xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision 67559900)
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 /*
542  * In the case that a guest uses storage keys
543  * faults should no longer be backed by zero pages
544  */
545 #define mm_forbids_zeropage mm_has_pgste
546 static inline int mm_uses_skeys(struct mm_struct *mm)
547 {
548 #ifdef CONFIG_PGSTE
549 	if (mm->context.uses_skeys)
550 		return 1;
551 #endif
552 	return 0;
553 }
554 
555 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
556 {
557 	union register_pair r1 = { .even = old, .odd = new, };
558 	unsigned long address = (unsigned long)ptr | 1;
559 
560 	asm volatile(
561 		"	csp	%[r1],%[address]"
562 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
563 		: [address] "d" (address)
564 		: "cc");
565 }
566 
567 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
568 {
569 	union register_pair r1 = { .even = old, .odd = new, };
570 	unsigned long address = (unsigned long)ptr | 1;
571 
572 	asm volatile(
573 		"	.insn	rre,0xb98a0000,%[r1],%[address]"
574 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
575 		: [address] "d" (address)
576 		: "cc");
577 }
578 
579 #define CRDTE_DTT_PAGE		0x00UL
580 #define CRDTE_DTT_SEGMENT	0x10UL
581 #define CRDTE_DTT_REGION3	0x14UL
582 #define CRDTE_DTT_REGION2	0x18UL
583 #define CRDTE_DTT_REGION1	0x1cUL
584 
585 static inline void crdte(unsigned long old, unsigned long new,
586 			 unsigned long *table, unsigned long dtt,
587 			 unsigned long address, unsigned long asce)
588 {
589 	union register_pair r1 = { .even = old, .odd = new, };
590 	union register_pair r2 = { .even = __pa(table) | dtt, .odd = address, };
591 
592 	asm volatile(".insn rrf,0xb98f0000,%[r1],%[r2],%[asce],0"
593 		     : [r1] "+&d" (r1.pair)
594 		     : [r2] "d" (r2.pair), [asce] "a" (asce)
595 		     : "memory", "cc");
596 }
597 
598 /*
599  * pgd/p4d/pud/pmd/pte query functions
600  */
601 static inline int pgd_folded(pgd_t pgd)
602 {
603 	return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
604 }
605 
606 static inline int pgd_present(pgd_t pgd)
607 {
608 	if (pgd_folded(pgd))
609 		return 1;
610 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
611 }
612 
613 static inline int pgd_none(pgd_t pgd)
614 {
615 	if (pgd_folded(pgd))
616 		return 0;
617 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
618 }
619 
620 static inline int pgd_bad(pgd_t pgd)
621 {
622 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1)
623 		return 0;
624 	return (pgd_val(pgd) & ~_REGION_ENTRY_BITS) != 0;
625 }
626 
627 static inline unsigned long pgd_pfn(pgd_t pgd)
628 {
629 	unsigned long origin_mask;
630 
631 	origin_mask = _REGION_ENTRY_ORIGIN;
632 	return (pgd_val(pgd) & origin_mask) >> PAGE_SHIFT;
633 }
634 
635 static inline int p4d_folded(p4d_t p4d)
636 {
637 	return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
638 }
639 
640 static inline int p4d_present(p4d_t p4d)
641 {
642 	if (p4d_folded(p4d))
643 		return 1;
644 	return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
645 }
646 
647 static inline int p4d_none(p4d_t p4d)
648 {
649 	if (p4d_folded(p4d))
650 		return 0;
651 	return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
652 }
653 
654 static inline unsigned long p4d_pfn(p4d_t p4d)
655 {
656 	unsigned long origin_mask;
657 
658 	origin_mask = _REGION_ENTRY_ORIGIN;
659 	return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
660 }
661 
662 static inline int pud_folded(pud_t pud)
663 {
664 	return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
665 }
666 
667 static inline int pud_present(pud_t pud)
668 {
669 	if (pud_folded(pud))
670 		return 1;
671 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
672 }
673 
674 static inline int pud_none(pud_t pud)
675 {
676 	if (pud_folded(pud))
677 		return 0;
678 	return pud_val(pud) == _REGION3_ENTRY_EMPTY;
679 }
680 
681 #define pud_leaf	pud_large
682 static inline int pud_large(pud_t pud)
683 {
684 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
685 		return 0;
686 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
687 }
688 
689 #define pmd_leaf	pmd_large
690 static inline int pmd_large(pmd_t pmd)
691 {
692 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
693 }
694 
695 static inline int pmd_bad(pmd_t pmd)
696 {
697 	if ((pmd_val(pmd) & _SEGMENT_ENTRY_TYPE_MASK) > 0 || pmd_large(pmd))
698 		return 1;
699 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
