xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision a03a8dbe20eff6d57aae3147577bf84b52aba4e6)
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 <linux/radix-tree.h>
34 #include <asm/bug.h>
35 #include <asm/page.h>
36 
37 extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
38 extern void paging_init(void);
39 extern void vmem_map_init(void);
40 
41 /*
42  * The S390 doesn't have any external MMU info: the kernel page
43  * tables contain all the necessary information.
44  */
45 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
46 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
47 
48 /*
49  * ZERO_PAGE is a global shared page that is always zero; used
50  * for zero-mapped memory areas etc..
51  */
52 
53 extern unsigned long empty_zero_page;
54 extern unsigned long zero_page_mask;
55 
56 #define ZERO_PAGE(vaddr) \
57 	(virt_to_page((void *)(empty_zero_page + \
58 	 (((unsigned long)(vaddr)) &zero_page_mask))))
59 #define __HAVE_COLOR_ZERO_PAGE
60 
61 /* TODO: s390 cannot support io_remap_pfn_range... */
62 #endif /* !__ASSEMBLY__ */
63 
64 /*
65  * PMD_SHIFT determines the size of the area a second-level page
66  * table can map
67  * PGDIR_SHIFT determines what a third-level page table entry can map
68  */
69 #ifndef CONFIG_64BIT
70 # define PMD_SHIFT	20
71 # define PUD_SHIFT	20
72 # define PGDIR_SHIFT	20
73 #else /* CONFIG_64BIT */
74 # define PMD_SHIFT	20
75 # define PUD_SHIFT	31
76 # define PGDIR_SHIFT	42
77 #endif /* CONFIG_64BIT */
78 
79 #define PMD_SIZE        (1UL << PMD_SHIFT)
80 #define PMD_MASK        (~(PMD_SIZE-1))
81 #define PUD_SIZE	(1UL << PUD_SHIFT)
82 #define PUD_MASK	(~(PUD_SIZE-1))
83 #define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
84 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
85 
86 /*
87  * entries per page directory level: the S390 is two-level, so
88  * we don't really have any PMD directory physically.
89  * for S390 segment-table entries are combined to one PGD
90  * that leads to 1024 pte per pgd
91  */
92 #define PTRS_PER_PTE	256
93 #ifndef CONFIG_64BIT
94 #define __PAGETABLE_PUD_FOLDED
95 #define PTRS_PER_PMD	1
96 #define __PAGETABLE_PMD_FOLDED
97 #define PTRS_PER_PUD	1
98 #else /* CONFIG_64BIT */
99 #define PTRS_PER_PMD	2048
100 #define PTRS_PER_PUD	2048
101 #endif /* CONFIG_64BIT */
102 #define PTRS_PER_PGD	2048
103 
104 #define FIRST_USER_ADDRESS  0UL
105 
106 #define pte_ERROR(e) \
107 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
108 #define pmd_ERROR(e) \
109 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
110 #define pud_ERROR(e) \
111 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
112 #define pgd_ERROR(e) \
113 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
114 
115 #ifndef __ASSEMBLY__
116 /*
117  * The vmalloc and module area will always be on the topmost area of the kernel
118  * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc and modules.
119  * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
120  * modules will reside. That makes sure that inter module branches always
121  * happen without trampolines and in addition the placement within a 2GB frame
122  * is branch prediction unit friendly.
123  */
124 extern unsigned long VMALLOC_START;
125 extern unsigned long VMALLOC_END;
126 extern struct page *vmemmap;
127 
128 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
129 
130 #ifdef CONFIG_64BIT
131 extern unsigned long MODULES_VADDR;
132 extern unsigned long MODULES_END;
133 #define MODULES_VADDR	MODULES_VADDR
134 #define MODULES_END	MODULES_END
135 #define MODULES_LEN	(1UL << 31)
136 #endif
137 
138 static inline int is_module_addr(void *addr)
139 {
140 #ifdef CONFIG_64BIT
141 	BUILD_BUG_ON(MODULES_LEN > (1UL << 31));
142 	if (addr < (void *)MODULES_VADDR)
143 		return 0;
144 	if (addr > (void *)MODULES_END)
145 		return 0;
146 #endif
147 	return 1;
148 }
149 
150 /*
151  * A 31 bit pagetable entry of S390 has following format:
152  *  |   PFRA          |    |  OS  |
153  * 0                   0IP0
154  * 00000000001111111111222222222233
155  * 01234567890123456789012345678901
156  *
157  * I Page-Invalid Bit:    Page is not available for address-translation
158  * P Page-Protection Bit: Store access not possible for page
159  *
160  * A 31 bit segmenttable entry of S390 has following format:
161  *  |   P-table origin      |  |PTL
162  * 0                         IC
163  * 00000000001111111111222222222233
164  * 01234567890123456789012345678901
165  *
166  * I Segment-Invalid Bit:    Segment is not available for address-translation
167  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
168  * PTL Page-Table-Length:    Page-table length (PTL+1*16 entries -> up to 256)
169  *
170  * The 31 bit segmenttable origin of S390 has following format:
171  *
172  *  |S-table origin   |     | STL |
173  * X                   **GPS
174  * 00000000001111111111222222222233
175  * 01234567890123456789012345678901
176  *
177  * X Space-Switch event:
178  * G Segment-Invalid Bit:     *
179  * P Private-Space Bit:       Segment is not private (PoP 3-30)
180  * S Storage-Alteration:
181  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
182  *
183  * A 64 bit pagetable entry of S390 has following format:
184  * |			 PFRA			      |0IPC|  OS  |
185  * 0000000000111111111122222222223333333333444444444455555555556666
186  * 0123456789012345678901234567890123456789012345678901234567890123
187  *
188  * I Page-Invalid Bit:    Page is not available for address-translation
189  * P Page-Protection Bit: Store access not possible for page
190  * C Change-bit override: HW is not required to set change bit
191  *
192  * A 64 bit segmenttable entry of S390 has following format:
193  * |        P-table origin                              |      TT
194  * 0000000000111111111122222222223333333333444444444455555555556666
195  * 0123456789012345678901234567890123456789012345678901234567890123
196  *
197  * I Segment-Invalid Bit:    Segment is not available for address-translation
198  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
199  * P Page-Protection Bit: Store access not possible for page
200  * TT Type 00
201  *
202  * A 64 bit region table entry of S390 has following format:
203  * |        S-table origin                             |   TF  TTTL
204  * 0000000000111111111122222222223333333333444444444455555555556666
205  * 0123456789012345678901234567890123456789012345678901234567890123
206  *
207  * I Segment-Invalid Bit:    Segment is not available for address-translation
208  * TT Type 01
209  * TF
210  * TL Table length
211  *
212  * The 64 bit regiontable origin of S390 has following format:
213  * |      region table origon                          |       DTTL
214  * 0000000000111111111122222222223333333333444444444455555555556666
215  * 0123456789012345678901234567890123456789012345678901234567890123
216  *
217  * X Space-Switch event:
218  * G Segment-Invalid Bit:
219  * P Private-Space Bit:
220  * S Storage-Alteration:
221  * R Real space
222  * TL Table-Length:
223  *
224  * A storage key has the following format:
225  * | ACC |F|R|C|0|
226  *  0   3 4 5 6 7
227  * ACC: access key
228  * F  : fetch protection bit
229  * R  : referenced bit
230  * C  : changed bit
231  */
232 
233 /* Hardware bits in the page table entry */
234 #define _PAGE_PROTECT	0x200		/* HW read-only bit  */
235 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
236 #define _PAGE_LARGE	0x800		/* Bit to mark a large pte */
237 
238 /* Software bits in the page table entry */
239 #define _PAGE_PRESENT	0x001		/* SW pte present bit */
240 #define _PAGE_TYPE	0x002		/* SW pte type bit */
241 #define _PAGE_YOUNG	0x004		/* SW pte young bit */
242 #define _PAGE_DIRTY	0x008		/* SW pte dirty bit */
243 #define _PAGE_READ	0x010		/* SW pte read bit */
244 #define _PAGE_WRITE	0x020		/* SW pte write bit */
245 #define _PAGE_SPECIAL	0x040		/* SW associated with special page */
246 #define _PAGE_UNUSED	0x080		/* SW bit for pgste usage state */
247 #define __HAVE_ARCH_PTE_SPECIAL
248 
249 /* Set of bits not changed in pte_modify */
250 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
251 				 _PAGE_YOUNG)
252 
253 /*
254  * handle_pte_fault uses pte_present and pte_none to find out the pte type
255  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
256  * distinguish present from not-present ptes. It is changed only with the page
257  * table lock held.
