blob: 1a97af31ef176e71d8138250c03c52b7ab0adf3f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _ASM_IA64_PGTABLE_H
2#define _ASM_IA64_PGTABLE_H
3
4/*
5 * This file contains the functions and defines necessary to modify and use
6 * the IA-64 page table tree.
7 *
8 * This hopefully works with any (fixed) IA-64 page-size, as defined
9 * in <asm/page.h>.
10 *
David Mosberger-Tangad597bd2005-06-08 10:45:00 -070011 * Copyright (C) 1998-2005 Hewlett-Packard Co
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * David Mosberger-Tang <davidm@hpl.hp.com>
13 */
14
Linus Torvalds1da177e2005-04-16 15:20:36 -070015
16#include <asm/mman.h>
17#include <asm/page.h>
18#include <asm/processor.h>
19#include <asm/system.h>
20#include <asm/types.h>
21
22#define IA64_MAX_PHYS_BITS 50 /* max. number of physical address bits (architected) */
23
24/*
25 * First, define the various bits in a PTE. Note that the PTE format
26 * matches the VHPT short format, the firt doubleword of the VHPD long
27 * format, and the first doubleword of the TLB insertion format.
28 */
29#define _PAGE_P_BIT 0
30#define _PAGE_A_BIT 5
31#define _PAGE_D_BIT 6
32
33#define _PAGE_P (1 << _PAGE_P_BIT) /* page present bit */
34#define _PAGE_MA_WB (0x0 << 2) /* write back memory attribute */
35#define _PAGE_MA_UC (0x4 << 2) /* uncacheable memory attribute */
36#define _PAGE_MA_UCE (0x5 << 2) /* UC exported attribute */
37#define _PAGE_MA_WC (0x6 << 2) /* write coalescing memory attribute */
38#define _PAGE_MA_NAT (0x7 << 2) /* not-a-thing attribute */
39#define _PAGE_MA_MASK (0x7 << 2)
40#define _PAGE_PL_0 (0 << 7) /* privilege level 0 (kernel) */
41#define _PAGE_PL_1 (1 << 7) /* privilege level 1 (unused) */
42#define _PAGE_PL_2 (2 << 7) /* privilege level 2 (unused) */
43#define _PAGE_PL_3 (3 << 7) /* privilege level 3 (user) */
44#define _PAGE_PL_MASK (3 << 7)
45#define _PAGE_AR_R (0 << 9) /* read only */
46#define _PAGE_AR_RX (1 << 9) /* read & execute */
47#define _PAGE_AR_RW (2 << 9) /* read & write */
48#define _PAGE_AR_RWX (3 << 9) /* read, write & execute */
49#define _PAGE_AR_R_RW (4 << 9) /* read / read & write */
50#define _PAGE_AR_RX_RWX (5 << 9) /* read & exec / read, write & exec */
51#define _PAGE_AR_RWX_RW (6 << 9) /* read, write & exec / read & write */
52#define _PAGE_AR_X_RX (7 << 9) /* exec & promote / read & exec */
53#define _PAGE_AR_MASK (7 << 9)
54#define _PAGE_AR_SHIFT 9
55#define _PAGE_A (1 << _PAGE_A_BIT) /* page accessed bit */
56#define _PAGE_D (1 << _PAGE_D_BIT) /* page dirty bit */
57#define _PAGE_PPN_MASK (((__IA64_UL(1) << IA64_MAX_PHYS_BITS) - 1) & ~0xfffUL)
58#define _PAGE_ED (__IA64_UL(1) << 52) /* exception deferral */
59#define _PAGE_PROTNONE (__IA64_UL(1) << 63)
60
61/* Valid only for a PTE with the present bit cleared: */
62#define _PAGE_FILE (1 << 1) /* see swap & file pte remarks below */
63
64#define _PFN_MASK _PAGE_PPN_MASK
