xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision e2c75e76)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  *  S390 version
4  *    Copyright IBM Corp. 1999, 2000
5  *    Author(s): Hartmut Penner (hp@de.ibm.com)
6  *               Ulrich Weigand (weigand@de.ibm.com)
7  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
8  *
9  *  Derived from "include/asm-i386/pgtable.h"
10  */
11 
12 #ifndef _ASM_S390_PGTABLE_H
13 #define _ASM_S390_PGTABLE_H
14 
15 #include <linux/sched.h>
16 #include <linux/mm_types.h>
17 #include <linux/page-flags.h>
18 #include <linux/radix-tree.h>
19 #include <linux/atomic.h>
20 #include <asm/bug.h>
21 #include <asm/page.h>
22 
23 extern pgd_t swapper_pg_dir[];
24 extern void paging_init(void);
25 
26 enum {
27 	PG_DIRECT_MAP_4K = 0,
28 	PG_DIRECT_MAP_1M,
29 	PG_DIRECT_MAP_2G,
30 	PG_DIRECT_MAP_MAX
31 };
32 
33 extern atomic_long_t direct_pages_count[PG_DIRECT_MAP_MAX];
34 
35 static inline void update_page_count(int level, long count)
36 {
37 	if (IS_ENABLED(CONFIG_PROC_FS))
38 		atomic_long_add(count, &direct_pages_count[level]);
39 }
40 
41 struct seq_file;
42 void arch_report_meminfo(struct seq_file *m);
43 
44 /*
45  * The S390 doesn't have any external MMU info: the kernel page
46  * tables contain all the necessary information.
47  */
48 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
49 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
50 
51 /*
52  * ZERO_PAGE is a global shared page that is always zero; used
53  * for zero-mapped memory areas etc..
54  */
55 
56 extern unsigned long empty_zero_page;
57 extern unsigned long zero_page_mask;
58 
59 #define ZERO_PAGE(vaddr) \
60 	(virt_to_page((void *)(empty_zero_page + \
61 	 (((unsigned long)(vaddr)) &zero_page_mask))))
62 #define __HAVE_COLOR_ZERO_PAGE
63 
64 /* TODO: s390 cannot support io_remap_pfn_range... */
65 
66 #define FIRST_USER_ADDRESS  0UL
67 
68 #define pte_ERROR(e) \
69 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
70 #define pmd_ERROR(e) \
71 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
72 #define pud_ERROR(e) \
73 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
74 #define p4d_ERROR(e) \
75 	printk("%s:%d: bad p4d %p.\n", __FILE__, __LINE__, (void *) p4d_val(e))
76 #define pgd_ERROR(e) \
77 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
78 
79 /*
80  * The vmalloc and module area will always be on the topmost area of the
81  * kernel mapping. We reserve 128GB (64bit) for vmalloc and modules.
82  * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
83  * modules will reside. That makes sure that inter module branches always
84  * happen without trampolines and in addition the placement within a 2GB frame
85  * is branch prediction unit friendly.
86  */
87 extern unsigned long VMALLOC_START;
88 extern unsigned long VMALLOC_END;
89 extern struct page *vmemmap;
90 
91 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
92 
93 extern unsigned long MODULES_VADDR;
94 extern unsigned long MODULES_END;
95 #define MODULES_VADDR	MODULES_VADDR
96 #define MODULES_END	MODULES_END
97 #define MODULES_LEN	(1UL << 31)
98 
99 static inline int is_module_addr(void *addr)
100 {
101 	BUILD_BUG_ON(MODULES_LEN > (1UL << 31));
102 	if (addr < (void *)MODULES_VADDR)
103 		return 0;
104 	if (addr > (void *)MODULES_END)
105 		return 0;
106 	return 1;
107 }
108 
109 /*
110  * A 64 bit pagetable entry of S390 has following format:
111  * |			 PFRA			      |0IPC|  OS  |
112  * 0000000000111111111122222222223333333333444444444455555555556666
113  * 0123456789012345678901234567890123456789012345678901234567890123
114  *
115  * I Page-Invalid Bit:    Page is not available for address-translation
116  * P Page-Protection Bit: Store access not possible for page
117  * C Change-bit override: HW is not required to set change bit
118  *
119  * A 64 bit segmenttable entry of S390 has following format:
120  * |        P-table origin                              |      TT
121  * 0000000000111111111122222222223333333333444444444455555555556666
122  * 0123456789012345678901234567890123456789012345678901234567890123
123  *
124  * I Segment-Invalid Bit:    Segment is not available for address-translation
125  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
126  * P Page-Protection Bit: Store access not possible for page
127  * TT Type 00
128  *
129  * A 64 bit region table entry of S390 has following format:
130  * |        S-table origin                             |   TF  TTTL
131  * 0000000000111111111122222222223333333333444444444455555555556666
132  * 0123456789012345678901234567890123456789012345678901234567890123
133  *
134  * I Segment-Invalid Bit:    Segment is not available for address-translation
135  * TT Type 01
136  * TF
137  * TL Table length
138  *
139  * The 64 bit regiontable origin of S390 has following format:
140  * |      region table origon                          |       DTTL
141  * 0000000000111111111122222222223333333333444444444455555555556666
142  * 0123456789012345678901234567890123456789012345678901234567890123
143  *
144  * X Space-Switch event:
145  * G Segment-Invalid Bit:
146  * P Private-Space Bit:
147  * S Storage-Alteration:
148  * R Real space
149  * TL Table-Length:
150  *
151  * A storage key has the following format:
152  * | ACC |F|R|C|0|
153  *  0   3 4 5 6 7
154  * ACC: access key
155  * F  : fetch protection bit
156  * R  : referenced bit
157  * C  : changed bit
158  */
159 
160 /* Hardware bits in the page table entry */
161 #define _PAGE_NOEXEC	0x100		/* HW no-execute bit  */
162 #define _PAGE_PROTECT	0x200		/* HW read-only bit  */
163 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
164 #define _PAGE_LARGE	0x800		/* Bit to mark a large pte */
165 
166 /* Software bits in the page table entry */
167 #define _PAGE_PRESENT	0x001		/* SW pte present bit */
168 #define _PAGE_YOUNG	0x004		/* SW pte young bit */
169 #define _PAGE_DIRTY	0x008		/* SW pte dirty bit */
170 #define _PAGE_READ	0x010		/* SW pte read bit */
171 #define _PAGE_WRITE	0x020		/* SW pte write bit */
172 #define _PAGE_SPECIAL	0x040		/* SW associated with special page */
173 #define _PAGE_UNUSED	0x080		/* SW bit for pgste usage state */
174 #define __HAVE_ARCH_PTE_SPECIAL
175 
176 #ifdef CONFIG_MEM_SOFT_DIRTY
177 #define _PAGE_SOFT_DIRTY 0x002		/* SW pte soft dirty bit */
178 #else
179 #define _PAGE_SOFT_DIRTY 0x000
180 #endif
181 
182 /* Set of bits not changed in pte_modify */
183 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
184 				 _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
185 
186 /*
187  * handle_pte_fault uses pte_present and pte_none to find out the pte type
188  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
189  * distinguish present from not-present ptes. It is changed only with the page
190  * table lock held.