700 }
701 
702 static inline int pud_bad(pud_t pud)
703 {
704 	unsigned long type = pud_val(pud) & _REGION_ENTRY_TYPE_MASK;
705 
706 	if (type > _REGION_ENTRY_TYPE_R3 || pud_large(pud))
707 		return 1;
708 	if (type < _REGION_ENTRY_TYPE_R3)
709 		return 0;
710 	return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
711 }
712 
713 static inline int p4d_bad(p4d_t p4d)
714 {
715 	unsigned long type = p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK;
716 
717 	if (type > _REGION_ENTRY_TYPE_R2)
718 		return 1;
719 	if (type < _REGION_ENTRY_TYPE_R2)
720 		return 0;
721 	return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
722 }
723 
724 static inline int pmd_present(pmd_t pmd)
725 {
726 	return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
727 }
728 
729 static inline int pmd_none(pmd_t pmd)
730 {
731 	return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
732 }
733 
734 #define pmd_write pmd_write
735 static inline int pmd_write(pmd_t pmd)
736 {
737 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
738 }
739 
740 #define pud_write pud_write
741 static inline int pud_write(pud_t pud)
742 {
743 	return (pud_val(pud) & _REGION3_ENTRY_WRITE) != 0;
744 }
745 
746 static inline int pmd_dirty(pmd_t pmd)
747 {
748 	return (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
749 }
750 
751 static inline int pmd_young(pmd_t pmd)
752 {
753 	return (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
754 }
755 
756 static inline int pte_present(pte_t pte)
757 {
758 	/* Bit pattern: (pte & 0x001) == 0x001 */
759 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
760 }
761 
762 static inline int pte_none(pte_t pte)
763 {
764 	/* Bit pattern: pte == 0x400 */
765 	return pte_val(pte) == _PAGE_INVALID;
766 }
767 
768 static inline int pte_swap(pte_t pte)
769 {
770 	/* Bit pattern: (pte & 0x201) == 0x200 */
771 	return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
772 		== _PAGE_PROTECT;
773 }
774 
775 static inline int pte_special(pte_t pte)
776 {
777 	return (pte_val(pte) & _PAGE_SPECIAL);
778 }
779 
780 #define __HAVE_ARCH_PTE_SAME
781 static inline int pte_same(pte_t a, pte_t b)
782 {
783 	return pte_val(a) == pte_val(b);
784 }
785 
786 #ifdef CONFIG_NUMA_BALANCING
787 static inline int pte_protnone(pte_t pte)
788 {
789 	return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
790 }
791 
792 static inline int pmd_protnone(pmd_t pmd)
793 {
794 	/* pmd_large(pmd) implies pmd_present(pmd) */
795 	return pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
796 }
797 #endif
798 
799 static inline int pte_soft_dirty(pte_t pte)
800 {
801 	return pte_val(pte) & _PAGE_SOFT_DIRTY;
802 }
803 #define pte_swp_soft_dirty pte_soft_dirty
804 
805 static inline pte_t pte_mksoft_dirty(pte_t pte)
806 {
807 	pte_val(pte) |= _PAGE_SOFT_DIRTY;
808 	return pte;
809 }
810 #define pte_swp_mksoft_dirty pte_mksoft_dirty
811 
812 static inline pte_t pte_clear_soft_dirty(pte_t pte)
813 {
814 	pte_val(pte) &= ~_PAGE_SOFT_DIRTY;
815 	return pte;
816 }
817 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
818 
819 static inline int pmd_soft_dirty(pmd_t pmd)
820 {
821 	return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
822 }
823 
824 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
825 {
826 	pmd_val(pmd) |= _SEGMENT_ENTRY_SOFT_DIRTY;
827 	return pmd;
828 }
829 
830 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
831 {
832 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_SOFT_DIRTY;
833 	return pmd;
834 }
835 
836 /*
837  * query functions pte_write/pte_dirty/pte_young only work if
838  * pte_present() is true. Undefined behaviour if not..
839  */
840 static inline int pte_write(pte_t pte)
841 {
842 	return (pte_val(pte) & _PAGE_WRITE) != 0;
843 }
844 
845 static inline int pte_dirty(pte_t pte)
846 {
847 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
848 }
849 
850 static inline int pte_young(pte_t pte)
851 {
852 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
853 }
854 
855 #define __HAVE_ARCH_PTE_UNUSED
856 static inline int pte_unused(pte_t pte)
857 {
858 	return pte_val(pte) & _PAGE_UNUSED;
859 }
860 
861 /*
862  * Extract the pgprot value from the given pte while at the same time making it
863  * usable for kernel address space mappings where fault driven dirty and
864  * young/old accounting is not supported, i.e _PAGE_PROTECT and _PAGE_INVALID
865  * must not be set.
