xref: /linux/arch/s390/include/asm/pgtable.h (revision b43ab901d671e3e3cad425ea5e9a3c74e266dcdd)
1 /*
2  *  include/asm-s390/pgtable.h
3  *
4  *  S390 version
5  *    Copyright (C) 1999,2000 IBM Deutschland Entwicklung GmbH, IBM Corporation
6  *    Author(s): Hartmut Penner (hp@de.ibm.com)
7  *               Ulrich Weigand (weigand@de.ibm.com)
8  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
9  *
10  *  Derived from "include/asm-i386/pgtable.h"
11  */
12 
13 #ifndef _ASM_S390_PGTABLE_H
14 #define _ASM_S390_PGTABLE_H
15 
16 /*
17  * The Linux memory management assumes a three-level page table setup. For
18  * s390 31 bit we "fold" the mid level into the top-level page table, so
19  * that we physically have the same two-level page table as the s390 mmu
20  * expects in 31 bit mode. For s390 64 bit we use three of the five levels
21  * the hardware provides (region first and region second tables are not
22  * used).
23  *
24  * The "pgd_xxx()" functions are trivial for a folded two-level
25  * setup: the pgd is never bad, and a pmd always exists (as it's folded
26  * into the pgd entry)
27  *
28  * This file contains the functions and defines necessary to modify and use
29  * the S390 page table tree.
30  */
31 #ifndef __ASSEMBLY__
32 #include <linux/sched.h>
33 #include <linux/mm_types.h>
34 #include <asm/bug.h>
35 #include <asm/page.h>
36 
37 extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
38 extern void paging_init(void);
39 extern void vmem_map_init(void);
40 extern void fault_init(void);
41 
42 /*
43  * The S390 doesn't have any external MMU info: the kernel page
44  * tables contain all the necessary information.
45  */
46 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
47 
48 /*
49  * ZERO_PAGE is a global shared page that is always zero; used
50  * for zero-mapped memory areas etc..
51  */
52 
53 extern unsigned long empty_zero_page;
54 extern unsigned long zero_page_mask;
55 
56 #define ZERO_PAGE(vaddr) \
57 	(virt_to_page((void *)(empty_zero_page + \
58 	 (((unsigned long)(vaddr)) &zero_page_mask))))
59 
60 #define is_zero_pfn is_zero_pfn
61 static inline int is_zero_pfn(unsigned long pfn)
62 {
63 	extern unsigned long zero_pfn;
64 	unsigned long offset_from_zero_pfn = pfn - zero_pfn;
65 	return offset_from_zero_pfn <= (zero_page_mask >> PAGE_SHIFT);
66 }
67 
68 #define my_zero_pfn(addr)	page_to_pfn(ZERO_PAGE(addr))
69 
70 #endif /* !__ASSEMBLY__ */
71 
72 /*
73  * PMD_SHIFT determines the size of the area a second-level page
74  * table can map
75  * PGDIR_SHIFT determines what a third-level page table entry can map
76  */
77 #ifndef __s390x__
78 # define PMD_SHIFT	20
79 # define PUD_SHIFT	20
80 # define PGDIR_SHIFT	20
81 #else /* __s390x__ */
82 # define PMD_SHIFT	20
83 # define PUD_SHIFT	31
84 # define PGDIR_SHIFT	42
85 #endif /* __s390x__ */
86 
87 #define PMD_SIZE        (1UL << PMD_SHIFT)
88 #define PMD_MASK        (~(PMD_SIZE-1))
89 #define PUD_SIZE	(1UL << PUD_SHIFT)
90 #define PUD_MASK	(~(PUD_SIZE-1))
91 #define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
92 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
93 
94 /*
95  * entries per page directory level: the S390 is two-level, so
96  * we don't really have any PMD directory physically.
97  * for S390 segment-table entries are combined to one PGD
98  * that leads to 1024 pte per pgd
99  */
100 #define PTRS_PER_PTE	256
101 #ifndef __s390x__
102 #define PTRS_PER_PMD	1
103 #define PTRS_PER_PUD	1
104 #else /* __s390x__ */
105 #define PTRS_PER_PMD	2048
106 #define PTRS_PER_PUD	2048
107 #endif /* __s390x__ */
108 #define PTRS_PER_PGD	2048
109 
110 #define FIRST_USER_ADDRESS  0
111 
112 #define pte_ERROR(e) \
113 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
114 #define pmd_ERROR(e) \
115 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
116 #define pud_ERROR(e) \
117 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
118 #define pgd_ERROR(e) \
119 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
120 
121 #ifndef __ASSEMBLY__
122 /*
123  * The vmalloc area will always be on the topmost area of the kernel
124  * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc,
125  * which should be enough for any sane case.
126  * By putting vmalloc at the top, we maximise the gap between physical
127  * memory and vmalloc to catch misplaced memory accesses. As a side
128  * effect, this also makes sure that 64 bit module code cannot be used
129  * as system call address.