65/* Mask of bits which may be changed by pte_modify(); the odd bits are there for _PAGE_PROTNONE */
66#define _PAGE_CHG_MASK (_PAGE_P | _PAGE_PROTNONE | _PAGE_PL_MASK | _PAGE_AR_MASK | _PAGE_ED)
67
68#define _PAGE_SIZE_4K 12
69#define _PAGE_SIZE_8K 13
70#define _PAGE_SIZE_16K 14
71#define _PAGE_SIZE_64K 16
72#define _PAGE_SIZE_256K 18
73#define _PAGE_SIZE_1M 20
74#define _PAGE_SIZE_4M 22
75#define _PAGE_SIZE_16M 24
76#define _PAGE_SIZE_64M 26
77#define _PAGE_SIZE_256M 28
78#define _PAGE_SIZE_1G 30
79#define _PAGE_SIZE_4G 32
80
81#define __ACCESS_BITS _PAGE_ED | _PAGE_A | _PAGE_P | _PAGE_MA_WB
82#define __DIRTY_BITS_NO_ED _PAGE_A | _PAGE_P | _PAGE_D | _PAGE_MA_WB
83#define __DIRTY_BITS _PAGE_ED | __DIRTY_BITS_NO_ED
84
85/*
Robin Holt837cd0b2005-11-11 09:35:43 -060086 * How many pointers will a page table level hold expressed in shift
87 */
88#define PTRS_PER_PTD_SHIFT (PAGE_SHIFT-3)
89
90/*
91 * Definitions for fourth level:
92 */
93#define PTRS_PER_PTE (__IA64_UL(1) << (PTRS_PER_PTD_SHIFT))
94
95/*
96 * Definitions for third level:
97 *
98 * PMD_SHIFT determines the size of the area a third-level page table
99 * can map.
100 */
101#define PMD_SHIFT (PAGE_SHIFT + (PTRS_PER_PTD_SHIFT))
102#define PMD_SIZE (1UL << PMD_SHIFT)
103#define PMD_MASK (~(PMD_SIZE-1))
104#define PTRS_PER_PMD (1UL << (PTRS_PER_PTD_SHIFT))
105
106#ifdef CONFIG_PGTABLE_4
107/*
108 * Definitions for second level:
109 *
110 * PUD_SHIFT determines the size of the area a second-level page table
111 * can map.
112 */
113#define PUD_SHIFT (PMD_SHIFT + (PTRS_PER_PTD_SHIFT))
114#define PUD_SIZE (1UL << PUD_SHIFT)
115#define PUD_MASK (~(PUD_SIZE-1))
116#define PTRS_PER_PUD (1UL << (PTRS_PER_PTD_SHIFT))
117#endif
118
119/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120 * Definitions for first level:
121 *
122 * PGDIR_SHIFT determines what a first-level page table entry can map.
123 */
Robin Holt837cd0b2005-11-11 09:35:43 -0600124#ifdef CONFIG_PGTABLE_4
125#define PGDIR_SHIFT (PUD_SHIFT + (PTRS_PER_PTD_SHIFT))
126#else
127#define PGDIR_SHIFT (PMD_SHIFT + (PTRS_PER_PTD_SHIFT))
128#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129#define PGDIR_SIZE (__IA64_UL(1) << PGDIR_SHIFT)
130#define PGDIR_MASK (~(PGDIR_SIZE-1))
Robin Holt837cd0b2005-11-11 09:35:43 -0600131#define PTRS_PER_PGD_SHIFT PTRS_PER_PTD_SHIFT
132#define PTRS_PER_PGD (1UL << PTRS_PER_PGD_SHIFT)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133#define USER_PTRS_PER_PGD (5*PTRS_PER_PGD/8) /* regions 0-4 are user regions */
Hugh Dickinsd455a362005-04-19 13:29:23 -0700134#define FIRST_USER_ADDRESS 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
136/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 * All the normal masks have the "page accessed" bits on, as any time
138 * they are used, the page is accessed. They are cleared only by the
139 * page-out routines.