191  *
192  * The following table gives the different possible bit combinations for
193  * the pte hardware and software bits in the last 12 bits of a pte
194  * (. unassigned bit, x don't care, t swap type):
195  *
196  *				842100000000
197  *				000084210000
198  *				000000008421
199  *				.IR.uswrdy.p
200  * empty			.10.00000000
201  * swap				.11..ttttt.0
202  * prot-none, clean, old	.11.xx0000.1
203  * prot-none, clean, young	.11.xx0001.1
204  * prot-none, dirty, old	.11.xx0010.1
205  * prot-none, dirty, young	.11.xx0011.1
206  * read-only, clean, old	.11.xx0100.1
207  * read-only, clean, young	.01.xx0101.1
208  * read-only, dirty, old	.11.xx0110.1
209  * read-only, dirty, young	.01.xx0111.1
210  * read-write, clean, old	.11.xx1100.1
211  * read-write, clean, young	.01.xx1101.1
212  * read-write, dirty, old	.10.xx1110.1
213  * read-write, dirty, young	.00.xx1111.1
214  * HW-bits: R read-only, I invalid
215  * SW-bits: p present, y young, d dirty, r read, w write, s special,
216  *	    u unused, l large
217  *
218  * pte_none    is true for the bit pattern .10.00000000, pte == 0x400
219  * pte_swap    is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
220  * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
221  */
222 
223 /* Bits in the segment/region table address-space-control-element */
224 #define _ASCE_ORIGIN		~0xfffUL/* region/segment table origin	    */
225 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
226 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
227 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
228 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
229 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
230 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
231 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
232 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
233 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
234 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
235 
236 /* Bits in the region table entry */
237 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
238 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
239 #define _REGION_ENTRY_NOEXEC	0x100	/* region no-execute bit	    */
240 #define _REGION_ENTRY_OFFSET	0xc0	/* region table offset		    */
241 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
242 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
243 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
244 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
245 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
246 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
247 
248 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
249 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
250 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
251 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
252 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
253 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
254 
255 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address	     */
256 #define _REGION3_ENTRY_DIRTY	0x2000	/* SW region dirty bit */
257 #define _REGION3_ENTRY_YOUNG	0x1000	/* SW region young bit */
258 #define _REGION3_ENTRY_LARGE	0x0400	/* RTTE-format control, large page  */
259 #define _REGION3_ENTRY_READ	0x0002	/* SW region read bit */
260 #define _REGION3_ENTRY_WRITE	0x0001	/* SW region write bit */
261 
262 #ifdef CONFIG_MEM_SOFT_DIRTY
263 #define _REGION3_ENTRY_SOFT_DIRTY 0x4000 /* SW region soft dirty bit */
264 #else
265 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
266 #endif
267 
268 #define _REGION_ENTRY_BITS	 0xfffffffffffff22fUL
269 #define _REGION_ENTRY_BITS_LARGE 0xffffffff8000fe2fUL
270 
271 /* Bits in the segment table entry */
272 #define _SEGMENT_ENTRY_BITS	0xfffffffffffffe33UL
273 #define _SEGMENT_ENTRY_BITS_LARGE 0xfffffffffff0ff33UL
274 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
275 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* page table origin		    */
276 #define _SEGMENT_ENTRY_PROTECT	0x200	/* segment protection bit	    */
277 #define _SEGMENT_ENTRY_NOEXEC	0x100	/* segment no-execute bit	    */
278 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
279 
280 #define _SEGMENT_ENTRY		(0)
281 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
282 
283 #define _SEGMENT_ENTRY_DIRTY	0x2000	/* SW segment dirty bit */
284 #define _SEGMENT_ENTRY_YOUNG	0x1000	/* SW segment young bit */
285 #define _SEGMENT_ENTRY_LARGE	0x0400	/* STE-format control, large page */
286 #define _SEGMENT_ENTRY_WRITE	0x0002	/* SW segment write bit */
287 #define _SEGMENT_ENTRY_READ	0x0001	/* SW segment read bit */
288 
289 #ifdef CONFIG_MEM_SOFT_DIRTY
290 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x4000 /* SW segment soft dirty bit */
291 #else
292 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
293 #endif
294 
295 #define _CRST_ENTRIES	2048	/* number of region/segment table entries */
296 #define _PAGE_ENTRIES	256	/* number of page table entries	*/
297 
298 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
299 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
300 
301 #define _REGION1_SHIFT	53
302 #define _REGION2_SHIFT	42
303 #define _REGION3_SHIFT	31
304 #define _SEGMENT_SHIFT	20
305 
306 #define _REGION1_INDEX	(0x7ffUL << _REGION1_SHIFT)
307 #define _REGION2_INDEX	(0x7ffUL << _REGION2_SHIFT)
308 #define _REGION3_INDEX	(0x7ffUL << _REGION3_SHIFT)
309 #define _SEGMENT_INDEX	(0x7ffUL << _SEGMENT_SHIFT)
310 #define _PAGE_INDEX	(0xffUL  << _PAGE_SHIFT)
311 
312 #define _REGION1_SIZE	(1UL << _REGION1_SHIFT)
313 #define _REGION2_SIZE	(1UL << _REGION2_SHIFT)
314 #define _REGION3_SIZE	(1UL << _REGION3_SHIFT)
315 #define _SEGMENT_SIZE	(1UL << _SEGMENT_SHIFT)
316 
317 #define _REGION1_MASK	(~(_REGION1_SIZE - 1))
318 #define _REGION2_MASK	(~(_REGION2_SIZE - 1))
319 #define _REGION3_MASK	(~(_REGION3_SIZE - 1))
