blob: 55f9d38e3bf8702d302054ead258bf588edf199b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifdef __KERNEL__
2#ifndef _PPC_PGTABLE_H
3#define _PPC_PGTABLE_H
4
5#include <asm-generic/4level-fixup.h>
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007
8#ifndef __ASSEMBLY__
9#include <linux/sched.h>
10#include <linux/threads.h>
11#include <asm/processor.h> /* For TASK_SIZE */
12#include <asm/mmu.h>
13#include <asm/page.h>
Marcelo Tosatti0ec57e52006-01-17 00:24:42 -020014#include <asm/io.h> /* For sub-arch specific PPC_PIN_SIZE */
Tim Schmielau8c65b4a2005-11-07 00:59:43 -080015struct mm_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -070016
17extern unsigned long va_to_phys(unsigned long address);
18extern pte_t *va_to_pte(unsigned long address);
19extern unsigned long ioremap_bot, ioremap_base;
20#endif /* __ASSEMBLY__ */
21
22/*
23 * The PowerPC MMU uses a hash table containing PTEs, together with
24 * a set of 16 segment registers (on 32-bit implementations), to define
25 * the virtual to physical address mapping.
26 *
27 * We use the hash table as an extended TLB, i.e. a cache of currently
28 * active mappings. We maintain a two-level page table tree, much
29 * like that used by the i386, for the sake of the Linux memory
30 * management code. Low-level assembler code in hashtable.S
31 * (procedure hash_page) is responsible for extracting ptes from the
32 * tree and putting them into the hash table when necessary, and
33 * updating the accessed and modified bits in the page table tree.
34 */
35
36/*
37 * The PowerPC MPC8xx uses a TLB with hardware assisted, software tablewalk.
38 * We also use the two level tables, but we can put the real bits in them
39 * needed for the TLB and tablewalk. These definitions require Mx_CTR.PPM = 0,
40 * Mx_CTR.PPCS = 0, and MD_CTR.TWAM = 1. The level 2 descriptor has
41 * additional page protection (when Mx_CTR.PPCS = 1) that allows TLB hit
42 * based upon user/super access. The TLB does not have accessed nor write
43 * protect. We assume that if the TLB get loaded with an entry it is
44 * accessed, and overload the changed bit for write protect. We use
45 * two bits in the software pte that are supposed to be set to zero in
46 * the TLB entry (24 and 25) for these indicators. Although the level 1
47 * descriptor contains the guarded and writethrough/copyback bits, we can
48 * set these at the page level since they get copied from the Mx_TWC
49 * register when the TLB entry is loaded. We will use bit 27 for guard, since
50 * that is where it exists in the MD_TWC, and bit 26 for writethrough.
51 * These will get masked from the level 2 descriptor at TLB load time, and
52 * copied to the MD_TWC before it gets loaded.
53 * Large page sizes added. We currently support two sizes, 4K and 8M.
54 * This also allows a TLB hander optimization because we can directly
55 * load the PMD into MD_TWC. The 8M pages are only used for kernel
56 * mapping of well known areas. The PMD (PGD) entries contain control
57 * flags in addition to the address, so care must be taken that the
58 * software no longer assumes these are only pointers.
59 */
60
61/*
62 * At present, all PowerPC 400-class processors share a similar TLB
63 * architecture. The instruction and data sides share a unified,
64 * 64-entry, fully-associative TLB which is maintained totally under
65 * software control. In addition, the instruction side has a
66 * hardware-managed, 4-entry, fully-associative TLB which serves as a
67 * first level to the shared TLB. These two TLBs are known as the UTLB
68 * and ITLB, respectively (see "mmu.h" for definitions).
69 */
70
71/*
72 * The normal case is that PTEs are 32-bits and we have a 1-page
73 * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages. -- paulus
74 *
75 * For any >32-bit physical address platform, we can use the following
76 * two level page table layout where the pgdir is 8KB and the MS 13 bits
77 * are an index to the second level table. The combined pgdir/pmd first
78 * level has 2048 entries and the second level has 512 64-bit PTE entries.
