
153
154static inline int pte_young(pte_t pte)
155{
156 return pte_flags(pte) & _PAGE_ACCESSED;
157}
158
159static inline int pte_write(pte_t pte)
160{
161 return pte_flags(pte) & _PAGE_RW;
162}
163
164static inline int pte_file(pte_t pte)
165{
166 return pte_flags(pte) & _PAGE_FILE;
167}
168
169static inline int pte_huge(pte_t pte)
170{
171 return pte_flags(pte) & _PAGE_PSE;
172}
173
174static inline int pte_global(pte_t pte)
175{
176 return pte_flags(pte) & _PAGE_GLOBAL;
177}
178
179static inline int pte_exec(pte_t pte)
180{
181 return !(pte_flags(pte) & _PAGE_NX);
182}
183
184static inline int pte_special(pte_t pte)
185{
186 return pte_val(pte) & _PAGE_SPECIAL;
187}
188
189static inline int pmd_large(pmd_t pte)
190{
191 return (pmd_val(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
192 (_PAGE_PSE | _PAGE_PRESENT);
193}
194
195static inline pte_t pte_mkclean(pte_t pte)
196{
197 return __pte(pte_val(pte) & ~_PAGE_DIRTY);
198}
199
200static inline pte_t pte_mkold(pte_t pte)
201{
202 return __pte(pte_val(pte) & ~_PAGE_ACCESSED);
203}
204
205static inline pte_t pte_wrprotect(pte_t pte)
206{
207 return __pte(pte_val(pte) & ~_PAGE_RW);
208}
209
210static inline pte_t pte_mkexec(pte_t pte)
211{
212 return __pte(pte_val(pte) & ~_PAGE_NX);
213}
214
215static inline pte_t pte_mkdirty(pte_t pte)
216{
217 return __pte(pte_val(pte) | _PAGE_DIRTY);
218}
219
220static inline pte_t pte_mkyoung(pte_t pte)
221{
222 return __pte(pte_val(pte) | _PAGE_ACCESSED);
223}
224
225static inline pte_t pte_mkwrite(pte_t pte)
226{
227 return __pte(pte_val(pte) | _PAGE_RW);
228}
229
230static inline pte_t pte_mkhuge(pte_t pte)
231{
232 return __pte(pte_val(pte) | _PAGE_PSE);
233}
234
235static inline pte_t pte_clrhuge(pte_t pte)
236{
237 return __pte(pte_val(pte) & ~_PAGE_PSE);
238}
239
240static inline pte_t pte_mkglobal(pte_t pte)
241{
242 return __pte(pte_val(pte) | _PAGE_GLOBAL);
243}
244
245static inline pte_t pte_clrglobal(pte_t pte)
246{
247 return __pte(pte_val(pte) & ~_PAGE_GLOBAL);
248}
249
250static inline pte_t pte_mkspecial(pte_t pte)
251{
252 return __pte(pte_val(pte) | _PAGE_SPECIAL);
253}
254
255extern pteval_t __supported_pte_mask;
256
257static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
258{
259 return __pte((((phys_addr_t)page_nr << PAGE_SHIFT) |
260 pgprot_val(pgprot)) & __supported_pte_mask);
261}
262
263static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
264{
265 return __pmd((((phys_addr_t)page_nr << PAGE_SHIFT) |
266 pgprot_val(pgprot)) & __supported_pte_mask);
267}
268
269static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
270{
271 pteval_t val = pte_val(pte);
272
273 /*
274 * Chop off the NX bit (if present), and add the NX portion of
275 * the newprot (if present):
276 */
277 val &= _PAGE_CHG_MASK;
278 val |= pgprot_val(newprot) & (~_PAGE_CHG_MASK) & __supported_pte_mask;
279
280 return __pte(val);
281}
282
283/* mprotect needs to preserve PAT bits when updating vm_page_prot */
284#define pgprot_modify pgprot_modify
285static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
286{
287 pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
288 pgprotval_t addbits = pgprot_val(newprot);
289 return __pgprot(preservebits | addbits);
290}
291
292#define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
293
294#define canon_pgprot(p) __pgprot(pgprot_val(p) & __supported_pte_mask)
295