130  */
131 extern unsigned long VMALLOC_START;
132 extern unsigned long VMALLOC_END;
133 extern struct page *vmemmap;
134 
135 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
136 
137 /*
138  * A 31 bit pagetable entry of S390 has following format:
139  *  |   PFRA          |    |  OS  |
140  * 0                   0IP0
141  * 00000000001111111111222222222233
142  * 01234567890123456789012345678901
143  *
144  * I Page-Invalid Bit:    Page is not available for address-translation
145  * P Page-Protection Bit: Store access not possible for page
146  *
147  * A 31 bit segmenttable entry of S390 has following format:
148  *  |   P-table origin      |  |PTL
149  * 0                         IC
150  * 00000000001111111111222222222233
151  * 01234567890123456789012345678901
152  *
153  * I Segment-Invalid Bit:    Segment is not available for address-translation
154  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
155  * PTL Page-Table-Length:    Page-table length (PTL+1*16 entries -> up to 256)
156  *
157  * The 31 bit segmenttable origin of S390 has following format:
158  *
159  *  |S-table origin   |     | STL |
160  * X                   **GPS
161  * 00000000001111111111222222222233
162  * 01234567890123456789012345678901
163  *
164  * X Space-Switch event:
165  * G Segment-Invalid Bit:     *
166  * P Private-Space Bit:       Segment is not private (PoP 3-30)
167  * S Storage-Alteration:
168  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
169  *
170  * A 64 bit pagetable entry of S390 has following format:
171  * |			 PFRA			      |0IPC|  OS  |
172  * 0000000000111111111122222222223333333333444444444455555555556666
173  * 0123456789012345678901234567890123456789012345678901234567890123
174  *
175  * I Page-Invalid Bit:    Page is not available for address-translation
176  * P Page-Protection Bit: Store access not possible for page
177  * C Change-bit override: HW is not required to set change bit
178  *
179  * A 64 bit segmenttable entry of S390 has following format:
180  * |        P-table origin                              |      TT
181  * 0000000000111111111122222222223333333333444444444455555555556666
182  * 0123456789012345678901234567890123456789012345678901234567890123
183  *
184  * I Segment-Invalid Bit:    Segment is not available for address-translation
185  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
186  * P Page-Protection Bit: Store access not possible for page
187  * TT Type 00
188  *
189  * A 64 bit region table entry of S390 has following format:
190  * |        S-table origin                             |   TF  TTTL
191  * 0000000000111111111122222222223333333333444444444455555555556666
192  * 0123456789012345678901234567890123456789012345678901234567890123
193  *
194  * I Segment-Invalid Bit:    Segment is not available for address-translation
195  * TT Type 01
196  * TF
197  * TL Table length
198  *
199  * The 64 bit regiontable origin of S390 has following format:
200  * |      region table origon                          |       DTTL
201  * 0000000000111111111122222222223333333333444444444455555555556666
202  * 0123456789012345678901234567890123456789012345678901234567890123
203  *
204  * X Space-Switch event:
205  * G Segment-Invalid Bit:
206  * P Private-Space Bit:
207  * S Storage-Alteration:
208  * R Real space
209  * TL Table-Length:
210  *
211  * A storage key has the following format:
212  * | ACC |F|R|C|0|
213  *  0   3 4 5 6 7
214  * ACC: access key
215  * F  : fetch protection bit
216  * R  : referenced bit
217  * C  : changed bit
218  */
219 
220 /* Hardware bits in the page table entry */
221 #define _PAGE_CO	0x100		/* HW Change-bit override */
222 #define _PAGE_RO	0x200		/* HW read-only bit  */
223 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
224 
225 /* Software bits in the page table entry */
226 #define _PAGE_SWT	0x001		/* SW pte type bit t */
227 #define _PAGE_SWX	0x002		/* SW pte type bit x */
228 #define _PAGE_SWC	0x004		/* SW pte changed bit (for KVM) */
229 #define _PAGE_SWR	0x008		/* SW pte referenced bit (for KVM) */
230 #define _PAGE_SPECIAL	0x010		/* SW associated with special page */
231 #define __HAVE_ARCH_PTE_SPECIAL
232 
233 /* Set of bits not changed in pte_modify */
234 #define _PAGE_CHG_MASK	(PAGE_MASK | _PAGE_SPECIAL | _PAGE_SWC | _PAGE_SWR)
235 
236 /* Six different types of pages. */
237 #define _PAGE_TYPE_EMPTY	0x400
238 #define _PAGE_TYPE_NONE		0x401
239 #define _PAGE_TYPE_SWAP		0x403
240 #define _PAGE_TYPE_FILE		0x601	/* bit 0x002 is used for offset !! */
241 #define _PAGE_TYPE_RO		0x200
242 #define _PAGE_TYPE_RW		0x000
243 
244 /*
245  * Only four types for huge pages, using the invalid bit and protection bit
246  * of a segment table entry.
247  */
248 #define _HPAGE_TYPE_EMPTY	0x020	/* _SEGMENT_ENTRY_INV */
249 #define _HPAGE_TYPE_NONE	0x220
250 #define _HPAGE_TYPE_RO		0x200	/* _SEGMENT_ENTRY_RO  */
251 #define _HPAGE_TYPE_RW		0x000
252 
253 /*
254  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
255  * pte_none and pte_file to find out the pte type WITHOUT holding the page
256  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
257  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
258  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
259  * This change is done while holding the lock, but the intermediate step
260  * of a previously valid pte with the hw invalid bit set can be observed by
261  * handle_pte_fault. That makes it necessary that all valid pte types with
262  * the hw invalid bit set must be distinguishable from the four pte types
263  * empty, none, swap and file.