79 * -Matt
80 */
81/* PMD_SHIFT determines the size of the area mapped by the PTE pages */
82#define PMD_SHIFT (PAGE_SHIFT + PTE_SHIFT)
83#define PMD_SIZE (1UL << PMD_SHIFT)
84#define PMD_MASK (~(PMD_SIZE-1))
85
86/* PGDIR_SHIFT determines what a top-level page table entry can map */
87#define PGDIR_SHIFT PMD_SHIFT
88#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
89#define PGDIR_MASK (~(PGDIR_SIZE-1))
90
91/*
92 * entries per page directory level: our page-table tree is two-level, so
93 * we don't really have any PMD directory.
94 */
95#define PTRS_PER_PTE (1 << PTE_SHIFT)
96#define PTRS_PER_PMD 1
97#define PTRS_PER_PGD (1 << (32 - PGDIR_SHIFT))
98
99#define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE)
Hugh Dickinsd455a362005-04-19 13:29:23 -0700100#define FIRST_USER_ADDRESS 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
102#define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT)
103#define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS)
104
105#define pte_ERROR(e) \
106 printk("%s:%d: bad pte "PTE_FMT".\n", __FILE__, __LINE__, pte_val(e))
107#define pmd_ERROR(e) \
108 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
109#define pgd_ERROR(e) \
110 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
111
112/*
113 * Just any arbitrary offset to the start of the vmalloc VM area: the
114 * current 64MB value just means that there will be a 64MB "hole" after the
115 * physical memory until the kernel virtual memory starts. That means that
116 * any out-of-bounds memory accesses will hopefully be caught.
117 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
118 * area for the same reason. ;)
119 *
120 * We no longer map larger than phys RAM with the BATs so we don't have
121 * to worry about the VMALLOC_OFFSET causing problems. We do have to worry
122 * about clashes between our early calls to ioremap() that start growing down
123 * from ioremap_base being run into the VM area allocations (growing upwards
124 * from VMALLOC_START). For this reason we have ioremap_bot to check when
125 * we actually run into our mappings setup in the early boot with the VM
126 * system. This really does become a problem for machines with good amounts
127 * of RAM. -- Cort
128 */
129#define VMALLOC_OFFSET (0x1000000) /* 16M */
Marcelo Tosatti0ec57e52006-01-17 00:24:42 -0200130#ifdef PPC_PIN_SIZE
131#define VMALLOC_START (((_ALIGN((long)high_memory, PPC_PIN_SIZE) + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132#else
133#define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
134#endif
135#define VMALLOC_END ioremap_bot
136
137/*
138 * Bits in a linux-style PTE. These match the bits in the
139 * (hardware-defined) PowerPC PTE as closely as possible.
140 */
141
142#if defined(CONFIG_40x)
143
144/* There are several potential gotchas here. The 40x hardware TLBLO
145 field looks like this:
146
147 0 1 2 3 4 ... 18 19 20 21 22 23 24 25 26 27 28 29 30 31
148 RPN..................... 0 0 EX WR ZSEL....... W I M G
149
150 Where possible we make the Linux PTE bits match up with this
151
152 - bits 20 and 21 must be cleared, because we use 4k pages (40x can
153 support down to 1k pages), this is done in the TLBMiss exception
154 handler.
155 - We use only zones 0 (for kernel pages) and 1 (for user pages)
156 of the 16 available. Bit 24-26 of the TLB are cleared in the TLB
157 miss handler. Bit 27 is PAGE_USER, thus selecting the correct
158 zone.
159 - PRESENT *must* be in the bottom two bits because swap cache
160 entries use the top 30 bits. Because 40x doesn't support SMP
161 anyway, M is irrelevant so we borrow it for PAGE_PRESENT. Bit 30
162 is cleared in the TLB miss handler before the TLB entry is loaded.