296#ifndef __ASSEMBLY__
297#define __HAVE_PHYS_MEM_ACCESS_PROT
298struct file;
299pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
300 unsigned long size, pgprot_t vma_prot);
301int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
302 unsigned long size, pgprot_t *vma_prot);
303#endif
304
305/* Install a pte for a particular vaddr in kernel space. */
306void set_pte_vaddr(unsigned long vaddr, pte_t pte);
307
308#ifdef CONFIG_X86_32
309extern void native_pagetable_setup_start(pgd_t *base);
310extern void native_pagetable_setup_done(pgd_t *base);
311#else
312static inline void native_pagetable_setup_start(pgd_t *base) {}
313static inline void native_pagetable_setup_done(pgd_t *base) {}
314#endif
315
316#ifdef CONFIG_PARAVIRT
317#include <asm/paravirt.h>
318#else /* !CONFIG_PARAVIRT */
319#define set_pte(ptep, pte) native_set_pte(ptep, pte)
320#define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
321
322#define set_pte_present(mm, addr, ptep, pte) \
323 native_set_pte_present(mm, addr, ptep, pte)
324#define set_pte_atomic(ptep, pte) \
325 native_set_pte_atomic(ptep, pte)
326
327#define set_pmd(pmdp, pmd) native_set_pmd(pmdp, pmd)
328
329#ifndef __PAGETABLE_PUD_FOLDED
330#define set_pgd(pgdp, pgd) native_set_pgd(pgdp, pgd)
331#define pgd_clear(pgd) native_pgd_clear(pgd)
332#endif
333
334#ifndef set_pud
335# define set_pud(pudp, pud) native_set_pud(pudp, pud)
336#endif
337
338#ifndef __PAGETABLE_PMD_FOLDED
339#define pud_clear(pud) native_pud_clear(pud)
340#endif
341
342#define pte_clear(mm, addr, ptep) native_pte_clear(mm, addr, ptep)
343#define pmd_clear(pmd) native_pmd_clear(pmd)
344
345#define pte_update(mm, addr, ptep) do { } while (0)
346#define pte_update_defer(mm, addr, ptep) do { } while (0)
347
348static inline void __init paravirt_pagetable_setup_start(pgd_t *base)
349{
350 native_pagetable_setup_start(base);
351}
352
353static inline void __init paravirt_pagetable_setup_done(pgd_t *base)
354{
355 native_pagetable_setup_done(base);
356}
357#endif /* CONFIG_PARAVIRT */
358
359#endif /* __ASSEMBLY__ */
360
361#ifdef CONFIG_X86_32
362# include "pgtable_32.h"
363#else
364# include "pgtable_64.h"
365#endif
366
367/*
368 * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
369 *
370 * this macro returns the index of the entry in the pgd page which would
371 * control the given virtual address
372 */
373#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
374
375/*
376 * pgd_offset() returns a (pgd_t *)
377 * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
378 */
379#define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
380/*
381 * a shortcut which implies the use of the kernel's pgd, instead
382 * of a process's
383 */
384#define pgd_offset_k(address) pgd_offset(&init_mm, (address))
385
386
387#define KERNEL_PGD_BOUNDARY pgd_index(PAGE_OFFSET)
388#define KERNEL_PGD_PTRS (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
389
390#ifndef __ASSEMBLY__
391
392enum {
393 PG_LEVEL_NONE,
394 PG_LEVEL_4K,
395 PG_LEVEL_2M,
396 PG_LEVEL_1G,
397 PG_LEVEL_NUM
398};
399
400#ifdef CONFIG_PROC_FS
401extern void update_page_count(int level, unsigned long pages);
402#else
403static inline void update_page_count(int level, unsigned long pages) { }
404#endif
405
406/*
407 * Helper function that returns the kernel pagetable entry controlling
408 * the virtual address 'address'. NULL means no pagetable entry present.
