xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision 046b212a)
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/sections.h>
21 #include <asm/bug.h>
22 #include <asm/page.h>
23 #include <asm/uv.h>
24 
25 extern pgd_t swapper_pg_dir[];
26 extern pgd_t invalid_pg_dir[];
27 extern void paging_init(void);
28 extern unsigned long s390_invalid_asce;
29 
30 enum {
31 	PG_DIRECT_MAP_4K = 0,
32 	PG_DIRECT_MAP_1M,
33 	PG_DIRECT_MAP_2G,
34 	PG_DIRECT_MAP_MAX
35 };
36 
37 extern atomic_long_t __bootdata_preserved(direct_pages_count[PG_DIRECT_MAP_MAX]);
38 
39 static inline void update_page_count(int level, long count)
40 {
41 	if (IS_ENABLED(CONFIG_PROC_FS))
42 		atomic_long_add(count, &direct_pages_count[level]);
43 }
44 
45 /*
46  * The S390 doesn't have any external MMU info: the kernel page
47  * tables contain all the necessary information.
48  */
49 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
50 #define update_mmu_cache_range(vmf, vma, addr, ptep, nr) do { } while (0)
51 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
52 
53 /*
54  * ZERO_PAGE is a global shared page that is always zero; used
55  * for zero-mapped memory areas etc..
56  */
57 
58 extern unsigned long empty_zero_page;
59 extern unsigned long zero_page_mask;
60 
61 #define ZERO_PAGE(vaddr) \
62 	(virt_to_page((void *)(empty_zero_page + \
63 	 (((unsigned long)(vaddr)) &zero_page_mask))))
64 #define __HAVE_COLOR_ZERO_PAGE
65 
66 /* TODO: s390 cannot support io_remap_pfn_range... */
67 
68 #define pte_ERROR(e) \
69 	pr_err("%s:%d: bad pte %016lx.\n", __FILE__, __LINE__, pte_val(e))
70 #define pmd_ERROR(e) \
71 	pr_err("%s:%d: bad pmd %016lx.\n", __FILE__, __LINE__, pmd_val(e))
72 #define pud_ERROR(e) \
73 	pr_err("%s:%d: bad pud %016lx.\n", __FILE__, __LINE__, pud_val(e))
74 #define p4d_ERROR(e) \
75 	pr_err("%s:%d: bad p4d %016lx.\n", __FILE__, __LINE__, p4d_val(e))
76 #define pgd_ERROR(e) \
77 	pr_err("%s:%d: bad pgd %016lx.\n", __FILE__, __LINE__, pgd_val(e))
78 
79 /*
80  * The vmalloc and module area will always be on the topmost area of the
81  * kernel mapping. 512GB are reserved for vmalloc by default.
82  * At the top of the vmalloc area a 2GB area is reserved where modules
83  * will reside. That makes sure that inter module branches always
84  * happen without trampolines and in addition the placement within a
85  * 2GB frame is branch prediction unit friendly.
86  */
87 extern unsigned long __bootdata_preserved(VMALLOC_START);
88 extern unsigned long __bootdata_preserved(VMALLOC_END);
89 #define VMALLOC_DEFAULT_SIZE	((512UL << 30) - MODULES_LEN)
90 extern struct page *__bootdata_preserved(vmemmap);
91 extern unsigned long __bootdata_preserved(vmemmap_size);
92 
93 extern unsigned long __bootdata_preserved(MODULES_VADDR);
94 extern unsigned long __bootdata_preserved(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 
175 #ifdef CONFIG_MEM_SOFT_DIRTY
176 #define _PAGE_SOFT_DIRTY 0x002		/* SW pte soft dirty bit */
177 #else
178 #define _PAGE_SOFT_DIRTY 0x000
179 #endif
180 
181 #define _PAGE_SW_BITS	0xffUL		/* All SW bits */
182 
183 #define _PAGE_SWP_EXCLUSIVE _PAGE_LARGE	/* SW pte exclusive swap bit */
184 
185 /* Set of bits not changed in pte_modify */
186 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
187 				 _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
188 
189 /*
190  * Mask of bits that must not be changed with RDP. Allow only _PAGE_PROTECT
191  * HW bit and all SW bits.
192  */
193 #define _PAGE_RDP_MASK		~(_PAGE_PROTECT | _PAGE_SW_BITS)
194 
195 /*
196  * handle_pte_fault uses pte_present and pte_none to find out the pte type
197  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
198  * distinguish present from not-present ptes. It is changed only with the page
199  * table lock held.
200  *
201  * The following table gives the different possible bit combinations for
202  * the pte hardware and software bits in the last 12 bits of a pte
203  * (. unassigned bit, x don't care, t swap type):
204  *
205  *				842100000000
206  *				000084210000
207  *				000000008421
208  *				.IR.uswrdy.p
209  * empty			.10.00000000
210  * swap				.11..ttttt.0
211  * prot-none, clean, old	.11.xx0000.1
212  * prot-none, clean, young	.11.xx0001.1
213  * prot-none, dirty, old	.11.xx0010.1
214  * prot-none, dirty, young	.11.xx0011.1
215  * read-only, clean, old	.11.xx0100.1
216  * read-only, clean, young	.01.xx0101.1
217  * read-only, dirty, old	.11.xx0110.1
218  * read-only, dirty, young	.01.xx0111.1
219  * read-write, clean, old	.11.xx1100.1
220  * read-write, clean, young	.01.xx1101.1
221  * read-write, dirty, old	.10.xx1110.1
222  * read-write, dirty, young	.00.xx1111.1
223  * HW-bits: R read-only, I invalid
224  * SW-bits: p present, y young, d dirty, r read, w write, s special,
225  *	    u unused, l large
226  *
227  * pte_none    is true for the bit pattern .10.00000000, pte == 0x400
228  * pte_swap    is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
229  * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
230  */
231 
232 /* Bits in the segment/region table address-space-control-element */
233 #define _ASCE_ORIGIN		~0xfffUL/* region/segment table origin	    */
234 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
235 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
236 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
237 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
238 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
239 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
240 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
241 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
242 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
243 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
244 
245 /* Bits in the region table entry */
246 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
247 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
248 #define _REGION_ENTRY_NOEXEC	0x100	/* region no-execute bit	    */
249 #define _REGION_ENTRY_OFFSET	0xc0	/* region table offset		    */
250 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
251 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region table type mask	    */
252 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
253 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
254 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
255 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
256 
257 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
258 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
259 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
260 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
261 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
262 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
263 
264 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address	     */
265 #define _REGION3_ENTRY_DIRTY	0x2000	/* SW region dirty bit */
266 #define _REGION3_ENTRY_YOUNG	0x1000	/* SW region young bit */
267 #define _REGION3_ENTRY_LARGE	0x0400	/* RTTE-format control, large page  */
