xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision 81d67439)
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 #ifndef __s390x__
297 
298 /* Bits in the segment table address-space-control-element */
299 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
300 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
301 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
302 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
303 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
304 
305 /* Bits in the segment table entry */
306 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
307 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
308 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
309 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
310 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
311 
312 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
313 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
314 
315 /* Page status table bits for virtualization */
316 #define RCP_ACC_BITS	0xf0000000UL
317 #define RCP_FP_BIT	0x08000000UL
318 #define RCP_PCL_BIT	0x00800000UL
319 #define RCP_HR_BIT	0x00400000UL
320 #define RCP_HC_BIT	0x00200000UL
321 #define RCP_GR_BIT	0x00040000UL
322 #define RCP_GC_BIT	0x00020000UL
323 
324 /* User dirty / referenced bit for KVM's migration feature */
325 #define KVM_UR_BIT	0x00008000UL
326 #define KVM_UC_BIT	0x00004000UL
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 /* Page status table bits for virtualization */
371 #define RCP_ACC_BITS	0xf000000000000000UL
372 #define RCP_FP_BIT	0x0800000000000000UL
373 #define RCP_PCL_BIT	0x0080000000000000UL
374 #define RCP_HR_BIT	0x0040000000000000UL
375 #define RCP_HC_BIT	0x0020000000000000UL
376 #define RCP_GR_BIT	0x0004000000000000UL
377 #define RCP_GC_BIT	0x0002000000000000UL
378 
379 /* User dirty / referenced bit for KVM's migration feature */
380 #define KVM_UR_BIT	0x0000800000000000UL
381 #define KVM_UC_BIT	0x0000400000000000UL
382 
383 #endif /* __s390x__ */
384 
385 /*
386  * A user page table pointer has the space-switch-event bit, the
387  * private-space-control bit and the storage-alteration-event-control
388  * bit set. A kernel page table pointer doesn't need them.
389  */
390 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
391 				 _ASCE_ALT_EVENT)
392 
393 /*
394  * Page protection definitions.
395  */
396 #define PAGE_NONE	__pgprot(_PAGE_TYPE_NONE)
397 #define PAGE_RO		__pgprot(_PAGE_TYPE_RO)
398 #define PAGE_RW		__pgprot(_PAGE_TYPE_RW)
399 
400 #define PAGE_KERNEL	PAGE_RW
401 #define PAGE_COPY	PAGE_RO
402 
403 /*
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.
407  */
408          /*xwr*/
409 #define __P000	PAGE_NONE
410 #define __P001	PAGE_RO
411 #define __P010	PAGE_RO
412 #define __P011	PAGE_RO
413 #define __P100	PAGE_RO
414 #define __P101	PAGE_RO
415 #define __P110	PAGE_RO
416 #define __P111	PAGE_RO
417 
418 #define __S000	PAGE_NONE
419 #define __S001	PAGE_RO
420 #define __S010	PAGE_RW
421 #define __S011	PAGE_RW
422 #define __S100	PAGE_RO
423 #define __S101	PAGE_RO
424 #define __S110	PAGE_RW
425 #define __S111	PAGE_RW
426 
427 static inline int mm_exclusive(struct mm_struct *mm)
428 {
429 	return likely(mm == current->active_mm &&
430 		      atomic_read(&mm->context.attach_count) <= 1);
431 }
432 
433 static inline int mm_has_pgste(struct mm_struct *mm)
434 {
435 #ifdef CONFIG_PGSTE
436 	if (unlikely(mm->context.has_pgste))
437 		return 1;
438 #endif
439 	return 0;
440 }
441 /*
442  * pgd/pmd/pte query functions
443  */
444 #ifndef __s390x__
445 
446 static inline int pgd_present(pgd_t pgd) { return 1; }
447 static inline int pgd_none(pgd_t pgd)    { return 0; }
448 static inline int pgd_bad(pgd_t pgd)     { return 0; }
449 
450 static inline int pud_present(pud_t pud) { return 1; }
451 static inline int pud_none(pud_t pud)	 { return 0; }
452 static inline int pud_bad(pud_t pud)	 { return 0; }
453 
454 #else /* __s390x__ */
455 
456 static inline int pgd_present(pgd_t pgd)
457 {
458 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
459 		return 1;
460 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
461 }
462 
463 static inline int pgd_none(pgd_t pgd)
464 {
465 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
466 		return 0;
467 	return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL;
468 }
469 
470 static inline int pgd_bad(pgd_t pgd)
471 {
472 	/*
473 	 * With dynamic page table levels the pgd can be a region table
474 	 * entry or a segment table entry. Check for the bit that are
475 	 * invalid for either table entry.
