xref: /linux/arch/s390/include/asm/pgtable.h (revision 0d456bad36d42d16022be045c8a53ddbb59ee478)
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 
39 /*
40  * The S390 doesn't have any external MMU info: the kernel page
41  * tables contain all the necessary information.
42  */
43 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
44 #define update_mmu_cache_pmd(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 #define __HAVE_COLOR_ZERO_PAGE
58 
59 #endif /* !__ASSEMBLY__ */
60 
61 /*
62  * PMD_SHIFT determines the size of the area a second-level page
63  * table can map
64  * PGDIR_SHIFT determines what a third-level page table entry can map
65  */
66 #ifndef CONFIG_64BIT
67 # define PMD_SHIFT	20
68 # define PUD_SHIFT	20
69 # define PGDIR_SHIFT	20
70 #else /* CONFIG_64BIT */
71 # define PMD_SHIFT	20
72 # define PUD_SHIFT	31
73 # define PGDIR_SHIFT	42
74 #endif /* CONFIG_64BIT */
75 
76 #define PMD_SIZE        (1UL << PMD_SHIFT)
77 #define PMD_MASK        (~(PMD_SIZE-1))
78 #define PUD_SIZE	(1UL << PUD_SHIFT)
79 #define PUD_MASK	(~(PUD_SIZE-1))
80 #define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
81 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
82 
83 /*
84  * entries per page directory level: the S390 is two-level, so
85  * we don't really have any PMD directory physically.
86  * for S390 segment-table entries are combined to one PGD
87  * that leads to 1024 pte per pgd
88  */
89 #define PTRS_PER_PTE	256
90 #ifndef CONFIG_64BIT
91 #define PTRS_PER_PMD	1
92 #define PTRS_PER_PUD	1
93 #else /* CONFIG_64BIT */
94 #define PTRS_PER_PMD	2048
95 #define PTRS_PER_PUD	2048
96 #endif /* CONFIG_64BIT */
97 #define PTRS_PER_PGD	2048
98 
99 #define FIRST_USER_ADDRESS  0
100 
101 #define pte_ERROR(e) \
102 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
103 #define pmd_ERROR(e) \
104 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
105 #define pud_ERROR(e) \
106 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
107 #define pgd_ERROR(e) \
108 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
109 
110 #ifndef __ASSEMBLY__
111 /*
112  * The vmalloc and module area will always be on the topmost area of the kernel
113  * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc and modules.
114  * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
115  * modules will reside. That makes sure that inter module branches always
116  * happen without trampolines and in addition the placement within a 2GB frame
117  * is branch prediction unit friendly.
118  */
119 extern unsigned long VMALLOC_START;
120 extern unsigned long VMALLOC_END;
121 extern struct page *vmemmap;
122 
123 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
124 
125 #ifdef CONFIG_64BIT
126 extern unsigned long MODULES_VADDR;
127 extern unsigned long MODULES_END;
128 #define MODULES_VADDR	MODULES_VADDR
129 #define MODULES_END	MODULES_END
130 #define MODULES_LEN	(1UL << 31)
131 #endif
132 
133 /*
134  * A 31 bit pagetable entry of S390 has following format:
135  *  |   PFRA          |    |  OS  |
136  * 0                   0IP0
137  * 00000000001111111111222222222233
138  * 01234567890123456789012345678901
139  *
140  * I Page-Invalid Bit:    Page is not available for address-translation
141  * P Page-Protection Bit: Store access not possible for page
142  *
143  * A 31 bit segmenttable entry of S390 has following format:
144  *  |   P-table origin      |  |PTL
145  * 0                         IC
146  * 00000000001111111111222222222233
147  * 01234567890123456789012345678901
148  *
149  * I Segment-Invalid Bit:    Segment is not available for address-translation
150  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
151  * PTL Page-Table-Length:    Page-table length (PTL+1*16 entries -> up to 256)
152  *
153  * The 31 bit segmenttable origin of S390 has following format:
154  *
155  *  |S-table origin   |     | STL |
156  * X                   **GPS
157  * 00000000001111111111222222222233
158  * 01234567890123456789012345678901
159  *
160  * X Space-Switch event:
161  * G Segment-Invalid Bit:     *
162  * P Private-Space Bit:       Segment is not private (PoP 3-30)
163  * S Storage-Alteration:
164  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
165  *
166  * A 64 bit pagetable entry of S390 has following format:
167  * |			 PFRA			      |0IPC|  OS  |
168  * 0000000000111111111122222222223333333333444444444455555555556666
169  * 0123456789012345678901234567890123456789012345678901234567890123
170  *
171  * I Page-Invalid Bit:    Page is not available for address-translation
172  * P Page-Protection Bit: Store access not possible for page
173  * C Change-bit override: HW is not required to set change bit
174  *
175  * A 64 bit segmenttable entry of S390 has following format:
176  * |        P-table origin                              |      TT
177  * 0000000000111111111122222222223333333333444444444455555555556666
178  * 0123456789012345678901234567890123456789012345678901234567890123
179  *
180  * I Segment-Invalid Bit:    Segment is not available for address-translation
181  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
182  * P Page-Protection Bit: Store access not possible for page
183  * TT Type 00
184  *
185  * A 64 bit region table entry of S390 has following format:
186  * |        S-table origin                             |   TF  TTTL
187  * 0000000000111111111122222222223333333333444444444455555555556666
188  * 0123456789012345678901234567890123456789012345678901234567890123
189  *
190  * I Segment-Invalid Bit:    Segment is not available for address-translation
191  * TT Type 01
192  * TF
193  * TL Table length
194  *
195  * The 64 bit regiontable origin of S390 has following format:
196  * |      region table origon                          |       DTTL
197  * 0000000000111111111122222222223333333333444444444455555555556666
198  * 0123456789012345678901234567890123456789012345678901234567890123
199  *
200  * X Space-Switch event:
201  * G Segment-Invalid Bit:
202  * P Private-Space Bit:
203  * S Storage-Alteration:
204  * R Real space
205  * TL Table-Length:
206  *
207  * A storage key has the following format:
208  * | ACC |F|R|C|0|
209  *  0   3 4 5 6 7
210  * ACC: access key
211  * F  : fetch protection bit
212  * R  : referenced bit
213  * C  : changed bit
214  */
215 
216 /* Hardware bits in the page table entry */
217 #define _PAGE_CO	0x100		/* HW Change-bit override */
218 #define _PAGE_RO	0x200		/* HW read-only bit  */
219 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
220 
221 /* Software bits in the page table entry */
222 #define _PAGE_SWT	0x001		/* SW pte type bit t */
223 #define _PAGE_SWX	0x002		/* SW pte type bit x */
224 #define _PAGE_SWC	0x004		/* SW pte changed bit (for KVM) */
225 #define _PAGE_SWR	0x008		/* SW pte referenced bit (for KVM) */
