xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision 95e9fd10)
1 /*
2  *  S390 version
3  *    Copyright IBM Corp. 1999, 2000
4  *    Author(s): Hartmut Penner (hp@de.ibm.com)
5  *               Ulrich Weigand (weigand@de.ibm.com)
6  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
7  *
8  *  Derived from "include/asm-i386/pgtable.h"
9  */
10 
11 #ifndef _ASM_S390_PGTABLE_H
12 #define _ASM_S390_PGTABLE_H
13 
14 /*
15  * The Linux memory management assumes a three-level page table setup. For
16  * s390 31 bit we "fold" the mid level into the top-level page table, so
17  * that we physically have the same two-level page table as the s390 mmu
18  * expects in 31 bit mode. For s390 64 bit we use three of the five levels
19  * the hardware provides (region first and region second tables are not
20  * used).
21  *
22  * The "pgd_xxx()" functions are trivial for a folded two-level
23  * setup: the pgd is never bad, and a pmd always exists (as it's folded
24  * into the pgd entry)
25  *
26  * This file contains the functions and defines necessary to modify and use
27  * the S390 page table tree.
28  */
29 #ifndef __ASSEMBLY__
30 #include <linux/sched.h>
31 #include <linux/mm_types.h>
32 #include <asm/bug.h>
33 #include <asm/page.h>
34 
35 extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
36 extern void paging_init(void);
37 extern void vmem_map_init(void);
38 extern void fault_init(void);
39 
40 /*
41  * The S390 doesn't have any external MMU info: the kernel page
42  * tables contain all the necessary information.
43  */
44 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
45 
46 /*
47  * ZERO_PAGE is a global shared page that is always zero; used
48  * for zero-mapped memory areas etc..
49  */
50 
51 extern unsigned long empty_zero_page;
52 extern unsigned long zero_page_mask;
53 
54 #define ZERO_PAGE(vaddr) \
55 	(virt_to_page((void *)(empty_zero_page + \
56 	 (((unsigned long)(vaddr)) &zero_page_mask))))
57 
58 #define is_zero_pfn is_zero_pfn
59 static inline int is_zero_pfn(unsigned long pfn)
60 {
61 	extern unsigned long zero_pfn;
62 	unsigned long offset_from_zero_pfn = pfn - zero_pfn;
63 	return offset_from_zero_pfn <= (zero_page_mask >> PAGE_SHIFT);
64 }
65 
66 #define my_zero_pfn(addr)	page_to_pfn(ZERO_PAGE(addr))
67 
68 #endif /* !__ASSEMBLY__ */
69 
70 /*
71  * PMD_SHIFT determines the size of the area a second-level page
72  * table can map
73  * PGDIR_SHIFT determines what a third-level page table entry can map
74  */
75 #ifndef CONFIG_64BIT
76 # define PMD_SHIFT	20
77 # define PUD_SHIFT	20
78 # define PGDIR_SHIFT	20
79 #else /* CONFIG_64BIT */
80 # define PMD_SHIFT	20
81 # define PUD_SHIFT	31
82 # define PGDIR_SHIFT	42
83 #endif /* CONFIG_64BIT */
84 
85 #define PMD_SIZE        (1UL << PMD_SHIFT)
86 #define PMD_MASK        (~(PMD_SIZE-1))
87 #define PUD_SIZE	(1UL << PUD_SHIFT)
88 #define PUD_MASK	(~(PUD_SIZE-1))
89 #define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
90 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
91 
92 /*
93  * entries per page directory level: the S390 is two-level, so
94  * we don't really have any PMD directory physically.
95  * for S390 segment-table entries are combined to one PGD
96  * that leads to 1024 pte per pgd
97  */
98 #define PTRS_PER_PTE	256
99 #ifndef CONFIG_64BIT
100 #define PTRS_PER_PMD	1
101 #define PTRS_PER_PUD	1
102 #else /* CONFIG_64BIT */
103 #define PTRS_PER_PMD	2048
104 #define PTRS_PER_PUD	2048
105 #endif /* CONFIG_64BIT */
106 #define PTRS_PER_PGD	2048
107 
108 #define FIRST_USER_ADDRESS  0
109 
110 #define pte_ERROR(e) \
111 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
112 #define pmd_ERROR(e) \
113 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
114 #define pud_ERROR(e) \
115 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
116 #define pgd_ERROR(e) \
117 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
118 
119 #ifndef __ASSEMBLY__
120 /*
121  * The vmalloc area will always be on the topmost area of the kernel
122  * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc,
123  * which should be enough for any sane case.
124  * By putting vmalloc at the top, we maximise the gap between physical
125  * memory and vmalloc to catch misplaced memory accesses. As a side
126  * effect, this also makes sure that 64 bit module code cannot be used
127  * as system call address.
