xref: /linux/arch/s390/include/asm/pgtable.h (revision 78c3925c048c752334873f56c3a3d1c9d53e0416)
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 /*
111  * A 64 bit pagetable entry of S390 has following format:
112  * |			 PFRA			      |0IPC|  OS  |
113  * 0000000000111111111122222222223333333333444444444455555555556666
114  * 0123456789012345678901234567890123456789012345678901234567890123
115  *
116  * I Page-Invalid Bit:    Page is not available for address-translation
117  * P Page-Protection Bit: Store access not possible for page
118  * C Change-bit override: HW is not required to set change bit
119  *
120  * A 64 bit segmenttable entry of S390 has following format:
121  * |        P-table origin                              |      TT
122  * 0000000000111111111122222222223333333333444444444455555555556666
123  * 0123456789012345678901234567890123456789012345678901234567890123
124  *
125  * I Segment-Invalid Bit:    Segment is not available for address-translation
126  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
127  * P Page-Protection Bit: Store access not possible for page
128  * TT Type 00
129  *
130  * A 64 bit region table entry of S390 has following format:
131  * |        S-table origin                             |   TF  TTTL
132  * 0000000000111111111122222222223333333333444444444455555555556666
133  * 0123456789012345678901234567890123456789012345678901234567890123
134  *
135  * I Segment-Invalid Bit:    Segment is not available for address-translation
136  * TT Type 01
137  * TF
138  * TL Table length
139  *
140  * The 64 bit regiontable origin of S390 has following format:
141  * |      region table origon                          |       DTTL
142  * 0000000000111111111122222222223333333333444444444455555555556666
143  * 0123456789012345678901234567890123456789012345678901234567890123
144  *
145  * X Space-Switch event:
146  * G Segment-Invalid Bit:
147  * P Private-Space Bit:
148  * S Storage-Alteration:
149  * R Real space
150  * TL Table-Length:
151  *
152  * A storage key has the following format:
153  * | ACC |F|R|C|0|
154  *  0   3 4 5 6 7
155  * ACC: access key
156  * F  : fetch protection bit
157  * R  : referenced bit
158  * C  : changed bit
159  */
160 
161 /* Hardware bits in the page table entry */
162 #define _PAGE_NOEXEC	0x100		/* HW no-execute bit  */
163 #define _PAGE_PROTECT	0x200		/* HW read-only bit  */
164 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
165 #define _PAGE_LARGE	0x800		/* Bit to mark a large pte */
166 
167 /* Software bits in the page table entry */
168 #define _PAGE_PRESENT	0x001		/* SW pte present bit */
169 #define _PAGE_YOUNG	0x004		/* SW pte young bit */
170 #define _PAGE_DIRTY	0x008		/* SW pte dirty bit */
171 #define _PAGE_READ	0x010		/* SW pte read bit */
172 #define _PAGE_WRITE	0x020		/* SW pte write bit */
173 #define _PAGE_SPECIAL	0x040		/* SW associated with special page */
174 #define _PAGE_UNUSED	0x080		/* SW bit for pgste usage state */
175 
176 #ifdef CONFIG_MEM_SOFT_DIRTY
177 #define _PAGE_SOFT_DIRTY 0x002		/* SW pte soft dirty bit */
178 #else
179 #define _PAGE_SOFT_DIRTY 0x000
180 #endif
181 
182 #define _PAGE_SW_BITS	0xffUL		/* All SW bits */
183 
184 #define _PAGE_SWP_EXCLUSIVE _PAGE_LARGE	/* SW pte exclusive swap bit */
185 
186 /* Set of bits not changed in pte_modify */
187 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
188 				 _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
189 
190 /*
191  * Mask of bits that must not be changed with RDP. Allow only _PAGE_PROTECT
192  * HW bit and all SW bits.
193  */
194 #define _PAGE_RDP_MASK		~(_PAGE_PROTECT | _PAGE_SW_BITS)
195 
196 /*
197  * handle_pte_fault uses pte_present and pte_none to find out the pte type
198  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
199  * distinguish present from not-present ptes. It is changed only with the page
200  * table lock held.
