mm: save soft-dirty bits on swapped pages
[deliverable/linux.git] / arch / x86 / include / asm / pgtable_types.h
CommitLineData
8d19c99f
JF
1#ifndef _ASM_X86_PGTABLE_DEFS_H
2#define _ASM_X86_PGTABLE_DEFS_H
3
4#include <linux/const.h>
e43623b4 5#include <asm/page_types.h>
8d19c99f
JF
6
7#define FIRST_USER_ADDRESS 0
8
9#define _PAGE_BIT_PRESENT 0 /* is present */
10#define _PAGE_BIT_RW 1 /* writeable */
11#define _PAGE_BIT_USER 2 /* userspace addressable */
12#define _PAGE_BIT_PWT 3 /* page write through */
13#define _PAGE_BIT_PCD 4 /* page cache disabled */
14#define _PAGE_BIT_ACCESSED 5 /* was accessed (raised by CPU) */
15#define _PAGE_BIT_DIRTY 6 /* was written to (raised by CPU) */
16#define _PAGE_BIT_PSE 7 /* 4 MB (or 2MB) page */
17#define _PAGE_BIT_PAT 7 /* on 4KB pages */
18#define _PAGE_BIT_GLOBAL 8 /* Global TLB entry PPro+ */
19#define _PAGE_BIT_UNUSED1 9 /* available for programmer */
20#define _PAGE_BIT_IOMAP 10 /* flag used to indicate IO mapping */
dfec072e 21#define _PAGE_BIT_HIDDEN 11 /* hidden by kmemcheck */
8d19c99f
JF
22#define _PAGE_BIT_PAT_LARGE 12 /* On 2MB or 1GB pages */
23#define _PAGE_BIT_SPECIAL _PAGE_BIT_UNUSED1
24#define _PAGE_BIT_CPA_TEST _PAGE_BIT_UNUSED1
5f6e8da7 25#define _PAGE_BIT_SPLITTING _PAGE_BIT_UNUSED1 /* only valid on a PSE pmd */
8d19c99f
JF
26#define _PAGE_BIT_NX 63 /* No execute: only valid after cpuid check */
27
28/* If _PAGE_BIT_PRESENT is clear, we use these: */
29/* - if the user mapped it with PROT_NONE; pte_present gives true */
30#define _PAGE_BIT_PROTNONE _PAGE_BIT_GLOBAL
31/* - set: nonlinear file mapping, saved PTE; unset:swap */
32#define _PAGE_BIT_FILE _PAGE_BIT_DIRTY
33
34#define _PAGE_PRESENT (_AT(pteval_t, 1) << _PAGE_BIT_PRESENT)
35#define _PAGE_RW (_AT(pteval_t, 1) << _PAGE_BIT_RW)
36#define _PAGE_USER (_AT(pteval_t, 1) << _PAGE_BIT_USER)
37#define _PAGE_PWT (_AT(pteval_t, 1) << _PAGE_BIT_PWT)
38#define _PAGE_PCD (_AT(pteval_t, 1) << _PAGE_BIT_PCD)
39#define _PAGE_ACCESSED (_AT(pteval_t, 1) << _PAGE_BIT_ACCESSED)
40#define _PAGE_DIRTY (_AT(pteval_t, 1) << _PAGE_BIT_DIRTY)
41#define _PAGE_PSE (_AT(pteval_t, 1) << _PAGE_BIT_PSE)
42#define _PAGE_GLOBAL (_AT(pteval_t, 1) << _PAGE_BIT_GLOBAL)
43#define _PAGE_UNUSED1 (_AT(pteval_t, 1) << _PAGE_BIT_UNUSED1)
44#define _PAGE_IOMAP (_AT(pteval_t, 1) << _PAGE_BIT_IOMAP)
8d19c99f
JF
45#define _PAGE_PAT (_AT(pteval_t, 1) << _PAGE_BIT_PAT)
46#define _PAGE_PAT_LARGE (_AT(pteval_t, 1) << _PAGE_BIT_PAT_LARGE)
47#define _PAGE_SPECIAL (_AT(pteval_t, 1) << _PAGE_BIT_SPECIAL)
48#define _PAGE_CPA_TEST (_AT(pteval_t, 1) << _PAGE_BIT_CPA_TEST)
5f6e8da7 49#define _PAGE_SPLITTING (_AT(pteval_t, 1) << _PAGE_BIT_SPLITTING)
8d19c99f
JF
50#define __HAVE_ARCH_PTE_SPECIAL
51
9d31c506
JF
52#ifdef CONFIG_KMEMCHECK
53#define _PAGE_HIDDEN (_AT(pteval_t, 1) << _PAGE_BIT_HIDDEN)
54#else
55#define _PAGE_HIDDEN (_AT(pteval_t, 0))
56#endif
57
0f8975ec
PE
58/*
59 * The same hidden bit is used by kmemcheck, but since kmemcheck
60 * works on kernel pages while soft-dirty engine on user space,
61 * they do not conflict with each other.
