xref: /linux/arch/s390/include/asm/pgtable.h (revision 492c826b9facefa84995f4dea917e301b5ee0884)
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 
132 extern unsigned long VMALLOC_START;
133 
134 #ifndef __s390x__
135 #define VMALLOC_SIZE	(96UL << 20)
136 #define VMALLOC_END	0x7e000000UL
137 #define VMEM_MAP_END	0x80000000UL
138 #else /* __s390x__ */
139 #define VMALLOC_SIZE	(128UL << 30)
140 #define VMALLOC_END	0x3e000000000UL
141 #define VMEM_MAP_END	0x40000000000UL
142 #endif /* __s390x__ */
143 
144 /*
145  * VMEM_MAX_PHYS is the highest physical address that can be added to the 1:1
146  * mapping. This needs to be calculated at compile time since the size of the
147  * VMEM_MAP is static but the size of struct page can change.
148  */
149 #define VMEM_MAX_PAGES	((VMEM_MAP_END - VMALLOC_END) / sizeof(struct page))
150 #define VMEM_MAX_PFN	min(VMALLOC_START >> PAGE_SHIFT, VMEM_MAX_PAGES)
151 #define VMEM_MAX_PHYS	((VMEM_MAX_PFN << PAGE_SHIFT) & ~((16 << 20) - 1))
152 #define vmemmap		((struct page *) VMALLOC_END)
153 
154 /*
155  * A 31 bit pagetable entry of S390 has following format:
156  *  |   PFRA          |    |  OS  |
157  * 0                   0IP0
158  * 00000000001111111111222222222233
159  * 01234567890123456789012345678901
160  *
161  * I Page-Invalid Bit:    Page is not available for address-translation
162  * P Page-Protection Bit: Store access not possible for page
163  *
164  * A 31 bit segmenttable entry of S390 has following format:
165  *  |   P-table origin      |  |PTL
166  * 0                         IC
167  * 00000000001111111111222222222233
168  * 01234567890123456789012345678901
169  *
170  * I Segment-Invalid Bit:    Segment is not available for address-translation
171  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
172  * PTL Page-Table-Length:    Page-table length (PTL+1*16 entries -> up to 256)
173  *
174  * The 31 bit segmenttable origin of S390 has following format:
175  *
176  *  |S-table origin   |     | STL |
177  * X                   **GPS
178  * 00000000001111111111222222222233
179  * 01234567890123456789012345678901
180  *
181  * X Space-Switch event:
182  * G Segment-Invalid Bit:     *
183  * P Private-Space Bit:       Segment is not private (PoP 3-30)
184  * S Storage-Alteration:
185  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
186  *
187  * A 64 bit pagetable entry of S390 has following format:
188  * |			 PFRA			      |0IPC|  OS  |
189  * 0000000000111111111122222222223333333333444444444455555555556666
190  * 0123456789012345678901234567890123456789012345678901234567890123
191  *
192  * I Page-Invalid Bit:    Page is not available for address-translation
193  * P Page-Protection Bit: Store access not possible for page
194  * C Change-bit override: HW is not required to set change bit
195  *
196  * A 64 bit segmenttable entry of S390 has following format:
197  * |        P-table origin                              |      TT
198  * 0000000000111111111122222222223333333333444444444455555555556666
199  * 0123456789012345678901234567890123456789012345678901234567890123
200  *
201  * I Segment-Invalid Bit:    Segment is not available for address-translation
202  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
203  * P Page-Protection Bit: Store access not possible for page
204  * TT Type 00
205  *
206  * A 64 bit region table entry of S390 has following format:
207  * |        S-table origin                             |   TF  TTTL
208  * 0000000000111111111122222222223333333333444444444455555555556666
209  * 0123456789012345678901234567890123456789012345678901234567890123
210  *
211  * I Segment-Invalid Bit:    Segment is not available for address-translation
212  * TT Type 01
213  * TF
214  * TL Table length
215  *
216  * The 64 bit regiontable origin of S390 has following format:
217  * |      region table origon                          |       DTTL
218  * 0000000000111111111122222222223333333333444444444455555555556666
219  * 0123456789012345678901234567890123456789012345678901234567890123
220  *
221  * X Space-Switch event:
222  * G Segment-Invalid Bit:
223  * P Private-Space Bit:
224  * S Storage-Alteration:
225  * R Real space
226  * TL Table-Length:
227  *
228  * A storage key has the following format:
229  * | ACC |F|R|C|0|
230  *  0   3 4 5 6 7
231  * ACC: access key
232  * F  : fetch protection bit
233  * R  : referenced bit
234  * C  : changed bit
235  */
236 
237 /* Hardware bits in the page table entry */
238 #define _PAGE_CO	0x100		/* HW Change-bit override */
239 #define _PAGE_RO	0x200		/* HW read-only bit  */
240 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
241 
242 /* Software bits in the page table entry */
243 #define _PAGE_SWT	0x001		/* SW pte type bit t */
244 #define _PAGE_SWX	0x002		/* SW pte type bit x */
245 #define _PAGE_SWC	0x004		/* SW pte changed bit (for KVM) */
246 #define _PAGE_SWR	0x008		/* SW pte referenced bit (for KVM) */
247 #define _PAGE_SPECIAL	0x010		/* SW associated with special page */
248 #define __HAVE_ARCH_PTE_SPECIAL
249 
250 /* Set of bits not changed in pte_modify */
251 #define _PAGE_CHG_MASK	(PAGE_MASK | _PAGE_SPECIAL | _PAGE_SWC | _PAGE_SWR)
252 
253 /* Six different types of pages. */
254 #define _PAGE_TYPE_EMPTY	0x400
255 #define _PAGE_TYPE_NONE		0x401
256 #define _PAGE_TYPE_SWAP		0x403
