xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision 99a15348)
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 void paging_init(void);
27 extern unsigned long s390_invalid_asce;
28 
29 enum {
30 	PG_DIRECT_MAP_4K = 0,
31 	PG_DIRECT_MAP_1M,
32 	PG_DIRECT_MAP_2G,
33 	PG_DIRECT_MAP_MAX
34 };
35 
36 extern atomic_long_t direct_pages_count[PG_DIRECT_MAP_MAX];
37 
38 static inline void update_page_count(int level, long count)
39 {
40 	if (IS_ENABLED(CONFIG_PROC_FS))
41 		atomic_long_add(count, &direct_pages_count[level]);
42 }
43 
44 struct seq_file;
45 void arch_report_meminfo(struct seq_file *m);
46 
47 /*
48  * The S390 doesn't have any external MMU info: the kernel page
49  * tables contain all the necessary information.
50  */
51 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
52 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
53 
54 /*
55  * ZERO_PAGE is a global shared page that is always zero; used
56  * for zero-mapped memory areas etc..
57  */
58 
59 extern unsigned long empty_zero_page;
60 extern unsigned long zero_page_mask;
61 
62 #define ZERO_PAGE(vaddr) \
63 	(virt_to_page((void *)(empty_zero_page + \
64 	 (((unsigned long)(vaddr)) &zero_page_mask))))
65 #define __HAVE_COLOR_ZERO_PAGE
66 
67 /* TODO: s390 cannot support io_remap_pfn_range... */
68 
69 #define pte_ERROR(e) \
70 	pr_err("%s:%d: bad pte %016lx.\n", __FILE__, __LINE__, pte_val(e))
71 #define pmd_ERROR(e) \
72 	pr_err("%s:%d: bad pmd %016lx.\n", __FILE__, __LINE__, pmd_val(e))
73 #define pud_ERROR(e) \
74 	pr_err("%s:%d: bad pud %016lx.\n", __FILE__, __LINE__, pud_val(e))
75 #define p4d_ERROR(e) \
76 	pr_err("%s:%d: bad p4d %016lx.\n", __FILE__, __LINE__, p4d_val(e))
77 #define pgd_ERROR(e) \
78 	pr_err("%s:%d: bad pgd %016lx.\n", __FILE__, __LINE__, pgd_val(e))
79 
80 /*
81  * The vmalloc and module area will always be on the topmost area of the
82  * kernel mapping. 512GB are reserved for vmalloc by default.
83  * At the top of the vmalloc area a 2GB area is reserved where modules
84  * will reside. That makes sure that inter module branches always
85  * happen without trampolines and in addition the placement within a
86  * 2GB frame is branch prediction unit friendly.
87  */
88 extern unsigned long __bootdata_preserved(VMALLOC_START);
89 extern unsigned long __bootdata_preserved(VMALLOC_END);
90 #define VMALLOC_DEFAULT_SIZE	((512UL << 30) - MODULES_LEN)
91 extern struct page *__bootdata_preserved(vmemmap);
92 extern unsigned long __bootdata_preserved(vmemmap_size);
93 
94 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
95 
96 extern unsigned long __bootdata_preserved(MODULES_VADDR);
97 extern unsigned long __bootdata_preserved(MODULES_END);
98 #define MODULES_VADDR	MODULES_VADDR
99 #define MODULES_END	MODULES_END
100 #define MODULES_LEN	(1UL << 31)
101 
102 static inline int is_module_addr(void *addr)
103 {
104 	BUILD_BUG_ON(MODULES_LEN > (1UL << 31));
105 	if (addr < (void *)MODULES_VADDR)
106 		return 0;
107 	if (addr > (void *)MODULES_END)
108 		return 0;
109 	return 1;
110 }
111 
112 /*
113  * A 64 bit pagetable entry of S390 has following format:
114  * |			 PFRA			      |0IPC|  OS  |
115  * 0000000000111111111122222222223333333333444444444455555555556666
116  * 0123456789012345678901234567890123456789012345678901234567890123
117  *
118  * I Page-Invalid Bit:    Page is not available for address-translation
119  * P Page-Protection Bit: Store access not possible for page
120  * C Change-bit override: HW is not required to set change bit
121  *
122  * A 64 bit segmenttable entry of S390 has following format:
123  * |        P-table origin                              |      TT
124  * 0000000000111111111122222222223333333333444444444455555555556666
125  * 0123456789012345678901234567890123456789012345678901234567890123
126  *
127  * I Segment-Invalid Bit:    Segment is not available for address-translation
128  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
129  * P Page-Protection Bit: Store access not possible for page
130  * TT Type 00
131  *
132  * A 64 bit region table entry of S390 has following format:
133  * |        S-table origin                             |   TF  TTTL
134  * 0000000000111111111122222222223333333333444444444455555555556666
135  * 0123456789012345678901234567890123456789012345678901234567890123
136  *
137  * I Segment-Invalid Bit:    Segment is not available for address-translation
138  * TT Type 01
139  * TF
140  * TL Table length
141  *
142  * The 64 bit regiontable origin of S390 has following format:
143  * |      region table origon                          |       DTTL
144  * 0000000000111111111122222222223333333333444444444455555555556666
145  * 0123456789012345678901234567890123456789012345678901234567890123
146  *
147  * X Space-Switch event:
148  * G Segment-Invalid Bit:
149  * P Private-Space Bit:
150  * S Storage-Alteration:
151  * R Real space
152  * TL Table-Length:
153  *
154  * A storage key has the following format:
155  * | ACC |F|R|C|0|
156  *  0   3 4 5 6 7
157  * ACC: access key
158  * F  : fetch protection bit
159  * R  : referenced bit
160  * C  : changed bit
161  */
162 
163 /* Hardware bits in the page table entry */
164 #define _PAGE_NOEXEC	0x100		/* HW no-execute bit  */
165 #define _PAGE_PROTECT	0x200		/* HW read-only bit  */
166 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
167 #define _PAGE_LARGE	0x800		/* Bit to mark a large pte */
168 
169 /* Software bits in the page table entry */
170 #define _PAGE_PRESENT	0x001		/* SW pte present bit */
171 #define _PAGE_YOUNG	0x004		/* SW pte young bit */
172 #define _PAGE_DIRTY	0x008		/* SW pte dirty bit */
173 #define _PAGE_READ	0x010		/* SW pte read bit */
174 #define _PAGE_WRITE	0x020		/* SW pte write bit */
175 #define _PAGE_SPECIAL	0x040		/* SW associated with special page */
176 #define _PAGE_UNUSED	0x080		/* SW bit for pgste usage state */
177 
178 #ifdef CONFIG_MEM_SOFT_DIRTY
179 #define _PAGE_SOFT_DIRTY 0x002		/* SW pte soft dirty bit */
180 #else
181 #define _PAGE_SOFT_DIRTY 0x000
182 #endif
183 
184 /* Set of bits not changed in pte_modify */
185 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
186 				 _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
187 
188 /*
189  * handle_pte_fault uses pte_present and pte_none to find out the pte type
190  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
191  * distinguish present from not-present ptes. It is changed only with the page
192  * table lock held.
