xref: /linux/arch/s390/include/asm/pgtable.h (revision 2eff01ee2881becc9daaa0d53477ec202136b1f4)
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/ctlreg.h>
22 #include <asm/bug.h>
23 #include <asm/page.h>
24 #include <asm/uv.h>
25 
26 extern pgd_t swapper_pg_dir[];
27 extern pgd_t invalid_pg_dir[];
28 extern void paging_init(void);
29 extern struct ctlreg s390_invalid_asce;
30 
31 enum {
32 	PG_DIRECT_MAP_4K = 0,
33 	PG_DIRECT_MAP_1M,
34 	PG_DIRECT_MAP_2G,
35 	PG_DIRECT_MAP_MAX
36 };
37 
38 extern atomic_long_t __bootdata_preserved(direct_pages_count[PG_DIRECT_MAP_MAX]);
39 
40 static inline void update_page_count(int level, long count)
41 {
42 	if (IS_ENABLED(CONFIG_PROC_FS))
43 		atomic_long_add(count, &direct_pages_count[level]);
44 }
45 
46 /*
47  * The S390 doesn't have any external MMU info: the kernel page
48  * tables contain all the necessary information.
49  */
50 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
51 #define update_mmu_cache_range(vmf, vma, addr, ptep, nr) 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 extern unsigned long __bootdata_preserved(MODULES_VADDR);
95 extern unsigned long __bootdata_preserved(MODULES_END);
96 #define MODULES_VADDR	MODULES_VADDR
97 #define MODULES_END	MODULES_END
98 #define MODULES_LEN	(1UL << 31)
99 
100 static inline int is_module_addr(void *addr)
101 {
102 	BUILD_BUG_ON(MODULES_LEN > (1UL << 31));
103 	if (addr < (void *)MODULES_VADDR)
104 		return 0;
105 	if (addr > (void *)MODULES_END)
106 		return 0;
107 	return 1;
108 }
109 
110 #ifdef CONFIG_KMSAN
111 #define KMSAN_VMALLOC_SIZE (VMALLOC_END - VMALLOC_START)
112 #define KMSAN_VMALLOC_SHADOW_START VMALLOC_END
113 #define KMSAN_VMALLOC_SHADOW_END (KMSAN_VMALLOC_SHADOW_START + KMSAN_VMALLOC_SIZE)
114 #define KMSAN_VMALLOC_ORIGIN_START KMSAN_VMALLOC_SHADOW_END
115 #define KMSAN_VMALLOC_ORIGIN_END (KMSAN_VMALLOC_ORIGIN_START + KMSAN_VMALLOC_SIZE)
116 #define KMSAN_MODULES_SHADOW_START KMSAN_VMALLOC_ORIGIN_END
117 #define KMSAN_MODULES_SHADOW_END (KMSAN_MODULES_SHADOW_START + MODULES_LEN)
118 #define KMSAN_MODULES_ORIGIN_START KMSAN_MODULES_SHADOW_END
119 #define KMSAN_MODULES_ORIGIN_END (KMSAN_MODULES_ORIGIN_START + MODULES_LEN)
120 #endif
121 
122 #ifdef CONFIG_RANDOMIZE_BASE
123 #define KASLR_LEN	(1UL << 31)
124 #else
125 #define KASLR_LEN	0UL
126 #endif
127 
128 /*
129  * A 64 bit pagetable entry of S390 has following format:
130  * |			 PFRA			      |0IPC|  OS  |
131  * 0000000000111111111122222222223333333333444444444455555555556666
132  * 0123456789012345678901234567890123456789012345678901234567890123
133  *
134  * I Page-Invalid Bit:    Page is not available for address-translation
135  * P Page-Protection Bit: Store access not possible for page
136  * C Change-bit override: HW is not required to set change bit
137  *
138  * A 64 bit segmenttable entry of S390 has following format:
139  * |        P-table origin                              |      TT
140  * 0000000000111111111122222222223333333333444444444455555555556666
141  * 0123456789012345678901234567890123456789012345678901234567890123
142  *
143  * I Segment-Invalid Bit:    Segment is not available for address-translation
144  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
145  * P Page-Protection Bit: Store access not possible for page
146  * TT Type 00
147  *
148  * A 64 bit region table entry of S390 has following format:
149  * |        S-table origin                             |   TF  TTTL
150  * 0000000000111111111122222222223333333333444444444455555555556666
151  * 0123456789012345678901234567890123456789012345678901234567890123
152  *
153  * I Segment-Invalid Bit:    Segment is not available for address-translation
154  * TT Type 01
155  * TF
156  * TL Table length
157  *
158  * The 64 bit regiontable origin of S390 has following format:
159  * |      region table origon                          |       DTTL
160  * 0000000000111111111122222222223333333333444444444455555555556666
161  * 0123456789012345678901234567890123456789012345678901234567890123
162  *
163  * X Space-Switch event:
164  * G Segment-Invalid Bit:
165  * P Private-Space Bit:
166  * S Storage-Alteration:
167  * R Real space
168  * TL Table-Length:
169  *
170  * A storage key has the following format:
171  * | ACC |F|R|C|0|
172  *  0   3 4 5 6 7
173  * ACC: access key
174  * F  : fetch protection bit
175  * R  : referenced bit
176  * C  : changed bit
177  */
178 
179 /* Hardware bits in the page table entry */
180 #define _PAGE_NOEXEC	0x100		/* HW no-execute bit  */
181 #define _PAGE_PROTECT	0x200		/* HW read-only bit  */
182 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
183 #define _PAGE_LARGE	0x800		/* Bit to mark a large pte */
184 
185 /* Software bits in the page table entry */
186 #define _PAGE_PRESENT	0x001		/* SW pte present bit */
187 #define _PAGE_YOUNG	0x004		/* SW pte young bit */
188 #define _PAGE_DIRTY	0x008		/* SW pte dirty bit */
189 #define _PAGE_READ	0x010		/* SW pte read bit */
190 #define _PAGE_WRITE	0x020		/* SW pte write bit */
191 #define _PAGE_SPECIAL	0x040		/* SW associated with special page */
192 #define _PAGE_UNUSED	0x080		/* SW bit for pgste usage state */
193 
194 #ifdef CONFIG_MEM_SOFT_DIRTY
195 #define _PAGE_SOFT_DIRTY 0x002		/* SW pte soft dirty bit */
196 #else
197 #define _PAGE_SOFT_DIRTY 0x000
198 #endif
199 
200 #define _PAGE_SW_BITS	0xffUL		/* All SW bits */
201 
202 #define _PAGE_SWP_EXCLUSIVE _PAGE_LARGE	/* SW pte exclusive swap bit */
203 
204 /* Set of bits not changed in pte_modify */
205 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
206 				 _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
207 
208 /*
209  * Mask of bits that must not be changed with RDP. Allow only _PAGE_PROTECT
210  * HW bit and all SW bits.
211  */
212 #define _PAGE_RDP_MASK		~(_PAGE_PROTECT | _PAGE_SW_BITS)
213 
214 /*
215  * handle_pte_fault uses pte_present and pte_none to find out the pte type
216  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
217  * distinguish present from not-present ptes. It is changed only with the page
218  * table lock held.
219  *
220  * The following table gives the different possible bit combinations for
221  * the pte hardware and software bits in the last 12 bits of a pte
222  * (. unassigned bit, x don't care, t swap type):
223  *
224  *				842100000000
225  *				000084210000
226  *				000000008421
227  *				.IR.uswrdy.p
228  * empty			.10.00000000
229  * swap				.11..ttttt.0
230  * prot-none, clean, old	.11.xx0000.1
231  * prot-none, clean, young	.11.xx0001.1
232  * prot-none, dirty, old	.11.xx0010.1
233  * prot-none, dirty, young	.11.xx0011.1
234  * read-only, clean, old	.11.xx0100.1
235  * read-only, clean, young	.01.xx0101.1
236  * read-only, dirty, old	.11.xx0110.1
237  * read-only, dirty, young	.01.xx0111.1
238  * read-write, clean, old	.11.xx1100.1
239  * read-write, clean, young	.01.xx1101.1
240  * read-write, dirty, old	.10.xx1110.1
241  * read-write, dirty, young	.00.xx1111.1
242  * HW-bits: R read-only, I invalid
243  * SW-bits: p present, y young, d dirty, r read, w write, s special,
244  *	    u unused, l large
245  *
246  * pte_none    is true for the bit pattern .10.00000000, pte == 0x400
247  * pte_swap    is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
248  * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
249  */
250 
251 /* Bits in the segment/region table address-space-control-element */
252 #define _ASCE_ORIGIN		~0xfffUL/* region/segment table origin	    */
253 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
254 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
255 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
256 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
257 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
258 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
259 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
260 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
261 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
262 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
263 
264 /* Bits in the region table entry */
265 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
266 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
267 #define _REGION_ENTRY_NOEXEC	0x100	/* region no-execute bit	    */
268 #define _REGION_ENTRY_OFFSET	0xc0	/* region table offset		    */
269 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
270 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region table type mask	    */
271 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
272 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
273 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
274 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
275 
276 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
277 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
278 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
279 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
280 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH | \
281 				 _REGION3_ENTRY_PRESENT)
282 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
283 
284 #define _REGION3_ENTRY_HARDWARE_BITS		0xfffffffffffff6ffUL
285 #define _REGION3_ENTRY_HARDWARE_BITS_LARGE	0xffffffff8001073cUL
286 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address	     */
287 #define _REGION3_ENTRY_DIRTY	0x2000	/* SW region dirty bit */
288 #define _REGION3_ENTRY_YOUNG	0x1000	/* SW region young bit */
289 #define _REGION3_ENTRY_COMM	0x0010	/* Common-Region, marks swap entry */
290 #define _REGION3_ENTRY_LARGE	0x0400	/* RTTE-format control, large page  */
291 #define _REGION3_ENTRY_WRITE	0x8000	/* SW region write bit */
292 #define _REGION3_ENTRY_READ	0x4000	/* SW region read bit */
293 
294 #ifdef CONFIG_MEM_SOFT_DIRTY
295 #define _REGION3_ENTRY_SOFT_DIRTY 0x0002 /* SW region soft dirty bit */
296 #else
297 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
298 #endif
299 
300 #define _REGION_ENTRY_BITS	 0xfffffffffffff22fUL
301 
302 /*
303  * SW region present bit. For non-leaf region-third-table entries, bits 62-63
304  * indicate the TABLE LENGTH and both must be set to 1. But such entries
305  * would always be considered as present, so it is safe to use bit 63 as
306  * PRESENT bit for PUD.