257 #define _PAGE_TYPE_FILE		0x601	/* bit 0x002 is used for offset !! */
258 #define _PAGE_TYPE_RO		0x200
259 #define _PAGE_TYPE_RW		0x000
260 
261 /*
262  * Only four types for huge pages, using the invalid bit and protection bit
263  * of a segment table entry.
264  */
265 #define _HPAGE_TYPE_EMPTY	0x020	/* _SEGMENT_ENTRY_INV */
266 #define _HPAGE_TYPE_NONE	0x220
267 #define _HPAGE_TYPE_RO		0x200	/* _SEGMENT_ENTRY_RO  */
268 #define _HPAGE_TYPE_RW		0x000
269 
270 /*
271  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
272  * pte_none and pte_file to find out the pte type WITHOUT holding the page
273  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
274  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
275  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
276  * This change is done while holding the lock, but the intermediate step
277  * of a previously valid pte with the hw invalid bit set can be observed by
278  * handle_pte_fault. That makes it necessary that all valid pte types with
279  * the hw invalid bit set must be distinguishable from the four pte types
280  * empty, none, swap and file.
281  *
282  *			irxt  ipte  irxt
283  * _PAGE_TYPE_EMPTY	1000   ->   1000
284  * _PAGE_TYPE_NONE	1001   ->   1001
285  * _PAGE_TYPE_SWAP	1011   ->   1011
286  * _PAGE_TYPE_FILE	11?1   ->   11?1
287  * _PAGE_TYPE_RO	0100   ->   1100
288  * _PAGE_TYPE_RW	0000   ->   1000
289  *
290  * pte_none is true for bits combinations 1000, 1010, 1100, 1110
291  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
292  * pte_file is true for bits combinations 1101, 1111
293  * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
294  */
295 
296 /* Page status table bits for virtualization */
297 #define RCP_ACC_BITS	0xf000000000000000UL
298 #define RCP_FP_BIT	0x0800000000000000UL
299 #define RCP_PCL_BIT	0x0080000000000000UL
300 #define RCP_HR_BIT	0x0040000000000000UL
301 #define RCP_HC_BIT	0x0020000000000000UL
302 #define RCP_GR_BIT	0x0004000000000000UL
303 #define RCP_GC_BIT	0x0002000000000000UL
304 
305 /* User dirty / referenced bit for KVM's migration feature */
306 #define KVM_UR_BIT	0x0000800000000000UL
307 #define KVM_UC_BIT	0x0000400000000000UL
308 
309 #ifndef __s390x__
310 
311 /* Bits in the segment table address-space-control-element */
312 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
313 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
314 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
315 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
316 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
317 
318 /* Bits in the segment table entry */
319 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
320 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
321 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
322 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
323 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
324 
325 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
326 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
327 
328 #else /* __s390x__ */
329 
330 /* Bits in the segment/region table address-space-control-element */
331 #define _ASCE_ORIGIN		~0xfffUL/* segment table origin		    */
332 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
333 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
334 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
335 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
336 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
337 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
338 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
339 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
340 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
341 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
342 
343 /* Bits in the region table entry */
344 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
345 #define _REGION_ENTRY_INV	0x20	/* invalid region table entry	    */
346 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
347 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
348 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
349 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
350 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
351 
352 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
353 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
354 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
355 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
356 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
357 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
358 
359 /* Bits in the segment table entry */
360 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* segment table origin		    */
361 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
362 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
363 
364 #define _SEGMENT_ENTRY		(0)
365 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
366 
367 #define _SEGMENT_ENTRY_LARGE	0x400	/* STE-format control, large page   */
368 #define _SEGMENT_ENTRY_CO	0x100	/* change-recording override   */
369 
370 #endif /* __s390x__ */
371 
372 /*
373  * A user page table pointer has the space-switch-event bit, the
374  * private-space-control bit and the storage-alteration-event-control
375  * bit set. A kernel page table pointer doesn't need them.