201  *
202  * The following table gives the different possible bit combinations for
203  * the pte hardware and software bits in the last 12 bits of a pte
204  * (. unassigned bit, x don't care, t swap type):
205  *
206  *				842100000000
207  *				000084210000
208  *				000000008421
209  *				.IR.uswrdy.p
210  * empty			.10.00000000
211  * swap				.11..ttttt.0
212  * prot-none, clean, old	.11.xx0000.1
213  * prot-none, clean, young	.11.xx0001.1
214  * prot-none, dirty, old	.11.xx0010.1
215  * prot-none, dirty, young	.11.xx0011.1
216  * read-only, clean, old	.11.xx0100.1
217  * read-only, clean, young	.01.xx0101.1
218  * read-only, dirty, old	.11.xx0110.1
219  * read-only, dirty, young	.01.xx0111.1
220  * read-write, clean, old	.11.xx1100.1
221  * read-write, clean, young	.01.xx1101.1
222  * read-write, dirty, old	.10.xx1110.1
223  * read-write, dirty, young	.00.xx1111.1
224  * HW-bits: R read-only, I invalid
225  * SW-bits: p present, y young, d dirty, r read, w write, s special,
226  *	    u unused, l large
227  *
228  * pte_none    is true for the bit pattern .10.00000000, pte == 0x400
229  * pte_swap    is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
230  * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
231  */
232 
233 /* Bits in the segment/region table address-space-control-element */
234 #define _ASCE_ORIGIN		~0xfffUL/* region/segment table origin	    */
235 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
236 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
237 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
238 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
239 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
240 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
241 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
242 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
243 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
244 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
245 
246 /* Bits in the region table entry */
247 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
248 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
249 #define _REGION_ENTRY_NOEXEC	0x100	/* region no-execute bit	    */
250 #define _REGION_ENTRY_OFFSET	0xc0	/* region table offset		    */
251 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
252 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region table type mask	    */
253 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
254 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
255 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
256 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
257 
258 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
259 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
260 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
261 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
262 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
263 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
264 
265 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address	     */
266 #define _REGION3_ENTRY_DIRTY	0x2000	/* SW region dirty bit */
267 #define _REGION3_ENTRY_YOUNG	0x1000	/* SW region young bit */
268 #define _REGION3_ENTRY_LARGE	0x0400	/* RTTE-format control, large page  */
269 #define _REGION3_ENTRY_READ	0x0002	/* SW region read bit */
270 #define _REGION3_ENTRY_WRITE	0x0001	/* SW region write bit */
271 
272 #ifdef CONFIG_MEM_SOFT_DIRTY
273 #define _REGION3_ENTRY_SOFT_DIRTY 0x4000 /* SW region soft dirty bit */
274 #else
275 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
276 #endif
277 
278 #define _REGION_ENTRY_BITS	 0xfffffffffffff22fUL
279 
280 /* Bits in the segment table entry */
281 #define _SEGMENT_ENTRY_BITS			0xfffffffffffffe33UL
282 #define _SEGMENT_ENTRY_HARDWARE_BITS		0xfffffffffffffe30UL
283 #define _SEGMENT_ENTRY_HARDWARE_BITS_LARGE	0xfffffffffff00730UL
284 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
285 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* page table origin		    */
286 #define _SEGMENT_ENTRY_PROTECT	0x200	/* segment protection bit	    */
287 #define _SEGMENT_ENTRY_NOEXEC	0x100	/* segment no-execute bit	    */
288 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
289 #define _SEGMENT_ENTRY_TYPE_MASK 0x0c	/* segment table type mask	    */
290 
291 #define _SEGMENT_ENTRY		(0)
292 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
293 