320 #define _SEGMENT_MASK	(~(_SEGMENT_SIZE - 1))
321 
322 #define PMD_SHIFT	_SEGMENT_SHIFT
323 #define PUD_SHIFT	_REGION3_SHIFT
324 #define P4D_SHIFT	_REGION2_SHIFT
325 #define PGDIR_SHIFT	_REGION1_SHIFT
326 
327 #define PMD_SIZE	_SEGMENT_SIZE
328 #define PUD_SIZE	_REGION3_SIZE
329 #define P4D_SIZE	_REGION2_SIZE
330 #define PGDIR_SIZE	_REGION1_SIZE
331 
332 #define PMD_MASK	_SEGMENT_MASK
333 #define PUD_MASK	_REGION3_MASK
334 #define P4D_MASK	_REGION2_MASK
335 #define PGDIR_MASK	_REGION1_MASK
336 
337 #define PTRS_PER_PTE	_PAGE_ENTRIES
338 #define PTRS_PER_PMD	_CRST_ENTRIES
339 #define PTRS_PER_PUD	_CRST_ENTRIES
340 #define PTRS_PER_P4D	_CRST_ENTRIES
341 #define PTRS_PER_PGD	_CRST_ENTRIES
342 
343 /*
344  * Segment table and region3 table entry encoding
345  * (R = read-only, I = invalid, y = young bit):
346  *				dy..R...I...wr
347  * prot-none, clean, old	00..1...1...00
348  * prot-none, clean, young	01..1...1...00
349  * prot-none, dirty, old	10..1...1...00
350  * prot-none, dirty, young	11..1...1...00
351  * read-only, clean, old	00..1...1...01
352  * read-only, clean, young	01..1...0...01
353  * read-only, dirty, old	10..1...1...01
354  * read-only, dirty, young	11..1...0...01
355  * read-write, clean, old	00..1...1...11
356  * read-write, clean, young	01..1...0...11
357  * read-write, dirty, old	10..0...1...11
358  * read-write, dirty, young	11..0...0...11
359  * The segment table origin is used to distinguish empty (origin==0) from
360  * read-write, old segment table entries (origin!=0)
361  * HW-bits: R read-only, I invalid
362  * SW-bits: y young, d dirty, r read, w write
363  */
364 
365 /* Page status table bits for virtualization */
366 #define PGSTE_ACC_BITS	0xf000000000000000UL
367 #define PGSTE_FP_BIT	0x0800000000000000UL
368 #define PGSTE_PCL_BIT	0x0080000000000000UL
369 #define PGSTE_HR_BIT	0x0040000000000000UL
370 #define PGSTE_HC_BIT	0x0020000000000000UL
371 #define PGSTE_GR_BIT	0x0004000000000000UL
372 #define PGSTE_GC_BIT	0x0002000000000000UL
373 #define PGSTE_UC_BIT	0x0000800000000000UL	/* user dirty (migration) */
374 #define PGSTE_IN_BIT	0x0000400000000000UL	/* IPTE notify bit */
375 #define PGSTE_VSIE_BIT	0x0000200000000000UL	/* ref'd in a shadow table */
376 
377 /* Guest Page State used for virtualization */
378 #define _PGSTE_GPS_ZERO			0x0000000080000000UL
379 #define _PGSTE_GPS_NODAT		0x0000000040000000UL
380 #define _PGSTE_GPS_USAGE_MASK		0x0000000003000000UL
381 #define _PGSTE_GPS_USAGE_STABLE		0x0000000000000000UL
382 #define _PGSTE_GPS_USAGE_UNUSED		0x0000000001000000UL
383 #define _PGSTE_GPS_USAGE_POT_VOLATILE	0x0000000002000000UL
384 #define _PGSTE_GPS_USAGE_VOLATILE	_PGSTE_GPS_USAGE_MASK
385 
386 /*
387  * A user page table pointer has the space-switch-event bit, the
388  * private-space-control bit and the storage-alteration-event-control
389  * bit set. A kernel page table pointer doesn't need them.
390  */
391 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
392 				 _ASCE_ALT_EVENT)
393 
394 /*
395  * Page protection definitions.
396  */
397 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID | _PAGE_PROTECT)
398 #define PAGE_RO		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
399 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
400 #define PAGE_RX		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
401 				 _PAGE_INVALID | _PAGE_PROTECT)
402 #define PAGE_RW		__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
403 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
404 #define PAGE_RWX	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
405 				 _PAGE_INVALID | _PAGE_PROTECT)
406 
407 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
408 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
409 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
410 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
411 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
412 				 _PAGE_PROTECT | _PAGE_NOEXEC)
413 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
414 				  _PAGE_YOUNG |	_PAGE_DIRTY)
415 
416 /*
417  * On s390 the page table entry has an invalid bit and a read-only bit.
418  * Read permission implies execute permission and write permission
419  * implies read permission.
420  */
421          /*xwr*/
422 #define __P000	PAGE_NONE
423 #define __P001	PAGE_RO
424 #define __P010	PAGE_RO
425 #define __P011	PAGE_RO
426 #define __P100	PAGE_RX
427 #define __P101	PAGE_RX
428 #define __P110	PAGE_RX
429 #define __P111	PAGE_RX
430 
431 #define __S000	PAGE_NONE
432 #define __S001	PAGE_RO
433 #define __S010	PAGE_RW
434 #define __S011	PAGE_RW
435 #define __S100	PAGE_RX
436 #define __S101	PAGE_RX
437 #define __S110	PAGE_RWX
438 #define __S111	PAGE_RWX
439 
440 /*
441  * Segment entry (large page) protection definitions.
442  */
443 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_INVALID | \
444 				 _SEGMENT_ENTRY_PROTECT)
445 #define SEGMENT_RO	__pgprot(_SEGMENT_ENTRY_PROTECT | \
446 				 _SEGMENT_ENTRY_READ | \
447 				 _SEGMENT_ENTRY_NOEXEC)
448 #define SEGMENT_RX	__pgprot(_SEGMENT_ENTRY_PROTECT | \
449 				 _SEGMENT_ENTRY_READ)
450 #define SEGMENT_RW	__pgprot(_SEGMENT_ENTRY_READ | \
451 				 _SEGMENT_ENTRY_WRITE | \
452 				 _SEGMENT_ENTRY_NOEXEC)
453 #define SEGMENT_RWX	__pgprot(_SEGMENT_ENTRY_READ | \
454 				 _SEGMENT_ENTRY_WRITE)
455 #define SEGMENT_KERNEL	__pgprot(_SEGMENT_ENTRY |	\
456 				 _SEGMENT_ENTRY_LARGE |	\
457 				 _SEGMENT_ENTRY_READ |	\
458 				 _SEGMENT_ENTRY_WRITE | \
459 				 _SEGMENT_ENTRY_YOUNG | \
460 				 _SEGMENT_ENTRY_DIRTY | \
461 				 _SEGMENT_ENTRY_NOEXEC)
462 #define SEGMENT_KERNEL_RO __pgprot(_SEGMENT_ENTRY |	\
463 				 _SEGMENT_ENTRY_LARGE |	\
464 				 _SEGMENT_ENTRY_READ |	\
465 				 _SEGMENT_ENTRY_YOUNG |	\
466 				 _SEGMENT_ENTRY_PROTECT | \
467 				 _SEGMENT_ENTRY_NOEXEC)
468 
469 /*
470  * Region3 entry (large page) protection definitions.