258  *
259  * The following table gives the different possible bit combinations for
260  * the pte hardware and software bits in the last 12 bits of a pte:
261  *
262  *				842100000000
263  *				000084210000
264  *				000000008421
265  *				.IR...wrdytp
266  * empty			.10...000000
267  * swap				.10...xxxx10
268  * file				.11...xxxxx0
269  * prot-none, clean, old	.11...000001
270  * prot-none, clean, young	.11...000101
271  * prot-none, dirty, old	.10...001001
272  * prot-none, dirty, young	.10...001101
273  * read-only, clean, old	.11...010001
274  * read-only, clean, young	.01...010101
275  * read-only, dirty, old	.11...011001
276  * read-only, dirty, young	.01...011101
277  * read-write, clean, old	.11...110001
278  * read-write, clean, young	.01...110101
279  * read-write, dirty, old	.10...111001
280  * read-write, dirty, young	.00...111101
281  *
282  * pte_present is true for the bit pattern .xx...xxxxx1, (pte & 0x001) == 0x001
283  * pte_none    is true for the bit pattern .10...xxxx00, (pte & 0x603) == 0x400
284  * pte_swap    is true for the bit pattern .10...xxxx10, (pte & 0x603) == 0x402
285  */
286 
287 #ifndef CONFIG_64BIT
288 
289 /* Bits in the segment table address-space-control-element */
290 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
291 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
292 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
293 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
294 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
295 
296 /* Bits in the segment table entry */
297 #define _SEGMENT_ENTRY_BITS	0x7fffffffUL	/* Valid segment table bits */
298 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
299 #define _SEGMENT_ENTRY_PROTECT	0x200	/* page protection bit		    */
300 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
301 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
302 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
303 
304 #define _SEGMENT_ENTRY_DIRTY	0	/* No sw dirty bit for 31-bit */
305 #define _SEGMENT_ENTRY_YOUNG	0	/* No sw young bit for 31-bit */
306 #define _SEGMENT_ENTRY_READ	0	/* No sw read bit for 31-bit */
307 #define _SEGMENT_ENTRY_WRITE	0	/* No sw write bit for 31-bit */
308 #define _SEGMENT_ENTRY_LARGE	0	/* No large pages for 31-bit */
309 #define _SEGMENT_ENTRY_BITS_LARGE 0
310 #define _SEGMENT_ENTRY_ORIGIN_LARGE 0
311 
312 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
313 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
314 
315 /*
316  * Segment table entry encoding (I = invalid, R = read-only bit):
317  *		..R...I.....
318  * prot-none	..1...1.....
319  * read-only	..1...0.....
320  * read-write	..0...0.....
321  * empty	..0...1.....
322  */
323 
324 /* Page status table bits for virtualization */
325 #define PGSTE_ACC_BITS	0xf0000000UL
326 #define PGSTE_FP_BIT	0x08000000UL
327 #define PGSTE_PCL_BIT	0x00800000UL
328 #define PGSTE_HR_BIT	0x00400000UL
329 #define PGSTE_HC_BIT	0x00200000UL
330 #define PGSTE_GR_BIT	0x00040000UL
331 #define PGSTE_GC_BIT	0x00020000UL
332 #define PGSTE_UC_BIT	0x00008000UL	/* user dirty (migration) */
333 #define PGSTE_IN_BIT	0x00004000UL	/* IPTE notify bit */
334 
335 #else /* CONFIG_64BIT */
336 
337 /* Bits in the segment/region table address-space-control-element */
338 #define _ASCE_ORIGIN		~0xfffUL/* segment table origin		    */
339 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
340 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
341 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
342 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
343 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
344 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
345 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
346 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
347 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
348 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
349 
350 /* Bits in the region table entry */
351 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
352 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
353 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
354 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
355 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
356 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
357 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
358 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
359 
360 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
361 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
362 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
363 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
364 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
365 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
366 
367 #define _REGION3_ENTRY_LARGE	0x400	/* RTTE-format control, large page  */
368 #define _REGION3_ENTRY_RO	0x200	/* page protection bit		    */
369 
370 /* Bits in the segment table entry */
371 #define _SEGMENT_ENTRY_BITS	0xfffffffffffffe33UL
372 #define _SEGMENT_ENTRY_BITS_LARGE 0xfffffffffff0ff33UL
373 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
374 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* segment table origin		    */
375 #define _SEGMENT_ENTRY_PROTECT	0x200	/* page protection bit		    */
376 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
377 
378 #define _SEGMENT_ENTRY		(0)
379 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
380 
381 #define _SEGMENT_ENTRY_DIRTY	0x2000	/* SW segment dirty bit */
382 #define _SEGMENT_ENTRY_YOUNG	0x1000	/* SW segment young bit */
383 #define _SEGMENT_ENTRY_SPLIT	0x0800	/* THP splitting bit */
384 #define _SEGMENT_ENTRY_LARGE	0x0400	/* STE-format control, large page */
385 #define _SEGMENT_ENTRY_READ	0x0002	/* SW segment read bit */
386 #define _SEGMENT_ENTRY_WRITE	0x0001	/* SW segment write bit */
387 
388 /*
389  * Segment table entry encoding (R = read-only, I = invalid, y = young bit):
390  *				dy..R...I...wr
391  * prot-none, clean, old	00..1...1...00
392  * prot-none, clean, young	01..1...1...00
393  * prot-none, dirty, old	10..1...1...00
394  * prot-none, dirty, young	11..1...1...00
395  * read-only, clean, old	00..1...1...01
396  * read-only, clean, young	01..1...0...01
397  * read-only, dirty, old	10..1...1...01
398  * read-only, dirty, young	11..1...0...01
399  * read-write, clean, old	00..1...1...11
400  * read-write, clean, young	01..1...0...11
401  * read-write, dirty, old	10..0...1...11
402  * read-write, dirty, young	11..0...0...11
403  * The segment table origin is used to distinguish empty (origin==0) from
404  * read-write, old segment table entries (origin!=0)
405  */
406 
407 #define _SEGMENT_ENTRY_SPLIT_BIT 11	/* THP splitting bit number */
408 
409 /* Page status table bits for virtualization */
410 #define PGSTE_ACC_BITS	0xf000000000000000UL
411 #define PGSTE_FP_BIT	0x0800000000000000UL
412 #define PGSTE_PCL_BIT	0x0080000000000000UL
413 #define PGSTE_HR_BIT	0x0040000000000000UL
414 #define PGSTE_HC_BIT	0x0020000000000000UL
415 #define PGSTE_GR_BIT	0x0004000000000000UL
416 #define PGSTE_GC_BIT	0x0002000000000000UL
417 #define PGSTE_UC_BIT	0x0000800000000000UL	/* user dirty (migration) */
418 #define PGSTE_IN_BIT	0x0000400000000000UL	/* IPTE notify bit */
419 
420 #endif /* CONFIG_64BIT */
421 
422 /* Guest Page State used for virtualization */
423 #define _PGSTE_GPS_ZERO		0x0000000080000000UL
424 #define _PGSTE_GPS_USAGE_MASK	0x0000000003000000UL
425 #define _PGSTE_GPS_USAGE_STABLE 0x0000000000000000UL
426 #define _PGSTE_GPS_USAGE_UNUSED 0x0000000001000000UL
427 
428 /*
429  * A user page table pointer has the space-switch-event bit, the
430  * private-space-control bit and the storage-alteration-event-control
431  * bit set. A kernel page table pointer doesn't need them.
