xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision 12eb4683)
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 <linux/page-flags.h>
33 #include <asm/bug.h>
34 #include <asm/page.h>
35 
36 extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
37 extern void paging_init(void);
38 extern void vmem_map_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 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
46 
47 /*
48  * ZERO_PAGE is a global shared page that is always zero; used
49  * for zero-mapped memory areas etc..
50  */
51 
52 extern unsigned long empty_zero_page;
53 extern unsigned long zero_page_mask;
54 
55 #define ZERO_PAGE(vaddr) \
56 	(virt_to_page((void *)(empty_zero_page + \
57 	 (((unsigned long)(vaddr)) &zero_page_mask))))
58 #define __HAVE_COLOR_ZERO_PAGE
59 
60 /* TODO: s390 cannot support io_remap_pfn_range... */
61 #endif /* !__ASSEMBLY__ */
62 
63 /*
64  * PMD_SHIFT determines the size of the area a second-level page
65  * table can map
66  * PGDIR_SHIFT determines what a third-level page table entry can map
67  */
68 #ifndef CONFIG_64BIT
69 # define PMD_SHIFT	20
70 # define PUD_SHIFT	20
71 # define PGDIR_SHIFT	20
72 #else /* CONFIG_64BIT */
73 # define PMD_SHIFT	20
74 # define PUD_SHIFT	31
75 # define PGDIR_SHIFT	42
76 #endif /* CONFIG_64BIT */
77 
78 #define PMD_SIZE        (1UL << PMD_SHIFT)
79 #define PMD_MASK        (~(PMD_SIZE-1))
80 #define PUD_SIZE	(1UL << PUD_SHIFT)
81 #define PUD_MASK	(~(PUD_SIZE-1))
82 #define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
83 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
84 
85 /*
86  * entries per page directory level: the S390 is two-level, so
87  * we don't really have any PMD directory physically.
88  * for S390 segment-table entries are combined to one PGD
89  * that leads to 1024 pte per pgd
90  */
91 #define PTRS_PER_PTE	256
92 #ifndef CONFIG_64BIT
93 #define PTRS_PER_PMD	1
94 #define PTRS_PER_PUD	1
95 #else /* CONFIG_64BIT */
96 #define PTRS_PER_PMD	2048
97 #define PTRS_PER_PUD	2048
98 #endif /* CONFIG_64BIT */
99 #define PTRS_PER_PGD	2048
100 
101 #define FIRST_USER_ADDRESS  0
102 
103 #define pte_ERROR(e) \
104 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
105 #define pmd_ERROR(e) \
106 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
107 #define pud_ERROR(e) \
108 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
109 #define pgd_ERROR(e) \
110 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
111 
112 #ifndef __ASSEMBLY__
113 /*
114  * The vmalloc and module area will always be on the topmost area of the kernel
115  * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc and modules.
116  * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
117  * modules will reside. That makes sure that inter module branches always
118  * happen without trampolines and in addition the placement within a 2GB frame
119  * is branch prediction unit friendly.
120  */
121 extern unsigned long VMALLOC_START;
122 extern unsigned long VMALLOC_END;
123 extern struct page *vmemmap;
124 
125 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
126 
127 #ifdef CONFIG_64BIT
128 extern unsigned long MODULES_VADDR;
129 extern unsigned long MODULES_END;
130 #define MODULES_VADDR	MODULES_VADDR
131 #define MODULES_END	MODULES_END
132 #define MODULES_LEN	(1UL << 31)
133 #endif
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_PROTECT	0x200		/* HW read-only bit  */
221 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
222 #define _PAGE_LARGE	0x800		/* Bit to mark a large pte */
223 
224 /* Software bits in the page table entry */
225 #define _PAGE_PRESENT	0x001		/* SW pte present bit */
226 #define _PAGE_TYPE	0x002		/* SW pte type bit */
227 #define _PAGE_YOUNG	0x004		/* SW pte young bit */
228 #define _PAGE_DIRTY	0x008		/* SW pte dirty bit */
229 #define _PAGE_READ	0x010		/* SW pte read bit */
230 #define _PAGE_WRITE	0x020		/* SW pte write bit */
231 #define _PAGE_SPECIAL	0x040		/* SW associated with special page */
232 #define __HAVE_ARCH_PTE_SPECIAL
233 
234 /* Set of bits not changed in pte_modify */
235 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_CO | \
236 				 _PAGE_DIRTY | _PAGE_YOUNG)
237 
238 /*
239  * handle_pte_fault uses pte_present, pte_none and pte_file to find out the
240  * pte type WITHOUT holding the page table lock. The _PAGE_PRESENT bit
241  * is used to distinguish present from not-present ptes. It is changed only
242  * with the page table lock held.
243  *
244  * The following table gives the different possible bit combinations for
245  * the pte hardware and software bits in the last 12 bits of a pte:
246  *
247  *				842100000000
248  *				000084210000
249  *				000000008421
250  *				.IR...wrdytp
251  * empty			.10...000000
252  * swap				.10...xxxx10
253  * file				.11...xxxxx0
254  * prot-none, clean, old	.11...000001
255  * prot-none, clean, young	.11...000101
256  * prot-none, dirty, old	.10...001001
257  * prot-none, dirty, young	.10...001101
258  * read-only, clean, old	.11...010001
259  * read-only, clean, young	.01...010101
260  * read-only, dirty, old	.11...011001
261  * read-only, dirty, young	.01...011101
262  * read-write, clean, old	.11...110001
263  * read-write, clean, young	.01...110101
264  * read-write, dirty, old	.10...111001
265  * read-write, dirty, young	.00...111101
266  *
267  * pte_present is true for the bit pattern .xx...xxxxx1, (pte & 0x001) == 0x001
268  * pte_none    is true for the bit pattern .10...xxxx00, (pte & 0x603) == 0x400
269  * pte_file    is true for the bit pattern .11...xxxxx0, (pte & 0x601) == 0x600
270  * pte_swap    is true for the bit pattern .10...xxxx10, (pte & 0x603) == 0x402
271  */
272 
273 #ifndef CONFIG_64BIT
274 
275 /* Bits in the segment table address-space-control-element */
276 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
277 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
278 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
279 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
280 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
281 
282 /* Bits in the segment table entry */
283 #define _SEGMENT_ENTRY_BITS	0x7fffffffUL	/* Valid segment table bits */
284 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
285 #define _SEGMENT_ENTRY_PROTECT	0x200	/* page protection bit		    */
286 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
287 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
288 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
289 #define _SEGMENT_ENTRY_NONE	_SEGMENT_ENTRY_PROTECT
290 
291 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
292 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
293 
294 /*
295  * Segment table entry encoding (I = invalid, R = read-only bit):
296  *		..R...I.....