471  */
472 
473 #define REGION3_KERNEL	__pgprot(_REGION_ENTRY_TYPE_R3 | \
474 				 _REGION3_ENTRY_LARGE |	 \
475 				 _REGION3_ENTRY_READ |	 \
476 				 _REGION3_ENTRY_WRITE |	 \
477 				 _REGION3_ENTRY_YOUNG |	 \
478 				 _REGION3_ENTRY_DIRTY | \
479 				 _REGION_ENTRY_NOEXEC)
480 #define REGION3_KERNEL_RO __pgprot(_REGION_ENTRY_TYPE_R3 | \
481 				   _REGION3_ENTRY_LARGE |  \
482 				   _REGION3_ENTRY_READ |   \
483 				   _REGION3_ENTRY_YOUNG |  \
484 				   _REGION_ENTRY_PROTECT | \
485 				   _REGION_ENTRY_NOEXEC)
486 
487 static inline int mm_has_pgste(struct mm_struct *mm)
488 {
489 #ifdef CONFIG_PGSTE
490 	if (unlikely(mm->context.has_pgste))
491 		return 1;
492 #endif
493 	return 0;
494 }
495 
496 static inline int mm_alloc_pgste(struct mm_struct *mm)
497 {
498 #ifdef CONFIG_PGSTE
499 	if (unlikely(mm->context.alloc_pgste))
500 		return 1;
501 #endif
502 	return 0;
503 }
504 
505 /*
506  * In the case that a guest uses storage keys
507  * faults should no longer be backed by zero pages
508  */
509 #define mm_forbids_zeropage mm_has_pgste
510 static inline int mm_use_skey(struct mm_struct *mm)
511 {
512 #ifdef CONFIG_PGSTE
513 	if (mm->context.use_skey)
514 		return 1;
515 #endif
516 	return 0;
517 }
518 
519 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
520 {
521 	register unsigned long reg2 asm("2") = old;
522 	register unsigned long reg3 asm("3") = new;
523 	unsigned long address = (unsigned long)ptr | 1;
524 
525 	asm volatile(
526 		"	csp	%0,%3"
527 		: "+d" (reg2), "+m" (*ptr)
528 		: "d" (reg3), "d" (address)
529 		: "cc");
530 }
531 
532 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
533 {
534 	register unsigned long reg2 asm("2") = old;
535 	register unsigned long reg3 asm("3") = new;
536 	unsigned long address = (unsigned long)ptr | 1;
537 
538 	asm volatile(
539 		"	.insn	rre,0xb98a0000,%0,%3"
540 		: "+d" (reg2), "+m" (*ptr)
541 		: "d" (reg3), "d" (address)
542 		: "cc");
543 }
544 
545 #define CRDTE_DTT_PAGE		0x00UL
546 #define CRDTE_DTT_SEGMENT	0x10UL
547 #define CRDTE_DTT_REGION3	0x14UL
548 #define CRDTE_DTT_REGION2	0x18UL
549 #define CRDTE_DTT_REGION1	0x1cUL
550 
551 static inline void crdte(unsigned long old, unsigned long new,
552 			 unsigned long table, unsigned long dtt,
553 			 unsigned long address, unsigned long asce)
554 {
555 	register unsigned long reg2 asm("2") = old;
556 	register unsigned long reg3 asm("3") = new;
557 	register unsigned long reg4 asm("4") = table | dtt;
558 	register unsigned long reg5 asm("5") = address;
559 
560 	asm volatile(".insn rrf,0xb98f0000,%0,%2,%4,0"
561 		     : "+d" (reg2)
562 		     : "d" (reg3), "d" (reg4), "d" (reg5), "a" (asce)
563 		     : "memory", "cc");
564 }
565 
566 /*
567  * pgd/p4d/pud/pmd/pte query functions
568  */
569 static inline int pgd_folded(pgd_t pgd)
570 {
571 	return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
572 }
573 
574 static inline int pgd_present(pgd_t pgd)
575 {
576 	if (pgd_folded(pgd))
577 		return 1;
578 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
579 }
580 
581 static inline int pgd_none(pgd_t pgd)
582 {
583 	if (pgd_folded(pgd))
584 		return 0;
585 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
586 }
587 
588 static inline int pgd_bad(pgd_t pgd)
589 {
590 	/*
591 	 * With dynamic page table levels the pgd can be a region table
592 	 * entry or a segment table entry. Check for the bit that are
593 	 * invalid for either table entry.
594 	 */
595 	unsigned long mask =
596 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
597 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
598 	return (pgd_val(pgd) & mask) != 0;
599 }
600 
601 static inline int p4d_folded(p4d_t p4d)
602 {
603 	return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
604 }
605 
606 static inline int p4d_present(p4d_t p4d)
607 {
608 	if (p4d_folded(p4d))
609 		return 1;
610 	return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
611 }
612 
613 static inline int p4d_none(p4d_t p4d)
614 {
615 	if (p4d_folded(p4d))
616 		return 0;
617 	return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
618 }
619 
620 static inline unsigned long p4d_pfn(p4d_t p4d)
621 {
622 	unsigned long origin_mask;
623 
624 	origin_mask = _REGION_ENTRY_ORIGIN;
625 	return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
626 }
627 
628 static inline int pud_folded(pud_t pud)
629 {
630 	return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
631 }
632 
633 static inline int pud_present(pud_t pud)
634 {
635 	if (pud_folded(pud))
636 		return 1;
637 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
638 }
639 
640 static inline int pud_none(pud_t pud)
641 {
642 	if (pud_folded(pud))
643 		return 0;
644 	return pud_val(pud) == _REGION3_ENTRY_EMPTY;
645 }
646 
647 static inline int pud_large(pud_t pud)
648 {
649 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
650 		return 0;
651 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
652 }
653 
654 static inline unsigned long pud_pfn(pud_t pud)
655 {
656 	unsigned long origin_mask;
657 
658 	origin_mask = _REGION_ENTRY_ORIGIN;
659 	if (pud_large(pud))
660 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
661 	return (pud_val(pud) & origin_mask) >> PAGE_SHIFT;
662 }
663 
664 static inline int pmd_large(pmd_t pmd)
665 {
666 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
667 }
668 
669 static inline int pmd_bad(pmd_t pmd)