866  */
867 static inline pgprot_t pte_pgprot(pte_t pte)
868 {
869 	unsigned long pte_flags = pte_val(pte) & _PAGE_CHG_MASK;
870 
871 	if (pte_write(pte))
872 		pte_flags |= pgprot_val(PAGE_KERNEL);
873 	else
874 		pte_flags |= pgprot_val(PAGE_KERNEL_RO);
875 	pte_flags |= pte_val(pte) & mio_wb_bit_mask;
876 
877 	return __pgprot(pte_flags);
878 }
879 
880 /*
881  * pgd/pmd/pte modification functions
882  */
883 
884 static inline void pgd_clear(pgd_t *pgd)
885 {
886 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
887 		pgd_val(*pgd) = _REGION1_ENTRY_EMPTY;
888 }
889 
890 static inline void p4d_clear(p4d_t *p4d)
891 {
892 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
893 		p4d_val(*p4d) = _REGION2_ENTRY_EMPTY;
894 }
895 
896 static inline void pud_clear(pud_t *pud)
897 {
898 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
899 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
900 }
901 
902 static inline void pmd_clear(pmd_t *pmdp)
903 {
904 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
905 }
906 
907 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
908 {
909 	pte_val(*ptep) = _PAGE_INVALID;
910 }
911 
912 /*
913  * The following pte modification functions only work if
914  * pte_present() is true. Undefined behaviour if not..
915  */
916 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
917 {
918 	pte_val(pte) &= _PAGE_CHG_MASK;
919 	pte_val(pte) |= pgprot_val(newprot);
920 	/*
921 	 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
922 	 * has the invalid bit set, clear it again for readable, young pages
923 	 */
924 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
925 		pte_val(pte) &= ~_PAGE_INVALID;
926 	/*
927 	 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
928 	 * protection bit set, clear it again for writable, dirty pages
929 	 */
930 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
931 		pte_val(pte) &= ~_PAGE_PROTECT;
932 	return pte;
933 }
934 
935 static inline pte_t pte_wrprotect(pte_t pte)
936 {
937 	pte_val(pte) &= ~_PAGE_WRITE;
938 	pte_val(pte) |= _PAGE_PROTECT;
939 	return pte;
940 }
941 
942 static inline pte_t pte_mkwrite(pte_t pte)
943 {
944 	pte_val(pte) |= _PAGE_WRITE;
945 	if (pte_val(pte) & _PAGE_DIRTY)
946 		pte_val(pte) &= ~_PAGE_PROTECT;
947 	return pte;
948 }
949 
950 static inline pte_t pte_mkclean(pte_t pte)
951 {
952 	pte_val(pte) &= ~_PAGE_DIRTY;
953 	pte_val(pte) |= _PAGE_PROTECT;
954 	return pte;
955 }
956 
957 static inline pte_t pte_mkdirty(pte_t pte)
958 {
959 	pte_val(pte) |= _PAGE_DIRTY | _PAGE_SOFT_DIRTY;
960 	if (pte_val(pte) & _PAGE_WRITE)
961 		pte_val(pte) &= ~_PAGE_PROTECT;
962 	return pte;
963 }
964 
965 static inline pte_t pte_mkold(pte_t pte)
966 {
967 	pte_val(pte) &= ~_PAGE_YOUNG;
968 	pte_val(pte) |= _PAGE_INVALID;
969 	return pte;
970 }
971 
972 static inline pte_t pte_mkyoung(pte_t pte)
973 {
974 	pte_val(pte) |= _PAGE_YOUNG;
975 	if (pte_val(pte) & _PAGE_READ)
976 		pte_val(pte) &= ~_PAGE_INVALID;
977 	return pte;
978 }
979 
980 static inline pte_t pte_mkspecial(pte_t pte)
981 {
982 	pte_val(pte) |= _PAGE_SPECIAL;
983 	return pte;
984 }
985 
986 #ifdef CONFIG_HUGETLB_PAGE
987 static inline pte_t pte_mkhuge(pte_t pte)
988 {
989 	pte_val(pte) |= _PAGE_LARGE;
990 	return pte;
991 }
992 #endif
993 
994 #define IPTE_GLOBAL	0
995 #define	IPTE_LOCAL	1
996 
997 #define IPTE_NODAT	0x400
998 #define IPTE_GUEST_ASCE	0x800
999 
1000 static __always_inline void __ptep_ipte(unsigned long address, pte_t *ptep,
1001 					unsigned long opt, unsigned long asce,
1002 					int local)
1003 {
1004 	unsigned long pto = __pa(ptep);
1005 
1006 	if (__builtin_constant_p(opt) && opt == 0) {
1007 		/* Invalidation + TLB flush for the pte */
1008 		asm volatile(