297  * prot-none	..1...1.....
298  * read-only	..1...0.....
299  * read-write	..0...0.....
300  * empty	..0...1.....
301  */
302 
303 /* Page status table bits for virtualization */
304 #define PGSTE_ACC_BITS	0xf0000000UL
305 #define PGSTE_FP_BIT	0x08000000UL
306 #define PGSTE_PCL_BIT	0x00800000UL
307 #define PGSTE_HR_BIT	0x00400000UL
308 #define PGSTE_HC_BIT	0x00200000UL
309 #define PGSTE_GR_BIT	0x00040000UL
310 #define PGSTE_GC_BIT	0x00020000UL
311 #define PGSTE_IN_BIT	0x00008000UL	/* IPTE notify bit */
312 
313 #else /* CONFIG_64BIT */
314 
315 /* Bits in the segment/region table address-space-control-element */
316 #define _ASCE_ORIGIN		~0xfffUL/* segment table origin		    */
317 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
318 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
319 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
320 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
321 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
322 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
323 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
324 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
325 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
326 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
327 
328 /* Bits in the region table entry */
329 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
330 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
331 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
332 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
333 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
334 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
335 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
336 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
337 
338 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
339 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
340 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
341 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
342 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
343 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
344 
345 #define _REGION3_ENTRY_LARGE	0x400	/* RTTE-format control, large page  */
346 #define _REGION3_ENTRY_RO	0x200	/* page protection bit		    */
347 #define _REGION3_ENTRY_CO	0x100	/* change-recording override	    */
348 
349 /* Bits in the segment table entry */
350 #define _SEGMENT_ENTRY_BITS	0xfffffffffffffe33UL
351 #define _SEGMENT_ENTRY_BITS_LARGE 0xfffffffffff1ff33UL
352 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
353 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* segment table origin		    */
354 #define _SEGMENT_ENTRY_PROTECT	0x200	/* page protection bit		    */
355 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
356 
357 #define _SEGMENT_ENTRY		(0)
358 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
359 
360 #define _SEGMENT_ENTRY_LARGE	0x400	/* STE-format control, large page   */
361 #define _SEGMENT_ENTRY_CO	0x100	/* change-recording override   */
362 #define _SEGMENT_ENTRY_SPLIT	0x001	/* THP splitting bit */
363 #define _SEGMENT_ENTRY_YOUNG	0x002	/* SW segment young bit */
364 #define _SEGMENT_ENTRY_NONE	_SEGMENT_ENTRY_YOUNG
365 
366 /*
367  * Segment table entry encoding (R = read-only, I = invalid, y = young bit):
368  *			..R...I...y.
369  * prot-none, old	..0...1...1.
370  * prot-none, young	..1...1...1.
371  * read-only, old	..1...1...0.
372  * read-only, young	..1...0...1.
373  * read-write, old	..0...1...0.
374  * read-write, young	..0...0...1.
375  * The segment table origin is used to distinguish empty (origin==0) from
376  * read-write, old segment table entries (origin!=0)
377  */
378 
379 #define _SEGMENT_ENTRY_SPLIT_BIT 0	/* THP splitting bit number */
380 
381 /* Set of bits not changed in pmd_modify */
382 #define _SEGMENT_CHG_MASK	(_SEGMENT_ENTRY_ORIGIN | _SEGMENT_ENTRY_LARGE \
383 				 | _SEGMENT_ENTRY_SPLIT | _SEGMENT_ENTRY_CO)
384 
385 /* Page status table bits for virtualization */
386 #define PGSTE_ACC_BITS	0xf000000000000000UL
387 #define PGSTE_FP_BIT	0x0800000000000000UL
388 #define PGSTE_PCL_BIT	0x0080000000000000UL
389 #define PGSTE_HR_BIT	0x0040000000000000UL
390 #define PGSTE_HC_BIT	0x0020000000000000UL
391 #define PGSTE_GR_BIT	0x0004000000000000UL
392 #define PGSTE_GC_BIT	0x0002000000000000UL
393 #define PGSTE_IN_BIT	0x0000800000000000UL	/* IPTE notify bit */
394 
395 #endif /* CONFIG_64BIT */
396 
397 /*
398  * A user page table pointer has the space-switch-event bit, the
399  * private-space-control bit and the storage-alteration-event-control
400  * bit set. A kernel page table pointer doesn't need them.
401  */
402 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
403 				 _ASCE_ALT_EVENT)
404 
405 /*
406  * Page protection definitions.
407  */
408 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID)
409 #define PAGE_READ	__pgprot(_PAGE_PRESENT | _PAGE_READ | \
410 				 _PAGE_INVALID | _PAGE_PROTECT)
411 #define PAGE_WRITE	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
412 				 _PAGE_INVALID | _PAGE_PROTECT)
413 
414 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
415 				 _PAGE_YOUNG | _PAGE_DIRTY)
416 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
417 				 _PAGE_YOUNG | _PAGE_DIRTY)
418 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
419 				 _PAGE_PROTECT)
420 
421 /*
422  * On s390 the page table entry has an invalid bit and a read-only bit.
423  * Read permission implies execute permission and write permission
424  * implies read permission.
425  */
426          /*xwr*/
427 #define __P000	PAGE_NONE
428 #define __P001	PAGE_READ
429 #define __P010	PAGE_READ
430 #define __P011	PAGE_READ
431 #define __P100	PAGE_READ
432 #define __P101	PAGE_READ
433 #define __P110	PAGE_READ
434 #define __P111	PAGE_READ
435 
436 #define __S000	PAGE_NONE
437 #define __S001	PAGE_READ
438 #define __S010	PAGE_WRITE
439 #define __S011	PAGE_WRITE
440 #define __S100	PAGE_READ
441 #define __S101	PAGE_READ
442 #define __S110	PAGE_WRITE
443 #define __S111	PAGE_WRITE
444 
445 /*
446  * Segment entry (large page) protection definitions.