307  */
308 #define _REGION3_ENTRY_PRESENT	0x0001
309 
310 /* Bits in the segment table entry */
311 #define _SEGMENT_ENTRY_BITS			0xfffffffffffffe3fUL
312 #define _SEGMENT_ENTRY_HARDWARE_BITS		0xfffffffffffffe3cUL
313 #define _SEGMENT_ENTRY_HARDWARE_BITS_LARGE	0xfffffffffff1073cUL
314 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
315 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* page table origin		    */
316 #define _SEGMENT_ENTRY_PROTECT	0x200	/* segment protection bit	    */
317 #define _SEGMENT_ENTRY_NOEXEC	0x100	/* segment no-execute bit	    */
318 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
319 #define _SEGMENT_ENTRY_TYPE_MASK 0x0c	/* segment table type mask	    */
320 
321 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PRESENT)
322 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
323 
324 #define _SEGMENT_ENTRY_DIRTY	0x2000	/* SW segment dirty bit */
325 #define _SEGMENT_ENTRY_YOUNG	0x1000	/* SW segment young bit */
326 
327 #define _SEGMENT_ENTRY_COMM	0x0010	/* Common-Segment, marks swap entry */
328 #define _SEGMENT_ENTRY_LARGE	0x0400	/* STE-format control, large page */
329 #define _SEGMENT_ENTRY_WRITE	0x8000	/* SW segment write bit */
330 #define _SEGMENT_ENTRY_READ	0x4000	/* SW segment read bit */
331 
332 #ifdef CONFIG_MEM_SOFT_DIRTY
333 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0002 /* SW segment soft dirty bit */
334 #else
335 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
336 #endif
337 
338 #define _SEGMENT_ENTRY_PRESENT	0x0001	/* SW segment present bit */
339 
340 /* Common bits in region and segment table entries, for swap entries */
341 #define _RST_ENTRY_COMM		0x0010	/* Common-Region/Segment, marks swap entry */
342 #define _RST_ENTRY_INVALID	0x0020	/* invalid region/segment table entry */
343 
344 #define _CRST_ENTRIES	2048	/* number of region/segment table entries */
345 #define _PAGE_ENTRIES	256	/* number of page table entries	*/
346 
347 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
348 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
349 
350 #define _REGION1_SHIFT	53
351 #define _REGION2_SHIFT	42
352 #define _REGION3_SHIFT	31
353 #define _SEGMENT_SHIFT	20
354 
355 #define _REGION1_INDEX	(0x7ffUL << _REGION1_SHIFT)
356 #define _REGION2_INDEX	(0x7ffUL << _REGION2_SHIFT)
357 #define _REGION3_INDEX	(0x7ffUL << _REGION3_SHIFT)
358 #define _SEGMENT_INDEX	(0x7ffUL << _SEGMENT_SHIFT)
359 #define _PAGE_INDEX	(0xffUL  << PAGE_SHIFT)
360 
361 #define _REGION1_SIZE	(1UL << _REGION1_SHIFT)
362 #define _REGION2_SIZE	(1UL << _REGION2_SHIFT)
363 #define _REGION3_SIZE	(1UL << _REGION3_SHIFT)
364 #define _SEGMENT_SIZE	(1UL << _SEGMENT_SHIFT)
365 
366 #define _REGION1_MASK	(~(_REGION1_SIZE - 1))
367 #define _REGION2_MASK	(~(_REGION2_SIZE - 1))
368 #define _REGION3_MASK	(~(_REGION3_SIZE - 1))
369 #define _SEGMENT_MASK	(~(_SEGMENT_SIZE - 1))
370 
371 #define PMD_SHIFT	_SEGMENT_SHIFT
372 #define PUD_SHIFT	_REGION3_SHIFT
373 #define P4D_SHIFT	_REGION2_SHIFT
374 #define PGDIR_SHIFT	_REGION1_SHIFT
375 
376 #define PMD_SIZE	_SEGMENT_SIZE
377 #define PUD_SIZE	_REGION3_SIZE
378 #define P4D_SIZE	_REGION2_SIZE
379 #define PGDIR_SIZE	_REGION1_SIZE
380 
381 #define PMD_MASK	_SEGMENT_MASK
382 #define PUD_MASK	_REGION3_MASK
383 #define P4D_MASK	_REGION2_MASK
384 #define PGDIR_MASK	_REGION1_MASK
385 
386 #define PTRS_PER_PTE	_PAGE_ENTRIES
387 #define PTRS_PER_PMD	_CRST_ENTRIES
388 #define PTRS_PER_PUD	_CRST_ENTRIES
389 #define PTRS_PER_P4D	_CRST_ENTRIES
390 #define PTRS_PER_PGD	_CRST_ENTRIES
391 
392 /*
393  * Segment table and region3 table entry encoding
394  * (R = read-only, I = invalid, y = young bit):
395  *				dy..R...I...wr
396  * prot-none, clean, old	00..1...1...00
397  * prot-none, clean, young	01..1...1...00
398  * prot-none, dirty, old	10..1...1...00
399  * prot-none, dirty, young	11..1...1...00
400  * read-only, clean, old	00..1...1...01
401  * read-only, clean, young	01..1...0...01
402  * read-only, dirty, old	10..1...1...01
403  * read-only, dirty, young	11..1...0...01
404  * read-write, clean, old	00..1...1...11
405  * read-write, clean, young	01..1...0...11
406  * read-write, dirty, old	10..0...1...11
407  * read-write, dirty, young	11..0...0...11
408  * The segment table origin is used to distinguish empty (origin==0) from
409  * read-write, old segment table entries (origin!=0)
410  * HW-bits: R read-only, I invalid
411  * SW-bits: y young, d dirty, r read, w write
412  */
413 
414 /* Page status table bits for virtualization */
415 #define PGSTE_ACC_BITS	0xf000000000000000UL
416 #define PGSTE_FP_BIT	0x0800000000000000UL
417 #define PGSTE_PCL_BIT	0x0080000000000000UL
418 #define PGSTE_HR_BIT	0x0040000000000000UL
419 #define PGSTE_HC_BIT	0x0020000000000000UL
420 #define PGSTE_GR_BIT	0x0004000000000000UL
421 #define PGSTE_GC_BIT	0x0002000000000000UL
422 #define PGSTE_UC_BIT	0x0000800000000000UL	/* user dirty (migration) */
423 #define PGSTE_IN_BIT	0x0000400000000000UL	/* IPTE notify bit */
424 #define PGSTE_VSIE_BIT	0x0000200000000000UL	/* ref'd in a shadow table */
425 
426 /* Guest Page State used for virtualization */
427 #define _PGSTE_GPS_ZERO			0x0000000080000000UL
428 #define _PGSTE_GPS_NODAT		0x0000000040000000UL
429 #define _PGSTE_GPS_USAGE_MASK		0x0000000003000000UL
430 #define _PGSTE_GPS_USAGE_STABLE		0x0000000000000000UL
431 #define _PGSTE_GPS_USAGE_UNUSED		0x0000000001000000UL
432 #define _PGSTE_GPS_USAGE_POT_VOLATILE	0x0000000002000000UL
433 #define _PGSTE_GPS_USAGE_VOLATILE	_PGSTE_GPS_USAGE_MASK
434 
435 /*
436  * A user page table pointer has the space-switch-event bit, the
437  * private-space-control bit and the storage-alteration-event-control
438  * bit set. A kernel page table pointer doesn't need them.