128  */
129 extern unsigned long VMALLOC_START;
130 extern unsigned long VMALLOC_END;
131 extern struct page *vmemmap;
132 
133 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
134 
135 /*
136  * A 31 bit pagetable entry of S390 has following format:
137  *  |   PFRA          |    |  OS  |
138  * 0                   0IP0
139  * 00000000001111111111222222222233
140  * 01234567890123456789012345678901
141  *
142  * I Page-Invalid Bit:    Page is not available for address-translation
143  * P Page-Protection Bit: Store access not possible for page
144  *
145  * A 31 bit segmenttable entry of S390 has following format:
146  *  |   P-table origin      |  |PTL
147  * 0                         IC
148  * 00000000001111111111222222222233
149  * 01234567890123456789012345678901
150  *
151  * I Segment-Invalid Bit:    Segment is not available for address-translation
152  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
153  * PTL Page-Table-Length:    Page-table length (PTL+1*16 entries -> up to 256)
154  *
155  * The 31 bit segmenttable origin of S390 has following format:
156  *
157  *  |S-table origin   |     | STL |
158  * X                   **GPS
159  * 00000000001111111111222222222233
160  * 01234567890123456789012345678901
161  *
162  * X Space-Switch event:
163  * G Segment-Invalid Bit:     *
164  * P Private-Space Bit:       Segment is not private (PoP 3-30)
165  * S Storage-Alteration:
166  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
167  *
168  * A 64 bit pagetable entry of S390 has following format:
169  * |			 PFRA			      |0IPC|  OS  |
170  * 0000000000111111111122222222223333333333444444444455555555556666
171  * 0123456789012345678901234567890123456789012345678901234567890123
172  *
173  * I Page-Invalid Bit:    Page is not available for address-translation
174  * P Page-Protection Bit: Store access not possible for page
175  * C Change-bit override: HW is not required to set change bit
176  *
177  * A 64 bit segmenttable entry of S390 has following format:
178  * |        P-table origin                              |      TT
179  * 0000000000111111111122222222223333333333444444444455555555556666
180  * 0123456789012345678901234567890123456789012345678901234567890123
181  *
182  * I Segment-Invalid Bit:    Segment is not available for address-translation
183  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
184  * P Page-Protection Bit: Store access not possible for page
185  * TT Type 00
186  *
187  * A 64 bit region table entry of S390 has following format:
188  * |        S-table origin                             |   TF  TTTL
189  * 0000000000111111111122222222223333333333444444444455555555556666
190  * 0123456789012345678901234567890123456789012345678901234567890123
191  *
192  * I Segment-Invalid Bit:    Segment is not available for address-translation
193  * TT Type 01
194  * TF
195  * TL Table length
196  *
197  * The 64 bit regiontable origin of S390 has following format:
198  * |      region table origon                          |       DTTL
199  * 0000000000111111111122222222223333333333444444444455555555556666
200  * 0123456789012345678901234567890123456789012345678901234567890123
201  *
202  * X Space-Switch event:
203  * G Segment-Invalid Bit:
204  * P Private-Space Bit:
205  * S Storage-Alteration:
206  * R Real space
207  * TL Table-Length:
208  *
209  * A storage key has the following format:
210  * | ACC |F|R|C|0|
211  *  0   3 4 5 6 7
212  * ACC: access key
213  * F  : fetch protection bit
214  * R  : referenced bit
215  * C  : changed bit
216  */
217 
218 /* Hardware bits in the page table entry */
219 #define _PAGE_CO	0x100		/* HW Change-bit override */
220 #define _PAGE_RO	0x200		/* HW read-only bit  */
221 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
222 
223 /* Software bits in the page table entry */
224 #define _PAGE_SWT	0x001		/* SW pte type bit t */
225 #define _PAGE_SWX	0x002		/* SW pte type bit x */
226 #define _PAGE_SWC	0x004		/* SW pte changed bit (for KVM) */
227 #define _PAGE_SWR	0x008		/* SW pte referenced bit (for KVM) */
228 #define _PAGE_SPECIAL	0x010		/* SW associated with special page */
229 #define __HAVE_ARCH_PTE_SPECIAL
230 
231 /* Set of bits not changed in pte_modify */
232 #define _PAGE_CHG_MASK	(PAGE_MASK | _PAGE_SPECIAL | _PAGE_SWC | _PAGE_SWR)
233 
234 /* Six different types of pages. */
235 #define _PAGE_TYPE_EMPTY	0x400
236 #define _PAGE_TYPE_NONE		0x401
237 #define _PAGE_TYPE_SWAP		0x403
238 #define _PAGE_TYPE_FILE		0x601	/* bit 0x002 is used for offset !! */
239 #define _PAGE_TYPE_RO		0x200
240 #define _PAGE_TYPE_RW		0x000
241 
242 /*
243  * Only four types for huge pages, using the invalid bit and protection bit
244  * of a segment table entry.
245  */
246 #define _HPAGE_TYPE_EMPTY	0x020	/* _SEGMENT_ENTRY_INV */
247 #define _HPAGE_TYPE_NONE	0x220
248 #define _HPAGE_TYPE_RO		0x200	/* _SEGMENT_ENTRY_RO  */
249 #define _HPAGE_TYPE_RW		0x000
250 
251 /*
252  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
253  * pte_none and pte_file to find out the pte type WITHOUT holding the page
254  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
255  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
256  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
257  * This change is done while holding the lock, but the intermediate step
258  * of a previously valid pte with the hw invalid bit set can be observed by
259  * handle_pte_fault. That makes it necessary that all valid pte types with
260  * the hw invalid bit set must be distinguishable from the four pte types
261  * empty, none, swap and file.