447  */
448 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_INVALID | \
449 				 _SEGMENT_ENTRY_NONE)
450 #define SEGMENT_READ	__pgprot(_SEGMENT_ENTRY_INVALID | \
451 				 _SEGMENT_ENTRY_PROTECT)
452 #define SEGMENT_WRITE	__pgprot(_SEGMENT_ENTRY_INVALID)
453 
454 static inline int mm_has_pgste(struct mm_struct *mm)
455 {
456 #ifdef CONFIG_PGSTE
457 	if (unlikely(mm->context.has_pgste))
458 		return 1;
459 #endif
460 	return 0;
461 }
462 /*
463  * pgd/pmd/pte query functions
464  */
465 #ifndef CONFIG_64BIT
466 
467 static inline int pgd_present(pgd_t pgd) { return 1; }
468 static inline int pgd_none(pgd_t pgd)    { return 0; }
469 static inline int pgd_bad(pgd_t pgd)     { return 0; }
470 
471 static inline int pud_present(pud_t pud) { return 1; }
472 static inline int pud_none(pud_t pud)	 { return 0; }
473 static inline int pud_large(pud_t pud)	 { return 0; }
474 static inline int pud_bad(pud_t pud)	 { return 0; }
475 
476 #else /* CONFIG_64BIT */
477 
478 static inline int pgd_present(pgd_t pgd)
479 {
480 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
481 		return 1;
482 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
483 }
484 
485 static inline int pgd_none(pgd_t pgd)
486 {
487 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
488 		return 0;
489 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
490 }
491 
492 static inline int pgd_bad(pgd_t pgd)
493 {
494 	/*
495 	 * With dynamic page table levels the pgd 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_INVALID &
501 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
502 	return (pgd_val(pgd) & mask) != 0;
503 }
504 
505 static inline int pud_present(pud_t pud)
506 {
507 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
508 		return 1;
509 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
510 }
511 
512 static inline int pud_none(pud_t pud)
513 {
514 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
515 		return 0;
516 	return (pud_val(pud) & _REGION_ENTRY_INVALID) != 0UL;
517 }
518 
519 static inline int pud_large(pud_t pud)
520 {
521 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
522 		return 0;
523 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
524 }
525 
526 static inline int pud_bad(pud_t pud)
527 {
528 	/*
529 	 * With dynamic page table levels the pud can be a region table
530 	 * entry or a segment table entry. Check for the bit that are
531 	 * invalid for either table entry.
532 	 */
533 	unsigned long mask =
534 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
535 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
536 	return (pud_val(pud) & mask) != 0;
537 }
538 
539 #endif /* CONFIG_64BIT */
540 
541 static inline int pmd_present(pmd_t pmd)
542 {
543 	return pmd_val(pmd) != _SEGMENT_ENTRY_INVALID;
544 }
545 
546 static inline int pmd_none(pmd_t pmd)
547 {
548 	return pmd_val(pmd) == _SEGMENT_ENTRY_INVALID;
549 }
550 
551 static inline int pmd_large(pmd_t pmd)
552 {
553 #ifdef CONFIG_64BIT
554 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
555 #else
556 	return 0;
557 #endif
558 }
559 
560 static inline int pmd_prot_none(pmd_t pmd)
561 {
562 	return (pmd_val(pmd) & _SEGMENT_ENTRY_INVALID) &&
563 		(pmd_val(pmd) & _SEGMENT_ENTRY_NONE);
564 }
565 
566 static inline int pmd_bad(pmd_t pmd)
567 {
568 #ifdef CONFIG_64BIT
569 	if (pmd_large(pmd))
570 		return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
571 #endif
572 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
573 }
574 
575 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
576 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
577 				 unsigned long addr, pmd_t *pmdp);
578 
579 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
580 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
581 				 unsigned long address, pmd_t *pmdp,
582 				 pmd_t entry, int dirty);
583 
584 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
585 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
586 				  unsigned long address, pmd_t *pmdp);
587 
588 #define __HAVE_ARCH_PMD_WRITE
589 static inline int pmd_write(pmd_t pmd)
590 {
591 	if (pmd_prot_none(pmd))
592 		return 0;
593 	return (pmd_val(pmd) & _SEGMENT_ENTRY_PROTECT) == 0;
594 }
595 
596 static inline int pmd_young(pmd_t pmd)
597 {
598 	int young = 0;
599 #ifdef CONFIG_64BIT
600 	if (pmd_prot_none(pmd))
601 		young = (pmd_val(pmd) & _SEGMENT_ENTRY_PROTECT) != 0;
602 	else
603 		young = (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
604 #endif
605 	return young;
606 }
607 
608 static inline int pte_present(pte_t pte)
609 {
610 	/* Bit pattern: (pte & 0x001) == 0x001 */
611 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
612 }
613 
614 static inline int pte_none(pte_t pte)
615 {
616 	/* Bit pattern: pte == 0x400 */
617 	return pte_val(pte) == _PAGE_INVALID;
618 }
619 
620 static inline int pte_file(pte_t pte)
621 {
622 	/* Bit pattern: (pte & 0x601) == 0x600 */
623 	return (pte_val(pte) & (_PAGE_INVALID | _PAGE_PROTECT | _PAGE_PRESENT))
624 		== (_PAGE_INVALID | _PAGE_PROTECT);
625 }
626 
627 static inline int pte_special(pte_t pte)
628 {
629 	return (pte_val(pte) & _PAGE_SPECIAL);
630 }
631 
632 #define __HAVE_ARCH_PTE_SAME
633 static inline int pte_same(pte_t a, pte_t b)
634 {
635 	return pte_val(a) == pte_val(b);
636 }
637 
638 static inline pgste_t pgste_get_lock(pte_t *ptep)
639 {
640 	unsigned long new = 0;
641 #ifdef CONFIG_PGSTE
642 	unsigned long old;
643 
644 	preempt_disable();
645 	asm(
646 		"	lg	%0,%2\n"
647 		"0:	lgr	%1,%0\n"
648 		"	nihh	%0,0xff7f\n"	/* clear PCL bit in old */
649 		"	oihh	%1,0x0080\n"	/* set PCL bit in new */
650 		"	csg	%0,%1,%2\n"
651 		"	jl	0b\n"
652 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
653 		: "Q" (ptep[PTRS_PER_PTE]) : "cc", "memory");
654 #endif
655 	return __pgste(new);
656 }
657 
658 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
659 {
660 #ifdef CONFIG_PGSTE
661 	asm(
662 		"	nihh	%1,0xff7f\n"	/* clear PCL bit */
663 		"	stg	%1,%0\n"
664 		: "=Q" (ptep[PTRS_PER_PTE])