163 - All other bits of the PTE are loaded into TLBLO without
164 modification, leaving us only the bits 20, 21, 24, 25, 26, 30 for
165 software PTE bits. We actually use use bits 21, 24, 25, and
166 30 respectively for the software bits: ACCESSED, DIRTY, RW, and
167 PRESENT.
168*/
169
170/* Definitions for 40x embedded chips. */
171#define _PAGE_GUARDED 0x001 /* G: page is guarded from prefetch */
172#define _PAGE_FILE 0x001 /* when !present: nonlinear file mapping */
173#define _PAGE_PRESENT 0x002 /* software: PTE contains a translation */
174#define _PAGE_NO_CACHE 0x004 /* I: caching is inhibited */
175#define _PAGE_WRITETHRU 0x008 /* W: caching is write-through */
176#define _PAGE_USER 0x010 /* matches one of the zone permission bits */
177#define _PAGE_RW 0x040 /* software: Writes permitted */
178#define _PAGE_DIRTY 0x080 /* software: dirty page */
179#define _PAGE_HWWRITE 0x100 /* hardware: Dirty & RW, set in exception */
180#define _PAGE_HWEXEC 0x200 /* hardware: EX permission */
181#define _PAGE_ACCESSED 0x400 /* software: R: page referenced */
182
183#define _PMD_PRESENT 0x400 /* PMD points to page of PTEs */
184#define _PMD_BAD 0x802
185#define _PMD_SIZE 0x0e0 /* size field, != 0 for large-page PMD entry */
186#define _PMD_SIZE_4M 0x0c0
187#define _PMD_SIZE_16M 0x0e0
188#define PMD_PAGE_SIZE(pmdval) (1024 << (((pmdval) & _PMD_SIZE) >> 4))
189
190#elif defined(CONFIG_44x)
191/*
192 * Definitions for PPC440
193 *
194 * Because of the 3 word TLB entries to support 36-bit addressing,
195 * the attribute are difficult to map in such a fashion that they
196 * are easily loaded during exception processing. I decided to
197 * organize the entry so the ERPN is the only portion in the
198 * upper word of the PTE and the attribute bits below are packed
199 * in as sensibly as they can be in the area below a 4KB page size
200 * oriented RPN. This at least makes it easy to load the RPN and
201 * ERPN fields in the TLB. -Matt
202 *
203 * Note that these bits preclude future use of a page size
204 * less than 4KB.
Matt Porter534afb92005-08-05 16:10:10 -0700205 *
206 *
207 * PPC 440 core has following TLB attribute fields;
208 *
209 * TLB1:
210 * 0 1 2 3 4 ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
211 * RPN................................. - - - - - - ERPN.......
212 *
213 * TLB2:
214 * 0 1 2 3 4 ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
215 * - - - - - - U0 U1 U2 U3 W I M G E - UX UW UR SX SW SR
216 *
217 * There are some constrains and options, to decide mapping software bits
218 * into TLB entry.
219 *
220 * - PRESENT *must* be in the bottom three bits because swap cache
221 * entries use the top 29 bits for TLB2.
222 *
223 * - FILE *must* be in the bottom three bits because swap cache
224 * entries use the top 29 bits for TLB2.
225 *
226 * - CACHE COHERENT bit (M) has no effect on PPC440 core, because it
227 * doesn't support SMP. So we can use this as software bit, like
228 * DIRTY.
229 *
Matt Porter021a52a2005-08-06 07:21:06 -0700230 * With the PPC 44x Linux implementation, the 0-11th LSBs of the PTE are used
231 * for memory protection related functions (see PTE structure in
232 * include/asm-ppc/mmu.h). The _PAGE_XXX definitions in this file map to the
233 * above bits. Note that the bit values are CPU specific, not architecture
234 * specific.
Matt Porter534afb92005-08-05 16:10:10 -0700235 *
Matt Porter021a52a2005-08-06 07:21:06 -0700236 * The kernel PTE entry holds an arch-dependent swp_entry structure under
237 * certain situations. In other words, in such situations some portion of
238 * the PTE bits are used as a swp_entry. In the PPC implementation, the
239 * 3-24th LSB are shared with swp_entry, however the 0-2nd three LSB still
240 * hold protection values. That means the three protection bits are
241 * reserved for both PTE and SWAP entry at the most significant three
242 * LSBs.
Matt Porter534afb92005-08-05 16:10:10 -0700243 *
Matt Porter021a52a2005-08-06 07:21:06 -0700244 * There are three protection bits available for SWAP entry:
Matt Porter534afb92005-08-05 16:10:10 -0700245 * _PAGE_PRESENT
246 * _PAGE_FILE
247 * _PAGE_HASHPTE (if HW has)
248 *
249 * So those three bits have to be inside of 0-2nd LSB of PTE.