62 */
63
64#ifdef CONFIG_MEM_SOFT_DIRTY
65#define _PAGE_SOFT_DIRTY (_AT(pteval_t, 1) << _PAGE_BIT_HIDDEN)
66#else
67#define _PAGE_SOFT_DIRTY (_AT(pteval_t, 0))
68#endif
69
179ef71c
CG
70/*
71 * Tracking soft dirty bit when a page goes to a swap is tricky.
72 * We need a bit which can be stored in pte _and_ not conflict
73 * with swap entry format. On x86 bits 6 and 7 are *not* involved
74 * into swap entry computation, but bit 6 is used for nonlinear
75 * file mapping, so we borrow bit 7 for soft dirty tracking.
76 */
77#ifdef CONFIG_MEM_SOFT_DIRTY
78#define _PAGE_SWP_SOFT_DIRTY _PAGE_PSE
79#else
80#define _PAGE_SWP_SOFT_DIRTY (_AT(pteval_t, 0))
81#endif
82
8d19c99f
JF
83#if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE)
84#define _PAGE_NX (_AT(pteval_t, 1) << _PAGE_BIT_NX)
85#else
86#define _PAGE_NX (_AT(pteval_t, 0))
87#endif
88
89#define _PAGE_FILE (_AT(pteval_t, 1) << _PAGE_BIT_FILE)
90#define _PAGE_PROTNONE (_AT(pteval_t, 1) << _PAGE_BIT_PROTNONE)
91
dbe4d203
AA
92/*
93 * _PAGE_NUMA indicates that this page will trigger a numa hinting
94 * minor page fault to gather numa placement statistics (see
95 * pte_numa()). The bit picked (8) is within the range between
96 * _PAGE_FILE (6) and _PAGE_PROTNONE (8) bits. Therefore, it doesn't
97 * require changes to the swp entry format because that bit is always
98 * zero when the pte is not present.
99 *
100 * The bit picked must be always zero when the pmd is present and not
101 * present, so that we don't lose information when we set it while
102 * atomically clearing the present bit.
103 *
104 * Because we shared the same bit (8) with _PAGE_PROTNONE this can be
105 * interpreted as _PAGE_NUMA only in places that _PAGE_PROTNONE
106 * couldn't reach, like handle_mm_fault() (see access_error in
107 * arch/x86/mm/fault.c, the vma protection must not be PROT_NONE for
108 * handle_mm_fault() to be invoked).