264  *
265  *			irxt  ipte  irxt
266  * _PAGE_TYPE_EMPTY	1000   ->   1000
267  * _PAGE_TYPE_NONE	1001   ->   1001
268  * _PAGE_TYPE_SWAP	1011   ->   1011
269  * _PAGE_TYPE_FILE	11?1   ->   11?1
270  * _PAGE_TYPE_RO	0100   ->   1100
271  * _PAGE_TYPE_RW	0000   ->   1000
272  *
273  * pte_none is true for bits combinations 1000, 1010, 1100, 1110
274  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
275  * pte_file is true for bits combinations 1101, 1111
276  * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
277  */
278 
279 #ifndef __s390x__
280 
281 /* Bits in the segment table address-space-control-element */
282 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
283 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
284 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
285 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
286 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
287 
288 /* Bits in the segment table entry */
289 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
290 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
291 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
292 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
293 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
294 
295 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
296 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
297 
298 /* Page status table bits for virtualization */
299 #define RCP_ACC_BITS	0xf0000000UL
300 #define RCP_FP_BIT	0x08000000UL
301 #define RCP_PCL_BIT	0x00800000UL
302 #define RCP_HR_BIT	0x00400000UL
303 #define RCP_HC_BIT	0x00200000UL
304 #define RCP_GR_BIT	0x00040000UL
305 #define RCP_GC_BIT	0x00020000UL
306 
307 /* User dirty / referenced bit for KVM's migration feature */
308 #define KVM_UR_BIT	0x00008000UL
309 #define KVM_UC_BIT	0x00004000UL
310 
311 #else /* __s390x__ */
312 
313 /* Bits in the segment/region table address-space-control-element */
314 #define _ASCE_ORIGIN		~0xfffUL/* segment table origin		    */
315 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
316 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
317 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
318 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
319 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
320 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
321 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
322 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
323 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
324 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
325 
326 /* Bits in the region table entry */
327 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
328 #define _REGION_ENTRY_INV	0x20	/* invalid region table entry	    */
329 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
330 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
331 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
332 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
333 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
334 
335 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
336 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
337 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
338 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
339 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
340 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
341 
342 /* Bits in the segment table entry */
343 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* segment table origin		    */
344 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
345 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
346 
347 #define _SEGMENT_ENTRY		(0)
348 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
349 
350 #define _SEGMENT_ENTRY_LARGE	0x400	/* STE-format control, large page   */
351 #define _SEGMENT_ENTRY_CO	0x100	/* change-recording override   */
352 
353 /* Page status table bits for virtualization */
354 #define RCP_ACC_BITS	0xf000000000000000UL
355 #define RCP_FP_BIT	0x0800000000000000UL
356 #define RCP_PCL_BIT	0x0080000000000000UL
357 #define RCP_HR_BIT	0x0040000000000000UL
358 #define RCP_HC_BIT	0x0020000000000000UL
359 #define RCP_GR_BIT	0x0004000000000000UL
360 #define RCP_GC_BIT	0x0002000000000000UL
361 
362 /* User dirty / referenced bit for KVM's migration feature */
363 #define KVM_UR_BIT	0x0000800000000000UL
364 #define KVM_UC_BIT	0x0000400000000000UL
365 
366 #endif /* __s390x__ */
367 
368 /*
369  * A user page table pointer has the space-switch-event bit, the
370  * private-space-control bit and the storage-alteration-event-control
371  * bit set. A kernel page table pointer doesn't need them.
372  */
373 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
374 				 _ASCE_ALT_EVENT)
375 
376 /*
377  * Page protection definitions.
378  */
379 #define PAGE_NONE	__pgprot(_PAGE_TYPE_NONE)
380 #define PAGE_RO		__pgprot(_PAGE_TYPE_RO)
381 #define PAGE_RW		__pgprot(_PAGE_TYPE_RW)
382 
383 #define PAGE_KERNEL	PAGE_RW
384 #define PAGE_COPY	PAGE_RO
385 
386 /*
387  * On s390 the page table entry has an invalid bit and a read-only bit.
388  * Read permission implies execute permission and write permission
389  * implies read permission.