439  */
440 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
441 				 _ASCE_ALT_EVENT)
442 
443 /*
444  * Page protection definitions.
445  */
446 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID | _PAGE_PROTECT)
447 #define PAGE_RO		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
448 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
449 #define PAGE_RX		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
450 				 _PAGE_INVALID | _PAGE_PROTECT)
451 #define PAGE_RW		__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
452 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
453 #define PAGE_RWX	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
454 				 _PAGE_INVALID | _PAGE_PROTECT)
455 
456 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
457 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
458 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
459 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
460 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
461 				 _PAGE_PROTECT | _PAGE_NOEXEC)
462 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
463 				  _PAGE_YOUNG |	_PAGE_DIRTY)
464 
465 /*
466  * On s390 the page table entry has an invalid bit and a read-only bit.
467  * Read permission implies execute permission and write permission
468  * implies read permission.
469  */
470          /*xwr*/
471 
472 /*
473  * Segment entry (large page) protection definitions.
474  */
475 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_PRESENT | \
476 				 _SEGMENT_ENTRY_INVALID | \
477 				 _SEGMENT_ENTRY_PROTECT)
478 #define SEGMENT_RO	__pgprot(_SEGMENT_ENTRY_PRESENT | \
479 				 _SEGMENT_ENTRY_PROTECT | \
480 				 _SEGMENT_ENTRY_READ | \
481 				 _SEGMENT_ENTRY_NOEXEC)
482 #define SEGMENT_RX	__pgprot(_SEGMENT_ENTRY_PRESENT | \
483 				 _SEGMENT_ENTRY_PROTECT | \
484 				 _SEGMENT_ENTRY_READ)
485 #define SEGMENT_RW	__pgprot(_SEGMENT_ENTRY_PRESENT | \
486 				 _SEGMENT_ENTRY_READ | \
487 				 _SEGMENT_ENTRY_WRITE | \
488 				 _SEGMENT_ENTRY_NOEXEC)
489 #define SEGMENT_RWX	__pgprot(_SEGMENT_ENTRY_PRESENT | \
490 				 _SEGMENT_ENTRY_READ | \
491 				 _SEGMENT_ENTRY_WRITE)
492 #define SEGMENT_KERNEL	__pgprot(_SEGMENT_ENTRY |	\
493 				 _SEGMENT_ENTRY_LARGE |	\
494 				 _SEGMENT_ENTRY_READ |	\
495 				 _SEGMENT_ENTRY_WRITE | \
496 				 _SEGMENT_ENTRY_YOUNG | \
497 				 _SEGMENT_ENTRY_DIRTY | \
498 				 _SEGMENT_ENTRY_NOEXEC)
499 #define SEGMENT_KERNEL_RO __pgprot(_SEGMENT_ENTRY |	\
500 				 _SEGMENT_ENTRY_LARGE |	\
501 				 _SEGMENT_ENTRY_READ |	\
502 				 _SEGMENT_ENTRY_YOUNG |	\
503 				 _SEGMENT_ENTRY_PROTECT | \
504 				 _SEGMENT_ENTRY_NOEXEC)
505 #define SEGMENT_KERNEL_EXEC __pgprot(_SEGMENT_ENTRY |	\
506 				 _SEGMENT_ENTRY_LARGE |	\
507 				 _SEGMENT_ENTRY_READ |	\
508 				 _SEGMENT_ENTRY_WRITE | \
509 				 _SEGMENT_ENTRY_YOUNG |	\
510 				 _SEGMENT_ENTRY_DIRTY)
511 
512 /*
513  * Region3 entry (large page) protection definitions.
514  */
515 
516 #define REGION3_KERNEL	__pgprot(_REGION_ENTRY_TYPE_R3 | \
517 				 _REGION3_ENTRY_PRESENT | \
518 				 _REGION3_ENTRY_LARGE |	 \
519 				 _REGION3_ENTRY_READ |	 \
520 				 _REGION3_ENTRY_WRITE |	 \
521 				 _REGION3_ENTRY_YOUNG |	 \
522 				 _REGION3_ENTRY_DIRTY | \
523 				 _REGION_ENTRY_NOEXEC)
524 #define REGION3_KERNEL_RO __pgprot(_REGION_ENTRY_TYPE_R3 | \
525 				   _REGION3_ENTRY_PRESENT | \
526 				   _REGION3_ENTRY_LARGE |  \
527 				   _REGION3_ENTRY_READ |   \
528 				   _REGION3_ENTRY_YOUNG |  \
529 				   _REGION_ENTRY_PROTECT | \
530 				   _REGION_ENTRY_NOEXEC)
531 #define REGION3_KERNEL_EXEC __pgprot(_REGION_ENTRY_TYPE_R3 | \
532 				 _REGION3_ENTRY_PRESENT | \
533 				 _REGION3_ENTRY_LARGE |	 \
534 				 _REGION3_ENTRY_READ |	 \
535 				 _REGION3_ENTRY_WRITE |	 \
536 				 _REGION3_ENTRY_YOUNG |	 \
537 				 _REGION3_ENTRY_DIRTY)
538 
539 static inline bool mm_p4d_folded(struct mm_struct *mm)
540 {
541 	return mm->context.asce_limit <= _REGION1_SIZE;
542 }
543 #define mm_p4d_folded(mm) mm_p4d_folded(mm)
544 
545 static inline bool mm_pud_folded(struct mm_struct *mm)
546 {
547 	return mm->context.asce_limit <= _REGION2_SIZE;
548 }
549 #define mm_pud_folded(mm) mm_pud_folded(mm)
550 
551 static inline bool mm_pmd_folded(struct mm_struct *mm)
552 {
553 	return mm->context.asce_limit <= _REGION3_SIZE;
554 }
555 #define mm_pmd_folded(mm) mm_pmd_folded(mm)
556 
557 static inline int mm_has_pgste(struct mm_struct *mm)
558 {
559 #ifdef CONFIG_PGSTE
560 	if (unlikely(mm->context.has_pgste))
561 		return 1;
562 #endif
563 	return 0;
564 }
565 
566 static inline int mm_is_protected(struct mm_struct *mm)
567 {
568 #ifdef CONFIG_PGSTE
569 	if (unlikely(atomic_read(&mm->context.protected_count)))
570 		return 1;
571 #endif
572 	return 0;
573 }
574 
575 static inline int mm_alloc_pgste(struct mm_struct *mm)
576 {
577 #ifdef CONFIG_PGSTE
578 	if (unlikely(mm->context.alloc_pgste))
579 		return 1;
580 #endif
581 	return 0;
582 }
583 
584 static inline pte_t clear_pte_bit(pte_t pte, pgprot_t prot)
585 {
586 	return __pte(pte_val(pte) & ~pgprot_val(prot));
587 }
588 
589 static inline pte_t set_pte_bit(pte_t pte, pgprot_t prot)
590 {
591 	return __pte(pte_val(pte) | pgprot_val(prot));
592 }
593 
594 static inline pmd_t clear_pmd_bit(pmd_t pmd, pgprot_t prot)
595 {
596 	return __pmd(pmd_val(pmd) & ~pgprot_val(prot));
597 }
598 
599 static inline pmd_t set_pmd_bit(pmd_t pmd, pgprot_t prot)
600 {
601 	return __pmd(pmd_val(pmd) | pgprot_val(prot));
602 }
603 
604 static inline pud_t clear_pud_bit(pud_t pud, pgprot_t prot)
605 {
606 	return __pud(pud_val(pud) & ~pgprot_val(prot));
607 }
608 
609 static inline pud_t set_pud_bit(pud_t pud, pgprot_t prot)
610 {
611 	return __pud(pud_val(pud) | pgprot_val(prot));
612 }
613 
614 /*
615  * As soon as the guest uses storage keys or enables PV, we deduplicate all
616  * mapped shared zeropages and prevent new shared zeropages from getting
617  * mapped.
618  */
619 #define mm_forbids_zeropage mm_forbids_zeropage
620 static inline int mm_forbids_zeropage(struct mm_struct *mm)
621 {
622 #ifdef CONFIG_PGSTE
623 	if (!mm->context.allow_cow_sharing)
624 		return 1;
625 #endif
626 	return 0;
627 }
628 
629 static inline int mm_uses_skeys(struct mm_struct *mm)
630 {
631 #ifdef CONFIG_PGSTE
632 	if (mm->context.uses_skeys)
633 		return 1;
634 #endif
635 	return 0;
636 }
637 
638 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
639 {
640 	union register_pair r1 = { .even = old, .odd = new, };
641 	unsigned long address = (unsigned long)ptr | 1;
642 
643 	asm volatile(
644 		"	csp	%[r1],%[address]"
645 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
646 		: [address] "d" (address)
647 		: "cc");
648 }
649 
650 /**
651  * cspg() - Compare and Swap and Purge (CSPG)
652  * @ptr: Pointer to the value to be exchanged
653  * @old: The expected old value
654  * @new: The new value
655  *
656  * Return: True if compare and swap was successful, otherwise false.