109 */
110#define _PAGE_NUMA _PAGE_PROTNONE
111
8d19c99f
JF
112#define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | \
113 _PAGE_ACCESSED | _PAGE_DIRTY)
114#define _KERNPG_TABLE (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | \
115 _PAGE_DIRTY)
116
117/* Set of bits not changed in pte_modify */
118#define _PAGE_CHG_MASK (PTE_PFN_MASK | _PAGE_PCD | _PAGE_PWT | \
119 _PAGE_SPECIAL | _PAGE_ACCESSED | _PAGE_DIRTY)
c489f125 120#define _HPAGE_CHG_MASK (_PAGE_CHG_MASK | _PAGE_PSE)
8d19c99f
JF
121
122#define _PAGE_CACHE_MASK (_PAGE_PCD | _PAGE_PWT)
123#define _PAGE_CACHE_WB (0)
124#define _PAGE_CACHE_WC (_PAGE_PWT)
125#define _PAGE_CACHE_UC_MINUS (_PAGE_PCD)
126#define _PAGE_CACHE_UC (_PAGE_PCD | _PAGE_PWT)
127
128#define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_ACCESSED)
129#define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | \
130 _PAGE_ACCESSED | _PAGE_NX)
131
132#define PAGE_SHARED_EXEC __pgprot(_PAGE_PRESENT | _PAGE_RW | \
133 _PAGE_USER | _PAGE_ACCESSED)
134#define PAGE_COPY_NOEXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | \
135 _PAGE_ACCESSED | _PAGE_NX)
136#define PAGE_COPY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | \
137 _PAGE_ACCESSED)
138#define PAGE_COPY PAGE_COPY_NOEXEC
139#define PAGE_READONLY __pgprot(_PAGE_PRESENT | _PAGE_USER | \
140 _PAGE_ACCESSED | _PAGE_NX)
141#define PAGE_READONLY_EXEC __pgprot(_PAGE_PRESENT | _PAGE_USER | \
142 _PAGE_ACCESSED)
143
144#define __PAGE_KERNEL_EXEC \
145 (_PAGE_PRESENT | _PAGE_RW | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_GLOBAL)
146#define __PAGE_KERNEL (__PAGE_KERNEL_EXEC | _PAGE_NX)
147
148#define __PAGE_KERNEL_RO (__PAGE_KERNEL & ~_PAGE_RW)
149#define __PAGE_KERNEL_RX (__PAGE_KERNEL_EXEC & ~_PAGE_RW)
150#define __PAGE_KERNEL_EXEC_NOCACHE (__PAGE_KERNEL_EXEC | _PAGE_PCD | _PAGE_PWT)
151#define __PAGE_KERNEL_WC (__PAGE_KERNEL | _PAGE_CACHE_WC)
152#define __PAGE_KERNEL_NOCACHE (__PAGE_KERNEL | _PAGE_PCD | _PAGE_PWT)
153#define __PAGE_KERNEL_UC_MINUS (__PAGE_KERNEL | _PAGE_PCD)
154#define __PAGE_KERNEL_VSYSCALL (__PAGE_KERNEL_RX | _PAGE_USER)
9fd67b4e 155#define __PAGE_KERNEL_VVAR (__PAGE_KERNEL_RO | _PAGE_USER)
d319bb79 156#define __PAGE_KERNEL_VVAR_NOCACHE (__PAGE_KERNEL_VVAR | _PAGE_PCD | _PAGE_PWT)
8d19c99f
JF
157#define __PAGE_KERNEL_LARGE (__PAGE_KERNEL | _PAGE_PSE)
158#define __PAGE_KERNEL_LARGE_NOCACHE (__PAGE_KERNEL | _PAGE_CACHE_UC | _PAGE_PSE)
159#define __PAGE_KERNEL_LARGE_EXEC (__PAGE_KERNEL_EXEC | _PAGE_PSE)
160
161#define __PAGE_KERNEL_IO (__PAGE_KERNEL | _PAGE_IOMAP)
162#define __PAGE_KERNEL_IO_NOCACHE (__PAGE_KERNEL_NOCACHE | _PAGE_IOMAP)
163#define __PAGE_KERNEL_IO_UC_MINUS (__PAGE_KERNEL_UC_MINUS | _PAGE_IOMAP)
164#define __PAGE_KERNEL_IO_WC (__PAGE_KERNEL_WC | _PAGE_IOMAP)
165
166#define PAGE_KERNEL __pgprot(__PAGE_KERNEL)
167#define PAGE_KERNEL_RO __pgprot(__PAGE_KERNEL_RO)
168#define PAGE_KERNEL_EXEC __pgprot(__PAGE_KERNEL_EXEC)
169#define PAGE_KERNEL_RX __pgprot(__PAGE_KERNEL_RX)
170#define PAGE_KERNEL_WC __pgprot(__PAGE_KERNEL_WC)
171#define PAGE_KERNEL_NOCACHE __pgprot(__PAGE_KERNEL_NOCACHE)