432  */
433 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
434 				 _ASCE_ALT_EVENT)
435 
436 /*
437  * Page protection definitions.
438  */
439 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID)
440 #define PAGE_READ	__pgprot(_PAGE_PRESENT | _PAGE_READ | \
441 				 _PAGE_INVALID | _PAGE_PROTECT)
442 #define PAGE_WRITE	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
443 				 _PAGE_INVALID | _PAGE_PROTECT)
444 
445 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
446 				 _PAGE_YOUNG | _PAGE_DIRTY)
447 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
448 				 _PAGE_YOUNG | _PAGE_DIRTY)
449 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
450 				 _PAGE_PROTECT)
451 
452 /*
453  * On s390 the page table entry has an invalid bit and a read-only bit.
454  * Read permission implies execute permission and write permission
455  * implies read permission.
456  */
457          /*xwr*/
458 #define __P000	PAGE_NONE
459 #define __P001	PAGE_READ
460 #define __P010	PAGE_READ
461 #define __P011	PAGE_READ
462 #define __P100	PAGE_READ
463 #define __P101	PAGE_READ
464 #define __P110	PAGE_READ
465 #define __P111	PAGE_READ
466 
467 #define __S000	PAGE_NONE
468 #define __S001	PAGE_READ
469 #define __S010	PAGE_WRITE
470 #define __S011	PAGE_WRITE
471 #define __S100	PAGE_READ
472 #define __S101	PAGE_READ
473 #define __S110	PAGE_WRITE
474 #define __S111	PAGE_WRITE
475 
476 /*
477  * Segment entry (large page) protection definitions.
478  */
479 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_INVALID | \
480 				 _SEGMENT_ENTRY_PROTECT)
481 #define SEGMENT_READ	__pgprot(_SEGMENT_ENTRY_PROTECT | \
482 				 _SEGMENT_ENTRY_READ)
483 #define SEGMENT_WRITE	__pgprot(_SEGMENT_ENTRY_READ | \
484 				 _SEGMENT_ENTRY_WRITE)
485 
486 static inline int mm_has_pgste(struct mm_struct *mm)
487 {
488 #ifdef CONFIG_PGSTE
489 	if (unlikely(mm->context.has_pgste))
490 		return 1;
491 #endif
492 	return 0;
493 }
494 
495 /*
496  * In the case that a guest uses storage keys
497  * faults should no longer be backed by zero pages
498  */
499 #define mm_forbids_zeropage mm_use_skey
500 static inline int mm_use_skey(struct mm_struct *mm)
501 {
502 #ifdef CONFIG_PGSTE
503 	if (mm->context.use_skey)
504 		return 1;
505 #endif
506 	return 0;
507 }
508 
509 /*
510  * pgd/pmd/pte query functions
511  */
512 #ifndef CONFIG_64BIT
513 
514 static inline int pgd_present(pgd_t pgd) { return 1; }
515 static inline int pgd_none(pgd_t pgd)    { return 0; }
516 static inline int pgd_bad(pgd_t pgd)     { return 0; }
517 
518 static inline int pud_present(pud_t pud) { return 1; }
519 static inline int pud_none(pud_t pud)	 { return 0; }
520 static inline int pud_large(pud_t pud)	 { return 0; }
521 static inline int pud_bad(pud_t pud)	 { return 0; }
522 
523 #else /* CONFIG_64BIT */
524 
525 static inline int pgd_present(pgd_t pgd)
526 {
527 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
528 		return 1;
529 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
530 }
531 
532 static inline int pgd_none(pgd_t pgd)
533 {
534 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
535 		return 0;
536 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
537 }
538 
539 static inline int pgd_bad(pgd_t pgd)
540 {
541 	/*
542 	 * With dynamic page table levels the pgd can be a region table
543 	 * entry or a segment table entry. Check for the bit that are
544 	 * invalid for either table entry.
545 	 */
546 	unsigned long mask =
547 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
548 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
549 	return (pgd_val(pgd) & mask) != 0;
550 }
551 
552 static inline int pud_present(pud_t pud)
553 {
554 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
555 		return 1;
556 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
557 }
558 
559 static inline int pud_none(pud_t pud)
560 {
561 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
562 		return 0;
563 	return (pud_val(pud) & _REGION_ENTRY_INVALID) != 0UL;
564 }
565 
566 static inline int pud_large(pud_t pud)
567 {
568 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
569 		return 0;
570 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
571 }
572 
573 static inline int pud_bad(pud_t pud)
574 {
575 	/*
576 	 * With dynamic page table levels the pud can be a region table
577 	 * entry or a segment table entry. Check for the bit that are
578 	 * invalid for either table entry.
579 	 */
580 	unsigned long mask =
581 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
582 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
583 	return (pud_val(pud) & mask) != 0;
584 }
585 
586 #endif /* CONFIG_64BIT */
587 
588 static inline int pmd_present(pmd_t pmd)
589 {
590 	return pmd_val(pmd) != _SEGMENT_ENTRY_INVALID;
591 }
592 
593 static inline int pmd_none(pmd_t pmd)
594 {
595 	return pmd_val(pmd) == _SEGMENT_ENTRY_INVALID;
596 }
597 
598 static inline int pmd_large(pmd_t pmd)
599 {
600 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
601 }
602 
603 static inline int pmd_pfn(pmd_t pmd)
604 {
605 	unsigned long origin_mask;
606 
607 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
608 	if (pmd_large(pmd))
609 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
610 	return (pmd_val(pmd) & origin_mask) >> PAGE_SHIFT;
611 }
612 
613 static inline int pmd_bad(pmd_t pmd)
614 {
615 	if (pmd_large(pmd))
616 		return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
617 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
618 }
619 
620 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
621 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
622 				 unsigned long addr, pmd_t *pmdp);
623 
624 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
625 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
626 				 unsigned long address, pmd_t *pmdp,
627 				 pmd_t entry, int dirty);
628 
629 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
630 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
631 				  unsigned long address, pmd_t *pmdp);
632 
633 #define __HAVE_ARCH_PMD_WRITE
634 static inline int pmd_write(pmd_t pmd)
635 {
636 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
637 }
638 
639 static inline int pmd_dirty(pmd_t pmd)
640 {
641 	int dirty = 1;
642 	if (pmd_large(pmd))
643 		dirty = (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
644 	return dirty;
645 }
646 
647 static inline int pmd_young(pmd_t pmd)
648 {
649 	int young = 1;
650 	if (pmd_large(pmd))
651 		young = (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
652 	return young;
653 }
654 
655 static inline int pte_present(pte_t pte)
656 {
657 	/* Bit pattern: (pte & 0x001) == 0x001 */
658 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
659 }
660 
661 static inline int pte_none(pte_t pte)
662 {
663 	/* Bit pattern: pte == 0x400 */
664 	return pte_val(pte) == _PAGE_INVALID;
