blob: 787b0a91d2553b951bcf9250d08760cff0a28e14 [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001/* SPDX-License-Identifier: GPL-2.0 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002#ifndef _ASM_IA64_PGTABLE_H
3#define _ASM_IA64_PGTABLE_H
4
5/*
6 * This file contains the functions and defines necessary to modify and use
7 * the IA-64 page table tree.
8 *
9 * This hopefully works with any (fixed) IA-64 page-size, as defined
10 * in <asm/page.h>.
11 *
David Mosberger-Tangad597bd2005-06-08 10:45:00 -070012 * Copyright (C) 1998-2005 Hewlett-Packard Co
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 * David Mosberger-Tang <davidm@hpl.hp.com>
14 */
15
Linus Torvalds1da177e2005-04-16 15:20:36 -070016
17#include <asm/mman.h>
18#include <asm/page.h>
19#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#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
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#define _PFN_MASK _PAGE_PPN_MASK
62/* Mask of bits which may be changed by pte_modify(); the odd bits are there for _PAGE_PROTNONE */
63#define _PAGE_CHG_MASK (_PAGE_P | _PAGE_PROTNONE | _PAGE_PL_MASK | _PAGE_AR_MASK | _PAGE_ED)
64
65#define _PAGE_SIZE_4K 12
66#define _PAGE_SIZE_8K 13
67#define _PAGE_SIZE_16K 14
68#define _PAGE_SIZE_64K 16
69#define _PAGE_SIZE_256K 18
70#define _PAGE_SIZE_1M 20
71#define _PAGE_SIZE_4M 22
72#define _PAGE_SIZE_16M 24
73#define _PAGE_SIZE_64M 26
74#define _PAGE_SIZE_256M 28
75#define _PAGE_SIZE_1G 30
76#define _PAGE_SIZE_4G 32
77
78#define __ACCESS_BITS _PAGE_ED | _PAGE_A | _PAGE_P | _PAGE_MA_WB
79#define __DIRTY_BITS_NO_ED _PAGE_A | _PAGE_P | _PAGE_D | _PAGE_MA_WB
80#define __DIRTY_BITS _PAGE_ED | __DIRTY_BITS_NO_ED
81
82/*
Robin Holt837cd0b2005-11-11 09:35:43 -060083 * How many pointers will a page table level hold expressed in shift
84 */
85#define PTRS_PER_PTD_SHIFT (PAGE_SHIFT-3)
86
87/*
88 * Definitions for fourth level:
89 */
90#define PTRS_PER_PTE (__IA64_UL(1) << (PTRS_PER_PTD_SHIFT))
91
92/*
93 * Definitions for third level:
94 *
95 * PMD_SHIFT determines the size of the area a third-level page table
96 * can map.
97 */
98#define PMD_SHIFT (PAGE_SHIFT + (PTRS_PER_PTD_SHIFT))
99#define PMD_SIZE (1UL << PMD_SHIFT)
100#define PMD_MASK (~(PMD_SIZE-1))
101#define PTRS_PER_PMD (1UL << (PTRS_PER_PTD_SHIFT))
102
Kirill A. Shutemov4d66bcc2015-04-14 15:45:45 -0700103#if CONFIG_PGTABLE_LEVELS == 4
Robin Holt837cd0b2005-11-11 09:35:43 -0600104/*
105 * Definitions for second level:
106 *
107 * PUD_SHIFT determines the size of the area a second-level page table
108 * can map.
109 */
110#define PUD_SHIFT (PMD_SHIFT + (PTRS_PER_PTD_SHIFT))
111#define PUD_SIZE (1UL << PUD_SHIFT)
112#define PUD_MASK (~(PUD_SIZE-1))
113#define PTRS_PER_PUD (1UL << (PTRS_PER_PTD_SHIFT))
114#endif
115
116/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 * Definitions for first level:
118 *
119 * PGDIR_SHIFT determines what a first-level page table entry can map.