268 #define _REGION3_ENTRY_READ	0x0002	/* SW region read bit */
269 #define _REGION3_ENTRY_WRITE	0x0001	/* SW region write bit */
270 
271 #ifdef CONFIG_MEM_SOFT_DIRTY
272 #define _REGION3_ENTRY_SOFT_DIRTY 0x4000 /* SW region soft dirty bit */
273 #else
274 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
275 #endif
276 
277 #define _REGION_ENTRY_BITS	 0xfffffffffffff22fUL
278 
279 /* Bits in the segment table entry */
280 #define _SEGMENT_ENTRY_BITS			0xfffffffffffffe33UL
281 #define _SEGMENT_ENTRY_HARDWARE_BITS		0xfffffffffffffe30UL
282 #define _SEGMENT_ENTRY_HARDWARE_BITS_LARGE	0xfffffffffff00730UL
283 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
284 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* page table origin		    */
285 #define _SEGMENT_ENTRY_PROTECT	0x200	/* segment protection bit	    */
286 #define _SEGMENT_ENTRY_NOEXEC	0x100	/* segment no-execute bit	    */
287 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
288 #define _SEGMENT_ENTRY_TYPE_MASK 0x0c	/* segment table type mask	    */
289 
290 #define _SEGMENT_ENTRY		(0)
291 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
292 
293 #define _SEGMENT_ENTRY_DIRTY	0x2000	/* SW segment dirty bit */
294 #define _SEGMENT_ENTRY_YOUNG	0x1000	/* SW segment young bit */
295 #define _SEGMENT_ENTRY_LARGE	0x0400	/* STE-format control, large page */
296 #define _SEGMENT_ENTRY_WRITE	0x0002	/* SW segment write bit */
297 #define _SEGMENT_ENTRY_READ	0x0001	/* SW segment read bit */
298 
299 #ifdef CONFIG_MEM_SOFT_DIRTY
300 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x4000 /* SW segment soft dirty bit */
301 #else
302 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
303 #endif
304 
305 #define _CRST_ENTRIES	2048	/* number of region/segment table entries */
306 #define _PAGE_ENTRIES	256	/* number of page table entries	*/
307 
308 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
309 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
310 
311 #define _REGION1_SHIFT	53
312 #define _REGION2_SHIFT	42
313 #define _REGION3_SHIFT	31
314 #define _SEGMENT_SHIFT	20
315 
316 #define _REGION1_INDEX	(0x7ffUL << _REGION1_SHIFT)
317 #define _REGION2_INDEX	(0x7ffUL << _REGION2_SHIFT)
318 #define _REGION3_INDEX	(0x7ffUL << _REGION3_SHIFT)
319 #define _SEGMENT_INDEX	(0x7ffUL << _SEGMENT_SHIFT)
320 #define _PAGE_INDEX	(0xffUL  << _PAGE_SHIFT)
321 
322 #define _REGION1_SIZE	(1UL << _REGION1_SHIFT)
323 #define _REGION2_SIZE	(1UL << _REGION2_SHIFT)
324 #define _REGION3_SIZE	(1UL << _REGION3_SHIFT)
325 #define _SEGMENT_SIZE	(1UL << _SEGMENT_SHIFT)
326 
327 #define _REGION1_MASK	(~(_REGION1_SIZE - 1))
328 #define _REGION2_MASK	(~(_REGION2_SIZE - 1))
329 #define _REGION3_MASK	(~(_REGION3_SIZE - 1))
330 #define _SEGMENT_MASK	(~(_SEGMENT_SIZE - 1))
331 
332 #define PMD_SHIFT	_SEGMENT_SHIFT
333 #define PUD_SHIFT	_REGION3_SHIFT
334 #define P4D_SHIFT	_REGION2_SHIFT
335 #define PGDIR_SHIFT	_REGION1_SHIFT
336 
337 #define PMD_SIZE	_SEGMENT_SIZE
338 #define PUD_SIZE	_REGION3_SIZE
339 #define P4D_SIZE	_REGION2_SIZE
340 #define PGDIR_SIZE	_REGION1_SIZE
341 
342 #define PMD_MASK	_SEGMENT_MASK
343 #define PUD_MASK	_REGION3_MASK
344 #define P4D_MASK	_REGION2_MASK
345 #define PGDIR_MASK	_REGION1_MASK
346 
347 #define PTRS_PER_PTE	_PAGE_ENTRIES
348 #define PTRS_PER_PMD	_CRST_ENTRIES
349 #define PTRS_PER_PUD	_CRST_ENTRIES
350 #define PTRS_PER_P4D	_CRST_ENTRIES
351 #define PTRS_PER_PGD	_CRST_ENTRIES
352 
353 /*
354  * Segment table and region3 table entry encoding
355  * (R = read-only, I = invalid, y = young bit):
356  *				dy..R...I...wr
357  * prot-none, clean, old	00..1...1...00
358  * prot-none, clean, young	01..1...1...00
359  * prot-none, dirty, old	10..1...1...00
360  * prot-none, dirty, young	11..1...1...00
361  * read-only, clean, old	00..1...1...01
362  * read-only, clean, young	01..1...0...01
363  * read-only, dirty, old	10..1...1...01
364  * read-only, dirty, young	11..1...0...01
365  * read-write, clean, old	00..1...1...11
366  * read-write, clean, young	01..1...0...11
367  * read-write, dirty, old	10..0...1...11
368  * read-write, dirty, young	11..0...0...11
369  * The segment table origin is used to distinguish empty (origin==0) from
370  * read-write, old segment table entries (origin!=0)
371  * HW-bits: R read-only, I invalid
372  * SW-bits: y young, d dirty, r read, w write
373  */
374 
375 /* Page status table bits for virtualization */
376 #define PGSTE_ACC_BITS	0xf000000000000000UL
377 #define PGSTE_FP_BIT	0x0800000000000000UL
378 #define PGSTE_PCL_BIT	0x0080000000000000UL
379 #define PGSTE_HR_BIT	0x0040000000000000UL
380 #define PGSTE_HC_BIT	0x0020000000000000UL
381 #define PGSTE_GR_BIT	0x0004000000000000UL
382 #define PGSTE_GC_BIT	0x0002000000000000UL
383 #define PGSTE_UC_BIT	0x0000800000000000UL	/* user dirty (migration) */
384 #define PGSTE_IN_BIT	0x0000400000000000UL	/* IPTE notify bit */
385 #define PGSTE_VSIE_BIT	0x0000200000000000UL	/* ref'd in a shadow table */
386 
387 /* Guest Page State used for virtualization */
388 #define _PGSTE_GPS_ZERO			0x0000000080000000UL
389 #define _PGSTE_GPS_NODAT		0x0000000040000000UL
390 #define _PGSTE_GPS_USAGE_MASK		0x0000000003000000UL
391 #define _PGSTE_GPS_USAGE_STABLE		0x0000000000000000UL
392 #define _PGSTE_GPS_USAGE_UNUSED		0x0000000001000000UL
393 #define _PGSTE_GPS_USAGE_POT_VOLATILE	0x0000000002000000UL
394 #define _PGSTE_GPS_USAGE_VOLATILE	_PGSTE_GPS_USAGE_MASK
395 
396 /*
397  * A user page table pointer has the space-switch-event bit, the
398  * private-space-control bit and the storage-alteration-event-control
399  * bit set. A kernel page table pointer doesn't need them.
400  */
401 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
402 				 _ASCE_ALT_EVENT)
403 
404 /*
405  * Page protection definitions.
406  */
407 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID | _PAGE_PROTECT)
408 #define PAGE_RO		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
409 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
410 #define PAGE_RX		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
411 				 _PAGE_INVALID | _PAGE_PROTECT)
412 #define PAGE_RW		__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
413 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
414 #define PAGE_RWX	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
415 				 _PAGE_INVALID | _PAGE_PROTECT)
416 
417 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
418 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
419 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
420 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
421 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
422 				 _PAGE_PROTECT | _PAGE_NOEXEC)
423 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
424 				  _PAGE_YOUNG |	_PAGE_DIRTY)
425 
426 /*
427  * On s390 the page table entry has an invalid bit and a read-only bit.
428  * Read permission implies execute permission and write permission
429  * implies read permission.
430  */
431          /*xwr*/
432 
433 /*
434  * Segment entry (large page) protection definitions.