294 #define _SEGMENT_ENTRY_DIRTY	0x2000	/* SW segment dirty bit */
295 #define _SEGMENT_ENTRY_YOUNG	0x1000	/* SW segment young bit */
296 #define _SEGMENT_ENTRY_LARGE	0x0400	/* STE-format control, large page */
297 #define _SEGMENT_ENTRY_WRITE	0x0002	/* SW segment write bit */
298 #define _SEGMENT_ENTRY_READ	0x0001	/* SW segment read bit */
299 
300 #ifdef CONFIG_MEM_SOFT_DIRTY
301 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x4000 /* SW segment soft dirty bit */
302 #else
303 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
304 #endif
305 
306 #define _CRST_ENTRIES	2048	/* number of region/segment table entries */
307 #define _PAGE_ENTRIES	256	/* number of page table entries	*/
308 
309 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
310 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
311 
312 #define _REGION1_SHIFT	53
313 #define _REGION2_SHIFT	42
314 #define _REGION3_SHIFT	31
315 #define _SEGMENT_SHIFT	20
316 
317 #define _REGION1_INDEX	(0x7ffUL << _REGION1_SHIFT)
318 #define _REGION2_INDEX	(0x7ffUL << _REGION2_SHIFT)
319 #define _REGION3_INDEX	(0x7ffUL << _REGION3_SHIFT)
320 #define _SEGMENT_INDEX	(0x7ffUL << _SEGMENT_SHIFT)
321 #define _PAGE_INDEX	(0xffUL  << _PAGE_SHIFT)
322 
323 #define _REGION1_SIZE	(1UL << _REGION1_SHIFT)
324 #define _REGION2_SIZE	(1UL << _REGION2_SHIFT)
325 #define _REGION3_SIZE	(1UL << _REGION3_SHIFT)
326 #define _SEGMENT_SIZE	(1UL << _SEGMENT_SHIFT)
327 
328 #define _REGION1_MASK	(~(_REGION1_SIZE - 1))
329 #define _REGION2_MASK	(~(_REGION2_SIZE - 1))
330 #define _REGION3_MASK	(~(_REGION3_SIZE - 1))
331 #define _SEGMENT_MASK	(~(_SEGMENT_SIZE - 1))
332 
333 #define PMD_SHIFT	_SEGMENT_SHIFT
334 #define PUD_SHIFT	_REGION3_SHIFT
335 #define P4D_SHIFT	_REGION2_SHIFT
336 #define PGDIR_SHIFT	_REGION1_SHIFT
337 
338 #define PMD_SIZE	_SEGMENT_SIZE
339 #define PUD_SIZE	_REGION3_SIZE
340 #define P4D_SIZE	_REGION2_SIZE
341 #define PGDIR_SIZE	_REGION1_SIZE
342 
343 #define PMD_MASK	_SEGMENT_MASK
344 #define PUD_MASK	_REGION3_MASK
345 #define P4D_MASK	_REGION2_MASK
346 #define PGDIR_MASK	_REGION1_MASK
347 
348 #define PTRS_PER_PTE	_PAGE_ENTRIES
349 #define PTRS_PER_PMD	_CRST_ENTRIES
350 #define PTRS_PER_PUD	_CRST_ENTRIES
351 #define PTRS_PER_P4D	_CRST_ENTRIES
352 #define PTRS_PER_PGD	_CRST_ENTRIES
353 
354 /*
355  * Segment table and region3 table entry encoding
356  * (R = read-only, I = invalid, y = young bit):
357  *				dy..R...I...wr
358  * prot-none, clean, old	00..1...1...00
359  * prot-none, clean, young	01..1...1...00
360  * prot-none, dirty, old	10..1...1...00
361  * prot-none, dirty, young	11..1...1...00
362  * read-only, clean, old	00..1...1...01
363  * read-only, clean, young	01..1...0...01
364  * read-only, dirty, old	10..1...1...01
365  * read-only, dirty, young	11..1...0...01
366  * read-write, clean, old	00..1...1...11
367  * read-write, clean, young	01..1...0...11
368  * read-write, dirty, old	10..0...1...11
369  * read-write, dirty, young	11..0...0...11
370  * The segment table origin is used to distinguish empty (origin==0) from
371  * read-write, old segment table entries (origin!=0)
372  * HW-bits: R read-only, I invalid
373  * SW-bits: y young, d dirty, r read, w write
374  */
375 
376 /* Page status table bits for virtualization */
377 #define PGSTE_ACC_BITS	0xf000000000000000UL
378 #define PGSTE_FP_BIT	0x0800000000000000UL
379 #define PGSTE_PCL_BIT	0x0080000000000000UL
380 #define PGSTE_HR_BIT	0x0040000000000000UL
381 #define PGSTE_HC_BIT	0x0020000000000000UL
382 #define PGSTE_GR_BIT	0x0004000000000000UL
383 #define PGSTE_GC_BIT	0x0002000000000000UL
384 #define PGSTE_UC_BIT	0x0000800000000000UL	/* user dirty (migration) */
385 #define PGSTE_IN_BIT	0x0000400000000000UL	/* IPTE notify bit */
386 #define PGSTE_VSIE_BIT	0x0000200000000000UL	/* ref'd in a shadow table */
387 
388 /* Guest Page State used for virtualization */
389 #define _PGSTE_GPS_ZERO			0x0000000080000000UL
390 #define _PGSTE_GPS_NODAT		0x0000000040000000UL
391 #define _PGSTE_GPS_USAGE_MASK		0x0000000003000000UL
392 #define _PGSTE_GPS_USAGE_STABLE		0x0000000000000000UL
393 #define _PGSTE_GPS_USAGE_UNUSED		0x0000000001000000UL
394 #define _PGSTE_GPS_USAGE_POT_VOLATILE	0x0000000002000000UL
395 #define _PGSTE_GPS_USAGE_VOLATILE	_PGSTE_GPS_USAGE_MASK
396 
397 /*
398  * A user page table pointer has the space-switch-event bit, the
399  * private-space-control bit and the storage-alteration-event-control
400  * bit set. A kernel page table pointer doesn't need them.