193  *
194  * The following table gives the different possible bit combinations for
195  * the pte hardware and software bits in the last 12 bits of a pte
196  * (. unassigned bit, x don't care, t swap type):
197  *
198  *				842100000000
199  *				000084210000
200  *				000000008421
201  *				.IR.uswrdy.p
202  * empty			.10.00000000
203  * swap				.11..ttttt.0
204  * prot-none, clean, old	.11.xx0000.1
205  * prot-none, clean, young	.11.xx0001.1
206  * prot-none, dirty, old	.11.xx0010.1
207  * prot-none, dirty, young	.11.xx0011.1
208  * read-only, clean, old	.11.xx0100.1
209  * read-only, clean, young	.01.xx0101.1
210  * read-only, dirty, old	.11.xx0110.1
211  * read-only, dirty, young	.01.xx0111.1
212  * read-write, clean, old	.11.xx1100.1
213  * read-write, clean, young	.01.xx1101.1
214  * read-write, dirty, old	.10.xx1110.1
215  * read-write, dirty, young	.00.xx1111.1
216  * HW-bits: R read-only, I invalid
217  * SW-bits: p present, y young, d dirty, r read, w write, s special,
218  *	    u unused, l large
219  *
220  * pte_none    is true for the bit pattern .10.00000000, pte == 0x400
221  * pte_swap    is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
222  * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
223  */
224 
225 /* Bits in the segment/region table address-space-control-element */
226 #define _ASCE_ORIGIN		~0xfffUL/* region/segment table origin	    */
227 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
228 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
229 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
230 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
231 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
232 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
233 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
234 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
235 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
236 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
237 
238 /* Bits in the region table entry */
239 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
240 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
241 #define _REGION_ENTRY_NOEXEC	0x100	/* region no-execute bit	    */
242 #define _REGION_ENTRY_OFFSET	0xc0	/* region table offset		    */
243 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
244 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region table type mask	    */
245 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
246 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
247 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
248 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
249 
250 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
251 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
252 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
253 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
254 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
255 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
256 
257 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address	     */
258 #define _REGION3_ENTRY_DIRTY	0x2000	/* SW region dirty bit */
259 #define _REGION3_ENTRY_YOUNG	0x1000	/* SW region young bit */
260 #define _REGION3_ENTRY_LARGE	0x0400	/* RTTE-format control, large page  */
261 #define _REGION3_ENTRY_READ	0x0002	/* SW region read bit */
262 #define _REGION3_ENTRY_WRITE	0x0001	/* SW region write bit */
263 
264 #ifdef CONFIG_MEM_SOFT_DIRTY
265 #define _REGION3_ENTRY_SOFT_DIRTY 0x4000 /* SW region soft dirty bit */
266 #else
267 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
268 #endif
269 
270 #define _REGION_ENTRY_BITS	 0xfffffffffffff22fUL
271 
272 /* Bits in the segment table entry */
273 #define _SEGMENT_ENTRY_BITS			0xfffffffffffffe33UL
274 #define _SEGMENT_ENTRY_HARDWARE_BITS		0xfffffffffffffe30UL
275 #define _SEGMENT_ENTRY_HARDWARE_BITS_LARGE	0xfffffffffff00730UL
276 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
277 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* page table origin		    */
278 #define _SEGMENT_ENTRY_PROTECT	0x200	/* segment protection bit	    */
279 #define _SEGMENT_ENTRY_NOEXEC	0x100	/* segment no-execute bit	    */
280 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
281 #define _SEGMENT_ENTRY_TYPE_MASK 0x0c	/* segment table type mask	    */
282 
283 #define _SEGMENT_ENTRY		(0)
284 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
285 
286 #define _SEGMENT_ENTRY_DIRTY	0x2000	/* SW segment dirty bit */
287 #define _SEGMENT_ENTRY_YOUNG	0x1000	/* SW segment young bit */
288 #define _SEGMENT_ENTRY_LARGE	0x0400	/* STE-format control, large page */
289 #define _SEGMENT_ENTRY_WRITE	0x0002	/* SW segment write bit */
290 #define _SEGMENT_ENTRY_READ	0x0001	/* SW segment read bit */
291 
292 #ifdef CONFIG_MEM_SOFT_DIRTY
293 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x4000 /* SW segment soft dirty bit */
294 #else
295 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
296 #endif
297 
298 #define _CRST_ENTRIES	2048	/* number of region/segment table entries */
299 #define _PAGE_ENTRIES	256	/* number of page table entries	*/
300 
301 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
302 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
303 
304 #define _REGION1_SHIFT	53
305 #define _REGION2_SHIFT	42
306 #define _REGION3_SHIFT	31
307 #define _SEGMENT_SHIFT	20
308 
309 #define _REGION1_INDEX	(0x7ffUL << _REGION1_SHIFT)
310 #define _REGION2_INDEX	(0x7ffUL << _REGION2_SHIFT)
311 #define _REGION3_INDEX	(0x7ffUL << _REGION3_SHIFT)
312 #define _SEGMENT_INDEX	(0x7ffUL << _SEGMENT_SHIFT)
313 #define _PAGE_INDEX	(0xffUL  << _PAGE_SHIFT)
314 
315 #define _REGION1_SIZE	(1UL << _REGION1_SHIFT)
316 #define _REGION2_SIZE	(1UL << _REGION2_SHIFT)
317 #define _REGION3_SIZE	(1UL << _REGION3_SHIFT)
318 #define _SEGMENT_SIZE	(1UL << _SEGMENT_SHIFT)
319 
320 #define _REGION1_MASK	(~(_REGION1_SIZE - 1))
321 #define _REGION2_MASK	(~(_REGION2_SIZE - 1))
322 #define _REGION3_MASK	(~(_REGION3_SIZE - 1))
323 #define _SEGMENT_MASK	(~(_SEGMENT_SIZE - 1))
324 
325 #define PMD_SHIFT	_SEGMENT_SHIFT
326 #define PUD_SHIFT	_REGION3_SHIFT
327 #define P4D_SHIFT	_REGION2_SHIFT
328 #define PGDIR_SHIFT	_REGION1_SHIFT
329 
330 #define PMD_SIZE	_SEGMENT_SIZE
331 #define PUD_SIZE	_REGION3_SIZE
332 #define P4D_SIZE	_REGION2_SIZE
333 #define PGDIR_SIZE	_REGION1_SIZE
334 
335 #define PMD_MASK	_SEGMENT_MASK
336 #define PUD_MASK	_REGION3_MASK
337 #define P4D_MASK	_REGION2_MASK
338 #define PGDIR_MASK	_REGION1_MASK
339 
340 #define PTRS_PER_PTE	_PAGE_ENTRIES
341 #define PTRS_PER_PMD	_CRST_ENTRIES
342 #define PTRS_PER_PUD	_CRST_ENTRIES
343 #define PTRS_PER_P4D	_CRST_ENTRIES
344 #define PTRS_PER_PGD	_CRST_ENTRIES
345 
346 /*
347  * Segment table and region3 table entry encoding
348  * (R = read-only, I = invalid, y = young bit):
349  *				dy..R...I...wr
350  * prot-none, clean, old	00..1...1...00
351  * prot-none, clean, young	01..1...1...00
352  * prot-none, dirty, old	10..1...1...00
353  * prot-none, dirty, young	11..1...1...00
354  * read-only, clean, old	00..1...1...01
355  * read-only, clean, young	01..1...0...01
356  * read-only, dirty, old	10..1...1...01
357  * read-only, dirty, young	11..1...0...01
358  * read-write, clean, old	00..1...1...11
359  * read-write, clean, young	01..1...0...11
360  * read-write, dirty, old	10..0...1...11
361  * read-write, dirty, young	11..0...0...11
362  * The segment table origin is used to distinguish empty (origin==0) from
363  * read-write, old segment table entries (origin!=0)
364  * HW-bits: R read-only, I invalid
365  * SW-bits: y young, d dirty, r read, w write
366  */
367 
368 /* Page status table bits for virtualization */
369 #define PGSTE_ACC_BITS	0xf000000000000000UL
370 #define PGSTE_FP_BIT	0x0800000000000000UL
371 #define PGSTE_PCL_BIT	0x0080000000000000UL
372 #define PGSTE_HR_BIT	0x0040000000000000UL
373 #define PGSTE_HC_BIT	0x0020000000000000UL
374 #define PGSTE_GR_BIT	0x0004000000000000UL
375 #define PGSTE_GC_BIT	0x0002000000000000UL
376 #define PGSTE_UC_BIT	0x0000800000000000UL	/* user dirty (migration) */
377 #define PGSTE_IN_BIT	0x0000400000000000UL	/* IPTE notify bit */
378 #define PGSTE_VSIE_BIT	0x0000200000000000UL	/* ref'd in a shadow table */
379 
380 /* Guest Page State used for virtualization */
381 #define _PGSTE_GPS_ZERO			0x0000000080000000UL
382 #define _PGSTE_GPS_NODAT		0x0000000040000000UL
383 #define _PGSTE_GPS_USAGE_MASK		0x0000000003000000UL
384 #define _PGSTE_GPS_USAGE_STABLE		0x0000000000000000UL
385 #define _PGSTE_GPS_USAGE_UNUSED		0x0000000001000000UL
386 #define _PGSTE_GPS_USAGE_POT_VOLATILE	0x0000000002000000UL
387 #define _PGSTE_GPS_USAGE_VOLATILE	_PGSTE_GPS_USAGE_MASK
388 
389 /*
390  * A user page table pointer has the space-switch-event bit, the
391  * private-space-control bit and the storage-alteration-event-control
392  * bit set. A kernel page table pointer doesn't need them.