435  */
436 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_INVALID | \
437 				 _SEGMENT_ENTRY_PROTECT)
438 #define SEGMENT_RO	__pgprot(_SEGMENT_ENTRY_PROTECT | \
439 				 _SEGMENT_ENTRY_READ | \
440 				 _SEGMENT_ENTRY_NOEXEC)
441 #define SEGMENT_RX	__pgprot(_SEGMENT_ENTRY_PROTECT | \
442 				 _SEGMENT_ENTRY_READ)
443 #define SEGMENT_RW	__pgprot(_SEGMENT_ENTRY_READ | \
444 				 _SEGMENT_ENTRY_WRITE | \
445 				 _SEGMENT_ENTRY_NOEXEC)
446 #define SEGMENT_RWX	__pgprot(_SEGMENT_ENTRY_READ | \
447 				 _SEGMENT_ENTRY_WRITE)
448 #define SEGMENT_KERNEL	__pgprot(_SEGMENT_ENTRY |	\
449 				 _SEGMENT_ENTRY_LARGE |	\
450 				 _SEGMENT_ENTRY_READ |	\
451 				 _SEGMENT_ENTRY_WRITE | \
452 				 _SEGMENT_ENTRY_YOUNG | \
453 				 _SEGMENT_ENTRY_DIRTY | \
454 				 _SEGMENT_ENTRY_NOEXEC)
455 #define SEGMENT_KERNEL_RO __pgprot(_SEGMENT_ENTRY |	\
456 				 _SEGMENT_ENTRY_LARGE |	\
457 				 _SEGMENT_ENTRY_READ |	\
458 				 _SEGMENT_ENTRY_YOUNG |	\
459 				 _SEGMENT_ENTRY_PROTECT | \
460 				 _SEGMENT_ENTRY_NOEXEC)
461 #define SEGMENT_KERNEL_EXEC __pgprot(_SEGMENT_ENTRY |	\
462 				 _SEGMENT_ENTRY_LARGE |	\
463 				 _SEGMENT_ENTRY_READ |	\
464 				 _SEGMENT_ENTRY_WRITE | \
465 				 _SEGMENT_ENTRY_YOUNG |	\
466 				 _SEGMENT_ENTRY_DIRTY)
467 
468 /*
469  * Region3 entry (large page) protection definitions.
470  */
471 
472 #define REGION3_KERNEL	__pgprot(_REGION_ENTRY_TYPE_R3 | \
473 				 _REGION3_ENTRY_LARGE |	 \
474 				 _REGION3_ENTRY_READ |	 \
475 				 _REGION3_ENTRY_WRITE |	 \
476 				 _REGION3_ENTRY_YOUNG |	 \
477 				 _REGION3_ENTRY_DIRTY | \
478 				 _REGION_ENTRY_NOEXEC)
479 #define REGION3_KERNEL_RO __pgprot(_REGION_ENTRY_TYPE_R3 | \
480 				   _REGION3_ENTRY_LARGE |  \
481 				   _REGION3_ENTRY_READ |   \
482 				   _REGION3_ENTRY_YOUNG |  \
483 				   _REGION_ENTRY_PROTECT | \
484 				   _REGION_ENTRY_NOEXEC)
485 #define REGION3_KERNEL_EXEC __pgprot(_REGION_ENTRY_TYPE_R3 | \
486 				 _REGION3_ENTRY_LARGE |	 \
487 				 _REGION3_ENTRY_READ |	 \
488 				 _REGION3_ENTRY_WRITE |	 \
489 				 _REGION3_ENTRY_YOUNG |	 \
490 				 _REGION3_ENTRY_DIRTY)
491 
492 static inline bool mm_p4d_folded(struct mm_struct *mm)
493 {
494 	return mm->context.asce_limit <= _REGION1_SIZE;
495 }
496 #define mm_p4d_folded(mm) mm_p4d_folded(mm)
497 
498 static inline bool mm_pud_folded(struct mm_struct *mm)
499 {
500 	return mm->context.asce_limit <= _REGION2_SIZE;
501 }
502 #define mm_pud_folded(mm) mm_pud_folded(mm)
503 
504 static inline bool mm_pmd_folded(struct mm_struct *mm)
505 {
506 	return mm->context.asce_limit <= _REGION3_SIZE;
507 }
508 #define mm_pmd_folded(mm) mm_pmd_folded(mm)
509 
510 static inline int mm_has_pgste(struct mm_struct *mm)
511 {
512 #ifdef CONFIG_PGSTE
513 	if (unlikely(mm->context.has_pgste))
514 		return 1;
515 #endif
516 	return 0;
517 }
518 
519 static inline int mm_is_protected(struct mm_struct *mm)
520 {
521 #ifdef CONFIG_PGSTE
522 	if (unlikely(atomic_read(&mm->context.protected_count)))
523 		return 1;
524 #endif
525 	return 0;
526 }
527 
528 static inline int mm_alloc_pgste(struct mm_struct *mm)
529 {
530 #ifdef CONFIG_PGSTE
531 	if (unlikely(mm->context.alloc_pgste))
532 		return 1;
533 #endif
534 	return 0;
535 }
536 
537 static inline pte_t clear_pte_bit(pte_t pte, pgprot_t prot)
538 {
539 	return __pte(pte_val(pte) & ~pgprot_val(prot));
540 }
541 
542 static inline pte_t set_pte_bit(pte_t pte, pgprot_t prot)
543 {
544 	return __pte(pte_val(pte) | pgprot_val(prot));
545 }
546 
547 static inline pmd_t clear_pmd_bit(pmd_t pmd, pgprot_t prot)
548 {
549 	return __pmd(pmd_val(pmd) & ~pgprot_val(prot));
550 }
551 
552 static inline pmd_t set_pmd_bit(pmd_t pmd, pgprot_t prot)
553 {
554 	return __pmd(pmd_val(pmd) | pgprot_val(prot));
555 }
556 
557 static inline pud_t clear_pud_bit(pud_t pud, pgprot_t prot)
558 {
559 	return __pud(pud_val(pud) & ~pgprot_val(prot));
560 }
561 
562 static inline pud_t set_pud_bit(pud_t pud, pgprot_t prot)
563 {
564 	return __pud(pud_val(pud) | pgprot_val(prot));
565 }
566 
567 /*
568  * In the case that a guest uses storage keys
569  * faults should no longer be backed by zero pages
570  */
571 #define mm_forbids_zeropage mm_has_pgste
572 static inline int mm_uses_skeys(struct mm_struct *mm)
573 {
574 #ifdef CONFIG_PGSTE
575 	if (mm->context.uses_skeys)
576 		return 1;
577 #endif
578 	return 0;
579 }
580 
581 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
582 {
583 	union register_pair r1 = { .even = old, .odd = new, };
584 	unsigned long address = (unsigned long)ptr | 1;
585 
586 	asm volatile(
587 		"	csp	%[r1],%[address]"
588 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
589 		: [address] "d" (address)
590 		: "cc");
591 }
592 
593 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
594 {
595 	union register_pair r1 = { .even = old, .odd = new, };
596 	unsigned long address = (unsigned long)ptr | 1;
597 
598 	asm volatile(
599 		"	cspg	%[r1],%[address]"
600 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
601 		: [address] "d" (address)
602 		: "cc");
603 }
604 
605 #define CRDTE_DTT_PAGE		0x00UL
606 #define CRDTE_DTT_SEGMENT	0x10UL
607 #define CRDTE_DTT_REGION3	0x14UL
608 #define CRDTE_DTT_REGION2	0x18UL
609 #define CRDTE_DTT_REGION1	0x1cUL
610 
611 static inline void crdte(unsigned long old, unsigned long new,
612 			 unsigned long *table, unsigned long dtt,
613 			 unsigned long address, unsigned long asce)
614 {
615 	union register_pair r1 = { .even = old, .odd = new, };
616 	union register_pair r2 = { .even = __pa(table) | dtt, .odd = address, };
617 
618 	asm volatile(".insn rrf,0xb98f0000,%[r1],%[r2],%[asce],0"
619 		     : [r1] "+&d" (r1.pair)
620 		     : [r2] "d" (r2.pair), [asce] "a" (asce)
621 		     : "memory", "cc");
622 }
623 
624 /*
625  * pgd/p4d/pud/pmd/pte query functions
626  */
627 static inline int pgd_folded(pgd_t pgd)
628 {
629 	return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
630 }
631 
632 static inline int pgd_present(pgd_t pgd)
633 {
634 	if (pgd_folded(pgd))
635 		return 1;
636 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
637 }
638 
639 static inline int pgd_none(pgd_t pgd)
640 {
641 	if (pgd_folded(pgd))
642 		return 0;
643 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
644 }
645 
646 static inline int pgd_bad(pgd_t pgd)
647 {
648 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1)
649 		return 0;
650 	return (pgd_val(pgd) & ~_REGION_ENTRY_BITS) != 0;
651 }
652 
653 static inline unsigned long pgd_pfn(pgd_t pgd)
654 {
655 	unsigned long origin_mask;
656 
657 	origin_mask = _REGION_ENTRY_ORIGIN;