476 	 */
477 	unsigned long mask =
478 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
479 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
480 	return (pgd_val(pgd) & mask) != 0;
481 }
482 
483 static inline int pud_present(pud_t pud)
484 {
485 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
486 		return 1;
487 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
488 }
489 
490 static inline int pud_none(pud_t pud)
491 {
492 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
493 		return 0;
494 	return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
495 }
496 
497 static inline int pud_bad(pud_t pud)
498 {
499 	/*
500 	 * With dynamic page table levels the pud can be a region table
501 	 * entry or a segment table entry. Check for the bit that are
502 	 * invalid for either table entry.
503 	 */
504 	unsigned long mask =
505 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
506 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
507 	return (pud_val(pud) & mask) != 0;
508 }
509 
510 #endif /* __s390x__ */
511 
512 static inline int pmd_present(pmd_t pmd)
513 {
514 	return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL;
515 }
516 
517 static inline int pmd_none(pmd_t pmd)
518 {
519 	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL;
520 }
521 
522 static inline int pmd_bad(pmd_t pmd)
523 {
524 	unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
525 	return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
526 }
527 
528 static inline int pte_none(pte_t pte)
529 {
530 	return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
531 }
532 
533 static inline int pte_present(pte_t pte)
534 {
535 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
536 	return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
537 		(!(pte_val(pte) & _PAGE_INVALID) &&
538 		 !(pte_val(pte) & _PAGE_SWT));
539 }
540 
541 static inline int pte_file(pte_t pte)
542 {
543 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
544 	return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
545 }
546 
547 static inline int pte_special(pte_t pte)
548 {
549 	return (pte_val(pte) & _PAGE_SPECIAL);
550 }
551 
552 #define __HAVE_ARCH_PTE_SAME
553 static inline int pte_same(pte_t a, pte_t b)
554 {
555 	return pte_val(a) == pte_val(b);
556 }
557 
558 static inline pgste_t pgste_get_lock(pte_t *ptep)
559 {
560 	unsigned long new = 0;
561 #ifdef CONFIG_PGSTE
562 	unsigned long old;
563 
564 	preempt_disable();
565 	asm(
566 		"	lg	%0,%2\n"
567 		"0:	lgr	%1,%0\n"
568 		"	nihh	%0,0xff7f\n"	/* clear RCP_PCL_BIT in old */
569 		"	oihh	%1,0x0080\n"	/* set RCP_PCL_BIT in new */
570 		"	csg	%0,%1,%2\n"
571 		"	jl	0b\n"
572 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
573 		: "Q" (ptep[PTRS_PER_PTE]) : "cc");
574 #endif
575 	return __pgste(new);
576 }
577 
578 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
579 {
580 #ifdef CONFIG_PGSTE
581 	asm(
582 		"	nihh	%1,0xff7f\n"	/* clear RCP_PCL_BIT */
583 		"	stg	%1,%0\n"
584 		: "=Q" (ptep[PTRS_PER_PTE])
585 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc");
586 	preempt_enable();
587 #endif
588 }
589 
590 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
591 {
592 #ifdef CONFIG_PGSTE
593 	unsigned long address, bits;
594 	unsigned char skey;
595 
596 	address = pte_val(*ptep) & PAGE_MASK;
597 	skey = page_get_storage_key(address);
598 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
599 	/* Clear page changed & referenced bit in the storage key */
600 	if (bits) {
601 		skey ^= bits;
602 		page_set_storage_key(address, skey, 1);
603 	}