665 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE])
666 		: "cc", "memory");
667 	preempt_enable();
668 #endif
669 }
670 
671 static inline pgste_t pgste_get(pte_t *ptep)
672 {
673 	unsigned long pgste = 0;
674 #ifdef CONFIG_PGSTE
675 	pgste = *(unsigned long *)(ptep + PTRS_PER_PTE);
676 #endif
677 	return __pgste(pgste);
678 }
679 
680 static inline void pgste_set(pte_t *ptep, pgste_t pgste)
681 {
682 #ifdef CONFIG_PGSTE
683 	*(pgste_t *)(ptep + PTRS_PER_PTE) = pgste;
684 #endif
685 }
686 
687 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
688 {
689 #ifdef CONFIG_PGSTE
690 	unsigned long address, bits, skey;
691 
692 	if (pte_val(*ptep) & _PAGE_INVALID)
693 		return pgste;
694 	address = pte_val(*ptep) & PAGE_MASK;
695 	skey = (unsigned long) page_get_storage_key(address);
696 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
697 	if (!(pgste_val(pgste) & PGSTE_HC_BIT) && (bits & _PAGE_CHANGED)) {
698 		/* Transfer dirty + referenced bit to host bits in pgste */
699 		pgste_val(pgste) |= bits << 52;
700 		page_set_storage_key(address, skey ^ bits, 0);
701 	} else if (!(pgste_val(pgste) & PGSTE_HR_BIT) &&
702 		   (bits & _PAGE_REFERENCED)) {
703 		/* Transfer referenced bit to host bit in pgste */
704 		pgste_val(pgste) |= PGSTE_HR_BIT;
705 		page_reset_referenced(address);
706 	}
707 	/* Transfer page changed & referenced bit to guest bits in pgste */
708 	pgste_val(pgste) |= bits << 48;		/* GR bit & GC bit */
709 	/* Copy page access key and fetch protection bit to pgste */
710 	pgste_val(pgste) &= ~(PGSTE_ACC_BITS | PGSTE_FP_BIT);
711 	pgste_val(pgste) |= (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
712 #endif
713 	return pgste;
714 
715 }
716 
717 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
718 {
719 #ifdef CONFIG_PGSTE
720 	if (pte_val(*ptep) & _PAGE_INVALID)
721 		return pgste;
722 	/* Get referenced bit from storage key */
723 	if (page_reset_referenced(pte_val(*ptep) & PAGE_MASK))
724 		pgste_val(pgste) |= PGSTE_HR_BIT | PGSTE_GR_BIT;
725 #endif
726 	return pgste;
727 }
728 
729 static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry)
730 {
731 #ifdef CONFIG_PGSTE
732 	unsigned long address;
733 	unsigned long nkey;
734 
735 	if (pte_val(entry) & _PAGE_INVALID)
736 		return;
737 	VM_BUG_ON(!(pte_val(*ptep) & _PAGE_INVALID));
738 	address = pte_val(entry) & PAGE_MASK;
739 	/*
740 	 * Set page access key and fetch protection bit from pgste.
741 	 * The guest C/R information is still in the PGSTE, set real
742 	 * key C/R to 0.
743 	 */
744 	nkey = (pgste_val(pgste) & (PGSTE_ACC_BITS | PGSTE_FP_BIT)) >> 56;
745 	page_set_storage_key(address, nkey, 0);
746 #endif
747 }
748 
749 static inline void pgste_set_pte(pte_t *ptep, pte_t entry)
750 {
751 	if (!MACHINE_HAS_ESOP &&
752 	    (pte_val(entry) & _PAGE_PRESENT) &&
753 	    (pte_val(entry) & _PAGE_WRITE)) {
754 		/*
755 		 * Without enhanced suppression-on-protection force
756 		 * the dirty bit on for all writable ptes.
757 		 */
758 		pte_val(entry) |= _PAGE_DIRTY;
759 		pte_val(entry) &= ~_PAGE_PROTECT;
760 	}
761 	*ptep = entry;
762 }
763 
764 /**
765  * struct gmap_struct - guest address space
766  * @mm: pointer to the parent mm_struct
767  * @table: pointer to the page directory
768  * @asce: address space control element for gmap page table
769  * @crst_list: list of all crst tables used in the guest address space
770  */
771 struct gmap {
772 	struct list_head list;
773 	struct mm_struct *mm;
774 	unsigned long *table;
775 	unsigned long asce;
776 	void *private;
777 	struct list_head crst_list;
778 };
779 
780 /**
781  * struct gmap_rmap - reverse mapping for segment table entries
782  * @gmap: pointer to the gmap_struct
783  * @entry: pointer to a segment table entry
784  * @vmaddr: virtual address in the guest address space
785  */
786 struct gmap_rmap {
787 	struct list_head list;
788 	struct gmap *gmap;
789 	unsigned long *entry;
790 	unsigned long vmaddr;
791 };
792 
793 /**
794  * struct gmap_pgtable - gmap information attached to a page table
795  * @vmaddr: address of the 1MB segment in the process virtual memory
796  * @mapper: list of segment table entries mapping a page table
797  */
798 struct gmap_pgtable {
799 	unsigned long vmaddr;
800 	struct list_head mapper;
801 };
802 
803 /**
804  * struct gmap_notifier - notify function block for page invalidation
805  * @notifier_call: address of callback function
806  */
807 struct gmap_notifier {
808 	struct list_head list;
809 	void (*notifier_call)(struct gmap *gmap, unsigned long address);
810 };
811 
812 struct gmap *gmap_alloc(struct mm_struct *mm);
813 void gmap_free(struct gmap *gmap);
814 void gmap_enable(struct gmap *gmap);
815 void gmap_disable(struct gmap *gmap);
816 int gmap_map_segment(struct gmap *gmap, unsigned long from,
817 		     unsigned long to, unsigned long len);
818 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
819 unsigned long __gmap_translate(unsigned long address, struct gmap *);
820 unsigned long gmap_translate(unsigned long address, struct gmap *);
821 unsigned long __gmap_fault(unsigned long address, struct gmap *);
822 unsigned long gmap_fault(unsigned long address, struct gmap *);
823 void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
824 
825 void gmap_register_ipte_notifier(struct gmap_notifier *);
826 void gmap_unregister_ipte_notifier(struct gmap_notifier *);
827 int gmap_ipte_notify(struct gmap *, unsigned long start, unsigned long len);
828 void gmap_do_ipte_notify(struct mm_struct *, unsigned long addr, pte_t *);
829 
830 static inline pgste_t pgste_ipte_notify(struct mm_struct *mm,
831 					unsigned long addr,
832 					pte_t *ptep, pgste_t pgste)
833 {
834 #ifdef CONFIG_PGSTE
835 	if (pgste_val(pgste) & PGSTE_IN_BIT) {
836 		pgste_val(pgste) &= ~PGSTE_IN_BIT;
837 		gmap_do_ipte_notify(mm, addr, ptep);
838 	}
839 #endif
840 	return pgste;
841 }
842 
843 /*
844  * Certain architectures need to do special things when PTEs
845  * within a page table are directly modified.  Thus, the following
846  * hook is made available.