376  */
377 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
378 				 _ASCE_ALT_EVENT)
379 
380 /*
381  * Page protection definitions.
382  */
383 #define PAGE_NONE	__pgprot(_PAGE_TYPE_NONE)
384 #define PAGE_RO		__pgprot(_PAGE_TYPE_RO)
385 #define PAGE_RW		__pgprot(_PAGE_TYPE_RW)
386 
387 #define PAGE_KERNEL	PAGE_RW
388 #define PAGE_COPY	PAGE_RO
389 
390 /*
391  * On s390 the page table entry has an invalid bit and a read-only bit.
392  * Read permission implies execute permission and write permission
393  * implies read permission.
394  */
395          /*xwr*/
396 #define __P000	PAGE_NONE
397 #define __P001	PAGE_RO
398 #define __P010	PAGE_RO
399 #define __P011	PAGE_RO
400 #define __P100	PAGE_RO
401 #define __P101	PAGE_RO
402 #define __P110	PAGE_RO
403 #define __P111	PAGE_RO
404 
405 #define __S000	PAGE_NONE
406 #define __S001	PAGE_RO
407 #define __S010	PAGE_RW
408 #define __S011	PAGE_RW
409 #define __S100	PAGE_RO
410 #define __S101	PAGE_RO
411 #define __S110	PAGE_RW
412 #define __S111	PAGE_RW
413 
414 static inline int mm_exclusive(struct mm_struct *mm)
415 {
416 	return likely(mm == current->active_mm &&
417 		      atomic_read(&mm->context.attach_count) <= 1);
418 }
419 
420 static inline int mm_has_pgste(struct mm_struct *mm)
421 {
422 #ifdef CONFIG_PGSTE
423 	if (unlikely(mm->context.has_pgste))
424 		return 1;
425 #endif
426 	return 0;
427 }
428 /*
429  * pgd/pmd/pte query functions
430  */
431 #ifndef __s390x__
432 
433 static inline int pgd_present(pgd_t pgd) { return 1; }
434 static inline int pgd_none(pgd_t pgd)    { return 0; }
435 static inline int pgd_bad(pgd_t pgd)     { return 0; }
436 
437 static inline int pud_present(pud_t pud) { return 1; }
438 static inline int pud_none(pud_t pud)	 { return 0; }
439 static inline int pud_bad(pud_t pud)	 { return 0; }
440 
441 #else /* __s390x__ */
442 
443 static inline int pgd_present(pgd_t pgd)
444 {
445 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
446 		return 1;
447 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
448 }
449 
450 static inline int pgd_none(pgd_t pgd)
451 {
452 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
453 		return 0;
454 	return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL;
455 }
456 
457 static inline int pgd_bad(pgd_t pgd)
458 {
459 	/*
460 	 * With dynamic page table levels the pgd can be a region table
461 	 * entry or a segment table entry. Check for the bit that are
462 	 * invalid for either table entry.