250 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 */
Matt Porter534afb92005-08-05 16:10:10 -0700252
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253#define _PAGE_PRESENT 0x00000001 /* S: PTE valid */
254#define _PAGE_RW 0x00000002 /* S: Write permission */
Matt Porter534afb92005-08-05 16:10:10 -0700255#define _PAGE_FILE 0x00000004 /* S: nonlinear file mapping */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256#define _PAGE_ACCESSED 0x00000008 /* S: Page referenced */
257#define _PAGE_HWWRITE 0x00000010 /* H: Dirty & RW */
258#define _PAGE_HWEXEC 0x00000020 /* H: Execute permission */
259#define _PAGE_USER 0x00000040 /* S: User page */
260#define _PAGE_ENDIAN 0x00000080 /* H: E bit */
261#define _PAGE_GUARDED 0x00000100 /* H: G bit */
Matt Porter534afb92005-08-05 16:10:10 -0700262#define _PAGE_DIRTY 0x00000200 /* S: Page dirty */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263#define _PAGE_NO_CACHE 0x00000400 /* H: I bit */
264#define _PAGE_WRITETHRU 0x00000800 /* H: W bit */
265
266/* TODO: Add large page lowmem mapping support */
267#define _PMD_PRESENT 0
268#define _PMD_PRESENT_MASK (PAGE_MASK)
269#define _PMD_BAD (~PAGE_MASK)
270
271/* ERPN in a PTE never gets cleared, ignore it */
272#define _PTE_NONE_MASK 0xffffffff00000000ULL
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274#elif defined(CONFIG_8xx)
275/* Definitions for 8xx embedded chips. */
276#define _PAGE_PRESENT 0x0001 /* Page is valid */
277#define _PAGE_FILE 0x0002 /* when !present: nonlinear file mapping */
278#define _PAGE_NO_CACHE 0x0002 /* I: cache inhibit */
279#define _PAGE_SHARED 0x0004 /* No ASID (context) compare */
280
281/* These five software bits must be masked out when the entry is loaded
282 * into the TLB.
283 */
284#define _PAGE_EXEC 0x0008 /* software: i-cache coherency required */
285#define _PAGE_GUARDED 0x0010 /* software: guarded access */
286#define _PAGE_DIRTY 0x0020 /* software: page changed */
287#define _PAGE_RW 0x0040 /* software: user write access allowed */
288#define _PAGE_ACCESSED 0x0080 /* software: page referenced */
289
290/* Setting any bits in the nibble with the follow two controls will
291 * require a TLB exception handler change. It is assumed unused bits
292 * are always zero.
293 */
294#define _PAGE_HWWRITE 0x0100 /* h/w write enable: never set in Linux PTE */
295#define _PAGE_USER 0x0800 /* One of the PP bits, the other is USER&~RW */
296
297#define _PMD_PRESENT 0x0001
298#define _PMD_BAD 0x0ff0
299#define _PMD_PAGE_MASK 0x000c
300#define _PMD_PAGE_8M 0x000c
301
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302#define _PTE_NONE_MASK _PAGE_ACCESSED
303
304#else /* CONFIG_6xx */
305/* Definitions for 60x, 740/750, etc. */
306#define _PAGE_PRESENT 0x001 /* software: pte contains a translation */
307#define _PAGE_HASHPTE 0x002 /* hash_page has made an HPTE for this pte */
308#define _PAGE_FILE 0x004 /* when !present: nonlinear file mapping */
309#define _PAGE_USER 0x004 /* usermode access allowed */
310#define _PAGE_GUARDED 0x008 /* G: prohibit speculative access */
311#define _PAGE_COHERENT 0x010 /* M: enforce memory coherence (SMP systems) */
312#define _PAGE_NO_CACHE 0x020 /* I: cache inhibit */
313#define _PAGE_WRITETHRU 0x040 /* W: cache write-through */
314#define _PAGE_DIRTY 0x080 /* C: page changed */
315#define _PAGE_ACCESSED 0x100 /* R: page referenced */
316#define _PAGE_EXEC 0x200 /* software: i-cache coherency required */
317#define _PAGE_RW 0x400 /* software: user write access allowed */
318
319#define _PTE_NONE_MASK _PAGE_HASHPTE
320
321#define _PMD_PRESENT 0
322#define _PMD_PRESENT_MASK (PAGE_MASK)
323#define _PMD_BAD (~PAGE_MASK)
324#endif
325
326/*
327 * Some bits are only used on some cpu families...