393  */
394 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
395 				 _ASCE_ALT_EVENT)
396 
397 /*
398  * Page protection definitions.
399  */
400 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID | _PAGE_PROTECT)
401 #define PAGE_RO		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
402 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
403 #define PAGE_RX		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
404 				 _PAGE_INVALID | _PAGE_PROTECT)
405 #define PAGE_RW		__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
406 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
407 #define PAGE_RWX	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
408 				 _PAGE_INVALID | _PAGE_PROTECT)
409 
410 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
411 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
412 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
413 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
414 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
415 				 _PAGE_PROTECT | _PAGE_NOEXEC)
416 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
417 				  _PAGE_YOUNG |	_PAGE_DIRTY)
418 
419 /*
420  * On s390 the page table entry has an invalid bit and a read-only bit.
421  * Read permission implies execute permission and write permission
422  * implies read permission.
423  */
424          /*xwr*/
425 #define __P000	PAGE_NONE
426 #define __P001	PAGE_RO
427 #define __P010	PAGE_RO
428 #define __P011	PAGE_RO
429 #define __P100	PAGE_RX
430 #define __P101	PAGE_RX
431 #define __P110	PAGE_RX
432 #define __P111	PAGE_RX
433 
434 #define __S000	PAGE_NONE
435 #define __S001	PAGE_RO
436 #define __S010	PAGE_RW
437 #define __S011	PAGE_RW
438 #define __S100	PAGE_RX
439 #define __S101	PAGE_RX
440 #define __S110	PAGE_RWX
441 #define __S111	PAGE_RWX
442 
443 /*
444  * Segment entry (large page) protection definitions.
445  */
446 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_INVALID | \
447 				 _SEGMENT_ENTRY_PROTECT)
448 #define SEGMENT_RO	__pgprot(_SEGMENT_ENTRY_PROTECT | \
449 				 _SEGMENT_ENTRY_READ | \
450 				 _SEGMENT_ENTRY_NOEXEC)
451 #define SEGMENT_RX	__pgprot(_SEGMENT_ENTRY_PROTECT | \
452 				 _SEGMENT_ENTRY_READ)
453 #define SEGMENT_RW	__pgprot(_SEGMENT_ENTRY_READ | \
454 				 _SEGMENT_ENTRY_WRITE | \
455 				 _SEGMENT_ENTRY_NOEXEC)
456 #define SEGMENT_RWX	__pgprot(_SEGMENT_ENTRY_READ | \
457 				 _SEGMENT_ENTRY_WRITE)
458 #define SEGMENT_KERNEL	__pgprot(_SEGMENT_ENTRY |	\
459 				 _SEGMENT_ENTRY_LARGE |	\
460 				 _SEGMENT_ENTRY_READ |	\
461 				 _SEGMENT_ENTRY_WRITE | \
462 				 _SEGMENT_ENTRY_YOUNG | \
463 				 _SEGMENT_ENTRY_DIRTY | \
464 				 _SEGMENT_ENTRY_NOEXEC)
465 #define SEGMENT_KERNEL_RO __pgprot(_SEGMENT_ENTRY |	\
466 				 _SEGMENT_ENTRY_LARGE |	\
467 				 _SEGMENT_ENTRY_READ |	\
468 				 _SEGMENT_ENTRY_YOUNG |	\
469 				 _SEGMENT_ENTRY_PROTECT | \
470 				 _SEGMENT_ENTRY_NOEXEC)
471 #define SEGMENT_KERNEL_EXEC __pgprot(_SEGMENT_ENTRY |	\
472 				 _SEGMENT_ENTRY_LARGE |	\
473 				 _SEGMENT_ENTRY_READ |	\
474 				 _SEGMENT_ENTRY_WRITE | \
475 				 _SEGMENT_ENTRY_YOUNG |	\
476 				 _SEGMENT_ENTRY_DIRTY)
477 
478 /*
479  * Region3 entry (large page) protection definitions.