463 	 */
464 	unsigned long mask =
465 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
466 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
467 	return (pgd_val(pgd) & mask) != 0;
468 }
469 
470 static inline int pud_present(pud_t pud)
471 {
472 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
473 		return 1;
474 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
475 }
476 
477 static inline int pud_none(pud_t pud)
478 {
479 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
480 		return 0;
481 	return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
482 }
483 
484 static inline int pud_bad(pud_t pud)
485 {
486 	/*
487 	 * With dynamic page table levels the pud can be a region table
488 	 * entry or a segment table entry. Check for the bit that are
489 	 * invalid for either table entry.
490 	 */
491 	unsigned long mask =
492 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
493 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
494 	return (pud_val(pud) & mask) != 0;
495 }
496 
497 #endif /* __s390x__ */
498 
499 static inline int pmd_present(pmd_t pmd)
500 {
501 	return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL;
502 }
503 
504 static inline int pmd_none(pmd_t pmd)
505 {
506 	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL;
507 }
508 
509 static inline int pmd_bad(pmd_t pmd)
510 {
511 	unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
512 	return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
513 }
514 
515 static inline int pte_none(pte_t pte)
516 {
517 	return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
518 }
519 
520 static inline int pte_present(pte_t pte)
521 {
522 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
523 	return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
524 		(!(pte_val(pte) & _PAGE_INVALID) &&
525 		 !(pte_val(pte) & _PAGE_SWT));
526 }
527 
528 static inline int pte_file(pte_t pte)
529 {
530 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
531 	return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
532 }
533 
534 static inline int pte_special(pte_t pte)
535 {
536 	return (pte_val(pte) & _PAGE_SPECIAL);
537 }
538 
539 #define __HAVE_ARCH_PTE_SAME
540 static inline int pte_same(pte_t a, pte_t b)
541 {
542 	return pte_val(a) == pte_val(b);
543 }
544 
545 static inline pgste_t pgste_get_lock(pte_t *ptep)
546 {
547 	unsigned long new = 0;
548 #ifdef CONFIG_PGSTE
549 	unsigned long old;
550 
551 	preempt_disable();
552 	asm(
553 		"	lg	%0,%2\n"
554 		"0:	lgr	%1,%0\n"
555 		"	nihh	%0,0xff7f\n"	/* clear RCP_PCL_BIT in old */
556 		"	oihh	%1,0x0080\n"	/* set RCP_PCL_BIT in new */
557 		"	csg	%0,%1,%2\n"
558 		"	jl	0b\n"
559 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
560 		: "Q" (ptep[PTRS_PER_PTE]) : "cc");
561 #endif
562 	return __pgste(new);
563 }
564 
565 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
566 {
567 #ifdef CONFIG_PGSTE
568 	asm(
569 		"	nihh	%1,0xff7f\n"	/* clear RCP_PCL_BIT */
570 		"	stg	%1,%0\n"
571 		: "=Q" (ptep[PTRS_PER_PTE])
572 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc");
573 	preempt_enable();
574 #endif
575 }
576 
577 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
578 {
579 #ifdef CONFIG_PGSTE
580 	unsigned long pfn, bits;
581 	unsigned char skey;
582 
583 	pfn = pte_val(*ptep) >> PAGE_SHIFT;
584 	skey = page_get_storage_key(pfn);
585 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
586 	/* Clear page changed & referenced bit in the storage key */
587 	if (bits) {
588 		skey ^= bits;
589 		page_set_storage_key(pfn, skey, 1);
590 	}
591 	/* Transfer page changed & referenced bit to guest bits in pgste */
592 	pgste_val(pgste) |= bits << 48;		/* RCP_GR_BIT & RCP_GC_BIT */
593 	/* Get host changed & referenced bits from pgste */
594 	bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52;
595 	/* Clear host bits in pgste. */
596 	pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT);
597 	pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT);
598 	/* Copy page access key and fetch protection bit to pgste */
599 	pgste_val(pgste) |=
600 		(unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
601 	/* Transfer changed and referenced to kvm user bits */
602 	pgste_val(pgste) |= bits << 45;		/* KVM_UR_BIT & KVM_UC_BIT */
603 	/* Transfer changed & referenced to pte sofware bits */
604 	pte_val(*ptep) |= bits << 1;		/* _PAGE_SWR & _PAGE_SWC */
605 #endif
606 	return pgste;
607 
608 }
609 
610 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
611 {
612 #ifdef CONFIG_PGSTE
613 	int young;
614 
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)
629 {
630 #ifdef CONFIG_PGSTE
631 	unsigned long pfn;
632 	unsigned long okey, nkey;
633 
634 	pfn = pte_val(*ptep) >> PAGE_SHIFT;
635 	okey = nkey = page_get_storage_key(pfn);
636 	nkey &= ~(_PAGE_ACC_BITS | _PAGE_FP_BIT);
637 	/* Set page access key and fetch protection bit from pgste */
638 	nkey |= (pgste_val(pgste) & (RCP_ACC_BITS | RCP_FP_BIT)) >> 56;
639 	if (okey != nkey)
640 		page_set_storage_key(pfn, nkey, 1);
641 #endif
642 }
643 
644 /*
645  * Certain architectures need to do special things when PTEs
646  * within a page table are directly modified.  Thus, the following
647  * hook is made available.