140 */
141#define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_A)
142#define PAGE_SHARED __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RW)
143#define PAGE_READONLY __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_R)
144#define PAGE_COPY __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_R)
145#define PAGE_COPY_EXEC __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX)
146#define PAGE_GATE __pgprot(__ACCESS_BITS | _PAGE_PL_0 | _PAGE_AR_X_RX)
147#define PAGE_KERNEL __pgprot(__DIRTY_BITS | _PAGE_PL_0 | _PAGE_AR_RWX)
148#define PAGE_KERNELRX __pgprot(__ACCESS_BITS | _PAGE_PL_0 | _PAGE_AR_RX)
Jes Sorensen0c72ea72009-02-25 10:38:52 -0600149#define PAGE_KERNEL_UC __pgprot(__DIRTY_BITS | _PAGE_PL_0 | _PAGE_AR_RWX | \
150 _PAGE_MA_UC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152# ifndef __ASSEMBLY__
153
Tim Schmielau8c65b4a2005-11-07 00:59:43 -0800154#include <linux/sched.h> /* for mm_struct */
Jiri Slaby1977f032007-10-18 23:40:25 -0700155#include <linux/bitops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156#include <asm/cacheflush.h>
157#include <asm/mmu_context.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
159/*
160 * Next come the mappings that determine how mmap() protection bits
161 * (PROT_EXEC, PROT_READ, PROT_WRITE, PROT_NONE) get implemented. The
162 * _P version gets used for a private shared memory segment, the _S
163 * version gets used for a shared memory segment with MAP_SHARED on.
164 * In a private shared memory segment, we do a copy-on-write if a task
165 * attempts to write to the page.
166 */
167 /* xwr */
168#define __P000 PAGE_NONE
169#define __P001 PAGE_READONLY
170#define __P010 PAGE_READONLY /* write to priv pg -> copy & make writable */
171#define __P011 PAGE_READONLY /* ditto */
172#define __P100 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_X_RX)
173#define __P101 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX)
174#define __P110 PAGE_COPY_EXEC
175#define __P111 PAGE_COPY_EXEC
176
177#define __S000 PAGE_NONE
178#define __S001 PAGE_READONLY
179#define __S010 PAGE_SHARED /* we don't have (and don't need) write-only */
180#define __S011 PAGE_SHARED
181#define __S100 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_X_RX)
182#define __S101 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX)
183#define __S110 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RWX)
184#define __S111 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RWX)
185
186#define pgd_ERROR(e) printk("%s:%d: bad pgd %016lx.\n", __FILE__, __LINE__, pgd_val(e))
Robin Holt837cd0b2005-11-11 09:35:43 -0600187#ifdef CONFIG_PGTABLE_4
188#define pud_ERROR(e) printk("%s:%d: bad pud %016lx.\n", __FILE__, __LINE__, pud_val(e))
189#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190#define pmd_ERROR(e) printk("%s:%d: bad pmd %016lx.\n", __FILE__, __LINE__, pmd_val(e))
191#define pte_ERROR(e) printk("%s:%d: bad pte %016lx.\n", __FILE__, __LINE__, pte_val(e))
192
193
194/*
195 * Some definitions to translate between mem_map, PTEs, and page addresses:
196 */
197
198
199/* Quick test to see if ADDR is a (potentially) valid physical address. */
200static inline long
201ia64_phys_addr_valid (unsigned long addr)
202{
203 return (addr & (local_cpu_data->unimpl_pa_mask)) == 0;
204}
205
206/*
207 * kern_addr_valid(ADDR) tests if ADDR is pointing to valid kernel
208 * memory. For the return value to be meaningful, ADDR must be >=
209 * PAGE_OFFSET. This operation can be relatively expensive (e.g.,
210 * require a hash-, or multi-level tree-lookup or something of that
211 * sort) but it guarantees to return TRUE only if accessing the page
212 * at that address does not cause an error. Note that there may be
213 * addresses for which kern_addr_valid() returns FALSE even though an
214 * access would not cause an error (e.g., this is typically true for
215 * memory mapped I/O regions.
216 *
217 * XXX Need to implement this for IA-64.