657  */
658 static inline bool cspg(unsigned long *ptr, unsigned long old, unsigned long new)
659 {
660 	union register_pair r1 = { .even = old, .odd = new, };
661 	unsigned long address = (unsigned long)ptr | 1;
662 
663 	asm volatile(
664 		"	cspg	%[r1],%[address]"
665 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
666 		: [address] "d" (address)
667 		: "cc");
668 	return old == r1.even;
669 }
670 
671 #define CRDTE_DTT_PAGE		0x00UL
672 #define CRDTE_DTT_SEGMENT	0x10UL
673 #define CRDTE_DTT_REGION3	0x14UL
674 #define CRDTE_DTT_REGION2	0x18UL
675 #define CRDTE_DTT_REGION1	0x1cUL
676 
677 /**
678  * crdte() - Compare and Replace DAT Table Entry
679  * @old:     The expected old value
680  * @new:     The new value
681  * @table:   Pointer to the value to be exchanged
682  * @dtt:     Table type of the table to be exchanged
683  * @address: The address mapped by the entry to be replaced
684  * @asce:    The ASCE of this entry
685  *
686  * Return: True if compare and replace was successful, otherwise false.
687  */
688 static inline bool crdte(unsigned long old, unsigned long new,
689 			 unsigned long *table, unsigned long dtt,
690 			 unsigned long address, unsigned long asce)
691 {
692 	union register_pair r1 = { .even = old, .odd = new, };
693 	union register_pair r2 = { .even = __pa(table) | dtt, .odd = address, };
694 
695 	asm volatile(".insn rrf,0xb98f0000,%[r1],%[r2],%[asce],0"
696 		     : [r1] "+&d" (r1.pair)
697 		     : [r2] "d" (r2.pair), [asce] "a" (asce)
698 		     : "memory", "cc");
699 	return old == r1.even;
700 }
701 
702 /*
703  * pgd/p4d/pud/pmd/pte query functions
704  */
705 static inline int pgd_folded(pgd_t pgd)
706 {
707 	return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
708 }
709 
710 static inline int pgd_present(pgd_t pgd)
711 {
712 	if (pgd_folded(pgd))
713 		return 1;
714 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
715 }
716 
717 static inline int pgd_none(pgd_t pgd)
718 {
719 	if (pgd_folded(pgd))
720 		return 0;
721 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
722 }
723 
724 static inline int pgd_bad(pgd_t pgd)
725 {
726 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1)
727 		return 0;
728 	return (pgd_val(pgd) & ~_REGION_ENTRY_BITS) != 0;
729 }
730 
731 static inline unsigned long pgd_pfn(pgd_t pgd)
732 {
733 	unsigned long origin_mask;
734 
735 	origin_mask = _REGION_ENTRY_ORIGIN;
736 	return (pgd_val(pgd) & origin_mask) >> PAGE_SHIFT;
737 }
738 
739 static inline int p4d_folded(p4d_t p4d)
740 {
741 	return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
742 }
743 
744 static inline int p4d_present(p4d_t p4d)
745 {
746 	if (p4d_folded(p4d))
747 		return 1;
748 	return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
749 }
750 
751 static inline int p4d_none(p4d_t p4d)
752 {
753 	if (p4d_folded(p4d))
754 		return 0;
755 	return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
756 }
757 
758 static inline unsigned long p4d_pfn(p4d_t p4d)
759 {
760 	unsigned long origin_mask;
761 
762 	origin_mask = _REGION_ENTRY_ORIGIN;
763 	return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
764 }
765 
766 static inline int pud_folded(pud_t pud)
767 {
768 	return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
769 }
770 
771 static inline int pud_present(pud_t pud)
772 {
773 	if (pud_folded(pud))
774 		return 1;
775 	return (pud_val(pud) & _REGION3_ENTRY_PRESENT) != 0;
776 }
777 
778 static inline int pud_none(pud_t pud)
779 {
780 	if (pud_folded(pud))
781 		return 0;
782 	return pud_val(pud) == _REGION3_ENTRY_EMPTY;
783 }
784 
785 #define pud_leaf pud_leaf
786 static inline bool pud_leaf(pud_t pud)
787 {
788 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
789 		return 0;
790 	return (pud_present(pud) && (pud_val(pud) & _REGION3_ENTRY_LARGE) != 0);
791 }
792 
793 static inline int pmd_present(pmd_t pmd)
794 {
795 	return (pmd_val(pmd) & _SEGMENT_ENTRY_PRESENT) != 0;
796 }
797 
798 #define pmd_leaf pmd_leaf
799 static inline bool pmd_leaf(pmd_t pmd)
800 {
801 	return (pmd_present(pmd) && (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0);
802 }
803 
804 static inline int pmd_bad(pmd_t pmd)
805 {
806 	if ((pmd_val(pmd) & _SEGMENT_ENTRY_TYPE_MASK) > 0 || pmd_leaf(pmd))
807 		return 1;
808 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
809 }
810 
811 static inline int pud_bad(pud_t pud)
812 {
813 	unsigned long type = pud_val(pud) & _REGION_ENTRY_TYPE_MASK;
814 
815 	if (type > _REGION_ENTRY_TYPE_R3 || pud_leaf(pud))
816 		return 1;
817 	if (type < _REGION_ENTRY_TYPE_R3)
818 		return 0;
819 	return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
820 }
821 
822 static inline int p4d_bad(p4d_t p4d)
823 {
824 	unsigned long type = p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK;
825 
826 	if (type > _REGION_ENTRY_TYPE_R2)
827 		return 1;
828 	if (type < _REGION_ENTRY_TYPE_R2)
829 		return 0;
830 	return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
831 }
832 
833 static inline int pmd_none(pmd_t pmd)
834 {
835 	return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
836 }
837 
838 #define pmd_write pmd_write
839 static inline int pmd_write(pmd_t pmd)
840 {
841 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
842 }
843 
844 #define pud_write pud_write
845 static inline int pud_write(pud_t pud)
846 {
847 	return (pud_val(pud) & _REGION3_ENTRY_WRITE) != 0;
848 }
849 
850 #define pmd_dirty pmd_dirty
851 static inline int pmd_dirty(pmd_t pmd)
852 {
853 	return (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
854 }
855 
856 #define pmd_young pmd_young
857 static inline int pmd_young(pmd_t pmd)
858 {
859 	return (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
860 }
861 
862 static inline int pte_present(pte_t pte)
863 {
864 	/* Bit pattern: (pte & 0x001) == 0x001 */
865 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
866 }
867 
868 static inline int pte_none(pte_t pte)
869 {
870 	/* Bit pattern: pte == 0x400 */
871 	return pte_val(pte) == _PAGE_INVALID;
872 }
873 
874 static inline int pte_swap(pte_t pte)
875 {
876 	/* Bit pattern: (pte & 0x201) == 0x200 */
877 	return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
878 		== _PAGE_PROTECT;
879 }
880 
881 static inline int pte_special(pte_t pte)
882 {
883 	return (pte_val(pte) & _PAGE_SPECIAL);
884 }
885 
886 #define __HAVE_ARCH_PTE_SAME
887 static inline int pte_same(pte_t a, pte_t b)
888 {
889 	return pte_val(a) == pte_val(b);
890 }
891 
892 #ifdef CONFIG_NUMA_BALANCING
893 static inline int pte_protnone(pte_t pte)
894 {
895 	return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
896 }
897 
898 static inline int pmd_protnone(pmd_t pmd)
899 {
900 	/* pmd_leaf(pmd) implies pmd_present(pmd) */
901 	return pmd_leaf(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
902 }
903 #endif
904 
905 static inline int pte_swp_exclusive(pte_t pte)
906 {
907 	return pte_val(pte) & _PAGE_SWP_EXCLUSIVE;
908 }
909 
910 static inline pte_t pte_swp_mkexclusive(pte_t pte)
911 {
912 	return set_pte_bit(pte, __pgprot(_PAGE_SWP_EXCLUSIVE));
913 }
914 
915 static inline pte_t pte_swp_clear_exclusive(pte_t pte)
916 {
917 	return clear_pte_bit(pte, __pgprot(_PAGE_SWP_EXCLUSIVE));
918 }
919 
920 static inline int pte_soft_dirty(pte_t pte)
921 {
922 	return pte_val(pte) & _PAGE_SOFT_DIRTY;
923 }
924 #define pte_swp_soft_dirty pte_soft_dirty
925 
926 static inline pte_t pte_mksoft_dirty(pte_t pte)
927 {
928 	return set_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
929 }
930 #define pte_swp_mksoft_dirty pte_mksoft_dirty
931 
932 static inline pte_t pte_clear_soft_dirty(pte_t pte)
933 {
934 	return clear_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
935 }
936 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
937 
938 static inline int pmd_soft_dirty(pmd_t pmd)
939 {
940 	return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
941 }
942 
943 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
944 {
945 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
946 }
947 
948 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
949 {
950 	return clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
951 }
952 
953 /*
954  * query functions pte_write/pte_dirty/pte_young only work if
955  * pte_present() is true. Undefined behaviour if not..