648  */
649 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
650 			      pte_t *ptep, pte_t entry)
651 {
652 	pgste_t pgste;
653 
654 	if (mm_has_pgste(mm)) {
655 		pgste = pgste_get_lock(ptep);
656 		pgste_set_pte(ptep, pgste);
657 		*ptep = entry;
658 		pgste_set_unlock(ptep, pgste);
659 	} else
660 		*ptep = entry;
661 }
662 
663 /*
664  * query functions pte_write/pte_dirty/pte_young only work if
665  * pte_present() is true. Undefined behaviour if not..
666  */
667 static inline int pte_write(pte_t pte)
668 {
669 	return (pte_val(pte) & _PAGE_RO) == 0;
670 }
671 
672 static inline int pte_dirty(pte_t pte)
673 {
674 #ifdef CONFIG_PGSTE
675 	if (pte_val(pte) & _PAGE_SWC)
676 		return 1;
677 #endif
678 	return 0;
679 }
680 
681 static inline int pte_young(pte_t pte)
682 {
683 #ifdef CONFIG_PGSTE
684 	if (pte_val(pte) & _PAGE_SWR)
685 		return 1;
686 #endif
687 	return 0;
688 }
689 
690 /*
691  * pgd/pmd/pte modification functions
692  */
693 
694 static inline void pgd_clear(pgd_t *pgd)
695 {
696 #ifdef __s390x__
697 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
698 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
699 #endif
700 }
701 
702 static inline void pud_clear(pud_t *pud)
703 {
704 #ifdef __s390x__
705 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
706 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
707 #endif
708 }
709 
710 static inline void pmd_clear(pmd_t *pmdp)
711 {
712 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
713 }
714 
715 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
716 {
717 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
718 }
719 
720 /*
721  * The following pte modification functions only work if
722  * pte_present() is true. Undefined behaviour if not..
723  */
724 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
725 {
726 	pte_val(pte) &= _PAGE_CHG_MASK;
727 	pte_val(pte) |= pgprot_val(newprot);
728 	return pte;
729 }
730 
731 static inline pte_t pte_wrprotect(pte_t pte)
732 {
733 	/* Do not clobber _PAGE_TYPE_NONE pages!  */
734 	if (!(pte_val(pte) & _PAGE_INVALID))
735 		pte_val(pte) |= _PAGE_RO;
736 	return pte;
737 }
738 
739 static inline pte_t pte_mkwrite(pte_t pte)
740 {
741 	pte_val(pte) &= ~_PAGE_RO;
742 	return pte;
743 }
744 
745 static inline pte_t pte_mkclean(pte_t pte)
746 {
747 #ifdef CONFIG_PGSTE
748 	pte_val(pte) &= ~_PAGE_SWC;
749 #endif
750 	return pte;
751 }
752 
753 static inline pte_t pte_mkdirty(pte_t pte)
754 {
755 	return pte;
756 }
757 
758 static inline pte_t pte_mkold(pte_t pte)
759 {
760 #ifdef CONFIG_PGSTE
761 	pte_val(pte) &= ~_PAGE_SWR;
762 #endif
763 	return pte;
764 }
765 
766 static inline pte_t pte_mkyoung(pte_t pte)
767 {
768 	return pte;
769 }
770 
771 static inline pte_t pte_mkspecial(pte_t pte)
772 {
773 	pte_val(pte) |= _PAGE_SPECIAL;
774 	return pte;
775 }
776 
777 #ifdef CONFIG_HUGETLB_PAGE
778 static inline pte_t pte_mkhuge(pte_t pte)
779 {
780 	/*
781 	 * PROT_NONE needs to be remapped from the pte type to the ste type.