847  */
848 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
849 			      pte_t *ptep, pte_t entry)
850 {
851 	pgste_t pgste;
852 
853 	if (mm_has_pgste(mm)) {
854 		pgste = pgste_get_lock(ptep);
855 		pgste_set_key(ptep, pgste, entry);
856 		pgste_set_pte(ptep, entry);
857 		pgste_set_unlock(ptep, pgste);
858 	} else {
859 		if (!(pte_val(entry) & _PAGE_INVALID) && MACHINE_HAS_EDAT1)
860 			pte_val(entry) |= _PAGE_CO;
861 		*ptep = entry;
862 	}
863 }
864 
865 /*
866  * query functions pte_write/pte_dirty/pte_young only work if
867  * pte_present() is true. Undefined behaviour if not..
868  */
869 static inline int pte_write(pte_t pte)
870 {
871 	return (pte_val(pte) & _PAGE_WRITE) != 0;
872 }
873 
874 static inline int pte_dirty(pte_t pte)
875 {
876 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
877 }
878 
879 static inline int pte_young(pte_t pte)
880 {
881 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
882 }
883 
884 /*
885  * pgd/pmd/pte modification functions
886  */
887 
888 static inline void pgd_clear(pgd_t *pgd)
889 {
890 #ifdef CONFIG_64BIT
891 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
892 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
893 #endif
894 }
895 
896 static inline void pud_clear(pud_t *pud)
897 {
898 #ifdef CONFIG_64BIT
899 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
900 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
901 #endif
902 }
903 
904 static inline void pmd_clear(pmd_t *pmdp)
905 {
906 	pmd_val(*pmdp) = _SEGMENT_ENTRY_INVALID;
907 }
908 
909 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
910 {
911 	pte_val(*ptep) = _PAGE_INVALID;
912 }
913 
914 /*
915  * The following pte modification functions only work if
916  * pte_present() is true. Undefined behaviour if not..
917  */
918 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
919 {
920 	pte_val(pte) &= _PAGE_CHG_MASK;
921 	pte_val(pte) |= pgprot_val(newprot);
922 	/*
923 	 * newprot for PAGE_NONE, PAGE_READ and PAGE_WRITE has the
924 	 * invalid bit set, clear it again for readable, young pages
925 	 */
926 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
927 		pte_val(pte) &= ~_PAGE_INVALID;
928 	/*
929 	 * newprot for PAGE_READ and PAGE_WRITE has the page protection
930 	 * bit set, clear it again for writable, dirty pages
931 	 */
932 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
933 		pte_val(pte) &= ~_PAGE_PROTECT;
934 	return pte;
935 }
936 
937 static inline pte_t pte_wrprotect(pte_t pte)
938 {
939 	pte_val(pte) &= ~_PAGE_WRITE;
940 	pte_val(pte) |= _PAGE_PROTECT;
941 	return pte;
942 }
943 
944 static inline pte_t pte_mkwrite(pte_t pte)
945 {
946 	pte_val(pte) |= _PAGE_WRITE;
947 	if (pte_val(pte) & _PAGE_DIRTY)
948 		pte_val(pte) &= ~_PAGE_PROTECT;
949 	return pte;
950 }
951 
952 static inline pte_t pte_mkclean(pte_t pte)
953 {
954 	pte_val(pte) &= ~_PAGE_DIRTY;
955 	pte_val(pte) |= _PAGE_PROTECT;
956 	return pte;
957 }
958 
959 static inline pte_t pte_mkdirty(pte_t pte)
960 {
961 	pte_val(pte) |= _PAGE_DIRTY;
962 	if (pte_val(pte) & _PAGE_WRITE)
963 		pte_val(pte) &= ~_PAGE_PROTECT;
964 	return pte;
965 }
966 
967 static inline pte_t pte_mkold(pte_t pte)
968 {
969 	pte_val(pte) &= ~_PAGE_YOUNG;
970 	pte_val(pte) |= _PAGE_INVALID;
971 	return pte;
972 }
973 
974 static inline pte_t pte_mkyoung(pte_t pte)
975 {
976 	pte_val(pte) |= _PAGE_YOUNG;
977 	if (pte_val(pte) & _PAGE_READ)
978 		pte_val(pte) &= ~_PAGE_INVALID;
979 	return pte;
980 }
981 
982 static inline pte_t pte_mkspecial(pte_t pte)
983 {
984 	pte_val(pte) |= _PAGE_SPECIAL;
985 	return pte;
986 }
987 
988 #ifdef CONFIG_HUGETLB_PAGE
989 static inline pte_t pte_mkhuge(pte_t pte)
990 {
991 	pte_val(pte) |= _PAGE_LARGE;
992 	return pte;
993 }
994 #endif
995 
996 /*
997  * Get (and clear) the user dirty bit for a pte.
998  */
999 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
1000 						 pte_t *ptep)
1001 {
1002 	pgste_t pgste;
1003 	int dirty = 0;
1004 
1005 	if (mm_has_pgste(mm)) {
1006 		pgste = pgste_get_lock(ptep);
1007 		pgste = pgste_update_all(ptep, pgste);
1008 		dirty = !!(pgste_val(pgste) & PGSTE_HC_BIT);
1009 		pgste_val(pgste) &= ~PGSTE_HC_BIT;
1010 		pgste_set_unlock(ptep, pgste);
1011 		return dirty;
1012 	}
1013 	return dirty;
1014 }
1015 
1016 /*
1017  * Get (and clear) the user referenced bit for a pte.