120 */
Kirill A. Shutemov4d66bcc2015-04-14 15:45:45 -0700121#if CONFIG_PGTABLE_LEVELS == 4
Robin Holt837cd0b2005-11-11 09:35:43 -0600122#define PGDIR_SHIFT (PUD_SHIFT + (PTRS_PER_PTD_SHIFT))
123#else
124#define PGDIR_SHIFT (PMD_SHIFT + (PTRS_PER_PTD_SHIFT))
125#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126#define PGDIR_SIZE (__IA64_UL(1) << PGDIR_SHIFT)
127#define PGDIR_MASK (~(PGDIR_SIZE-1))
Robin Holt837cd0b2005-11-11 09:35:43 -0600128#define PTRS_PER_PGD_SHIFT PTRS_PER_PTD_SHIFT
129#define PTRS_PER_PGD (1UL << PTRS_PER_PGD_SHIFT)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130#define USER_PTRS_PER_PGD (5*PTRS_PER_PGD/8) /* regions 0-4 are user regions */
Kirill A. Shutemovd016bf72015-02-11 15:26:41 -0800131#define FIRST_USER_ADDRESS 0UL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
133/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 * All the normal masks have the "page accessed" bits on, as any time
135 * they are used, the page is accessed. They are cleared only by the
136 * page-out routines.
137 */
138#define PAGE_NONE __pgprot(_PAGE_PROTNONE | _PAGE_A)
139#define PAGE_SHARED __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RW)
140#define PAGE_READONLY __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_R)
141#define PAGE_COPY __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_R)
142#define PAGE_COPY_EXEC __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX)
143#define PAGE_GATE __pgprot(__ACCESS_BITS | _PAGE_PL_0 | _PAGE_AR_X_RX)
144#define PAGE_KERNEL __pgprot(__DIRTY_BITS | _PAGE_PL_0 | _PAGE_AR_RWX)
145#define PAGE_KERNELRX __pgprot(__ACCESS_BITS | _PAGE_PL_0 | _PAGE_AR_RX)
Jes Sorensen0c72ea72009-02-25 10:38:52 -0600146#define PAGE_KERNEL_UC __pgprot(__DIRTY_BITS | _PAGE_PL_0 | _PAGE_AR_RWX | \
147 _PAGE_MA_UC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
149# ifndef __ASSEMBLY__
150
Ingo Molnar589ee622017-02-04 00:16:44 +0100151#include <linux/sched/mm.h> /* for mm_struct */
Jiri Slaby1977f032007-10-18 23:40:25 -0700152#include <linux/bitops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153#include <asm/cacheflush.h>
154#include <asm/mmu_context.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
156/*
157 * Next come the mappings that determine how mmap() protection bits
158 * (PROT_EXEC, PROT_READ, PROT_WRITE, PROT_NONE) get implemented. The
159 * _P version gets used for a private shared memory segment, the _S
160 * version gets used for a shared memory segment with MAP_SHARED on.
161 * In a private shared memory segment, we do a copy-on-write if a task
162 * attempts to write to the page.
163 */
164 /* xwr */
165#define __P000 PAGE_NONE
166#define __P001 PAGE_READONLY
167#define __P010 PAGE_READONLY /* write to priv pg -> copy & make writable */
168#define __P011 PAGE_READONLY /* ditto */
169#define __P100 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_X_RX)
170#define __P101 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX)
171#define __P110 PAGE_COPY_EXEC
172#define __P111 PAGE_COPY_EXEC
173
174#define __S000 PAGE_NONE
175#define __S001 PAGE_READONLY
176#define __S010 PAGE_SHARED /* we don't have (and don't need) write-only */
177#define __S011 PAGE_SHARED
178#define __S100 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_X_RX)
179#define __S101 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RX)
180#define __S110 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RWX)
181#define __S111 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RWX)
182
183#define pgd_ERROR(e) printk("%s:%d: bad pgd %016lx.\n", __FILE__, __LINE__, pgd_val(e))
Kirill A. Shutemov4d66bcc2015-04-14 15:45:45 -0700184#if CONFIG_PGTABLE_LEVELS == 4
Robin Holt837cd0b2005-11-11 09:35:43 -0600185#define pud_ERROR(e) printk("%s:%d: bad pud %016lx.\n", __FILE__, __LINE__, pud_val(e))
186#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187#define pmd_ERROR(e) printk("%s:%d: bad pmd %016lx.\n", __FILE__, __LINE__, pmd_val(e))
188#define pte_ERROR(e) printk("%s:%d: bad pte %016lx.\n", __FILE__, __LINE__, pte_val(e))
189
190
191/*
192 * Some definitions to translate between mem_map, PTEs, and page addresses:
193 */
194
195
196/* Quick test to see if ADDR is a (potentially) valid physical address. */
197static inline long
198ia64_phys_addr_valid (unsigned long addr)
199{
200 return (addr & (local_cpu_data->unimpl_pa_mask)) == 0;
201}
202
203/*
204 * kern_addr_valid(ADDR) tests if ADDR is pointing to valid kernel
205 * memory. For the return value to be meaningful, ADDR must be >=
206 * PAGE_OFFSET. This operation can be relatively expensive (e.g.,
207 * require a hash-, or multi-level tree-lookup or something of that
208 * sort) but it guarantees to return TRUE only if accessing the page
209 * at that address does not cause an error. Note that there may be
210 * addresses for which kern_addr_valid() returns FALSE even though an
211 * access would not cause an error (e.g., this is typically true for
212 * memory mapped I/O regions.