956  */
957 static inline int pte_write(pte_t pte)
958 {
959 	return (pte_val(pte) & _PAGE_WRITE) != 0;
960 }
961 
962 static inline int pte_dirty(pte_t pte)
963 {
964 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
965 }
966 
967 static inline int pte_young(pte_t pte)
968 {
969 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
970 }
971 
972 #define __HAVE_ARCH_PTE_UNUSED
973 static inline int pte_unused(pte_t pte)
974 {
975 	return pte_val(pte) & _PAGE_UNUSED;
976 }
977 
978 /*
979  * Extract the pgprot value from the given pte while at the same time making it
980  * usable for kernel address space mappings where fault driven dirty and
981  * young/old accounting is not supported, i.e _PAGE_PROTECT and _PAGE_INVALID
982  * must not be set.
983  */
984 #define pte_pgprot pte_pgprot
985 static inline pgprot_t pte_pgprot(pte_t pte)
986 {
987 	unsigned long pte_flags = pte_val(pte) & _PAGE_CHG_MASK;
988 
989 	if (pte_write(pte))
990 		pte_flags |= pgprot_val(PAGE_KERNEL);
991 	else
992 		pte_flags |= pgprot_val(PAGE_KERNEL_RO);
993 	pte_flags |= pte_val(pte) & mio_wb_bit_mask;
994 
995 	return __pgprot(pte_flags);
996 }
997 
998 /*
999  * pgd/pmd/pte modification functions
1000  */
1001 
1002 static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
1003 {
1004 	WRITE_ONCE(*pgdp, pgd);
1005 }
1006 
1007 static inline void set_p4d(p4d_t *p4dp, p4d_t p4d)
1008 {
1009 	WRITE_ONCE(*p4dp, p4d);
1010 }
1011 
1012 static inline void set_pud(pud_t *pudp, pud_t pud)
1013 {
1014 	WRITE_ONCE(*pudp, pud);
1015 }
1016 
1017 static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
1018 {
1019 	WRITE_ONCE(*pmdp, pmd);
1020 }
1021 
1022 static inline void set_pte(pte_t *ptep, pte_t pte)
1023 {
1024 	WRITE_ONCE(*ptep, pte);
1025 }
1026 
1027 static inline void pgd_clear(pgd_t *pgd)
1028 {
1029 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
1030 		set_pgd(pgd, __pgd(_REGION1_ENTRY_EMPTY));
1031 }
1032 
1033 static inline void p4d_clear(p4d_t *p4d)
1034 {
1035 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1036 		set_p4d(p4d, __p4d(_REGION2_ENTRY_EMPTY));
1037 }
1038 
1039 static inline void pud_clear(pud_t *pud)
1040 {
1041 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1042 		set_pud(pud, __pud(_REGION3_ENTRY_EMPTY));
1043 }
1044 
1045 static inline void pmd_clear(pmd_t *pmdp)
1046 {
1047 	set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1048 }
1049 
1050 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
1051 {
1052 	set_pte(ptep, __pte(_PAGE_INVALID));
1053 }
1054 
1055 /*
1056  * The following pte modification functions only work if
1057  * pte_present() is true. Undefined behaviour if not..
1058  */
1059 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
1060 {
1061 	pte = clear_pte_bit(pte, __pgprot(~_PAGE_CHG_MASK));
1062 	pte = set_pte_bit(pte, newprot);
1063 	/*
1064 	 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
1065 	 * has the invalid bit set, clear it again for readable, young pages
1066 	 */
1067 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
1068 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
1069 	/*
1070 	 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
1071 	 * protection bit set, clear it again for writable, dirty pages
1072 	 */
1073 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
1074 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1075 	return pte;
1076 }
1077 
1078 static inline pte_t pte_wrprotect(pte_t pte)
1079 {
1080 	pte = clear_pte_bit(pte, __pgprot(_PAGE_WRITE));
1081 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1082 }
1083 
1084 static inline pte_t pte_mkwrite_novma(pte_t pte)
1085 {
1086 	pte = set_pte_bit(pte, __pgprot(_PAGE_WRITE));
1087 	if (pte_val(pte) & _PAGE_DIRTY)
1088 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1089 	return pte;
1090 }
1091 
1092 static inline pte_t pte_mkclean(pte_t pte)
1093 {
1094 	pte = clear_pte_bit(pte, __pgprot(_PAGE_DIRTY));
1095 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1096 }
1097 
1098 static inline pte_t pte_mkdirty(pte_t pte)
1099 {
1100 	pte = set_pte_bit(pte, __pgprot(_PAGE_DIRTY | _PAGE_SOFT_DIRTY));
1101 	if (pte_val(pte) & _PAGE_WRITE)
1102 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1103 	return pte;
1104 }
1105 
1106 static inline pte_t pte_mkold(pte_t pte)
1107 {
1108 	pte = clear_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1109 	return set_pte_bit(pte, __pgprot(_PAGE_INVALID));
1110 }
1111 
1112 static inline pte_t pte_mkyoung(pte_t pte)
1113 {
1114 	pte = set_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1115 	if (pte_val(pte) & _PAGE_READ)
1116 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
1117 	return pte;
1118 }
1119 
1120 static inline pte_t pte_mkspecial(pte_t pte)
1121 {
1122 	return set_pte_bit(pte, __pgprot(_PAGE_SPECIAL));
1123 }
1124 
1125 #ifdef CONFIG_HUGETLB_PAGE
1126 static inline pte_t pte_mkhuge(pte_t pte)
1127 {
1128 	return set_pte_bit(pte, __pgprot(_PAGE_LARGE));
1129 }
1130 #endif
1131 
1132 #define IPTE_GLOBAL	0
1133 #define	IPTE_LOCAL	1
1134 
1135 #define IPTE_NODAT	0x400
1136 #define IPTE_GUEST_ASCE	0x800
1137 
1138 static __always_inline void __ptep_rdp(unsigned long addr, pte_t *ptep,
1139 				       unsigned long opt, unsigned long asce,
1140 				       int local)
1141 {
1142 	unsigned long pto;
1143 
1144 	pto = __pa(ptep) & ~(PTRS_PER_PTE * sizeof(pte_t) - 1);
1145 	asm volatile(".insn rrf,0xb98b0000,%[r1],%[r2],%[asce],%[m4]"
1146 		     : "+m" (*ptep)
1147 		     : [r1] "a" (pto), [r2] "a" ((addr & PAGE_MASK) | opt),
1148 		       [asce] "a" (asce), [m4] "i" (local));
1149 }
1150 
1151 static __always_inline void __ptep_ipte(unsigned long address, pte_t *ptep,
1152 					unsigned long opt, unsigned long asce,
1153 					int local)
1154 {
1155 	unsigned long pto = __pa(ptep);
1156 
1157 	if (__builtin_constant_p(opt) && opt == 0) {
1158 		/* Invalidation + TLB flush for the pte */
1159 		asm volatile(
1160 			"	ipte	%[r1],%[r2],0,%[m4]"
1161 			: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
1162 			  [m4] "i" (local));
1163 		return;
1164 	}
1165 
1166 	/* Invalidate ptes with options + TLB flush of the ptes */
1167 	opt = opt | (asce & _ASCE_ORIGIN);
1168 	asm volatile(
1169 		"	ipte	%[r1],%[r2],%[r3],%[m4]"
1170 		: [r2] "+a" (address), [r3] "+a" (opt)
1171 		: [r1] "a" (pto), [m4] "i" (local) : "memory");
1172 }
1173 
1174 static __always_inline void __ptep_ipte_range(unsigned long address, int nr,
1175 					      pte_t *ptep, int local)
1176 {
1177 	unsigned long pto = __pa(ptep);
1178 
1179 	/* Invalidate a range of ptes + TLB flush of the ptes */
1180 	do {
1181 		asm volatile(
1182 			"	ipte %[r1],%[r2],%[r3],%[m4]"
1183 			: [r2] "+a" (address), [r3] "+a" (nr)
1184 			: [r1] "a" (pto), [m4] "i" (local) : "memory");
1185 	} while (nr != 255);
1186 }
1187 
1188 /*
1189  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1190  * both clear the TLB for the unmapped pte. The reason is that
1191  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1192  * to modify an active pte. The sequence is
1193  *   1) ptep_get_and_clear
1194  *   2) set_pte_at
1195  *   3) flush_tlb_range
1196  * On s390 the tlb needs to get flushed with the modification of the pte
1197  * if the pte is active. The only way how this can be implemented is to
1198  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1199  * is a nop.