480  */
481 
482 #define REGION3_KERNEL	__pgprot(_REGION_ENTRY_TYPE_R3 | \
483 				 _REGION3_ENTRY_LARGE |	 \
484 				 _REGION3_ENTRY_READ |	 \
485 				 _REGION3_ENTRY_WRITE |	 \
486 				 _REGION3_ENTRY_YOUNG |	 \
487 				 _REGION3_ENTRY_DIRTY | \
488 				 _REGION_ENTRY_NOEXEC)
489 #define REGION3_KERNEL_RO __pgprot(_REGION_ENTRY_TYPE_R3 | \
490 				   _REGION3_ENTRY_LARGE |  \
491 				   _REGION3_ENTRY_READ |   \
492 				   _REGION3_ENTRY_YOUNG |  \
493 				   _REGION_ENTRY_PROTECT | \
494 				   _REGION_ENTRY_NOEXEC)
495 
496 static inline bool mm_p4d_folded(struct mm_struct *mm)
497 {
498 	return mm->context.asce_limit <= _REGION1_SIZE;
499 }
500 #define mm_p4d_folded(mm) mm_p4d_folded(mm)
501 
502 static inline bool mm_pud_folded(struct mm_struct *mm)
503 {
504 	return mm->context.asce_limit <= _REGION2_SIZE;
505 }
506 #define mm_pud_folded(mm) mm_pud_folded(mm)
507 
508 static inline bool mm_pmd_folded(struct mm_struct *mm)
509 {
510 	return mm->context.asce_limit <= _REGION3_SIZE;
511 }
512 #define mm_pmd_folded(mm) mm_pmd_folded(mm)
513 
514 static inline int mm_has_pgste(struct mm_struct *mm)
515 {
516 #ifdef CONFIG_PGSTE
517 	if (unlikely(mm->context.has_pgste))
518 		return 1;
519 #endif
520 	return 0;
521 }
522 
523 static inline int mm_is_protected(struct mm_struct *mm)
524 {
525 #ifdef CONFIG_PGSTE
526 	if (unlikely(atomic_read(&mm->context.is_protected)))
527 		return 1;
528 #endif
529 	return 0;
530 }
531 
532 static inline int mm_alloc_pgste(struct mm_struct *mm)
533 {
534 #ifdef CONFIG_PGSTE
535 	if (unlikely(mm->context.alloc_pgste))
536 		return 1;
537 #endif
538 	return 0;
539 }
540 
541 static inline pte_t clear_pte_bit(pte_t pte, pgprot_t prot)
542 {
543 	return __pte(pte_val(pte) & ~pgprot_val(prot));
544 }
545 
546 static inline pte_t set_pte_bit(pte_t pte, pgprot_t prot)
547 {
548 	return __pte(pte_val(pte) | pgprot_val(prot));
549 }
550 
551 static inline pmd_t clear_pmd_bit(pmd_t pmd, pgprot_t prot)
552 {
553 	return __pmd(pmd_val(pmd) & ~pgprot_val(prot));
554 }
555 
556 static inline pmd_t set_pmd_bit(pmd_t pmd, pgprot_t prot)
557 {
558 	return __pmd(pmd_val(pmd) | pgprot_val(prot));
559 }
560 
561 static inline pud_t clear_pud_bit(pud_t pud, pgprot_t prot)
562 {
563 	return __pud(pud_val(pud) & ~pgprot_val(prot));
564 }
565 
566 static inline pud_t set_pud_bit(pud_t pud, pgprot_t prot)
567 {
568 	return __pud(pud_val(pud) | pgprot_val(prot));
569 }
570 
571 /*
572  * In the case that a guest uses storage keys
573  * faults should no longer be backed by zero pages
574  */
575 #define mm_forbids_zeropage mm_has_pgste
576 static inline int mm_uses_skeys(struct mm_struct *mm)
577 {
578 #ifdef CONFIG_PGSTE
579 	if (mm->context.uses_skeys)
580 		return 1;
581 #endif
582 	return 0;
583 }
584 
585 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
586 {
587 	union register_pair r1 = { .even = old, .odd = new, };
588 	unsigned long address = (unsigned long)ptr | 1;
589 
590 	asm volatile(
591 		"	csp	%[r1],%[address]"
592 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
593 		: [address] "d" (address)
594 		: "cc");
595 }
596 
597 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
598 {
599 	union register_pair r1 = { .even = old, .odd = new, };
600 	unsigned long address = (unsigned long)ptr | 1;
601 
602 	asm volatile(
603 		"	cspg	%[r1],%[address]"
604 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
605 		: [address] "d" (address)
606 		: "cc");
607 }
608 
609 #define CRDTE_DTT_PAGE		0x00UL
610 #define CRDTE_DTT_SEGMENT	0x10UL
611 #define CRDTE_DTT_REGION3	0x14UL
612 #define CRDTE_DTT_REGION2	0x18UL
613 #define CRDTE_DTT_REGION1	0x1cUL
614 
615 static inline void crdte(unsigned long old, unsigned long new,
616 			 unsigned long *table, unsigned long dtt,
617 			 unsigned long address, unsigned long asce)
618 {
619 	union register_pair r1 = { .even = old, .odd = new, };
620 	union register_pair r2 = { .even = __pa(table) | dtt, .odd = address, };
621 
622 	asm volatile(".insn rrf,0xb98f0000,%[r1],%[r2],%[asce],0"
623 		     : [r1] "+&d" (r1.pair)
624 		     : [r2] "d" (r2.pair), [asce] "a" (asce)
625 		     : "memory", "cc");
626 }
627 
628 /*
629  * pgd/p4d/pud/pmd/pte query functions
630  */
631 static inline int pgd_folded(pgd_t pgd)
632 {
633 	return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
634 }
635 
636 static inline int pgd_present(pgd_t pgd)
637 {
638 	if (pgd_folded(pgd))
639 		return 1;
640 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
641 }
642 
643 static inline int pgd_none(pgd_t pgd)
644 {
645 	if (pgd_folded(pgd))
646 		return 0;
647 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
648 }
649 
650 static inline int pgd_bad(pgd_t pgd)
651 {
652 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1)
653 		return 0;
654 	return (pgd_val(pgd) & ~_REGION_ENTRY_BITS) != 0;
655 }
656 
657 static inline unsigned long pgd_pfn(pgd_t pgd)
658 {
659 	unsigned long origin_mask;
660 
661 	origin_mask = _REGION_ENTRY_ORIGIN;
662 	return (pgd_val(pgd) & origin_mask) >> PAGE_SHIFT;
663 }
664 
665 static inline int p4d_folded(p4d_t p4d)
666 {
667 	return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
668 }
669 
670 static inline int p4d_present(p4d_t p4d)
671 {
672 	if (p4d_folded(p4d))
673 		return 1;
674 	return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
675 }
676 
677 static inline int p4d_none(p4d_t p4d)
678 {
679 	if (p4d_folded(p4d))
680 		return 0;
681 	return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
682 }
683 
684 static inline unsigned long p4d_pfn(p4d_t p4d)
685 {
686 	unsigned long origin_mask;
687 
688 	origin_mask = _REGION_ENTRY_ORIGIN;
689 	return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
690 }
691 
692 static inline int pud_folded(pud_t pud)
693 {
694 	return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
695 }
696 
697 static inline int pud_present(pud_t pud)
698 {
699 	if (pud_folded(pud))
700 		return 1;
701 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
702 }
703 
704 static inline int pud_none(pud_t pud)
705 {
706 	if (pud_folded(pud))
707 		return 0;
708 	return pud_val(pud) == _REGION3_ENTRY_EMPTY;
709 }
710 
711 #define pud_leaf	pud_large
712 static inline int pud_large(pud_t pud)
713 {
714 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
715 		return 0;
716 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
717 }
718 
719 #define pmd_leaf	pmd_large
720 static inline int pmd_large(pmd_t pmd)
721 {
722 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
723 }
724 
725 static inline int pmd_bad(pmd_t pmd)
726 {
727 	if ((pmd_val(pmd) & _SEGMENT_ENTRY_TYPE_MASK) > 0 || pmd_large(pmd))
728 		return 1;
729 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
730 }
731 
732 static inline int pud_bad(pud_t pud)
733 {
734 	unsigned long type = pud_val(pud) & _REGION_ENTRY_TYPE_MASK;
735 
736 	if (type > _REGION_ENTRY_TYPE_R3 || pud_large(pud))
737 		return 1;
738 	if (type < _REGION_ENTRY_TYPE_R3)
739 		return 0;
740 	return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
741 }
742 
743 static inline int p4d_bad(p4d_t p4d)
744 {
745 	unsigned long type = p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK;
746 
747 	if (type > _REGION_ENTRY_TYPE_R2)
748 		return 1;
749 	if (type < _REGION_ENTRY_TYPE_R2)
750 		return 0;
751 	return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
752 }
753 
754 static inline int pmd_present(pmd_t pmd)
755 {
756 	return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
757 }
758 
759 static inline int pmd_none(pmd_t pmd)
760 {
761 	return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
762 }
763 
764 #define pmd_write pmd_write
765 static inline int pmd_write(pmd_t pmd)
766 {
767 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
768 }
769 
770 #define pud_write pud_write
771 static inline int pud_write(pud_t pud)
772 {
773 	return (pud_val(pud) & _REGION3_ENTRY_WRITE) != 0;
774 }
775 
776 static inline int pmd_dirty(pmd_t pmd)
777 {
778 	return (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
779 }
780 
781 static inline int pmd_young(pmd_t pmd)
782 {
783 	return (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
784 }
785 
786 static inline int pte_present(pte_t pte)
787 {
788 	/* Bit pattern: (pte & 0x001) == 0x001 */
789 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
790 }
791 
792 static inline int pte_none(pte_t pte)
793 {
794 	/* Bit pattern: pte == 0x400 */
795 	return pte_val(pte) == _PAGE_INVALID;
796 }
797 
798 static inline int pte_swap(pte_t pte)
799 {
800 	/* Bit pattern: (pte & 0x201) == 0x200 */
801 	return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
802 		== _PAGE_PROTECT;
803 }
804 
805 static inline int pte_special(pte_t pte)
806 {
807 	return (pte_val(pte) & _PAGE_SPECIAL);
808 }
809 
810 #define __HAVE_ARCH_PTE_SAME
811 static inline int pte_same(pte_t a, pte_t b)
812 {
813 	return pte_val(a) == pte_val(b);
814 }
815 
816 #ifdef CONFIG_NUMA_BALANCING
817 static inline int pte_protnone(pte_t pte)
818 {
819 	return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
820 }
821 
822 static inline int pmd_protnone(pmd_t pmd)
823 {
824 	/* pmd_large(pmd) implies pmd_present(pmd) */
825 	return pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
826 }
827 #endif
828 
829 static inline int pte_soft_dirty(pte_t pte)
830 {
831 	return pte_val(pte) & _PAGE_SOFT_DIRTY;
832 }
833 #define pte_swp_soft_dirty pte_soft_dirty
834 
835 static inline pte_t pte_mksoft_dirty(pte_t pte)
836 {
837 	return set_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
838 }
839 #define pte_swp_mksoft_dirty pte_mksoft_dirty
840 
841 static inline pte_t pte_clear_soft_dirty(pte_t pte)
842 {
843 	return clear_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
844 }
845 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
846 
847 static inline int pmd_soft_dirty(pmd_t pmd)
848 {
849 	return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
850 }
851 
852 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
853 {
854 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
855 }
856 
857 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
858 {
859 	return clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
860 }
861 
862 /*
863  * query functions pte_write/pte_dirty/pte_young only work if
864  * pte_present() is true. Undefined behaviour if not..