665 }
666 
667 static inline int pte_swap(pte_t pte)
668 {
669 	/* Bit pattern: (pte & 0x603) == 0x402 */
670 	return (pte_val(pte) & (_PAGE_INVALID | _PAGE_PROTECT |
671 				_PAGE_TYPE | _PAGE_PRESENT))
672 		== (_PAGE_INVALID | _PAGE_TYPE);
673 }
674 
675 static inline int pte_special(pte_t pte)
676 {
677 	return (pte_val(pte) & _PAGE_SPECIAL);
678 }
679 
680 #define __HAVE_ARCH_PTE_SAME
681 static inline int pte_same(pte_t a, pte_t b)
682 {
683 	return pte_val(a) == pte_val(b);
684 }
685 
686 static inline pgste_t pgste_get_lock(pte_t *ptep)
687 {
688 	unsigned long new = 0;
689 #ifdef CONFIG_PGSTE
690 	unsigned long old;
691 
692 	preempt_disable();
693 	asm(
694 		"	lg	%0,%2\n"
695 		"0:	lgr	%1,%0\n"
696 		"	nihh	%0,0xff7f\n"	/* clear PCL bit in old */
697 		"	oihh	%1,0x0080\n"	/* set PCL bit in new */
698 		"	csg	%0,%1,%2\n"
699 		"	jl	0b\n"
700 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
701 		: "Q" (ptep[PTRS_PER_PTE]) : "cc", "memory");
702 #endif
703 	return __pgste(new);
704 }
705 
706 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
707 {
708 #ifdef CONFIG_PGSTE
709 	asm(
710 		"	nihh	%1,0xff7f\n"	/* clear PCL bit */
711 		"	stg	%1,%0\n"
712 		: "=Q" (ptep[PTRS_PER_PTE])
713 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE])
714 		: "cc", "memory");
715 	preempt_enable();
716 #endif
717 }
718 
719 static inline pgste_t pgste_get(pte_t *ptep)
720 {
721 	unsigned long pgste = 0;
722 #ifdef CONFIG_PGSTE
723 	pgste = *(unsigned long *)(ptep + PTRS_PER_PTE);
724 #endif
725 	return __pgste(pgste);
726 }
727 
728 static inline void pgste_set(pte_t *ptep, pgste_t pgste)
729 {
730 #ifdef CONFIG_PGSTE
731 	*(pgste_t *)(ptep + PTRS_PER_PTE) = pgste;
732 #endif
733 }
734 
735 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste,
736 				       struct mm_struct *mm)
737 {
738 #ifdef CONFIG_PGSTE
739 	unsigned long address, bits, skey;
740 
741 	if (!mm_use_skey(mm) || pte_val(*ptep) & _PAGE_INVALID)
742 		return pgste;
743 	address = pte_val(*ptep) & PAGE_MASK;
744 	skey = (unsigned long) page_get_storage_key(address);
745 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
746 	/* Transfer page changed & referenced bit to guest bits in pgste */
747 	pgste_val(pgste) |= bits << 48;		/* GR bit & GC bit */
748 	/* Copy page access key and fetch protection bit to pgste */
749 	pgste_val(pgste) &= ~(PGSTE_ACC_BITS | PGSTE_FP_BIT);
750 	pgste_val(pgste) |= (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
751 #endif
752 	return pgste;
753 
754 }
755 
756 static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry,
757 				 struct mm_struct *mm)
758 {
759 #ifdef CONFIG_PGSTE
760 	unsigned long address;
761 	unsigned long nkey;
762 
763 	if (!mm_use_skey(mm) || pte_val(entry) & _PAGE_INVALID)
764 		return;
765 	VM_BUG_ON(!(pte_val(*ptep) & _PAGE_INVALID));
766 	address = pte_val(entry) & PAGE_MASK;
767 	/*
768 	 * Set page access key and fetch protection bit from pgste.
769 	 * The guest C/R information is still in the PGSTE, set real
770 	 * key C/R to 0.
771 	 */
772 	nkey = (pgste_val(pgste) & (PGSTE_ACC_BITS | PGSTE_FP_BIT)) >> 56;
773 	nkey |= (pgste_val(pgste) & (PGSTE_GR_BIT | PGSTE_GC_BIT)) >> 48;
774 	page_set_storage_key(address, nkey, 0);
775 #endif
776 }
777 
778 static inline pgste_t pgste_set_pte(pte_t *ptep, pgste_t pgste, pte_t entry)
779 {
780 	if ((pte_val(entry) & _PAGE_PRESENT) &&
781 	    (pte_val(entry) & _PAGE_WRITE) &&
782 	    !(pte_val(entry) & _PAGE_INVALID)) {
783 		if (!MACHINE_HAS_ESOP) {
784 			/*
785 			 * Without enhanced suppression-on-protection force
786 			 * the dirty bit on for all writable ptes.
787 			 */
788 			pte_val(entry) |= _PAGE_DIRTY;
789 			pte_val(entry) &= ~_PAGE_PROTECT;
790 		}
791 		if (!(pte_val(entry) & _PAGE_PROTECT))
792 			/* This pte allows write access, set user-dirty */
793 			pgste_val(pgste) |= PGSTE_UC_BIT;
794 	}
795 	*ptep = entry;
796 	return pgste;
797 }
798 
799 /**
800  * struct gmap_struct - guest address space
801  * @crst_list: list of all crst tables used in the guest address space
802  * @mm: pointer to the parent mm_struct
803  * @guest_to_host: radix tree with guest to host address translation
804  * @host_to_guest: radix tree with pointer to segment table entries
805  * @guest_table_lock: spinlock to protect all entries in the guest page table
806  * @table: pointer to the page directory
807  * @asce: address space control element for gmap page table
808  * @pfault_enabled: defines if pfaults are applicable for the guest
809  */
810 struct gmap {
811 	struct list_head list;
812 	struct list_head crst_list;
813 	struct mm_struct *mm;
814 	struct radix_tree_root guest_to_host;
815 	struct radix_tree_root host_to_guest;
816 	spinlock_t guest_table_lock;
817 	unsigned long *table;
818 	unsigned long asce;
819 	unsigned long asce_end;
820 	void *private;
821 	bool pfault_enabled;
822 };
823 
824 /**
825  * struct gmap_notifier - notify function block for page invalidation
826  * @notifier_call: address of callback function
827  */
828 struct gmap_notifier {
829 	struct list_head list;
830 	void (*notifier_call)(struct gmap *gmap, unsigned long gaddr);
831 };
832 
833 struct gmap *gmap_alloc(struct mm_struct *mm, unsigned long limit);
834 void gmap_free(struct gmap *gmap);
835 void gmap_enable(struct gmap *gmap);
836 void gmap_disable(struct gmap *gmap);
837 int gmap_map_segment(struct gmap *gmap, unsigned long from,
838 		     unsigned long to, unsigned long len);
839 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
840 unsigned long __gmap_translate(struct gmap *, unsigned long gaddr);
841 unsigned long gmap_translate(struct gmap *, unsigned long gaddr);
842 int __gmap_link(struct gmap *gmap, unsigned long gaddr, unsigned long vmaddr);
843 int gmap_fault(struct gmap *, unsigned long gaddr, unsigned int fault_flags);
844 void gmap_discard(struct gmap *, unsigned long from, unsigned long to);
845 void __gmap_zap(struct gmap *, unsigned long gaddr);
846 bool gmap_test_and_clear_dirty(unsigned long address, struct gmap *);
847 
848 
849 void gmap_register_ipte_notifier(struct gmap_notifier *);
850 void gmap_unregister_ipte_notifier(struct gmap_notifier *);
851 int gmap_ipte_notify(struct gmap *, unsigned long start, unsigned long len);
852 void gmap_do_ipte_notify(struct mm_struct *, unsigned long addr, pte_t *);
853 
854 static inline pgste_t pgste_ipte_notify(struct mm_struct *mm,
855 					unsigned long addr,
856 					pte_t *ptep, pgste_t pgste)
857 {
858 #ifdef CONFIG_PGSTE
859 	if (pgste_val(pgste) & PGSTE_IN_BIT) {
860 		pgste_val(pgste) &= ~PGSTE_IN_BIT;
861 		gmap_do_ipte_notify(mm, addr, ptep);
862 	}
863 #endif
864 	return pgste;
865 }
866 
867 /*
868  * Certain architectures need to do special things when PTEs
869  * within a page table are directly modified.  Thus, the following
870  * hook is made available.