226 #define _PAGE_SPECIAL	0x010		/* SW associated with special page */
227 #define __HAVE_ARCH_PTE_SPECIAL
228 
229 /* Set of bits not changed in pte_modify */
230 #define _PAGE_CHG_MASK	(PAGE_MASK | _PAGE_SPECIAL | _PAGE_SWC | _PAGE_SWR)
231 
232 /* Six different types of pages. */
233 #define _PAGE_TYPE_EMPTY	0x400
234 #define _PAGE_TYPE_NONE		0x401
235 #define _PAGE_TYPE_SWAP		0x403
236 #define _PAGE_TYPE_FILE		0x601	/* bit 0x002 is used for offset !! */
237 #define _PAGE_TYPE_RO		0x200
238 #define _PAGE_TYPE_RW		0x000
239 
240 /*
241  * Only four types for huge pages, using the invalid bit and protection bit
242  * of a segment table entry.
243  */
244 #define _HPAGE_TYPE_EMPTY	0x020	/* _SEGMENT_ENTRY_INV */
245 #define _HPAGE_TYPE_NONE	0x220
246 #define _HPAGE_TYPE_RO		0x200	/* _SEGMENT_ENTRY_RO  */
247 #define _HPAGE_TYPE_RW		0x000
248 
249 /*
250  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
251  * pte_none and pte_file to find out the pte type WITHOUT holding the page
252  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
253  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
254  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
255  * This change is done while holding the lock, but the intermediate step
256  * of a previously valid pte with the hw invalid bit set can be observed by
257  * handle_pte_fault. That makes it necessary that all valid pte types with
258  * the hw invalid bit set must be distinguishable from the four pte types
259  * empty, none, swap and file.
260  *
261  *			irxt  ipte  irxt
262  * _PAGE_TYPE_EMPTY	1000   ->   1000
263  * _PAGE_TYPE_NONE	1001   ->   1001
264  * _PAGE_TYPE_SWAP	1011   ->   1011
265  * _PAGE_TYPE_FILE	11?1   ->   11?1
266  * _PAGE_TYPE_RO	0100   ->   1100
267  * _PAGE_TYPE_RW	0000   ->   1000
268  *
269  * pte_none is true for bits combinations 1000, 1010, 1100, 1110
270  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
271  * pte_file is true for bits combinations 1101, 1111
272  * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
273  */
274 
275 #ifndef CONFIG_64BIT
276 
277 /* Bits in the segment table address-space-control-element */
278 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
279 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
280 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
281 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
282 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
283 
284 /* Bits in the segment table entry */
285 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
286 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
287 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
288 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
289 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
290 
291 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
292 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
293 
294 /* Page status table bits for virtualization */
295 #define RCP_ACC_BITS	0xf0000000UL
296 #define RCP_FP_BIT	0x08000000UL
297 #define RCP_PCL_BIT	0x00800000UL
298 #define RCP_HR_BIT	0x00400000UL
299 #define RCP_HC_BIT	0x00200000UL
300 #define RCP_GR_BIT	0x00040000UL
301 #define RCP_GC_BIT	0x00020000UL
302 
303 /* User dirty / referenced bit for KVM's migration feature */
304 #define KVM_UR_BIT	0x00008000UL
305 #define KVM_UC_BIT	0x00004000UL
306 
307 #else /* CONFIG_64BIT */
308 
309 /* Bits in the segment/region table address-space-control-element */
310 #define _ASCE_ORIGIN		~0xfffUL/* segment table origin		    */
311 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
312 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
313 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
314 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
315 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
316 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
317 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
318 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
319 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
320 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
321 
322 /* Bits in the region table entry */
323 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
324 #define _REGION_ENTRY_INV	0x20	/* invalid region table entry	    */
325 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
326 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
327 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
328 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
329 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
330 
331 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
332 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
333 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
334 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
335 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
336 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
337 
338 #define _REGION3_ENTRY_LARGE	0x400	/* RTTE-format control, large page  */
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 #define _SEGMENT_ENTRY_SPLIT_BIT 0	/* THP splitting bit number */
351 #define _SEGMENT_ENTRY_SPLIT	(1UL << _SEGMENT_ENTRY_SPLIT_BIT)
352 
353 /* Set of bits not changed in pmd_modify */
354 #define _SEGMENT_CHG_MASK	(_SEGMENT_ENTRY_ORIGIN | _SEGMENT_ENTRY_LARGE \
355 				 | _SEGMENT_ENTRY_SPLIT | _SEGMENT_ENTRY_CO)
356 
357 /* Page status table bits for virtualization */
358 #define RCP_ACC_BITS	0xf000000000000000UL
359 #define RCP_FP_BIT	0x0800000000000000UL
360 #define RCP_PCL_BIT	0x0080000000000000UL
361 #define RCP_HR_BIT	0x0040000000000000UL
362 #define RCP_HC_BIT	0x0020000000000000UL
363 #define RCP_GR_BIT	0x0004000000000000UL
364 #define RCP_GC_BIT	0x0002000000000000UL
365 
366 /* User dirty / referenced bit for KVM's migration feature */
367 #define KVM_UR_BIT	0x0000800000000000UL
368 #define KVM_UC_BIT	0x0000400000000000UL
369 
370 #endif /* CONFIG_64BIT */
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 CONFIG_64BIT
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_large(pud_t pud)	 { return 0; }
440 static inline int pud_bad(pud_t pud)	 { return 0; }
441 
442 #else /* CONFIG_64BIT */
443 
444 static inline int pgd_present(pgd_t pgd)
445 {
446 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
447 		return 1;
448 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
449 }
450 
451 static inline int pgd_none(pgd_t pgd)
452 {
453 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
454 		return 0;
455 	return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL;
456 }
457 
458 static inline int pgd_bad(pgd_t pgd)
459 {
460 	/*
461 	 * With dynamic page table levels the pgd can be a region table
462 	 * entry or a segment table entry. Check for the bit that are
463 	 * invalid for either table entry.