401  */
402 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
403 				 _ASCE_ALT_EVENT)
404 
405 /*
406  * Page protection definitions.
407  */
408 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID | _PAGE_PROTECT)
409 #define PAGE_RO		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
410 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
411 #define PAGE_RX		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
412 				 _PAGE_INVALID | _PAGE_PROTECT)
413 #define PAGE_RW		__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
414 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
415 #define PAGE_RWX	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
416 				 _PAGE_INVALID | _PAGE_PROTECT)
417 
418 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
419 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
420 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
421 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
422 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
423 				 _PAGE_PROTECT | _PAGE_NOEXEC)
424 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
425 				  _PAGE_YOUNG |	_PAGE_DIRTY)
426 
427 /*
428  * On s390 the page table entry has an invalid bit and a read-only bit.
429  * Read permission implies execute permission and write permission
430  * implies read permission.
431  */
432          /*xwr*/
433 
434 /*
435  * Segment entry (large page) protection definitions.
436  */
437 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_INVALID | \
438 				 _SEGMENT_ENTRY_PROTECT)
439 #define SEGMENT_RO	__pgprot(_SEGMENT_ENTRY_PROTECT | \
440 				 _SEGMENT_ENTRY_READ | \
441 				 _SEGMENT_ENTRY_NOEXEC)
442 #define SEGMENT_RX	__pgprot(_SEGMENT_ENTRY_PROTECT | \
443 				 _SEGMENT_ENTRY_READ)
444 #define SEGMENT_RW	__pgprot(_SEGMENT_ENTRY_READ | \
445 				 _SEGMENT_ENTRY_WRITE | \
446 				 _SEGMENT_ENTRY_NOEXEC)
447 #define SEGMENT_RWX	__pgprot(_SEGMENT_ENTRY_READ | \
448 				 _SEGMENT_ENTRY_WRITE)
449 #define SEGMENT_KERNEL	__pgprot(_SEGMENT_ENTRY |	\
450 				 _SEGMENT_ENTRY_LARGE |	\
451 				 _SEGMENT_ENTRY_READ |	\
452 				 _SEGMENT_ENTRY_WRITE | \
453 				 _SEGMENT_ENTRY_YOUNG | \
454 				 _SEGMENT_ENTRY_DIRTY | \
455 				 _SEGMENT_ENTRY_NOEXEC)
456 #define SEGMENT_KERNEL_RO __pgprot(_SEGMENT_ENTRY |	\
457 				 _SEGMENT_ENTRY_LARGE |	\
458 				 _SEGMENT_ENTRY_READ |	\
459 				 _SEGMENT_ENTRY_YOUNG |	\
460 				 _SEGMENT_ENTRY_PROTECT | \
461 				 _SEGMENT_ENTRY_NOEXEC)
462 #define SEGMENT_KERNEL_EXEC __pgprot(_SEGMENT_ENTRY |	\
463 				 _SEGMENT_ENTRY_LARGE |	\
464 				 _SEGMENT_ENTRY_READ |	\
465 				 _SEGMENT_ENTRY_WRITE | \
466 				 _SEGMENT_ENTRY_YOUNG |	\
467 				 _SEGMENT_ENTRY_DIRTY)
468 
469 /*
470  * Region3 entry (large page) protection definitions.