1009 			"	.insn	rrf,0xb2210000,%[r1],%[r2],0,%[m4]"
1010 			: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
1011 			  [m4] "i" (local));
1012 		return;
1013 	}
1014 
1015 	/* Invalidate ptes with options + TLB flush of the ptes */
1016 	opt = opt | (asce & _ASCE_ORIGIN);
1017 	asm volatile(
1018 		"	.insn	rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
1019 		: [r2] "+a" (address), [r3] "+a" (opt)
1020 		: [r1] "a" (pto), [m4] "i" (local) : "memory");
1021 }
1022 
1023 static __always_inline void __ptep_ipte_range(unsigned long address, int nr,
1024 					      pte_t *ptep, int local)
1025 {
1026 	unsigned long pto = __pa(ptep);
1027 
1028 	/* Invalidate a range of ptes + TLB flush of the ptes */
1029 	do {
1030 		asm volatile(
1031 			"       .insn rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
1032 			: [r2] "+a" (address), [r3] "+a" (nr)
1033 			: [r1] "a" (pto), [m4] "i" (local) : "memory");
1034 	} while (nr != 255);
1035 }
1036 
1037 /*
1038  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1039  * both clear the TLB for the unmapped pte. The reason is that
1040  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1041  * to modify an active pte. The sequence is
1042  *   1) ptep_get_and_clear
1043  *   2) set_pte_at
1044  *   3) flush_tlb_range
1045  * On s390 the tlb needs to get flushed with the modification of the pte
1046  * if the pte is active. The only way how this can be implemented is to
1047  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1048  * is a nop.
1049  */
1050 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1051 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1052 
1053 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1054 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1055 					    unsigned long addr, pte_t *ptep)
1056 {
1057 	pte_t pte = *ptep;
1058 
1059 	pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1060 	return pte_young(pte);
1061 }
1062 
1063 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1064 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1065 					 unsigned long address, pte_t *ptep)
1066 {
1067 	return ptep_test_and_clear_young(vma, address, ptep);
1068 }
1069 
1070 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1071 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1072 				       unsigned long addr, pte_t *ptep)
1073 {
1074 	pte_t res;
1075 
1076 	res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1077 	/* At this point the reference through the mapping is still present */
1078 	if (mm_is_protected(mm) && pte_present(res))
1079 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1080 	return res;
1081 }
1082 
1083 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1084 pte_t ptep_modify_prot_start(struct vm_area_struct *, unsigned long, pte_t *);
1085 void ptep_modify_prot_commit(struct vm_area_struct *, unsigned long,
1086 			     pte_t *, pte_t, pte_t);
1087 
1088 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1089 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1090 				     unsigned long addr, pte_t *ptep)
1091 {
1092 	pte_t res;
1093 
1094 	res = ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1095 	/* At this point the reference through the mapping is still present */
1096 	if (mm_is_protected(vma->vm_mm) && pte_present(res))
1097 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1098 	return res;
1099 }
1100 
1101 /*
1102  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1103  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1104  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1105  * cannot be accessed while the batched unmap is running. In this case
1106  * full==1 and a simple pte_clear is enough. See tlb.h.