464 	 */
465 	unsigned long mask =
466 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
467 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
468 	return (pgd_val(pgd) & mask) != 0;
469 }
470 
471 static inline int pud_present(pud_t pud)
472 {
473 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
474 		return 1;
475 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
476 }
477 
478 static inline int pud_none(pud_t pud)
479 {
480 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
481 		return 0;
482 	return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
483 }
484 
485 static inline int pud_large(pud_t pud)
486 {
487 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
488 		return 0;
489 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
490 }
491 
492 static inline int pud_bad(pud_t pud)
493 {
494 	/*
495 	 * With dynamic page table levels the pud can be a region table
496 	 * entry or a segment table entry. Check for the bit that are
497 	 * invalid for either table entry.
498 	 */
499 	unsigned long mask =
500 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
501 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
502 	return (pud_val(pud) & mask) != 0;
503 }
504 
505 #endif /* CONFIG_64BIT */
506 
507 static inline int pmd_present(pmd_t pmd)
508 {
509 	unsigned long mask = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO;
510 	return (pmd_val(pmd) & mask) == _HPAGE_TYPE_NONE ||
511 	       !(pmd_val(pmd) & _SEGMENT_ENTRY_INV);
512 }
513 
514 static inline int pmd_none(pmd_t pmd)
515 {
516 	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) &&
517 	       !(pmd_val(pmd) & _SEGMENT_ENTRY_RO);
518 }
519 
520 static inline int pmd_large(pmd_t pmd)
521 {
522 #ifdef CONFIG_64BIT
523 	return !!(pmd_val(pmd) & _SEGMENT_ENTRY_LARGE);
524 #else
525 	return 0;
526 #endif
527 }
528 
529 static inline int pmd_bad(pmd_t pmd)
530 {
531 	unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
532 	return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
533 }
534 
535 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
536 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
537 				 unsigned long addr, pmd_t *pmdp);
538 
539 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
540 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
541 				 unsigned long address, pmd_t *pmdp,
542 				 pmd_t entry, int dirty);
543 
544 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
545 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
546 				  unsigned long address, pmd_t *pmdp);
547 
548 #define __HAVE_ARCH_PMD_WRITE
549 static inline int pmd_write(pmd_t pmd)
550 {
551 	return (pmd_val(pmd) & _SEGMENT_ENTRY_RO) == 0;
552 }
553 
554 static inline int pmd_young(pmd_t pmd)
555 {
556 	return 0;
557 }
558 
559 static inline int pte_none(pte_t pte)
560 {
561 	return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
562 }
563 
564 static inline int pte_present(pte_t pte)
565 {
566 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
567 	return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
568 		(!(pte_val(pte) & _PAGE_INVALID) &&
569 		 !(pte_val(pte) & _PAGE_SWT));
570 }
571 
572 static inline int pte_file(pte_t pte)
573 {
574 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
575 	return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
576 }
577 
578 static inline int pte_special(pte_t pte)
579 {
580 	return (pte_val(pte) & _PAGE_SPECIAL);
581 }
582 
583 #define __HAVE_ARCH_PTE_SAME
584 static inline int pte_same(pte_t a, pte_t b)
585 {
586 	return pte_val(a) == pte_val(b);
587 }
588 
589 static inline pgste_t pgste_get_lock(pte_t *ptep)
590 {
591 	unsigned long new = 0;
592 #ifdef CONFIG_PGSTE
593 	unsigned long old;
594 
595 	preempt_disable();
596 	asm(
597 		"	lg	%0,%2\n"
598 		"0:	lgr	%1,%0\n"
599 		"	nihh	%0,0xff7f\n"	/* clear RCP_PCL_BIT in old */
600 		"	oihh	%1,0x0080\n"	/* set RCP_PCL_BIT in new */
601 		"	csg	%0,%1,%2\n"
602 		"	jl	0b\n"
603 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
604 		: "Q" (ptep[PTRS_PER_PTE]) : "cc");
605 #endif
606 	return __pgste(new);
607 }
608 
609 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
610 {
611 #ifdef CONFIG_PGSTE
612 	asm(
613 		"	nihh	%1,0xff7f\n"	/* clear RCP_PCL_BIT */
614 		"	stg	%1,%0\n"
615 		: "=Q" (ptep[PTRS_PER_PTE])
616 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE]) : "cc");
617 	preempt_enable();
618 #endif
619 }
620 