1200  */
1201 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1202 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1203 
1204 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1205 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1206 					    unsigned long addr, pte_t *ptep)
1207 {
1208 	pte_t pte = *ptep;
1209 
1210 	pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1211 	return pte_young(pte);
1212 }
1213 
1214 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1215 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1216 					 unsigned long address, pte_t *ptep)
1217 {
1218 	return ptep_test_and_clear_young(vma, address, ptep);
1219 }
1220 
1221 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1222 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1223 				       unsigned long addr, pte_t *ptep)
1224 {
1225 	pte_t res;
1226 
1227 	res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1228 	/* At this point the reference through the mapping is still present */
1229 	if (mm_is_protected(mm) && pte_present(res))
1230 		uv_convert_from_secure_pte(res);
1231 	return res;
1232 }
1233 
1234 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1235 pte_t ptep_modify_prot_start(struct vm_area_struct *, unsigned long, pte_t *);
1236 void ptep_modify_prot_commit(struct vm_area_struct *, unsigned long,
1237 			     pte_t *, pte_t, pte_t);
1238 
1239 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1240 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1241 				     unsigned long addr, pte_t *ptep)
1242 {
1243 	pte_t res;
1244 
1245 	res = ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1246 	/* At this point the reference through the mapping is still present */
1247 	if (mm_is_protected(vma->vm_mm) && pte_present(res))
1248 		uv_convert_from_secure_pte(res);
1249 	return res;
1250 }
1251 
1252 /*
1253  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1254  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1255  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1256  * cannot be accessed while the batched unmap is running. In this case
1257  * full==1 and a simple pte_clear is enough. See tlb.h.
1258  */
1259 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1260 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1261 					    unsigned long addr,
1262 					    pte_t *ptep, int full)
1263 {
1264 	pte_t res;
1265 
1266 	if (full) {
1267 		res = *ptep;
1268 		set_pte(ptep, __pte(_PAGE_INVALID));
1269 	} else {
1270 		res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1271 	}
1272 	/* Nothing to do */
1273 	if (!mm_is_protected(mm) || !pte_present(res))
1274 		return res;
1275 	/*
1276 	 * At this point the reference through the mapping is still present.
1277 	 * The notifier should have destroyed all protected vCPUs at this
1278 	 * point, so the destroy should be successful.
1279 	 */
1280 	if (full && !uv_destroy_pte(res))
1281 		return res;
1282 	/*
1283 	 * If something went wrong and the page could not be destroyed, or
1284 	 * if this is not a mm teardown, the slower export is used as
1285 	 * fallback instead.
1286 	 */
1287 	uv_convert_from_secure_pte(res);
1288 	return res;
1289 }
1290 
1291 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1292 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1293 				      unsigned long addr, pte_t *ptep)
1294 {
1295 	pte_t pte = *ptep;
1296 
1297 	if (pte_write(pte))
1298 		ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1299 }
1300 
1301 /*
1302  * Check if PTEs only differ in _PAGE_PROTECT HW bit, but also allow SW PTE
1303  * bits in the comparison. Those might change e.g. because of dirty and young
1304  * tracking.
1305  */
1306 static inline int pte_allow_rdp(pte_t old, pte_t new)
1307 {
1308 	/*
1309 	 * Only allow changes from RO to RW
1310 	 */
1311 	if (!(pte_val(old) & _PAGE_PROTECT) || pte_val(new) & _PAGE_PROTECT)
1312 		return 0;
1313 
1314 	return (pte_val(old) & _PAGE_RDP_MASK) == (pte_val(new) & _PAGE_RDP_MASK);
1315 }
1316 
1317 static inline void flush_tlb_fix_spurious_fault(struct vm_area_struct *vma,
1318 						unsigned long address,
1319 						pte_t *ptep)
1320 {
1321 	/*
1322 	 * RDP might not have propagated the PTE protection reset to all CPUs,
1323 	 * so there could be spurious TLB protection faults.
1324 	 * NOTE: This will also be called when a racing pagetable update on
1325 	 * another thread already installed the correct PTE. Both cases cannot
1326 	 * really be distinguished.
1327 	 * Therefore, only do the local TLB flush when RDP can be used, and the
1328 	 * PTE does not have _PAGE_PROTECT set, to avoid unnecessary overhead.
1329 	 * A local RDP can be used to do the flush.
1330 	 */
1331 	if (MACHINE_HAS_RDP && !(pte_val(*ptep) & _PAGE_PROTECT))
1332 		__ptep_rdp(address, ptep, 0, 0, 1);
1333 }
1334 #define flush_tlb_fix_spurious_fault flush_tlb_fix_spurious_fault
1335 
1336 void ptep_reset_dat_prot(struct mm_struct *mm, unsigned long addr, pte_t *ptep,
1337 			 pte_t new);
1338 
1339 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1340 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1341 					unsigned long addr, pte_t *ptep,
1342 					pte_t entry, int dirty)
1343 {
1344 	if (pte_same(*ptep, entry))
1345 		return 0;
1346 	if (MACHINE_HAS_RDP && !mm_has_pgste(vma->vm_mm) && pte_allow_rdp(*ptep, entry))
1347 		ptep_reset_dat_prot(vma->vm_mm, addr, ptep, entry);
1348 	else
1349 		ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1350 	return 1;
1351 }
1352 
1353 /*
1354  * Additional functions to handle KVM guest page tables
1355  */
1356 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1357 		     pte_t *ptep, pte_t entry);
1358 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1359 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1360 		 pte_t *ptep, unsigned long bits);
1361 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1362 		    pte_t *ptep, int prot, unsigned long bit);
1363 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1364 		     pte_t *ptep , int reset);
1365 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1366 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1367 		    pte_t *sptep, pte_t *tptep, pte_t pte);
1368 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1369 
1370 bool ptep_test_and_clear_uc(struct mm_struct *mm, unsigned long address,
1371 			    pte_t *ptep);
1372 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1373 			  unsigned char key, bool nq);
1374 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1375 			       unsigned char key, unsigned char *oldkey,
1376 			       bool nq, bool mr, bool mc);
1377 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1378 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1379 			  unsigned char *key);
1380 
1381 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1382 				unsigned long bits, unsigned long value);
1383 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1384 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1385 			unsigned long *oldpte, unsigned long *oldpgste);
1386 void gmap_pmdp_csp(struct mm_struct *mm, unsigned long vmaddr);
1387 void gmap_pmdp_invalidate(struct mm_struct *mm, unsigned long vmaddr);
1388 void gmap_pmdp_idte_local(struct mm_struct *mm, unsigned long vmaddr);
1389 void gmap_pmdp_idte_global(struct mm_struct *mm, unsigned long vmaddr);
1390 
1391 #define pgprot_writecombine	pgprot_writecombine
1392 pgprot_t pgprot_writecombine(pgprot_t prot);
1393 
1394 #define pgprot_writethrough	pgprot_writethrough
1395 pgprot_t pgprot_writethrough(pgprot_t prot);
1396 
1397 #define PFN_PTE_SHIFT		PAGE_SHIFT
1398 
1399 /*
1400  * Set multiple PTEs to consecutive pages with a single call.  All PTEs
1401  * are within the same folio, PMD and VMA.
1402  */
1403 static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
1404 			      pte_t *ptep, pte_t entry, unsigned int nr)
1405 {
1406 	if (pte_present(entry))
1407 		entry = clear_pte_bit(entry, __pgprot(_PAGE_UNUSED));
1408 	if (mm_has_pgste(mm)) {
1409 		for (;;) {
1410 			ptep_set_pte_at(mm, addr, ptep, entry);
1411 			if (--nr == 0)
1412 				break;
1413 			ptep++;
1414 			entry = __pte(pte_val(entry) + PAGE_SIZE);
1415 			addr += PAGE_SIZE;
1416 		}
1417 	} else {
1418 		for (;;) {
1419 			set_pte(ptep, entry);
1420 			if (--nr == 0)
1421 				break;
1422 			ptep++;
1423 			entry = __pte(pte_val(entry) + PAGE_SIZE);
1424 		}
1425 	}
1426 }
1427 #define set_ptes set_ptes
1428 
1429 /*
1430  * Conversion functions: convert a page and protection to a page entry,
1431  * and a page entry and page directory to the page they refer to.