1018  */
1019 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
1020 						 pte_t *ptep)
1021 {
1022 	pgste_t pgste;
1023 	int young = 0;
1024 
1025 	if (mm_has_pgste(mm)) {
1026 		pgste = pgste_get_lock(ptep);
1027 		pgste = pgste_update_young(ptep, pgste);
1028 		young = !!(pgste_val(pgste) & PGSTE_HR_BIT);
1029 		pgste_val(pgste) &= ~PGSTE_HR_BIT;
1030 		pgste_set_unlock(ptep, pgste);
1031 	}
1032 	return young;
1033 }
1034 
1035 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
1036 {
1037 	if (!(pte_val(*ptep) & _PAGE_INVALID)) {
1038 #ifndef CONFIG_64BIT
1039 		/* pto must point to the start of the segment table */
1040 		pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
1041 #else
1042 		/* ipte in zarch mode can do the math */
1043 		pte_t *pto = ptep;
1044 #endif
1045 		asm volatile(
1046 			"	ipte	%2,%3"
1047 			: "=m" (*ptep) : "m" (*ptep),
1048 			  "a" (pto), "a" (address));
1049 	}
1050 }
1051 
1052 static inline void ptep_flush_lazy(struct mm_struct *mm,
1053 				   unsigned long address, pte_t *ptep)
1054 {
1055 	int active = (mm == current->active_mm) ? 1 : 0;
1056 
1057 	if (atomic_read(&mm->context.attach_count) > active)
1058 		__ptep_ipte(address, ptep);
1059 	else
1060 		mm->context.flush_mm = 1;
1061 }
1062 
1063 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1064 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1065 					    unsigned long addr, pte_t *ptep)
1066 {
1067 	pgste_t pgste;
1068 	pte_t pte;
1069 	int young;
1070 
1071 	if (mm_has_pgste(vma->vm_mm)) {
1072 		pgste = pgste_get_lock(ptep);
1073 		pgste = pgste_ipte_notify(vma->vm_mm, addr, ptep, pgste);
1074 	}
1075 
1076 	pte = *ptep;
1077 	__ptep_ipte(addr, ptep);
1078 	young = pte_young(pte);
1079 	pte = pte_mkold(pte);
1080 
1081 	if (mm_has_pgste(vma->vm_mm)) {
1082 		pgste_set_pte(ptep, pte);
1083 		pgste_set_unlock(ptep, pgste);
1084 	} else
1085 		*ptep = pte;
1086 
1087 	return young;
1088 }
1089 
1090 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1091 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1092 					 unsigned long address, pte_t *ptep)
1093 {
1094 	return ptep_test_and_clear_young(vma, address, ptep);
1095 }
1096 
1097 /*
1098  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1099  * both clear the TLB for the unmapped pte. The reason is that
1100  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1101  * to modify an active pte. The sequence is
1102  *   1) ptep_get_and_clear
1103  *   2) set_pte_at
1104  *   3) flush_tlb_range
1105  * On s390 the tlb needs to get flushed with the modification of the pte
1106  * if the pte is active. The only way how this can be implemented is to
1107  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1108  * is a nop.
1109  */
1110 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1111 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1112 				       unsigned long address, pte_t *ptep)
1113 {
1114 	pgste_t pgste;
1115 	pte_t pte;
1116 
1117 	if (mm_has_pgste(mm)) {
1118 		pgste = pgste_get_lock(ptep);
1119 		pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1120 	}
1121 
1122 	pte = *ptep;
1123 	ptep_flush_lazy(mm, address, ptep);
1124 	pte_val(*ptep) = _PAGE_INVALID;
1125 
1126 	if (mm_has_pgste(mm)) {
1127 		pgste = pgste_update_all(&pte, pgste);
1128 		pgste_set_unlock(ptep, pgste);
1129 	}
1130 	return pte;
1131 }
1132 
1133 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1134 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
1135 					   unsigned long address,
1136 					   pte_t *ptep)
1137 {
1138 	pgste_t pgste;
1139 	pte_t pte;
1140 
1141 	if (mm_has_pgste(mm)) {
1142 		pgste = pgste_get_lock(ptep);
1143 		pgste_ipte_notify(mm, address, ptep, pgste);
1144 	}
1145 
1146 	pte = *ptep;
1147 	ptep_flush_lazy(mm, address, ptep);
1148 	pte_val(*ptep) |= _PAGE_INVALID;
1149 
1150 	if (mm_has_pgste(mm)) {
1151 		pgste = pgste_update_all(&pte, pgste);
1152 		pgste_set(ptep, pgste);
1153 	}
1154 	return pte;
1155 }
1156 
1157 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
1158 					   unsigned long address,
1159 					   pte_t *ptep, pte_t pte)
1160 {
1161 	pgste_t pgste;
1162 
1163 	if (mm_has_pgste(mm)) {
1164 		pgste = pgste_get(ptep);
1165 		pgste_set_key(ptep, pgste, pte);
1166 		pgste_set_pte(ptep, pte);
1167 		pgste_set_unlock(ptep, pgste);
1168 	} else
1169 		*ptep = pte;
1170 }
1171 
1172 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1173 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1174 				     unsigned long address, pte_t *ptep)
1175 {
1176 	pgste_t pgste;
1177 	pte_t pte;
1178 
1179 	if (mm_has_pgste(vma->vm_mm)) {
1180 		pgste = pgste_get_lock(ptep);
1181 		pgste = pgste_ipte_notify(vma->vm_mm, address, ptep, pgste);
1182 	}
1183 
1184 	pte = *ptep;
1185 	__ptep_ipte(address, ptep);
1186 	pte_val(*ptep) = _PAGE_INVALID;
1187 
1188 	if (mm_has_pgste(vma->vm_mm)) {
1189 		pgste = pgste_update_all(&pte, pgste);
1190 		pgste_set_unlock(ptep, pgste);
1191 	}
1192 	return pte;
1193 }
1194 
1195 /*
1196  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1197  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1198  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1199  * cannot be accessed while the batched unmap is running. In this case
1200  * full==1 and a simple pte_clear is enough. See tlb.h.