262  *
263  *			irxt  ipte  irxt
264  * _PAGE_TYPE_EMPTY	1000   ->   1000
265  * _PAGE_TYPE_NONE	1001   ->   1001
266  * _PAGE_TYPE_SWAP	1011   ->   1011
267  * _PAGE_TYPE_FILE	11?1   ->   11?1
268  * _PAGE_TYPE_RO	0100   ->   1100
269  * _PAGE_TYPE_RW	0000   ->   1000
270  *
271  * pte_none is true for bits combinations 1000, 1010, 1100, 1110
272  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
273  * pte_file is true for bits combinations 1101, 1111
274  * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
275  */
276 
277 #ifndef CONFIG_64BIT
278 
279 /* Bits in the segment table address-space-control-element */
280 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
281 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
282 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
283 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
284 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
285 
286 /* Bits in the segment table entry */
287 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
288 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
289 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
290 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
291 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
292 
293 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
294 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
295 
296 /* Page status table bits for virtualization */
297 #define RCP_ACC_BITS	0xf0000000UL
298 #define RCP_FP_BIT	0x08000000UL
299 #define RCP_PCL_BIT	0x00800000UL
300 #define RCP_HR_BIT	0x00400000UL
301 #define RCP_HC_BIT	0x00200000UL
302 #define RCP_GR_BIT	0x00040000UL
303 #define RCP_GC_BIT	0x00020000UL
304 
305 /* User dirty / referenced bit for KVM's migration feature */
306 #define KVM_UR_BIT	0x00008000UL
307 #define KVM_UC_BIT	0x00004000UL
308 
309 #else /* CONFIG_64BIT */
310 
311 /* Bits in the segment/region table address-space-control-element */
312 #define _ASCE_ORIGIN		~0xfffUL/* segment table origin		    */
313 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
314 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
315 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
316 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
317 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
318 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
319 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
320 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
321 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
322 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
323 
324 /* Bits in the region table entry */
325 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
326 #define _REGION_ENTRY_INV	0x20	/* invalid region table entry	    */
327 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
328 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
329 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
330 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
331 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
332 
333 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
334 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
335 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
336 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
337 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
338 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
339 
340 /* Bits in the segment table entry */
341 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* segment table origin		    */
342 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
343 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
344 
345 #define _SEGMENT_ENTRY		(0)
346 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
347 
348 #define _SEGMENT_ENTRY_LARGE	0x400	/* STE-format control, large page   */
349 #define _SEGMENT_ENTRY_CO	0x100	/* change-recording override   */
350 
351 /* Page status table bits for virtualization */
352 #define RCP_ACC_BITS	0xf000000000000000UL
353 #define RCP_FP_BIT	0x0800000000000000UL
354 #define RCP_PCL_BIT	0x0080000000000000UL
355 #define RCP_HR_BIT	0x0040000000000000UL
356 #define RCP_HC_BIT	0x0020000000000000UL
357 #define RCP_GR_BIT	0x0004000000000000UL
358 #define RCP_GC_BIT	0x0002000000000000UL
359 
360 /* User dirty / referenced bit for KVM's migration feature */
361 #define KVM_UR_BIT	0x0000800000000000UL
362 #define KVM_UC_BIT	0x0000400000000000UL
363 
364 #endif /* CONFIG_64BIT */
365 
366 /*
367  * A user page table pointer has the space-switch-event bit, the
368  * private-space-control bit and the storage-alteration-event-control
369  * bit set. A kernel page table pointer doesn't need them.
370  */
371 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
372 				 _ASCE_ALT_EVENT)
373 
374 /*
375  * Page protection definitions.
376  */
377 #define PAGE_NONE	__pgprot(_PAGE_TYPE_NONE)
378 #define PAGE_RO		__pgprot(_PAGE_TYPE_RO)
379 #define PAGE_RW		__pgprot(_PAGE_TYPE_RW)
380 
381 #define PAGE_KERNEL	PAGE_RW
382 #define PAGE_COPY	PAGE_RO
383 
384 /*
385  * On s390 the page table entry has an invalid bit and a read-only bit.
386  * Read permission implies execute permission and write permission
387  * implies read permission.