213 *
214 * XXX Need to implement this for IA-64.
215 */
216#define kern_addr_valid(addr) (1)
217
218
219/*
220 * Now come the defines and routines to manage and access the three-level
221 * page table.
222 */
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Peter Chubb0a41e252005-08-16 19:54:00 -0700225#define VMALLOC_START (RGN_BASE(RGN_GATE) + 0x200000000UL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226#ifdef CONFIG_VIRTUAL_MEM_MAP
Peter Chubb0a41e252005-08-16 19:54:00 -0700227# define VMALLOC_END_INIT (RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 9)))
Tejun Heo126b3fc2009-10-02 13:28:55 +0900228extern unsigned long VMALLOC_END;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229#else
Christoph Lameteref229c52007-10-16 01:24:15 -0700230#if defined(CONFIG_SPARSEMEM) && defined(CONFIG_SPARSEMEM_VMEMMAP)
231/* SPARSEMEM_VMEMMAP uses half of vmalloc... */
232# define VMALLOC_END (RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 10)))
233# define vmemmap ((struct page *)VMALLOC_END)
234#else
Peter Chubb0a41e252005-08-16 19:54:00 -0700235# define VMALLOC_END (RGN_BASE(RGN_GATE) + (1UL << (4*PAGE_SHIFT - 9)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236#endif
Christoph Lameteref229c52007-10-16 01:24:15 -0700237#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
239/* fs/proc/kcore.c */
Peter Chubb0a41e252005-08-16 19:54:00 -0700240#define kc_vaddr_to_offset(v) ((v) - RGN_BASE(RGN_GATE))
241#define kc_offset_to_vaddr(o) ((o) + RGN_BASE(RGN_GATE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
Robin Holt837cd0b2005-11-11 09:35:43 -0600243#define RGN_MAP_SHIFT (PGDIR_SHIFT + PTRS_PER_PGD_SHIFT - 3)
244#define RGN_MAP_LIMIT ((1UL << RGN_MAP_SHIFT) - PAGE_SIZE) /* per region addr limit */
245
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246/*
247 * Conversion functions: convert page frame number (pfn) and a protection value to a page
248 * table entry (pte).