1107  */
1108 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1109 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1110 					    unsigned long addr,
1111 					    pte_t *ptep, int full)
1112 {
1113 	pte_t res;
1114 
1115 	if (full) {
1116 		res = *ptep;
1117 		*ptep = __pte(_PAGE_INVALID);
1118 	} else {
1119 		res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1120 	}
1121 	/* At this point the reference through the mapping is still present */
1122 	if (mm_is_protected(mm) && pte_present(res))
1123 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1124 	return res;
1125 }
1126 
1127 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1128 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1129 				      unsigned long addr, pte_t *ptep)
1130 {
1131 	pte_t pte = *ptep;
1132 
1133 	if (pte_write(pte))
1134 		ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1135 }
1136 
1137 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1138 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1139 					unsigned long addr, pte_t *ptep,
1140 					pte_t entry, int dirty)
1141 {
1142 	if (pte_same(*ptep, entry))
1143 		return 0;
1144 	ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1145 	return 1;
1146 }
1147 
1148 /*
1149  * Additional functions to handle KVM guest page tables
1150  */
1151 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1152 		     pte_t *ptep, pte_t entry);
1153 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1154 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1155 		 pte_t *ptep, unsigned long bits);
1156 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1157 		    pte_t *ptep, int prot, unsigned long bit);
1158 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1159 		     pte_t *ptep , int reset);
1160 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1161 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1162 		    pte_t *sptep, pte_t *tptep, pte_t pte);
1163 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1164 
1165 bool ptep_test_and_clear_uc(struct mm_struct *mm, unsigned long address,
1166 			    pte_t *ptep);
1167 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1168 			  unsigned char key, bool nq);
1169 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1170 			       unsigned char key, unsigned char *oldkey,
1171 			       bool nq, bool mr, bool mc);
1172 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1173 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1174 			  unsigned char *key);
1175 
1176 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1177 				unsigned long bits, unsigned long value);
1178 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1179 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1180 			unsigned long *oldpte, unsigned long *oldpgste);
1181 void gmap_pmdp_csp(struct mm_struct *mm, unsigned long vmaddr);
1182 void gmap_pmdp_invalidate(struct mm_struct *mm, unsigned long vmaddr);
1183 void gmap_pmdp_idte_local(struct mm_struct *mm, unsigned long vmaddr);
1184 void gmap_pmdp_idte_global(struct mm_struct *mm, unsigned long vmaddr);
1185 
1186 #define pgprot_writecombine	pgprot_writecombine
1187 pgprot_t pgprot_writecombine(pgprot_t prot);
1188 
1189 #define pgprot_writethrough	pgprot_writethrough
1190 pgprot_t pgprot_writethrough(pgprot_t prot);
1191 
1192 /*
1193  * Certain architectures need to do special things when PTEs
1194  * within a page table are directly modified.  Thus, the following
1195  * hook is made available.
1196  */
1197 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
1198 			      pte_t *ptep, pte_t entry)
1199 {
1200 	if (pte_present(entry))
1201 		pte_val(entry) &= ~_PAGE_UNUSED;
1202 	if (mm_has_pgste(mm))
1203 		ptep_set_pte_at(mm, addr, ptep, entry);
1204 	else
1205 		*ptep = entry;
1206 }
1207 
1208 /*
1209  * Conversion functions: convert a page and protection to a page entry,
1210  * and a page entry and page directory to the page they refer to.
1211  */
1212 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1213 {
1214 	pte_t __pte;
1215 
1216 	pte_val(__pte) = physpage | pgprot_val(pgprot);
1217 	if (!MACHINE_HAS_NX)
1218 		pte_val(__pte) &= ~_PAGE_NOEXEC;
1219 	return pte_mkyoung(__pte);
1220 }
1221 
1222 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1223 {
1224 	unsigned long physpage = page_to_phys(page);
1225 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1226 
1227 	if (pte_write(__pte) && PageDirty(page))
1228 		__pte = pte_mkdirty(__pte);
1229 	return __pte;
1230 }
1231 
1232 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1233 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1234 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1235 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1236 
1237 #define p4d_deref(pud) ((unsigned long)__va(p4d_val(pud) & _REGION_ENTRY_ORIGIN))
1238 #define pgd_deref(pgd) ((unsigned long)__va(pgd_val(pgd) & _REGION_ENTRY_ORIGIN))
1239 
1240 static inline unsigned long pmd_deref(pmd_t pmd)
1241 {
1242 	unsigned long origin_mask;
1243 
1244 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
1245 	if (pmd_large(pmd))
1246 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
1247 	return (unsigned long)__va(pmd_val(pmd) & origin_mask);
1248 }
1249 
1250 static inline unsigned long pmd_pfn(pmd_t pmd)
1251 {
1252 	return __pa(pmd_deref(pmd)) >> PAGE_SHIFT;
1253 }
1254 
1255 static inline unsigned long pud_deref(pud_t pud)
1256 {
1257 	unsigned long origin_mask;
1258 
1259 	origin_mask = _REGION_ENTRY_ORIGIN;
1260 	if (pud_large(pud))
1261 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
1262 	return (unsigned long)__va(pud_val(pud) & origin_mask);
1263 }
1264 
1265 static inline unsigned long pud_pfn(pud_t pud)
1266 {
1267 	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
1268 }
1269 
1270 /*
1271  * The pgd_offset function *always* adds the index for the top-level
1272  * region/segment table. This is done to get a sequence like the
1273  * following to work:
1274  *	pgdp = pgd_offset(current->mm, addr);
1275  *	pgd = READ_ONCE(*pgdp);
1276  *	p4dp = p4d_offset(&pgd, addr);
1277  *	...
1278  * The subsequent p4d_offset, pud_offset and pmd_offset functions
1279  * only add an index if they dereferenced the pointer.