1201  */
1202 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1203 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1204 					    unsigned long address,
1205 					    pte_t *ptep, int full)
1206 {
1207 	pgste_t pgste;
1208 	pte_t pte;
1209 
1210 	if (!full && mm_has_pgste(mm)) {
1211 		pgste = pgste_get_lock(ptep);
1212 		pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1213 	}
1214 
1215 	pte = *ptep;
1216 	if (!full)
1217 		ptep_flush_lazy(mm, address, ptep);
1218 	pte_val(*ptep) = _PAGE_INVALID;
1219 
1220 	if (!full && mm_has_pgste(mm)) {
1221 		pgste = pgste_update_all(&pte, pgste);
1222 		pgste_set_unlock(ptep, pgste);
1223 	}
1224 	return pte;
1225 }
1226 
1227 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1228 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1229 				       unsigned long address, pte_t *ptep)
1230 {
1231 	pgste_t pgste;
1232 	pte_t pte = *ptep;
1233 
1234 	if (pte_write(pte)) {
1235 		if (mm_has_pgste(mm)) {
1236 			pgste = pgste_get_lock(ptep);
1237 			pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1238 		}
1239 
1240 		ptep_flush_lazy(mm, address, ptep);
1241 		pte = pte_wrprotect(pte);
1242 
1243 		if (mm_has_pgste(mm)) {
1244 			pgste_set_pte(ptep, pte);
1245 			pgste_set_unlock(ptep, pgste);
1246 		} else
1247 			*ptep = pte;
1248 	}
1249 	return pte;
1250 }
1251 
1252 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1253 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1254 					unsigned long address, pte_t *ptep,
1255 					pte_t entry, int dirty)
1256 {
1257 	pgste_t pgste;
1258 
1259 	if (pte_same(*ptep, entry))
1260 		return 0;
1261 	if (mm_has_pgste(vma->vm_mm)) {
1262 		pgste = pgste_get_lock(ptep);
1263 		pgste = pgste_ipte_notify(vma->vm_mm, address, ptep, pgste);
1264 	}
1265 
1266 	__ptep_ipte(address, ptep);
1267 
1268 	if (mm_has_pgste(vma->vm_mm)) {
1269 		pgste_set_pte(ptep, entry);
1270 		pgste_set_unlock(ptep, pgste);
1271 	} else
1272 		*ptep = entry;
1273 	return 1;
1274 }
1275 
1276 /*
1277  * Conversion functions: convert a page and protection to a page entry,
1278  * and a page entry and page directory to the page they refer to.
1279  */
1280 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1281 {
1282 	pte_t __pte;
1283 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1284 	return pte_mkyoung(__pte);
1285 }
1286 
1287 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1288 {
1289 	unsigned long physpage = page_to_phys(page);
1290 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1291 
1292 	if (pte_write(__pte) && PageDirty(page))
1293 		__pte = pte_mkdirty(__pte);
1294 	return __pte;
1295 }
1296 
1297 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1298 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1299 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1300 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1301 
1302 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1303 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1304 
1305 #ifndef CONFIG_64BIT
1306 
1307 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1308 #define pud_deref(pmd) ({ BUG(); 0UL; })
1309 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1310 
1311 #define pud_offset(pgd, address) ((pud_t *) pgd)
1312 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1313 
1314 #else /* CONFIG_64BIT */
1315 
1316 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1317 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1318 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1319 
1320 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1321 {
1322 	pud_t *pud = (pud_t *) pgd;
1323 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1324 		pud = (pud_t *) pgd_deref(*pgd);
1325 	return pud  + pud_index(address);
1326 }
1327 
1328 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1329 {
1330 	pmd_t *pmd = (pmd_t *) pud;
1331 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1332 		pmd = (pmd_t *) pud_deref(*pud);
1333 	return pmd + pmd_index(address);
1334 }
1335 
1336 #endif /* CONFIG_64BIT */
1337 
1338 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1339 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1340 #define pte_page(x) pfn_to_page(pte_pfn(x))
1341 
1342 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1343 
1344 /* Find an entry in the lowest level page table.. */
1345 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1346 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1347 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1348 #define pte_unmap(pte) do { } while (0)
1349 
1350 static inline void __pmd_idte(unsigned long address, pmd_t *pmdp)
1351 {
1352 	unsigned long sto = (unsigned long) pmdp -
1353 			    pmd_index(address) * sizeof(pmd_t);
1354 
1355 	if (!(pmd_val(*pmdp) & _SEGMENT_ENTRY_INVALID)) {
1356 		asm volatile(
1357 			"	.insn	rrf,0xb98e0000,%2,%3,0,0"
1358 			: "=m" (*pmdp)
1359 			: "m" (*pmdp), "a" (sto),
1360 			  "a" ((address & HPAGE_MASK))
1361 			: "cc"
1362 		);
1363 	}
1364 }
1365 
1366 static inline void __pmd_csp(pmd_t *pmdp)
1367 {
1368 	register unsigned long reg2 asm("2") = pmd_val(*pmdp);
1369 	register unsigned long reg3 asm("3") = pmd_val(*pmdp) |
1370 					       _SEGMENT_ENTRY_INVALID;
1371 	register unsigned long reg4 asm("4") = ((unsigned long) pmdp) + 5;
1372 
1373 	asm volatile(
1374 		"	csp %1,%3"
1375 		: "=m" (*pmdp)
1376 		: "d" (reg2), "d" (reg3), "d" (reg4), "m" (*pmdp) : "cc");
1377 }
1378 
1379 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1380 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1381 {
1382 	/*
1383 	 * pgprot is PAGE_NONE, PAGE_READ, or PAGE_WRITE (see __Pxxx / __Sxxx)
1384 	 * Convert to segment table entry format.