871  */
872 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
873 			      pte_t *ptep, pte_t entry)
874 {
875 	pgste_t pgste;
876 
877 	if (mm_has_pgste(mm)) {
878 		pgste = pgste_get_lock(ptep);
879 		pgste_val(pgste) &= ~_PGSTE_GPS_ZERO;
880 		pgste_set_key(ptep, pgste, entry, mm);
881 		pgste = pgste_set_pte(ptep, pgste, entry);
882 		pgste_set_unlock(ptep, pgste);
883 	} else {
884 		*ptep = entry;
885 	}
886 }
887 
888 /*
889  * query functions pte_write/pte_dirty/pte_young only work if
890  * pte_present() is true. Undefined behaviour if not..
891  */
892 static inline int pte_write(pte_t pte)
893 {
894 	return (pte_val(pte) & _PAGE_WRITE) != 0;
895 }
896 
897 static inline int pte_dirty(pte_t pte)
898 {
899 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
900 }
901 
902 static inline int pte_young(pte_t pte)
903 {
904 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
905 }
906 
907 #define __HAVE_ARCH_PTE_UNUSED
908 static inline int pte_unused(pte_t pte)
909 {
910 	return pte_val(pte) & _PAGE_UNUSED;
911 }
912 
913 /*
914  * pgd/pmd/pte modification functions
915  */
916 
917 static inline void pgd_clear(pgd_t *pgd)
918 {
919 #ifdef CONFIG_64BIT
920 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
921 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
922 #endif
923 }
924 
925 static inline void pud_clear(pud_t *pud)
926 {
927 #ifdef CONFIG_64BIT
928 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
929 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
930 #endif
931 }
932 
933 static inline void pmd_clear(pmd_t *pmdp)
934 {
935 	pmd_val(*pmdp) = _SEGMENT_ENTRY_INVALID;
936 }
937 
938 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
939 {
940 	pte_val(*ptep) = _PAGE_INVALID;
941 }
942 
943 /*
944  * The following pte modification functions only work if
945  * pte_present() is true. Undefined behaviour if not..
946  */
947 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
948 {
949 	pte_val(pte) &= _PAGE_CHG_MASK;
950 	pte_val(pte) |= pgprot_val(newprot);
951 	/*
952 	 * newprot for PAGE_NONE, PAGE_READ and PAGE_WRITE has the
953 	 * invalid bit set, clear it again for readable, young pages
954 	 */
955 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
956 		pte_val(pte) &= ~_PAGE_INVALID;
957 	/*
958 	 * newprot for PAGE_READ and PAGE_WRITE has the page protection
959 	 * bit set, clear it again for writable, dirty pages
960 	 */
961 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
962 		pte_val(pte) &= ~_PAGE_PROTECT;
963 	return pte;
964 }
965 
966 static inline pte_t pte_wrprotect(pte_t pte)
967 {
968 	pte_val(pte) &= ~_PAGE_WRITE;
969 	pte_val(pte) |= _PAGE_PROTECT;
970 	return pte;
971 }
972 
973 static inline pte_t pte_mkwrite(pte_t pte)
974 {
975 	pte_val(pte) |= _PAGE_WRITE;
976 	if (pte_val(pte) & _PAGE_DIRTY)
977 		pte_val(pte) &= ~_PAGE_PROTECT;
978 	return pte;
979 }
980 
981 static inline pte_t pte_mkclean(pte_t pte)
982 {
983 	pte_val(pte) &= ~_PAGE_DIRTY;
984 	pte_val(pte) |= _PAGE_PROTECT;
985 	return pte;
986 }
987 
988 static inline pte_t pte_mkdirty(pte_t pte)
989 {
990 	pte_val(pte) |= _PAGE_DIRTY;
991 	if (pte_val(pte) & _PAGE_WRITE)
992 		pte_val(pte) &= ~_PAGE_PROTECT;
993 	return pte;
994 }
995 
996 static inline pte_t pte_mkold(pte_t pte)
997 {
998 	pte_val(pte) &= ~_PAGE_YOUNG;
999 	pte_val(pte) |= _PAGE_INVALID;
1000 	return pte;
1001 }
1002 
1003 static inline pte_t pte_mkyoung(pte_t pte)
1004 {
1005 	pte_val(pte) |= _PAGE_YOUNG;
1006 	if (pte_val(pte) & _PAGE_READ)
1007 		pte_val(pte) &= ~_PAGE_INVALID;
1008 	return pte;
1009 }
1010 
1011 static inline pte_t pte_mkspecial(pte_t pte)
1012 {
1013 	pte_val(pte) |= _PAGE_SPECIAL;
1014 	return pte;
1015 }
1016 
1017 #ifdef CONFIG_HUGETLB_PAGE
1018 static inline pte_t pte_mkhuge(pte_t pte)
1019 {
1020 	pte_val(pte) |= _PAGE_LARGE;
1021 	return pte;
1022 }
1023 #endif
1024 
1025 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
1026 {
1027 	unsigned long pto = (unsigned long) ptep;
1028 
1029 #ifndef CONFIG_64BIT
1030 	/* pto in ESA mode must point to the start of the segment table */
1031 	pto &= 0x7ffffc00;
1032 #endif
1033 	/* Invalidation + global TLB flush for the pte */
1034 	asm volatile(
1035 		"	ipte	%2,%3"
1036 		: "=m" (*ptep) : "m" (*ptep), "a" (pto), "a" (address));
1037 }
1038 
1039 static inline void __ptep_ipte_local(unsigned long address, pte_t *ptep)
1040 {
1041 	unsigned long pto = (unsigned long) ptep;
1042 
1043 #ifndef CONFIG_64BIT
1044 	/* pto in ESA mode must point to the start of the segment table */
1045 	pto &= 0x7ffffc00;
1046 #endif
1047 	/* Invalidation + local TLB flush for the pte */
1048 	asm volatile(
1049 		"	.insn rrf,0xb2210000,%2,%3,0,1"
1050 		: "=m" (*ptep) : "m" (*ptep), "a" (pto), "a" (address));
1051 }
1052 
1053 static inline void __ptep_ipte_range(unsigned long address, int nr, pte_t *ptep)
1054 {
1055 	unsigned long pto = (unsigned long) ptep;
1056 
1057 #ifndef CONFIG_64BIT
1058 	/* pto in ESA mode must point to the start of the segment table */
1059 	pto &= 0x7ffffc00;
1060 #endif
1061 	/* Invalidate a range of ptes + global TLB flush of the ptes */
1062 	do {
1063 		asm volatile(
1064 			"	.insn rrf,0xb2210000,%2,%0,%1,0"
1065 			: "+a" (address), "+a" (nr) : "a" (pto) : "memory");
1066 	} while (nr != 255);
1067 }
1068 
1069 static inline void ptep_flush_direct(struct mm_struct *mm,
1070 				     unsigned long address, pte_t *ptep)
1071 {
1072 	int active, count;
1073 
1074 	if (pte_val(*ptep) & _PAGE_INVALID)
1075 		return;
1076 	active = (mm == current->active_mm) ? 1 : 0;
1077 	count = atomic_add_return(0x10000, &mm->context.attach_count);
1078 	if (MACHINE_HAS_TLB_LC && (count & 0xffff) <= active &&
1079 	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id())))
1080 		__ptep_ipte_local(address, ptep);
1081 	else
1082 		__ptep_ipte(address, ptep);
1083 	atomic_sub(0x10000, &mm->context.attach_count);
1084 }
1085 
1086 static inline void ptep_flush_lazy(struct mm_struct *mm,
1087 				   unsigned long address, pte_t *ptep)
1088 {
1089 	int active, count;
1090 
1091 	if (pte_val(*ptep) & _PAGE_INVALID)
1092 		return;
1093 	active = (mm == current->active_mm) ? 1 : 0;
1094 	count = atomic_add_return(0x10000, &mm->context.attach_count);
1095 	if ((count & 0xffff) <= active) {
1096 		pte_val(*ptep) |= _PAGE_INVALID;
1097 		mm->context.flush_mm = 1;
1098 	} else
1099 		__ptep_ipte(address, ptep);
1100 	atomic_sub(0x10000, &mm->context.attach_count);
1101 }
1102 
1103 /*
1104  * Get (and clear) the user dirty bit for a pte.