218 */
219#define kern_addr_valid(addr) (1)
220
221
222/*
223 * Now come the defines and routines to manage and access the three-level
224 * page table.
225 */
226
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227
Peter Chubb0a41e252005-08-16 19:54:00 -0700228#define VMALLOC_START (RGN_BASE(RGN_GATE) + 0x200000000UL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229#ifdef CONFIG_VIRTUAL_MEM_MAP
Peter Chubb0a41e252005-08-16 19:54:00 -0700230# define VMALLOC_END_INIT (RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 9)))
Tejun Heo126b3fc2009-10-02 13:28:55 +0900231extern unsigned long VMALLOC_END;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232#else
Christoph Lameteref229c52007-10-16 01:24:15 -0700233#if defined(CONFIG_SPARSEMEM) && defined(CONFIG_SPARSEMEM_VMEMMAP)
234/* SPARSEMEM_VMEMMAP uses half of vmalloc... */
235# define VMALLOC_END (RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 10)))
236# define vmemmap ((struct page *)VMALLOC_END)
237#else
Peter Chubb0a41e252005-08-16 19:54:00 -0700238# define VMALLOC_END (RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 9)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239#endif
Christoph Lameteref229c52007-10-16 01:24:15 -0700240#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
242/* fs/proc/kcore.c */
Peter Chubb0a41e252005-08-16 19:54:00 -0700243#define kc_vaddr_to_offset(v) ((v) - RGN_BASE(RGN_GATE))
244#define kc_offset_to_vaddr(o) ((o) + RGN_BASE(RGN_GATE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245
Robin Holt837cd0b2005-11-11 09:35:43 -0600246#define RGN_MAP_SHIFT (PGDIR_SHIFT + PTRS_PER_PGD_SHIFT - 3)
247#define RGN_MAP_LIMIT ((1UL << RGN_MAP_SHIFT) - PAGE_SIZE) /* per region addr limit */
248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249/*
250 * Conversion functions: convert page frame number (pfn) and a protection value to a page
251 * table entry (pte).
252 */
253#define pfn_pte(pfn, pgprot) \
254({ pte_t __pte; pte_val(__pte) = ((pfn) << PAGE_SHIFT) | pgprot_val(pgprot); __pte; })
255
256/* Extract pfn from pte. */
257#define pte_pfn(_pte) ((pte_val(_pte) & _PFN_MASK) >> PAGE_SHIFT)
258
259#define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
260
261/* This takes a physical page address that is used by the remapping functions */
262#define mk_pte_phys(physpage, pgprot) \
263({ pte_t __pte; pte_val(__pte) = physpage + pgprot_val(pgprot); __pte; })
264
265#define pte_modify(_pte, newprot) \
266 (__pte((pte_val(_pte) & ~_PAGE_CHG_MASK) | (pgprot_val(newprot) & _PAGE_CHG_MASK)))
267
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268#define pte_none(pte) (!pte_val(pte))
269#define pte_present(pte) (pte_val(pte) & (_PAGE_P | _PAGE_PROTNONE))
270#define pte_clear(mm,addr,pte) (pte_val(*(pte)) = 0UL)
271/* pte_page() returns the "struct page *" corresponding to the PTE: */
272#define pte_page(pte) virt_to_page(((pte_val(pte) & _PFN_MASK) + PAGE_OFFSET))
273
274#define pmd_none(pmd) (!pmd_val(pmd))
275#define pmd_bad(pmd) (!ia64_phys_addr_valid(pmd_val(pmd)))
276#define pmd_present(pmd) (pmd_val(pmd) != 0UL)
277#define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0UL)
Dave McCracken46a82b22006-09-25 23:31:48 -0700278#define pmd_page_vaddr(pmd) ((unsigned long) __va(pmd_val(pmd) & _PFN_MASK))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279#define pmd_page(pmd) virt_to_page((pmd_val(pmd) + PAGE_OFFSET))
