mm/hugetlb: add more arch-defined huge_pte functions
[linux-2.6-block.git] / arch / s390 / include / asm / pgtable.h
CommitLineData
1da177e4 1/*
1da177e4 2 * S390 version
a53c8fab 3 * Copyright IBM Corp. 1999, 2000
1da177e4
LT
4 * Author(s): Hartmut Penner (hp@de.ibm.com)
5 * Ulrich Weigand (weigand@de.ibm.com)
6 * Martin Schwidefsky (schwidefsky@de.ibm.com)
7 *
8 * Derived from "include/asm-i386/pgtable.h"
9 */
10
11#ifndef _ASM_S390_PGTABLE_H
12#define _ASM_S390_PGTABLE_H
13
1da177e4
LT
14/*
15 * The Linux memory management assumes a three-level page table setup. For
16 * s390 31 bit we "fold" the mid level into the top-level page table, so
17 * that we physically have the same two-level page table as the s390 mmu
18 * expects in 31 bit mode. For s390 64 bit we use three of the five levels
19 * the hardware provides (region first and region second tables are not
20 * used).
21 *
22 * The "pgd_xxx()" functions are trivial for a folded two-level
23 * setup: the pgd is never bad, and a pmd always exists (as it's folded
24 * into the pgd entry)
25 *
26 * This file contains the functions and defines necessary to modify and use
27 * the S390 page table tree.
28 */
29#ifndef __ASSEMBLY__
9789db08 30#include <linux/sched.h>
2dcea57a 31#include <linux/mm_types.h>
abf09bed 32#include <linux/page-flags.h>
1da177e4 33#include <asm/bug.h>
b2fa47e6 34#include <asm/page.h>
1da177e4 35
1da177e4
LT
36extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
37extern void paging_init(void);
2b67fc46 38extern void vmem_map_init(void);
1da177e4
LT
39
40/*
41 * The S390 doesn't have any external MMU info: the kernel page
42 * tables contain all the necessary information.
43 */
4b3073e1 44#define update_mmu_cache(vma, address, ptep) do { } while (0)
b113da65 45#define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
1da177e4
LT
46
47/*
238ec4ef 48 * ZERO_PAGE is a global shared page that is always zero; used
1da177e4
LT
49 * for zero-mapped memory areas etc..
50 */
238ec4ef
MS
51
52extern unsigned long empty_zero_page;
53extern unsigned long zero_page_mask;
54
55#define ZERO_PAGE(vaddr) \
56 (virt_to_page((void *)(empty_zero_page + \
57 (((unsigned long)(vaddr)) &zero_page_mask))))
816422ad 58#define __HAVE_COLOR_ZERO_PAGE
238ec4ef 59
4f2e2903
LT
60/* TODO: s390 cannot support io_remap_pfn_range... */
61#define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
62 remap_pfn_range(vma, vaddr, pfn, size, prot)
63
1da177e4
LT
64#endif /* !__ASSEMBLY__ */
65
66/*
67 * PMD_SHIFT determines the size of the area a second-level page
68 * table can map
69 * PGDIR_SHIFT determines what a third-level page table entry can map
70 */
f4815ac6 71#ifndef CONFIG_64BIT
146e4b3c
MS
72# define PMD_SHIFT 20
73# define PUD_SHIFT 20
74# define PGDIR_SHIFT 20
f4815ac6 75#else /* CONFIG_64BIT */
146e4b3c 76# define PMD_SHIFT 20
190a1d72 77# define PUD_SHIFT 31
5a216a20 78# define PGDIR_SHIFT 42
f4815ac6 79#endif /* CONFIG_64BIT */
1da177e4
LT
80
81#define PMD_SIZE (1UL << PMD_SHIFT)
82#define PMD_MASK (~(PMD_SIZE-1))
190a1d72
MS
83#define PUD_SIZE (1UL << PUD_SHIFT)
84#define PUD_MASK (~(PUD_SIZE-1))
5a216a20
MS
85#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
86#define PGDIR_MASK (~(PGDIR_SIZE-1))
1da177e4
LT
87
88/*
89 * entries per page directory level: the S390 is two-level, so
90 * we don't really have any PMD directory physically.
91 * for S390 segment-table entries are combined to one PGD
92 * that leads to 1024 pte per pgd
93 */
146e4b3c 94#define PTRS_PER_PTE 256
f4815ac6 95#ifndef CONFIG_64BIT
146e4b3c 96#define PTRS_PER_PMD 1
5a216a20 97#define PTRS_PER_PUD 1
f4815ac6 98#else /* CONFIG_64BIT */
146e4b3c 99#define PTRS_PER_PMD 2048
5a216a20 100#define PTRS_PER_PUD 2048
f4815ac6 101#endif /* CONFIG_64BIT */
146e4b3c 102#define PTRS_PER_PGD 2048
1da177e4 103
d455a369
HD
104#define FIRST_USER_ADDRESS 0
105
1da177e4
LT
106#define pte_ERROR(e) \
107 printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
108#define pmd_ERROR(e) \
109 printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
190a1d72
MS
110#define pud_ERROR(e) \
111 printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
1da177e4
LT
112#define pgd_ERROR(e) \
113 printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
114
115#ifndef __ASSEMBLY__
116/*
c972cc60
HC
117 * The vmalloc and module area will always be on the topmost area of the kernel
118 * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc and modules.
119 * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
120 * modules will reside. That makes sure that inter module branches always
121 * happen without trampolines and in addition the placement within a 2GB frame
122 * is branch prediction unit friendly.