328 */
329#ifndef _PAGE_HASHPTE
330#define _PAGE_HASHPTE 0
331#endif
332#ifndef _PTE_NONE_MASK
333#define _PTE_NONE_MASK 0
334#endif
335#ifndef _PAGE_SHARED
336#define _PAGE_SHARED 0
337#endif
338#ifndef _PAGE_HWWRITE
339#define _PAGE_HWWRITE 0
340#endif
341#ifndef _PAGE_HWEXEC
342#define _PAGE_HWEXEC 0
343#endif
344#ifndef _PAGE_EXEC
345#define _PAGE_EXEC 0
346#endif
347#ifndef _PMD_PRESENT_MASK
348#define _PMD_PRESENT_MASK _PMD_PRESENT
349#endif
350#ifndef _PMD_SIZE
351#define _PMD_SIZE 0
352#define PMD_PAGE_SIZE(pmd) bad_call_to_PMD_PAGE_SIZE()
353#endif
354
355#define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
356
357/*
358 * Note: the _PAGE_COHERENT bit automatically gets set in the hardware
359 * PTE if CONFIG_SMP is defined (hash_page does this); there is no need
360 * to have it in the Linux PTE, and in fact the bit could be reused for
361 * another purpose. -- paulus.
362 */
363
364#ifdef CONFIG_44x
365#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_GUARDED)
366#else
367#define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED)
368#endif
369#define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY | _PAGE_HWWRITE)
370#define _PAGE_KERNEL (_PAGE_BASE | _PAGE_SHARED | _PAGE_WRENABLE)
371
372#ifdef CONFIG_PPC_STD_MMU
373/* On standard PPC MMU, no user access implies kernel read/write access,
374 * so to write-protect kernel memory we must turn on user access */
375#define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED | _PAGE_USER)
376#else
377#define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED)
378#endif
379
380#define _PAGE_IO (_PAGE_KERNEL | _PAGE_NO_CACHE | _PAGE_GUARDED)
381#define _PAGE_RAM (_PAGE_KERNEL | _PAGE_HWEXEC)
382
383#if defined(CONFIG_KGDB) || defined(CONFIG_XMON) || defined(CONFIG_BDI_SWITCH)
384/* We want the debuggers to be able to set breakpoints anywhere, so
385 * don't write protect the kernel text */
386#define _PAGE_RAM_TEXT _PAGE_RAM
387#else
388#define _PAGE_RAM_TEXT (_PAGE_KERNEL_RO | _PAGE_HWEXEC)
389#endif
390
391#define PAGE_NONE __pgprot(_PAGE_BASE)
392#define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
393#define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
394#define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW)
395#define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW | _PAGE_EXEC)
396#define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
397#define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
398
399#define PAGE_KERNEL __pgprot(_PAGE_RAM)
400#define PAGE_KERNEL_NOCACHE __pgprot(_PAGE_IO)
401
402/*
403 * The PowerPC can only do execute protection on a segment (256MB) basis,
404 * not on a page basis. So we consider execute permission the same as read.
405 * Also, write permissions imply read permissions.
406 * This is the closest we can get..
407 */
408#define __P000 PAGE_NONE
409#define __P001 PAGE_READONLY_X
410#define __P010 PAGE_COPY
411#define __P011 PAGE_COPY_X
412#define __P100 PAGE_READONLY
413#define __P101 PAGE_READONLY_X
414#define __P110 PAGE_COPY
415#define __P111 PAGE_COPY_X
416
417#define __S000 PAGE_NONE
418#define __S001 PAGE_READONLY_X
419#define __S010 PAGE_SHARED
420#define __S011 PAGE_SHARED_X
421#define __S100 PAGE_READONLY
422#define __S101 PAGE_READONLY_X
423#define __S110 PAGE_SHARED
424#define __S111 PAGE_SHARED_X
425
426#ifndef __ASSEMBLY__
427/* Make sure we get a link error if PMD_PAGE_SIZE is ever called on a
428 * kernel without large page PMD support */
429extern unsigned long bad_call_to_PMD_PAGE_SIZE(void);
430
431/*
432 * Conversions between PTE values and page frame numbers.