1432  */
1433 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1434 {
1435 	pte_t __pte;
1436 
1437 	__pte = __pte(physpage | pgprot_val(pgprot));
1438 	if (!MACHINE_HAS_NX)
1439 		__pte = clear_pte_bit(__pte, __pgprot(_PAGE_NOEXEC));
1440 	return pte_mkyoung(__pte);
1441 }
1442 
1443 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1444 {
1445 	unsigned long physpage = page_to_phys(page);
1446 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1447 
1448 	if (pte_write(__pte) && PageDirty(page))
1449 		__pte = pte_mkdirty(__pte);
1450 	return __pte;
1451 }
1452 
1453 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1454 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1455 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1456 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1457 
1458 #define p4d_deref(pud) ((unsigned long)__va(p4d_val(pud) & _REGION_ENTRY_ORIGIN))
1459 #define pgd_deref(pgd) ((unsigned long)__va(pgd_val(pgd) & _REGION_ENTRY_ORIGIN))
1460 
1461 static inline unsigned long pmd_deref(pmd_t pmd)
1462 {
1463 	unsigned long origin_mask;
1464 
1465 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
1466 	if (pmd_leaf(pmd))
1467 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
1468 	return (unsigned long)__va(pmd_val(pmd) & origin_mask);
1469 }
1470 
1471 static inline unsigned long pmd_pfn(pmd_t pmd)
1472 {
1473 	return __pa(pmd_deref(pmd)) >> PAGE_SHIFT;
1474 }
1475 
1476 static inline unsigned long pud_deref(pud_t pud)
1477 {
1478 	unsigned long origin_mask;
1479 
1480 	origin_mask = _REGION_ENTRY_ORIGIN;
1481 	if (pud_leaf(pud))
1482 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
1483 	return (unsigned long)__va(pud_val(pud) & origin_mask);
1484 }
1485 
1486 #define pud_pfn pud_pfn
1487 static inline unsigned long pud_pfn(pud_t pud)
1488 {
1489 	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
1490 }
1491 
1492 /*
1493  * The pgd_offset function *always* adds the index for the top-level
1494  * region/segment table. This is done to get a sequence like the
1495  * following to work:
1496  *	pgdp = pgd_offset(current->mm, addr);
1497  *	pgd = READ_ONCE(*pgdp);
1498  *	p4dp = p4d_offset(&pgd, addr);
1499  *	...
1500  * The subsequent p4d_offset, pud_offset and pmd_offset functions
1501  * only add an index if they dereferenced the pointer.
1502  */
1503 static inline pgd_t *pgd_offset_raw(pgd_t *pgd, unsigned long address)
1504 {
1505 	unsigned long rste;
1506 	unsigned int shift;
1507 
1508 	/* Get the first entry of the top level table */
1509 	rste = pgd_val(*pgd);
1510 	/* Pick up the shift from the table type of the first entry */
1511 	shift = ((rste & _REGION_ENTRY_TYPE_MASK) >> 2) * 11 + 20;
1512 	return pgd + ((address >> shift) & (PTRS_PER_PGD - 1));
1513 }
1514 
1515 #define pgd_offset(mm, address) pgd_offset_raw(READ_ONCE((mm)->pgd), address)
1516 
1517 static inline p4d_t *p4d_offset_lockless(pgd_t *pgdp, pgd_t pgd, unsigned long address)
1518 {
1519 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R1)
1520 		return (p4d_t *) pgd_deref(pgd) + p4d_index(address);
1521 	return (p4d_t *) pgdp;
1522 }
1523 #define p4d_offset_lockless p4d_offset_lockless
1524 
1525 static inline p4d_t *p4d_offset(pgd_t *pgdp, unsigned long address)
1526 {
1527 	return p4d_offset_lockless(pgdp, *pgdp, address);
1528 }
1529 
1530 static inline pud_t *pud_offset_lockless(p4d_t *p4dp, p4d_t p4d, unsigned long address)
1531 {
1532 	if ((p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R2)
1533 		return (pud_t *) p4d_deref(p4d) + pud_index(address);
1534 	return (pud_t *) p4dp;
1535 }
1536 #define pud_offset_lockless pud_offset_lockless
1537 
1538 static inline pud_t *pud_offset(p4d_t *p4dp, unsigned long address)
1539 {
1540 	return pud_offset_lockless(p4dp, *p4dp, address);
1541 }
1542 #define pud_offset pud_offset
1543 
1544 static inline pmd_t *pmd_offset_lockless(pud_t *pudp, pud_t pud, unsigned long address)
1545 {
1546 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R3)
1547 		return (pmd_t *) pud_deref(pud) + pmd_index(address);
1548 	return (pmd_t *) pudp;
1549 }
1550 #define pmd_offset_lockless pmd_offset_lockless
1551 
1552 static inline pmd_t *pmd_offset(pud_t *pudp, unsigned long address)
1553 {
1554 	return pmd_offset_lockless(pudp, *pudp, address);
1555 }
1556 #define pmd_offset pmd_offset
1557 
1558 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
1559 {
1560 	return (unsigned long) pmd_deref(pmd);
1561 }
1562 
1563 static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
1564 {
1565 	return end <= current->mm->context.asce_limit;
1566 }
1567 #define gup_fast_permitted gup_fast_permitted
1568 
1569 #define pfn_pte(pfn, pgprot)	mk_pte_phys(((pfn) << PAGE_SHIFT), (pgprot))
1570 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1571 #define pte_page(x) pfn_to_page(pte_pfn(x))
1572 
1573 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1574 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1575 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
1576 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
1577 
1578 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1579 {
1580 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1581 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1582 }
1583 
1584 static inline pmd_t pmd_mkwrite_novma(pmd_t pmd)
1585 {
1586 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1587 	if (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY)
1588 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1589 	return pmd;
1590 }
1591 
1592 static inline pmd_t pmd_mkclean(pmd_t pmd)
1593 {
1594 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY));
1595 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1596 }
1597 
1598 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1599 {
1600 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_SOFT_DIRTY));
1601 	if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1602 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1603 	return pmd;
1604 }
1605 
1606 static inline pud_t pud_wrprotect(pud_t pud)
1607 {
1608 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1609 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1610 }
1611 
1612 static inline pud_t pud_mkwrite(pud_t pud)
1613 {
1614 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1615 	if (pud_val(pud) & _REGION3_ENTRY_DIRTY)
1616 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1617 	return pud;
1618 }
1619 
1620 static inline pud_t pud_mkclean(pud_t pud)
1621 {
1622 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY));
1623 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1624 }
1625 
1626 static inline pud_t pud_mkdirty(pud_t pud)
1627 {
1628 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY | _REGION3_ENTRY_SOFT_DIRTY));
1629 	if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1630 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1631 	return pud;
1632 }
1633 
1634 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1635 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1636 {
1637 	/*
1638 	 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1639 	 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1640 	 */
1641 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1642 		return pgprot_val(SEGMENT_NONE);
1643 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1644 		return pgprot_val(SEGMENT_RO);
1645 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1646 		return pgprot_val(SEGMENT_RX);
1647 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1648 		return pgprot_val(SEGMENT_RW);
1649 	return pgprot_val(SEGMENT_RWX);
1650 }
1651 
1652 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1653 {
1654 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1655 	if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1656 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1657 	return pmd;
1658 }
1659 
1660 static inline pmd_t pmd_mkold(pmd_t pmd)
1661 {
1662 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1663 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1664 }
1665 
1666 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1667 {
1668 	unsigned long mask;
1669 
1670 	mask  = _SEGMENT_ENTRY_ORIGIN_LARGE;
1671 	mask |= _SEGMENT_ENTRY_DIRTY;
1672 	mask |= _SEGMENT_ENTRY_YOUNG;
1673 	mask |=	_SEGMENT_ENTRY_LARGE;
1674 	mask |= _SEGMENT_ENTRY_SOFT_DIRTY;
1675 	pmd = __pmd(pmd_val(pmd) & mask);
1676 	pmd = set_pmd_bit(pmd, __pgprot(massage_pgprot_pmd(newprot)));
1677 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1678 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1679 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1680 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1681 	return pmd;
1682 }
1683 
1684 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1685 {
1686 	return __pmd(physpage + massage_pgprot_pmd(pgprot));
1687 }
1688 
1689 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1690 
1691 static inline void __pmdp_csp(pmd_t *pmdp)
1692 {
1693 	csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1694 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1695 }
1696 
1697 #define IDTE_GLOBAL	0
1698 #define IDTE_LOCAL	1
1699 
1700 #define IDTE_PTOA	0x0800
1701 #define IDTE_NODAT	0x1000
1702 #define IDTE_GUEST_ASCE	0x2000
1703 
1704 static __always_inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1705 					unsigned long opt, unsigned long asce,
1706 					int local)
1707 {
1708 	unsigned long sto;
1709 
1710 	sto = __pa(pmdp) - pmd_index(addr) * sizeof(pmd_t);
1711 	if (__builtin_constant_p(opt) && opt == 0) {
1712 		/* flush without guest asce */
1713 		asm volatile(
1714 			"	idte	%[r1],0,%[r2],%[m4]"
1715 			: "+m" (*pmdp)
1716 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1717 			  [m4] "i" (local)
1718 			: "cc" );
1719 	} else {
1720 		/* flush with guest asce */
1721 		asm volatile(