8b62bc96 123 */
239a6425 124extern unsigned long VMALLOC_START;
14045ebf
MS
125extern unsigned long VMALLOC_END;
126extern struct page *vmemmap;
239a6425 127
14045ebf 128#define VMEM_MAX_PHYS ((unsigned long) vmemmap)
5fd9c6e2 129
c972cc60
HC
130#ifdef CONFIG_64BIT
131extern unsigned long MODULES_VADDR;
132extern unsigned long MODULES_END;
133#define MODULES_VADDR MODULES_VADDR
134#define MODULES_END MODULES_END
135#define MODULES_LEN (1UL << 31)
136#endif
137
1da177e4
LT
138/*
139 * A 31 bit pagetable entry of S390 has following format:
140 * | PFRA | | OS |
141 * 0 0IP0
142 * 00000000001111111111222222222233
143 * 01234567890123456789012345678901
144 *
145 * I Page-Invalid Bit: Page is not available for address-translation
146 * P Page-Protection Bit: Store access not possible for page
147 *
148 * A 31 bit segmenttable entry of S390 has following format:
149 * | P-table origin | |PTL
150 * 0 IC
151 * 00000000001111111111222222222233
152 * 01234567890123456789012345678901
153 *
154 * I Segment-Invalid Bit: Segment is not available for address-translation
155 * C Common-Segment Bit: Segment is not private (PoP 3-30)
156 * PTL Page-Table-Length: Page-table length (PTL+1*16 entries -> up to 256)
157 *
158 * The 31 bit segmenttable origin of S390 has following format:
159 *
160 * |S-table origin | | STL |
161 * X **GPS
162 * 00000000001111111111222222222233
163 * 01234567890123456789012345678901
164 *
165 * X Space-Switch event:
166 * G Segment-Invalid Bit: *
167 * P Private-Space Bit: Segment is not private (PoP 3-30)
168 * S Storage-Alteration:
169 * STL Segment-Table-Length: Segment-table length (STL+1*16 entries -> up to 2048)
170 *
171 * A 64 bit pagetable entry of S390 has following format:
6a985c61 172 * | PFRA |0IPC| OS |
1da177e4
LT
173 * 0000000000111111111122222222223333333333444444444455555555556666
174 * 0123456789012345678901234567890123456789012345678901234567890123
175 *
176 * I Page-Invalid Bit: Page is not available for address-translation
177 * P Page-Protection Bit: Store access not possible for page
6a985c61 178 * C Change-bit override: HW is not required to set change bit
1da177e4
LT
179 *
180 * A 64 bit segmenttable entry of S390 has following format:
181 * | P-table origin | TT
182 * 0000000000111111111122222222223333333333444444444455555555556666
183 * 0123456789012345678901234567890123456789012345678901234567890123
184 *
185 * I Segment-Invalid Bit: Segment is not available for address-translation
186 * C Common-Segment Bit: Segment is not private (PoP 3-30)
187 * P Page-Protection Bit: Store access not possible for page
188 * TT Type 00
189 *
190 * A 64 bit region table entry of S390 has following format:
191 * | S-table origin | TF TTTL
192 * 0000000000111111111122222222223333333333444444444455555555556666
193 * 0123456789012345678901234567890123456789012345678901234567890123
194 *
195 * I Segment-Invalid Bit: Segment is not available for address-translation
196 * TT Type 01
197 * TF
190a1d72 198 * TL Table length
1da177e4
LT
199 *
200 * The 64 bit regiontable origin of S390 has following format:
201 * | region table origon | DTTL
202 * 0000000000111111111122222222223333333333444444444455555555556666
203 * 0123456789012345678901234567890123456789012345678901234567890123
204 *
205 * X Space-Switch event:
206 * G Segment-Invalid Bit:
207 * P Private-Space Bit:
208 * S Storage-Alteration:
209 * R Real space
210 * TL Table-Length:
211 *
212 * A storage key has the following format:
213 * | ACC |F|R|C|0|
214 * 0 3 4 5 6 7
215 * ACC: access key
216 * F : fetch protection bit
217 * R : referenced bit
218 * C : changed bit
219 */
220
221/* Hardware bits in the page table entry */
6a985c61 222#define _PAGE_CO 0x100 /* HW Change-bit override */
83377484
MS
223#define _PAGE_RO 0x200 /* HW read-only bit */
224#define _PAGE_INVALID 0x400 /* HW invalid bit */
3610cce8
MS
225
226/* Software bits in the page table entry */
83377484
MS
227#define _PAGE_SWT 0x001 /* SW pte type bit t */
228#define _PAGE_SWX 0x002 /* SW pte type bit x */
abf09bed
MS
229#define _PAGE_SWC 0x004 /* SW pte changed bit */
230#define _PAGE_SWR 0x008 /* SW pte referenced bit */
231#define _PAGE_SWW 0x010 /* SW pte write bit */
232#define _PAGE_SPECIAL 0x020 /* SW associated with special page */
a08cb629 233#define __HAVE_ARCH_PTE_SPECIAL
1da177e4 234
138c9021 235/* Set of bits not changed in pte_modify */
abf09bed
MS
236#define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_SPECIAL | _PAGE_CO | \
237 _PAGE_SWC | _PAGE_SWR)
138c9021 238
83377484 239/* Six different types of pages. */
9282ed92
GS
240#define _PAGE_TYPE_EMPTY 0x400
241#define _PAGE_TYPE_NONE 0x401
83377484
MS
242#define _PAGE_TYPE_SWAP 0x403
243#define _PAGE_TYPE_FILE 0x601 /* bit 0x002 is used for offset !! */
9282ed92
GS
244#define _PAGE_TYPE_RO 0x200
245#define _PAGE_TYPE_RW 0x000
1da177e4 246
53492b1d
GS
247/*
248 * Only four types for huge pages, using the invalid bit and protection bit
249 * of a segment table entry.
250 */
251#define _HPAGE_TYPE_EMPTY 0x020 /* _SEGMENT_ENTRY_INV */
252#define _HPAGE_TYPE_NONE 0x220
253#define _HPAGE_TYPE_RO 0x200 /* _SEGMENT_ENTRY_RO */
254#define _HPAGE_TYPE_RW 0x000
255
83377484
MS
256/*
257 * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
258 * pte_none and pte_file to find out the pte type WITHOUT holding the page
259 * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
260 * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
261 * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
262 * This change is done while holding the lock, but the intermediate step
263 * of a previously valid pte with the hw invalid bit set can be observed by
264 * handle_pte_fault. That makes it necessary that all valid pte types with
265 * the hw invalid bit set must be distinguishable from the four pte types
266 * empty, none, swap and file.
267 *
268 * irxt ipte irxt
269 * _PAGE_TYPE_EMPTY 1000 -> 1000
270 * _PAGE_TYPE_NONE 1001 -> 1001
271 * _PAGE_TYPE_SWAP 1011 -> 1011
272 * _PAGE_TYPE_FILE 11?1 -> 11?1
273 * _PAGE_TYPE_RO 0100 -> 1100
274 * _PAGE_TYPE_RW 0000 -> 1000
275 *
c1821c2e 276 * pte_none is true for bits combinations 1000, 1010, 1100, 1110
83377484
MS
277 * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
278 * pte_file is true for bits combinations 1101, 1111
c1821c2e 279 * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
83377484
MS
280 */
281
f4815ac6 282#ifndef CONFIG_64BIT
1da177e4 283
3610cce8
MS
284/* Bits in the segment table address-space-control-element */
285#define _ASCE_SPACE_SWITCH 0x80000000UL /* space switch event */
286#define _ASCE_ORIGIN_MASK 0x7ffff000UL /* segment table origin */
287#define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
288#define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
289#define _ASCE_TABLE_LENGTH 0x7f /* 128 x 64 entries = 8k */
1da177e4 290
3610cce8
MS
291/* Bits in the segment table entry */
292#define _SEGMENT_ENTRY_ORIGIN 0x7fffffc0UL /* page table origin */
80217147 293#define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */
3610cce8
MS
294#define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */
295#define _SEGMENT_ENTRY_COMMON 0x10 /* common segment bit */
296#define _SEGMENT_ENTRY_PTL 0x0f /* page table length */
1da177e4 297
3610cce8
MS
298#define _SEGMENT_ENTRY (_SEGMENT_ENTRY_PTL)
299#define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
1da177e4 300
6c61cfe9
MS
301/* Page status table bits for virtualization */
302#define RCP_ACC_BITS 0xf0000000UL