280
281#define pud_none(pud) (!pud_val(pud))
282#define pud_bad(pud) (!ia64_phys_addr_valid(pud_val(pud)))
283#define pud_present(pud) (pud_val(pud) != 0UL)
284#define pud_clear(pudp) (pud_val(*(pudp)) = 0UL)
Dave McCracken46a82b22006-09-25 23:31:48 -0700285#define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & _PFN_MASK))
286#define pud_page(pud) virt_to_page((pud_val(pud) + PAGE_OFFSET))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Robin Holt837cd0b2005-11-11 09:35:43 -0600288#ifdef CONFIG_PGTABLE_4
289#define pgd_none(pgd) (!pgd_val(pgd))
290#define pgd_bad(pgd) (!ia64_phys_addr_valid(pgd_val(pgd)))
291#define pgd_present(pgd) (pgd_val(pgd) != 0UL)
292#define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0UL)
Dave McCracken46a82b22006-09-25 23:31:48 -0700293#define pgd_page_vaddr(pgd) ((unsigned long) __va(pgd_val(pgd) & _PFN_MASK))
294#define pgd_page(pgd) virt_to_page((pgd_val(pgd) + PAGE_OFFSET))
Robin Holt837cd0b2005-11-11 09:35:43 -0600295#endif
296
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297/*
298 * The following have defined behavior only work if pte_present() is true.
299 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300#define pte_write(pte) ((unsigned) (((pte_val(pte) & _PAGE_AR_MASK) >> _PAGE_AR_SHIFT) - 2) <= 4)
301#define pte_exec(pte) ((pte_val(pte) & _PAGE_AR_RX) != 0)
302#define pte_dirty(pte) ((pte_val(pte) & _PAGE_D) != 0)
303#define pte_young(pte) ((pte_val(pte) & _PAGE_A) != 0)
304#define pte_file(pte) ((pte_val(pte) & _PAGE_FILE) != 0)
Nick Piggin7e675132008-04-28 02:13:00 -0700305#define pte_special(pte) 0
306
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307/*
308 * Note: we convert AR_RWX to AR_RX and AR_RW to AR_R by clearing the 2nd bit in the
309 * access rights:
310 */
311#define pte_wrprotect(pte) (__pte(pte_val(pte) & ~_PAGE_AR_RW))
312#define pte_mkwrite(pte) (__pte(pte_val(pte) | _PAGE_AR_RW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313#define pte_mkold(pte) (__pte(pte_val(pte) & ~_PAGE_A))
314#define pte_mkyoung(pte) (__pte(pte_val(pte) | _PAGE_A))
315#define pte_mkclean(pte) (__pte(pte_val(pte) & ~_PAGE_D))
316#define pte_mkdirty(pte) (__pte(pte_val(pte) | _PAGE_D))
Zhang, Yanmin8f860592006-03-22 00:08:50 -0800317#define pte_mkhuge(pte) (__pte(pte_val(pte)))
Nick Piggin7e675132008-04-28 02:13:00 -0700318#define pte_mkspecial(pte) (pte)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319
320/*
KAMEZAWA Hiroyuki954ffcb2007-10-16 01:25:44 -0700321 * Because ia64's Icache and Dcache is not coherent (on a cpu), we need to
322 * sync icache and dcache when we insert *new* executable page.
323 * __ia64_sync_icache_dcache() check Pg_arch_1 bit and flush icache
324 * if necessary.
325 *
326 * set_pte() is also called by the kernel, but we can expect that the kernel
327 * flushes icache explicitly if necessary.
328 */
329#define pte_present_exec_user(pte)\
330 ((pte_val(pte) & (_PAGE_P | _PAGE_PL_MASK | _PAGE_AR_RX)) == \
331 (_PAGE_P | _PAGE_PL_3 | _PAGE_AR_RX))
332
333extern void __ia64_sync_icache_dcache(pte_t pteval);
334static inline void set_pte(pte_t *ptep, pte_t pteval)
335{
336 /* page is present && page is user && page is executable
337 * && (page swapin or new page or page migraton
338 * || copy_on_write with page copying.)