782 	 * The HW invalid bit is also different for pte and ste. The pte
783 	 * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE
784 	 * bit, so we don't have to clear it.
785 	 */
786 	if (pte_val(pte) & _PAGE_INVALID) {
787 		if (pte_val(pte) & _PAGE_SWT)
788 			pte_val(pte) |= _HPAGE_TYPE_NONE;
789 		pte_val(pte) |= _SEGMENT_ENTRY_INV;
790 	}
791 	/*
792 	 * Clear SW pte bits SWT and SWX, there are no SW bits in a segment
793 	 * table entry.
794 	 */
795 	pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX);
796 	/*
797 	 * Also set the change-override bit because we don't need dirty bit
798 	 * tracking for hugetlbfs pages.
799 	 */
800 	pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
801 	return pte;
802 }
803 #endif
804 
805 /*
806  * Get (and clear) the user dirty bit for a pte.
807  */
808 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
809 						 pte_t *ptep)
810 {
811 	pgste_t pgste;
812 	int dirty = 0;
813 
814 	if (mm_has_pgste(mm)) {
815 		pgste = pgste_get_lock(ptep);
816 		pgste = pgste_update_all(ptep, pgste);
817 		dirty = !!(pgste_val(pgste) & KVM_UC_BIT);
818 		pgste_val(pgste) &= ~KVM_UC_BIT;
819 		pgste_set_unlock(ptep, pgste);
820 		return dirty;
821 	}
822 	return dirty;
823 }
824 
825 /*
826  * Get (and clear) the user referenced bit for a pte.
827  */
828 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
829 						 pte_t *ptep)
830 {
831 	pgste_t pgste;
832 	int young = 0;
833 
834 	if (mm_has_pgste(mm)) {
835 		pgste = pgste_get_lock(ptep);
836 		pgste = pgste_update_young(ptep, pgste);
837 		young = !!(pgste_val(pgste) & KVM_UR_BIT);
838 		pgste_val(pgste) &= ~KVM_UR_BIT;
839 		pgste_set_unlock(ptep, pgste);
840 	}
841 	return young;
842 }
843 
844 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
845 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
846 					    unsigned long addr, pte_t *ptep)
847 {
848 	pgste_t pgste;
849 	pte_t pte;
850 
851 	if (mm_has_pgste(vma->vm_mm)) {
852 		pgste = pgste_get_lock(ptep);
853 		pgste = pgste_update_young(ptep, pgste);
854 		pte = *ptep;
855 		*ptep = pte_mkold(pte);
856 		pgste_set_unlock(ptep, pgste);
857 		return pte_young(pte);
858 	}
859 	return 0;
860 }
861 
862 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
863 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
864 					 unsigned long address, pte_t *ptep)
865 {
866 	/* No need to flush TLB
867 	 * On s390 reference bits are in storage key and never in TLB
868 	 * With virtualization we handle the reference bit, without we
869 	 * we can simply return */
870 	return ptep_test_and_clear_young(vma, address, ptep);
871 }
872 
873 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
874 {
875 	if (!(pte_val(*ptep) & _PAGE_INVALID)) {
876 #ifndef __s390x__
877 		/* pto must point to the start of the segment table */
878 		pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
879 #else
880 		/* ipte in zarch mode can do the math */
881 		pte_t *pto = ptep;
882 #endif
883 		asm volatile(
884 			"	ipte	%2,%3"
885 			: "=m" (*ptep) : "m" (*ptep),
886 			  "a" (pto), "a" (address));
887 	}
888 }
889 
890 /*
891  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
892  * both clear the TLB for the unmapped pte. The reason is that
893  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
894  * to modify an active pte. The sequence is
895  *   1) ptep_get_and_clear
896  *   2) set_pte_at
897  *   3) flush_tlb_range
898  * On s390 the tlb needs to get flushed with the modification of the pte
899  * if the pte is active. The only way how this can be implemented is to
900  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
901  * is a nop.