388  */
389          /*xwr*/
390 #define __P000	PAGE_NONE
391 #define __P001	PAGE_RO
392 #define __P010	PAGE_RO
393 #define __P011	PAGE_RO
394 #define __P100	PAGE_RO
395 #define __P101	PAGE_RO
396 #define __P110	PAGE_RO
397 #define __P111	PAGE_RO
398 
399 #define __S000	PAGE_NONE
400 #define __S001	PAGE_RO
401 #define __S010	PAGE_RW
402 #define __S011	PAGE_RW
403 #define __S100	PAGE_RO
404 #define __S101	PAGE_RO
405 #define __S110	PAGE_RW
406 #define __S111	PAGE_RW
407 
408 static inline int mm_exclusive(struct mm_struct *mm)
409 {
410 	return likely(mm == current->active_mm &&
411 		      atomic_read(&mm->context.attach_count) <= 1);
412 }
413 
414 static inline int mm_has_pgste(struct mm_struct *mm)
415 {
416 #ifdef CONFIG_PGSTE
417 	if (unlikely(mm->context.has_pgste))
418 		return 1;
419 #endif
420 	return 0;
421 }
422 /*
423  * pgd/pmd/pte query functions
424  */
425 #ifndef CONFIG_64BIT
426 
427 static inline int pgd_present(pgd_t pgd) { return 1; }
428 static inline int pgd_none(pgd_t pgd)    { return 0; }
429 static inline int pgd_bad(pgd_t pgd)     { return 0; }
430 
431 static inline int pud_present(pud_t pud) { return 1; }
432 static inline int pud_none(pud_t pud)	 { return 0; }
433 static inline int pud_bad(pud_t pud)	 { return 0; }
434 
435 #else /* CONFIG_64BIT */
436 
437 static inline int pgd_present(pgd_t pgd)
438 {
439 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
440 		return 1;
441 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
442 }
443 
444 static inline int pgd_none(pgd_t pgd)
445 {
446 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
447 		return 0;
448 	return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL;
449 }
450 
451 static inline int pgd_bad(pgd_t pgd)
452 {
453 	/*
454 	 * With dynamic page table levels the pgd can be a region table
455 	 * entry or a segment table entry. Check for the bit that are
456 	 * invalid for either table entry.
457 	 */
458 	unsigned long mask =
459 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
460 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
461 	return (pgd_val(pgd) & mask) != 0;
462 }
463 
464 static inline int pud_present(pud_t pud)
465 {
466 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
467 		return 1;
468 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
469 }
470 
471 static inline int pud_none(pud_t pud)
472 {
473 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
474 		return 0;
475 	return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
476 }
477 
478 static inline int pud_bad(pud_t pud)
479 {
480 	/*
481 	 * With dynamic page table levels the pud can be a region table
482 	 * entry or a segment table entry. Check for the bit that are
483 	 * invalid for either table entry.
484 	 */
485 	unsigned long mask =
486 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
487 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
488 	return (pud_val(pud) & mask) != 0;
489 }
490 
491 #endif /* CONFIG_64BIT */
492 
493 static inline int pmd_present(pmd_t pmd)
494 {
495 	return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL;
496 }
497 
498 static inline int pmd_none(pmd_t pmd)
499 {
500 	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL;
501 }
502 
503 static inline int pmd_bad(pmd_t pmd)
504 {
505 	unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
506 	return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
507 }
508 
509 static inline int pte_none(pte_t pte)
510 {
511 	return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
512 }
513 
514 static inline int pte_present(pte_t pte)
515 {
516 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
517 	return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
518 		(!(pte_val(pte) & _PAGE_INVALID) &&
519 		 !(pte_val(pte) & _PAGE_SWT));
520 }
521 
522 static inline int pte_file(pte_t pte)
523 {
524 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
525 	return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
526 }
527 
528 static inline int pte_special(pte_t pte)
529 {
530 	return (pte_val(pte) & _PAGE_SPECIAL);
531 }
532 
533 #define __HAVE_ARCH_PTE_SAME
534 static inline int pte_same(pte_t a, pte_t b)
535 {
536 	return pte_val(a) == pte_val(b);
537 }
538 
539 static inline pgste_t pgste_get_lock(pte_t *ptep)
540 {
541 	unsigned long new = 0;
542 #ifdef CONFIG_PGSTE
543 	unsigned long old;
544 
545 	preempt_disable();
546 	asm(
547 		"	lg	%0,%2\n"
548 		"0:	lgr	%1,%0\n"
549 		"	nihh	%0,0xff7f\n"	/* clear RCP_PCL_BIT in old */
550 		"	oihh	%1,0x0080\n"	/* set RCP_PCL_BIT in new */
551 		"	csg	%0,%1,%2\n"
552 		"	jl	0b\n"
553 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
554 		: "Q" (ptep[PTRS_PER_PTE]) : "cc");
555 #endif
556 	return __pgste(new);
557 }
558 
559 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
560 {
561 #ifdef CONFIG_PGSTE
562 	asm(
563 		"	nihh	%1,0xff7f\n"	/* clear RCP_PCL_BIT */