621 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
622 {
623 #ifdef CONFIG_PGSTE
624 	unsigned long address, bits;
625 	unsigned char skey;
626 
627 	if (!pte_present(*ptep))
628 		return pgste;
629 	address = pte_val(*ptep) & PAGE_MASK;
630 	skey = page_get_storage_key(address);
631 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
632 	/* Clear page changed & referenced bit in the storage key */
633 	if (bits & _PAGE_CHANGED)
634 		page_set_storage_key(address, skey ^ bits, 1);
635 	else if (bits)
636 		page_reset_referenced(address);
637 	/* Transfer page changed & referenced bit to guest bits in pgste */
638 	pgste_val(pgste) |= bits << 48;		/* RCP_GR_BIT & RCP_GC_BIT */
639 	/* Get host changed & referenced bits from pgste */
640 	bits |= (pgste_val(pgste) & (RCP_HR_BIT | RCP_HC_BIT)) >> 52;
641 	/* Clear host bits in pgste. */
642 	pgste_val(pgste) &= ~(RCP_HR_BIT | RCP_HC_BIT);
643 	pgste_val(pgste) &= ~(RCP_ACC_BITS | RCP_FP_BIT);
644 	/* Copy page access key and fetch protection bit to pgste */
645 	pgste_val(pgste) |=
646 		(unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
647 	/* Transfer changed and referenced to kvm user bits */
648 	pgste_val(pgste) |= bits << 45;		/* KVM_UR_BIT & KVM_UC_BIT */
649 	/* Transfer changed & referenced to pte sofware bits */
650 	pte_val(*ptep) |= bits << 1;		/* _PAGE_SWR & _PAGE_SWC */
651 #endif
652 	return pgste;
653 
654 }
655 
656 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
657 {
658 #ifdef CONFIG_PGSTE
659 	int young;
660 
661 	if (!pte_present(*ptep))
662 		return pgste;
663 	young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK);
664 	/* Transfer page referenced bit to pte software bit (host view) */
665 	if (young || (pgste_val(pgste) & RCP_HR_BIT))
666 		pte_val(*ptep) |= _PAGE_SWR;
667 	/* Clear host referenced bit in pgste. */
668 	pgste_val(pgste) &= ~RCP_HR_BIT;
669 	/* Transfer page referenced bit to guest bit in pgste */
670 	pgste_val(pgste) |= (unsigned long) young << 50; /* set RCP_GR_BIT */
671 #endif
672 	return pgste;
673 
674 }
675 
676 static inline void pgste_set_pte(pte_t *ptep, pgste_t pgste, pte_t entry)
677 {
678 #ifdef CONFIG_PGSTE
679 	unsigned long address;
680 	unsigned long okey, nkey;
681 
682 	if (!pte_present(entry))
683 		return;
684 	address = pte_val(entry) & PAGE_MASK;
685 	okey = nkey = page_get_storage_key(address);
686 	nkey &= ~(_PAGE_ACC_BITS | _PAGE_FP_BIT);
687 	/* Set page access key and fetch protection bit from pgste */
688 	nkey |= (pgste_val(pgste) & (RCP_ACC_BITS | RCP_FP_BIT)) >> 56;
689 	if (okey != nkey)
690 		page_set_storage_key(address, nkey, 1);
691 #endif
692 }
693 
694 /**
695  * struct gmap_struct - guest address space
696  * @mm: pointer to the parent mm_struct
697  * @table: pointer to the page directory
698  * @asce: address space control element for gmap page table
699  * @crst_list: list of all crst tables used in the guest address space
700  */
701 struct gmap {
702 	struct list_head list;
703 	struct mm_struct *mm;
704 	unsigned long *table;
705 	unsigned long asce;
706 	struct list_head crst_list;
707 };
708 
709 /**
710  * struct gmap_rmap - reverse mapping for segment table entries
711  * @next: pointer to the next gmap_rmap structure in the list
712  * @entry: pointer to a segment table entry
713  */
714 struct gmap_rmap {
715 	struct list_head list;
716 	unsigned long *entry;
717 };
718 
719 /**
720  * struct gmap_pgtable - gmap information attached to a page table
721  * @vmaddr: address of the 1MB segment in the process virtual memory
722  * @mapper: list of segment table entries maping a page table
723  */
724 struct gmap_pgtable {
725 	unsigned long vmaddr;
726 	struct list_head mapper;
727 };
728 
729 struct gmap *gmap_alloc(struct mm_struct *mm);
730 void gmap_free(struct gmap *gmap);
731 void gmap_enable(struct gmap *gmap);
732 void gmap_disable(struct gmap *gmap);
733 int gmap_map_segment(struct gmap *gmap, unsigned long from,
734 		     unsigned long to, unsigned long length);
735 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
736 unsigned long __gmap_fault(unsigned long address, struct gmap *);
737 unsigned long gmap_fault(unsigned long address, struct gmap *);
738 void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
739 
740 /*
741  * Certain architectures need to do special things when PTEs
742  * within a page table are directly modified.  Thus, the following
743  * hook is made available.