865  */
866 static inline int pte_write(pte_t pte)
867 {
868 	return (pte_val(pte) & _PAGE_WRITE) != 0;
869 }
870 
871 static inline int pte_dirty(pte_t pte)
872 {
873 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
874 }
875 
876 static inline int pte_young(pte_t pte)
877 {
878 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
879 }
880 
881 #define __HAVE_ARCH_PTE_UNUSED
882 static inline int pte_unused(pte_t pte)
883 {
884 	return pte_val(pte) & _PAGE_UNUSED;
885 }
886 
887 /*
888  * Extract the pgprot value from the given pte while at the same time making it
889  * usable for kernel address space mappings where fault driven dirty and
890  * young/old accounting is not supported, i.e _PAGE_PROTECT and _PAGE_INVALID
891  * must not be set.
892  */
893 static inline pgprot_t pte_pgprot(pte_t pte)
894 {
895 	unsigned long pte_flags = pte_val(pte) & _PAGE_CHG_MASK;
896 
897 	if (pte_write(pte))
898 		pte_flags |= pgprot_val(PAGE_KERNEL);
899 	else
900 		pte_flags |= pgprot_val(PAGE_KERNEL_RO);
901 	pte_flags |= pte_val(pte) & mio_wb_bit_mask;
902 
903 	return __pgprot(pte_flags);
904 }
905 
906 /*
907  * pgd/pmd/pte modification functions
908  */
909 
910 static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
911 {
912 	WRITE_ONCE(*pgdp, pgd);
913 }
914 
915 static inline void set_p4d(p4d_t *p4dp, p4d_t p4d)
916 {
917 	WRITE_ONCE(*p4dp, p4d);
918 }
919 
920 static inline void set_pud(pud_t *pudp, pud_t pud)
921 {
922 	WRITE_ONCE(*pudp, pud);
923 }
924 
925 static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
926 {
927 	WRITE_ONCE(*pmdp, pmd);
928 }
929 
930 static inline void set_pte(pte_t *ptep, pte_t pte)
931 {
932 	WRITE_ONCE(*ptep, pte);
933 }
934 
935 static inline void pgd_clear(pgd_t *pgd)
936 {
937 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
938 		set_pgd(pgd, __pgd(_REGION1_ENTRY_EMPTY));
939 }
940 
941 static inline void p4d_clear(p4d_t *p4d)
942 {
943 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
944 		set_p4d(p4d, __p4d(_REGION2_ENTRY_EMPTY));
945 }
946 
947 static inline void pud_clear(pud_t *pud)
948 {
949 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
950 		set_pud(pud, __pud(_REGION3_ENTRY_EMPTY));
951 }
952 
953 static inline void pmd_clear(pmd_t *pmdp)
954 {
955 	set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
956 }
957 
958 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
959 {
960 	set_pte(ptep, __pte(_PAGE_INVALID));
961 }
962 
963 /*
964  * The following pte modification functions only work if
965  * pte_present() is true. Undefined behaviour if not..
966  */
967 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
968 {
969 	pte = clear_pte_bit(pte, __pgprot(~_PAGE_CHG_MASK));
970 	pte = set_pte_bit(pte, newprot);
971 	/*
972 	 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
973 	 * has the invalid bit set, clear it again for readable, young pages
974 	 */
975 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
976 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
977 	/*
978 	 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
979 	 * protection bit set, clear it again for writable, dirty pages
980 	 */
981 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
982 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
983 	return pte;
984 }
985 
986 static inline pte_t pte_wrprotect(pte_t pte)
987 {
988 	pte = clear_pte_bit(pte, __pgprot(_PAGE_WRITE));
989 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
990 }
991 
992 static inline pte_t pte_mkwrite(pte_t pte)
993 {
994 	pte = set_pte_bit(pte, __pgprot(_PAGE_WRITE));
995 	if (pte_val(pte) & _PAGE_DIRTY)
996 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
997 	return pte;
998 }
999 
1000 static inline pte_t pte_mkclean(pte_t pte)
1001 {
1002 	pte = clear_pte_bit(pte, __pgprot(_PAGE_DIRTY));
1003 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1004 }
1005 
1006 static inline pte_t pte_mkdirty(pte_t pte)
1007 {
1008 	pte = set_pte_bit(pte, __pgprot(_PAGE_DIRTY | _PAGE_SOFT_DIRTY));
1009 	if (pte_val(pte) & _PAGE_WRITE)
1010 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1011 	return pte;
1012 }
1013 
1014 static inline pte_t pte_mkold(pte_t pte)
1015 {
1016 	pte = clear_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1017 	return set_pte_bit(pte, __pgprot(_PAGE_INVALID));
1018 }
1019 
1020 static inline pte_t pte_mkyoung(pte_t pte)
1021 {
1022 	pte = set_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1023 	if (pte_val(pte) & _PAGE_READ)
1024 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
1025 	return pte;
1026 }
1027 
1028 static inline pte_t pte_mkspecial(pte_t pte)
1029 {
1030 	return set_pte_bit(pte, __pgprot(_PAGE_SPECIAL));
1031 }
1032 
1033 #ifdef CONFIG_HUGETLB_PAGE
1034 static inline pte_t pte_mkhuge(pte_t pte)
1035 {
1036 	return set_pte_bit(pte, __pgprot(_PAGE_LARGE));
1037 }
1038 #endif
1039 
1040 #define IPTE_GLOBAL	0
1041 #define	IPTE_LOCAL	1
1042 
1043 #define IPTE_NODAT	0x400
1044 #define IPTE_GUEST_ASCE	0x800
1045 
1046 static __always_inline void __ptep_ipte(unsigned long address, pte_t *ptep,
1047 					unsigned long opt, unsigned long asce,
1048 					int local)
1049 {
1050 	unsigned long pto = __pa(ptep);
1051 
1052 	if (__builtin_constant_p(opt) && opt == 0) {
1053 		/* Invalidation + TLB flush for the pte */
1054 		asm volatile(
1055 			"	ipte	%[r1],%[r2],0,%[m4]"
1056 			: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
1057 			  [m4] "i" (local));
1058 		return;
1059 	}
1060 
1061 	/* Invalidate ptes with options + TLB flush of the ptes */
1062 	opt = opt | (asce & _ASCE_ORIGIN);
1063 	asm volatile(
1064 		"	ipte	%[r1],%[r2],%[r3],%[m4]"
1065 		: [r2] "+a" (address), [r3] "+a" (opt)
1066 		: [r1] "a" (pto), [m4] "i" (local) : "memory");
1067 }
1068 
1069 static __always_inline void __ptep_ipte_range(unsigned long address, int nr,
1070 					      pte_t *ptep, int local)
1071 {
1072 	unsigned long pto = __pa(ptep);
1073 
1074 	/* Invalidate a range of ptes + TLB flush of the ptes */
1075 	do {
1076 		asm volatile(
1077 			"	ipte %[r1],%[r2],%[r3],%[m4]"
1078 			: [r2] "+a" (address), [r3] "+a" (nr)
1079 			: [r1] "a" (pto), [m4] "i" (local) : "memory");
1080 	} while (nr != 255);
1081 }
1082 
1083 /*
1084  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1085  * both clear the TLB for the unmapped pte. The reason is that
1086  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1087  * to modify an active pte. The sequence is
1088  *   1) ptep_get_and_clear
1089  *   2) set_pte_at
1090  *   3) flush_tlb_range
1091  * On s390 the tlb needs to get flushed with the modification of the pte
1092  * if the pte is active. The only way how this can be implemented is to
1093  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1094  * is a nop.