564 		"	stg	%1,%0\n"
565 		: "=Q" (ptep[PTRS_PER_PTE])
566 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc");
567 	preempt_enable();
568 #endif
569 }
570 
571 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
572 {
573 #ifdef CONFIG_PGSTE
574 	unsigned long address, bits;
575 	unsigned char skey;
576 
577 	if (!pte_present(*ptep))
578 		return pgste;
579 	address = pte_val(*ptep) & PAGE_MASK;
580 	skey = page_get_storage_key(address);
581 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
582 	/* Clear page changed & referenced bit in the storage key */
583 	if (bits & _PAGE_CHANGED)
584 		page_set_storage_key(address, skey ^ bits, 1);
585 	else if (bits)
586 		page_reset_referenced(address);
587 	/* Transfer page changed & referenced bit to guest bits in pgste */
588 	pgste_val(pgste) |= bits << 48;		/* RCP_GR_BIT & RCP_GC_BIT */
589 	/* Get host changed & referenced bits from pgste */
590 	bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52;
591 	/* Clear host bits in pgste. */
592 	pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT);
593 	pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT);
594 	/* Copy page access key and fetch protection bit to pgste */
595 	pgste_val(pgste) |=
596 		(unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
597 	/* Transfer changed and referenced to kvm user bits */
598 	pgste_val(pgste) |= bits << 45;		/* KVM_UR_BIT & KVM_UC_BIT */
599 	/* Transfer changed & referenced to pte sofware bits */
600 	pte_val(*ptep) |= bits << 1;		/* _PAGE_SWR & _PAGE_SWC */
601 #endif
602 	return pgste;
603 
604 }
605 
606 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
607 {
608 #ifdef CONFIG_PGSTE
609 	int young;
610 
611 	if (!pte_present(*ptep))
612 		return pgste;
613 	young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK);
614 	/* Transfer page referenced bit to pte software bit (host view) */
615 	if (young || (pgste_val(pgste) & RCP_HR_BIT))
616 		pte_val(*ptep) |= _PAGE_SWR;
617 	/* Clear host referenced bit in pgste. */
618 	pgste_val(pgste) &= ~RCP_HR_BIT;
619 	/* Transfer page referenced bit to guest bit in pgste */
620 	pgste_val(pgste) |= (unsigned long) young << 50; /* set RCP_GR_BIT */
621 #endif
622 	return pgste;
623 
624 }
625 
626 static inline void pgste_set_pte(pte_t *ptep, pgste_t pgste, pte_t entry)
627 {
628 #ifdef CONFIG_PGSTE
629 	unsigned long address;
630 	unsigned long okey, nkey;
631 
632 	if (!pte_present(entry))
633 		return;
634 	address = pte_val(entry) & PAGE_MASK;
635 	okey = nkey = page_get_storage_key(address);
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(address, nkey, 1);
641 #endif
642 }
643 
644 /**
645  * struct gmap_struct - guest address space
646  * @mm: pointer to the parent mm_struct
647  * @table: pointer to the page directory
648  * @asce: address space control element for gmap page table
649  * @crst_list: list of all crst tables used in the guest address space
650  */
651 struct gmap {
652 	struct list_head list;
653 	struct mm_struct *mm;
654 	unsigned long *table;
655 	unsigned long asce;
656 	struct list_head crst_list;
657 };
658 
659 /**
660  * struct gmap_rmap - reverse mapping for segment table entries
661  * @next: pointer to the next gmap_rmap structure in the list
662  * @entry: pointer to a segment table entry
663  */
664 struct gmap_rmap {
665 	struct list_head list;
666 	unsigned long *entry;
667 };
668 
669 /**
670  * struct gmap_pgtable - gmap information attached to a page table
671  * @vmaddr: address of the 1MB segment in the process virtual memory
672  * @mapper: list of segment table entries maping a page table
673  */
674 struct gmap_pgtable {
675 	unsigned long vmaddr;
676 	struct list_head mapper;
677 };
678 
679 struct gmap *gmap_alloc(struct mm_struct *mm);
680 void gmap_free(struct gmap *gmap);
681 void gmap_enable(struct gmap *gmap);
682 void gmap_disable(struct gmap *gmap);
683 int gmap_map_segment(struct gmap *gmap, unsigned long from,
684 		     unsigned long to, unsigned long length);
685 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
686 unsigned long __gmap_fault(unsigned long address, struct gmap *);
687 unsigned long gmap_fault(unsigned long address, struct gmap *);
688 void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
689 
690 /*
691  * Certain architectures need to do special things when PTEs
692  * within a page table are directly modified.  Thus, the following
693  * hook is made available.
694  */
695 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
696 			      pte_t *ptep, pte_t entry)
697 {
698 	pgste_t pgste;
699 
700 	if (mm_has_pgste(mm)) {
701 		pgste = pgste_get_lock(ptep);
702 		pgste_set_pte(ptep, pgste, entry);
703 		*ptep = entry;
704 		pgste_set_unlock(ptep, pgste);
705 	} else
706 		*ptep = entry;
707 }
708 
709 /*
710  * query functions pte_write/pte_dirty/pte_young only work if
711  * pte_present() is true. Undefined behaviour if not..