744  */
745 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
746 			      pte_t *ptep, pte_t entry)
747 {
748 	pgste_t pgste;
749 
750 	if (mm_has_pgste(mm)) {
751 		pgste = pgste_get_lock(ptep);
752 		pgste_set_pte(ptep, pgste, entry);
753 		*ptep = entry;
754 		pgste_set_unlock(ptep, pgste);
755 	} else
756 		*ptep = entry;
757 }
758 
759 /*
760  * query functions pte_write/pte_dirty/pte_young only work if
761  * pte_present() is true. Undefined behaviour if not..
762  */
763 static inline int pte_write(pte_t pte)
764 {
765 	return (pte_val(pte) & _PAGE_RO) == 0;
766 }
767 
768 static inline int pte_dirty(pte_t pte)
769 {
770 #ifdef CONFIG_PGSTE
771 	if (pte_val(pte) & _PAGE_SWC)
772 		return 1;
773 #endif
774 	return 0;
775 }
776 
777 static inline int pte_young(pte_t pte)
778 {
779 #ifdef CONFIG_PGSTE
780 	if (pte_val(pte) & _PAGE_SWR)
781 		return 1;
782 #endif
783 	return 0;
784 }
785 
786 /*
787  * pgd/pmd/pte modification functions
788  */
789 
790 static inline void pgd_clear(pgd_t *pgd)
791 {
792 #ifdef CONFIG_64BIT
793 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
794 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
795 #endif
796 }
797 
798 static inline void pud_clear(pud_t *pud)
799 {
800 #ifdef CONFIG_64BIT
801 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
802 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
803 #endif
804 }
805 
806 static inline void pmd_clear(pmd_t *pmdp)
807 {
808 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
809 }
810 
811 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
812 {
813 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
814 }
815 
816 /*
817  * The following pte modification functions only work if
818  * pte_present() is true. Undefined behaviour if not..
819  */
820 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
821 {
822 	pte_val(pte) &= _PAGE_CHG_MASK;
823 	pte_val(pte) |= pgprot_val(newprot);
824 	return pte;
825 }
826 
827 static inline pte_t pte_wrprotect(pte_t pte)
828 {
829 	/* Do not clobber _PAGE_TYPE_NONE pages!  */
830 	if (!(pte_val(pte) & _PAGE_INVALID))
831 		pte_val(pte) |= _PAGE_RO;
832 	return pte;
833 }
834 
835 static inline pte_t pte_mkwrite(pte_t pte)
836 {
837 	pte_val(pte) &= ~_PAGE_RO;
838 	return pte;
839 }
840 
841 static inline pte_t pte_mkclean(pte_t pte)
842 {
843 #ifdef CONFIG_PGSTE
844 	pte_val(pte) &= ~_PAGE_SWC;
845 #endif
846 	return pte;
847 }
848 
849 static inline pte_t pte_mkdirty(pte_t pte)
850 {
851 	return pte;
852 }
853 
854 static inline pte_t pte_mkold(pte_t pte)
855 {
856 #ifdef CONFIG_PGSTE
857 	pte_val(pte) &= ~_PAGE_SWR;
858 #endif
859 	return pte;
860 }
861 
862 static inline pte_t pte_mkyoung(pte_t pte)
863 {
864 	return pte;
865 }
866 
867 static inline pte_t pte_mkspecial(pte_t pte)
868 {
869 	pte_val(pte) |= _PAGE_SPECIAL;
870 	return pte;
871 }
872 
873 #ifdef CONFIG_HUGETLB_PAGE
874 static inline pte_t pte_mkhuge(pte_t pte)
875 {
876 	/*
877 	 * PROT_NONE needs to be remapped from the pte type to the ste type.
878 	 * The HW invalid bit is also different for pte and ste. The pte
879 	 * invalid bit happens to be the same as the ste _SEGMENT_ENTRY_LARGE
880 	 * bit, so we don't have to clear it.
881 	 */
882 	if (pte_val(pte) & _PAGE_INVALID) {
883 		if (pte_val(pte) & _PAGE_SWT)
884 			pte_val(pte) |= _HPAGE_TYPE_NONE;
885 		pte_val(pte) |= _SEGMENT_ENTRY_INV;
886 	}
887 	/*
888 	 * Clear SW pte bits SWT and SWX, there are no SW bits in a segment
889 	 * table entry.
890 	 */
891 	pte_val(pte) &= ~(_PAGE_SWT | _PAGE_SWX);
892 	/*
893 	 * Also set the change-override bit because we don't need dirty bit
894 	 * tracking for hugetlbfs pages.
895 	 */
896 	pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
897 	return pte;
898 }
899 #endif
900 
901 /*
902  * Get (and clear) the user dirty bit for a pte.
903  */
904 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
905 						 pte_t *ptep)
906 {
907 	pgste_t pgste;
908 	int dirty = 0;
909 
910 	if (mm_has_pgste(mm)) {
911 		pgste = pgste_get_lock(ptep);
912 		pgste = pgste_update_all(ptep, pgste);
913 		dirty = !!(pgste_val(pgste) & KVM_UC_BIT);
914 		pgste_val(pgste) &= ~KVM_UC_BIT;
915 		pgste_set_unlock(ptep, pgste);
916 		return dirty;
917 	}
918 	return dirty;
919 }
920 
921 /*
922  * Get (and clear) the user referenced bit for a pte.