303#define RCP_FP_BIT 0x08000000UL
304#define RCP_PCL_BIT 0x00800000UL
305#define RCP_HR_BIT 0x00400000UL
306#define RCP_HC_BIT 0x00200000UL
307#define RCP_GR_BIT 0x00040000UL
308#define RCP_GC_BIT 0x00020000UL
309
310/* User dirty / referenced bit for KVM's migration feature */
311#define KVM_UR_BIT 0x00008000UL
312#define KVM_UC_BIT 0x00004000UL
313
f4815ac6 314#else /* CONFIG_64BIT */
1da177e4 315
3610cce8
MS
316/* Bits in the segment/region table address-space-control-element */
317#define _ASCE_ORIGIN ~0xfffUL/* segment table origin */
318#define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
319#define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
320#define _ASCE_SPACE_SWITCH 0x40 /* space switch event */
321#define _ASCE_REAL_SPACE 0x20 /* real space control */
322#define _ASCE_TYPE_MASK 0x0c /* asce table type mask */
323#define _ASCE_TYPE_REGION1 0x0c /* region first table type */
324#define _ASCE_TYPE_REGION2 0x08 /* region second table type */
325#define _ASCE_TYPE_REGION3 0x04 /* region third table type */
326#define _ASCE_TYPE_SEGMENT 0x00 /* segment table type */
327#define _ASCE_TABLE_LENGTH 0x03 /* region table length */
328
329/* Bits in the region table entry */
330#define _REGION_ENTRY_ORIGIN ~0xfffUL/* region/segment table origin */
abf09bed 331#define _REGION_ENTRY_RO 0x200 /* region protection bit */
3610cce8
MS
332#define _REGION_ENTRY_INV 0x20 /* invalid region table entry */
333#define _REGION_ENTRY_TYPE_MASK 0x0c /* region/segment table type mask */
334#define _REGION_ENTRY_TYPE_R1 0x0c /* region first table type */
335#define _REGION_ENTRY_TYPE_R2 0x08 /* region second table type */
336#define _REGION_ENTRY_TYPE_R3 0x04 /* region third table type */
337#define _REGION_ENTRY_LENGTH 0x03 /* region third length */
338
339#define _REGION1_ENTRY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
340#define _REGION1_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
341#define _REGION2_ENTRY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
342#define _REGION2_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
343#define _REGION3_ENTRY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
344#define _REGION3_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
345
18da2369 346#define _REGION3_ENTRY_LARGE 0x400 /* RTTE-format control, large page */
1819ed1f
HC
347#define _REGION3_ENTRY_RO 0x200 /* page protection bit */
348#define _REGION3_ENTRY_CO 0x100 /* change-recording override */
18da2369 349
1da177e4 350/* Bits in the segment table entry */
ea81531d 351#define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address */
3610cce8
MS
352#define _SEGMENT_ENTRY_ORIGIN ~0x7ffUL/* segment table origin */
353#define _SEGMENT_ENTRY_RO 0x200 /* page protection bit */
354#define _SEGMENT_ENTRY_INV 0x20 /* invalid segment table entry */
1da177e4 355
3610cce8
MS
356#define _SEGMENT_ENTRY (0)
357#define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INV)
358
53492b1d
GS
359#define _SEGMENT_ENTRY_LARGE 0x400 /* STE-format control, large page */
360#define _SEGMENT_ENTRY_CO 0x100 /* change-recording override */
75077afb
GS
361#define _SEGMENT_ENTRY_SPLIT_BIT 0 /* THP splitting bit number */
362#define _SEGMENT_ENTRY_SPLIT (1UL << _SEGMENT_ENTRY_SPLIT_BIT)
53492b1d 363
1ae1c1d0
GS
364/* Set of bits not changed in pmd_modify */
365#define _SEGMENT_CHG_MASK (_SEGMENT_ENTRY_ORIGIN | _SEGMENT_ENTRY_LARGE \
366 | _SEGMENT_ENTRY_SPLIT | _SEGMENT_ENTRY_CO)
367
6c61cfe9
MS
368/* Page status table bits for virtualization */
369#define RCP_ACC_BITS 0xf000000000000000UL
370#define RCP_FP_BIT 0x0800000000000000UL
371#define RCP_PCL_BIT 0x0080000000000000UL
372#define RCP_HR_BIT 0x0040000000000000UL
373#define RCP_HC_BIT 0x0020000000000000UL
374#define RCP_GR_BIT 0x0004000000000000UL
375#define RCP_GC_BIT 0x0002000000000000UL
376
377/* User dirty / referenced bit for KVM's migration feature */
378#define KVM_UR_BIT 0x0000800000000000UL
379#define KVM_UC_BIT 0x0000400000000000UL
380
f4815ac6 381#endif /* CONFIG_64BIT */
1da177e4
LT
382
383/*
3610cce8
MS
384 * A user page table pointer has the space-switch-event bit, the
385 * private-space-control bit and the storage-alteration-event-control
386 * bit set. A kernel page table pointer doesn't need them.
1da177e4 387 */
3610cce8
MS
388#define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
389 _ASCE_ALT_EVENT)
1da177e4 390
1da177e4 391/*
9282ed92 392 * Page protection definitions.
1da177e4 393 */
9282ed92
GS
394#define PAGE_NONE __pgprot(_PAGE_TYPE_NONE)
395#define PAGE_RO __pgprot(_PAGE_TYPE_RO)
abf09bed
MS
396#define PAGE_RW __pgprot(_PAGE_TYPE_RO | _PAGE_SWW)
397#define PAGE_RWC __pgprot(_PAGE_TYPE_RW | _PAGE_SWW | _PAGE_SWC)
9282ed92 398
abf09bed 399#define PAGE_KERNEL PAGE_RWC
bddb7ae2 400#define PAGE_SHARED PAGE_KERNEL
9282ed92 401#define PAGE_COPY PAGE_RO
1da177e4
LT
402
403/*
043d0708
MS
404 * On s390 the page table entry has an invalid bit and a read-only bit.
405 * Read permission implies execute permission and write permission
406 * implies read permission.
1da177e4
LT
407 */
408 /*xwr*/
9282ed92
GS
409#define __P000 PAGE_NONE
410#define __P001 PAGE_RO
411#define __P010 PAGE_RO
412#define __P011 PAGE_RO
043d0708
MS
413#define __P100 PAGE_RO
414#define __P101 PAGE_RO
415#define __P110 PAGE_RO
416#define __P111 PAGE_RO
9282ed92
GS
417
418#define __S000 PAGE_NONE
419#define __S001 PAGE_RO
420#define __S010 PAGE_RW
421#define __S011 PAGE_RW
043d0708
MS
422#define __S100 PAGE_RO
423#define __S101 PAGE_RO
424#define __S110 PAGE_RW
425#define __S111 PAGE_RW
1da177e4 426
106c992a
GS
427/*
428 * Segment entry (large page) protection definitions.
429 */
430#define SEGMENT_NONE __pgprot(_HPAGE_TYPE_NONE)
431#define SEGMENT_RO __pgprot(_HPAGE_TYPE_RO)
432#define SEGMENT_RW __pgprot(_HPAGE_TYPE_RW)
433
b2fa47e6 434static inline int mm_exclusive(struct mm_struct *mm)
1da177e4 435{
b2fa47e6
MS
436 return likely(mm == current->active_mm &&
437 atomic_read(&mm->context.attach_count) <= 1);
1da177e4 438}
1da177e4 439
b2fa47e6
MS
440static inline int mm_has_pgste(struct mm_struct *mm)
441{
442#ifdef CONFIG_PGSTE
443 if (unlikely(mm->context.has_pgste))
444 return 1;
445#endif
446 return 0;
447}
1da177e4
LT
448/*
449 * pgd/pmd/pte query functions
450 */
f4815ac6 451#ifndef CONFIG_64BIT
1da177e4 452
4448aaf0
AB
453static inline int pgd_present(pgd_t pgd) { return 1; }
454static inline int pgd_none(pgd_t pgd) { return 0; }
455static inline int pgd_bad(pgd_t pgd) { return 0; }
1da177e4 456
190a1d72
MS
457static inline int pud_present(pud_t pud) { return 1; }
458static inline int pud_none(pud_t pud) { return 0; }
18da2369 459static inline int pud_large(pud_t pud) { return 0; }
190a1d72
MS
460static inline int pud_bad(pud_t pud) { return 0; }
461
f4815ac6 462#else /* CONFIG_64BIT */
1da177e4 463
5a216a20
MS
464static inline int pgd_present(pgd_t pgd)
465{
6252d702
MS
466 if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
467 return 1;
5a216a20
MS
468 return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
469}
470
471static inline int pgd_none(pgd_t pgd)
472{
6252d702
MS
473 if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
474 return 0;
5a216a20
MS
475 return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL;
476}
477
478static inline int pgd_bad(pgd_t pgd)
479{
6252d702
MS
480 /*
481 * With dynamic page table levels the pgd can be a region table
482 * entry or a segment table entry. Check for the bit that are
483 * invalid for either table entry.