1280  */
1281 static inline pgd_t *pgd_offset_raw(pgd_t *pgd, unsigned long address)
1282 {
1283 	unsigned long rste;
1284 	unsigned int shift;
1285 
1286 	/* Get the first entry of the top level table */
1287 	rste = pgd_val(*pgd);
1288 	/* Pick up the shift from the table type of the first entry */
1289 	shift = ((rste & _REGION_ENTRY_TYPE_MASK) >> 2) * 11 + 20;
1290 	return pgd + ((address >> shift) & (PTRS_PER_PGD - 1));
1291 }
1292 
1293 #define pgd_offset(mm, address) pgd_offset_raw(READ_ONCE((mm)->pgd), address)
1294 
1295 static inline p4d_t *p4d_offset_lockless(pgd_t *pgdp, pgd_t pgd, unsigned long address)
1296 {
1297 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R1)
1298 		return (p4d_t *) pgd_deref(pgd) + p4d_index(address);
1299 	return (p4d_t *) pgdp;
1300 }
1301 #define p4d_offset_lockless p4d_offset_lockless
1302 
1303 static inline p4d_t *p4d_offset(pgd_t *pgdp, unsigned long address)
1304 {
1305 	return p4d_offset_lockless(pgdp, *pgdp, address);
1306 }
1307 
1308 static inline pud_t *pud_offset_lockless(p4d_t *p4dp, p4d_t p4d, unsigned long address)
1309 {
1310 	if ((p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R2)
1311 		return (pud_t *) p4d_deref(p4d) + pud_index(address);
1312 	return (pud_t *) p4dp;
1313 }
1314 #define pud_offset_lockless pud_offset_lockless
1315 
1316 static inline pud_t *pud_offset(p4d_t *p4dp, unsigned long address)
1317 {
1318 	return pud_offset_lockless(p4dp, *p4dp, address);
1319 }
1320 #define pud_offset pud_offset
1321 
1322 static inline pmd_t *pmd_offset_lockless(pud_t *pudp, pud_t pud, unsigned long address)
1323 {
1324 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R3)
1325 		return (pmd_t *) pud_deref(pud) + pmd_index(address);
1326 	return (pmd_t *) pudp;
1327 }
1328 #define pmd_offset_lockless pmd_offset_lockless
1329 
1330 static inline pmd_t *pmd_offset(pud_t *pudp, unsigned long address)
1331 {
1332 	return pmd_offset_lockless(pudp, *pudp, address);
1333 }
1334 #define pmd_offset pmd_offset
1335 
1336 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
1337 {
1338 	return (unsigned long) pmd_deref(pmd);
1339 }
1340 
1341 static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
1342 {
1343 	return end <= current->mm->context.asce_limit;
1344 }
1345 #define gup_fast_permitted gup_fast_permitted
1346 
1347 #define pfn_pte(pfn, pgprot)	mk_pte_phys(((pfn) << PAGE_SHIFT), (pgprot))
1348 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1349 #define pte_page(x) pfn_to_page(pte_pfn(x))
1350 
1351 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1352 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1353 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
1354 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
1355 
1356 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1357 {
1358 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_WRITE;
1359 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1360 	return pmd;
1361 }
1362 
1363 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1364 {
1365 	pmd_val(pmd) |= _SEGMENT_ENTRY_WRITE;
1366 	if (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY)
1367 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1368 	return pmd;
1369 }
1370 
1371 static inline pmd_t pmd_mkclean(pmd_t pmd)
1372 {
1373 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_DIRTY;
1374 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1375 	return pmd;
1376 }
1377 
1378 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1379 {
1380 	pmd_val(pmd) |= _SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_SOFT_DIRTY;
1381 	if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1382 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1383 	return pmd;
1384 }
1385 
1386 static inline pud_t pud_wrprotect(pud_t pud)
1387 {
1388 	pud_val(pud) &= ~_REGION3_ENTRY_WRITE;
1389 	pud_val(pud) |= _REGION_ENTRY_PROTECT;
1390 	return pud;
1391 }
1392 
1393 static inline pud_t pud_mkwrite(pud_t pud)
1394 {
1395 	pud_val(pud) |= _REGION3_ENTRY_WRITE;
1396 	if (pud_val(pud) & _REGION3_ENTRY_DIRTY)
1397 		pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1398 	return pud;
1399 }
1400 
1401 static inline pud_t pud_mkclean(pud_t pud)
1402 {
1403 	pud_val(pud) &= ~_REGION3_ENTRY_DIRTY;
1404 	pud_val(pud) |= _REGION_ENTRY_PROTECT;
1405 	return pud;
1406 }
1407 
1408 static inline pud_t pud_mkdirty(pud_t pud)
1409 {
1410 	pud_val(pud) |= _REGION3_ENTRY_DIRTY | _REGION3_ENTRY_SOFT_DIRTY;
1411 	if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1412 		pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1413 	return pud;
1414 }
1415 
1416 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1417 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1418 {
1419 	/*
1420 	 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1421 	 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1422 	 */
1423 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1424 		return pgprot_val(SEGMENT_NONE);
1425 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1426 		return pgprot_val(SEGMENT_RO);
1427 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1428 		return pgprot_val(SEGMENT_RX);