172#define PAGE_KERNEL_UC_MINUS __pgprot(__PAGE_KERNEL_UC_MINUS)
173#define PAGE_KERNEL_EXEC_NOCACHE __pgprot(__PAGE_KERNEL_EXEC_NOCACHE)
174#define PAGE_KERNEL_LARGE __pgprot(__PAGE_KERNEL_LARGE)
175#define PAGE_KERNEL_LARGE_NOCACHE __pgprot(__PAGE_KERNEL_LARGE_NOCACHE)
176#define PAGE_KERNEL_LARGE_EXEC __pgprot(__PAGE_KERNEL_LARGE_EXEC)
177#define PAGE_KERNEL_VSYSCALL __pgprot(__PAGE_KERNEL_VSYSCALL)
9fd67b4e 178#define PAGE_KERNEL_VVAR __pgprot(__PAGE_KERNEL_VVAR)
d319bb79 179#define PAGE_KERNEL_VVAR_NOCACHE __pgprot(__PAGE_KERNEL_VVAR_NOCACHE)
8d19c99f
JF
180
181#define PAGE_KERNEL_IO __pgprot(__PAGE_KERNEL_IO)
182#define PAGE_KERNEL_IO_NOCACHE __pgprot(__PAGE_KERNEL_IO_NOCACHE)
183#define PAGE_KERNEL_IO_UC_MINUS __pgprot(__PAGE_KERNEL_IO_UC_MINUS)
184#define PAGE_KERNEL_IO_WC __pgprot(__PAGE_KERNEL_IO_WC)
185
186/* xwr */
187#define __P000 PAGE_NONE
188#define __P001 PAGE_READONLY
189#define __P010 PAGE_COPY
190#define __P011 PAGE_COPY
191#define __P100 PAGE_READONLY_EXEC
192#define __P101 PAGE_READONLY_EXEC
193#define __P110 PAGE_COPY_EXEC
194#define __P111 PAGE_COPY_EXEC
195
196#define __S000 PAGE_NONE
197#define __S001 PAGE_READONLY
198#define __S010 PAGE_SHARED
199#define __S011 PAGE_SHARED
200#define __S100 PAGE_READONLY_EXEC
201#define __S101 PAGE_READONLY_EXEC
202#define __S110 PAGE_SHARED_EXEC
203#define __S111 PAGE_SHARED_EXEC
204
205/*
206 * early identity mapping pte attrib macros.
207 */
208#ifdef CONFIG_X86_64
209#define __PAGE_KERNEL_IDENT_LARGE_EXEC __PAGE_KERNEL_LARGE_EXEC
210#else
211/*
212 * For PDE_IDENT_ATTR include USER bit. As the PDE and PTE protection
213 * bits are combined, this will alow user to access the high address mapped
214 * VDSO in the presence of CONFIG_COMPAT_VDSO
215 */
216#define PTE_IDENT_ATTR 0x003 /* PRESENT+RW */
217#define PDE_IDENT_ATTR 0x067 /* PRESENT+RW+USER+DIRTY+ACCESSED */
218#define PGD_IDENT_ATTR 0x001 /* PRESENT (no other attributes) */
219#endif
220
54321d94 221#ifdef CONFIG_X86_32
a1ce3928 222# include <asm/pgtable_32_types.h>
54321d94 223#else
a1ce3928 224# include <asm/pgtable_64_types.h>
54321d94
JF
225#endif
226
8d19c99f
JF
227#ifndef __ASSEMBLY__
228
54321d94
JF
229#include <linux/types.h>
230
9b3651cb
JF
231/* PTE_PFN_MASK extracts the PFN from a (pte|pmd|pud|pgd)val_t */
232#define PTE_PFN_MASK ((pteval_t)PHYSICAL_PAGE_MASK)
233
234/* PTE_FLAGS_MASK extracts the flags from a (pte|pmd|pud|pgd)val_t */
235#define PTE_FLAGS_MASK (~PTE_PFN_MASK)
236
54321d94
JF
237typedef struct pgprot { pgprotval_t pgprot; } pgprot_t;
238
239typedef struct { pgdval_t pgd; } pgd_t;
240
241static inline pgd_t native_make_pgd(pgdval_t val)
242{
243 return (pgd_t) { val };
244}
245
246static inline pgdval_t native_pgd_val(pgd_t pgd)
247{
248 return pgd.pgd;
249}
250
251static inline pgdval_t pgd_flags(pgd_t pgd)
252{
253 return native_pgd_val(pgd) & PTE_FLAGS_MASK;
254}
255
256#if PAGETABLE_LEVELS > 3
257typedef struct { pudval_t pud; } pud_t;
258
259static inline pud_t native_make_pud(pmdval_t val)
260{
261 return (pud_t) { val };
262}
263
264static inline pudval_t native_pud_val(pud_t pud)
265{
266 return pud.pud;
267}
268#else