433 */
434
Kumar Galab464fce2005-04-16 15:24:21 -0700435/* in some case we want to additionaly adjust where the pfn is in the pte to
436 * allow room for more flags */
437#define PFN_SHIFT_OFFSET (PAGE_SHIFT)
438
439#define pte_pfn(x) (pte_val(x) >> PFN_SHIFT_OFFSET)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440#define pte_page(x) pfn_to_page(pte_pfn(x))
441
Kumar Galab464fce2005-04-16 15:24:21 -0700442#define pfn_pte(pfn, prot) __pte(((pte_basic_t)(pfn) << PFN_SHIFT_OFFSET) |\
443 pgprot_val(prot))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444#define mk_pte(page, prot) pfn_pte(page_to_pfn(page), prot)
445
446/*
447 * ZERO_PAGE is a global shared page that is always zero: used
448 * for zero-mapped memory areas etc..
449 */
450extern unsigned long empty_zero_page[1024];
451#define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
452
453#endif /* __ASSEMBLY__ */
454
455#define pte_none(pte) ((pte_val(pte) & ~_PTE_NONE_MASK) == 0)
456#define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
457#define pte_clear(mm,addr,ptep) do { set_pte_at((mm), (addr), (ptep), __pte(0)); } while (0)
458
459#define pmd_none(pmd) (!pmd_val(pmd))
460#define pmd_bad(pmd) (pmd_val(pmd) & _PMD_BAD)
461#define pmd_present(pmd) (pmd_val(pmd) & _PMD_PRESENT_MASK)
462#define pmd_clear(pmdp) do { pmd_val(*(pmdp)) = 0; } while (0)
463
464#ifndef __ASSEMBLY__
465/*
466 * The "pgd_xxx()" functions here are trivial for a folded two-level
467 * setup: the pgd is never bad, and a pmd always exists (as it's folded
468 * into the pgd entry)
469 */
470static inline int pgd_none(pgd_t pgd) { return 0; }
471static inline int pgd_bad(pgd_t pgd) { return 0; }
472static inline int pgd_present(pgd_t pgd) { return 1; }
473#define pgd_clear(xp) do { } while (0)
474
Dave McCracken46a82b22006-09-25 23:31:48 -0700475#define pgd_page_vaddr(pgd) \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK))
477
478/*
479 * The following only work if pte_present() is true.
480 * Undefined behaviour if not..
481 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY; }
484static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED; }
485static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE; }
Nick Piggin7e675132008-04-28 02:13:00 -0700486static inline int pte_special(pte_t pte) { return 0; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
488static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; }
489static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; }
490
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491static inline pte_t pte_wrprotect(pte_t pte) {
492 pte_val(pte) &= ~(_PAGE_RW | _PAGE_HWWRITE); return pte; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493static inline pte_t pte_mkclean(pte_t pte) {
494 pte_val(pte) &= ~(_PAGE_DIRTY | _PAGE_HWWRITE); return pte; }
495static inline pte_t pte_mkold(pte_t pte) {
496 pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
497
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498static inline pte_t pte_mkwrite(pte_t pte) {
499 pte_val(pte) |= _PAGE_RW; return pte; }
500static inline pte_t pte_mkdirty(pte_t pte) {
501 pte_val(pte) |= _PAGE_DIRTY; return pte; }
502static inline pte_t pte_mkyoung(pte_t pte) {
503 pte_val(pte) |= _PAGE_ACCESSED; return pte; }
Nick Piggin7e675132008-04-28 02:13:00 -0700504static inline pte_t pte_mkspecial(pte_t pte) {
505 return pte; }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
507static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
508{
509 pte_val(pte) = (pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot);
510 return pte;
511}
512
513/*
514 * When flushing the tlb entry for a page, we also need to flush the hash
515 * table entry. flush_hash_pages is assembler (for speed) in hashtable.S.
516 */
517extern int flush_hash_pages(unsigned context, unsigned long va,
518 unsigned long pmdval, int count);
519
520/* Add an HPTE to the hash table */
521extern void add_hash_page(unsigned context, unsigned long va,
522 unsigned long pmdval);
523
524/*
525 * Atomic PTE updates.