1385 	 */
1386 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1387 		return pgprot_val(SEGMENT_NONE);
1388 	if (pgprot_val(pgprot) == pgprot_val(PAGE_READ))
1389 		return pgprot_val(SEGMENT_READ);
1390 	return pgprot_val(SEGMENT_WRITE);
1391 }
1392 
1393 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1394 {
1395 #ifdef CONFIG_64BIT
1396 	if (pmd_prot_none(pmd)) {
1397 		pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1398 	} else {
1399 		pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1400 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1401 	}
1402 #endif
1403 	return pmd;
1404 }
1405 
1406 static inline pmd_t pmd_mkold(pmd_t pmd)
1407 {
1408 #ifdef CONFIG_64BIT
1409 	if (pmd_prot_none(pmd)) {
1410 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1411 	} else {
1412 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1413 		pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1414 	}
1415 #endif
1416 	return pmd;
1417 }
1418 
1419 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1420 {
1421 	int young;
1422 
1423 	young = pmd_young(pmd);
1424 	pmd_val(pmd) &= _SEGMENT_CHG_MASK;
1425 	pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1426 	if (young)
1427 		pmd = pmd_mkyoung(pmd);
1428 	return pmd;
1429 }
1430 
1431 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1432 {
1433 	pmd_t __pmd;
1434 	pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1435 	return pmd_mkyoung(__pmd);
1436 }
1437 
1438 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1439 {
1440 	/* Do not clobber PROT_NONE segments! */
1441 	if (!pmd_prot_none(pmd))
1442 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1443 	return pmd;
1444 }
1445 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1446 
1447 static inline void pmdp_flush_lazy(struct mm_struct *mm,
1448 				   unsigned long address, pmd_t *pmdp)
1449 {
1450 	int active = (mm == current->active_mm) ? 1 : 0;
1451 
1452 	if ((atomic_read(&mm->context.attach_count) & 0xffff) > active)
1453 		__pmd_idte(address, pmdp);
1454 	else
1455 		mm->context.flush_mm = 1;
1456 }
1457 
1458 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1459 
1460 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1461 extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1462 				       pgtable_t pgtable);
1463 
1464 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1465 extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1466 
1467 static inline int pmd_trans_splitting(pmd_t pmd)
1468 {
1469 	return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT;
1470 }
1471 
1472 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1473 			      pmd_t *pmdp, pmd_t entry)
1474 {
1475 	if (!(pmd_val(entry) & _SEGMENT_ENTRY_INVALID) && MACHINE_HAS_EDAT1)
1476 		pmd_val(entry) |= _SEGMENT_ENTRY_CO;
1477 	*pmdp = entry;
1478 }
1479 
1480 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1481 {
1482 	pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1483 	return pmd;
1484 }
1485 
1486 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1487 {
1488 	/* Do not clobber PROT_NONE segments! */
1489 	if (!pmd_prot_none(pmd))
1490 		pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1491 	return pmd;
1492 }
1493 
1494 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1495 {
1496 	/* No dirty bit in the segment table entry. */
1497 	return pmd;
1498 }
1499 
1500 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1501 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1502 					    unsigned long address, pmd_t *pmdp)
1503 {
1504 	pmd_t pmd;
1505 
1506 	pmd = *pmdp;
1507 	__pmd_idte(address, pmdp);
1508 	*pmdp = pmd_mkold(pmd);
1509 	return pmd_young(pmd);
1510 }
1511 
1512 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
1513 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
1514 				       unsigned long address, pmd_t *pmdp)
1515 {
1516 	pmd_t pmd = *pmdp;
1517 
1518 	__pmd_idte(address, pmdp);
1519 	pmd_clear(pmdp);
1520 	return pmd;
1521 }
1522 
1523 #define __HAVE_ARCH_PMDP_CLEAR_FLUSH
1524 static inline pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
1525 				     unsigned long address, pmd_t *pmdp)
1526 {
1527 	return pmdp_get_and_clear(vma->vm_mm, address, pmdp);
1528 }
1529 
1530 #define __HAVE_ARCH_PMDP_INVALIDATE
1531 static inline void pmdp_invalidate(struct vm_area_struct *vma,
1532 				   unsigned long address, pmd_t *pmdp)
1533 {
1534 	__pmd_idte(address, pmdp);
1535 }
1536 
1537 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1538 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1539 				      unsigned long address, pmd_t *pmdp)
1540 {
1541 	pmd_t pmd = *pmdp;
1542 
1543 	if (pmd_write(pmd)) {
1544 		__pmd_idte(address, pmdp);
1545 		set_pmd_at(mm, address, pmdp, pmd_wrprotect(pmd));
1546 	}
1547 }
1548 
1549 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1550 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1551 
1552 static inline int pmd_trans_huge(pmd_t pmd)
1553 {
1554 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1555 }
1556 
1557 static inline int has_transparent_hugepage(void)
1558 {
1559 	return MACHINE_HAS_HPAGE ? 1 : 0;
1560 }
1561 
1562 static inline unsigned long pmd_pfn(pmd_t pmd)
1563 {
1564 	return pmd_val(pmd) >> PAGE_SHIFT;
1565 }
1566 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1567 
1568 /*
1569  * 31 bit swap entry format:
1570  * A page-table entry has some bits we have to treat in a special way.
1571  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1572  * exception will occur instead of a page translation exception. The
1573  * specifiation exception has the bad habit not to store necessary
1574  * information in the lowcore.
1575  * Bits 21, 22, 30 and 31 are used to indicate the page type.
1576  * A swap pte is indicated by bit pattern (pte & 0x603) == 0x402
1577  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1578  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1579  * plus 24 for the offset.
1580  * 0|     offset        |0110|o|type |00|
1581  * 0 0000000001111111111 2222 2 22222 33
1582  * 0 1234567890123456789 0123 4 56789 01
1583  *
1584  * 64 bit swap entry format:
1585  * A page-table entry has some bits we have to treat in a special way.
1586  * Bits 52 and bit 55 have to be zero, otherwise an specification
1587  * exception will occur instead of a page translation exception. The
1588  * specifiation exception has the bad habit not to store necessary
1589  * information in the lowcore.
1590  * Bits 53, 54, 62 and 63 are used to indicate the page type.
1591  * A swap pte is indicated by bit pattern (pte & 0x603) == 0x402
1592  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1593  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1594  * plus 56 for the offset.
1595  * |                      offset                        |0110|o|type |00|
1596  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1597  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1598  */
1599 #ifndef CONFIG_64BIT
1600 #define __SWP_OFFSET_MASK (~0UL >> 12)
1601 #else
1602 #define __SWP_OFFSET_MASK (~0UL >> 11)
1603 #endif
1604 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1605 {
1606 	pte_t pte;
1607 	offset &= __SWP_OFFSET_MASK;
1608 	pte_val(pte) = _PAGE_INVALID | _PAGE_TYPE | ((type & 0x1f) << 2) |
1609 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1610 	return pte;
1611 }
1612 
1613 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1614 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1615 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1616 
1617 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1618 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1619 
1620 #ifndef CONFIG_64BIT
1621 # define PTE_FILE_MAX_BITS	26
1622 #else /* CONFIG_64BIT */
1623 # define PTE_FILE_MAX_BITS	59
1624 #endif /* CONFIG_64BIT */
1625 
1626 #define pte_to_pgoff(__pte) \
1627 	((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1628 
1629 #define pgoff_to_pte(__off) \
1630 	((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1631 		   | _PAGE_INVALID | _PAGE_PROTECT })
1632 
1633 #endif /* !__ASSEMBLY__ */
1634 
1635 #define kern_addr_valid(addr)   (1)
1636 
1637 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1638 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1639 extern int s390_enable_sie(void);
1640 
1641 /*
1642  * No page table caches to initialise
1643  */
1644 static inline void pgtable_cache_init(void) { }
1645 static inline void check_pgt_cache(void) { }
1646 
1647 #include <asm-generic/pgtable.h>
1648 
1649 #endif /* _S390_PAGE_H */
1650