249 */
250#define pfn_pte(pfn, pgprot) \
251({ pte_t __pte; pte_val(__pte) = ((pfn) << PAGE_SHIFT) | pgprot_val(pgprot); __pte; })
252
253/* Extract pfn from pte. */
254#define pte_pfn(_pte) ((pte_val(_pte) & _PFN_MASK) >> PAGE_SHIFT)
255
256#define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
257
258/* This takes a physical page address that is used by the remapping functions */
259#define mk_pte_phys(physpage, pgprot) \
260({ pte_t __pte; pte_val(__pte) = physpage + pgprot_val(pgprot); __pte; })
261
262#define pte_modify(_pte, newprot) \
263 (__pte((pte_val(_pte) & ~_PAGE_CHG_MASK) | (pgprot_val(newprot) & _PAGE_CHG_MASK)))
264
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265#define pte_none(pte) (!pte_val(pte))
266#define pte_present(pte) (pte_val(pte) & (_PAGE_P | _PAGE_PROTNONE))
267#define pte_clear(mm,addr,pte) (pte_val(*(pte)) = 0UL)
268/* pte_page() returns the "struct page *" corresponding to the PTE: */
269#define pte_page(pte) virt_to_page(((pte_val(pte) & _PFN_MASK) + PAGE_OFFSET))
270
271#define pmd_none(pmd) (!pmd_val(pmd))
272#define pmd_bad(pmd) (!ia64_phys_addr_valid(pmd_val(pmd)))
273#define pmd_present(pmd) (pmd_val(pmd) != 0UL)
274#define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0UL)
Dave McCracken46a82b22006-09-25 23:31:48 -0700275#define pmd_page_vaddr(pmd) ((unsigned long) __va(pmd_val(pmd) & _PFN_MASK))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276#define pmd_page(pmd) virt_to_page((pmd_val(pmd) + PAGE_OFFSET))
277
278#define pud_none(pud) (!pud_val(pud))
279#define pud_bad(pud) (!ia64_phys_addr_valid(pud_val(pud)))
280#define pud_present(pud) (pud_val(pud) != 0UL)
281#define pud_clear(pudp) (pud_val(*(pudp)) = 0UL)
Dave McCracken46a82b22006-09-25 23:31:48 -0700282#define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & _PFN_MASK))
283#define pud_page(pud) virt_to_page((pud_val(pud) + PAGE_OFFSET))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
Kirill A. Shutemov4d66bcc2015-04-14 15:45:45 -0700285#if CONFIG_PGTABLE_LEVELS == 4
Mike Rapoportc03ab9e2020-06-04 16:46:31 -0700286#define p4d_none(p4d) (!p4d_val(p4d))
287#define p4d_bad(p4d) (!ia64_phys_addr_valid(p4d_val(p4d)))
288#define p4d_present(p4d) (p4d_val(p4d) != 0UL)
289#define p4d_clear(p4dp) (p4d_val(*(p4dp)) = 0UL)
290#define p4d_page_vaddr(p4d) ((unsigned long) __va(p4d_val(p4d) & _PFN_MASK))
291#define p4d_page(p4d) virt_to_page((p4d_val(p4d) + PAGE_OFFSET))
Robin Holt837cd0b2005-11-11 09:35:43 -0600292#endif
293
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294/*
295 * The following have defined behavior only work if pte_present() is true.
296 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297#define pte_write(pte) ((unsigned) (((pte_val(pte) & _PAGE_AR_MASK) >> _PAGE_AR_SHIFT) - 2) <= 4)
298#define pte_exec(pte) ((pte_val(pte) & _PAGE_AR_RX) != 0)
299#define pte_dirty(pte) ((pte_val(pte) & _PAGE_D) != 0)
300#define pte_young(pte) ((pte_val(pte) & _PAGE_A) != 0)
Nick Piggin7e675132008-04-28 02:13:00 -0700301
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302/*
303 * Note: we convert AR_RWX to AR_RX and AR_RW to AR_R by clearing the 2nd bit in the
304 * access rights:
305 */
306#define pte_wrprotect(pte) (__pte(pte_val(pte) & ~_PAGE_AR_RW))
307#define pte_mkwrite(pte) (__pte(pte_val(pte) | _PAGE_AR_RW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308#define pte_mkold(pte) (__pte(pte_val(pte) & ~_PAGE_A))
309#define pte_mkyoung(pte) (__pte(pte_val(pte) | _PAGE_A))
310#define pte_mkclean(pte) (__pte(pte_val(pte) & ~_PAGE_D))
311#define pte_mkdirty(pte) (__pte(pte_val(pte) | _PAGE_D))
Zhang, Yanmin8f860592006-03-22 00:08:50 -0800312#define pte_mkhuge(pte) (__pte(pte_val(pte)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
314/*
KAMEZAWA Hiroyuki954ffcb2007-10-16 01:25:44 -0700315 * Because ia64's Icache and Dcache is not coherent (on a cpu), we need to
316 * sync icache and dcache when we insert *new* executable page.
317 * __ia64_sync_icache_dcache() check Pg_arch_1 bit and flush icache
318 * if necessary.