484 */
5a216a20 485 unsigned long mask =
6252d702 486 ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
5a216a20
MS
487 ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
488 return (pgd_val(pgd) & mask) != 0;
489}
190a1d72
MS
490
491static inline int pud_present(pud_t pud)
1da177e4 492{
6252d702
MS
493 if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
494 return 1;
0d017923 495 return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
1da177e4
LT
496}
497
190a1d72 498static inline int pud_none(pud_t pud)
1da177e4 499{
6252d702
MS
500 if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
501 return 0;
0d017923 502 return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
1da177e4
LT
503}
504
18da2369
HC
505static inline int pud_large(pud_t pud)
506{
507 if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
508 return 0;
509 return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
510}
511
190a1d72 512static inline int pud_bad(pud_t pud)
1da177e4 513{
6252d702
MS
514 /*
515 * With dynamic page table levels the pud can be a region table
516 * entry or a segment table entry. Check for the bit that are
517 * invalid for either table entry.
518 */
5a216a20 519 unsigned long mask =
6252d702 520 ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
5a216a20
MS
521 ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
522 return (pud_val(pud) & mask) != 0;
1da177e4
LT
523}
524
f4815ac6 525#endif /* CONFIG_64BIT */
3610cce8 526
4448aaf0 527static inline int pmd_present(pmd_t pmd)
1da177e4 528{
d8e7a33d
GS
529 unsigned long mask = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO;
530 return (pmd_val(pmd) & mask) == _HPAGE_TYPE_NONE ||
531 !(pmd_val(pmd) & _SEGMENT_ENTRY_INV);
1da177e4
LT
532}
533
4448aaf0 534static inline int pmd_none(pmd_t pmd)
1da177e4 535{
d8e7a33d
GS
536 return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) &&
537 !(pmd_val(pmd) & _SEGMENT_ENTRY_RO);
1da177e4
LT
538}
539
378b1e7a
HC
540static inline int pmd_large(pmd_t pmd)
541{
542#ifdef CONFIG_64BIT
543 return !!(pmd_val(pmd) & _SEGMENT_ENTRY_LARGE);
544#else
545 return 0;
546#endif
547}
548
4448aaf0 549static inline int pmd_bad(pmd_t pmd)
1da177e4 550{
3610cce8
MS
551 unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
552 return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
1da177e4
LT
553}
554
75077afb
GS
555#define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
556extern void pmdp_splitting_flush(struct vm_area_struct *vma,
557 unsigned long addr, pmd_t *pmdp);
558
1ae1c1d0
GS
559#define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
560extern int pmdp_set_access_flags(struct vm_area_struct *vma,
561 unsigned long address, pmd_t *pmdp,
562 pmd_t entry, int dirty);
563
564#define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
565extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
566 unsigned long address, pmd_t *pmdp);
567
568#define __HAVE_ARCH_PMD_WRITE
569static inline int pmd_write(pmd_t pmd)
570{
571 return (pmd_val(pmd) & _SEGMENT_ENTRY_RO) == 0;
572}
573
574static inline int pmd_young(pmd_t pmd)
575{
576 return 0;
577}
578
4448aaf0 579static inline int pte_none(pte_t pte)
1da177e4 580{
83377484 581 return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
1da177e4
LT
582}
583
4448aaf0 584static inline int pte_present(pte_t pte)
1da177e4 585{
83377484
MS
586 unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
587 return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
588 (!(pte_val(pte) & _PAGE_INVALID) &&
589 !(pte_val(pte) & _PAGE_SWT));
1da177e4
LT
590}
591
4448aaf0 592static inline int pte_file(pte_t pte)
1da177e4 593{
83377484
MS
594 unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
595 return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
1da177e4
LT
596}
597
7e675137
NP
598static inline int pte_special(pte_t pte)
599{
a08cb629 600 return (pte_val(pte) & _PAGE_SPECIAL);
7e675137
NP
601}
602
ba8a9229 603#define __HAVE_ARCH_PTE_SAME
b2fa47e6
MS
604static inline int pte_same(pte_t a, pte_t b)
605{
606 return pte_val(a) == pte_val(b);
607}
1da177e4 608
b2fa47e6 609static inline pgste_t pgste_get_lock(pte_t *ptep)
5b7baf05 610{
b2fa47e6 611 unsigned long new = 0;
5b7baf05 612#ifdef CONFIG_PGSTE
b2fa47e6
MS
613 unsigned long old;
614
5b7baf05 615 preempt_disable();
b2fa47e6
MS
616 asm(
617 " lg %0,%2\n"
618 "0: lgr %1,%0\n"
619 " nihh %0,0xff7f\n" /* clear RCP_PCL_BIT in old */
620 " oihh %1,0x0080\n" /* set RCP_PCL_BIT in new */
621 " csg %0,%1,%2\n"
622 " jl 0b\n"
623 : "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
624 : "Q" (ptep[PTRS_PER_PTE]) : "cc");
5b7baf05 625#endif
b2fa47e6 626 return __pgste(new);
5b7baf05
CB
627}
628
b2fa47e6 629static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
5b7baf05
CB
630{
631#ifdef CONFIG_PGSTE
b2fa47e6
MS
632 asm(
633 " nihh %1,0xff7f\n" /* clear RCP_PCL_BIT */
634 " stg %1,%0\n"
635 : "=Q" (ptep[PTRS_PER_PTE])
636 : "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc");
5b7baf05
CB
637 preempt_enable();
638#endif
639}
640
b2fa47e6 641static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
5b7baf05
CB
642{
643#ifdef CONFIG_PGSTE
a43a9d93 644 unsigned long address, bits;
b2fa47e6
MS
645 unsigned char skey;
646
09b53883
MS
647 if (!pte_present(*ptep))
648 return pgste;
a43a9d93
HC
649 address = pte_val(*ptep) & PAGE_MASK;
650 skey = page_get_storage_key(address);
b2fa47e6
MS
651 bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
652 /* Clear page changed & referenced bit in the storage key */
7c81878b 653 if (bits & _PAGE_CHANGED)
abf09bed 654 page_set_storage_key(address, skey ^ bits, 0);
7c81878b
CO
655 else if (bits)
656 page_reset_referenced(address);
b2fa47e6
MS
657 /* Transfer page changed & referenced bit to guest bits in pgste */
658 pgste_val(pgste) |= bits << 48; /* RCP_GR_BIT & RCP_GC_BIT */
659 /* Get host changed & referenced bits from pgste */
660 bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52;
abf09bed
MS
661 /* Transfer page changed & referenced bit to kvm user bits */
662 pgste_val(pgste) |= bits << 45; /* KVM_UR_BIT & KVM_UC_BIT */
663 /* Clear relevant host bits in pgste. */
b2fa47e6
MS
664 pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT);
665 pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT);
666 /* Copy page access key and fetch protection bit to pgste */
667 pgste_val(pgste) |=
668 (unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
abf09bed
MS
669 /* Transfer referenced bit to pte */
670 pte_val(*ptep) |= (bits & _PAGE_REFERENCED) << 1;
b2fa47e6
MS
671#endif
672 return pgste;
673
674}
675
676static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
677{
678#ifdef CONFIG_PGSTE
679 int young;
680
09b53883
MS
681 if (!pte_present(*ptep))
682 return pgste;
abf09bed 683 /* Get referenced bit from storage key */
b2fa47e6 684 young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK);
abf09bed
MS
685 if (young)
686 pgste_val(pgste) |= RCP_GR_BIT;
687 /* Get host referenced bit from pgste */
688 if (pgste_val(pgste) & RCP_HR_BIT) {
689 pgste_val(pgste) &= ~RCP_HR_BIT;
690 young = 1;
691 }
692 /* Transfer referenced bit to kvm user bits and pte */
693 if (young) {
694 pgste_val(pgste) |= KVM_UR_BIT;
b2fa47e6 695 pte_val(*ptep) |= _PAGE_SWR;
abf09bed 696 }
b2fa47e6
MS
697#endif
698 return pgste;
b2fa47e6
MS
699}
700
abf09bed 701static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry)
b2fa47e6
MS
702{
703#ifdef CONFIG_PGSTE
a43a9d93 704 unsigned long address;
b2fa47e6
MS
705 unsigned long okey, nkey;
706
09b53883
MS
707 if (!pte_present(entry))
708 return;
709 address = pte_val(entry) & PAGE_MASK;
a43a9d93 710 okey = nkey = page_get_storage_key(address);
b2fa47e6
MS
711 nkey &= ~(_PAGE_ACC_BITS | _PAGE_FP_BIT);
712 /* Set page access key and fetch protection bit from pgste */
713 nkey |= (pgste_val(pgste) & (RCP_ACC_BITS | RCP_FP_BIT)) >> 56;
714 if (okey != nkey)
abf09bed 715 page_set_storage_key(address, nkey, 0);
5b7baf05
CB
716#endif
717}
718
abf09bed
MS
719static inline void pgste_set_pte(pte_t *ptep, pte_t entry)
720{
721 if (!MACHINE_HAS_ESOP && (pte_val(entry) & _PAGE_SWW)) {
722 /*
723 * Without enhanced suppression-on-protection force
724 * the dirty bit on for all writable ptes.