658 	return (pgd_val(pgd) & origin_mask) >> PAGE_SHIFT;
659 }
660 
661 static inline int p4d_folded(p4d_t p4d)
662 {
663 	return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
664 }
665 
666 static inline int p4d_present(p4d_t p4d)
667 {
668 	if (p4d_folded(p4d))
669 		return 1;
670 	return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
671 }
672 
673 static inline int p4d_none(p4d_t p4d)
674 {
675 	if (p4d_folded(p4d))
676 		return 0;
677 	return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
678 }
679 
680 static inline unsigned long p4d_pfn(p4d_t p4d)
681 {
682 	unsigned long origin_mask;
683 
684 	origin_mask = _REGION_ENTRY_ORIGIN;
685 	return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
686 }
687 
688 static inline int pud_folded(pud_t pud)
689 {
690 	return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
691 }
692 
693 static inline int pud_present(pud_t pud)
694 {
695 	if (pud_folded(pud))
696 		return 1;
697 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
698 }
699 
700 static inline int pud_none(pud_t pud)
701 {
702 	if (pud_folded(pud))
703 		return 0;
704 	return pud_val(pud) == _REGION3_ENTRY_EMPTY;
705 }
706 
707 #define pud_leaf	pud_large
708 static inline int pud_large(pud_t pud)
709 {
710 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
711 		return 0;
712 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
713 }
714 
715 #define pmd_leaf	pmd_large
716 static inline int pmd_large(pmd_t pmd)
717 {
718 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
719 }
720 
721 static inline int pmd_bad(pmd_t pmd)
722 {
723 	if ((pmd_val(pmd) & _SEGMENT_ENTRY_TYPE_MASK) > 0 || pmd_large(pmd))
724 		return 1;
725 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
726 }
727 
728 static inline int pud_bad(pud_t pud)
729 {
730 	unsigned long type = pud_val(pud) & _REGION_ENTRY_TYPE_MASK;
731 
732 	if (type > _REGION_ENTRY_TYPE_R3 || pud_large(pud))
733 		return 1;
734 	if (type < _REGION_ENTRY_TYPE_R3)
735 		return 0;
736 	return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
737 }
738 
739 static inline int p4d_bad(p4d_t p4d)
740 {
741 	unsigned long type = p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK;
742 
743 	if (type > _REGION_ENTRY_TYPE_R2)
744 		return 1;
745 	if (type < _REGION_ENTRY_TYPE_R2)
746 		return 0;
747 	return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
748 }
749 
750 static inline int pmd_present(pmd_t pmd)
751 {
752 	return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
753 }
754 
755 static inline int pmd_none(pmd_t pmd)
756 {
757 	return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
758 }
759 
760 #define pmd_write pmd_write
761 static inline int pmd_write(pmd_t pmd)
762 {
763 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
764 }
765 
766 #define pud_write pud_write
767 static inline int pud_write(pud_t pud)
768 {
769 	return (pud_val(pud) & _REGION3_ENTRY_WRITE) != 0;
770 }
771 
772 static inline int pmd_dirty(pmd_t pmd)
773 {
774 	return (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
775 }
776 
777 #define pmd_young pmd_young
778 static inline int pmd_young(pmd_t pmd)
779 {
780 	return (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
781 }
782 
783 static inline int pte_present(pte_t pte)
784 {
785 	/* Bit pattern: (pte & 0x001) == 0x001 */
786 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
787 }
788 
789 static inline int pte_none(pte_t pte)
790 {
791 	/* Bit pattern: pte == 0x400 */
792 	return pte_val(pte) == _PAGE_INVALID;
793 }
794 
795 static inline int pte_swap(pte_t pte)
796 {
797 	/* Bit pattern: (pte & 0x201) == 0x200 */
798 	return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
799 		== _PAGE_PROTECT;
800 }
801 
802 static inline int pte_special(pte_t pte)
803 {
804 	return (pte_val(pte) & _PAGE_SPECIAL);
805 }
806 
807 #define __HAVE_ARCH_PTE_SAME
808 static inline int pte_same(pte_t a, pte_t b)
809 {
810 	return pte_val(a) == pte_val(b);
811 }
812 
813 #ifdef CONFIG_NUMA_BALANCING
814 static inline int pte_protnone(pte_t pte)
815 {
816 	return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
817 }
818 
819 static inline int pmd_protnone(pmd_t pmd)
820 {
821 	/* pmd_large(pmd) implies pmd_present(pmd) */
822 	return pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
823 }
824 #endif
825 
826 static inline int pte_swp_exclusive(pte_t pte)
827 {
828 	return pte_val(pte) & _PAGE_SWP_EXCLUSIVE;
829 }
830 
831 static inline pte_t pte_swp_mkexclusive(pte_t pte)
832 {
833 	return set_pte_bit(pte, __pgprot(_PAGE_SWP_EXCLUSIVE));
834 }
835 
836 static inline pte_t pte_swp_clear_exclusive(pte_t pte)
837 {
838 	return clear_pte_bit(pte, __pgprot(_PAGE_SWP_EXCLUSIVE));
839 }
840 
841 static inline int pte_soft_dirty(pte_t pte)
842 {
843 	return pte_val(pte) & _PAGE_SOFT_DIRTY;
844 }
845 #define pte_swp_soft_dirty pte_soft_dirty
846 
847 static inline pte_t pte_mksoft_dirty(pte_t pte)
848 {
849 	return set_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
850 }
851 #define pte_swp_mksoft_dirty pte_mksoft_dirty
852 
853 static inline pte_t pte_clear_soft_dirty(pte_t pte)
854 {
855 	return clear_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
856 }
857 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
858 
859 static inline int pmd_soft_dirty(pmd_t pmd)
860 {
861 	return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
862 }
863 
864 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
865 {
866 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
867 }
868 
869 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
870 {
871 	return clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
872 }
873 
874 /*
875  * query functions pte_write/pte_dirty/pte_young only work if
876  * pte_present() is true. Undefined behaviour if not..
877  */
878 static inline int pte_write(pte_t pte)
879 {
880 	return (pte_val(pte) & _PAGE_WRITE) != 0;
881 }
882 
883 static inline int pte_dirty(pte_t pte)
884 {
885 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
886 }
887 
888 static inline int pte_young(pte_t pte)
889 {
890 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
891 }
892 
893 #define __HAVE_ARCH_PTE_UNUSED
894 static inline int pte_unused(pte_t pte)
895 {
896 	return pte_val(pte) & _PAGE_UNUSED;
897 }
898 
899 /*
900  * Extract the pgprot value from the given pte while at the same time making it
901  * usable for kernel address space mappings where fault driven dirty and
902  * young/old accounting is not supported, i.e _PAGE_PROTECT and _PAGE_INVALID
903  * must not be set.