339 */
340 if (pte_present_exec_user(pteval) &&
341 (!pte_present(*ptep) ||
342 pte_pfn(*ptep) != pte_pfn(pteval)))
343 /* load_module() calles flush_icache_range() explicitly*/
344 __ia64_sync_icache_dcache(pteval);
345 *ptep = pteval;
346}
347
348#define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
349
350/*
Bjorn Helgaas32e62c62006-05-05 17:19:50 -0600351 * Make page protection values cacheable, uncacheable, or write-
352 * combining. Note that "protection" is really a misnomer here as the
353 * protection value contains the memory attribute bits, dirty bits, and
354 * various other bits as well.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 */
Bjorn Helgaas32e62c62006-05-05 17:19:50 -0600356#define pgprot_cacheable(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_WB)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357#define pgprot_noncached(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_UC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358#define pgprot_writecombine(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_WC)
359
Bjorn Helgaas32e62c62006-05-05 17:19:50 -0600360struct file;
361extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
362 unsigned long size, pgprot_t vma_prot);
363#define __HAVE_PHYS_MEM_ACCESS_PROT
364
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365static inline unsigned long
366pgd_index (unsigned long address)
367{
368 unsigned long region = address >> 61;
369 unsigned long l1index = (address >> PGDIR_SHIFT) & ((PTRS_PER_PGD >> 3) - 1);
370
371 return (region << (PAGE_SHIFT - 6)) | l1index;
372}
373
374/* The offset in the 1-level directory is given by the 3 region bits
375 (61..63) and the level-1 bits. */
376static inline pgd_t*
KOSAKI Motohiroe4b05d42008-04-09 12:26:10 +0900377pgd_offset (const struct mm_struct *mm, unsigned long address)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378{
379 return mm->pgd + pgd_index(address);
380}
381
382/* In the kernel's mapped region we completely ignore the region number
383 (since we know it's in region number 5). */
384#define pgd_offset_k(addr) \
385 (init_mm.pgd + (((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1)))
386
387/* Look up a pgd entry in the gate area. On IA-64, the gate-area
388 resides in the kernel-mapped segment, hence we use pgd_offset_k()
389 here. */
390#define pgd_offset_gate(mm, addr) pgd_offset_k(addr)
391
Robin Holt837cd0b2005-11-11 09:35:43 -0600392#ifdef CONFIG_PGTABLE_4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393/* Find an entry in the second-level page table.. */
Robin Holt837cd0b2005-11-11 09:35:43 -0600394#define pud_offset(dir,addr) \
Dave McCracken46a82b22006-09-25 23:31:48 -0700395 ((pud_t *) pgd_page_vaddr(*(dir)) + (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1)))
Robin Holt837cd0b2005-11-11 09:35:43 -0600396#endif
397
398/* Find an entry in the third-level page table.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399#define pmd_offset(dir,addr) \
Dave McCracken46a82b22006-09-25 23:31:48 -0700400 ((pmd_t *) pud_page_vaddr(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
402/*
403 * Find an entry in the third-level page table. This looks more complicated than it
404 * should be because some platforms place page tables in high memory.