269#include <asm-generic/pgtable-nopud.h>
270
271static inline pudval_t native_pud_val(pud_t pud)
272{
273 return native_pgd_val(pud.pgd);
274}
275#endif
276
277#if PAGETABLE_LEVELS > 2
278typedef struct { pmdval_t pmd; } pmd_t;
279
280static inline pmd_t native_make_pmd(pmdval_t val)
281{
282 return (pmd_t) { val };
283}
284
285static inline pmdval_t native_pmd_val(pmd_t pmd)
286{
287 return pmd.pmd;
288}
289#else
290#include <asm-generic/pgtable-nopmd.h>
291
292static inline pmdval_t native_pmd_val(pmd_t pmd)
293{
294 return native_pgd_val(pmd.pud.pgd);
295}
296#endif
297
298static inline pudval_t pud_flags(pud_t pud)
299{
300 return native_pud_val(pud) & PTE_FLAGS_MASK;
301}
302
303static inline pmdval_t pmd_flags(pmd_t pmd)
304{
305 return native_pmd_val(pmd) & PTE_FLAGS_MASK;
306}
307
308static inline pte_t native_make_pte(pteval_t val)
309{
310 return (pte_t) { .pte = val };
311}
312
313static inline pteval_t native_pte_val(pte_t pte)
314{
315 return pte.pte;
316}
317
318static inline pteval_t pte_flags(pte_t pte)
319{
320 return native_pte_val(pte) & PTE_FLAGS_MASK;
321}
322
323#define pgprot_val(x) ((x).pgprot)
324#define __pgprot(x) ((pgprot_t) { (x) } )
325
326
327typedef struct page *pgtable_t;
328
8d19c99f 329extern pteval_t __supported_pte_mask;
c44c9ec0 330extern void set_nx(void);
54321d94 331extern int nx_enabled;
8d19c99f
JF
332
333#define pgprot_writecombine pgprot_writecombine
334extern pgprot_t pgprot_writecombine(pgprot_t prot);
335
336/* Indicate that x86 has its own track and untrack pfn vma functions */
337#define __HAVE_PFNMAP_TRACKING
338
339#define __HAVE_PHYS_MEM_ACCESS_PROT
340struct file;
341pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
342 unsigned long size, pgprot_t vma_prot);
343int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
344 unsigned long size, pgprot_t *vma_prot);
345
346/* Install a pte for a particular vaddr in kernel space. */
347void set_pte_vaddr(unsigned long vaddr, pte_t pte);
348
349#ifdef CONFIG_X86_32
7737b215 350extern void native_pagetable_init(void);
8d19c99f 351#else
843b8ed2 352#define native_pagetable_init paging_init
8d19c99f
JF
353#endif
354
355struct seq_file;
356extern void arch_report_meminfo(struct seq_file *m);
357
4cbeb51b 358enum pg_level {
8d19c99f
JF
359 PG_LEVEL_NONE,
360 PG_LEVEL_4K,
361 PG_LEVEL_2M,
362 PG_LEVEL_1G,
363 PG_LEVEL_NUM
364};
365
366#ifdef CONFIG_PROC_FS
367extern void update_page_count(int level, unsigned long pages);
368#else
369static inline void update_page_count(int level, unsigned long pages) { }
370#endif
371
372/*
373 * Helper function that returns the kernel pagetable entry controlling
374 * the virtual address 'address'. NULL means no pagetable entry present.
375 * NOTE: the return type is pte_t but if the pmd is PSE then we return it
376 * as a pte too.
377 */
378extern pte_t *lookup_address(unsigned long address, unsigned int *level);
d7656534 379extern phys_addr_t slow_virt_to_phys(void *__address);
8d19c99f
JF
380
381#endif /* !__ASSEMBLY__ */
382
383#endif /* _ASM_X86_PGTABLE_DEFS_H */
This page took 0.270937 seconds and 5 git commands to generate.