904  */
905 static inline pgprot_t pte_pgprot(pte_t pte)
906 {
907 	unsigned long pte_flags = pte_val(pte) & _PAGE_CHG_MASK;
908 
909 	if (pte_write(pte))
910 		pte_flags |= pgprot_val(PAGE_KERNEL);
911 	else
912 		pte_flags |= pgprot_val(PAGE_KERNEL_RO);
913 	pte_flags |= pte_val(pte) & mio_wb_bit_mask;
914 
915 	return __pgprot(pte_flags);
916 }
917 
918 /*
919  * pgd/pmd/pte modification functions
920  */
921 
922 static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
923 {
924 	WRITE_ONCE(*pgdp, pgd);
925 }
926 
927 static inline void set_p4d(p4d_t *p4dp, p4d_t p4d)
928 {
929 	WRITE_ONCE(*p4dp, p4d);
930 }
931 
932 static inline void set_pud(pud_t *pudp, pud_t pud)
933 {
934 	WRITE_ONCE(*pudp, pud);
935 }
936 
937 static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
938 {
939 	WRITE_ONCE(*pmdp, pmd);
940 }
941 
942 static inline void set_pte(pte_t *ptep, pte_t pte)
943 {
944 	WRITE_ONCE(*ptep, pte);
945 }
946 
947 static inline void pgd_clear(pgd_t *pgd)
948 {
949 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
950 		set_pgd(pgd, __pgd(_REGION1_ENTRY_EMPTY));
951 }
952 
953 static inline void p4d_clear(p4d_t *p4d)
954 {
955 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
956 		set_p4d(p4d, __p4d(_REGION2_ENTRY_EMPTY));
957 }
958 
959 static inline void pud_clear(pud_t *pud)
960 {
961 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
962 		set_pud(pud, __pud(_REGION3_ENTRY_EMPTY));
963 }
964 
965 static inline void pmd_clear(pmd_t *pmdp)
966 {
967 	set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
968 }
969 
970 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
971 {
972 	set_pte(ptep, __pte(_PAGE_INVALID));
973 }
974 
975 /*
976  * The following pte modification functions only work if
977  * pte_present() is true. Undefined behaviour if not..
978  */
979 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
980 {
981 	pte = clear_pte_bit(pte, __pgprot(~_PAGE_CHG_MASK));
982 	pte = set_pte_bit(pte, newprot);
983 	/*
984 	 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
985 	 * has the invalid bit set, clear it again for readable, young pages
986 	 */
987 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
988 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
989 	/*
990 	 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
991 	 * protection bit set, clear it again for writable, dirty pages
992 	 */
993 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
994 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
995 	return pte;
996 }
997 
998 static inline pte_t pte_wrprotect(pte_t pte)
999 {
1000 	pte = clear_pte_bit(pte, __pgprot(_PAGE_WRITE));
1001 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1002 }
1003 
1004 static inline pte_t pte_mkwrite_novma(pte_t pte)
1005 {
1006 	pte = set_pte_bit(pte, __pgprot(_PAGE_WRITE));
1007 	if (pte_val(pte) & _PAGE_DIRTY)
1008 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1009 	return pte;
1010 }
1011 
1012 static inline pte_t pte_mkclean(pte_t pte)
1013 {
1014 	pte = clear_pte_bit(pte, __pgprot(_PAGE_DIRTY));
1015 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1016 }
1017 
1018 static inline pte_t pte_mkdirty(pte_t pte)
1019 {
1020 	pte = set_pte_bit(pte, __pgprot(_PAGE_DIRTY | _PAGE_SOFT_DIRTY));
1021 	if (pte_val(pte) & _PAGE_WRITE)
1022 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1023 	return pte;
1024 }
1025 
1026 static inline pte_t pte_mkold(pte_t pte)
1027 {
1028 	pte = clear_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1029 	return set_pte_bit(pte, __pgprot(_PAGE_INVALID));
1030 }
1031 
1032 static inline pte_t pte_mkyoung(pte_t pte)
1033 {
1034 	pte = set_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1035 	if (pte_val(pte) & _PAGE_READ)
1036 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
1037 	return pte;
1038 }
1039 
1040 static inline pte_t pte_mkspecial(pte_t pte)
1041 {
1042 	return set_pte_bit(pte, __pgprot(_PAGE_SPECIAL));
1043 }
1044 
1045 #ifdef CONFIG_HUGETLB_PAGE
1046 static inline pte_t pte_mkhuge(pte_t pte)
1047 {
1048 	return set_pte_bit(pte, __pgprot(_PAGE_LARGE));
1049 }
1050 #endif
1051 
1052 #define IPTE_GLOBAL	0
1053 #define	IPTE_LOCAL	1
1054 
1055 #define IPTE_NODAT	0x400
1056 #define IPTE_GUEST_ASCE	0x800
1057 
1058 static __always_inline void __ptep_rdp(unsigned long addr, pte_t *ptep,
1059 				       unsigned long opt, unsigned long asce,
1060 				       int local)
1061 {
1062 	unsigned long pto;
1063 
1064 	pto = __pa(ptep) & ~(PTRS_PER_PTE * sizeof(pte_t) - 1);
1065 	asm volatile(".insn rrf,0xb98b0000,%[r1],%[r2],%[asce],%[m4]"
1066 		     : "+m" (*ptep)
1067 		     : [r1] "a" (pto), [r2] "a" ((addr & PAGE_MASK) | opt),
1068 		       [asce] "a" (asce), [m4] "i" (local));
1069 }
1070 
1071 static __always_inline void __ptep_ipte(unsigned long address, pte_t *ptep,
1072 					unsigned long opt, unsigned long asce,
1073 					int local)
1074 {
1075 	unsigned long pto = __pa(ptep);
1076 
1077 	if (__builtin_constant_p(opt) && opt == 0) {
1078 		/* Invalidation + TLB flush for the pte */
1079 		asm volatile(
1080 			"	ipte	%[r1],%[r2],0,%[m4]"
1081 			: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
1082 			  [m4] "i" (local));
1083 		return;
1084 	}
1085 
1086 	/* Invalidate ptes with options + TLB flush of the ptes */
1087 	opt = opt | (asce & _ASCE_ORIGIN);
1088 	asm volatile(
1089 		"	ipte	%[r1],%[r2],%[r3],%[m4]"
1090 		: [r2] "+a" (address), [r3] "+a" (opt)
1091 		: [r1] "a" (pto), [m4] "i" (local) : "memory");
1092 }
1093 
1094 static __always_inline void __ptep_ipte_range(unsigned long address, int nr,
1095 					      pte_t *ptep, int local)
1096 {
1097 	unsigned long pto = __pa(ptep);
1098 
1099 	/* Invalidate a range of ptes + TLB flush of the ptes */
1100 	do {
1101 		asm volatile(
1102 			"	ipte %[r1],%[r2],%[r3],%[m4]"
1103 			: [r2] "+a" (address), [r3] "+a" (nr)
1104 			: [r1] "a" (pto), [m4] "i" (local) : "memory");
1105 	} while (nr != 255);
1106 }
1107 
1108 /*
1109  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1110  * both clear the TLB for the unmapped pte. The reason is that
1111  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1112  * to modify an active pte. The sequence is
1113  *   1) ptep_get_and_clear
1114  *   2) set_pte_at
1115  *   3) flush_tlb_range
1116  * On s390 the tlb needs to get flushed with the modification of the pte
1117  * if the pte is active. The only way how this can be implemented is to
1118  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1119  * is a nop.