604 	/* Transfer page changed & referenced bit to guest bits in pgste */
605 	pgste_val(pgste) |= bits << 48;		/* RCP_GR_BIT & RCP_GC_BIT */
606 	/* Get host changed & referenced bits from pgste */
607 	bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52;
608 	/* Clear host bits in pgste. */
609 	pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT);
610 	pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT);
611 	/* Copy page access key and fetch protection bit to pgste */
612 	pgste_val(pgste) |=
613 		(unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
614 	/* Transfer changed and referenced to kvm user bits */
615 	pgste_val(pgste) |= bits << 45;		/* KVM_UR_BIT & KVM_UC_BIT */
616 	/* Transfer changed & referenced to pte sofware bits */
617 	pte_val(*ptep) |= bits << 1;		/* _PAGE_SWR & _PAGE_SWC */
618 #endif
619 	return pgste;
620 
621 }
622 
623 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
624 {
625 #ifdef CONFIG_PGSTE
626 	int young;
627 
628 	young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK);
629 	/* Transfer page referenced bit to pte software bit (host view) */
630 	if (young || (pgste_val(pgste) & RCP_HR_BIT))
631 		pte_val(*ptep) |= _PAGE_SWR;
632 	/* Clear host referenced bit in pgste. */
633 	pgste_val(pgste) &= ~RCP_HR_BIT;
634 	/* Transfer page referenced bit to guest bit in pgste */
635 	pgste_val(pgste) |= (unsigned long) young << 50; /* set RCP_GR_BIT */
636 #endif
637 	return pgste;
638 
639 }
640 
641 static inline void pgste_set_pte(pte_t *ptep, pgste_t pgste)
642 {
643 #ifdef CONFIG_PGSTE
644 	unsigned long address;
645 	unsigned long okey, nkey;
646 
647 	address = pte_val(*ptep) & PAGE_MASK;
648 	okey = nkey = page_get_storage_key(address);
649 	nkey &= ~(_PAGE_ACC_BITS | _PAGE_FP_BIT);
650 	/* Set page access key and fetch protection bit from pgste */
651 	nkey |= (pgste_val(pgste) & (RCP_ACC_BITS | RCP_FP_BIT)) >> 56;
652 	if (okey != nkey)
653 		page_set_storage_key(address, nkey, 1);
654 #endif
655 }
656 
657 /**
658  * struct gmap_struct - guest address space
659  * @mm: pointer to the parent mm_struct
660  * @table: pointer to the page directory
661  * @crst_list: list of all crst tables used in the guest address space
662  */
663 struct gmap {
664 	struct list_head list;
665 	struct mm_struct *mm;
666 	unsigned long *table;
667 	struct list_head crst_list;
668 };
669 
670 /**
671  * struct gmap_rmap - reverse mapping for segment table entries
672  * @next: pointer to the next gmap_rmap structure in the list
673  * @entry: pointer to a segment table entry
674  */
675 struct gmap_rmap {
676 	struct list_head list;
677 	unsigned long *entry;
678 };
679 
680 /**
681  * struct gmap_pgtable - gmap information attached to a page table
682  * @vmaddr: address of the 1MB segment in the process virtual memory
683  * @mapper: list of segment table entries maping a page table
684  */
685 struct gmap_pgtable {
686 	unsigned long vmaddr;
687 	struct list_head mapper;
688 };
689 
690 struct gmap *gmap_alloc(struct mm_struct *mm);
691 void gmap_free(struct gmap *gmap);
692 void gmap_enable(struct gmap *gmap);
693 void gmap_disable(struct gmap *gmap);
694 int gmap_map_segment(struct gmap *gmap, unsigned long from,
695 		     unsigned long to, unsigned long length);
696 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
697 unsigned long gmap_fault(unsigned long address, struct gmap *);
698 
699 /*
700  * Certain architectures need to do special things when PTEs
701  * within a page table are directly modified.  Thus, the following
702  * hook is made available.