725 */
726 pte_val(entry) |= _PAGE_SWC;
727 pte_val(entry) &= ~_PAGE_RO;
728 }
729 *ptep = entry;
730}
731
e5992f2e
MS
732/**
733 * struct gmap_struct - guest address space
734 * @mm: pointer to the parent mm_struct
735 * @table: pointer to the page directory
480e5926 736 * @asce: address space control element for gmap page table
e5992f2e
MS
737 * @crst_list: list of all crst tables used in the guest address space
738 */
739struct gmap {
740 struct list_head list;
741 struct mm_struct *mm;
742 unsigned long *table;
480e5926 743 unsigned long asce;
e5992f2e
MS
744 struct list_head crst_list;
745};
746
747/**
748 * struct gmap_rmap - reverse mapping for segment table entries
749 * @next: pointer to the next gmap_rmap structure in the list
750 * @entry: pointer to a segment table entry
751 */
752struct gmap_rmap {
753 struct list_head list;
754 unsigned long *entry;
755};
756
757/**
758 * struct gmap_pgtable - gmap information attached to a page table
759 * @vmaddr: address of the 1MB segment in the process virtual memory
760 * @mapper: list of segment table entries maping a page table
761 */
762struct gmap_pgtable {
763 unsigned long vmaddr;
764 struct list_head mapper;
765};
766
767struct gmap *gmap_alloc(struct mm_struct *mm);
768void gmap_free(struct gmap *gmap);
769void gmap_enable(struct gmap *gmap);
770void gmap_disable(struct gmap *gmap);
771int gmap_map_segment(struct gmap *gmap, unsigned long from,
772 unsigned long to, unsigned long length);
773int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
c5034945
HC
774unsigned long __gmap_translate(unsigned long address, struct gmap *);
775unsigned long gmap_translate(unsigned long address, struct gmap *);
499069e1 776unsigned long __gmap_fault(unsigned long address, struct gmap *);
e5992f2e 777unsigned long gmap_fault(unsigned long address, struct gmap *);
388186bc 778void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
e5992f2e 779
b2fa47e6
MS
780/*
781 * Certain architectures need to do special things when PTEs
782 * within a page table are directly modified. Thus, the following
783 * hook is made available.
784 */
785static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
786 pte_t *ptep, pte_t entry)
787{
788 pgste_t pgste;
789
790 if (mm_has_pgste(mm)) {
791 pgste = pgste_get_lock(ptep);
abf09bed
MS
792 pgste_set_key(ptep, pgste, entry);
793 pgste_set_pte(ptep, entry);
b2fa47e6 794 pgste_set_unlock(ptep, pgste);
abf09bed
MS
795 } else {
796 if (!(pte_val(entry) & _PAGE_INVALID) && MACHINE_HAS_EDAT1)
797 pte_val(entry) |= _PAGE_CO;
b2fa47e6 798 *ptep = entry;
abf09bed 799 }
b2fa47e6
MS
800}
801
1da177e4
LT
802/*
803 * query functions pte_write/pte_dirty/pte_young only work if
804 * pte_present() is true. Undefined behaviour if not..
805 */
4448aaf0 806static inline int pte_write(pte_t pte)
1da177e4 807{
abf09bed 808 return (pte_val(pte) & _PAGE_SWW) != 0;
1da177e4
LT
809}
810
4448aaf0 811static inline int pte_dirty(pte_t pte)
1da177e4 812{
abf09bed 813 return (pte_val(pte) & _PAGE_SWC) != 0;
1da177e4
LT
814}
815
4448aaf0 816static inline int pte_young(pte_t pte)
1da177e4 817{
b2fa47e6
MS
818#ifdef CONFIG_PGSTE
819 if (pte_val(pte) & _PAGE_SWR)
820 return 1;
821#endif
1da177e4
LT
822 return 0;
823}
824
1da177e4
LT
825/*
826 * pgd/pmd/pte modification functions
827 */
828
b2fa47e6 829static inline void pgd_clear(pgd_t *pgd)
5a216a20 830{
f4815ac6 831#ifdef CONFIG_64BIT
6252d702
MS
832 if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
833 pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
b2fa47e6 834#endif
5a216a20
MS
835}
836
b2fa47e6 837static inline void pud_clear(pud_t *pud)
1da177e4 838{
f4815ac6 839#ifdef CONFIG_64BIT
6252d702
MS
840 if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
841 pud_val(*pud) = _REGION3_ENTRY_EMPTY;
b2fa47e6 842#endif
1da177e4
LT
843}
844
b2fa47e6 845static inline void pmd_clear(pmd_t *pmdp)
1da177e4 846{
3610cce8 847 pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
1da177e4
LT
848}
849
4448aaf0 850static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
1da177e4 851{
9282ed92 852 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1da177e4
LT
853}
854
855/*
856 * The following pte modification functions only work if
857 * pte_present() is true. Undefined behaviour if not..
858 */
4448aaf0 859static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
1da177e4 860{
138c9021 861 pte_val(pte) &= _PAGE_CHG_MASK;
1da177e4 862 pte_val(pte) |= pgprot_val(newprot);
abf09bed
MS
863 if ((pte_val(pte) & _PAGE_SWC) && (pte_val(pte) & _PAGE_SWW))
864 pte_val(pte) &= ~_PAGE_RO;
1da177e4
LT
865 return pte;
866}
867
4448aaf0 868static inline pte_t pte_wrprotect(pte_t pte)
1da177e4 869{
abf09bed 870 pte_val(pte) &= ~_PAGE_SWW;
9282ed92 871 /* Do not clobber _PAGE_TYPE_NONE pages! */
1da177e4
LT
872 if (!(pte_val(pte) & _PAGE_INVALID))
873 pte_val(pte) |= _PAGE_RO;
874 return pte;
875}
876
4448aaf0 877static inline pte_t pte_mkwrite(pte_t pte)
1da177e4 878{
abf09bed
MS
879 pte_val(pte) |= _PAGE_SWW;
880 if (pte_val(pte) & _PAGE_SWC)
881 pte_val(pte) &= ~_PAGE_RO;
1da177e4
LT
882 return pte;
883}
884
4448aaf0 885static inline pte_t pte_mkclean(pte_t pte)
1da177e4 886{
b2fa47e6 887 pte_val(pte) &= ~_PAGE_SWC;
abf09bed
MS
888 /* Do not clobber _PAGE_TYPE_NONE pages! */
889 if (!(pte_val(pte) & _PAGE_INVALID))
890 pte_val(pte) |= _PAGE_RO;
1da177e4
LT
891 return pte;
892}
893
4448aaf0 894static inline pte_t pte_mkdirty(pte_t pte)
1da177e4 895{
abf09bed
MS
896 pte_val(pte) |= _PAGE_SWC;
897 if (pte_val(pte) & _PAGE_SWW)
898 pte_val(pte) &= ~_PAGE_RO;
1da177e4
LT
899 return pte;
900}
901
4448aaf0 902static inline pte_t pte_mkold(pte_t pte)
1da177e4 903{
b2fa47e6
MS
904#ifdef CONFIG_PGSTE
905 pte_val(pte) &= ~_PAGE_SWR;
906#endif
1da177e4
LT
907 return pte;
908}
909
4448aaf0 910static inline pte_t pte_mkyoung(pte_t pte)
1da177e4 911{
1da177e4
LT
912 return pte;
913}
914
7e675137
NP
915static inline pte_t pte_mkspecial(pte_t pte)
916{
a08cb629 917 pte_val(pte) |= _PAGE_SPECIAL;
7e675137
NP
918 return pte;
919}
920
84afdcee
HC
921#ifdef CONFIG_HUGETLB_PAGE
922static inline pte_t pte_mkhuge(pte_t pte)
923{
84afdcee
HC
924 pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
925 return pte;
926}
927#endif
928
15e86b0c 929/*
b2fa47e6 930 * Get (and clear) the user dirty bit for a pte.