1120  */
1121 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1122 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1123 
1124 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1125 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1126 					    unsigned long addr, pte_t *ptep)
1127 {
1128 	pte_t pte = *ptep;
1129 
1130 	pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1131 	return pte_young(pte);
1132 }
1133 
1134 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1135 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1136 					 unsigned long address, pte_t *ptep)
1137 {
1138 	return ptep_test_and_clear_young(vma, address, ptep);
1139 }
1140 
1141 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1142 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1143 				       unsigned long addr, pte_t *ptep)
1144 {
1145 	pte_t res;
1146 
1147 	res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1148 	/* At this point the reference through the mapping is still present */
1149 	if (mm_is_protected(mm) && pte_present(res))
1150 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1151 	return res;
1152 }
1153 
1154 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1155 pte_t ptep_modify_prot_start(struct vm_area_struct *, unsigned long, pte_t *);
1156 void ptep_modify_prot_commit(struct vm_area_struct *, unsigned long,
1157 			     pte_t *, pte_t, pte_t);
1158 
1159 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1160 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1161 				     unsigned long addr, pte_t *ptep)
1162 {
1163 	pte_t res;
1164 
1165 	res = ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1166 	/* At this point the reference through the mapping is still present */
1167 	if (mm_is_protected(vma->vm_mm) && pte_present(res))
1168 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1169 	return res;
1170 }
1171 
1172 /*
1173  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1174  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1175  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1176  * cannot be accessed while the batched unmap is running. In this case
1177  * full==1 and a simple pte_clear is enough. See tlb.h.
1178  */
1179 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1180 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1181 					    unsigned long addr,
1182 					    pte_t *ptep, int full)
1183 {
1184 	pte_t res;
1185 
1186 	if (full) {
1187 		res = *ptep;
1188 		set_pte(ptep, __pte(_PAGE_INVALID));
1189 	} else {
1190 		res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1191 	}
1192 	/* Nothing to do */
1193 	if (!mm_is_protected(mm) || !pte_present(res))
1194 		return res;
1195 	/*
1196 	 * At this point the reference through the mapping is still present.
1197 	 * The notifier should have destroyed all protected vCPUs at this
1198 	 * point, so the destroy should be successful.
1199 	 */
1200 	if (full && !uv_destroy_owned_page(pte_val(res) & PAGE_MASK))
1201 		return res;
1202 	/*
1203 	 * If something went wrong and the page could not be destroyed, or
1204 	 * if this is not a mm teardown, the slower export is used as
1205 	 * fallback instead.
1206 	 */
1207 	uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1208 	return res;
1209 }
1210 
1211 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1212 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1213 				      unsigned long addr, pte_t *ptep)
1214 {
1215 	pte_t pte = *ptep;
1216 
1217 	if (pte_write(pte))
1218 		ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1219 }
1220 
1221 /*
1222  * Check if PTEs only differ in _PAGE_PROTECT HW bit, but also allow SW PTE
1223  * bits in the comparison. Those might change e.g. because of dirty and young
1224  * tracking.
1225  */
1226 static inline int pte_allow_rdp(pte_t old, pte_t new)
1227 {
1228 	/*
1229 	 * Only allow changes from RO to RW
1230 	 */
1231 	if (!(pte_val(old) & _PAGE_PROTECT) || pte_val(new) & _PAGE_PROTECT)
1232 		return 0;
1233 
1234 	return (pte_val(old) & _PAGE_RDP_MASK) == (pte_val(new) & _PAGE_RDP_MASK);
1235 }
1236 
1237 static inline void flush_tlb_fix_spurious_fault(struct vm_area_struct *vma,
1238 						unsigned long address,
1239 						pte_t *ptep)
1240 {
1241 	/*
1242 	 * RDP might not have propagated the PTE protection reset to all CPUs,
1243 	 * so there could be spurious TLB protection faults.
1244 	 * NOTE: This will also be called when a racing pagetable update on
1245 	 * another thread already installed the correct PTE. Both cases cannot
1246 	 * really be distinguished.
1247 	 * Therefore, only do the local TLB flush when RDP can be used, and the
1248 	 * PTE does not have _PAGE_PROTECT set, to avoid unnecessary overhead.
1249 	 * A local RDP can be used to do the flush.
1250 	 */
1251 	if (MACHINE_HAS_RDP && !(pte_val(*ptep) & _PAGE_PROTECT))
1252 		__ptep_rdp(address, ptep, 0, 0, 1);
1253 }
1254 #define flush_tlb_fix_spurious_fault flush_tlb_fix_spurious_fault
1255 
1256 void ptep_reset_dat_prot(struct mm_struct *mm, unsigned long addr, pte_t *ptep,
1257 			 pte_t new);
1258 
1259 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1260 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1261 					unsigned long addr, pte_t *ptep,
1262 					pte_t entry, int dirty)
1263 {
1264 	if (pte_same(*ptep, entry))
1265 		return 0;
1266 	if (MACHINE_HAS_RDP && !mm_has_pgste(vma->vm_mm) && pte_allow_rdp(*ptep, entry))
1267 		ptep_reset_dat_prot(vma->vm_mm, addr, ptep, entry);
1268 	else
1269 		ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1270 	return 1;
1271 }
1272 
1273 /*
1274  * Additional functions to handle KVM guest page tables
1275  */
1276 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1277 		     pte_t *ptep, pte_t entry);
1278 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1279 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1280 		 pte_t *ptep, unsigned long bits);
1281 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1282 		    pte_t *ptep, int prot, unsigned long bit);
1283 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1284 		     pte_t *ptep , int reset);
1285 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1286 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1287 		    pte_t *sptep, pte_t *tptep, pte_t pte);
1288 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1289 
1290 bool ptep_test_and_clear_uc(struct mm_struct *mm, unsigned long address,
1291 			    pte_t *ptep);
1292 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1293 			  unsigned char key, bool nq);
1294 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1295 			       unsigned char key, unsigned char *oldkey,
1296 			       bool nq, bool mr, bool mc);
1297 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1298 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1299 			  unsigned char *key);
1300 
1301 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1302 				unsigned long bits, unsigned long value);
1303 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1304 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1305 			unsigned long *oldpte, unsigned long *oldpgste);
1306 void gmap_pmdp_csp(struct mm_struct *mm, unsigned long vmaddr);
1307 void gmap_pmdp_invalidate(struct mm_struct *mm, unsigned long vmaddr);
1308 void gmap_pmdp_idte_local(struct mm_struct *mm, unsigned long vmaddr);
1309 void gmap_pmdp_idte_global(struct mm_struct *mm, unsigned long vmaddr);
1310 
1311 #define pgprot_writecombine	pgprot_writecombine
1312 pgprot_t pgprot_writecombine(pgprot_t prot);
1313 
1314 #define pgprot_writethrough	pgprot_writethrough
1315 pgprot_t pgprot_writethrough(pgprot_t prot);
1316 
1317 /*
1318  * Set multiple PTEs to consecutive pages with a single call.  All PTEs
1319  * are within the same folio, PMD and VMA.
1320  */
1321 static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
1322 			      pte_t *ptep, pte_t entry, unsigned int nr)
1323 {
1324 	if (pte_present(entry))
1325 		entry = clear_pte_bit(entry, __pgprot(_PAGE_UNUSED));
1326 	if (mm_has_pgste(mm)) {
1327 		for (;;) {
1328 			ptep_set_pte_at(mm, addr, ptep, entry);
1329 			if (--nr == 0)
1330 				break;
1331 			ptep++;
1332 			entry = __pte(pte_val(entry) + PAGE_SIZE);
1333 			addr += PAGE_SIZE;
1334 		}
1335 	} else {
1336 		for (;;) {
1337 			set_pte(ptep, entry);
1338 			if (--nr == 0)
1339 				break;
1340 			ptep++;
1341 			entry = __pte(pte_val(entry) + PAGE_SIZE);
1342 		}
1343 	}
1344 }
1345 #define set_ptes set_ptes
1346 
1347 /*
1348  * Conversion functions: convert a page and protection to a page entry,
1349  * and a page entry and page directory to the page they refer to.