1429 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1430 		return pgprot_val(SEGMENT_RW);
1431 	return pgprot_val(SEGMENT_RWX);
1432 }
1433 
1434 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1435 {
1436 	pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1437 	if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1438 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1439 	return pmd;
1440 }
1441 
1442 static inline pmd_t pmd_mkold(pmd_t pmd)
1443 {
1444 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1445 	pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1446 	return pmd;
1447 }
1448 
1449 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1450 {
1451 	pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN_LARGE |
1452 		_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_YOUNG |
1453 		_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_SOFT_DIRTY;
1454 	pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1455 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1456 		pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1457 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1458 		pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1459 	return pmd;
1460 }
1461 
1462 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1463 {
1464 	pmd_t __pmd;
1465 	pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1466 	return __pmd;
1467 }
1468 
1469 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1470 
1471 static inline void __pmdp_csp(pmd_t *pmdp)
1472 {
1473 	csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1474 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1475 }
1476 
1477 #define IDTE_GLOBAL	0
1478 #define IDTE_LOCAL	1
1479 
1480 #define IDTE_PTOA	0x0800
1481 #define IDTE_NODAT	0x1000
1482 #define IDTE_GUEST_ASCE	0x2000
1483 
1484 static __always_inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1485 					unsigned long opt, unsigned long asce,
1486 					int local)
1487 {
1488 	unsigned long sto;
1489 
1490 	sto = __pa(pmdp) - pmd_index(addr) * sizeof(pmd_t);
1491 	if (__builtin_constant_p(opt) && opt == 0) {
1492 		/* flush without guest asce */
1493 		asm volatile(
1494 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1495 			: "+m" (*pmdp)
1496 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1497 			  [m4] "i" (local)
1498 			: "cc" );
1499 	} else {
1500 		/* flush with guest asce */
1501 		asm volatile(
1502 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1503 			: "+m" (*pmdp)
1504 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1505 			  [r3] "a" (asce), [m4] "i" (local)
1506 			: "cc" );
1507 	}
1508 }
1509 
1510 static __always_inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1511 					unsigned long opt, unsigned long asce,
1512 					int local)
1513 {
1514 	unsigned long r3o;
1515 
1516 	r3o = __pa(pudp) - pud_index(addr) * sizeof(pud_t);
1517 	r3o |= _ASCE_TYPE_REGION3;
1518 	if (__builtin_constant_p(opt) && opt == 0) {
1519 		/* flush without guest asce */
1520 		asm volatile(
1521 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1522 			: "+m" (*pudp)
1523 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1524 			  [m4] "i" (local)
1525 			: "cc");
1526 	} else {
1527 		/* flush with guest asce */
1528 		asm volatile(
1529 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1530 			: "+m" (*pudp)
1531 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1532 			  [r3] "a" (asce), [m4] "i" (local)
1533 			: "cc" );
1534 	}
1535 }
1536 
1537 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1538 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1539 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1540 
1541 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1542 
1543 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1544 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1545 				pgtable_t pgtable);
1546 
1547 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1548 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1549 
1550 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1551 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1552 					unsigned long addr, pmd_t *pmdp,
1553 					pmd_t entry, int dirty)
1554 {
1555 	VM_BUG_ON(addr & ~HPAGE_MASK);
1556 
1557 	entry = pmd_mkyoung(entry);
1558 	if (dirty)
1559 		entry = pmd_mkdirty(entry);
1560 	if (pmd_val(*pmdp) == pmd_val(entry))
1561 		return 0;
1562 	pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1563 	return 1;
1564 }
1565 
1566 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1567 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1568 					    unsigned long addr, pmd_t *pmdp)
1569 {
1570 	pmd_t pmd = *pmdp;
1571 
1572 	pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1573 	return pmd_young(pmd);
1574 }
1575 
1576 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1577 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1578 					 unsigned long addr, pmd_t *pmdp)
1579 {
1580 	VM_BUG_ON(addr & ~HPAGE_MASK);
1581 	return pmdp_test_and_clear_young(vma, addr, pmdp);
1582 }
1583 
1584 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1585 			      pmd_t *pmdp, pmd_t entry)
1586 {