15e86b0c 931 */
b2fa47e6
MS
932static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
933 pte_t *ptep)
15e86b0c 934{
b2fa47e6
MS
935 pgste_t pgste;
936 int dirty = 0;
937
938 if (mm_has_pgste(mm)) {
939 pgste = pgste_get_lock(ptep);
940 pgste = pgste_update_all(ptep, pgste);
941 dirty = !!(pgste_val(pgste) & KVM_UC_BIT);
942 pgste_val(pgste) &= ~KVM_UC_BIT;
943 pgste_set_unlock(ptep, pgste);
944 return dirty;
15e86b0c 945 }
15e86b0c
FF
946 return dirty;
947}
b2fa47e6
MS
948
949/*
950 * Get (and clear) the user referenced bit for a pte.
951 */
952static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
953 pte_t *ptep)
954{
955 pgste_t pgste;
956 int young = 0;
957
958 if (mm_has_pgste(mm)) {
959 pgste = pgste_get_lock(ptep);
960 pgste = pgste_update_young(ptep, pgste);
961 young = !!(pgste_val(pgste) & KVM_UR_BIT);
962 pgste_val(pgste) &= ~KVM_UR_BIT;
963 pgste_set_unlock(ptep, pgste);
964 }
965 return young;
966}
15e86b0c 967
ba8a9229
MS
968#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
969static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
970 unsigned long addr, pte_t *ptep)
1da177e4 971{
b2fa47e6
MS
972 pgste_t pgste;
973 pte_t pte;
974
975 if (mm_has_pgste(vma->vm_mm)) {
976 pgste = pgste_get_lock(ptep);
977 pgste = pgste_update_young(ptep, pgste);
978 pte = *ptep;
979 *ptep = pte_mkold(pte);
980 pgste_set_unlock(ptep, pgste);
981 return pte_young(pte);
982 }
1da177e4
LT
983 return 0;
984}
985
ba8a9229
MS
986#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
987static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
988 unsigned long address, pte_t *ptep)
1da177e4 989{
5b7baf05
CB
990 /* No need to flush TLB
991 * On s390 reference bits are in storage key and never in TLB
992 * With virtualization we handle the reference bit, without we
993 * we can simply return */
5b7baf05 994 return ptep_test_and_clear_young(vma, address, ptep);
1da177e4
LT
995}
996
9282ed92 997static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
1da177e4 998{
9282ed92 999 if (!(pte_val(*ptep) & _PAGE_INVALID)) {
f4815ac6 1000#ifndef CONFIG_64BIT
146e4b3c 1001 /* pto must point to the start of the segment table */
1da177e4 1002 pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
9282ed92
GS
1003#else
1004 /* ipte in zarch mode can do the math */
1005 pte_t *pto = ptep;
1006#endif
94c12cc7
MS
1007 asm volatile(
1008 " ipte %2,%3"
1009 : "=m" (*ptep) : "m" (*ptep),
1010 "a" (pto), "a" (address));
1da177e4 1011 }
9282ed92
GS
1012}
1013
ba8a9229
MS
1014/*
1015 * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1016 * both clear the TLB for the unmapped pte. The reason is that
1017 * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1018 * to modify an active pte. The sequence is
1019 * 1) ptep_get_and_clear
1020 * 2) set_pte_at
1021 * 3) flush_tlb_range
1022 * On s390 the tlb needs to get flushed with the modification of the pte
1023 * if the pte is active. The only way how this can be implemented is to
1024 * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1025 * is a nop.
1026 */
1027#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
b2fa47e6
MS
1028static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1029 unsigned long address, pte_t *ptep)
1030{
1031 pgste_t pgste;
1032 pte_t pte;
1033
1034 mm->context.flush_mm = 1;
1035 if (mm_has_pgste(mm))
1036 pgste = pgste_get_lock(ptep);
1037
1038 pte = *ptep;
1039 if (!mm_exclusive(mm))
1040 __ptep_ipte(address, ptep);
1041 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1042
1043 if (mm_has_pgste(mm)) {
1044 pgste = pgste_update_all(&pte, pgste);
1045 pgste_set_unlock(ptep, pgste);
1046 }
1047 return pte;
1048}
1049
1050#define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1051static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
1052 unsigned long address,
1053 pte_t *ptep)
1054{
1055 pte_t pte;
1056
1057 mm->context.flush_mm = 1;
1058 if (mm_has_pgste(mm))
1059 pgste_get_lock(ptep);
1060
1061 pte = *ptep;
1062 if (!mm_exclusive(mm))
1063 __ptep_ipte(address, ptep);
1064 return pte;
1065}
1066
1067static inline void ptep_modify_prot_commit(struct mm_struct *mm,
1068 unsigned long address,
1069 pte_t *ptep, pte_t pte)
1070{
abf09bed
MS
1071 if (mm_has_pgste(mm)) {
1072 pgste_set_pte(ptep, pte);
b2fa47e6 1073 pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE));
abf09bed
MS
1074 } else
1075 *ptep = pte;
b2fa47e6 1076}
ba8a9229
MS
1077
1078#define __HAVE_ARCH_PTEP_CLEAR_FLUSH
f0e47c22
MS
1079static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1080 unsigned long address, pte_t *ptep)
1081{
b2fa47e6
MS
1082 pgste_t pgste;
1083 pte_t pte;
1084
1085 if (mm_has_pgste(vma->vm_mm))
1086 pgste = pgste_get_lock(ptep);
1087
1088 pte = *ptep;
1089 __ptep_ipte(address, ptep);
1090 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1091
1092 if (mm_has_pgste(vma->vm_mm)) {
1093 pgste = pgste_update_all(&pte, pgste);
1094 pgste_set_unlock(ptep, pgste);
1095 }
1da177e4
LT
1096 return pte;
1097}
1098
ba8a9229
MS
1099/*
1100 * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1101 * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1102 * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1103 * cannot be accessed while the batched unmap is running. In this case
1104 * full==1 and a simple pte_clear is enough. See tlb.h.