390  */
391          /*xwr*/
392 #define __P000	PAGE_NONE
393 #define __P001	PAGE_RO
394 #define __P010	PAGE_RO
395 #define __P011	PAGE_RO
396 #define __P100	PAGE_RO
397 #define __P101	PAGE_RO
398 #define __P110	PAGE_RO
399 #define __P111	PAGE_RO
400 
401 #define __S000	PAGE_NONE
402 #define __S001	PAGE_RO
403 #define __S010	PAGE_RW
404 #define __S011	PAGE_RW
405 #define __S100	PAGE_RO
406 #define __S101	PAGE_RO
407 #define __S110	PAGE_RW
408 #define __S111	PAGE_RW
409 
410 static inline int mm_exclusive(struct mm_struct *mm)
411 {
412 	return likely(mm == current->active_mm &&
413 		      atomic_read(&mm->context.attach_count) <= 1);
414 }
415 
416 static inline int mm_has_pgste(struct mm_struct *mm)
417 {
418 #ifdef CONFIG_PGSTE
419 	if (unlikely(mm->context.has_pgste))
420 		return 1;
421 #endif
422 	return 0;
423 }
424 /*
425  * pgd/pmd/pte query functions
426  */
427 #ifndef __s390x__
428 
429 static inline int pgd_present(pgd_t pgd) { return 1; }
430 static inline int pgd_none(pgd_t pgd)    { return 0; }
431 static inline int pgd_bad(pgd_t pgd)     { return 0; }
432 
433 static inline int pud_present(pud_t pud) { return 1; }
434 static inline int pud_none(pud_t pud)	 { return 0; }
435 static inline int pud_bad(pud_t pud)	 { return 0; }
436 
437 #else /* __s390x__ */
438 
439 static inline int pgd_present(pgd_t pgd)
440 {
441 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
442 		return 1;
443 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
444 }
445 
446 static inline int pgd_none(pgd_t pgd)
447 {
448 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
449 		return 0;
450 	return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL;
451 }
452 
453 static inline int pgd_bad(pgd_t pgd)
454 {
455 	/*
456 	 * With dynamic page table levels the pgd can be a region table
457 	 * entry or a segment table entry. Check for the bit that are
458 	 * invalid for either table entry.
459 	 */
460 	unsigned long mask =
461 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
462 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
463 	return (pgd_val(pgd) & mask) != 0;
464 }
465 
466 static inline int pud_present(pud_t pud)
467 {
468 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
469 		return 1;
470 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
471 }
472 
473 static inline int pud_none(pud_t pud)
474 {
475 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
476 		return 0;
477 	return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
478 }
479 
480 static inline int pud_bad(pud_t pud)
481 {
482 	/*
483 	 * With dynamic page table levels the pud can be a region table
484 	 * entry or a segment table entry. Check for the bit that are
485 	 * invalid for either table entry.
486 	 */
487 	unsigned long mask =
488 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
489 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
490 	return (pud_val(pud) & mask) != 0;
491 }
492 
493 #endif /* __s390x__ */
494 
495 static inline int pmd_present(pmd_t pmd)
496 {
497 	return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL;
498 }
499 
500 static inline int pmd_none(pmd_t pmd)
501 {
502 	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL;
503 }
504 
505 static inline int pmd_bad(pmd_t pmd)
506 {
507 	unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
508 	return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
509 }
510 
511 static inline int pte_none(pte_t pte)
512 {
513 	return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
514 }
515 
516 static inline int pte_present(pte_t pte)
517 {
518 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
519 	return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
520 		(!(pte_val(pte) & _PAGE_INVALID) &&
521 		 !(pte_val(pte) & _PAGE_SWT));
522 }
523 
524 static inline int pte_file(pte_t pte)
525 {
526 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
527 	return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
528 }
529 
530 static inline int pte_special(pte_t pte)
531 {
532 	return (pte_val(pte) & _PAGE_SPECIAL);
533 }
534 
535 #define __HAVE_ARCH_PTE_SAME
536 static inline int pte_same(pte_t a, pte_t b)
537 {
538 	return pte_val(a) == pte_val(b);
539 }
540 
541 static inline pgste_t pgste_get_lock(pte_t *ptep)
542 {
543 	unsigned long new = 0;
544 #ifdef CONFIG_PGSTE
545 	unsigned long old;
546 
547 	preempt_disable();
548 	asm(
549 		"	lg	%0,%2\n"
550 		"0:	lgr	%1,%0\n"
551 		"	nihh	%0,0xff7f\n"	/* clear RCP_PCL_BIT in old */
552 		"	oihh	%1,0x0080\n"	/* set RCP_PCL_BIT in new */
553 		"	csg	%0,%1,%2\n"
554 		"	jl	0b\n"
555 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
556 		: "Q" (ptep[PTRS_PER_PTE]) : "cc");
557 #endif
558 	return __pgste(new);
559 }
560 
561 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
562 {
563 #ifdef CONFIG_PGSTE
564 	asm(
565 		"	nihh	%1,0xff7f\n"	/* clear RCP_PCL_BIT */
566 		"	stg	%1,%0\n"
567 		: "=Q" (ptep[PTRS_PER_PTE])
568 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc");