1350  */
1351 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1352 {
1353 	pte_t __pte;
1354 
1355 	__pte = __pte(physpage | pgprot_val(pgprot));
1356 	if (!MACHINE_HAS_NX)
1357 		__pte = clear_pte_bit(__pte, __pgprot(_PAGE_NOEXEC));
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 p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1373 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1374 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1375 
1376 #define p4d_deref(pud) ((unsigned long)__va(p4d_val(pud) & _REGION_ENTRY_ORIGIN))
1377 #define pgd_deref(pgd) ((unsigned long)__va(pgd_val(pgd) & _REGION_ENTRY_ORIGIN))
1378 
1379 static inline unsigned long pmd_deref(pmd_t pmd)
1380 {
1381 	unsigned long origin_mask;
1382 
1383 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
1384 	if (pmd_large(pmd))
1385 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
1386 	return (unsigned long)__va(pmd_val(pmd) & origin_mask);
1387 }
1388 
1389 static inline unsigned long pmd_pfn(pmd_t pmd)
1390 {
1391 	return __pa(pmd_deref(pmd)) >> PAGE_SHIFT;
1392 }
1393 
1394 static inline unsigned long pud_deref(pud_t pud)
1395 {
1396 	unsigned long origin_mask;
1397 
1398 	origin_mask = _REGION_ENTRY_ORIGIN;
1399 	if (pud_large(pud))
1400 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
1401 	return (unsigned long)__va(pud_val(pud) & origin_mask);
1402 }
1403 
1404 static inline unsigned long pud_pfn(pud_t pud)
1405 {
1406 	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
1407 }
1408 
1409 /*
1410  * The pgd_offset function *always* adds the index for the top-level
1411  * region/segment table. This is done to get a sequence like the
1412  * following to work:
1413  *	pgdp = pgd_offset(current->mm, addr);
1414  *	pgd = READ_ONCE(*pgdp);
1415  *	p4dp = p4d_offset(&pgd, addr);
1416  *	...
1417  * The subsequent p4d_offset, pud_offset and pmd_offset functions
1418  * only add an index if they dereferenced the pointer.
1419  */
1420 static inline pgd_t *pgd_offset_raw(pgd_t *pgd, unsigned long address)
1421 {
1422 	unsigned long rste;
1423 	unsigned int shift;
1424 
1425 	/* Get the first entry of the top level table */
1426 	rste = pgd_val(*pgd);
1427 	/* Pick up the shift from the table type of the first entry */
1428 	shift = ((rste & _REGION_ENTRY_TYPE_MASK) >> 2) * 11 + 20;
1429 	return pgd + ((address >> shift) & (PTRS_PER_PGD - 1));
1430 }
1431 
1432 #define pgd_offset(mm, address) pgd_offset_raw(READ_ONCE((mm)->pgd), address)
1433 
1434 static inline p4d_t *p4d_offset_lockless(pgd_t *pgdp, pgd_t pgd, unsigned long address)
1435 {
1436 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R1)
1437 		return (p4d_t *) pgd_deref(pgd) + p4d_index(address);
1438 	return (p4d_t *) pgdp;
1439 }
1440 #define p4d_offset_lockless p4d_offset_lockless
1441 
1442 static inline p4d_t *p4d_offset(pgd_t *pgdp, unsigned long address)
1443 {
1444 	return p4d_offset_lockless(pgdp, *pgdp, address);
1445 }
1446 
1447 static inline pud_t *pud_offset_lockless(p4d_t *p4dp, p4d_t p4d, unsigned long address)
1448 {
1449 	if ((p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R2)
1450 		return (pud_t *) p4d_deref(p4d) + pud_index(address);
1451 	return (pud_t *) p4dp;
1452 }
1453 #define pud_offset_lockless pud_offset_lockless
1454 
1455 static inline pud_t *pud_offset(p4d_t *p4dp, unsigned long address)
1456 {
1457 	return pud_offset_lockless(p4dp, *p4dp, address);
1458 }
1459 #define pud_offset pud_offset
1460 
1461 static inline pmd_t *pmd_offset_lockless(pud_t *pudp, pud_t pud, unsigned long address)
1462 {
1463 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R3)
1464 		return (pmd_t *) pud_deref(pud) + pmd_index(address);
1465 	return (pmd_t *) pudp;
1466 }
1467 #define pmd_offset_lockless pmd_offset_lockless
1468 
1469 static inline pmd_t *pmd_offset(pud_t *pudp, unsigned long address)
1470 {
1471 	return pmd_offset_lockless(pudp, *pudp, address);
1472 }
1473 #define pmd_offset pmd_offset
1474 
1475 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
1476 {
1477 	return (unsigned long) pmd_deref(pmd);
1478 }
1479 
1480 static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
1481 {
1482 	return end <= current->mm->context.asce_limit;
1483 }
1484 #define gup_fast_permitted gup_fast_permitted
1485 
1486 #define pfn_pte(pfn, pgprot)	mk_pte_phys(((pfn) << PAGE_SHIFT), (pgprot))
1487 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1488 #define pte_page(x) pfn_to_page(pte_pfn(x))
1489 
1490 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1491 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1492 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
1493 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
1494 
1495 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1496 {
1497 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1498 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1499 }
1500 
1501 static inline pmd_t pmd_mkwrite_novma(pmd_t pmd)
1502 {
1503 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1504 	if (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY)
1505 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1506 	return pmd;
1507 }
1508 
1509 static inline pmd_t pmd_mkclean(pmd_t pmd)
1510 {
1511 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY));
1512 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1513 }
1514 
1515 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1516 {
1517 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_SOFT_DIRTY));
1518 	if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1519 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1520 	return pmd;
1521 }
1522 
1523 static inline pud_t pud_wrprotect(pud_t pud)
1524 {
1525 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1526 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1527 }
1528 
1529 static inline pud_t pud_mkwrite(pud_t pud)
1530 {
1531 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1532 	if (pud_val(pud) & _REGION3_ENTRY_DIRTY)
1533 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1534 	return pud;
1535 }
1536 
1537 static inline pud_t pud_mkclean(pud_t pud)
1538 {
1539 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY));
1540 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1541 }
1542 
1543 static inline pud_t pud_mkdirty(pud_t pud)
1544 {
1545 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY | _REGION3_ENTRY_SOFT_DIRTY));
1546 	if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1547 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1548 	return pud;
1549 }
1550 
1551 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1552 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1553 {
1554 	/*
1555 	 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1556 	 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1557 	 */
1558 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1559 		return pgprot_val(SEGMENT_NONE);
1560 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1561 		return pgprot_val(SEGMENT_RO);
1562 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1563 		return pgprot_val(SEGMENT_RX);
1564 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1565 		return pgprot_val(SEGMENT_RW);
1566 	return pgprot_val(SEGMENT_RWX);
1567 }
1568 
1569 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1570 {
1571 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1572 	if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1573 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1574 	return pmd;
1575 }
1576 
1577 static inline pmd_t pmd_mkold(pmd_t pmd)
1578 {
1579 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1580 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1581 }
1582 
1583 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1584 {
1585 	unsigned long mask;
1586 
1587 	mask  = _SEGMENT_ENTRY_ORIGIN_LARGE;
1588 	mask |= _SEGMENT_ENTRY_DIRTY;
1589 	mask |= _SEGMENT_ENTRY_YOUNG;
1590 	mask |=	_SEGMENT_ENTRY_LARGE;
1591 	mask |= _SEGMENT_ENTRY_SOFT_DIRTY;
1592 	pmd = __pmd(pmd_val(pmd) & mask);
1593 	pmd = set_pmd_bit(pmd, __pgprot(massage_pgprot_pmd(newprot)));
1594 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1595 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1596 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1597 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1598 	return pmd;
1599 }
1600 
1601 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1602 {
1603 	return __pmd(physpage + massage_pgprot_pmd(pgprot));
1604 }
1605 
1606 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1607 
1608 static inline void __pmdp_csp(pmd_t *pmdp)
1609 {
1610 	csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1611 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1612 }
1613 
1614 #define IDTE_GLOBAL	0