670 {
671 	if (pmd_large(pmd))
672 		return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
673 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
674 }
675 
676 static inline int pud_bad(pud_t pud)
677 {
678 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
679 		return pmd_bad(__pmd(pud_val(pud)));
680 	if (pud_large(pud))
681 		return (pud_val(pud) & ~_REGION_ENTRY_BITS_LARGE) != 0;
682 	return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
683 }
684 
685 static inline int p4d_bad(p4d_t p4d)
686 {
687 	if ((p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
688 		return pud_bad(__pud(p4d_val(p4d)));
689 	return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
690 }
691 
692 static inline int pmd_present(pmd_t pmd)
693 {
694 	return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
695 }
696 
697 static inline int pmd_none(pmd_t pmd)
698 {
699 	return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
700 }
701 
702 static inline unsigned long pmd_pfn(pmd_t pmd)
703 {
704 	unsigned long origin_mask;
705 
706 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
707 	if (pmd_large(pmd))
708 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
709 	return (pmd_val(pmd) & origin_mask) >> PAGE_SHIFT;
710 }
711 
712 #define pmd_write pmd_write
713 static inline int pmd_write(pmd_t pmd)
714 {
715 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
716 }
717 
718 static inline int pmd_dirty(pmd_t pmd)
719 {
720 	int dirty = 1;
721 	if (pmd_large(pmd))
722 		dirty = (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
723 	return dirty;
724 }
725 
726 static inline int pmd_young(pmd_t pmd)
727 {
728 	int young = 1;
729 	if (pmd_large(pmd))
730 		young = (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
731 	return young;
732 }
733 
734 static inline int pte_present(pte_t pte)
735 {
736 	/* Bit pattern: (pte & 0x001) == 0x001 */
737 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
738 }
739 
740 static inline int pte_none(pte_t pte)
741 {
742 	/* Bit pattern: pte == 0x400 */
743 	return pte_val(pte) == _PAGE_INVALID;
744 }
745 
746 static inline int pte_swap(pte_t pte)
747 {
748 	/* Bit pattern: (pte & 0x201) == 0x200 */
749 	return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
750 		== _PAGE_PROTECT;
751 }
752 
753 static inline int pte_special(pte_t pte)
754 {
755 	return (pte_val(pte) & _PAGE_SPECIAL);
756 }
757 
758 #define __HAVE_ARCH_PTE_SAME
759 static inline int pte_same(pte_t a, pte_t b)
760 {
761 	return pte_val(a) == pte_val(b);
762 }
763 
764 #ifdef CONFIG_NUMA_BALANCING
765 static inline int pte_protnone(pte_t pte)
766 {
767 	return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
768 }
769 
770 static inline int pmd_protnone(pmd_t pmd)
771 {
772 	/* pmd_large(pmd) implies pmd_present(pmd) */
773 	return pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
774 }
775 #endif
776 
777 static inline int pte_soft_dirty(pte_t pte)
778 {
779 	return pte_val(pte) & _PAGE_SOFT_DIRTY;
780 }
781 #define pte_swp_soft_dirty pte_soft_dirty
782 
783 static inline pte_t pte_mksoft_dirty(pte_t pte)
784 {
785 	pte_val(pte) |= _PAGE_SOFT_DIRTY;
786 	return pte;
787 }
788 #define pte_swp_mksoft_dirty pte_mksoft_dirty
789 
790 static inline pte_t pte_clear_soft_dirty(pte_t pte)
791 {
792 	pte_val(pte) &= ~_PAGE_SOFT_DIRTY;
793 	return pte;
794 }
795 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
796 
797 static inline int pmd_soft_dirty(pmd_t pmd)
798 {
799 	return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
800 }
801 
802 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
803 {
804 	pmd_val(pmd) |= _SEGMENT_ENTRY_SOFT_DIRTY;
805 	return pmd;
806 }
807 
808 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
809 {
810 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_SOFT_DIRTY;
811 	return pmd;
812 }
813 
814 /*
815  * query functions pte_write/pte_dirty/pte_young only work if
816  * pte_present() is true. Undefined behaviour if not..
817  */
818 static inline int pte_write(pte_t pte)
819 {
820 	return (pte_val(pte) & _PAGE_WRITE) != 0;
821 }
822 
823 static inline int pte_dirty(pte_t pte)
824 {
825 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
826 }
827 
828 static inline int pte_young(pte_t pte)
829 {
830 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
831 }
832 
833 #define __HAVE_ARCH_PTE_UNUSED
834 static inline int pte_unused(pte_t pte)
835 {
836 	return pte_val(pte) & _PAGE_UNUSED;
837 }
838 
839 /*
840  * pgd/pmd/pte modification functions
841  */
842 
843 static inline void pgd_clear(pgd_t *pgd)
844 {
845 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
846 		pgd_val(*pgd) = _REGION1_ENTRY_EMPTY;
847 }
848 
849 static inline void p4d_clear(p4d_t *p4d)
850 {
851 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
852 		p4d_val(*p4d) = _REGION2_ENTRY_EMPTY;
853 }
854 
855 static inline void pud_clear(pud_t *pud)
856 {
857 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
858 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
859 }
860 
861 static inline void pmd_clear(pmd_t *pmdp)
862 {
863 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
864 }
865 
866 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
867 {
868 	pte_val(*ptep) = _PAGE_INVALID;
869 }
870 
871 /*
872  * The following pte modification functions only work if
873  * pte_present() is true. Undefined behaviour if not..