1095  */
1096 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1097 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1098 
1099 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1100 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1101 					    unsigned long addr, pte_t *ptep)
1102 {
1103 	pte_t pte = *ptep;
1104 
1105 	pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1106 	return pte_young(pte);
1107 }
1108 
1109 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1110 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1111 					 unsigned long address, pte_t *ptep)
1112 {
1113 	return ptep_test_and_clear_young(vma, address, ptep);
1114 }
1115 
1116 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1117 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1118 				       unsigned long addr, pte_t *ptep)
1119 {
1120 	pte_t res;
1121 
1122 	res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1123 	/* At this point the reference through the mapping is still present */
1124 	if (mm_is_protected(mm) && pte_present(res))
1125 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1126 	return res;
1127 }
1128 
1129 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1130 pte_t ptep_modify_prot_start(struct vm_area_struct *, unsigned long, pte_t *);
1131 void ptep_modify_prot_commit(struct vm_area_struct *, unsigned long,
1132 			     pte_t *, pte_t, pte_t);
1133 
1134 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1135 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1136 				     unsigned long addr, pte_t *ptep)
1137 {
1138 	pte_t res;
1139 
1140 	res = ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1141 	/* At this point the reference through the mapping is still present */
1142 	if (mm_is_protected(vma->vm_mm) && pte_present(res))
1143 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1144 	return res;
1145 }
1146 
1147 /*
1148  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1149  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1150  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1151  * cannot be accessed while the batched unmap is running. In this case
1152  * full==1 and a simple pte_clear is enough. See tlb.h.
1153  */
1154 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1155 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1156 					    unsigned long addr,
1157 					    pte_t *ptep, int full)
1158 {
1159 	pte_t res;
1160 
1161 	if (full) {
1162 		res = *ptep;
1163 		set_pte(ptep, __pte(_PAGE_INVALID));
1164 	} else {
1165 		res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1166 	}
1167 	/* At this point the reference through the mapping is still present */
1168 	if (mm_is_protected(mm) && pte_present(res))
1169 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1170 	return res;
1171 }
1172 
1173 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1174 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1175 				      unsigned long addr, pte_t *ptep)
1176 {
1177 	pte_t pte = *ptep;
1178 
1179 	if (pte_write(pte))
1180 		ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1181 }
1182 
1183 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1184 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1185 					unsigned long addr, pte_t *ptep,
1186 					pte_t entry, int dirty)
1187 {
1188 	if (pte_same(*ptep, entry))
1189 		return 0;
1190 	ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1191 	return 1;
1192 }
1193 
1194 /*
1195  * Additional functions to handle KVM guest page tables
1196  */
1197 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1198 		     pte_t *ptep, pte_t entry);
1199 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1200 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1201 		 pte_t *ptep, unsigned long bits);
1202 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1203 		    pte_t *ptep, int prot, unsigned long bit);
1204 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1205 		     pte_t *ptep , int reset);
1206 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1207 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1208 		    pte_t *sptep, pte_t *tptep, pte_t pte);
1209 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1210 
1211 bool ptep_test_and_clear_uc(struct mm_struct *mm, unsigned long address,
1212 			    pte_t *ptep);
1213 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1214 			  unsigned char key, bool nq);
1215 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1216 			       unsigned char key, unsigned char *oldkey,
1217 			       bool nq, bool mr, bool mc);
1218 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1219 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1220 			  unsigned char *key);
1221 
1222 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1223 				unsigned long bits, unsigned long value);
1224 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1225 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1226 			unsigned long *oldpte, unsigned long *oldpgste);
1227 void gmap_pmdp_csp(struct mm_struct *mm, unsigned long vmaddr);
1228 void gmap_pmdp_invalidate(struct mm_struct *mm, unsigned long vmaddr);
1229 void gmap_pmdp_idte_local(struct mm_struct *mm, unsigned long vmaddr);
1230 void gmap_pmdp_idte_global(struct mm_struct *mm, unsigned long vmaddr);
1231 
1232 #define pgprot_writecombine	pgprot_writecombine
1233 pgprot_t pgprot_writecombine(pgprot_t prot);
1234 
1235 #define pgprot_writethrough	pgprot_writethrough
1236 pgprot_t pgprot_writethrough(pgprot_t prot);
1237 
1238 /*
1239  * Certain architectures need to do special things when PTEs
1240  * within a page table are directly modified.  Thus, the following
1241  * hook is made available.
1242  */
1243 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
1244 			      pte_t *ptep, pte_t entry)
1245 {
1246 	if (pte_present(entry))
1247 		entry = clear_pte_bit(entry, __pgprot(_PAGE_UNUSED));
1248 	if (mm_has_pgste(mm))
1249 		ptep_set_pte_at(mm, addr, ptep, entry);
1250 	else
1251 		set_pte(ptep, entry);
1252 }
1253 
1254 /*
1255  * Conversion functions: convert a page and protection to a page entry,
1256  * and a page entry and page directory to the page they refer to.