703  */
704 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
705 			      pte_t *ptep, pte_t entry)
706 {
707 	pgste_t pgste;
708 
709 	if (mm_has_pgste(mm)) {
710 		pgste = pgste_get_lock(ptep);
711 		pgste_set_pte(ptep, pgste);
712 		*ptep = entry;
713 		pgste_set_unlock(ptep, pgste);
714 	} else
715 		*ptep = entry;
716 }
717 
718 /*
719  * query functions pte_write/pte_dirty/pte_young only work if
720  * pte_present() is true. Undefined behaviour if not..
721  */
722 static inline int pte_write(pte_t pte)
723 {
724 	return (pte_val(pte) & _PAGE_RO) == 0;
725 }
726 
727 static inline int pte_dirty(pte_t pte)
728 {
729 #ifdef CONFIG_PGSTE
730 	if (pte_val(pte) & _PAGE_SWC)
731 		return 1;
732 #endif
733 	return 0;
734 }
735 
736 static inline int pte_young(pte_t pte)
737 {
738 #ifdef CONFIG_PGSTE
739 	if (pte_val(pte) & _PAGE_SWR)
740 		return 1;
741 #endif
742 	return 0;
743 }
744 
745 /*
746  * pgd/pmd/pte modification functions
747  */
748 
749 static inline void pgd_clear(pgd_t *pgd)
750 {
751 #ifdef __s390x__
752 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
753 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
754 #endif
755 }
756 
757 static inline void pud_clear(pud_t *pud)
758 {
759 #ifdef __s390x__
760 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
761 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
762 #endif
763 }
764 
765 static inline void pmd_clear(pmd_t *pmdp)
766 {
767 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
768 }
769 
770 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
771 {
772 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
773 }
774 
775 /*
776  * The following pte modification functions only work if
777  * pte_present() is true. Undefined behaviour if not..
778  */
779 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
780 {
781 	pte_val(pte) &= _PAGE_CHG_MASK;
782 	pte_val(pte) |= pgprot_val(newprot);
783 	return pte;
784 }
785 
786 static inline pte_t pte_wrprotect(pte_t pte)
787 {
788 	/* Do not clobber _PAGE_TYPE_NONE pages!  */
789 	if (!(pte_val(pte) & _PAGE_INVALID))
790 		pte_val(pte) |= _PAGE_RO;
791 	return pte;
792 }
793 
794 static inline pte_t pte_mkwrite(pte_t pte)
795 {
796 	pte_val(pte) &= ~_PAGE_RO;
797 	return pte;
798 }
799 
800 static inline pte_t pte_mkclean(pte_t pte)
801 {
802 #ifdef CONFIG_PGSTE
803 	pte_val(pte) &= ~_PAGE_SWC;
804 #endif
805 	return pte;
806 }
807 
808 static inline pte_t pte_mkdirty(pte_t pte)
809 {
810 	return pte;
811 }
812 
813 static inline pte_t pte_mkold(pte_t pte)
814 {
815 #ifdef CONFIG_PGSTE
816 	pte_val(pte) &= ~_PAGE_SWR;
817 #endif
818 	return pte;
819 }
820 
821 static inline pte_t pte_mkyoung(pte_t pte)
822 {
823 	return pte;
824 }
825 
826 static inline pte_t pte_mkspecial(pte_t pte)
827 {
828 	pte_val(pte) |= _PAGE_SPECIAL;
829 	return pte;
830 }
831 
832 #ifdef CONFIG_HUGETLB_PAGE
833 static inline pte_t pte_mkhuge(pte_t pte)
834 {
835 	/*
836 	 * PROT_NONE needs to be remapped from the pte type to the ste type.
837 	 * The HW invalid bit is also different for pte and ste. The pte
838 	 * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE
839 	 * bit, so we don't have to clear it.
840 	 */
841 	if (pte_val(pte) & _PAGE_INVALID) {
842 		if (pte_val(pte) & _PAGE_SWT)
843 			pte_val(pte) |= _HPAGE_TYPE_NONE;
844 		pte_val(pte) |= _SEGMENT_ENTRY_INV;
845 	}
846 	/*
847 	 * Clear SW pte bits SWT and SWX, there are no SW bits in a segment
848 	 * table entry.