471  */
472 
473 #define REGION3_KERNEL	__pgprot(_REGION_ENTRY_TYPE_R3 | \
474 				 _REGION3_ENTRY_LARGE |	 \
475 				 _REGION3_ENTRY_READ |	 \
476 				 _REGION3_ENTRY_WRITE |	 \
477 				 _REGION3_ENTRY_YOUNG |	 \
478 				 _REGION3_ENTRY_DIRTY | \
479 				 _REGION_ENTRY_NOEXEC)
480 #define REGION3_KERNEL_RO __pgprot(_REGION_ENTRY_TYPE_R3 | \
481 				   _REGION3_ENTRY_LARGE |  \
482 				   _REGION3_ENTRY_READ |   \
483 				   _REGION3_ENTRY_YOUNG |  \
484 				   _REGION_ENTRY_PROTECT | \
485 				   _REGION_ENTRY_NOEXEC)
486 #define REGION3_KERNEL_EXEC __pgprot(_REGION_ENTRY_TYPE_R3 | \
487 				 _REGION3_ENTRY_LARGE |	 \
488 				 _REGION3_ENTRY_READ |	 \
489 				 _REGION3_ENTRY_WRITE |	 \
490 				 _REGION3_ENTRY_YOUNG |	 \
491 				 _REGION3_ENTRY_DIRTY)
492 
493 static inline bool mm_p4d_folded(struct mm_struct *mm)
494 {
495 	return mm->context.asce_limit <= _REGION1_SIZE;
496 }
497 #define mm_p4d_folded(mm) mm_p4d_folded(mm)
498 
499 static inline bool mm_pud_folded(struct mm_struct *mm)
500 {
501 	return mm->context.asce_limit <= _REGION2_SIZE;
502 }
503 #define mm_pud_folded(mm) mm_pud_folded(mm)
504 
505 static inline bool mm_pmd_folded(struct mm_struct *mm)
506 {
507 	return mm->context.asce_limit <= _REGION3_SIZE;
508 }
509 #define mm_pmd_folded(mm) mm_pmd_folded(mm)
510 
511 static inline int mm_has_pgste(struct mm_struct *mm)
512 {
513 #ifdef CONFIG_PGSTE
514 	if (unlikely(mm->context.has_pgste))
515 		return 1;
516 #endif
517 	return 0;
518 }
519 
520 static inline int mm_is_protected(struct mm_struct *mm)
521 {
522 #ifdef CONFIG_PGSTE
523 	if (unlikely(atomic_read(&mm->context.protected_count)))
524 		return 1;
525 #endif
526 	return 0;
527 }
528 
529 static inline int mm_alloc_pgste(struct mm_struct *mm)
530 {
531 #ifdef CONFIG_PGSTE
532 	if (unlikely(mm->context.alloc_pgste))
533 		return 1;
534 #endif
535 	return 0;
536 }
537 
538 static inline pte_t clear_pte_bit(pte_t pte, pgprot_t prot)
539 {
540 	return __pte(pte_val(pte) & ~pgprot_val(prot));
541 }
542 
543 static inline pte_t set_pte_bit(pte_t pte, pgprot_t prot)
544 {
545 	return __pte(pte_val(pte) | pgprot_val(prot));
546 }
547 
548 static inline pmd_t clear_pmd_bit(pmd_t pmd, pgprot_t prot)
549 {
550 	return __pmd(pmd_val(pmd) & ~pgprot_val(prot));
551 }
552 
553 static inline pmd_t set_pmd_bit(pmd_t pmd, pgprot_t prot)
554 {
555 	return __pmd(pmd_val(pmd) | pgprot_val(prot));
556 }
557 
558 static inline pud_t clear_pud_bit(pud_t pud, pgprot_t prot)
559 {
560 	return __pud(pud_val(pud) & ~pgprot_val(prot));
561 }
562 
563 static inline pud_t set_pud_bit(pud_t pud, pgprot_t prot)
564 {
565 	return __pud(pud_val(pud) | pgprot_val(prot));
566 }
567 
568 /*
569  * In the case that a guest uses storage keys
570  * faults should no longer be backed by zero pages
571  */
572 #define mm_forbids_zeropage mm_has_pgste
573 static inline int mm_uses_skeys(struct mm_struct *mm)
574 {
575 #ifdef CONFIG_PGSTE
576 	if (mm->context.uses_skeys)
577 		return 1;
578 #endif
579 	return 0;
580 }
581 
582 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
583 {
584 	union register_pair r1 = { .even = old, .odd = new, };
585 	unsigned long address = (unsigned long)ptr | 1;
586 
587 	asm volatile(
588 		"	csp	%[r1],%[address]"
589 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
590 		: [address] "d" (address)
591 		: "cc");
592 }
593 
594 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
595 {
596 	union register_pair r1 = { .even = old, .odd = new, };
597 	unsigned long address = (unsigned long)ptr | 1;
598 
599 	asm volatile(
600 		"	cspg	%[r1],%[address]"
601 		: [r1] "+&d" (r1.pair), "+m" (*ptr)