319 *
320 * set_pte() is also called by the kernel, but we can expect that the kernel
321 * flushes icache explicitly if necessary.
322 */
323#define pte_present_exec_user(pte)\
324 ((pte_val(pte) & (_PAGE_P | _PAGE_PL_MASK | _PAGE_AR_RX)) == \
325 (_PAGE_P | _PAGE_PL_3 | _PAGE_AR_RX))
326
327extern void __ia64_sync_icache_dcache(pte_t pteval);
328static inline void set_pte(pte_t *ptep, pte_t pteval)
329{
330 /* page is present && page is user && page is executable
331 * && (page swapin or new page or page migraton
332 * || copy_on_write with page copying.)
333 */
334 if (pte_present_exec_user(pteval) &&
335 (!pte_present(*ptep) ||
336 pte_pfn(*ptep) != pte_pfn(pteval)))
337 /* load_module() calles flush_icache_range() explicitly*/
338 __ia64_sync_icache_dcache(pteval);
339 *ptep = pteval;
340}
341
342#define set_pte_at(mm,addr,ptep,pteval) set_pte(ptep,pteval)
343
344/*
Bjorn Helgaas32e62c62006-05-05 17:19:50 -0600345 * Make page protection values cacheable, uncacheable, or write-
346 * combining. Note that "protection" is really a misnomer here as the
347 * protection value contains the memory attribute bits, dirty bits, and
348 * various other bits as well.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 */
Bjorn Helgaas32e62c62006-05-05 17:19:50 -0600350#define pgprot_cacheable(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_WB)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351#define pgprot_noncached(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_UC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352#define pgprot_writecombine(prot) __pgprot((pgprot_val(prot) & ~_PAGE_MA_MASK) | _PAGE_MA_WC)
353
Bjorn Helgaas32e62c62006-05-05 17:19:50 -0600354struct file;
355extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
356 unsigned long size, pgprot_t vma_prot);
357#define __HAVE_PHYS_MEM_ACCESS_PROT
358
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359static inline unsigned long
360pgd_index (unsigned long address)
361{
362 unsigned long region = address >> 61;
363 unsigned long l1index = (address >> PGDIR_SHIFT) & ((PTRS_PER_PGD >> 3) - 1);
364
365 return (region << (PAGE_SHIFT - 6)) | l1index;
366}
367
368/* The offset in the 1-level directory is given by the 3 region bits
369 (61..63) and the level-1 bits. */
370static inline pgd_t*
KOSAKI Motohiroe4b05d42008-04-09 12:26:10 +0900371pgd_offset (const struct mm_struct *mm, unsigned long address)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372{
373 return mm->pgd + pgd_index(address);
374}
375
376/* In the kernel's mapped region we completely ignore the region number
377 (since we know it's in region number 5). */
378#define pgd_offset_k(addr) \
379 (init_mm.pgd + (((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1)))
380
381/* Look up a pgd entry in the gate area. On IA-64, the gate-area
382 resides in the kernel-mapped segment, hence we use pgd_offset_k()
383 here. */
384#define pgd_offset_gate(mm, addr) pgd_offset_k(addr)
385
Kirill A. Shutemov4d66bcc2015-04-14 15:45:45 -0700386#if CONFIG_PGTABLE_LEVELS == 4
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387/* Find an entry in the second-level page table.. */
Robin Holt837cd0b2005-11-11 09:35:43 -0600388#define pud_offset(dir,addr) \
Mike Rapoportc03ab9e2020-06-04 16:46:31 -0700389 ((pud_t *) p4d_page_vaddr(*(dir)) + (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1)))
Robin Holt837cd0b2005-11-11 09:35:43 -0600390#endif
391
392/* Find an entry in the third-level page table.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393#define pmd_offset(dir,addr) \
Dave McCracken46a82b22006-09-25 23:31:48 -0700394 ((pmd_t *) pud_page_vaddr(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
396/*
397 * Find an entry in the third-level page table. This looks more complicated than it
398 * should be because some platforms place page tables in high memory.