849 	 */
850 	pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX);
851 	/*
852 	 * Also set the change-override bit because we don't need dirty bit
853 	 * tracking for hugetlbfs pages.
854 	 */
855 	pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
856 	return pte;
857 }
858 #endif
859 
860 /*
861  * Get (and clear) the user dirty bit for a pte.
862  */
863 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
864 						 pte_t *ptep)
865 {
866 	pgste_t pgste;
867 	int dirty = 0;
868 
869 	if (mm_has_pgste(mm)) {
870 		pgste = pgste_get_lock(ptep);
871 		pgste = pgste_update_all(ptep, pgste);
872 		dirty = !!(pgste_val(pgste) & KVM_UC_BIT);
873 		pgste_val(pgste) &= ~KVM_UC_BIT;
874 		pgste_set_unlock(ptep, pgste);
875 		return dirty;
876 	}
877 	return dirty;
878 }
879 
880 /*
881  * Get (and clear) the user referenced bit for a pte.
882  */
883 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
884 						 pte_t *ptep)
885 {
886 	pgste_t pgste;
887 	int young = 0;
888 
889 	if (mm_has_pgste(mm)) {
890 		pgste = pgste_get_lock(ptep);
891 		pgste = pgste_update_young(ptep, pgste);
892 		young = !!(pgste_val(pgste) & KVM_UR_BIT);
893 		pgste_val(pgste) &= ~KVM_UR_BIT;
894 		pgste_set_unlock(ptep, pgste);
895 	}
896 	return young;
897 }
898 
899 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
900 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
901 					    unsigned long addr, pte_t *ptep)
902 {
903 	pgste_t pgste;
904 	pte_t pte;
905 
906 	if (mm_has_pgste(vma->vm_mm)) {
907 		pgste = pgste_get_lock(ptep);
908 		pgste = pgste_update_young(ptep, pgste);
909 		pte = *ptep;
910 		*ptep = pte_mkold(pte);
911 		pgste_set_unlock(ptep, pgste);
912 		return pte_young(pte);
913 	}
914 	return 0;
915 }
916 
917 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
918 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
919 					 unsigned long address, pte_t *ptep)
920 {
921 	/* No need to flush TLB
922 	 * On s390 reference bits are in storage key and never in TLB
923 	 * With virtualization we handle the reference bit, without we
924 	 * we can simply return */
925 	return ptep_test_and_clear_young(vma, address, ptep);
926 }
927 
928 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
929 {
930 	if (!(pte_val(*ptep) & _PAGE_INVALID)) {
931 #ifndef __s390x__
932 		/* pto must point to the start of the segment table */
933 		pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
934 #else
935 		/* ipte in zarch mode can do the math */
936 		pte_t *pto = ptep;
937 #endif
938 		asm volatile(
939 			"	ipte	%2,%3"
940 			: "=m" (*ptep) : "m" (*ptep),
941 			  "a" (pto), "a" (address));
942 	}
943 }
944 
945 /*
946  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
947  * both clear the TLB for the unmapped pte. The reason is that
948  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
949  * to modify an active pte. The sequence is
950  *   1) ptep_get_and_clear
951  *   2) set_pte_at
952  *   3) flush_tlb_range
953  * On s390 the tlb needs to get flushed with the modification of the pte
954  * if the pte is active. The only way how this can be implemented is to
955  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
956  * is a nop.