569 	preempt_enable();
570 #endif
571 }
572 
573 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
574 {
575 #ifdef CONFIG_PGSTE
576 	unsigned long address, bits;
577 	unsigned char skey;
578 
579 	if (!pte_present(*ptep))
580 		return pgste;
581 	address = pte_val(*ptep) & PAGE_MASK;
582 	skey = page_get_storage_key(address);
583 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
584 	/* Clear page changed & referenced bit in the storage key */
585 	if (bits & _PAGE_CHANGED)
586 		page_set_storage_key(address, skey ^ bits, 1);
587 	else if (bits)
588 		page_reset_referenced(address);
589 	/* Transfer page changed & referenced bit to guest bits in pgste */
590 	pgste_val(pgste) |= bits << 48;		/* RCP_GR_BIT & RCP_GC_BIT */
591 	/* Get host changed & referenced bits from pgste */
592 	bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52;
593 	/* Clear host bits in pgste. */
594 	pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT);
595 	pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT);
596 	/* Copy page access key and fetch protection bit to pgste */
597 	pgste_val(pgste) |=
598 		(unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
599 	/* Transfer changed and referenced to kvm user bits */
600 	pgste_val(pgste) |= bits << 45;		/* KVM_UR_BIT & KVM_UC_BIT */
601 	/* Transfer changed & referenced to pte sofware bits */
602 	pte_val(*ptep) |= bits << 1;		/* _PAGE_SWR & _PAGE_SWC */
603 #endif
604 	return pgste;
605 
606 }
607 
608 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
609 {
610 #ifdef CONFIG_PGSTE
611 	int young;
612 
613 	if (!pte_present(*ptep))
614 		return pgste;
615 	young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK);
616 	/* Transfer page referenced bit to pte software bit (host view) */
617 	if (young || (pgste_val(pgste) & RCP_HR_BIT))
618 		pte_val(*ptep) |= _PAGE_SWR;
619 	/* Clear host referenced bit in pgste. */
620 	pgste_val(pgste) &= ~RCP_HR_BIT;
621 	/* Transfer page referenced bit to guest bit in pgste */
622 	pgste_val(pgste) |= (unsigned long) young << 50; /* set RCP_GR_BIT */
623 #endif
624 	return pgste;
625 
626 }
627 
628 static inline void pgste_set_pte(pte_t *ptep, pgste_t pgste, pte_t entry)
629 {
630 #ifdef CONFIG_PGSTE
631 	unsigned long address;
632 	unsigned long okey, nkey;
633 
634 	if (!pte_present(entry))
635 		return;
636 	address = pte_val(entry) & PAGE_MASK;
637 	okey = nkey = page_get_storage_key(address);
638 	nkey &= ~(_PAGE_ACC_BITS | _PAGE_FP_BIT);
639 	/* Set page access key and fetch protection bit from pgste */
640 	nkey |= (pgste_val(pgste) & (RCP_ACC_BITS | RCP_FP_BIT)) >> 56;
641 	if (okey != nkey)
642 		page_set_storage_key(address, nkey, 1);
643 #endif
644 }
645 
646 /**
647  * struct gmap_struct - guest address space
648  * @mm: pointer to the parent mm_struct
649  * @table: pointer to the page directory
650  * @asce: address space control element for gmap page table
651  * @crst_list: list of all crst tables used in the guest address space
652  */
653 struct gmap {
654 	struct list_head list;
655 	struct mm_struct *mm;
656 	unsigned long *table;
657 	unsigned long asce;
658 	struct list_head crst_list;
659 };
660 
661 /**
662  * struct gmap_rmap - reverse mapping for segment table entries
663  * @next: pointer to the next gmap_rmap structure in the list
664  * @entry: pointer to a segment table entry
665  */
666 struct gmap_rmap {
667 	struct list_head list;
668 	unsigned long *entry;
669 };
670 
671 /**
672  * struct gmap_pgtable - gmap information attached to a page table
673  * @vmaddr: address of the 1MB segment in the process virtual memory
674  * @mapper: list of segment table entries maping a page table
675  */
676 struct gmap_pgtable {
677 	unsigned long vmaddr;
678 	struct list_head mapper;
679 };
680 
681 struct gmap *gmap_alloc(struct mm_struct *mm);
682 void gmap_free(struct gmap *gmap);
683 void gmap_enable(struct gmap *gmap);
684 void gmap_disable(struct gmap *gmap);
685 int gmap_map_segment(struct gmap *gmap, unsigned long from,
686 		     unsigned long to, unsigned long length);
687 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
688 unsigned long __gmap_fault(unsigned long address, struct gmap *);
689 unsigned long gmap_fault(unsigned long address, struct gmap *);
690 void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
691 
692 /*
693  * Certain architectures need to do special things when PTEs
694  * within a page table are directly modified.  Thus, the following
695  * hook is made available.
696  */
697 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
698 			      pte_t *ptep, pte_t entry)
699 {
700 	pgste_t pgste;
701 
702 	if (mm_has_pgste(mm)) {
703 		pgste = pgste_get_lock(ptep);
704 		pgste_set_pte(ptep, pgste, entry);
705 		*ptep = entry;
706 		pgste_set_unlock(ptep, pgste);
707 	} else
708 		*ptep = entry;
709 }
710 
711 /*
712  * query functions pte_write/pte_dirty/pte_young only work if
713  * pte_present() is true. Undefined behaviour if not..