923  */
924 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
925 						 pte_t *ptep)
926 {
927 	pgste_t pgste;
928 	int young = 0;
929 
930 	if (mm_has_pgste(mm)) {
931 		pgste = pgste_get_lock(ptep);
932 		pgste = pgste_update_young(ptep, pgste);
933 		young = !!(pgste_val(pgste) & KVM_UR_BIT);
934 		pgste_val(pgste) &= ~KVM_UR_BIT;
935 		pgste_set_unlock(ptep, pgste);
936 	}
937 	return young;
938 }
939 
940 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
941 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
942 					    unsigned long addr, pte_t *ptep)
943 {
944 	pgste_t pgste;
945 	pte_t pte;
946 
947 	if (mm_has_pgste(vma->vm_mm)) {
948 		pgste = pgste_get_lock(ptep);
949 		pgste = pgste_update_young(ptep, pgste);
950 		pte = *ptep;
951 		*ptep = pte_mkold(pte);
952 		pgste_set_unlock(ptep, pgste);
953 		return pte_young(pte);
954 	}
955 	return 0;
956 }
957 
958 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
959 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
960 					 unsigned long address, pte_t *ptep)
961 {
962 	/* No need to flush TLB
963 	 * On s390 reference bits are in storage key and never in TLB
964 	 * With virtualization we handle the reference bit, without we
965 	 * we can simply return */
966 	return ptep_test_and_clear_young(vma, address, ptep);
967 }
968 
969 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
970 {
971 	if (!(pte_val(*ptep) & _PAGE_INVALID)) {
972 #ifndef CONFIG_64BIT
973 		/* pto must point to the start of the segment table */
974 		pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
975 #else
976 		/* ipte in zarch mode can do the math */
977 		pte_t *pto = ptep;
978 #endif
979 		asm volatile(
980 			"	ipte	%2,%3"
981 			: "=m" (*ptep) : "m" (*ptep),
982 			  "a" (pto), "a" (address));
983 	}
984 }
985 
986 /*
987  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
988  * both clear the TLB for the unmapped pte. The reason is that
989  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
990  * to modify an active pte. The sequence is
991  *   1) ptep_get_and_clear
992  *   2) set_pte_at
993  *   3) flush_tlb_range
994  * On s390 the tlb needs to get flushed with the modification of the pte
995  * if the pte is active. The only way how this can be implemented is to
996  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
997  * is a nop.
998  */
999 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1000 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1001 				       unsigned long address, pte_t *ptep)
1002 {
1003 	pgste_t pgste;
1004 	pte_t pte;
1005 
1006 	mm->context.flush_mm = 1;
1007 	if (mm_has_pgste(mm))
1008 		pgste = pgste_get_lock(ptep);
1009 
1010 	pte = *ptep;
1011 	if (!mm_exclusive(mm))
1012 		__ptep_ipte(address, ptep);
1013 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1014 
1015 	if (mm_has_pgste(mm)) {
1016 		pgste = pgste_update_all(&pte, pgste);
1017 		pgste_set_unlock(ptep, pgste);
1018 	}
1019 	return pte;
1020 }
1021 
1022 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1023 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
1024 					   unsigned long address,
1025 					   pte_t *ptep)
1026 {
1027 	pte_t pte;
1028 
1029 	mm->context.flush_mm = 1;
1030 	if (mm_has_pgste(mm))
1031 		pgste_get_lock(ptep);
1032 
1033 	pte = *ptep;
1034 	if (!mm_exclusive(mm))
1035 		__ptep_ipte(address, ptep);
1036 	return pte;
1037 }
1038 
1039 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
1040 					   unsigned long address,
1041 					   pte_t *ptep, pte_t pte)
1042 {
1043 	*ptep = pte;
1044 	if (mm_has_pgste(mm))
1045 		pgste_set_unlock(ptep, *(pgste_t *)(ptep + PTRS_PER_PTE));
1046 }
1047 
1048 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1049 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1050 				     unsigned long address, pte_t *ptep)
1051 {
1052 	pgste_t pgste;
1053 	pte_t pte;
1054 
1055 	if (mm_has_pgste(vma->vm_mm))
1056 		pgste = pgste_get_lock(ptep);
1057 
1058 	pte = *ptep;
1059 	__ptep_ipte(address, ptep);
1060 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1061 
1062 	if (mm_has_pgste(vma->vm_mm)) {
1063 		pgste = pgste_update_all(&pte, pgste);
1064 		pgste_set_unlock(ptep, pgste);
1065 	}
1066 	return pte;
1067 }
1068 
1069 /*
1070  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1071  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1072  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1073  * cannot be accessed while the batched unmap is running. In this case
1074  * full==1 and a simple pte_clear is enough. See tlb.h.
1075  */
1076 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1077 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1078 					    unsigned long address,
1079 					    pte_t *ptep, int full)
1080 {
1081 	pgste_t pgste;
1082 	pte_t pte;
1083 
1084 	if (mm_has_pgste(mm))
1085 		pgste = pgste_get_lock(ptep);
1086 
1087 	pte = *ptep;
1088 	if (!full)
1089 		__ptep_ipte(address, ptep);
1090 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1091 
1092 	if (mm_has_pgste(mm)) {
1093 		pgste = pgste_update_all(&pte, pgste);
1094 		pgste_set_unlock(ptep, pgste);
1095 	}
1096 	return pte;
1097 }
1098 
1099 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1100 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1101 				       unsigned long address, pte_t *ptep)
1102 {
1103 	pgste_t pgste;
1104 	pte_t pte = *ptep;
1105 
1106 	if (pte_write(pte)) {
1107 		mm->context.flush_mm = 1;
1108 		if (mm_has_pgste(mm))
1109 			pgste = pgste_get_lock(ptep);
1110 
1111 		if (!mm_exclusive(mm))
1112 			__ptep_ipte(address, ptep);
1113 		*ptep = pte_wrprotect(pte);
1114 
1115 		if (mm_has_pgste(mm))
1116 			pgste_set_unlock(ptep, pgste);
1117 	}
1118 	return pte;
1119 }
1120 
1121 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1122 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1123 					unsigned long address, pte_t *ptep,
1124 					pte_t entry, int dirty)
1125 {
1126 	pgste_t pgste;
1127 
1128 	if (pte_same(*ptep, entry))
1129 		return 0;
1130 	if (mm_has_pgste(vma->vm_mm))
1131 		pgste = pgste_get_lock(ptep);
1132 
1133 	__ptep_ipte(address, ptep);
1134 	*ptep = entry;
1135 
1136 	if (mm_has_pgste(vma->vm_mm))
1137 		pgste_set_unlock(ptep, pgste);
1138 	return 1;
1139 }
1140 
1141 /*
1142  * Conversion functions: convert a page and protection to a page entry,
1143  * and a page entry and page directory to the page they refer to.