1722 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1723 			: "+m" (*pmdp)
1724 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1725 			  [r3] "a" (asce), [m4] "i" (local)
1726 			: "cc" );
1727 	}
1728 }
1729 
1730 static __always_inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1731 					unsigned long opt, unsigned long asce,
1732 					int local)
1733 {
1734 	unsigned long r3o;
1735 
1736 	r3o = __pa(pudp) - pud_index(addr) * sizeof(pud_t);
1737 	r3o |= _ASCE_TYPE_REGION3;
1738 	if (__builtin_constant_p(opt) && opt == 0) {
1739 		/* flush without guest asce */
1740 		asm volatile(
1741 			"	idte	%[r1],0,%[r2],%[m4]"
1742 			: "+m" (*pudp)
1743 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1744 			  [m4] "i" (local)
1745 			: "cc");
1746 	} else {
1747 		/* flush with guest asce */
1748 		asm volatile(
1749 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1750 			: "+m" (*pudp)
1751 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1752 			  [r3] "a" (asce), [m4] "i" (local)
1753 			: "cc" );
1754 	}
1755 }
1756 
1757 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1758 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1759 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1760 
1761 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1762 
1763 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1764 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1765 				pgtable_t pgtable);
1766 
1767 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1768 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1769 
1770 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1771 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1772 					unsigned long addr, pmd_t *pmdp,
1773 					pmd_t entry, int dirty)
1774 {
1775 	VM_BUG_ON(addr & ~HPAGE_MASK);
1776 
1777 	entry = pmd_mkyoung(entry);
1778 	if (dirty)
1779 		entry = pmd_mkdirty(entry);
1780 	if (pmd_val(*pmdp) == pmd_val(entry))
1781 		return 0;
1782 	pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1783 	return 1;
1784 }
1785 
1786 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1787 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1788 					    unsigned long addr, pmd_t *pmdp)
1789 {
1790 	pmd_t pmd = *pmdp;
1791 
1792 	pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1793 	return pmd_young(pmd);
1794 }
1795 
1796 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1797 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1798 					 unsigned long addr, pmd_t *pmdp)
1799 {
1800 	VM_BUG_ON(addr & ~HPAGE_MASK);
1801 	return pmdp_test_and_clear_young(vma, addr, pmdp);
1802 }
1803 
1804 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1805 			      pmd_t *pmdp, pmd_t entry)
1806 {
1807 	if (!MACHINE_HAS_NX)
1808 		entry = clear_pmd_bit(entry, __pgprot(_SEGMENT_ENTRY_NOEXEC));
1809 	set_pmd(pmdp, entry);
1810 }
1811 
1812 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1813 {
1814 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_LARGE));
1815 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1816 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1817 }
1818 
1819 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1820 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1821 					    unsigned long addr, pmd_t *pmdp)
1822 {
1823 	return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1824 }
1825 
1826 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1827 static inline pmd_t pmdp_huge_get_and_clear_full(struct vm_area_struct *vma,
1828 						 unsigned long addr,
1829 						 pmd_t *pmdp, int full)
1830 {
1831 	if (full) {
1832 		pmd_t pmd = *pmdp;
1833 		set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1834 		return pmd;
1835 	}
1836 	return pmdp_xchg_lazy(vma->vm_mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1837 }
1838 
1839 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1840 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1841 					  unsigned long addr, pmd_t *pmdp)
1842 {
1843 	return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1844 }
1845 
1846 #define __HAVE_ARCH_PMDP_INVALIDATE
1847 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1848 				   unsigned long addr, pmd_t *pmdp)
1849 {
1850 	pmd_t pmd;
1851 
1852 	VM_WARN_ON_ONCE(!pmd_present(*pmdp));
1853 	pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1854 	return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1855 }
1856 
1857 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1858 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1859 				      unsigned long addr, pmd_t *pmdp)
1860 {
1861 	pmd_t pmd = *pmdp;
1862 
1863 	if (pmd_write(pmd))
1864 		pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1865 }
1866 
1867 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1868 					unsigned long address,
1869 					pmd_t *pmdp)
1870 {
1871 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1872 }
1873 #define pmdp_collapse_flush pmdp_collapse_flush
1874 
1875 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(((pfn) << PAGE_SHIFT), (pgprot))
1876 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1877 
1878 static inline int pmd_trans_huge(pmd_t pmd)
1879 {
1880 	return pmd_leaf(pmd);
1881 }
1882 
1883 #define has_transparent_hugepage has_transparent_hugepage
1884 static inline int has_transparent_hugepage(void)
1885 {
1886 	return MACHINE_HAS_EDAT1 ? 1 : 0;
1887 }
1888 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1889 
1890 /*
1891  * 64 bit swap entry format:
1892  * A page-table entry has some bits we have to treat in a special way.
1893  * Bits 54 and 63 are used to indicate the page type. Bit 53 marks the pte
1894  * as invalid.
1895  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1896  * |			  offset			|E11XX|type |S0|
1897  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1898  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1899  *
1900  * Bits 0-51 store the offset.
1901  * Bit 52 (E) is used to remember PG_anon_exclusive.
1902  * Bits 57-61 store the type.
1903  * Bit 62 (S) is used for softdirty tracking.
1904  * Bits 55 and 56 (X) are unused.
1905  */
1906 
1907 #define __SWP_OFFSET_MASK	((1UL << 52) - 1)
1908 #define __SWP_OFFSET_SHIFT	12
1909 #define __SWP_TYPE_MASK		((1UL << 5) - 1)
1910 #define __SWP_TYPE_SHIFT	2
1911 
1912 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1913 {
1914 	unsigned long pteval;
1915 
1916 	pteval = _PAGE_INVALID | _PAGE_PROTECT;
1917 	pteval |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1918 	pteval |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1919 	return __pte(pteval);
1920 }
1921 
1922 static inline unsigned long __swp_type(swp_entry_t entry)
1923 {
1924 	return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1925 }
1926 
1927 static inline unsigned long __swp_offset(swp_entry_t entry)
1928 {
1929 	return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1930 }
1931 
1932 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1933 {
1934 	return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1935 }
1936 
1937 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1938 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1939 
1940 /*
1941  * 64 bit swap entry format for REGION3 and SEGMENT table entries (RSTE)
1942  * Bits 59 and 63 are used to indicate the swap entry. Bit 58 marks the rste
1943  * as invalid.
1944  * A swap entry is indicated by bit pattern (rste & 0x011) == 0x010
1945  * |			  offset			|Xtype |11TT|S0|
1946  * |0000000000111111111122222222223333333333444444444455|555555|5566|66|
1947  * |0123456789012345678901234567890123456789012345678901|234567|8901|23|
1948  *
1949  * Bits 0-51 store the offset.
1950  * Bits 53-57 store the type.
1951  * Bit 62 (S) is used for softdirty tracking.
1952  * Bits 60-61 (TT) indicate the table type: 0x01 for REGION3 and 0x00 for SEGMENT.
1953  * Bit 52 (X) is unused.
1954  */
1955 
1956 #define __SWP_OFFSET_MASK_RSTE	((1UL << 52) - 1)
1957 #define __SWP_OFFSET_SHIFT_RSTE	12
1958 #define __SWP_TYPE_MASK_RSTE		((1UL << 5) - 1)
1959 #define __SWP_TYPE_SHIFT_RSTE	6
1960 
1961 /*
1962  * TT bits set to 0x00 == SEGMENT. For REGION3 entries, caller must add R3
1963  * bits 0x01. See also __set_huge_pte_at().
1964  */
1965 static inline unsigned long mk_swap_rste(unsigned long type, unsigned long offset)
1966 {
1967 	unsigned long rste;
1968 
1969 	rste = _RST_ENTRY_INVALID | _RST_ENTRY_COMM;
1970 	rste |= (offset & __SWP_OFFSET_MASK_RSTE) << __SWP_OFFSET_SHIFT_RSTE;
1971 	rste |= (type & __SWP_TYPE_MASK_RSTE) << __SWP_TYPE_SHIFT_RSTE;
1972 	return rste;
1973 }
1974 
1975 static inline unsigned long __swp_type_rste(swp_entry_t entry)
1976 {
1977 	return (entry.val >> __SWP_TYPE_SHIFT_RSTE) & __SWP_TYPE_MASK_RSTE;
1978 }
1979 
1980 static inline unsigned long __swp_offset_rste(swp_entry_t entry)
1981 {
1982 	return (entry.val >> __SWP_OFFSET_SHIFT_RSTE) & __SWP_OFFSET_MASK_RSTE;
1983 }
1984 
1985 #define __rste_to_swp_entry(rste)	((swp_entry_t) { rste })
1986 
1987 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1988 extern void vmem_remove_mapping(unsigned long start, unsigned long size);
1989 extern int __vmem_map_4k_page(unsigned long addr, unsigned long phys, pgprot_t prot, bool alloc);
1990 extern int vmem_map_4k_page(unsigned long addr, unsigned long phys, pgprot_t prot);
1991 extern void vmem_unmap_4k_page(unsigned long addr);
1992 extern pte_t *vmem_get_alloc_pte(unsigned long addr, bool alloc);
1993 extern int s390_enable_sie(void);
1994 extern int s390_enable_skey(void);
1995 extern void s390_reset_cmma(struct mm_struct *mm);
1996 
1997 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1998 #define HAVE_ARCH_UNMAPPED_AREA
1999 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
2000 
2001 #define pmd_pgtable(pmd) \
2002 	((pgtable_t)__va(pmd_val(pmd) & -sizeof(pte_t)*PTRS_PER_PTE))
2003 
2004 #endif /* _S390_PAGE_H */
2005