714  */
715 static inline int pte_write(pte_t pte)
716 {
717 	return (pte_val(pte) & _PAGE_RO) == 0;
718 }
719 
720 static inline int pte_dirty(pte_t pte)
721 {
722 #ifdef CONFIG_PGSTE
723 	if (pte_val(pte) & _PAGE_SWC)
724 		return 1;
725 #endif
726 	return 0;
727 }
728 
729 static inline int pte_young(pte_t pte)
730 {
731 #ifdef CONFIG_PGSTE
732 	if (pte_val(pte) & _PAGE_SWR)
733 		return 1;
734 #endif
735 	return 0;
736 }
737 
738 /*
739  * pgd/pmd/pte modification functions
740  */
741 
742 static inline void pgd_clear(pgd_t *pgd)
743 {
744 #ifdef __s390x__
745 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
746 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
747 #endif
748 }
749 
750 static inline void pud_clear(pud_t *pud)
751 {
752 #ifdef __s390x__
753 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
754 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
755 #endif
756 }
757 
758 static inline void pmd_clear(pmd_t *pmdp)
759 {
760 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
761 }
762 
763 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
764 {
765 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
766 }
767 
768 /*
769  * The following pte modification functions only work if
770  * pte_present() is true. Undefined behaviour if not..
771  */
772 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
773 {
774 	pte_val(pte) &= _PAGE_CHG_MASK;
775 	pte_val(pte) |= pgprot_val(newprot);
776 	return pte;
777 }
778 
779 static inline pte_t pte_wrprotect(pte_t pte)
780 {
781 	/* Do not clobber _PAGE_TYPE_NONE pages!  */
782 	if (!(pte_val(pte) & _PAGE_INVALID))
783 		pte_val(pte) |= _PAGE_RO;
784 	return pte;
785 }
786 
787 static inline pte_t pte_mkwrite(pte_t pte)
788 {
789 	pte_val(pte) &= ~_PAGE_RO;
790 	return pte;
791 }
792 
793 static inline pte_t pte_mkclean(pte_t pte)
794 {
795 #ifdef CONFIG_PGSTE
796 	pte_val(pte) &= ~_PAGE_SWC;
797 #endif
798 	return pte;
799 }
800 
801 static inline pte_t pte_mkdirty(pte_t pte)
802 {
803 	return pte;
804 }
805 
806 static inline pte_t pte_mkold(pte_t pte)
807 {
808 #ifdef CONFIG_PGSTE
809 	pte_val(pte) &= ~_PAGE_SWR;
810 #endif
811 	return pte;
812 }
813 
814 static inline pte_t pte_mkyoung(pte_t pte)
815 {
816 	return pte;
817 }
818 
819 static inline pte_t pte_mkspecial(pte_t pte)
820 {
821 	pte_val(pte) |= _PAGE_SPECIAL;
822 	return pte;
823 }
824 
825 #ifdef CONFIG_HUGETLB_PAGE
826 static inline pte_t pte_mkhuge(pte_t pte)
827 {
828 	/*
829 	 * PROT_NONE needs to be remapped from the pte type to the ste type.
830 	 * The HW invalid bit is also different for pte and ste. The pte
831 	 * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE
832 	 * bit, so we don't have to clear it.
833 	 */
834 	if (pte_val(pte) & _PAGE_INVALID) {
835 		if (pte_val(pte) & _PAGE_SWT)
836 			pte_val(pte) |= _HPAGE_TYPE_NONE;
837 		pte_val(pte) |= _SEGMENT_ENTRY_INV;
838 	}
839 	/*
840 	 * Clear SW pte bits SWT and SWX, there are no SW bits in a segment
841 	 * table entry.
842 	 */
843 	pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX);
844 	/*
845 	 * Also set the change-override bit because we don't need dirty bit
846 	 * tracking for hugetlbfs pages.
847 	 */
848 	pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
849 	return pte;
850 }
851 #endif
852 
853 /*
854  * Get (and clear) the user dirty bit for a pte.
855  */
856 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
857 						 pte_t *ptep)
858 {
859 	pgste_t pgste;
860 	int dirty = 0;
861 
862 	if (mm_has_pgste(mm)) {
863 		pgste = pgste_get_lock(ptep);
864 		pgste = pgste_update_all(ptep, pgste);
865 		dirty = !!(pgste_val(pgste) & KVM_UC_BIT);
866 		pgste_val(pgste) &= ~KVM_UC_BIT;
867 		pgste_set_unlock(ptep, pgste);
868 		return dirty;
869 	}
870 	return dirty;
871 }
872 
873 /*
874  * Get (and clear) the user referenced bit for a pte.