712  */
713 static inline int pte_write(pte_t pte)
714 {
715 	return (pte_val(pte) & _PAGE_RO) == 0;
716 }
717 
718 static inline int pte_dirty(pte_t pte)
719 {
720 #ifdef CONFIG_PGSTE
721 	if (pte_val(pte) & _PAGE_SWC)
722 		return 1;
723 #endif
724 	return 0;
725 }
726 
727 static inline int pte_young(pte_t pte)
728 {
729 #ifdef CONFIG_PGSTE
730 	if (pte_val(pte) & _PAGE_SWR)
731 		return 1;
732 #endif
733 	return 0;
734 }
735 
736 /*
737  * pgd/pmd/pte modification functions
738  */
739 
740 static inline void pgd_clear(pgd_t *pgd)
741 {
742 #ifdef CONFIG_64BIT
743 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
744 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
745 #endif
746 }
747 
748 static inline void pud_clear(pud_t *pud)
749 {
750 #ifdef CONFIG_64BIT
751 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
752 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
753 #endif
754 }
755 
756 static inline void pmd_clear(pmd_t *pmdp)
757 {
758 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
759 }
760 
761 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
762 {
763 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
764 }
765 
766 /*
767  * The following pte modification functions only work if
768  * pte_present() is true. Undefined behaviour if not..
769  */
770 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
771 {
772 	pte_val(pte) &= _PAGE_CHG_MASK;
773 	pte_val(pte) |= pgprot_val(newprot);
774 	return pte;
775 }
776 
777 static inline pte_t pte_wrprotect(pte_t pte)
778 {
779 	/* Do not clobber _PAGE_TYPE_NONE pages!  */
780 	if (!(pte_val(pte) & _PAGE_INVALID))
781 		pte_val(pte) |= _PAGE_RO;
782 	return pte;
783 }
784 
785 static inline pte_t pte_mkwrite(pte_t pte)
786 {
787 	pte_val(pte) &= ~_PAGE_RO;
788 	return pte;
789 }
790 
791 static inline pte_t pte_mkclean(pte_t pte)
792 {
793 #ifdef CONFIG_PGSTE
794 	pte_val(pte) &= ~_PAGE_SWC;
795 #endif
796 	return pte;
797 }
798 
799 static inline pte_t pte_mkdirty(pte_t pte)
800 {
801 	return pte;
802 }
803 
804 static inline pte_t pte_mkold(pte_t pte)
805 {
806 #ifdef CONFIG_PGSTE
807 	pte_val(pte) &= ~_PAGE_SWR;
808 #endif
809 	return pte;
810 }
811 
812 static inline pte_t pte_mkyoung(pte_t pte)
813 {
814 	return pte;
815 }
816 
817 static inline pte_t pte_mkspecial(pte_t pte)
818 {
819 	pte_val(pte) |= _PAGE_SPECIAL;
820 	return pte;
821 }
822 
823 #ifdef CONFIG_HUGETLB_PAGE
824 static inline pte_t pte_mkhuge(pte_t pte)
825 {
826 	/*
827 	 * PROT_NONE needs to be remapped from the pte type to the ste type.
828 	 * The HW invalid bit is also different for pte and ste. The pte
829 	 * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE
830 	 * bit, so we don't have to clear it.
831 	 */
832 	if (pte_val(pte) & _PAGE_INVALID) {
833 		if (pte_val(pte) & _PAGE_SWT)
834 			pte_val(pte) |= _HPAGE_TYPE_NONE;
835 		pte_val(pte) |= _SEGMENT_ENTRY_INV;
836 	}
837 	/*
838 	 * Clear SW pte bits SWT and SWX, there are no SW bits in a segment
839 	 * table entry.
840 	 */
841 	pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX);
842 	/*
843 	 * Also set the change-override bit because we don't need dirty bit
844 	 * tracking for hugetlbfs pages.
845 	 */
846 	pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
847 	return pte;
848 }
849 #endif
850 
851 /*
852  * Get (and clear) the user dirty bit for a pte.
853  */
854 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
855 						 pte_t *ptep)
856 {
857 	pgste_t pgste;
858 	int dirty = 0;
859 
860 	if (mm_has_pgste(mm)) {
861 		pgste = pgste_get_lock(ptep);
862 		pgste = pgste_update_all(ptep, pgste);
863 		dirty = !!(pgste_val(pgste) & KVM_UC_BIT);
864 		pgste_val(pgste) &= ~KVM_UC_BIT;
865 		pgste_set_unlock(ptep, pgste);
866 		return dirty;
867 	}
868 	return dirty;
869 }
870 
871 /*
872  * Get (and clear) the user referenced bit for a pte.