902  */
903 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
904 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
905 				       unsigned long address, pte_t *ptep)
906 {
907 	pgste_t pgste;
908 	pte_t pte;
909 
910 	mm->context.flush_mm = 1;
911 	if (mm_has_pgste(mm))
912 		pgste = pgste_get_lock(ptep);
913 
914 	pte = *ptep;
915 	if (!mm_exclusive(mm))
916 		__ptep_ipte(address, ptep);
917 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
918 
919 	if (mm_has_pgste(mm)) {
920 		pgste = pgste_update_all(&pte, pgste);
921 		pgste_set_unlock(ptep, pgste);
922 	}
923 	return pte;
924 }
925 
926 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
927 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
928 					   unsigned long address,
929 					   pte_t *ptep)
930 {
931 	pte_t pte;
932 
933 	mm->context.flush_mm = 1;
934 	if (mm_has_pgste(mm))
935 		pgste_get_lock(ptep);
936 
937 	pte = *ptep;
938 	if (!mm_exclusive(mm))
939 		__ptep_ipte(address, ptep);
940 	return pte;
941 }
942 
943 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
944 					   unsigned long address,
945 					   pte_t *ptep, pte_t pte)
946 {
947 	*ptep = pte;
948 	if (mm_has_pgste(mm))
949 		pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE));
950 }
951 
952 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
953 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
954 				     unsigned long address, pte_t *ptep)
955 {
956 	pgste_t pgste;
957 	pte_t pte;
958 
959 	if (mm_has_pgste(vma->vm_mm))
960 		pgste = pgste_get_lock(ptep);
961 
962 	pte = *ptep;
963 	__ptep_ipte(address, ptep);
964 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
965 
966 	if (mm_has_pgste(vma->vm_mm)) {
967 		pgste = pgste_update_all(&pte, pgste);
968 		pgste_set_unlock(ptep, pgste);
969 	}
970 	return pte;
971 }
972 
973 /*
974  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
975  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
976  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
977  * cannot be accessed while the batched unmap is running. In this case
978  * full==1 and a simple pte_clear is enough. See tlb.h.
979  */
980 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
981 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
982 					    unsigned long address,
983 					    pte_t *ptep, int full)
984 {
985 	pgste_t pgste;
986 	pte_t pte;
987 
988 	if (mm_has_pgste(mm))
989 		pgste = pgste_get_lock(ptep);
990 
991 	pte = *ptep;
992 	if (!full)
993 		__ptep_ipte(address, ptep);
994 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
995 
996 	if (mm_has_pgste(mm)) {
997 		pgste = pgste_update_all(&pte, pgste);
998 		pgste_set_unlock(ptep, pgste);
999 	}
1000 	return pte;
1001 }
1002 
1003 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1004 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1005 				       unsigned long address, pte_t *ptep)
1006 {
1007 	pgste_t pgste;
1008 	pte_t pte = *ptep;
1009 
1010 	if (pte_write(pte)) {
1011 		mm->context.flush_mm = 1;
1012 		if (mm_has_pgste(mm))
1013 			pgste = pgste_get_lock(ptep);
1014 
1015 		if (!mm_exclusive(mm))
1016 			__ptep_ipte(address, ptep);
1017 		*ptep = pte_wrprotect(pte);
1018 
1019 		if (mm_has_pgste(mm))
1020 			pgste_set_unlock(ptep, pgste);
1021 	}
1022 	return pte;
1023 }
1024 
1025 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1026 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1027 					unsigned long address, pte_t *ptep,
1028 					pte_t entry, int dirty)
1029 {
1030 	pgste_t pgste;
1031 
1032 	if (pte_same(*ptep, entry))
1033 		return 0;
1034 	if (mm_has_pgste(vma->vm_mm))
1035 		pgste = pgste_get_lock(ptep);
1036 
1037 	__ptep_ipte(address, ptep);
1038 	*ptep = entry;
1039 
1040 	if (mm_has_pgste(vma->vm_mm))
1041 		pgste_set_unlock(ptep, pgste);
1042 	return 1;
1043 }
1044 
1045 /*
1046  * Conversion functions: convert a page and protection to a page entry,
1047  * and a page entry and page directory to the page they refer to.