602 		: [address] "d" (address)
603 		: "cc");
604 }
605 
606 #define CRDTE_DTT_PAGE		0x00UL
607 #define CRDTE_DTT_SEGMENT	0x10UL
608 #define CRDTE_DTT_REGION3	0x14UL
609 #define CRDTE_DTT_REGION2	0x18UL
610 #define CRDTE_DTT_REGION1	0x1cUL
611 
612 static inline void crdte(unsigned long old, unsigned long new,
613 			 unsigned long *table, unsigned long dtt,
614 			 unsigned long address, unsigned long asce)
615 {
616 	union register_pair r1 = { .even = old, .odd = new, };
617 	union register_pair r2 = { .even = __pa(table) | dtt, .odd = address, };
618 
619 	asm volatile(".insn rrf,0xb98f0000,%[r1],%[r2],%[asce],0"
620 		     : [r1] "+&d" (r1.pair)
621 		     : [r2] "d" (r2.pair), [asce] "a" (asce)
622 		     : "memory", "cc");
623 }
624 
625 /*
626  * pgd/p4d/pud/pmd/pte query functions
627  */
628 static inline int pgd_folded(pgd_t pgd)
629 {
630 	return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
631 }
632 
633 static inline int pgd_present(pgd_t pgd)
634 {
635 	if (pgd_folded(pgd))
636 		return 1;
637 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
638 }
639 
640 static inline int pgd_none(pgd_t pgd)
641 {
642 	if (pgd_folded(pgd))
643 		return 0;
644 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
645 }
646 
647 static inline int pgd_bad(pgd_t pgd)
648 {
649 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1)
650 		return 0;
651 	return (pgd_val(pgd) & ~_REGION_ENTRY_BITS) != 0;
652 }
653 
654 static inline unsigned long pgd_pfn(pgd_t pgd)
655 {
656 	unsigned long origin_mask;
657 
658 	origin_mask = _REGION_ENTRY_ORIGIN;
659 	return (pgd_val(pgd) & origin_mask) >> PAGE_SHIFT;
660 }
661 
662 static inline int p4d_folded(p4d_t p4d)
663 {
664 	return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
665 }
666 
667 static inline int p4d_present(p4d_t p4d)
668 {
669 	if (p4d_folded(p4d))
670 		return 1;
671 	return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
672 }
673 
674 static inline int p4d_none(p4d_t p4d)
675 {
676 	if (p4d_folded(p4d))
677 		return 0;
678 	return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
679 }
680 
681 static inline unsigned long p4d_pfn(p4d_t p4d)
682 {
683 	unsigned long origin_mask;
684 
685 	origin_mask = _REGION_ENTRY_ORIGIN;
686 	return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
687 }
688 
689 static inline int pud_folded(pud_t pud)
690 {
691 	return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
692 }
693 
694 static inline int pud_present(pud_t pud)
695 {
696 	if (pud_folded(pud))
697 		return 1;
698 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
699 }
700 
701 static inline int pud_none(pud_t pud)
702 {
703 	if (pud_folded(pud))
704 		return 0;
705 	return pud_val(pud) == _REGION3_ENTRY_EMPTY;
706 }
707 
708 #define pud_leaf pud_leaf
709 static inline bool pud_leaf(pud_t pud)
710 {
711 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
712 		return 0;
713 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
714 }
715 
716 #define pmd_leaf pmd_leaf
717 static inline bool pmd_leaf(pmd_t pmd)
718 {
719 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
720 }
721 
722 static inline int pmd_bad(pmd_t pmd)
723 {
724 	if ((pmd_val(pmd) & _SEGMENT_ENTRY_TYPE_MASK) > 0 || pmd_leaf(pmd))
725 		return 1;
726 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
727 }
728 
729 static inline int pud_bad(pud_t pud)
730 {
731 	unsigned long type = pud_val(pud) & _REGION_ENTRY_TYPE_MASK;
732 
733 	if (type > _REGION_ENTRY_TYPE_R3 || pud_leaf(pud))
734 		return 1;
735 	if (type < _REGION_ENTRY_TYPE_R3)
736 		return 0;
737 	return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