405 */
406#define pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
Dave McCracken46a82b22006-09-25 23:31:48 -0700407#define pte_offset_kernel(dir,addr) ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408#define pte_offset_map(dir,addr) pte_offset_kernel(dir, addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409#define pte_unmap(pte) do { } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410
411/* atomic versions of the some PTE manipulations: */
412
413static inline int
414ptep_test_and_clear_young (struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
415{
416#ifdef CONFIG_SMP
417 if (!pte_young(*ptep))
418 return 0;
419 return test_and_clear_bit(_PAGE_A_BIT, ptep);
420#else
421 pte_t pte = *ptep;
422 if (!pte_young(pte))
423 return 0;
424 set_pte_at(vma->vm_mm, addr, ptep, pte_mkold(pte));
425 return 1;
426#endif
427}
428
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429static inline pte_t
430ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
431{
432#ifdef CONFIG_SMP
433 return __pte(xchg((long *) ptep, 0));
434#else
435 pte_t pte = *ptep;
436 pte_clear(mm, addr, ptep);
437 return pte;
438#endif
439}
440
441static inline void
442ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
443{
444#ifdef CONFIG_SMP
445 unsigned long new, old;
446
447 do {
448 old = pte_val(*ptep);
449 new = pte_val(pte_wrprotect(__pte (old)));
450 } while (cmpxchg((unsigned long *) ptep, old, new) != old);
451#else
452 pte_t old_pte = *ptep;
453 set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
454#endif
455}
456
457static inline int
458pte_same (pte_t a, pte_t b)
459{
460 return pte_val(a) == pte_val(b);
461}
462
Russell King4b3073e2009-12-18 16:40:18 +0000463#define update_mmu_cache(vma, address, ptep) do { } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
465extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
466extern void paging_init (void);
467
468/*
469 * Note: The macros below rely on the fact that MAX_SWAPFILES_SHIFT <= number of
470 * bits in the swap-type field of the swap pte. It would be nice to
471 * enforce that, but we can't easily include <linux/swap.h> here.
472 * (Of course, better still would be to define MAX_SWAPFILES_SHIFT here...).
473 *
474 * Format of swap pte:
475 * bit 0 : present bit (must be zero)
476 * bit 1 : _PAGE_FILE (must be zero)
477 * bits 2- 8: swap-type
478 * bits 9-62: swap offset
479 * bit 63 : _PAGE_PROTNONE bit
480 *
481 * Format of file pte:
482 * bit 0 : present bit (must be zero)
483 * bit 1 : _PAGE_FILE (must be one)
484 * bits 2-62: file_offset/PAGE_SIZE
485 * bit 63 : _PAGE_PROTNONE bit
486 */
487#define __swp_type(entry) (((entry).val >> 2) & 0x7f)
488#define __swp_offset(entry) (((entry).val << 1) >> 10)
489#define __swp_entry(type,offset) ((swp_entry_t) { ((type) << 2) | ((long) (offset) << 9) })
490#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
491#define __swp_entry_to_pte(x) ((pte_t) { (x).val })
492
493#define PTE_FILE_MAX_BITS 61
494#define pte_to_pgoff(pte) ((pte_val(pte) << 1) >> 3)
495#define pgoff_to_pte(off) ((pte_t) { ((off) << 2) | _PAGE_FILE })
496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
498 remap_pfn_range(vma, vaddr, pfn, size, prot)
499
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500/*
501 * ZERO_PAGE is a global shared page that is always zero: used
502 * for zero-mapped memory areas etc..
503 */
504extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
505extern struct page *zero_page_memmap_ptr;
506#define ZERO_PAGE(vaddr) (zero_page_memmap_ptr)
507
508/* We provide our own get_unmapped_area to cope with VA holes for userland */
509#define HAVE_ARCH_UNMAPPED_AREA
510
511#ifdef CONFIG_HUGETLB_PAGE
512#define HUGETLB_PGDIR_SHIFT (HPAGE_SHIFT + 2*(PAGE_SHIFT-3))
513#define HUGETLB_PGDIR_SIZE (__IA64_UL(1) << HUGETLB_PGDIR_SHIFT)
514#define HUGETLB_PGDIR_MASK (~(HUGETLB_PGDIR_SIZE-1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515#endif
516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
518#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
519/*
520 * Update PTEP with ENTRY, which is guaranteed to be a less
521 * restrictive PTE. That is, ENTRY may have the ACCESSED, DIRTY, and
522 * WRITABLE bits turned on, when the value at PTEP did not. The
523 * WRITABLE bit may only be turned if SAFELY_WRITABLE is TRUE.
524 *
525 * SAFELY_WRITABLE is TRUE if we can update the value at PTEP without
526 * having to worry about races. On SMP machines, there are only two
527 * cases where this is true:
528 *
529 * (1) *PTEP has the PRESENT bit turned OFF
530 * (2) ENTRY has the DIRTY bit turned ON
531 *
532 * On ia64, we could implement this routine with a cmpxchg()-loop
533 * which ORs in the _PAGE_A/_PAGE_D bit if they're set in ENTRY.