1048  */
1049 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1050 {
1051 	pte_t __pte;
1052 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1053 	return __pte;
1054 }
1055 
1056 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1057 {
1058 	unsigned long physpage = page_to_phys(page);
1059 
1060 	return mk_pte_phys(physpage, pgprot);
1061 }
1062 
1063 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1064 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1065 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1066 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1067 
1068 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1069 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1070 
1071 #ifndef __s390x__
1072 
1073 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1074 #define pud_deref(pmd) ({ BUG(); 0UL; })
1075 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1076 
1077 #define pud_offset(pgd, address) ((pud_t *) pgd)
1078 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1079 
1080 #else /* __s390x__ */
1081 
1082 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1083 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1084 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1085 
1086 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1087 {
1088 	pud_t *pud = (pud_t *) pgd;
1089 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1090 		pud = (pud_t *) pgd_deref(*pgd);
1091 	return pud  + pud_index(address);
1092 }
1093 
1094 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1095 {
1096 	pmd_t *pmd = (pmd_t *) pud;
1097 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1098 		pmd = (pmd_t *) pud_deref(*pud);
1099 	return pmd + pmd_index(address);
1100 }
1101 
1102 #endif /* __s390x__ */
1103 
1104 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1105 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1106 #define pte_page(x) pfn_to_page(pte_pfn(x))
1107 
1108 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1109 
1110 /* Find an entry in the lowest level page table.. */
1111 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1112 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1113 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1114 #define pte_unmap(pte) do { } while (0)
1115 
1116 /*
1117  * 31 bit swap entry format:
1118  * A page-table entry has some bits we have to treat in a special way.
1119  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1120  * exception will occur instead of a page translation exception. The
1121  * specifiation exception has the bad habit not to store necessary
1122  * information in the lowcore.
1123  * Bit 21 and bit 22 are the page invalid bit and the page protection
1124  * bit. We set both to indicate a swapped page.
1125  * Bit 30 and 31 are used to distinguish the different page types. For
1126  * a swapped page these bits need to be zero.
1127  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1128  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1129  * plus 24 for the offset.
1130  * 0|     offset        |0110|o|type |00|
1131  * 0 0000000001111111111 2222 2 22222 33
1132  * 0 1234567890123456789 0123 4 56789 01
1133  *
1134  * 64 bit swap entry format:
1135  * A page-table entry has some bits we have to treat in a special way.
1136  * Bits 52 and bit 55 have to be zero, otherwise an specification
1137  * exception will occur instead of a page translation exception. The
1138  * specifiation exception has the bad habit not to store necessary
1139  * information in the lowcore.
1140  * Bit 53 and bit 54 are the page invalid bit and the page protection
1141  * bit. We set both to indicate a swapped page.
1142  * Bit 62 and 63 are used to distinguish the different page types. For
1143  * a swapped page these bits need to be zero.
1144  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1145  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1146  * plus 56 for the offset.
1147  * |                      offset                        |0110|o|type |00|
1148  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1149  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1150  */
1151 #ifndef __s390x__
1152 #define __SWP_OFFSET_MASK (~0UL >> 12)
1153 #else
1154 #define __SWP_OFFSET_MASK (~0UL >> 11)
1155 #endif
1156 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1157 {
1158 	pte_t pte;
1159 	offset &= __SWP_OFFSET_MASK;
1160 	pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
1161 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1162 	return pte;
1163 }
1164 
1165 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1166 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1167 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1168 
1169 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1170 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1171 
1172 #ifndef __s390x__
1173 # define PTE_FILE_MAX_BITS	26
1174 #else /* __s390x__ */
1175 # define PTE_FILE_MAX_BITS	59
1176 #endif /* __s390x__ */
1177 
1178 #define pte_to_pgoff(__pte) \
1179 	((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1180 
1181 #define pgoff_to_pte(__off) \
1182 	((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1183 		   | _PAGE_TYPE_FILE })
1184 
1185 #endif /* !__ASSEMBLY__ */
1186 
1187 #define kern_addr_valid(addr)   (1)
1188 
1189 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1190 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1191 extern int s390_enable_sie(void);
1192 
1193 /*
1194  * No page table caches to initialise
1195  */
1196 #define pgtable_cache_init()	do { } while (0)
1197 
1198 #include <asm-generic/pgtable.h>
1199 
1200 #endif /* _S390_PAGE_H */
1201