874  */
875 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
876 {
877 	pte_val(pte) &= _PAGE_CHG_MASK;
878 	pte_val(pte) |= pgprot_val(newprot);
879 	/*
880 	 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
881 	 * has the invalid bit set, clear it again for readable, young pages
882 	 */
883 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
884 		pte_val(pte) &= ~_PAGE_INVALID;
885 	/*
886 	 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
887 	 * protection bit set, clear it again for writable, dirty pages
888 	 */
889 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
890 		pte_val(pte) &= ~_PAGE_PROTECT;
891 	return pte;
892 }
893 
894 static inline pte_t pte_wrprotect(pte_t pte)
895 {
896 	pte_val(pte) &= ~_PAGE_WRITE;
897 	pte_val(pte) |= _PAGE_PROTECT;
898 	return pte;
899 }
900 
901 static inline pte_t pte_mkwrite(pte_t pte)
902 {
903 	pte_val(pte) |= _PAGE_WRITE;
904 	if (pte_val(pte) & _PAGE_DIRTY)
905 		pte_val(pte) &= ~_PAGE_PROTECT;
906 	return pte;
907 }
908 
909 static inline pte_t pte_mkclean(pte_t pte)
910 {
911 	pte_val(pte) &= ~_PAGE_DIRTY;
912 	pte_val(pte) |= _PAGE_PROTECT;
913 	return pte;
914 }
915 
916 static inline pte_t pte_mkdirty(pte_t pte)
917 {
918 	pte_val(pte) |= _PAGE_DIRTY | _PAGE_SOFT_DIRTY;
919 	if (pte_val(pte) & _PAGE_WRITE)
920 		pte_val(pte) &= ~_PAGE_PROTECT;
921 	return pte;
922 }
923 
924 static inline pte_t pte_mkold(pte_t pte)
925 {
926 	pte_val(pte) &= ~_PAGE_YOUNG;
927 	pte_val(pte) |= _PAGE_INVALID;
928 	return pte;
929 }
930 
931 static inline pte_t pte_mkyoung(pte_t pte)
932 {
933 	pte_val(pte) |= _PAGE_YOUNG;
934 	if (pte_val(pte) & _PAGE_READ)
935 		pte_val(pte) &= ~_PAGE_INVALID;
936 	return pte;
937 }
938 
939 static inline pte_t pte_mkspecial(pte_t pte)
940 {
941 	pte_val(pte) |= _PAGE_SPECIAL;
942 	return pte;
943 }
944 
945 #ifdef CONFIG_HUGETLB_PAGE
946 static inline pte_t pte_mkhuge(pte_t pte)
947 {
948 	pte_val(pte) |= _PAGE_LARGE;
949 	return pte;
950 }
951 #endif
952 
953 #define IPTE_GLOBAL	0
954 #define	IPTE_LOCAL	1
955 
956 #define IPTE_NODAT	0x400
957 #define IPTE_GUEST_ASCE	0x800
958 
959 static inline void __ptep_ipte(unsigned long address, pte_t *ptep,
960 			       unsigned long opt, unsigned long asce,
961 			       int local)
962 {
963 	unsigned long pto = (unsigned long) ptep;
964 
965 	if (__builtin_constant_p(opt) && opt == 0) {
966 		/* Invalidation + TLB flush for the pte */
967 		asm volatile(
968 			"	.insn	rrf,0xb2210000,%[r1],%[r2],0,%[m4]"
969 			: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
970 			  [m4] "i" (local));
971 		return;
972 	}
973 
974 	/* Invalidate ptes with options + TLB flush of the ptes */
975 	opt = opt | (asce & _ASCE_ORIGIN);
976 	asm volatile(
977 		"	.insn	rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
978 		: [r2] "+a" (address), [r3] "+a" (opt)
979 		: [r1] "a" (pto), [m4] "i" (local) : "memory");
980 }
981 
982 static inline void __ptep_ipte_range(unsigned long address, int nr,
983 				     pte_t *ptep, int local)
984 {
985 	unsigned long pto = (unsigned long) ptep;
986 
987 	/* Invalidate a range of ptes + TLB flush of the ptes */
988 	do {
989 		asm volatile(
990 			"       .insn rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
991 			: [r2] "+a" (address), [r3] "+a" (nr)
992 			: [r1] "a" (pto), [m4] "i" (local) : "memory");
993 	} while (nr != 255);
994 }
995 
996 /*
997  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
998  * both clear the TLB for the unmapped pte. The reason is that
999  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1000  * to modify an active pte. The sequence is
1001  *   1) ptep_get_and_clear
1002  *   2) set_pte_at
1003  *   3) flush_tlb_range
1004  * On s390 the tlb needs to get flushed with the modification of the pte
1005  * if the pte is active. The only way how this can be implemented is to
1006  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1007  * is a nop.
1008  */
1009 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1010 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1011 
1012 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1013 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1014 					    unsigned long addr, pte_t *ptep)
1015 {
1016 	pte_t pte = *ptep;
1017 
1018 	pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1019 	return pte_young(pte);
1020 }
1021 
1022 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1023 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1024 					 unsigned long address, pte_t *ptep)
1025 {
1026 	return ptep_test_and_clear_young(vma, address, ptep);
1027 }
1028 
1029 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1030 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1031 				       unsigned long addr, pte_t *ptep)
1032 {
1033 	return ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1034 }
1035 
1036 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1037 pte_t ptep_modify_prot_start(struct mm_struct *, unsigned long, pte_t *);
1038 void ptep_modify_prot_commit(struct mm_struct *, unsigned long, pte_t *, pte_t);
1039 
1040 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1041 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1042 				     unsigned long addr, pte_t *ptep)
1043 {
1044 	return ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1045 }
1046 
1047 /*
1048  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1049  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1050  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1051  * cannot be accessed while the batched unmap is running. In this case
1052  * full==1 and a simple pte_clear is enough. See tlb.h.
1053  */
1054 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1055 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1056 					    unsigned long addr,
1057 					    pte_t *ptep, int full)
1058 {
1059 	if (full) {
1060 		pte_t pte = *ptep;
1061 		*ptep = __pte(_PAGE_INVALID);
1062 		return pte;
1063 	}
1064 	return ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1065 }
1066 
1067 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1068 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1069 				      unsigned long addr, pte_t *ptep)
1070 {
1071 	pte_t pte = *ptep;
1072 
1073 	if (pte_write(pte))
1074 		ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1075 }
1076 
1077 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1078 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1079 					unsigned long addr, pte_t *ptep,
1080 					pte_t entry, int dirty)
1081 {
1082 	if (pte_same(*ptep, entry))
1083 		return 0;
1084 	ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1085 	return 1;
1086 }
1087 
1088 /*
1089  * Additional functions to handle KVM guest page tables
1090  */
1091 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1092 		     pte_t *ptep, pte_t entry);
1093 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1094 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1095 		 pte_t *ptep, unsigned long bits);
1096 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1097 		    pte_t *ptep, int prot, unsigned long bit);
1098 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1099 		     pte_t *ptep , int reset);
1100 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1101 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1102 		    pte_t *sptep, pte_t *tptep, pte_t pte);
1103 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1104 
1105 bool test_and_clear_guest_dirty(struct mm_struct *mm, unsigned long address);
1106 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1107 			  unsigned char key, bool nq);
1108 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1109 			       unsigned char key, unsigned char *oldkey,
1110 			       bool nq, bool mr, bool mc);
1111 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1112 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1113 			  unsigned char *key);
1114 
1115 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1116 				unsigned long bits, unsigned long value);
1117 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1118 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1119 			unsigned long *oldpte, unsigned long *oldpgste);
1120 
1121 /*
1122  * Certain architectures need to do special things when PTEs
1123  * within a page table are directly modified.  Thus, the following
1124  * hook is made available.
1125  */
1126 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
1127 			      pte_t *ptep, pte_t entry)
1128 {
1129 	if (!MACHINE_HAS_NX)
1130 		pte_val(entry) &= ~_PAGE_NOEXEC;
1131 	if (pte_present(entry))
1132 		pte_val(entry) &= ~_PAGE_UNUSED;
1133 	if (mm_has_pgste(mm))
1134 		ptep_set_pte_at(mm, addr, ptep, entry);
1135 	else
1136 		*ptep = entry;
1137 }
1138 
1139 /*
1140  * Conversion functions: convert a page and protection to a page entry,
1141  * and a page entry and page directory to the page they refer to.