957  */
958 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
959 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
960 				       unsigned long address, pte_t *ptep)
961 {
962 	pgste_t pgste;
963 	pte_t pte;
964 
965 	mm->context.flush_mm = 1;
966 	if (mm_has_pgste(mm))
967 		pgste = pgste_get_lock(ptep);
968 
969 	pte = *ptep;
970 	if (!mm_exclusive(mm))
971 		__ptep_ipte(address, ptep);
972 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
973 
974 	if (mm_has_pgste(mm)) {
975 		pgste = pgste_update_all(&pte, pgste);
976 		pgste_set_unlock(ptep, pgste);
977 	}
978 	return pte;
979 }
980 
981 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
982 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
983 					   unsigned long address,
984 					   pte_t *ptep)
985 {
986 	pte_t pte;
987 
988 	mm->context.flush_mm = 1;
989 	if (mm_has_pgste(mm))
990 		pgste_get_lock(ptep);
991 
992 	pte = *ptep;
993 	if (!mm_exclusive(mm))
994 		__ptep_ipte(address, ptep);
995 	return pte;
996 }
997 
998 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
999 					   unsigned long address,
1000 					   pte_t *ptep, pte_t pte)
1001 {
1002 	*ptep = pte;
1003 	if (mm_has_pgste(mm))
1004 		pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE));
1005 }
1006 
1007 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1008 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1009 				     unsigned long address, pte_t *ptep)
1010 {
1011 	pgste_t pgste;
1012 	pte_t pte;
1013 
1014 	if (mm_has_pgste(vma->vm_mm))
1015 		pgste = pgste_get_lock(ptep);
1016 
1017 	pte = *ptep;
1018 	__ptep_ipte(address, ptep);
1019 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1020 
1021 	if (mm_has_pgste(vma->vm_mm)) {
1022 		pgste = pgste_update_all(&pte, pgste);
1023 		pgste_set_unlock(ptep, pgste);
1024 	}
1025 	return pte;
1026 }
1027 
1028 /*
1029  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1030  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1031  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1032  * cannot be accessed while the batched unmap is running. In this case
1033  * full==1 and a simple pte_clear is enough. See tlb.h.
1034  */
1035 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1036 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1037 					    unsigned long address,
1038 					    pte_t *ptep, int full)
1039 {
1040 	pgste_t pgste;
1041 	pte_t pte;
1042 
1043 	if (mm_has_pgste(mm))
1044 		pgste = pgste_get_lock(ptep);
1045 
1046 	pte = *ptep;
1047 	if (!full)
1048 		__ptep_ipte(address, ptep);
1049 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1050 
1051 	if (mm_has_pgste(mm)) {
1052 		pgste = pgste_update_all(&pte, pgste);
1053 		pgste_set_unlock(ptep, pgste);
1054 	}
1055 	return pte;
1056 }
1057 
1058 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1059 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1060 				       unsigned long address, pte_t *ptep)
1061 {
1062 	pgste_t pgste;
1063 	pte_t pte = *ptep;
1064 
1065 	if (pte_write(pte)) {
1066 		mm->context.flush_mm = 1;
1067 		if (mm_has_pgste(mm))
1068 			pgste = pgste_get_lock(ptep);
1069 
1070 		if (!mm_exclusive(mm))
1071 			__ptep_ipte(address, ptep);
1072 		*ptep = pte_wrprotect(pte);
1073 
1074 		if (mm_has_pgste(mm))
1075 			pgste_set_unlock(ptep, pgste);
1076 	}
1077 	return pte;
1078 }
1079 
1080 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1081 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1082 					unsigned long address, pte_t *ptep,
1083 					pte_t entry, int dirty)
1084 {
1085 	pgste_t pgste;
1086 
1087 	if (pte_same(*ptep, entry))
1088 		return 0;
1089 	if (mm_has_pgste(vma->vm_mm))
1090 		pgste = pgste_get_lock(ptep);
1091 
1092 	__ptep_ipte(address, ptep);
1093 	*ptep = entry;
1094 
1095 	if (mm_has_pgste(vma->vm_mm))
1096 		pgste_set_unlock(ptep, pgste);
1097 	return 1;
1098 }
1099 
1100 /*
1101  * Conversion functions: convert a page and protection to a page entry,
1102  * and a page entry and page directory to the page they refer to.