1105  */
1106 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
1107 						 unsigned long addr,
1108 						 pte_t *ptep)
1109 {
1110 	pgste_t pgste;
1111 	pte_t pte;
1112 	int dirty;
1113 
1114 	if (!mm_has_pgste(mm))
1115 		return 0;
1116 	pgste = pgste_get_lock(ptep);
1117 	dirty = !!(pgste_val(pgste) & PGSTE_UC_BIT);
1118 	pgste_val(pgste) &= ~PGSTE_UC_BIT;
1119 	pte = *ptep;
1120 	if (dirty && (pte_val(pte) & _PAGE_PRESENT)) {
1121 		pgste = pgste_ipte_notify(mm, addr, ptep, pgste);
1122 		__ptep_ipte(addr, ptep);
1123 		if (MACHINE_HAS_ESOP || !(pte_val(pte) & _PAGE_WRITE))
1124 			pte_val(pte) |= _PAGE_PROTECT;
1125 		else
1126 			pte_val(pte) |= _PAGE_INVALID;
1127 		*ptep = pte;
1128 	}
1129 	pgste_set_unlock(ptep, pgste);
1130 	return dirty;
1131 }
1132 
1133 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1134 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1135 					    unsigned long addr, pte_t *ptep)
1136 {
1137 	pgste_t pgste;
1138 	pte_t pte, oldpte;
1139 	int young;
1140 
1141 	if (mm_has_pgste(vma->vm_mm)) {
1142 		pgste = pgste_get_lock(ptep);
1143 		pgste = pgste_ipte_notify(vma->vm_mm, addr, ptep, pgste);
1144 	}
1145 
1146 	oldpte = pte = *ptep;
1147 	ptep_flush_direct(vma->vm_mm, addr, ptep);
1148 	young = pte_young(pte);
1149 	pte = pte_mkold(pte);
1150 
1151 	if (mm_has_pgste(vma->vm_mm)) {
1152 		pgste = pgste_update_all(&oldpte, pgste, vma->vm_mm);
1153 		pgste = pgste_set_pte(ptep, pgste, pte);
1154 		pgste_set_unlock(ptep, pgste);
1155 	} else
1156 		*ptep = pte;
1157 
1158 	return young;
1159 }
1160 
1161 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1162 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1163 					 unsigned long address, pte_t *ptep)
1164 {
1165 	return ptep_test_and_clear_young(vma, address, ptep);
1166 }
1167 
1168 /*
1169  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1170  * both clear the TLB for the unmapped pte. The reason is that
1171  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1172  * to modify an active pte. The sequence is
1173  *   1) ptep_get_and_clear
1174  *   2) set_pte_at
1175  *   3) flush_tlb_range
1176  * On s390 the tlb needs to get flushed with the modification of the pte
1177  * if the pte is active. The only way how this can be implemented is to
1178  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1179  * is a nop.
1180  */
1181 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1182 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1183 				       unsigned long address, pte_t *ptep)
1184 {
1185 	pgste_t pgste;
1186 	pte_t pte;
1187 
1188 	if (mm_has_pgste(mm)) {
1189 		pgste = pgste_get_lock(ptep);
1190 		pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1191 	}
1192 
1193 	pte = *ptep;
1194 	ptep_flush_lazy(mm, address, ptep);
1195 	pte_val(*ptep) = _PAGE_INVALID;
1196 
1197 	if (mm_has_pgste(mm)) {
1198 		pgste = pgste_update_all(&pte, pgste, mm);
1199 		pgste_set_unlock(ptep, pgste);
1200 	}
1201 	return pte;
1202 }
1203 
1204 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1205 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
1206 					   unsigned long address,
1207 					   pte_t *ptep)
1208 {
1209 	pgste_t pgste;
1210 	pte_t pte;
1211 
1212 	if (mm_has_pgste(mm)) {
1213 		pgste = pgste_get_lock(ptep);
1214 		pgste_ipte_notify(mm, address, ptep, pgste);
1215 	}
1216 
1217 	pte = *ptep;
1218 	ptep_flush_lazy(mm, address, ptep);
1219 
1220 	if (mm_has_pgste(mm)) {
1221 		pgste = pgste_update_all(&pte, pgste, mm);
1222 		pgste_set(ptep, pgste);
1223 	}
1224 	return pte;
1225 }
1226 
1227 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
1228 					   unsigned long address,
1229 					   pte_t *ptep, pte_t pte)
1230 {
1231 	pgste_t pgste;
1232 
1233 	if (mm_has_pgste(mm)) {
1234 		pgste = pgste_get(ptep);
1235 		pgste_set_key(ptep, pgste, pte, mm);
1236 		pgste = pgste_set_pte(ptep, pgste, pte);
1237 		pgste_set_unlock(ptep, pgste);
1238 	} else
1239 		*ptep = pte;
1240 }
1241 
1242 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1243 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1244 				     unsigned long address, pte_t *ptep)
1245 {
1246 	pgste_t pgste;
1247 	pte_t pte;
1248 
1249 	if (mm_has_pgste(vma->vm_mm)) {
1250 		pgste = pgste_get_lock(ptep);
1251 		pgste = pgste_ipte_notify(vma->vm_mm, address, ptep, pgste);
1252 	}
1253 
1254 	pte = *ptep;
1255 	ptep_flush_direct(vma->vm_mm, address, ptep);
1256 	pte_val(*ptep) = _PAGE_INVALID;
1257 
1258 	if (mm_has_pgste(vma->vm_mm)) {
1259 		if ((pgste_val(pgste) & _PGSTE_GPS_USAGE_MASK) ==
1260 		    _PGSTE_GPS_USAGE_UNUSED)
1261 			pte_val(pte) |= _PAGE_UNUSED;
1262 		pgste = pgste_update_all(&pte, pgste, vma->vm_mm);
1263 		pgste_set_unlock(ptep, pgste);
1264 	}
1265 	return pte;
1266 }
1267 
1268 /*
1269  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1270  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1271  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1272  * cannot be accessed while the batched unmap is running. In this case
1273  * full==1 and a simple pte_clear is enough. See tlb.h.