1144  */
1145 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1146 {
1147 	pte_t __pte;
1148 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1149 	return __pte;
1150 }
1151 
1152 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1153 {
1154 	unsigned long physpage = page_to_phys(page);
1155 
1156 	return mk_pte_phys(physpage, pgprot);
1157 }
1158 
1159 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1160 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1161 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1162 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1163 
1164 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1165 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1166 
1167 #ifndef CONFIG_64BIT
1168 
1169 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1170 #define pud_deref(pmd) ({ BUG(); 0UL; })
1171 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1172 
1173 #define pud_offset(pgd, address) ((pud_t *) pgd)
1174 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1175 
1176 #else /* CONFIG_64BIT */
1177 
1178 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1179 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1180 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1181 
1182 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1183 {
1184 	pud_t *pud = (pud_t *) pgd;
1185 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1186 		pud = (pud_t *) pgd_deref(*pgd);
1187 	return pud  + pud_index(address);
1188 }
1189 
1190 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1191 {
1192 	pmd_t *pmd = (pmd_t *) pud;
1193 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1194 		pmd = (pmd_t *) pud_deref(*pud);
1195 	return pmd + pmd_index(address);
1196 }
1197 
1198 #endif /* CONFIG_64BIT */
1199 
1200 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1201 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1202 #define pte_page(x) pfn_to_page(pte_pfn(x))
1203 
1204 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1205 
1206 /* Find an entry in the lowest level page table.. */
1207 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1208 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1209 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1210 #define pte_unmap(pte) do { } while (0)
1211 
1212 static inline void __pmd_idte(unsigned long address, pmd_t *pmdp)
1213 {
1214 	unsigned long sto = (unsigned long) pmdp -
1215 			    pmd_index(address) * sizeof(pmd_t);
1216 
1217 	if (!(pmd_val(*pmdp) & _SEGMENT_ENTRY_INV)) {
1218 		asm volatile(
1219 			"	.insn	rrf,0xb98e0000,%2,%3,0,0"
1220 			: "=m" (*pmdp)
1221 			: "m" (*pmdp), "a" (sto),
1222 			  "a" ((address & HPAGE_MASK))
1223 			: "cc"
1224 		);
1225 	}
1226 }
1227 
1228 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1229 
1230 #define SEGMENT_NONE	__pgprot(_HPAGE_TYPE_NONE)
1231 #define SEGMENT_RO	__pgprot(_HPAGE_TYPE_RO)
1232 #define SEGMENT_RW	__pgprot(_HPAGE_TYPE_RW)
1233 
1234 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1235 extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable);
1236 
1237 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1238 extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm);
1239 
1240 static inline int pmd_trans_splitting(pmd_t pmd)
1241 {
1242 	return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT;
1243 }
1244 
1245 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1246 			      pmd_t *pmdp, pmd_t entry)
1247 {
1248 	*pmdp = entry;
1249 }
1250 
1251 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1252 {
1253 	/*
1254 	 * pgprot is PAGE_NONE, PAGE_RO, or PAGE_RW (see __Pxxx / __Sxxx)
1255 	 * Convert to segment table entry format.