1142  */
1143 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1144 {
1145 	pte_t __pte;
1146 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1147 	return pte_mkyoung(__pte);
1148 }
1149 
1150 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1151 {
1152 	unsigned long physpage = page_to_phys(page);
1153 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1154 
1155 	if (pte_write(__pte) && PageDirty(page))
1156 		__pte = pte_mkdirty(__pte);
1157 	return __pte;
1158 }
1159 
1160 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1161 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1162 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1163 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1164 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1165 
1166 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1167 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1168 
1169 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1170 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1171 #define p4d_deref(pud) (p4d_val(pud) & _REGION_ENTRY_ORIGIN)
1172 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1173 
1174 static inline p4d_t *p4d_offset(pgd_t *pgd, unsigned long address)
1175 {
1176 	p4d_t *p4d = (p4d_t *) pgd;
1177 
1178 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
1179 		p4d = (p4d_t *) pgd_deref(*pgd);
1180 	return p4d + p4d_index(address);
1181 }
1182 
1183 static inline pud_t *pud_offset(p4d_t *p4d, unsigned long address)
1184 {
1185 	pud_t *pud = (pud_t *) p4d;
1186 
1187 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1188 		pud = (pud_t *) p4d_deref(*p4d);
1189 	return pud + pud_index(address);
1190 }
1191 
1192 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1193 {
1194 	pmd_t *pmd = (pmd_t *) pud;
1195 
1196 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1197 		pmd = (pmd_t *) pud_deref(*pud);
1198 	return pmd + pmd_index(address);
1199 }
1200 
1201 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1202 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1203 #define pte_page(x) pfn_to_page(pte_pfn(x))
1204 
1205 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1206 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1207 #define p4d_page(pud) pfn_to_page(p4d_pfn(p4d))
1208 
1209 /* Find an entry in the lowest level page table.. */
1210 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1211 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1212 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1213 #define pte_unmap(pte) do { } while (0)
1214 
1215 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1216 {
1217 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_WRITE;
1218 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1219 	return pmd;
1220 }
1221 
1222 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1223 {
1224 	pmd_val(pmd) |= _SEGMENT_ENTRY_WRITE;
1225 	if (pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1226 		return pmd;
1227 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1228 	return pmd;
1229 }
1230 
1231 static inline pmd_t pmd_mkclean(pmd_t pmd)
1232 {
1233 	if (pmd_large(pmd)) {
1234 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_DIRTY;
1235 		pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1236 	}
1237 	return pmd;
1238 }
1239 
1240 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1241 {
1242 	if (pmd_large(pmd)) {
1243 		pmd_val(pmd) |= _SEGMENT_ENTRY_DIRTY |
1244 				_SEGMENT_ENTRY_SOFT_DIRTY;
1245 		if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1246 			pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1247 	}
1248 	return pmd;
1249 }
1250 
1251 static inline pud_t pud_wrprotect(pud_t pud)
1252 {
1253 	pud_val(pud) &= ~_REGION3_ENTRY_WRITE;
1254 	pud_val(pud) |= _REGION_ENTRY_PROTECT;
1255 	return pud;
1256 }
1257 
1258 static inline pud_t pud_mkwrite(pud_t pud)
1259 {
1260 	pud_val(pud) |= _REGION3_ENTRY_WRITE;
1261 	if (pud_large(pud) && !(pud_val(pud) & _REGION3_ENTRY_DIRTY))
1262 		return pud;
1263 	pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1264 	return pud;
1265 }
1266 
1267 static inline pud_t pud_mkclean(pud_t pud)
1268 {
1269 	if (pud_large(pud)) {
1270 		pud_val(pud) &= ~_REGION3_ENTRY_DIRTY;
1271 		pud_val(pud) |= _REGION_ENTRY_PROTECT;
1272 	}
1273 	return pud;
1274 }
1275 
1276 static inline pud_t pud_mkdirty(pud_t pud)
1277 {
1278 	if (pud_large(pud)) {
1279 		pud_val(pud) |= _REGION3_ENTRY_DIRTY |
1280 				_REGION3_ENTRY_SOFT_DIRTY;
1281 		if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1282 			pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1283 	}
1284 	return pud;
1285 }
1286 
1287 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1288 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1289 {
1290 	/*
1291 	 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1292 	 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1293 	 */
1294 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1295 		return pgprot_val(SEGMENT_NONE);
1296 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1297 		return pgprot_val(SEGMENT_RO);
1298 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1299 		return pgprot_val(SEGMENT_RX);
1300 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1301 		return pgprot_val(SEGMENT_RW);
1302 	return pgprot_val(SEGMENT_RWX);
1303 }
1304 
1305 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1306 {
1307 	if (pmd_large(pmd)) {
1308 		pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1309 		if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1310 			pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1311 	}
1312 	return pmd;
1313 }
1314 
1315 static inline pmd_t pmd_mkold(pmd_t pmd)
1316 {
1317 	if (pmd_large(pmd)) {
1318 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1319 		pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1320 	}
1321 	return pmd;
1322 }
1323 
1324 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1325 {
1326 	if (pmd_large(pmd)) {
1327 		pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN_LARGE |
1328 			_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_YOUNG |
1329 			_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_SOFT_DIRTY;
1330 		pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1331 		if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1332 			pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1333 		if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1334 			pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1335 		return pmd;
1336 	}
1337 	pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN;
1338 	pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1339 	return pmd;
1340 }
1341 
1342 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1343 {
1344 	pmd_t __pmd;
1345 	pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1346 	return __pmd;
1347 }
1348 
1349 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1350 
1351 static inline void __pmdp_csp(pmd_t *pmdp)
1352 {
1353 	csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1354 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1355 }
1356 
1357 #define IDTE_GLOBAL	0
1358 #define IDTE_LOCAL	1
1359 
1360 #define IDTE_PTOA	0x0800
1361 #define IDTE_NODAT	0x1000
1362 #define IDTE_GUEST_ASCE	0x2000
1363 
1364 static inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1365 			       unsigned long opt, unsigned long asce,
1366 			       int local)
1367 {
1368 	unsigned long sto;
1369 
1370 	sto = (unsigned long) pmdp - pmd_index(addr) * sizeof(pmd_t);
1371 	if (__builtin_constant_p(opt) && opt == 0) {
1372 		/* flush without guest asce */
1373 		asm volatile(
1374 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1375 			: "+m" (*pmdp)
1376 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1377 			  [m4] "i" (local)
1378 			: "cc" );
1379 	} else {
1380 		/* flush with guest asce */
1381 		asm volatile(