399 */
400#define pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
Dave McCracken46a82b22006-09-25 23:31:48 -0700401#define pte_offset_kernel(dir,addr) ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402#define pte_offset_map(dir,addr) pte_offset_kernel(dir, addr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403#define pte_unmap(pte) do { } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
405/* atomic versions of the some PTE manipulations: */
406
407static inline int
408ptep_test_and_clear_young (struct vm_area_struct *vma, unsigned long addr, pte_t *ptep)
409{
410#ifdef CONFIG_SMP
411 if (!pte_young(*ptep))
412 return 0;
413 return test_and_clear_bit(_PAGE_A_BIT, ptep);
414#else
415 pte_t pte = *ptep;
416 if (!pte_young(pte))
417 return 0;
418 set_pte_at(vma->vm_mm, addr, ptep, pte_mkold(pte));
419 return 1;
420#endif
421}
422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423static inline pte_t
424ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
425{
426#ifdef CONFIG_SMP
427 return __pte(xchg((long *) ptep, 0));
428#else
429 pte_t pte = *ptep;
430 pte_clear(mm, addr, ptep);
431 return pte;
432#endif
433}
434
435static inline void
436ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
437{
438#ifdef CONFIG_SMP
439 unsigned long new, old;
440
441 do {
442 old = pte_val(*ptep);
443 new = pte_val(pte_wrprotect(__pte (old)));
444 } while (cmpxchg((unsigned long *) ptep, old, new) != old);
445#else
446 pte_t old_pte = *ptep;
447 set_pte_at(mm, addr, ptep, pte_wrprotect(old_pte));
448#endif
449}
450
451static inline int
452pte_same (pte_t a, pte_t b)
453{
454 return pte_val(a) == pte_val(b);
455}
456
Russell King4b3073e2009-12-18 16:40:18 +0000457#define update_mmu_cache(vma, address, ptep) do { } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458
459extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
460extern void paging_init (void);
461
462/*
463 * Note: The macros below rely on the fact that MAX_SWAPFILES_SHIFT <= number of
464 * bits in the swap-type field of the swap pte. It would be nice to
465 * enforce that, but we can't easily include <linux/swap.h> here.
466 * (Of course, better still would be to define MAX_SWAPFILES_SHIFT here...).
467 *
468 * Format of swap pte:
469 * bit 0 : present bit (must be zero)
Kirill A. Shutemov636a0022015-02-10 14:10:36 -0800470 * bits 1- 7: swap-type
471 * bits 8-62: swap offset
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 * bit 63 : _PAGE_PROTNONE bit
473 */
Kirill A. Shutemov636a0022015-02-10 14:10:36 -0800474#define __swp_type(entry) (((entry).val >> 1) & 0x7f)
475#define __swp_offset(entry) (((entry).val << 1) >> 9)
476#define __swp_entry(type,offset) ((swp_entry_t) { ((type) << 1) | ((long) (offset) << 8) })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
478#define __swp_entry_to_pte(x) ((pte_t) { (x).val })
479
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480/*
481 * ZERO_PAGE is a global shared page that is always zero: used
482 * for zero-mapped memory areas etc..
483 */
484extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
485extern struct page *zero_page_memmap_ptr;
486#define ZERO_PAGE(vaddr) (zero_page_memmap_ptr)
487
488/* We provide our own get_unmapped_area to cope with VA holes for userland */
489#define HAVE_ARCH_UNMAPPED_AREA
490
491#ifdef CONFIG_HUGETLB_PAGE
492#define HUGETLB_PGDIR_SHIFT (HPAGE_SHIFT + 2*(PAGE_SHIFT-3))
493#define HUGETLB_PGDIR_SIZE (__IA64_UL(1) << HUGETLB_PGDIR_SHIFT)
494#define HUGETLB_PGDIR_MASK (~(HUGETLB_PGDIR_SIZE-1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495#endif
496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497
498#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
499/*
500 * Update PTEP with ENTRY, which is guaranteed to be a less
501 * restrictive PTE. That is, ENTRY may have the ACCESSED, DIRTY, and
502 * WRITABLE bits turned on, when the value at PTEP did not. The
503 * WRITABLE bit may only be turned if SAFELY_WRITABLE is TRUE.