875  */
876 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
877 						 pte_t *ptep)
878 {
879 	pgste_t pgste;
880 	int young = 0;
881 
882 	if (mm_has_pgste(mm)) {
883 		pgste = pgste_get_lock(ptep);
884 		pgste = pgste_update_young(ptep, pgste);
885 		young = !!(pgste_val(pgste) & KVM_UR_BIT);
886 		pgste_val(pgste) &= ~KVM_UR_BIT;
887 		pgste_set_unlock(ptep, pgste);
888 	}
889 	return young;
890 }
891 
892 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
893 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
894 					    unsigned long addr, pte_t *ptep)
895 {
896 	pgste_t pgste;
897 	pte_t pte;
898 
899 	if (mm_has_pgste(vma->vm_mm)) {
900 		pgste = pgste_get_lock(ptep);
901 		pgste = pgste_update_young(ptep, pgste);
902 		pte = *ptep;
903 		*ptep = pte_mkold(pte);
904 		pgste_set_unlock(ptep, pgste);
905 		return pte_young(pte);
906 	}
907 	return 0;
908 }
909 
910 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
911 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
912 					 unsigned long address, pte_t *ptep)
913 {
914 	/* No need to flush TLB
915 	 * On s390 reference bits are in storage key and never in TLB
916 	 * With virtualization we handle the reference bit, without we
917 	 * we can simply return */
918 	return ptep_test_and_clear_young(vma, address, ptep);
919 }
920 
921 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
922 {
923 	if (!(pte_val(*ptep) & _PAGE_INVALID)) {
924 #ifndef __s390x__
925 		/* pto must point to the start of the segment table */
926 		pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
927 #else
928 		/* ipte in zarch mode can do the math */
929 		pte_t *pto = ptep;
930 #endif
931 		asm volatile(
932 			"	ipte	%2,%3"
933 			: "=m" (*ptep) : "m" (*ptep),
934 			  "a" (pto), "a" (address));
935 	}
936 }
937 
938 /*
939  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
940  * both clear the TLB for the unmapped pte. The reason is that
941  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
942  * to modify an active pte. The sequence is
943  *   1) ptep_get_and_clear
944  *   2) set_pte_at
945  *   3) flush_tlb_range
946  * On s390 the tlb needs to get flushed with the modification of the pte
947  * if the pte is active. The only way how this can be implemented is to
948  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
949  * is a nop.
950  */
951 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
952 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
953 				       unsigned long address, pte_t *ptep)
954 {
955 	pgste_t pgste;
956 	pte_t pte;
957 
958 	mm->context.flush_mm = 1;
959 	if (mm_has_pgste(mm))
960 		pgste = pgste_get_lock(ptep);
961 
962 	pte = *ptep;
963 	if (!mm_exclusive(mm))
964 		__ptep_ipte(address, ptep);
965 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
966 
967 	if (mm_has_pgste(mm)) {
968 		pgste = pgste_update_all(&pte, pgste);
969 		pgste_set_unlock(ptep, pgste);
970 	}
971 	return pte;
972 }
973 
974 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
975 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
976 					   unsigned long address,
977 					   pte_t *ptep)
978 {
979 	pte_t pte;
980 
981 	mm->context.flush_mm = 1;
982 	if (mm_has_pgste(mm))
983 		pgste_get_lock(ptep);
984 
985 	pte = *ptep;
986 	if (!mm_exclusive(mm))
987 		__ptep_ipte(address, ptep);
988 	return pte;
989 }
990 
991 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
992 					   unsigned long address,
993 					   pte_t *ptep, pte_t pte)
994 {
995 	*ptep = pte;
996 	if (mm_has_pgste(mm))
997 		pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE));
998 }
999 
1000 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1001 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1002 				     unsigned long address, pte_t *ptep)
1003 {
1004 	pgste_t pgste;
1005 	pte_t pte;
1006 
1007 	if (mm_has_pgste(vma->vm_mm))
1008 		pgste = pgste_get_lock(ptep);
1009 
1010 	pte = *ptep;
1011 	__ptep_ipte(address, ptep);
1012 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1013 
1014 	if (mm_has_pgste(vma->vm_mm)) {
1015 		pgste = pgste_update_all(&pte, pgste);
1016 		pgste_set_unlock(ptep, pgste);
1017 	}
1018 	return pte;
1019 }
1020 
1021 /*
1022  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1023  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1024  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1025  * cannot be accessed while the batched unmap is running. In this case
1026  * full==1 and a simple pte_clear is enough. See tlb.h.
1027  */
1028 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1029 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1030 					    unsigned long address,
1031 					    pte_t *ptep, int full)
1032 {
1033 	pgste_t pgste;
1034 	pte_t pte;
1035 
1036 	if (mm_has_pgste(mm))
1037 		pgste = pgste_get_lock(ptep);
1038 
1039 	pte = *ptep;
1040 	if (!full)
1041 		__ptep_ipte(address, ptep);
1042 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1043 
1044 	if (mm_has_pgste(mm)) {
1045 		pgste = pgste_update_all(&pte, pgste);
1046 		pgste_set_unlock(ptep, pgste);
1047 	}
1048 	return pte;
1049 }
1050 
1051 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1052 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1053 				       unsigned long address, pte_t *ptep)
1054 {
1055 	pgste_t pgste;
1056 	pte_t pte = *ptep;
1057 
1058 	if (pte_write(pte)) {
1059 		mm->context.flush_mm = 1;
1060 		if (mm_has_pgste(mm))
1061 			pgste = pgste_get_lock(ptep);
1062 
1063 		if (!mm_exclusive(mm))
1064 			__ptep_ipte(address, ptep);
1065 		*ptep = pte_wrprotect(pte);
1066 
1067 		if (mm_has_pgste(mm))
1068 			pgste_set_unlock(ptep, pgste);
1069 	}
1070 	return pte;
1071 }
1072 
1073 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1074 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1075 					unsigned long address, pte_t *ptep,
1076 					pte_t entry, int dirty)
1077 {
1078 	pgste_t pgste;
1079 
1080 	if (pte_same(*ptep, entry))
1081 		return 0;
1082 	if (mm_has_pgste(vma->vm_mm))
1083 		pgste = pgste_get_lock(ptep);
1084 
1085 	__ptep_ipte(address, ptep);
1086 	*ptep = entry;
1087 
1088 	if (mm_has_pgste(vma->vm_mm))
1089 		pgste_set_unlock(ptep, pgste);
1090 	return 1;
1091 }
1092 
1093 /*
1094  * Conversion functions: convert a page and protection to a page entry,
1095  * and a page entry and page directory to the page they refer to.