1274  */
1275 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1276 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1277 					    unsigned long address,
1278 					    pte_t *ptep, int full)
1279 {
1280 	pgste_t pgste;
1281 	pte_t pte;
1282 
1283 	if (!full && mm_has_pgste(mm)) {
1284 		pgste = pgste_get_lock(ptep);
1285 		pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1286 	}
1287 
1288 	pte = *ptep;
1289 	if (!full)
1290 		ptep_flush_lazy(mm, address, ptep);
1291 	pte_val(*ptep) = _PAGE_INVALID;
1292 
1293 	if (!full && mm_has_pgste(mm)) {
1294 		pgste = pgste_update_all(&pte, pgste, mm);
1295 		pgste_set_unlock(ptep, pgste);
1296 	}
1297 	return pte;
1298 }
1299 
1300 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1301 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1302 				       unsigned long address, pte_t *ptep)
1303 {
1304 	pgste_t pgste;
1305 	pte_t pte = *ptep;
1306 
1307 	if (pte_write(pte)) {
1308 		if (mm_has_pgste(mm)) {
1309 			pgste = pgste_get_lock(ptep);
1310 			pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1311 		}
1312 
1313 		ptep_flush_lazy(mm, address, ptep);
1314 		pte = pte_wrprotect(pte);
1315 
1316 		if (mm_has_pgste(mm)) {
1317 			pgste = pgste_set_pte(ptep, pgste, pte);
1318 			pgste_set_unlock(ptep, pgste);
1319 		} else
1320 			*ptep = pte;
1321 	}
1322 	return pte;
1323 }
1324 
1325 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1326 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1327 					unsigned long address, pte_t *ptep,
1328 					pte_t entry, int dirty)
1329 {
1330 	pgste_t pgste;
1331 
1332 	if (pte_same(*ptep, entry))
1333 		return 0;
1334 	if (mm_has_pgste(vma->vm_mm)) {
1335 		pgste = pgste_get_lock(ptep);
1336 		pgste = pgste_ipte_notify(vma->vm_mm, address, ptep, pgste);
1337 	}
1338 
1339 	ptep_flush_direct(vma->vm_mm, address, ptep);
1340 
1341 	if (mm_has_pgste(vma->vm_mm)) {
1342 		pgste_set_key(ptep, pgste, entry, vma->vm_mm);
1343 		pgste = pgste_set_pte(ptep, pgste, entry);
1344 		pgste_set_unlock(ptep, pgste);
1345 	} else
1346 		*ptep = entry;
1347 	return 1;
1348 }
1349 
1350 /*
1351  * Conversion functions: convert a page and protection to a page entry,
1352  * and a page entry and page directory to the page they refer to.
1353  */
1354 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1355 {
1356 	pte_t __pte;
1357 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1358 	return pte_mkyoung(__pte);
1359 }
1360 
1361 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1362 {
1363 	unsigned long physpage = page_to_phys(page);
1364 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1365 
1366 	if (pte_write(__pte) && PageDirty(page))
1367 		__pte = pte_mkdirty(__pte);
1368 	return __pte;
1369 }
1370 
1371 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1372 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1373 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1374 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1375 
1376 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1377 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1378 
1379 #ifndef CONFIG_64BIT
1380 
1381 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1382 #define pud_deref(pmd) ({ BUG(); 0UL; })
1383 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1384 
1385 #define pud_offset(pgd, address) ((pud_t *) pgd)
1386 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1387 
1388 #else /* CONFIG_64BIT */
1389 
1390 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1391 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1392 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1393 
1394 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1395 {
1396 	pud_t *pud = (pud_t *) pgd;
1397 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1398 		pud = (pud_t *) pgd_deref(*pgd);
1399 	return pud  + pud_index(address);
1400 }
1401 
1402 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1403 {
1404 	pmd_t *pmd = (pmd_t *) pud;
1405 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1406 		pmd = (pmd_t *) pud_deref(*pud);
1407 	return pmd + pmd_index(address);
1408 }
1409 
1410 #endif /* CONFIG_64BIT */
1411 
1412 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1413 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1414 #define pte_page(x) pfn_to_page(pte_pfn(x))
1415 
1416 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1417 
1418 /* Find an entry in the lowest level page table.. */
1419 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1420 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1421 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1422 #define pte_unmap(pte) do { } while (0)
1423 
1424 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1425 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1426 {
1427 	/*
1428 	 * pgprot is PAGE_NONE, PAGE_READ, or PAGE_WRITE (see __Pxxx / __Sxxx)
1429 	 * Convert to segment table entry format.
1430 	 */
1431 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1432 		return pgprot_val(SEGMENT_NONE);
1433 	if (pgprot_val(pgprot) == pgprot_val(PAGE_READ))
1434 		return pgprot_val(SEGMENT_READ);
1435 	return pgprot_val(SEGMENT_WRITE);
1436 }
1437 
1438 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1439 {
1440 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_WRITE;
1441 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1442 	return pmd;
1443 }
1444 
1445 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1446 {
1447 	pmd_val(pmd) |= _SEGMENT_ENTRY_WRITE;
1448 	if (pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1449 		return pmd;
1450 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1451 	return pmd;
1452 }
1453 
1454 static inline pmd_t pmd_mkclean(pmd_t pmd)
1455 {
1456 	if (pmd_large(pmd)) {
1457 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_DIRTY;
1458 		pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1459 	}
1460 	return pmd;
1461 }
1462 
1463 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1464 {
1465 	if (pmd_large(pmd)) {
1466 		pmd_val(pmd) |= _SEGMENT_ENTRY_DIRTY;
1467 		if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1468 			pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1469 	}
1470 	return pmd;
1471 }
1472 
1473 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1474 {
1475 	if (pmd_large(pmd)) {
1476 		pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1477 		if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1478 			pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1479 	}
1480 	return pmd;
1481 }
1482 
1483 static inline pmd_t pmd_mkold(pmd_t pmd)
1484 {
1485 	if (pmd_large(pmd)) {
1486 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1487 		pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1488 	}
1489 	return pmd;
1490 }
1491 
1492 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1493 {
1494 	if (pmd_large(pmd)) {
1495 		pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN_LARGE |
1496 			_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_YOUNG |
1497 			_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_SPLIT;
1498 		pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1499 		if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1500 			pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1501 		if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1502 			pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1503 		return pmd;
1504 	}
1505 	pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN;
1506 	pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1507 	return pmd;
1508 }
1509 
1510 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1511 {
1512 	pmd_t __pmd;
1513 	pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1514 	return __pmd;
1515 }
1516 
1517 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1518 
1519 static inline void __pmdp_csp(pmd_t *pmdp)
1520 {
1521 	register unsigned long reg2 asm("2") = pmd_val(*pmdp);
1522 	register unsigned long reg3 asm("3") = pmd_val(*pmdp) |
1523 					       _SEGMENT_ENTRY_INVALID;
1524 	register unsigned long reg4 asm("4") = ((unsigned long) pmdp) + 5;
1525 
1526 	asm volatile(
1527 		"	csp %1,%3"
1528 		: "=m" (*pmdp)
1529 		: "d" (reg2), "d" (reg3), "d" (reg4), "m" (*pmdp) : "cc");
1530 }
1531 
1532 static inline void __pmdp_idte(unsigned long address, pmd_t *pmdp)