1105 */
1106#define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1107static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
b2fa47e6 1108 unsigned long address,
ba8a9229 1109 pte_t *ptep, int full)
1da177e4 1110{
b2fa47e6
MS
1111 pgste_t pgste;
1112 pte_t pte;
1113
1114 if (mm_has_pgste(mm))
1115 pgste = pgste_get_lock(ptep);
ba8a9229 1116
b2fa47e6
MS
1117 pte = *ptep;
1118 if (!full)
1119 __ptep_ipte(address, ptep);
1120 pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1121
1122 if (mm_has_pgste(mm)) {
1123 pgste = pgste_update_all(&pte, pgste);
1124 pgste_set_unlock(ptep, pgste);
1125 }
ba8a9229 1126 return pte;
1da177e4
LT
1127}
1128
ba8a9229 1129#define __HAVE_ARCH_PTEP_SET_WRPROTECT
b2fa47e6
MS
1130static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1131 unsigned long address, pte_t *ptep)
1132{
1133 pgste_t pgste;
1134 pte_t pte = *ptep;
1135
1136 if (pte_write(pte)) {
1137 mm->context.flush_mm = 1;
1138 if (mm_has_pgste(mm))
1139 pgste = pgste_get_lock(ptep);
1140
1141 if (!mm_exclusive(mm))
1142 __ptep_ipte(address, ptep);
abf09bed 1143 pte = pte_wrprotect(pte);
b2fa47e6 1144
abf09bed
MS
1145 if (mm_has_pgste(mm)) {
1146 pgste_set_pte(ptep, pte);
b2fa47e6 1147 pgste_set_unlock(ptep, pgste);
abf09bed
MS
1148 } else
1149 *ptep = pte;
b2fa47e6
MS
1150 }
1151 return pte;
1152}
ba8a9229
MS
1153
1154#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
b2fa47e6
MS
1155static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1156 unsigned long address, pte_t *ptep,
1157 pte_t entry, int dirty)
1158{
1159 pgste_t pgste;
1160
1161 if (pte_same(*ptep, entry))
1162 return 0;
1163 if (mm_has_pgste(vma->vm_mm))
1164 pgste = pgste_get_lock(ptep);
1165
1166 __ptep_ipte(address, ptep);
b2fa47e6 1167
abf09bed
MS
1168 if (mm_has_pgste(vma->vm_mm)) {
1169 pgste_set_pte(ptep, entry);
b2fa47e6 1170 pgste_set_unlock(ptep, pgste);
abf09bed
MS
1171 } else
1172 *ptep = entry;
b2fa47e6
MS
1173 return 1;
1174}
1da177e4 1175
1da177e4
LT
1176/*
1177 * Conversion functions: convert a page and protection to a page entry,
1178 * and a page entry and page directory to the page they refer to.
1179 */
1180static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1181{
1182 pte_t __pte;
1183 pte_val(__pte) = physpage + pgprot_val(pgprot);
1184 return __pte;
1185}
1186
2dcea57a
HC
1187static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1188{
0b2b6e1d 1189 unsigned long physpage = page_to_phys(page);
abf09bed 1190 pte_t __pte = mk_pte_phys(physpage, pgprot);
1da177e4 1191
abf09bed
MS
1192 if ((pte_val(__pte) & _PAGE_SWW) && PageDirty(page)) {
1193 pte_val(__pte) |= _PAGE_SWC;
1194 pte_val(__pte) &= ~_PAGE_RO;
1195 }
1196 return __pte;
2dcea57a
HC
1197}
1198
190a1d72
MS
1199#define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1200#define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1201#define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1202#define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1da177e4 1203
190a1d72
MS
1204#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1205#define pgd_offset_k(address) pgd_offset(&init_mm, address)
1da177e4 1206
f4815ac6 1207#ifndef CONFIG_64BIT
1da177e4 1208
190a1d72
MS
1209#define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1210#define pud_deref(pmd) ({ BUG(); 0UL; })
1211#define pgd_deref(pmd) ({ BUG(); 0UL; })
46a82b2d 1212
190a1d72
MS
1213#define pud_offset(pgd, address) ((pud_t *) pgd)
1214#define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1da177e4 1215
f4815ac6 1216#else /* CONFIG_64BIT */
1da177e4 1217
190a1d72
MS
1218#define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1219#define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
5a216a20 1220#define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1da177e4 1221
5a216a20
MS
1222static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1223{
6252d702
MS
1224 pud_t *pud = (pud_t *) pgd;
1225 if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1226 pud = (pud_t *) pgd_deref(*pgd);
5a216a20
MS
1227 return pud + pud_index(address);
1228}
1da177e4 1229
190a1d72 1230static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1da177e4 1231{
6252d702
MS
1232 pmd_t *pmd = (pmd_t *) pud;
1233 if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1234 pmd = (pmd_t *) pud_deref(*pud);
190a1d72 1235 return pmd + pmd_index(address);
1da177e4
LT
1236}
1237
f4815ac6 1238#endif /* CONFIG_64BIT */
1da177e4 1239
190a1d72
MS
1240#define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1241#define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1242#define pte_page(x) pfn_to_page(pte_pfn(x))
1da177e4 1243
190a1d72 1244#define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1da177e4 1245
190a1d72
MS
1246/* Find an entry in the lowest level page table.. */
1247#define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1248#define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1da177e4 1249#define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1da177e4 1250#define pte_unmap(pte) do { } while (0)
1da177e4 1251
1ae1c1d0
GS
1252static inline void __pmd_idte(unsigned long address, pmd_t *pmdp)
1253{
1254 unsigned long sto = (unsigned long) pmdp -
1255 pmd_index(address) * sizeof(pmd_t);
1256
1257 if (!(pmd_val(*pmdp) & _SEGMENT_ENTRY_INV)) {
1258 asm volatile(
1259 " .insn rrf,0xb98e0000,%2,%3,0,0"
1260 : "=m" (*pmdp)
1261 : "m" (*pmdp), "a" (sto),
1262 "a" ((address & HPAGE_MASK))
1263 : "cc"
1264 );
1265 }
1266}
1267
106c992a 1268#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1ae1c1d0
GS
1269static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1270{
d8e7a33d
GS
1271 /*
1272 * pgprot is PAGE_NONE, PAGE_RO, or PAGE_RW (see __Pxxx / __Sxxx)
1273 * Convert to segment table entry format.