1382 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1383 			: "+m" (*pmdp)
1384 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1385 			  [r3] "a" (asce), [m4] "i" (local)
1386 			: "cc" );
1387 	}
1388 }
1389 
1390 static inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1391 			       unsigned long opt, unsigned long asce,
1392 			       int local)
1393 {
1394 	unsigned long r3o;
1395 
1396 	r3o = (unsigned long) pudp - pud_index(addr) * sizeof(pud_t);
1397 	r3o |= _ASCE_TYPE_REGION3;
1398 	if (__builtin_constant_p(opt) && opt == 0) {
1399 		/* flush without guest asce */
1400 		asm volatile(
1401 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1402 			: "+m" (*pudp)
1403 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1404 			  [m4] "i" (local)
1405 			: "cc");
1406 	} else {
1407 		/* flush with guest asce */
1408 		asm volatile(
1409 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1410 			: "+m" (*pudp)
1411 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1412 			  [r3] "a" (asce), [m4] "i" (local)
1413 			: "cc" );
1414 	}
1415 }
1416 
1417 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1418 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1419 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1420 
1421 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1422 
1423 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1424 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1425 				pgtable_t pgtable);
1426 
1427 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1428 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1429 
1430 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1431 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1432 					unsigned long addr, pmd_t *pmdp,
1433 					pmd_t entry, int dirty)
1434 {
1435 	VM_BUG_ON(addr & ~HPAGE_MASK);
1436 
1437 	entry = pmd_mkyoung(entry);
1438 	if (dirty)
1439 		entry = pmd_mkdirty(entry);
1440 	if (pmd_val(*pmdp) == pmd_val(entry))
1441 		return 0;
1442 	pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1443 	return 1;
1444 }
1445 
1446 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1447 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1448 					    unsigned long addr, pmd_t *pmdp)
1449 {
1450 	pmd_t pmd = *pmdp;
1451 
1452 	pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1453 	return pmd_young(pmd);
1454 }
1455 
1456 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1457 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1458 					 unsigned long addr, pmd_t *pmdp)
1459 {
1460 	VM_BUG_ON(addr & ~HPAGE_MASK);
1461 	return pmdp_test_and_clear_young(vma, addr, pmdp);
1462 }
1463 
1464 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1465 			      pmd_t *pmdp, pmd_t entry)
1466 {
1467 	if (!MACHINE_HAS_NX)
1468 		pmd_val(entry) &= ~_SEGMENT_ENTRY_NOEXEC;
1469 	*pmdp = entry;
1470 }
1471 
1472 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1473 {
1474 	pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1475 	pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1476 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1477 	return pmd;
1478 }
1479 
1480 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1481 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1482 					    unsigned long addr, pmd_t *pmdp)
1483 {
1484 	return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1485 }
1486 
1487 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1488 static inline pmd_t pmdp_huge_get_and_clear_full(struct mm_struct *mm,
1489 						 unsigned long addr,
1490 						 pmd_t *pmdp, int full)
1491 {
1492 	if (full) {
1493 		pmd_t pmd = *pmdp;
1494 		*pmdp = __pmd(_SEGMENT_ENTRY_EMPTY);
1495 		return pmd;
1496 	}
1497 	return pmdp_xchg_lazy(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1498 }
1499 
1500 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1501 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1502 					  unsigned long addr, pmd_t *pmdp)
1503 {
1504 	return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1505 }
1506 
1507 #define __HAVE_ARCH_PMDP_INVALIDATE
1508 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1509 				   unsigned long addr, pmd_t *pmdp)
1510 {
1511 	pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1512 
1513 	return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1514 }
1515 
1516 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1517 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1518 				      unsigned long addr, pmd_t *pmdp)
1519 {
1520 	pmd_t pmd = *pmdp;
1521 
1522 	if (pmd_write(pmd))
1523 		pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1524 }
1525 
1526 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1527 					unsigned long address,
1528 					pmd_t *pmdp)
1529 {
1530 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1531 }
1532 #define pmdp_collapse_flush pmdp_collapse_flush
1533 
1534 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1535 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1536 
1537 static inline int pmd_trans_huge(pmd_t pmd)
1538 {
1539 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1540 }
1541 
1542 #define has_transparent_hugepage has_transparent_hugepage
1543 static inline int has_transparent_hugepage(void)
1544 {
1545 	return MACHINE_HAS_EDAT1 ? 1 : 0;
1546 }
1547 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1548 
1549 /*
1550  * 64 bit swap entry format:
1551  * A page-table entry has some bits we have to treat in a special way.
1552  * Bits 52 and bit 55 have to be zero, otherwise a specification
1553  * exception will occur instead of a page translation exception. The
1554  * specification exception has the bad habit not to store necessary
1555  * information in the lowcore.
1556  * Bits 54 and 63 are used to indicate the page type.
1557  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1558  * This leaves the bits 0-51 and bits 56-62 to store type and offset.
1559  * We use the 5 bits from 57-61 for the type and the 52 bits from 0-51
1560  * for the offset.
1561  * |			  offset			|01100|type |00|
1562  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1563  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1564  */
1565 
1566 #define __SWP_OFFSET_MASK	((1UL << 52) - 1)
1567 #define __SWP_OFFSET_SHIFT	12
1568 #define __SWP_TYPE_MASK		((1UL << 5) - 1)
1569 #define __SWP_TYPE_SHIFT	2
1570 
1571 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1572 {
1573 	pte_t pte;
1574 
1575 	pte_val(pte) = _PAGE_INVALID | _PAGE_PROTECT;
1576 	pte_val(pte) |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1577 	pte_val(pte) |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1578 	return pte;
1579 }
1580 
1581 static inline unsigned long __swp_type(swp_entry_t entry)
1582 {
1583 	return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1584 }
1585 
1586 static inline unsigned long __swp_offset(swp_entry_t entry)
1587 {
1588 	return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1589 }
1590 
1591 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1592 {
1593 	return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1594 }
1595 
1596 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1597 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1598 
1599 #define kern_addr_valid(addr)   (1)
1600 
1601 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1602 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1603 extern int s390_enable_sie(void);
1604 extern int s390_enable_skey(void);
1605 extern void s390_reset_cmma(struct mm_struct *mm);
1606 
1607 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1608 #define HAVE_ARCH_UNMAPPED_AREA
1609 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1610 
1611 /*
1612  * No page table caches to initialise
1613  */
1614 static inline void pgtable_cache_init(void) { }
1615 static inline void check_pgt_cache(void) { }
1616 
1617 #include <asm-generic/pgtable.h>
1618 
1619 #endif /* _S390_PAGE_H */
1620