534 * However, like on x86, we can get a more streamlined version by
535 * observing that it is OK to drop ACCESSED bit updates when
536 * SAFELY_WRITABLE is FALSE. Besides being rare, all that would do is
537 * result in an extra Access-bit fault, which would then turn on the
538 * ACCESSED bit in the low-level fault handler (iaccess_bit or
539 * daccess_bit in ivt.S).
540 */
541#ifdef CONFIG_SMP
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700542# define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __safely_writable) \
543({ \
544 int __changed = !pte_same(*(__ptep), __entry); \
545 if (__changed && __safely_writable) { \
546 set_pte(__ptep, __entry); \
547 flush_tlb_page(__vma, __addr); \
548 } \
549 __changed; \
550})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551#else
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700552# define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __safely_writable) \
553({ \
554 int __changed = !pte_same(*(__ptep), __entry); \
Martin Schwidefskyf0e47c22007-07-17 04:03:03 -0700555 if (__changed) { \
556 set_pte_at((__vma)->vm_mm, (__addr), __ptep, __entry); \
557 flush_tlb_page(__vma, __addr); \
558 } \
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700559 __changed; \
560})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561#endif
562
563# ifdef CONFIG_VIRTUAL_MEM_MAP
564 /* arch mem_map init routine is needed due to holes in a virtual mem_map */
565# define __HAVE_ARCH_MEMMAP_INIT
566 extern void memmap_init (unsigned long size, int nid, unsigned long zone,
567 unsigned long start_pfn);
568# endif /* CONFIG_VIRTUAL_MEM_MAP */
569# endif /* !__ASSEMBLY__ */
570
571/*
572 * Identity-mapped regions use a large page size. We'll call such large pages
573 * "granules". If you can think of a better name that's unambiguous, let me
574 * know...
575 */
576#if defined(CONFIG_IA64_GRANULE_64MB)
577# define IA64_GRANULE_SHIFT _PAGE_SIZE_64M
578#elif defined(CONFIG_IA64_GRANULE_16MB)
579# define IA64_GRANULE_SHIFT _PAGE_SIZE_16M
580#endif
581#define IA64_GRANULE_SIZE (1 << IA64_GRANULE_SHIFT)
582/*
583 * log2() of the page size we use to map the kernel image (IA64_TR_KERNEL):
584 */
585#define KERNEL_TR_PAGE_SHIFT _PAGE_SIZE_64M
586#define KERNEL_TR_PAGE_SIZE (1 << KERNEL_TR_PAGE_SHIFT)
587
588/*
589 * No page table caches to initialise
590 */
591#define pgtable_cache_init() do { } while (0)
592
593/* These tell get_user_pages() that the first gate page is accessible from user-level. */
594#define FIXADDR_USER_START GATE_ADDR
David Mosberger-Tangad597bd2005-06-08 10:45:00 -0700595#ifdef HAVE_BUGGY_SEGREL
596# define FIXADDR_USER_END (GATE_ADDR + 2*PAGE_SIZE)
597#else
598# define FIXADDR_USER_END (GATE_ADDR + 2*PERCPU_PAGE_SIZE)
599#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
603#define __HAVE_ARCH_PTEP_SET_WRPROTECT
604#define __HAVE_ARCH_PTE_SAME
605#define __HAVE_ARCH_PGD_OFFSET_GATE
KAMEZAWA Hiroyuki954ffcb2007-10-16 01:25:44 -0700606
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
Robin Holt837cd0b2005-11-11 09:35:43 -0600608#ifndef CONFIG_PGTABLE_4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609#include <asm-generic/pgtable-nopud.h>
Robin Holt837cd0b2005-11-11 09:35:43 -0600610#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611#include <asm-generic/pgtable.h>
612
613#endif /* _ASM_IA64_PGTABLE_H */