1257  */
1258 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1259 {
1260 	pte_t __pte;
1261 
1262 	__pte = __pte(physpage | pgprot_val(pgprot));
1263 	if (!MACHINE_HAS_NX)
1264 		__pte = clear_pte_bit(__pte, __pgprot(_PAGE_NOEXEC));
1265 	return pte_mkyoung(__pte);
1266 }
1267 
1268 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1269 {
1270 	unsigned long physpage = page_to_phys(page);
1271 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1272 
1273 	if (pte_write(__pte) && PageDirty(page))
1274 		__pte = pte_mkdirty(__pte);
1275 	return __pte;
1276 }
1277 
1278 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1279 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1280 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1281 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1282 
1283 #define p4d_deref(pud) ((unsigned long)__va(p4d_val(pud) & _REGION_ENTRY_ORIGIN))
1284 #define pgd_deref(pgd) ((unsigned long)__va(pgd_val(pgd) & _REGION_ENTRY_ORIGIN))
1285 
1286 static inline unsigned long pmd_deref(pmd_t pmd)
1287 {
1288 	unsigned long origin_mask;
1289 
1290 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
1291 	if (pmd_large(pmd))
1292 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
1293 	return (unsigned long)__va(pmd_val(pmd) & origin_mask);
1294 }
1295 
1296 static inline unsigned long pmd_pfn(pmd_t pmd)
1297 {
1298 	return __pa(pmd_deref(pmd)) >> PAGE_SHIFT;
1299 }
1300 
1301 static inline unsigned long pud_deref(pud_t pud)
1302 {
1303 	unsigned long origin_mask;
1304 
1305 	origin_mask = _REGION_ENTRY_ORIGIN;
1306 	if (pud_large(pud))
1307 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
1308 	return (unsigned long)__va(pud_val(pud) & origin_mask);
1309 }
1310 
1311 static inline unsigned long pud_pfn(pud_t pud)
1312 {
1313 	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
1314 }
1315 
1316 /*
1317  * The pgd_offset function *always* adds the index for the top-level
1318  * region/segment table. This is done to get a sequence like the
1319  * following to work:
1320  *	pgdp = pgd_offset(current->mm, addr);
1321  *	pgd = READ_ONCE(*pgdp);
1322  *	p4dp = p4d_offset(&pgd, addr);
1323  *	...
1324  * The subsequent p4d_offset, pud_offset and pmd_offset functions
1325  * only add an index if they dereferenced the pointer.
1326  */
1327 static inline pgd_t *pgd_offset_raw(pgd_t *pgd, unsigned long address)
1328 {
1329 	unsigned long rste;
1330 	unsigned int shift;
1331 
1332 	/* Get the first entry of the top level table */
1333 	rste = pgd_val(*pgd);
1334 	/* Pick up the shift from the table type of the first entry */
1335 	shift = ((rste & _REGION_ENTRY_TYPE_MASK) >> 2) * 11 + 20;
1336 	return pgd + ((address >> shift) & (PTRS_PER_PGD - 1));
1337 }
1338 
1339 #define pgd_offset(mm, address) pgd_offset_raw(READ_ONCE((mm)->pgd), address)
1340 
1341 static inline p4d_t *p4d_offset_lockless(pgd_t *pgdp, pgd_t pgd, unsigned long address)
1342 {
1343 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R1)
1344 		return (p4d_t *) pgd_deref(pgd) + p4d_index(address);
1345 	return (p4d_t *) pgdp;
1346 }
1347 #define p4d_offset_lockless p4d_offset_lockless
1348 
1349 static inline p4d_t *p4d_offset(pgd_t *pgdp, unsigned long address)
1350 {
1351 	return p4d_offset_lockless(pgdp, *pgdp, address);
1352 }
1353 
1354 static inline pud_t *pud_offset_lockless(p4d_t *p4dp, p4d_t p4d, unsigned long address)
1355 {
1356 	if ((p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R2)
1357 		return (pud_t *) p4d_deref(p4d) + pud_index(address);
1358 	return (pud_t *) p4dp;
1359 }
1360 #define pud_offset_lockless pud_offset_lockless
1361 
1362 static inline pud_t *pud_offset(p4d_t *p4dp, unsigned long address)
1363 {
1364 	return pud_offset_lockless(p4dp, *p4dp, address);
1365 }
1366 #define pud_offset pud_offset
1367 
1368 static inline pmd_t *pmd_offset_lockless(pud_t *pudp, pud_t pud, unsigned long address)
1369 {
1370 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R3)
1371 		return (pmd_t *) pud_deref(pud) + pmd_index(address);
1372 	return (pmd_t *) pudp;
1373 }
1374 #define pmd_offset_lockless pmd_offset_lockless
1375 
1376 static inline pmd_t *pmd_offset(pud_t *pudp, unsigned long address)
1377 {
1378 	return pmd_offset_lockless(pudp, *pudp, address);
1379 }
1380 #define pmd_offset pmd_offset
1381 
1382 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
1383 {
1384 	return (unsigned long) pmd_deref(pmd);
1385 }
1386 
1387 static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
1388 {
1389 	return end <= current->mm->context.asce_limit;
1390 }
1391 #define gup_fast_permitted gup_fast_permitted
1392 
1393 #define pfn_pte(pfn, pgprot)	mk_pte_phys(((pfn) << PAGE_SHIFT), (pgprot))
1394 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1395 #define pte_page(x) pfn_to_page(pte_pfn(x))
1396 
1397 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1398 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1399 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
1400 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
1401 
1402 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1403 {
1404 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1405 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1406 }
1407 
1408 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1409 {
1410 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1411 	if (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY)
1412 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1413 	return pmd;
1414 }
1415 
1416 static inline pmd_t pmd_mkclean(pmd_t pmd)
1417 {
1418 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY));
1419 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1420 }
1421 
1422 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1423 {
1424 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_SOFT_DIRTY));
1425 	if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1426 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1427 	return pmd;
1428 }
1429 
1430 static inline pud_t pud_wrprotect(pud_t pud)
1431 {
1432 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1433 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1434 }
1435 
1436 static inline pud_t pud_mkwrite(pud_t pud)
1437 {
1438 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1439 	if (pud_val(pud) & _REGION3_ENTRY_DIRTY)
1440 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1441 	return pud;
1442 }
1443 
1444 static inline pud_t pud_mkclean(pud_t pud)
1445 {
1446 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY));
1447 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1448 }
1449 
1450 static inline pud_t pud_mkdirty(pud_t pud)
1451 {
1452 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY | _REGION3_ENTRY_SOFT_DIRTY));
1453 	if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1454 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1455 	return pud;
1456 }
1457 
1458 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1459 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1460 {
1461 	/*
1462 	 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1463 	 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1464 	 */
1465 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1466 		return pgprot_val(SEGMENT_NONE);
1467 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1468 		return pgprot_val(SEGMENT_RO);
1469 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1470 		return pgprot_val(SEGMENT_RX);
1471 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1472 		return pgprot_val(SEGMENT_RW);
1473 	return pgprot_val(SEGMENT_RWX);
1474 }
1475 
1476 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1477 {
1478 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1479 	if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1480 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1481 	return pmd;
1482 }
1483 
1484 static inline pmd_t pmd_mkold(pmd_t pmd)
1485 {
1486 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1487 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1488 }
1489 
1490 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1491 {
1492 	unsigned long mask;
1493 
1494 	mask  = _SEGMENT_ENTRY_ORIGIN_LARGE;
1495 	mask |= _SEGMENT_ENTRY_DIRTY;
1496 	mask |= _SEGMENT_ENTRY_YOUNG;
1497 	mask |=	_SEGMENT_ENTRY_LARGE;
1498 	mask |= _SEGMENT_ENTRY_SOFT_DIRTY;
1499 	pmd = __pmd(pmd_val(pmd) & mask);
1500 	pmd = set_pmd_bit(pmd, __pgprot(massage_pgprot_pmd(newprot)));
1501 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1502 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1503 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1504 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1505 	return pmd;
1506 }
1507 
1508 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1509 {
1510 	return __pmd(physpage + massage_pgprot_pmd(pgprot));
1511 }
1512 
1513 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1514 