1103  */
1104 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1105 {
1106 	pte_t __pte;
1107 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1108 	return __pte;
1109 }
1110 
1111 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1112 {
1113 	unsigned long physpage = page_to_phys(page);
1114 
1115 	return mk_pte_phys(physpage, pgprot);
1116 }
1117 
1118 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1119 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1120 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1121 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1122 
1123 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1124 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1125 
1126 #ifndef __s390x__
1127 
1128 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1129 #define pud_deref(pmd) ({ BUG(); 0UL; })
1130 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1131 
1132 #define pud_offset(pgd, address) ((pud_t *) pgd)
1133 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1134 
1135 #else /* __s390x__ */
1136 
1137 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1138 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1139 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1140 
1141 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1142 {
1143 	pud_t *pud = (pud_t *) pgd;
1144 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1145 		pud = (pud_t *) pgd_deref(*pgd);
1146 	return pud  + pud_index(address);
1147 }
1148 
1149 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1150 {
1151 	pmd_t *pmd = (pmd_t *) pud;
1152 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1153 		pmd = (pmd_t *) pud_deref(*pud);
1154 	return pmd + pmd_index(address);
1155 }
1156 
1157 #endif /* __s390x__ */
1158 
1159 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1160 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1161 #define pte_page(x) pfn_to_page(pte_pfn(x))
1162 
1163 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1164 
1165 /* Find an entry in the lowest level page table.. */
1166 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1167 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1168 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1169 #define pte_unmap(pte) do { } while (0)
1170 
1171 /*
1172  * 31 bit swap entry format:
1173  * A page-table entry has some bits we have to treat in a special way.
1174  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1175  * exception will occur instead of a page translation exception. The
1176  * specifiation exception has the bad habit not to store necessary
1177  * information in the lowcore.
1178  * Bit 21 and bit 22 are the page invalid bit and the page protection
1179  * bit. We set both to indicate a swapped page.
1180  * Bit 30 and 31 are used to distinguish the different page types. For
1181  * a swapped page these bits need to be zero.
1182  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1183  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1184  * plus 24 for the offset.
1185  * 0|     offset        |0110|o|type |00|
1186  * 0 0000000001111111111 2222 2 22222 33
1187  * 0 1234567890123456789 0123 4 56789 01
1188  *
1189  * 64 bit swap entry format:
1190  * A page-table entry has some bits we have to treat in a special way.
1191  * Bits 52 and bit 55 have to be zero, otherwise an specification
1192  * exception will occur instead of a page translation exception. The
1193  * specifiation exception has the bad habit not to store necessary
1194  * information in the lowcore.
1195  * Bit 53 and bit 54 are the page invalid bit and the page protection
1196  * bit. We set both to indicate a swapped page.
1197  * Bit 62 and 63 are used to distinguish the different page types. For
1198  * a swapped page these bits need to be zero.
1199  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1200  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1201  * plus 56 for the offset.
1202  * |                      offset                        |0110|o|type |00|
1203  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1204  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1205  */
1206 #ifndef __s390x__
1207 #define __SWP_OFFSET_MASK (~0UL >> 12)
1208 #else
1209 #define __SWP_OFFSET_MASK (~0UL >> 11)
1210 #endif
1211 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1212 {
1213 	pte_t pte;
1214 	offset &= __SWP_OFFSET_MASK;
1215 	pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
1216 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1217 	return pte;
1218 }
1219 
1220 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1221 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1222 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1223 
1224 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1225 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1226 
1227 #ifndef __s390x__
1228 # define PTE_FILE_MAX_BITS	26
1229 #else /* __s390x__ */
1230 # define PTE_FILE_MAX_BITS	59
1231 #endif /* __s390x__ */
1232 
1233 #define pte_to_pgoff(__pte) \
1234 	((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1235 
1236 #define pgoff_to_pte(__off) \
1237 	((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1238 		   | _PAGE_TYPE_FILE })
1239 
1240 #endif /* !__ASSEMBLY__ */
1241 
1242 #define kern_addr_valid(addr)   (1)
1243 
1244 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1245 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1246 extern int s390_enable_sie(void);
1247 
1248 /*
1249  * No page table caches to initialise
1250  */
1251 #define pgtable_cache_init()	do { } while (0)
1252 
1253 #include <asm-generic/pgtable.h>
1254 
1255 #endif /* _S390_PAGE_H */
1256