1533 {
1534 	unsigned long sto;
1535 
1536 	sto = (unsigned long) pmdp - pmd_index(address) * sizeof(pmd_t);
1537 	asm volatile(
1538 		"	.insn	rrf,0xb98e0000,%2,%3,0,0"
1539 		: "=m" (*pmdp)
1540 		: "m" (*pmdp), "a" (sto), "a" ((address & HPAGE_MASK))
1541 		: "cc" );
1542 }
1543 
1544 static inline void __pmdp_idte_local(unsigned long address, pmd_t *pmdp)
1545 {
1546 	unsigned long sto;
1547 
1548 	sto = (unsigned long) pmdp - pmd_index(address) * sizeof(pmd_t);
1549 	asm volatile(
1550 		"	.insn	rrf,0xb98e0000,%2,%3,0,1"
1551 		: "=m" (*pmdp)
1552 		: "m" (*pmdp), "a" (sto), "a" ((address & HPAGE_MASK))
1553 		: "cc" );
1554 }
1555 
1556 static inline void pmdp_flush_direct(struct mm_struct *mm,
1557 				     unsigned long address, pmd_t *pmdp)
1558 {
1559 	int active, count;
1560 
1561 	if (pmd_val(*pmdp) & _SEGMENT_ENTRY_INVALID)
1562 		return;
1563 	if (!MACHINE_HAS_IDTE) {
1564 		__pmdp_csp(pmdp);
1565 		return;
1566 	}
1567 	active = (mm == current->active_mm) ? 1 : 0;
1568 	count = atomic_add_return(0x10000, &mm->context.attach_count);
1569 	if (MACHINE_HAS_TLB_LC && (count & 0xffff) <= active &&
1570 	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id())))
1571 		__pmdp_idte_local(address, pmdp);
1572 	else
1573 		__pmdp_idte(address, pmdp);
1574 	atomic_sub(0x10000, &mm->context.attach_count);
1575 }
1576 
1577 static inline void pmdp_flush_lazy(struct mm_struct *mm,
1578 				   unsigned long address, pmd_t *pmdp)
1579 {
1580 	int active, count;
1581 
1582 	if (pmd_val(*pmdp) & _SEGMENT_ENTRY_INVALID)
1583 		return;
1584 	active = (mm == current->active_mm) ? 1 : 0;
1585 	count = atomic_add_return(0x10000, &mm->context.attach_count);
1586 	if ((count & 0xffff) <= active) {
1587 		pmd_val(*pmdp) |= _SEGMENT_ENTRY_INVALID;
1588 		mm->context.flush_mm = 1;
1589 	} else if (MACHINE_HAS_IDTE)
1590 		__pmdp_idte(address, pmdp);
1591 	else
1592 		__pmdp_csp(pmdp);
1593 	atomic_sub(0x10000, &mm->context.attach_count);
1594 }
1595 
1596 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1597 
1598 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1599 extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1600 				       pgtable_t pgtable);
1601 
1602 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1603 extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1604 
1605 static inline int pmd_trans_splitting(pmd_t pmd)
1606 {
1607 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) &&
1608 		(pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT);
1609 }
1610 
1611 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1612 			      pmd_t *pmdp, pmd_t entry)
1613 {
1614 	*pmdp = entry;
1615 }
1616 
1617 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1618 {
1619 	pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1620 	pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1621 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1622 	return pmd;
1623 }
1624 
1625 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1626 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1627 					    unsigned long address, pmd_t *pmdp)
1628 {
1629 	pmd_t pmd;
1630 
1631 	pmd = *pmdp;
1632 	pmdp_flush_direct(vma->vm_mm, address, pmdp);
1633 	*pmdp = pmd_mkold(pmd);
1634 	return pmd_young(pmd);
1635 }
1636 
1637 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
1638 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
1639 				       unsigned long address, pmd_t *pmdp)
1640 {
1641 	pmd_t pmd = *pmdp;
1642 
1643 	pmdp_flush_direct(mm, address, pmdp);
1644 	pmd_clear(pmdp);
1645 	return pmd;
1646 }
1647 
1648 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR_FULL
1649 static inline pmd_t pmdp_get_and_clear_full(struct mm_struct *mm,
1650 					    unsigned long address,
1651 					    pmd_t *pmdp, int full)
1652 {
1653 	pmd_t pmd = *pmdp;
1654 
1655 	if (!full)
1656 		pmdp_flush_lazy(mm, address, pmdp);
1657 	pmd_clear(pmdp);
1658 	return pmd;
1659 }
1660 
1661 #define __HAVE_ARCH_PMDP_CLEAR_FLUSH
1662 static inline pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
1663 				     unsigned long address, pmd_t *pmdp)
1664 {
1665 	return pmdp_get_and_clear(vma->vm_mm, address, pmdp);
1666 }
1667 
1668 #define __HAVE_ARCH_PMDP_INVALIDATE
1669 static inline void pmdp_invalidate(struct vm_area_struct *vma,
1670 				   unsigned long address, pmd_t *pmdp)
1671 {
1672 	pmdp_flush_direct(vma->vm_mm, address, pmdp);
1673 }
1674 
1675 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1676 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1677 				      unsigned long address, pmd_t *pmdp)
1678 {
1679 	pmd_t pmd = *pmdp;
1680 
1681 	if (pmd_write(pmd)) {
1682 		pmdp_flush_direct(mm, address, pmdp);
1683 		set_pmd_at(mm, address, pmdp, pmd_wrprotect(pmd));
1684 	}
1685 }
1686 
1687 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1688 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1689 
1690 static inline int pmd_trans_huge(pmd_t pmd)
1691 {
1692 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1693 }
1694 
1695 static inline int has_transparent_hugepage(void)
1696 {
1697 	return MACHINE_HAS_HPAGE ? 1 : 0;
1698 }
1699 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1700 
1701 /*
1702  * 31 bit swap entry format:
1703  * A page-table entry has some bits we have to treat in a special way.
1704  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1705  * exception will occur instead of a page translation exception. The
1706  * specifiation exception has the bad habit not to store necessary
1707  * information in the lowcore.
1708  * Bits 21, 22, 30 and 31 are used to indicate the page type.
1709  * A swap pte is indicated by bit pattern (pte & 0x603) == 0x402
1710  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1711  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1712  * plus 24 for the offset.
1713  * 0|     offset        |0110|o|type |00|
1714  * 0 0000000001111111111 2222 2 22222 33
1715  * 0 1234567890123456789 0123 4 56789 01
1716  *
1717  * 64 bit swap entry format:
1718  * A page-table entry has some bits we have to treat in a special way.
1719  * Bits 52 and bit 55 have to be zero, otherwise an specification
1720  * exception will occur instead of a page translation exception. The
1721  * specifiation exception has the bad habit not to store necessary
1722  * information in the lowcore.
1723  * Bits 53, 54, 62 and 63 are used to indicate the page type.
1724  * A swap pte is indicated by bit pattern (pte & 0x603) == 0x402
1725  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1726  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1727  * plus 56 for the offset.
1728  * |                      offset                        |0110|o|type |00|
1729  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1730  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1731  */
1732 #ifndef CONFIG_64BIT
1733 #define __SWP_OFFSET_MASK (~0UL >> 12)
1734 #else
1735 #define __SWP_OFFSET_MASK (~0UL >> 11)
1736 #endif
1737 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1738 {
1739 	pte_t pte;
1740 	offset &= __SWP_OFFSET_MASK;
1741 	pte_val(pte) = _PAGE_INVALID | _PAGE_TYPE | ((type & 0x1f) << 2) |
1742 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1743 	return pte;
1744 }
1745 
1746 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1747 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1748 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1749 
1750 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1751 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1752 
1753 #endif /* !__ASSEMBLY__ */
1754 
1755 #define kern_addr_valid(addr)   (1)
1756 
1757 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1758 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1759 extern int s390_enable_sie(void);
1760 extern int s390_enable_skey(void);
1761 extern void s390_reset_cmma(struct mm_struct *mm);
1762 
1763 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1764 #define HAVE_ARCH_UNMAPPED_AREA
1765 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1766 
1767 /*
1768  * No page table caches to initialise
1769  */
1770 static inline void pgtable_cache_init(void) { }
1771 static inline void check_pgt_cache(void) { }
1772 
1773 #include <asm-generic/pgtable.h>
1774 
1775 #endif /* _S390_PAGE_H */
1776