1274 */
1275 if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1276 return pgprot_val(SEGMENT_NONE);
1277 if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1278 return pgprot_val(SEGMENT_RO);
1279 return pgprot_val(SEGMENT_RW);
1ae1c1d0
GS
1280}
1281
1282static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1283{
1284 pmd_val(pmd) &= _SEGMENT_CHG_MASK;
1285 pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1286 return pmd;
1287}
1288
106c992a 1289static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1ae1c1d0 1290{
106c992a
GS
1291 pmd_t __pmd;
1292 pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1293 return __pmd;
1ae1c1d0
GS
1294}
1295
1296static inline pmd_t pmd_mkwrite(pmd_t pmd)
1297{
d8e7a33d
GS
1298 /* Do not clobber _HPAGE_TYPE_NONE pages! */
1299 if (!(pmd_val(pmd) & _SEGMENT_ENTRY_INV))
1300 pmd_val(pmd) &= ~_SEGMENT_ENTRY_RO;
1ae1c1d0
GS
1301 return pmd;
1302}
106c992a
GS
1303#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1304
1305#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1306
1307#define __HAVE_ARCH_PGTABLE_DEPOSIT
1308extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable);
1309
1310#define __HAVE_ARCH_PGTABLE_WITHDRAW
1311extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm);
1312
1313static inline int pmd_trans_splitting(pmd_t pmd)
1314{
1315 return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT;
1316}
1317
1318static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1319 pmd_t *pmdp, pmd_t entry)
1320{
1321 if (!(pmd_val(entry) & _SEGMENT_ENTRY_INV) && MACHINE_HAS_EDAT1)
1322 pmd_val(entry) |= _SEGMENT_ENTRY_CO;
1323 *pmdp = entry;
1324}
1325
1326static inline pmd_t pmd_mkhuge(pmd_t pmd)
1327{
1328 pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1329 return pmd;
1330}
1ae1c1d0
GS
1331
1332static inline pmd_t pmd_wrprotect(pmd_t pmd)
1333{
1334 pmd_val(pmd) |= _SEGMENT_ENTRY_RO;
1335 return pmd;
1336}
1337
1338static inline pmd_t pmd_mkdirty(pmd_t pmd)
1339{
1340 /* No dirty bit in the segment table entry. */
1341 return pmd;
1342}
1343
1344static inline pmd_t pmd_mkold(pmd_t pmd)
1345{
1346 /* No referenced bit in the segment table entry. */
1347 return pmd;
1348}
1349
1350static inline pmd_t pmd_mkyoung(pmd_t pmd)
1351{
1352 /* No referenced bit in the segment table entry. */
1353 return pmd;
1354}
1355
1356#define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1357static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1358 unsigned long address, pmd_t *pmdp)
1359{
1360 unsigned long pmd_addr = pmd_val(*pmdp) & HPAGE_MASK;
1361 long tmp, rc;
1362 int counter;
1363
1364 rc = 0;
1365 if (MACHINE_HAS_RRBM) {
1366 counter = PTRS_PER_PTE >> 6;
1367 asm volatile(
1368 "0: .insn rre,0xb9ae0000,%0,%3\n" /* rrbm */
1369 " ogr %1,%0\n"
1370 " la %3,0(%4,%3)\n"
1371 " brct %2,0b\n"
1372 : "=&d" (tmp), "+&d" (rc), "+d" (counter),
1373 "+a" (pmd_addr)
1374 : "a" (64 * 4096UL) : "cc");
1375 rc = !!rc;
1376 } else {
1377 counter = PTRS_PER_PTE;
1378 asm volatile(
1379 "0: rrbe 0,%2\n"
1380 " la %2,0(%3,%2)\n"
1381 " brc 12,1f\n"
1382 " lhi %0,1\n"
1383 "1: brct %1,0b\n"
1384 : "+d" (rc), "+d" (counter), "+a" (pmd_addr)
1385 : "a" (4096UL) : "cc");
1386 }
1387 return rc;
1388}
1389
1390#define __HAVE_ARCH_PMDP_GET_AND_CLEAR
1391static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
1392 unsigned long address, pmd_t *pmdp)
1393{
1394 pmd_t pmd = *pmdp;
1395
1396 __pmd_idte(address, pmdp);
1397 pmd_clear(pmdp);
1398 return pmd;
1399}
1400
1401#define __HAVE_ARCH_PMDP_CLEAR_FLUSH
1402static inline pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
1403 unsigned long address, pmd_t *pmdp)
1404{
1405 return pmdp_get_and_clear(vma->vm_mm, address, pmdp);
1406}
1407
1408#define __HAVE_ARCH_PMDP_INVALIDATE
1409static inline void pmdp_invalidate(struct vm_area_struct *vma,
1410 unsigned long address, pmd_t *pmdp)
1411{
1412 __pmd_idte(address, pmdp);
1413}
1414
be328650
GS
1415#define __HAVE_ARCH_PMDP_SET_WRPROTECT
1416static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1417 unsigned long address, pmd_t *pmdp)
1418{
1419 pmd_t pmd = *pmdp;
1420
1421 if (pmd_write(pmd)) {
1422 __pmd_idte(address, pmdp);
1423 set_pmd_at(mm, address, pmdp, pmd_wrprotect(pmd));
1424 }
1425}
1426
1ae1c1d0
GS
1427#define pfn_pmd(pfn, pgprot) mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1428#define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
1429
1430static inline int pmd_trans_huge(pmd_t pmd)
1431{
1432 return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1433}
1434
1435static inline int has_transparent_hugepage(void)
1436{
1437 return MACHINE_HAS_HPAGE ? 1 : 0;
1438}
1439
1440static inline unsigned long pmd_pfn(pmd_t pmd)
1441{
171c4006 1442 return pmd_val(pmd) >> PAGE_SHIFT;
1ae1c1d0 1443}
75077afb
GS
1444#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1445
1da177e4
LT
1446/*
1447 * 31 bit swap entry format:
1448 * A page-table entry has some bits we have to treat in a special way.
1449 * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1450 * exception will occur instead of a page translation exception. The
1451 * specifiation exception has the bad habit not to store necessary
1452 * information in the lowcore.
1453 * Bit 21 and bit 22 are the page invalid bit and the page protection
1454 * bit. We set both to indicate a swapped page.
1455 * Bit 30 and 31 are used to distinguish the different page types. For
1456 * a swapped page these bits need to be zero.
1457 * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1458 * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1459 * plus 24 for the offset.
1460 * 0| offset |0110|o|type |00|
1461 * 0 0000000001111111111 2222 2 22222 33
1462 * 0 1234567890123456789 0123 4 56789 01
1463 *
1464 * 64 bit swap entry format:
1465 * A page-table entry has some bits we have to treat in a special way.
1466 * Bits 52 and bit 55 have to be zero, otherwise an specification
1467 * exception will occur instead of a page translation exception. The
1468 * specifiation exception has the bad habit not to store necessary
1469 * information in the lowcore.
1470 * Bit 53 and bit 54 are the page invalid bit and the page protection
1471 * bit. We set both to indicate a swapped page.
1472 * Bit 62 and 63 are used to distinguish the different page types. For
1473 * a swapped page these bits need to be zero.
1474 * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1475 * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1476 * plus 56 for the offset.
1477 * | offset |0110|o|type |00|
1478 * 0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1479 * 0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1480 */
f4815ac6 1481#ifndef CONFIG_64BIT
1da177e4
LT
1482#define __SWP_OFFSET_MASK (~0UL >> 12)
1483#else
1484#define __SWP_OFFSET_MASK (~0UL >> 11)
1485#endif
4448aaf0 1486static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1da177e4
LT
1487{
1488 pte_t pte;
1489 offset &= __SWP_OFFSET_MASK;
9282ed92 1490 pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
1da177e4
LT
1491 ((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1492 return pte;
1493}
1494
1495#define __swp_type(entry) (((entry).val >> 2) & 0x1f)
1496#define __swp_offset(entry) (((entry).val >> 11) | (((entry).val >> 7) & 1))
1497#define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1498
1499#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
1500#define __swp_entry_to_pte(x) ((pte_t) { (x).val })
1501
f4815ac6 1502#ifndef CONFIG_64BIT
1da177e4 1503# define PTE_FILE_MAX_BITS 26
f4815ac6 1504#else /* CONFIG_64BIT */
1da177e4 1505# define PTE_FILE_MAX_BITS 59
f4815ac6 1506#endif /* CONFIG_64BIT */
1da177e4
LT
1507
1508#define pte_to_pgoff(__pte) \
1509 ((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1510
1511#define pgoff_to_pte(__off) \
1512 ((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
9282ed92 1513 | _PAGE_TYPE_FILE })
1da177e4
LT
1514
1515#endif /* !__ASSEMBLY__ */
1516
1517#define kern_addr_valid(addr) (1)
1518
17f34580
HC
1519extern int vmem_add_mapping(unsigned long start, unsigned long size);
1520extern int vmem_remove_mapping(unsigned long start, unsigned long size);
402b0862 1521extern int s390_enable_sie(void);
f4eb07c1 1522
1da177e4
LT
1523/*
1524 * No page table caches to initialise
1525 */
765a0cac
HC
1526static inline void pgtable_cache_init(void) { }
1527static inline void check_pgt_cache(void) { }
1da177e4 1528
1da177e4
LT
1529#include <asm-generic/pgtable.h>
1530
1531#endif /* _S390_PAGE_H */