873  */
874 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
875 						 pte_t *ptep)
876 {
877 	pgste_t pgste;
878 	int young = 0;
879 
880 	if (mm_has_pgste(mm)) {
881 		pgste = pgste_get_lock(ptep);
882 		pgste = pgste_update_young(ptep, pgste);
883 		young = !!(pgste_val(pgste) & KVM_UR_BIT);
884 		pgste_val(pgste) &= ~KVM_UR_BIT;
885 		pgste_set_unlock(ptep, pgste);
886 	}
887 	return young;
888 }
889 
890 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
891 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
892 					    unsigned long addr, pte_t *ptep)
893 {
894 	pgste_t pgste;
895 	pte_t pte;
896 
897 	if (mm_has_pgste(vma->vm_mm)) {
898 		pgste = pgste_get_lock(ptep);
899 		pgste = pgste_update_young(ptep, pgste);
900 		pte = *ptep;
901 		*ptep = pte_mkold(pte);
902 		pgste_set_unlock(ptep, pgste);
903 		return pte_young(pte);
904 	}
905 	return 0;
906 }
907 
908 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
909 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
910 					 unsigned long address, pte_t *ptep)
911 {
912 	/* No need to flush TLB
913 	 * On s390 reference bits are in storage key and never in TLB
914 	 * With virtualization we handle the reference bit, without we
915 	 * we can simply return */
916 	return ptep_test_and_clear_young(vma, address, ptep);
917 }
918 
919 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
920 {
921 	if (!(pte_val(*ptep) & _PAGE_INVALID)) {
922 #ifndef CONFIG_64BIT
923 		/* pto must point to the start of the segment table */
924 		pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
925 #else
926 		/* ipte in zarch mode can do the math */
927 		pte_t *pto = ptep;
928 #endif
929 		asm volatile(
930 			"	ipte	%2,%3"
931 			: "=m" (*ptep) : "m" (*ptep),
932 			  "a" (pto), "a" (address));
933 	}
934 }
935 
936 /*
937  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
938  * both clear the TLB for the unmapped pte. The reason is that
939  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
940  * to modify an active pte. The sequence is
941  *   1) ptep_get_and_clear
942  *   2) set_pte_at
943  *   3) flush_tlb_range
944  * On s390 the tlb needs to get flushed with the modification of the pte
945  * if the pte is active. The only way how this can be implemented is to
946  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
947  * is a nop.
948  */
949 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
950 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
951 				       unsigned long address, pte_t *ptep)
952 {
953 	pgste_t pgste;
954 	pte_t pte;
955 
956 	mm->context.flush_mm = 1;
957 	if (mm_has_pgste(mm))
958 		pgste = pgste_get_lock(ptep);
959 
960 	pte = *ptep;
961 	if (!mm_exclusive(mm))
962 		__ptep_ipte(address, ptep);
963 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
964 
965 	if (mm_has_pgste(mm)) {
966 		pgste = pgste_update_all(&pte, pgste);
967 		pgste_set_unlock(ptep, pgste);
968 	}
969 	return pte;
970 }
971 
972 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
973 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
974 					   unsigned long address,
975 					   pte_t *ptep)
976 {
977 	pte_t pte;
978 
979 	mm->context.flush_mm = 1;
980 	if (mm_has_pgste(mm))
981 		pgste_get_lock(ptep);
982 
983 	pte = *ptep;
984 	if (!mm_exclusive(mm))
985 		__ptep_ipte(address, ptep);
986 	return pte;
987 }
988 
989 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
990 					   unsigned long address,
991 					   pte_t *ptep, pte_t pte)
992 {
993 	*ptep = pte;
994 	if (mm_has_pgste(mm))
995 		pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE));
996 }
997 
998 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
999 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1000 				     unsigned long address, pte_t *ptep)
1001 {
1002 	pgste_t pgste;
1003 	pte_t pte;
1004 
1005 	if (mm_has_pgste(vma->vm_mm))
1006 		pgste = pgste_get_lock(ptep);
1007 
1008 	pte = *ptep;
1009 	__ptep_ipte(address, ptep);
1010 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1011 
1012 	if (mm_has_pgste(vma->vm_mm)) {
1013 		pgste = pgste_update_all(&pte, pgste);
1014 		pgste_set_unlock(ptep, pgste);
1015 	}
1016 	return pte;
1017 }
1018 
1019 /*
1020  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1021  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1022  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1023  * cannot be accessed while the batched unmap is running. In this case
1024  * full==1 and a simple pte_clear is enough. See tlb.h.
1025  */
1026 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1027 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1028 					    unsigned long address,
1029 					    pte_t *ptep, int full)
1030 {
1031 	pgste_t pgste;
1032 	pte_t pte;
1033 
1034 	if (mm_has_pgste(mm))
1035 		pgste = pgste_get_lock(ptep);
1036 
1037 	pte = *ptep;
1038 	if (!full)
1039 		__ptep_ipte(address, ptep);
1040 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1041 
1042 	if (mm_has_pgste(mm)) {
1043 		pgste = pgste_update_all(&pte, pgste);
1044 		pgste_set_unlock(ptep, pgste);
1045 	}
1046 	return pte;
1047 }
1048 
1049 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1050 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1051 				       unsigned long address, pte_t *ptep)
1052 {
1053 	pgste_t pgste;
1054 	pte_t pte = *ptep;
1055 
1056 	if (pte_write(pte)) {
1057 		mm->context.flush_mm = 1;
1058 		if (mm_has_pgste(mm))
1059 			pgste = pgste_get_lock(ptep);
1060 
1061 		if (!mm_exclusive(mm))
1062 			__ptep_ipte(address, ptep);
1063 		*ptep = pte_wrprotect(pte);
1064 
1065 		if (mm_has_pgste(mm))
1066 			pgste_set_unlock(ptep, pgste);
1067 	}
1068 	return pte;
1069 }
1070 
1071 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1072 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1073 					unsigned long address, pte_t *ptep,
1074 					pte_t entry, int dirty)
1075 {
1076 	pgste_t pgste;
1077 
1078 	if (pte_same(*ptep, entry))
1079 		return 0;
1080 	if (mm_has_pgste(vma->vm_mm))
1081 		pgste = pgste_get_lock(ptep);
1082 
1083 	__ptep_ipte(address, ptep);
1084 	*ptep = entry;
1085 
1086 	if (mm_has_pgste(vma->vm_mm))
1087 		pgste_set_unlock(ptep, pgste);
1088 	return 1;
1089 }
1090 
1091 /*
1092  * Conversion functions: convert a page and protection to a page entry,
1093  * and a page entry and page directory to the page they refer to.