1515 static inline void __pmdp_csp(pmd_t *pmdp)
1516 {
1517 	csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1518 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1519 }
1520 
1521 #define IDTE_GLOBAL	0
1522 #define IDTE_LOCAL	1
1523 
1524 #define IDTE_PTOA	0x0800
1525 #define IDTE_NODAT	0x1000
1526 #define IDTE_GUEST_ASCE	0x2000
1527 
1528 static __always_inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1529 					unsigned long opt, unsigned long asce,
1530 					int local)
1531 {
1532 	unsigned long sto;
1533 
1534 	sto = __pa(pmdp) - pmd_index(addr) * sizeof(pmd_t);
1535 	if (__builtin_constant_p(opt) && opt == 0) {
1536 		/* flush without guest asce */
1537 		asm volatile(
1538 			"	idte	%[r1],0,%[r2],%[m4]"
1539 			: "+m" (*pmdp)
1540 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1541 			  [m4] "i" (local)
1542 			: "cc" );
1543 	} else {
1544 		/* flush with guest asce */
1545 		asm volatile(
1546 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1547 			: "+m" (*pmdp)
1548 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1549 			  [r3] "a" (asce), [m4] "i" (local)
1550 			: "cc" );
1551 	}
1552 }
1553 
1554 static __always_inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1555 					unsigned long opt, unsigned long asce,
1556 					int local)
1557 {
1558 	unsigned long r3o;
1559 
1560 	r3o = __pa(pudp) - pud_index(addr) * sizeof(pud_t);
1561 	r3o |= _ASCE_TYPE_REGION3;
1562 	if (__builtin_constant_p(opt) && opt == 0) {
1563 		/* flush without guest asce */
1564 		asm volatile(
1565 			"	idte	%[r1],0,%[r2],%[m4]"
1566 			: "+m" (*pudp)
1567 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1568 			  [m4] "i" (local)
1569 			: "cc");
1570 	} else {
1571 		/* flush with guest asce */
1572 		asm volatile(
1573 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1574 			: "+m" (*pudp)
1575 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1576 			  [r3] "a" (asce), [m4] "i" (local)
1577 			: "cc" );
1578 	}
1579 }
1580 
1581 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1582 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1583 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1584 
1585 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1586 
1587 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1588 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1589 				pgtable_t pgtable);
1590 
1591 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1592 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1593 
1594 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1595 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1596 					unsigned long addr, pmd_t *pmdp,
1597 					pmd_t entry, int dirty)
1598 {
1599 	VM_BUG_ON(addr & ~HPAGE_MASK);
1600 
1601 	entry = pmd_mkyoung(entry);
1602 	if (dirty)
1603 		entry = pmd_mkdirty(entry);
1604 	if (pmd_val(*pmdp) == pmd_val(entry))
1605 		return 0;
1606 	pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1607 	return 1;
1608 }
1609 
1610 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1611 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1612 					    unsigned long addr, pmd_t *pmdp)
1613 {
1614 	pmd_t pmd = *pmdp;
1615 
1616 	pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1617 	return pmd_young(pmd);
1618 }
1619 
1620 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1621 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1622 					 unsigned long addr, pmd_t *pmdp)
1623 {
1624 	VM_BUG_ON(addr & ~HPAGE_MASK);
1625 	return pmdp_test_and_clear_young(vma, addr, pmdp);
1626 }
1627 
1628 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1629 			      pmd_t *pmdp, pmd_t entry)
1630 {
1631 	if (!MACHINE_HAS_NX)
1632 		entry = clear_pmd_bit(entry, __pgprot(_SEGMENT_ENTRY_NOEXEC));
1633 	set_pmd(pmdp, entry);
1634 }
1635 
1636 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1637 {
1638 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_LARGE));
1639 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1640 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1641 }
1642 
1643 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1644 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1645 					    unsigned long addr, pmd_t *pmdp)
1646 {
1647 	return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1648 }
1649 
1650 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1651 static inline pmd_t pmdp_huge_get_and_clear_full(struct vm_area_struct *vma,
1652 						 unsigned long addr,
1653 						 pmd_t *pmdp, int full)
1654 {
1655 	if (full) {
1656 		pmd_t pmd = *pmdp;
1657 		set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1658 		return pmd;
1659 	}
1660 	return pmdp_xchg_lazy(vma->vm_mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1661 }
1662 
1663 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1664 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1665 					  unsigned long addr, pmd_t *pmdp)
1666 {
1667 	return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1668 }
1669 
1670 #define __HAVE_ARCH_PMDP_INVALIDATE
1671 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1672 				   unsigned long addr, pmd_t *pmdp)
1673 {
1674 	pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1675 
1676 	return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1677 }
1678 
1679 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1680 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1681 				      unsigned long addr, pmd_t *pmdp)
1682 {
1683 	pmd_t pmd = *pmdp;
1684 
1685 	if (pmd_write(pmd))
1686 		pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1687 }
1688 
1689 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1690 					unsigned long address,
1691 					pmd_t *pmdp)
1692 {
1693 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1694 }
1695 #define pmdp_collapse_flush pmdp_collapse_flush
1696 
1697 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(((pfn) << PAGE_SHIFT), (pgprot))
1698 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1699 
1700 static inline int pmd_trans_huge(pmd_t pmd)
1701 {
1702 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1703 }
1704 
1705 #define has_transparent_hugepage has_transparent_hugepage
1706 static inline int has_transparent_hugepage(void)
1707 {
1708 	return MACHINE_HAS_EDAT1 ? 1 : 0;
1709 }
1710 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1711 
1712 /*
1713  * 64 bit swap entry format:
1714  * A page-table entry has some bits we have to treat in a special way.
1715  * Bits 52 and bit 55 have to be zero, otherwise a specification
1716  * exception will occur instead of a page translation exception. The
1717  * specification exception has the bad habit not to store necessary
1718  * information in the lowcore.
1719  * Bits 54 and 63 are used to indicate the page type.
1720  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1721  * This leaves the bits 0-51 and bits 56-62 to store type and offset.
1722  * We use the 5 bits from 57-61 for the type and the 52 bits from 0-51
1723  * for the offset.
1724  * |			  offset			|01100|type |00|
1725  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1726  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1727  */
1728 
1729 #define __SWP_OFFSET_MASK	((1UL << 52) - 1)
1730 #define __SWP_OFFSET_SHIFT	12
1731 #define __SWP_TYPE_MASK		((1UL << 5) - 1)
1732 #define __SWP_TYPE_SHIFT	2
1733 
1734 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1735 {
1736 	unsigned long pteval;
1737 
1738 	pteval = _PAGE_INVALID | _PAGE_PROTECT;
1739 	pteval |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1740 	pteval |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1741 	return __pte(pteval);
1742 }
1743 
1744 static inline unsigned long __swp_type(swp_entry_t entry)
1745 {
1746 	return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1747 }
1748 
1749 static inline unsigned long __swp_offset(swp_entry_t entry)
1750 {
1751 	return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1752 }
1753 
1754 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1755 {
1756 	return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1757 }
1758 
1759 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1760 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1761 
1762 #define kern_addr_valid(addr)   (1)
1763 
1764 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1765 extern void vmem_remove_mapping(unsigned long start, unsigned long size);
1766 extern int s390_enable_sie(void);
1767 extern int s390_enable_skey(void);
1768 extern void s390_reset_cmma(struct mm_struct *mm);
1769 
1770 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1771 #define HAVE_ARCH_UNMAPPED_AREA
1772 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1773 
1774 #define pmd_pgtable(pmd) \
1775 	((pgtable_t)__va(pmd_val(pmd) & -sizeof(pte_t)*PTRS_PER_PTE))
1776 
1777 #endif /* _S390_PAGE_H */
1778