1587 	if (!MACHINE_HAS_NX)
1588 		pmd_val(entry) &= ~_SEGMENT_ENTRY_NOEXEC;
1589 	*pmdp = entry;
1590 }
1591 
1592 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1593 {
1594 	pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1595 	pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1596 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1597 	return pmd;
1598 }
1599 
1600 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1601 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1602 					    unsigned long addr, pmd_t *pmdp)
1603 {
1604 	return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1605 }
1606 
1607 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1608 static inline pmd_t pmdp_huge_get_and_clear_full(struct vm_area_struct *vma,
1609 						 unsigned long addr,
1610 						 pmd_t *pmdp, int full)
1611 {
1612 	if (full) {
1613 		pmd_t pmd = *pmdp;
1614 		*pmdp = __pmd(_SEGMENT_ENTRY_EMPTY);
1615 		return pmd;
1616 	}
1617 	return pmdp_xchg_lazy(vma->vm_mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1618 }
1619 
1620 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1621 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1622 					  unsigned long addr, pmd_t *pmdp)
1623 {
1624 	return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1625 }
1626 
1627 #define __HAVE_ARCH_PMDP_INVALIDATE
1628 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1629 				   unsigned long addr, pmd_t *pmdp)
1630 {
1631 	pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1632 
1633 	return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1634 }
1635 
1636 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1637 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1638 				      unsigned long addr, pmd_t *pmdp)
1639 {
1640 	pmd_t pmd = *pmdp;
1641 
1642 	if (pmd_write(pmd))
1643 		pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1644 }
1645 
1646 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1647 					unsigned long address,
1648 					pmd_t *pmdp)
1649 {
1650 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1651 }
1652 #define pmdp_collapse_flush pmdp_collapse_flush
1653 
1654 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(((pfn) << PAGE_SHIFT), (pgprot))
1655 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1656 
1657 static inline int pmd_trans_huge(pmd_t pmd)
1658 {
1659 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1660 }
1661 
1662 #define has_transparent_hugepage has_transparent_hugepage
1663 static inline int has_transparent_hugepage(void)
1664 {
1665 	return MACHINE_HAS_EDAT1 ? 1 : 0;
1666 }
1667 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1668 
1669 /*
1670  * 64 bit swap entry format:
1671  * A page-table entry has some bits we have to treat in a special way.
1672  * Bits 52 and bit 55 have to be zero, otherwise a specification
1673  * exception will occur instead of a page translation exception. The
1674  * specification exception has the bad habit not to store necessary
1675  * information in the lowcore.
1676  * Bits 54 and 63 are used to indicate the page type.
1677  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1678  * This leaves the bits 0-51 and bits 56-62 to store type and offset.
1679  * We use the 5 bits from 57-61 for the type and the 52 bits from 0-51
1680  * for the offset.
1681  * |			  offset			|01100|type |00|
1682  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1683  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1684  */
1685 
1686 #define __SWP_OFFSET_MASK	((1UL << 52) - 1)
1687 #define __SWP_OFFSET_SHIFT	12
1688 #define __SWP_TYPE_MASK		((1UL << 5) - 1)
1689 #define __SWP_TYPE_SHIFT	2
1690 
1691 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1692 {
1693 	pte_t pte;
1694 
1695 	pte_val(pte) = _PAGE_INVALID | _PAGE_PROTECT;
1696 	pte_val(pte) |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1697 	pte_val(pte) |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1698 	return pte;
1699 }
1700 
1701 static inline unsigned long __swp_type(swp_entry_t entry)
1702 {
1703 	return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1704 }
1705 
1706 static inline unsigned long __swp_offset(swp_entry_t entry)
1707 {
1708 	return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1709 }
1710 
1711 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1712 {
1713 	return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1714 }
1715 
1716 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1717 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1718 
1719 #define kern_addr_valid(addr)   (1)
1720 
1721 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1722 extern void vmem_remove_mapping(unsigned long start, unsigned long size);
1723 extern int s390_enable_sie(void);
1724 extern int s390_enable_skey(void);
1725 extern void s390_reset_cmma(struct mm_struct *mm);
1726 
1727 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1728 #define HAVE_ARCH_UNMAPPED_AREA
1729 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1730 
1731 #define pmd_pgtable(pmd) \
1732 	((pgtable_t)__va(pmd_val(pmd) & -sizeof(pte_t)*PTRS_PER_PTE))
1733 
1734 #endif /* _S390_PAGE_H */
1735