526 *
527 * pte_update clears and sets bit atomically, and returns
Kumar Gala7a1e3352005-04-16 15:24:20 -0700528 * the old pte value. In the 64-bit PTE case we lock around the
529 * low PTE word since we expect ALL flag bits to be there
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 */
Kumar Gala7a1e3352005-04-16 15:24:20 -0700531#ifndef CONFIG_PTE_64BIT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532static inline unsigned long pte_update(pte_t *p, unsigned long clr,
533 unsigned long set)
534{
535 unsigned long old, tmp;
536
537 __asm__ __volatile__("\
5381: lwarx %0,0,%3\n\
539 andc %1,%0,%4\n\
540 or %1,%1,%5\n"
541 PPC405_ERR77(0,%3)
542" stwcx. %1,0,%3\n\
543 bne- 1b"
544 : "=&r" (old), "=&r" (tmp), "=m" (*p)
Kumar Gala7a1e3352005-04-16 15:24:20 -0700545 : "r" (p), "r" (clr), "r" (set), "m" (*p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 : "cc" );
547 return old;
548}
Kumar Gala7a1e3352005-04-16 15:24:20 -0700549#else
550static inline unsigned long long pte_update(pte_t *p, unsigned long clr,
551 unsigned long set)
552{
553 unsigned long long old;
554 unsigned long tmp;
555
556 __asm__ __volatile__("\
5571: lwarx %L0,0,%4\n\
558 lwzx %0,0,%3\n\
559 andc %1,%L0,%5\n\
560 or %1,%1,%6\n"
561 PPC405_ERR77(0,%3)
562" stwcx. %1,0,%4\n\
563 bne- 1b"
564 : "=&r" (old), "=&r" (tmp), "=m" (*p)
565 : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p)
566 : "cc" );
567 return old;
568}
569#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
571/*
572 * set_pte stores a linux PTE into the linux page table.
573 * On machines which use an MMU hash table we avoid changing the
574 * _PAGE_HASHPTE bit.
575 */
576static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
577 pte_t *ptep, pte_t pte)
578{
579#if _PAGE_HASHPTE != 0
580 pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte) & ~_PAGE_HASHPTE);
581#else
582 *ptep = pte;
583#endif
584}
585
586/*
587 * 2.6 calles this without flushing the TLB entry, this is wrong
588 * for our hash-based implementation, we fix that up here
589 */
590#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
591static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep)
592{
593 unsigned long old;
594 old = pte_update(ptep, _PAGE_ACCESSED, 0);
595#if _PAGE_HASHPTE != 0
596 if (old & _PAGE_HASHPTE) {
597 unsigned long ptephys = __pa(ptep) & PAGE_MASK;
598 flush_hash_pages(context, addr, ptephys, 1);
599 }
600#endif
601 return (old & _PAGE_ACCESSED) != 0;
602}
603#define ptep_test_and_clear_young(__vma, __addr, __ptep) \
Paul Mackerras6218a762006-06-11 14:15:17 +1000604 __ptep_test_and_clear_young((__vma)->vm_mm->context.id, __addr, __ptep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
607static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
608 pte_t *ptep)
609{
610 return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0));
611}
612
613#define __HAVE_ARCH_PTEP_SET_WRPROTECT
614static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
615 pte_t *ptep)
616{
617 pte_update(ptep, (_PAGE_RW | _PAGE_HWWRITE), 0);
618}
619
620#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
621static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty)
622{
623 unsigned long bits = pte_val(entry) &
624 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW);
625 pte_update(ptep, 0, bits);
626}
627
628#define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
Benjamin Herrenschmidt8dab5242007-06-16 10:16:12 -0700629({ \
630 int __changed = !pte_same(*(__ptep), __entry); \
631 if (__changed) { \
632 __ptep_set_access_flags(__ptep, __entry, __dirty); \
633 flush_tlb_page_nohash(__vma, __address); \
634 } \
635 __changed; \
636})
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637
638/*
639 * Macro to mark a page protection value as "uncacheable".