504 *
505 * SAFELY_WRITABLE is TRUE if we can update the value at PTEP without
506 * having to worry about races. On SMP machines, there are only two
507 * cases where this is true:
508 *
509 * (1) *PTEP has the PRESENT bit turned OFF
510 * (2) ENTRY has the DIRTY bit turned ON
511 *
512 * On ia64, we could implement this routine with a cmpxchg()-loop
513 * which ORs in the _PAGE_A/_PAGE_D bit if they're set in ENTRY.
514 * However, like on x86, we can get a more streamlined version by
515 * observing that it is OK to drop ACCESSED bit updates when
516 * SAFELY_WRITABLE is FALSE. Besides being rare, all that would do is
517 * result in an extra Access-bit fault, which would then turn on the
518 * ACCESSED bit in the low-level fault handler (iaccess_bit or
519 * daccess_bit in ivt.S).
520 */
521#ifdef CONFIG_SMP
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700522# define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __safely_writable) \
523({ \
524 int __changed = !pte_same(*(__ptep), __entry); \
525 if (__changed && __safely_writable) { \
526 set_pte(__ptep, __entry); \
527 flush_tlb_page(__vma, __addr); \
528 } \
529 __changed; \
530})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531#else
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700532# define ptep_set_access_flags(__vma, __addr, __ptep, __entry, __safely_writable) \
533({ \
534 int __changed = !pte_same(*(__ptep), __entry); \
Martin Schwidefskyf0e47c22007-07-17 04:03:03 -0700535 if (__changed) { \
536 set_pte_at((__vma)->vm_mm, (__addr), __ptep, __entry); \
537 flush_tlb_page(__vma, __addr); \
538 } \
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700539 __changed; \
540})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541#endif
542
543# ifdef CONFIG_VIRTUAL_MEM_MAP
544 /* arch mem_map init routine is needed due to holes in a virtual mem_map */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 extern void memmap_init (unsigned long size, int nid, unsigned long zone,
546 unsigned long start_pfn);
547# endif /* CONFIG_VIRTUAL_MEM_MAP */
548# endif /* !__ASSEMBLY__ */
549
550/*
551 * Identity-mapped regions use a large page size. We'll call such large pages
552 * "granules". If you can think of a better name that's unambiguous, let me
553 * know...
554 */
555#if defined(CONFIG_IA64_GRANULE_64MB)
556# define IA64_GRANULE_SHIFT _PAGE_SIZE_64M
557#elif defined(CONFIG_IA64_GRANULE_16MB)
558# define IA64_GRANULE_SHIFT _PAGE_SIZE_16M
559#endif
560#define IA64_GRANULE_SIZE (1 << IA64_GRANULE_SHIFT)
561/*
562 * log2() of the page size we use to map the kernel image (IA64_TR_KERNEL):
563 */
564#define KERNEL_TR_PAGE_SHIFT _PAGE_SIZE_64M
565#define KERNEL_TR_PAGE_SIZE (1 << KERNEL_TR_PAGE_SHIFT)
566
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567/* These tell get_user_pages() that the first gate page is accessible from user-level. */
568#define FIXADDR_USER_START GATE_ADDR
David Mosberger-Tangad597bd2005-06-08 10:45:00 -0700569#ifdef HAVE_BUGGY_SEGREL
570# define FIXADDR_USER_END (GATE_ADDR + 2*PAGE_SIZE)
571#else
572# define FIXADDR_USER_END (GATE_ADDR + 2*PERCPU_PAGE_SIZE)
573#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574
575#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
577#define __HAVE_ARCH_PTEP_SET_WRPROTECT
578#define __HAVE_ARCH_PTE_SAME
579#define __HAVE_ARCH_PGD_OFFSET_GATE
KAMEZAWA Hiroyuki954ffcb2007-10-16 01:25:44 -0700580
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
Kirill A. Shutemov4d66bcc2015-04-14 15:45:45 -0700582#if CONFIG_PGTABLE_LEVELS == 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583#include <asm-generic/pgtable-nopud.h>
Robin Holt837cd0b2005-11-11 09:35:43 -0600584#endif
Mike Rapoportc03ab9e2020-06-04 16:46:31 -0700585#include <asm-generic/pgtable-nop4d.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586#include <asm-generic/pgtable.h>
587
588#endif /* _ASM_IA64_PGTABLE_H */