1256 	 */
1257 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1258 		return pgprot_val(SEGMENT_NONE);
1259 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1260 		return pgprot_val(SEGMENT_RO);
1261 	return pgprot_val(SEGMENT_RW);
1262 }
1263 
1264 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1265 {
1266 	pmd_val(pmd) &= _SEGMENT_CHG_MASK;
1267 	pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1268 	return pmd;
1269 }
1270 
1271 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1272 {
1273 	pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1274 	return pmd;
1275 }
1276 
1277 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1278 {
1279 	/* Do not clobber _HPAGE_TYPE_NONE pages! */
1280 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_INV))
1281 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_RO;
1282 	return pmd;
1283 }
1284 
1285 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1286 {
1287 	pmd_val(pmd) |= _SEGMENT_ENTRY_RO;
1288 	return pmd;
1289 }
1290 
1291 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1292 {
1293 	/* No dirty bit in the segment table entry. */
1294 	return pmd;
1295 }
1296 
1297 static inline pmd_t pmd_mkold(pmd_t pmd)
1298 {
1299 	/* No referenced bit in the segment table entry. */
1300 	return pmd;
1301 }
1302 
1303 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1304 {
1305 	/* No referenced bit in the segment table entry. */
1306 	return pmd;
1307 }
1308 
1309 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1310 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1311 					    unsigned long address, pmd_t *pmdp)
1312 {
1313 	unsigned long pmd_addr = pmd_val(*pmdp) & HPAGE_MASK;
1314 	long tmp, rc;
1315 	int counter;
1316 
1317 	rc = 0;
1318 	if (MACHINE_HAS_RRBM) {
1319 		counter = PTRS_PER_PTE >> 6;
1320 		asm volatile(
1321 			"0:	.insn	rre,0xb9ae0000,%0,%3\n"	/* rrbm */
1322 			"	ogr	%1,%0\n"
1323 			"	la	%3,0(%4,%3)\n"
1324 			"	brct	%2,0b\n"
1325 			: "=&d" (tmp), "+&d" (rc), "+d" (counter),
1326 			  "+a" (pmd_addr)
1327 			: "a" (64 * 4096UL) : "cc");
1328 		rc = !!rc;
1329 	} else {
1330 		counter = PTRS_PER_PTE;
1331 		asm volatile(
1332 			"0:	rrbe	0,%2\n"
1333 			"	la	%2,0(%3,%2)\n"
1334 			"	brc	12,1f\n"
1335 			"	lhi	%0,1\n"
1336 			"1:	brct	%1,0b\n"
1337 			: "+d" (rc), "+d" (counter), "+a" (pmd_addr)
1338 			: "a" (4096UL) : "cc");
1339 	}
1340 	return rc;
1341 }
1342 
1343 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
1344 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
1345 				       unsigned long address, pmd_t *pmdp)
1346 {
1347 	pmd_t pmd = *pmdp;
1348 
1349 	__pmd_idte(address, pmdp);
1350 	pmd_clear(pmdp);
1351 	return pmd;
1352 }
1353 
1354 #define __HAVE_ARCH_PMDP_CLEAR_FLUSH
1355 static inline pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
1356 				     unsigned long address, pmd_t *pmdp)
1357 {
1358 	return pmdp_get_and_clear(vma->vm_mm, address, pmdp);
1359 }
1360 
1361 #define __HAVE_ARCH_PMDP_INVALIDATE
1362 static inline void pmdp_invalidate(struct vm_area_struct *vma,
1363 				   unsigned long address, pmd_t *pmdp)
1364 {
1365 	__pmd_idte(address, pmdp);
1366 }
1367 
1368 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1369 {
1370 	pmd_t __pmd;
1371 	pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1372 	return __pmd;
1373 }
1374 
1375 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1376 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1377 
1378 static inline int pmd_trans_huge(pmd_t pmd)
1379 {
1380 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1381 }
1382 
1383 static inline int has_transparent_hugepage(void)
1384 {
1385 	return MACHINE_HAS_HPAGE ? 1 : 0;
1386 }
1387 
1388 static inline unsigned long pmd_pfn(pmd_t pmd)
1389 {
1390 	if (pmd_trans_huge(pmd))
1391 		return pmd_val(pmd) >> HPAGE_SHIFT;
1392 	else
1393 		return pmd_val(pmd) >> PAGE_SHIFT;
1394 }
1395 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1396 
1397 /*
1398  * 31 bit swap entry format:
1399  * A page-table entry has some bits we have to treat in a special way.
1400  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1401  * exception will occur instead of a page translation exception. The
1402  * specifiation exception has the bad habit not to store necessary
1403  * information in the lowcore.
1404  * Bit 21 and bit 22 are the page invalid bit and the page protection
1405  * bit. We set both to indicate a swapped page.
1406  * Bit 30 and 31 are used to distinguish the different page types. For
1407  * a swapped page these bits need to be zero.
1408  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1409  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1410  * plus 24 for the offset.
1411  * 0|     offset        |0110|o|type |00|
1412  * 0 0000000001111111111 2222 2 22222 33
1413  * 0 1234567890123456789 0123 4 56789 01
1414  *
1415  * 64 bit swap entry format:
1416  * A page-table entry has some bits we have to treat in a special way.
1417  * Bits 52 and bit 55 have to be zero, otherwise an specification
1418  * exception will occur instead of a page translation exception. The
1419  * specifiation exception has the bad habit not to store necessary
1420  * information in the lowcore.
1421  * Bit 53 and bit 54 are the page invalid bit and the page protection
1422  * bit. We set both to indicate a swapped page.
1423  * Bit 62 and 63 are used to distinguish the different page types. For
1424  * a swapped page these bits need to be zero.
1425  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1426  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1427  * plus 56 for the offset.
1428  * |                      offset                        |0110|o|type |00|
1429  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1430  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1431  */
1432 #ifndef CONFIG_64BIT
1433 #define __SWP_OFFSET_MASK (~0UL >> 12)
1434 #else
1435 #define __SWP_OFFSET_MASK (~0UL >> 11)
1436 #endif
1437 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1438 {
1439 	pte_t pte;
1440 	offset &= __SWP_OFFSET_MASK;
1441 	pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
1442 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1443 	return pte;
1444 }
1445 
1446 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1447 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1448 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1449 
1450 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1451 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1452 
1453 #ifndef CONFIG_64BIT
1454 # define PTE_FILE_MAX_BITS	26
1455 #else /* CONFIG_64BIT */
1456 # define PTE_FILE_MAX_BITS	59
1457 #endif /* CONFIG_64BIT */
1458 
1459 #define pte_to_pgoff(__pte) \
1460 	((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1461 
1462 #define pgoff_to_pte(__off) \
1463 	((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1464 		   | _PAGE_TYPE_FILE })
1465 
1466 #endif /* !__ASSEMBLY__ */
1467 
1468 #define kern_addr_valid(addr)   (1)
1469 
1470 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1471 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1472 extern int s390_enable_sie(void);
1473 
1474 /*
1475  * No page table caches to initialise
1476  */
1477 #define pgtable_cache_init()	do { } while (0)
1478 
1479 #include <asm-generic/pgtable.h>
1480 
1481 #endif /* _S390_PAGE_H */
1482