1096  */
1097 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1098 {
1099 	pte_t __pte;
1100 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1101 	return __pte;
1102 }
1103 
1104 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1105 {
1106 	unsigned long physpage = page_to_phys(page);
1107 
1108 	return mk_pte_phys(physpage, pgprot);
1109 }
1110 
1111 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1112 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1113 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1114 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1115 
1116 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1117 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1118 
1119 #ifndef __s390x__
1120 
1121 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1122 #define pud_deref(pmd) ({ BUG(); 0UL; })
1123 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1124 
1125 #define pud_offset(pgd, address) ((pud_t *) pgd)
1126 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1127 
1128 #else /* __s390x__ */
1129 
1130 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1131 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1132 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1133 
1134 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1135 {
1136 	pud_t *pud = (pud_t *) pgd;
1137 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1138 		pud = (pud_t *) pgd_deref(*pgd);
1139 	return pud  + pud_index(address);
1140 }
1141 
1142 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1143 {
1144 	pmd_t *pmd = (pmd_t *) pud;
1145 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1146 		pmd = (pmd_t *) pud_deref(*pud);
1147 	return pmd + pmd_index(address);
1148 }
1149 
1150 #endif /* __s390x__ */
1151 
1152 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1153 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1154 #define pte_page(x) pfn_to_page(pte_pfn(x))
1155 
1156 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1157 
1158 /* Find an entry in the lowest level page table.. */
1159 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1160 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1161 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1162 #define pte_unmap(pte) do { } while (0)
1163 
1164 /*
1165  * 31 bit swap entry format:
1166  * A page-table entry has some bits we have to treat in a special way.
1167  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1168  * exception will occur instead of a page translation exception. The
1169  * specifiation exception has the bad habit not to store necessary
1170  * information in the lowcore.
1171  * Bit 21 and bit 22 are the page invalid bit and the page protection
1172  * bit. We set both to indicate a swapped page.
1173  * Bit 30 and 31 are used to distinguish the different page types. For
1174  * a swapped page these bits need to be zero.
1175  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1176  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1177  * plus 24 for the offset.
1178  * 0|     offset        |0110|o|type |00|
1179  * 0 0000000001111111111 2222 2 22222 33
1180  * 0 1234567890123456789 0123 4 56789 01
1181  *
1182  * 64 bit swap entry format:
1183  * A page-table entry has some bits we have to treat in a special way.
1184  * Bits 52 and bit 55 have to be zero, otherwise an specification
1185  * exception will occur instead of a page translation exception. The
1186  * specifiation exception has the bad habit not to store necessary
1187  * information in the lowcore.
1188  * Bit 53 and bit 54 are the page invalid bit and the page protection
1189  * bit. We set both to indicate a swapped page.
1190  * Bit 62 and 63 are used to distinguish the different page types. For
1191  * a swapped page these bits need to be zero.
1192  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1193  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1194  * plus 56 for the offset.
1195  * |                      offset                        |0110|o|type |00|
1196  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1197  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1198  */
1199 #ifndef __s390x__
1200 #define __SWP_OFFSET_MASK (~0UL >> 12)
1201 #else
1202 #define __SWP_OFFSET_MASK (~0UL >> 11)
1203 #endif
1204 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1205 {
1206 	pte_t pte;
1207 	offset &= __SWP_OFFSET_MASK;
1208 	pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
1209 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1210 	return pte;
1211 }
1212 
1213 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1214 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1215 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1216 
1217 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1218 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1219 
1220 #ifndef __s390x__
1221 # define PTE_FILE_MAX_BITS	26
1222 #else /* __s390x__ */
1223 # define PTE_FILE_MAX_BITS	59
1224 #endif /* __s390x__ */
1225 
1226 #define pte_to_pgoff(__pte) \
1227 	((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1228 
1229 #define pgoff_to_pte(__off) \
1230 	((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1231 		   | _PAGE_TYPE_FILE })
1232 
1233 #endif /* !__ASSEMBLY__ */
1234 
1235 #define kern_addr_valid(addr)   (1)
1236 
1237 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1238 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1239 extern int s390_enable_sie(void);
1240 
1241 /*
1242  * No page table caches to initialise
1243  */
1244 #define pgtable_cache_init()	do { } while (0)
1245 
1246 #include <asm-generic/pgtable.h>
1247 
1248 #endif /* _S390_PAGE_H */
1249