640 */
641#define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
642
643struct file;
Roland Dreier8b150472005-10-28 17:46:18 -0700644extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 unsigned long size, pgprot_t vma_prot);
646#define __HAVE_PHYS_MEM_ACCESS_PROT
647
648#define __HAVE_ARCH_PTE_SAME
649#define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0)
650
651/*
652 * Note that on Book E processors, the pmd contains the kernel virtual
653 * (lowmem) address of the pte page. The physical address is less useful
654 * because everything runs with translation enabled (even the TLB miss
655 * handler). On everything else the pmd contains the physical address
656 * of the pte page. -- paulus
657 */
658#ifndef CONFIG_BOOKE
Dave McCracken46a82b22006-09-25 23:31:48 -0700659#define pmd_page_vaddr(pmd) \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
661#define pmd_page(pmd) \
662 (mem_map + (pmd_val(pmd) >> PAGE_SHIFT))
663#else
Dave McCracken46a82b22006-09-25 23:31:48 -0700664#define pmd_page_vaddr(pmd) \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 ((unsigned long) (pmd_val(pmd) & PAGE_MASK))
666#define pmd_page(pmd) \
667 (mem_map + (__pa(pmd_val(pmd)) >> PAGE_SHIFT))
668#endif
669
670/* to find an entry in a kernel page-table-directory */
671#define pgd_offset_k(address) pgd_offset(&init_mm, address)
672
673/* to find an entry in a page-table-directory */
674#define pgd_index(address) ((address) >> PGDIR_SHIFT)
675#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
676
677/* Find an entry in the second-level page table.. */
678static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
679{
680 return (pmd_t *) dir;
681}
682
683/* Find an entry in the third-level page table.. */
684#define pte_index(address) \
685 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
686#define pte_offset_kernel(dir, addr) \
Dave McCracken46a82b22006-09-25 23:31:48 -0700687 ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688#define pte_offset_map(dir, addr) \
689 ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE0) + pte_index(addr))
690#define pte_offset_map_nested(dir, addr) \
691 ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE1) + pte_index(addr))
692
693#define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0)
694#define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1)
695
696extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
697
698extern void paging_init(void);
699
700/*
701 * Encode and decode a swap entry.
702 * Note that the bits we use in a PTE for representing a swap entry
703 * must not include the _PAGE_PRESENT bit, the _PAGE_FILE bit, or the
704 *_PAGE_HASHPTE bit (if used). -- paulus
705 */
706#define __swp_type(entry) ((entry).val & 0x1f)
707#define __swp_offset(entry) ((entry).val >> 5)
708#define __swp_entry(type, offset) ((swp_entry_t) { (type) | ((offset) << 5) })
709#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> 3 })
710#define __swp_entry_to_pte(x) ((pte_t) { (x).val << 3 })
711
712/* Encode and decode a nonlinear file mapping entry */
713#define PTE_FILE_MAX_BITS 29
714#define pte_to_pgoff(pte) (pte_val(pte) >> 3)
715#define pgoff_to_pte(off) ((pte_t) { ((off) << 3) | _PAGE_FILE })
716
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717/* Values for nocacheflag and cmode */
718/* These are not used by the APUS kernel_map, but prevents
719 compilation errors. */
720#define KERNELMAP_FULL_CACHING 0
721#define KERNELMAP_NOCACHE_SER 1
722#define KERNELMAP_NOCACHE_NONSER 2
723#define KERNELMAP_NO_COPYBACK 3
724
725/*
726 * Map some physical address range into the kernel address space.
727 */
728extern unsigned long kernel_map(unsigned long paddr, unsigned long size,
729 int nocacheflag, unsigned long *memavailp );
730
731/*
732 * Set cache mode of (kernel space) address range.
733 */
734extern void kernel_set_cachemode (unsigned long address, unsigned long size,
735 unsigned int cmode);
736
737/* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
738#define kern_addr_valid(addr) (1)
739
740#ifdef CONFIG_PHYS_64BIT
741extern int remap_pfn_range(struct vm_area_struct *vma, unsigned long from,
742 unsigned long paddr, unsigned long size, pgprot_t prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
744static inline int io_remap_pfn_range(struct vm_area_struct *vma,
745 unsigned long vaddr,
746 unsigned long pfn,
747 unsigned long size,
748 pgprot_t prot)
749{
750 phys_addr_t paddr64 = fixup_bigphys_addr(pfn << PAGE_SHIFT, size);
751 return remap_pfn_range(vma, vaddr, paddr64 >> PAGE_SHIFT, size, prot);
752}
753#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
755 remap_pfn_range(vma, vaddr, pfn, size, prot)
756#endif
757
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758/*
759 * No page table caches to initialise
760 */
761#define pgtable_cache_init() do { } while (0)
762
Eugene Suroveginbab70a42006-03-28 10:13:12 -0800763extern int get_pteptr(struct mm_struct *mm, unsigned long addr, pte_t **ptep,
764 pmd_t **pmdp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765
766#include <asm-generic/pgtable.h>
767
768#endif /* !__ASSEMBLY__ */
769
770#endif /* _PPC_PGTABLE_H */
771#endif /* __KERNEL__ */