1094  */
1095 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1096 {
1097 	pte_t __pte;
1098 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1099 	return __pte;
1100 }
1101 
1102 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1103 {
1104 	unsigned long physpage = page_to_phys(page);
1105 
1106 	return mk_pte_phys(physpage, pgprot);
1107 }
1108 
1109 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1110 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1111 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1112 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1113 
1114 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1115 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1116 
1117 #ifndef CONFIG_64BIT
1118 
1119 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1120 #define pud_deref(pmd) ({ BUG(); 0UL; })
1121 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1122 
1123 #define pud_offset(pgd, address) ((pud_t *) pgd)
1124 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1125 
1126 #else /* CONFIG_64BIT */
1127 
1128 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1129 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1130 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1131 
1132 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1133 {
1134 	pud_t *pud = (pud_t *) pgd;
1135 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1136 		pud = (pud_t *) pgd_deref(*pgd);
1137 	return pud  + pud_index(address);
1138 }
1139 
1140 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1141 {
1142 	pmd_t *pmd = (pmd_t *) pud;
1143 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1144 		pmd = (pmd_t *) pud_deref(*pud);
1145 	return pmd + pmd_index(address);
1146 }
1147 
1148 #endif /* CONFIG_64BIT */
1149 
1150 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1151 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1152 #define pte_page(x) pfn_to_page(pte_pfn(x))
1153 
1154 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1155 
1156 /* Find an entry in the lowest level page table.. */
1157 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1158 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1159 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1160 #define pte_unmap(pte) do { } while (0)
1161 
1162 /*
1163  * 31 bit swap entry format:
1164  * A page-table entry has some bits we have to treat in a special way.
1165  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1166  * exception will occur instead of a page translation exception. The
1167  * specifiation exception has the bad habit not to store necessary
1168  * information in the lowcore.
1169  * Bit 21 and bit 22 are the page invalid bit and the page protection
1170  * bit. We set both to indicate a swapped page.
1171  * Bit 30 and 31 are used to distinguish the different page types. For
1172  * a swapped page these bits need to be zero.
1173  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1174  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1175  * plus 24 for the offset.
1176  * 0|     offset        |0110|o|type |00|
1177  * 0 0000000001111111111 2222 2 22222 33
1178  * 0 1234567890123456789 0123 4 56789 01
1179  *
1180  * 64 bit swap entry format:
1181  * A page-table entry has some bits we have to treat in a special way.
1182  * Bits 52 and bit 55 have to be zero, otherwise an specification
1183  * exception will occur instead of a page translation exception. The
1184  * specifiation exception has the bad habit not to store necessary
1185  * information in the lowcore.
1186  * Bit 53 and bit 54 are the page invalid bit and the page protection
1187  * bit. We set both to indicate a swapped page.
1188  * Bit 62 and 63 are used to distinguish the different page types. For
1189  * a swapped page these bits need to be zero.
1190  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1191  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1192  * plus 56 for the offset.
1193  * |                      offset                        |0110|o|type |00|
1194  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1195  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1196  */
1197 #ifndef CONFIG_64BIT
1198 #define __SWP_OFFSET_MASK (~0UL >> 12)
1199 #else
1200 #define __SWP_OFFSET_MASK (~0UL >> 11)
1201 #endif
1202 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1203 {
1204 	pte_t pte;
1205 	offset &= __SWP_OFFSET_MASK;
1206 	pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
1207 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1208 	return pte;
1209 }
1210 
1211 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1212 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1213 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1214 
1215 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1216 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1217 
1218 #ifndef CONFIG_64BIT
1219 # define PTE_FILE_MAX_BITS	26
1220 #else /* CONFIG_64BIT */
1221 # define PTE_FILE_MAX_BITS	59
1222 #endif /* CONFIG_64BIT */
1223 
1224 #define pte_to_pgoff(__pte) \
1225 	((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1226 
1227 #define pgoff_to_pte(__off) \
1228 	((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1229 		   | _PAGE_TYPE_FILE })
1230 
1231 #endif /* !__ASSEMBLY__ */
1232 
1233 #define kern_addr_valid(addr)   (1)
1234 
1235 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1236 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1237 extern int s390_enable_sie(void);
1238 
1239 /*
1240  * No page table caches to initialise
1241  */
1242 #define pgtable_cache_init()	do { } while (0)
1243 
1244 #include <asm-generic/pgtable.h>
1245 
1246 #endif /* _S390_PAGE_H */
1247