1615 #define IDTE_LOCAL	1
1616 
1617 #define IDTE_PTOA	0x0800
1618 #define IDTE_NODAT	0x1000
1619 #define IDTE_GUEST_ASCE	0x2000
1620 
1621 static __always_inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1622 					unsigned long opt, unsigned long asce,
1623 					int local)
1624 {
1625 	unsigned long sto;
1626 
1627 	sto = __pa(pmdp) - pmd_index(addr) * sizeof(pmd_t);
1628 	if (__builtin_constant_p(opt) && opt == 0) {
1629 		/* flush without guest asce */
1630 		asm volatile(
1631 			"	idte	%[r1],0,%[r2],%[m4]"
1632 			: "+m" (*pmdp)
1633 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1634 			  [m4] "i" (local)
1635 			: "cc" );
1636 	} else {
1637 		/* flush with guest asce */
1638 		asm volatile(
1639 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1640 			: "+m" (*pmdp)
1641 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1642 			  [r3] "a" (asce), [m4] "i" (local)
1643 			: "cc" );
1644 	}
1645 }
1646 
1647 static __always_inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1648 					unsigned long opt, unsigned long asce,
1649 					int local)
1650 {
1651 	unsigned long r3o;
1652 
1653 	r3o = __pa(pudp) - pud_index(addr) * sizeof(pud_t);
1654 	r3o |= _ASCE_TYPE_REGION3;
1655 	if (__builtin_constant_p(opt) && opt == 0) {
1656 		/* flush without guest asce */
1657 		asm volatile(
1658 			"	idte	%[r1],0,%[r2],%[m4]"
1659 			: "+m" (*pudp)
1660 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1661 			  [m4] "i" (local)
1662 			: "cc");
1663 	} else {
1664 		/* flush with guest asce */
1665 		asm volatile(
1666 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1667 			: "+m" (*pudp)
1668 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1669 			  [r3] "a" (asce), [m4] "i" (local)
1670 			: "cc" );
1671 	}
1672 }
1673 
1674 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1675 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1676 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1677 
1678 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1679 
1680 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1681 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1682 				pgtable_t pgtable);
1683 
1684 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1685 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1686 
1687 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1688 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1689 					unsigned long addr, pmd_t *pmdp,
1690 					pmd_t entry, int dirty)
1691 {
1692 	VM_BUG_ON(addr & ~HPAGE_MASK);
1693 
1694 	entry = pmd_mkyoung(entry);
1695 	if (dirty)
1696 		entry = pmd_mkdirty(entry);
1697 	if (pmd_val(*pmdp) == pmd_val(entry))
1698 		return 0;
1699 	pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1700 	return 1;
1701 }
1702 
1703 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1704 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1705 					    unsigned long addr, pmd_t *pmdp)
1706 {
1707 	pmd_t pmd = *pmdp;
1708 
1709 	pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1710 	return pmd_young(pmd);
1711 }
1712 
1713 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1714 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1715 					 unsigned long addr, pmd_t *pmdp)
1716 {
1717 	VM_BUG_ON(addr & ~HPAGE_MASK);
1718 	return pmdp_test_and_clear_young(vma, addr, pmdp);
1719 }
1720 
1721 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1722 			      pmd_t *pmdp, pmd_t entry)
1723 {
1724 	if (!MACHINE_HAS_NX)
1725 		entry = clear_pmd_bit(entry, __pgprot(_SEGMENT_ENTRY_NOEXEC));
1726 	set_pmd(pmdp, entry);
1727 }
1728 
1729 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1730 {
1731 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_LARGE));
1732 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1733 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1734 }
1735 
1736 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1737 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1738 					    unsigned long addr, pmd_t *pmdp)
1739 {
1740 	return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1741 }
1742 
1743 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1744 static inline pmd_t pmdp_huge_get_and_clear_full(struct vm_area_struct *vma,
1745 						 unsigned long addr,
1746 						 pmd_t *pmdp, int full)
1747 {
1748 	if (full) {
1749 		pmd_t pmd = *pmdp;
1750 		set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1751 		return pmd;
1752 	}
1753 	return pmdp_xchg_lazy(vma->vm_mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1754 }
1755 
1756 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1757 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1758 					  unsigned long addr, pmd_t *pmdp)
1759 {
1760 	return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1761 }
1762 
1763 #define __HAVE_ARCH_PMDP_INVALIDATE
1764 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1765 				   unsigned long addr, pmd_t *pmdp)
1766 {
1767 	pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1768 
1769 	return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1770 }
1771 
1772 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1773 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1774 				      unsigned long addr, pmd_t *pmdp)
1775 {
1776 	pmd_t pmd = *pmdp;
1777 
1778 	if (pmd_write(pmd))
1779 		pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1780 }
1781 
1782 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1783 					unsigned long address,
1784 					pmd_t *pmdp)
1785 {
1786 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1787 }
1788 #define pmdp_collapse_flush pmdp_collapse_flush
1789 
1790 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(((pfn) << PAGE_SHIFT), (pgprot))
1791 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1792 
1793 static inline int pmd_trans_huge(pmd_t pmd)
1794 {
1795 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1796 }
1797 
1798 #define has_transparent_hugepage has_transparent_hugepage
1799 static inline int has_transparent_hugepage(void)
1800 {
1801 	return MACHINE_HAS_EDAT1 ? 1 : 0;
1802 }
1803 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1804 
1805 /*
1806  * 64 bit swap entry format:
1807  * A page-table entry has some bits we have to treat in a special way.
1808  * Bits 54 and 63 are used to indicate the page type. Bit 53 marks the pte
1809  * as invalid.
1810  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1811  * |			  offset			|E11XX|type |S0|
1812  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1813  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1814  *
1815  * Bits 0-51 store the offset.
1816  * Bit 52 (E) is used to remember PG_anon_exclusive.
1817  * Bits 57-61 store the type.
1818  * Bit 62 (S) is used for softdirty tracking.
1819  * Bits 55 and 56 (X) are unused.
1820  */
1821 
1822 #define __SWP_OFFSET_MASK	((1UL << 52) - 1)
1823 #define __SWP_OFFSET_SHIFT	12
1824 #define __SWP_TYPE_MASK		((1UL << 5) - 1)
1825 #define __SWP_TYPE_SHIFT	2
1826 
1827 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1828 {
1829 	unsigned long pteval;
1830 
1831 	pteval = _PAGE_INVALID | _PAGE_PROTECT;
1832 	pteval |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1833 	pteval |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1834 	return __pte(pteval);
1835 }
1836 
1837 static inline unsigned long __swp_type(swp_entry_t entry)
1838 {
1839 	return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1840 }
1841 
1842 static inline unsigned long __swp_offset(swp_entry_t entry)
1843 {
1844 	return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1845 }
1846 
1847 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1848 {
1849 	return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1850 }
1851 
1852 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1853 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1854 
1855 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1856 extern void vmem_remove_mapping(unsigned long start, unsigned long size);
1857 extern int __vmem_map_4k_page(unsigned long addr, unsigned long phys, pgprot_t prot, bool alloc);
1858 extern int vmem_map_4k_page(unsigned long addr, unsigned long phys, pgprot_t prot);
1859 extern void vmem_unmap_4k_page(unsigned long addr);
1860 extern pte_t *vmem_get_alloc_pte(unsigned long addr, bool alloc);
1861 extern int s390_enable_sie(void);
1862 extern int s390_enable_skey(void);
1863 extern void s390_reset_cmma(struct mm_struct *mm);
1864 
1865 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1866 #define HAVE_ARCH_UNMAPPED_AREA
1867 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1868 
1869 #define pmd_pgtable(pmd) \
1870 	((pgtable_t)__va(pmd_val(pmd) & -sizeof(pte_t)*PTRS_PER_PTE))
1871 
1872 #endif /* _S390_PAGE_H */
1873