409 * NOTE: the return type is pte_t but if the pmd is PSE then we return it
410 * as a pte too.
411 */
412extern pte_t *lookup_address(unsigned long address, unsigned int *level);
413
414/* local pte updates need not use xchg for locking */
415static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
416{
417 pte_t res = *ptep;
418
419 /* Pure native function needs no input for mm, addr */
420 native_pte_clear(NULL, 0, ptep);
421 return res;
422}
423
424static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
425 pte_t *ptep , pte_t pte)
426{
427 native_set_pte(ptep, pte);
428}
429
430#ifndef CONFIG_PARAVIRT
431/*
432 * Rules for using pte_update - it must be called after any PTE update which
433 * has not been done using the set_pte / clear_pte interfaces. It is used by
434 * shadow mode hypervisors to resynchronize the shadow page tables. Kernel PTE
435 * updates should either be sets, clears, or set_pte_atomic for P->P
436 * transitions, which means this hook should only be called for user PTEs.
437 * This hook implies a P->P protection or access change has taken place, which
438 * requires a subsequent TLB flush. The notification can optionally be delayed
439 * until the TLB flush event by using the pte_update_defer form of the
440 * interface, but care must be taken to assure that the flush happens while
441 * still holding the same page table lock so that the shadow and primary pages
442 * do not become out of sync on SMP.
443 */
444#define pte_update(mm, addr, ptep) do { } while (0)
445#define pte_update_defer(mm, addr, ptep) do { } while (0)
446#endif
447
448/*
449 * We only update the dirty/accessed state if we set
450 * the dirty bit by hand in the kernel, since the hardware
451 * will do the accessed bit for us, and we don't want to
452 * race with other CPU's that might be updating the dirty
453 * bit at the same time.
454 */
455struct vm_area_struct;
456
457#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
458extern int ptep_set_access_flags(struct vm_area_struct *vma,
459 unsigned long address, pte_t *ptep,
460 pte_t entry, int dirty);
461
462#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
463extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
464 unsigned long addr, pte_t *ptep);
465
466#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
467extern int ptep_clear_flush_young(struct vm_area_struct *vma,
468 unsigned long address, pte_t *ptep);
469
470#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
471static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
472 pte_t *ptep)
473{
474 pte_t pte = native_ptep_get_and_clear(ptep);
475 pte_update(mm, addr, ptep);
476 return pte;
477}
478
479#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
480static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
481 unsigned long addr, pte_t *ptep,
482 int full)
483{
484 pte_t pte;
485 if (full) {
486 /*
487 * Full address destruction in progress; paravirt does not
488 * care about updates and native needs no locking
489 */
490 pte = native_local_ptep_get_and_clear(ptep);
491 } else {
492 pte = ptep_get_and_clear(mm, addr, ptep);
493 }
494 return pte;
495}
496
497#define __HAVE_ARCH_PTEP_SET_WRPROTECT
498static inline void ptep_set_wrprotect(struct mm_struct *mm,
499 unsigned long addr, pte_t *ptep)
500{
501 clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
502 pte_update(mm, addr, ptep);
503}
504
505/*
506 * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
507 *
508 * dst - pointer to pgd range anwhere on a pgd page
509 * src - ""
510 * count - the number of pgds to copy.
511 *
512 * dst and src can be on the same page, but the range must not overlap,
513 * and must not cross a page boundary.
514 */
515static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
516{
517 memcpy(dst, src, count * sizeof(pgd_t));
518}
519
520
521#include <asm-generic/pgtable.h>
522#endif /* __ASSEMBLY__ */
523
524#endif /* _ASM_X86_PGTABLE_H */