738 }
739 
740 static inline int p4d_bad(p4d_t p4d)
741 {
742 	unsigned long type = p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK;
743 
744 	if (type > _REGION_ENTRY_TYPE_R2)
745 		return 1;
746 	if (type < _REGION_ENTRY_TYPE_R2)
747 		return 0;
748 	return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
749 }
750 
751 static inline int pmd_present(pmd_t pmd)
752 {
753 	return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
754 }
755 
756 static inline int pmd_none(pmd_t pmd)
757 {
758 	return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
759 }
760 
761 #define pmd_write pmd_write
762 static inline int pmd_write(pmd_t pmd)
763 {
764 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
765 }
766 
767 #define pud_write pud_write
768 static inline int pud_write(pud_t pud)
769 {
770 	return (pud_val(pud) & _REGION3_ENTRY_WRITE) != 0;
771 }
772 
773 #define pmd_dirty pmd_dirty
774 static inline int pmd_dirty(pmd_t pmd)
775 {
776 	return (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
777 }
778 
779 #define pmd_young pmd_young
780 static inline int pmd_young(pmd_t pmd)
781 {
782 	return (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
783 }
784 
785 static inline int pte_present(pte_t pte)
786 {
787 	/* Bit pattern: (pte & 0x001) == 0x001 */
788 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
789 }
790 
791 static inline int pte_none(pte_t pte)
792 {
793 	/* Bit pattern: pte == 0x400 */
794 	return pte_val(pte) == _PAGE_INVALID;
795 }
796 
797 static inline int pte_swap(pte_t pte)
798 {
799 	/* Bit pattern: (pte & 0x201) == 0x200 */
800 	return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
801 		== _PAGE_PROTECT;
802 }
803 
804 static inline int pte_special(pte_t pte)
805 {
806 	return (pte_val(pte) & _PAGE_SPECIAL);
807 }
808 
809 #define __HAVE_ARCH_PTE_SAME
810 static inline int pte_same(pte_t a, pte_t b)
811 {
812 	return pte_val(a) == pte_val(b);
813 }
814 
815 #ifdef CONFIG_NUMA_BALANCING
816 static inline int pte_protnone(pte_t pte)
817 {
818 	return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
819 }
820 
821 static inline int pmd_protnone(pmd_t pmd)
822 {
823 	/* pmd_leaf(pmd) implies pmd_present(pmd) */
824 	return pmd_leaf(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
825 }
826 #endif
827 
828 static inline int pte_swp_exclusive(pte_t pte)
829 {
830 	return pte_val(pte) & _PAGE_SWP_EXCLUSIVE;
831 }
832 
833 static inline pte_t pte_swp_mkexclusive(pte_t pte)
834 {
835 	return set_pte_bit(pte, __pgprot(_PAGE_SWP_EXCLUSIVE));
836 }
837 
838 static inline pte_t pte_swp_clear_exclusive(pte_t pte)
839 {
840 	return clear_pte_bit(pte, __pgprot(_PAGE_SWP_EXCLUSIVE));
841 }
842 
843 static inline int pte_soft_dirty(pte_t pte)
844 {
845 	return pte_val(pte) & _PAGE_SOFT_DIRTY;
846 }
847 #define pte_swp_soft_dirty pte_soft_dirty
848 
849 static inline pte_t pte_mksoft_dirty(pte_t pte)
850 {
851 	return set_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
852 }
853 #define pte_swp_mksoft_dirty pte_mksoft_dirty
854 
855 static inline pte_t pte_clear_soft_dirty(pte_t pte)
856 {
857 	return clear_pte_bit(pte, __pgprot(_PAGE_SOFT_DIRTY));
858 }
859 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
860 
861 static inline int pmd_soft_dirty(pmd_t pmd)
862 {
863 	return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
864 }
865 
866 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
867 {
868 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
869 }
870 
871 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
872 {
873 	return clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_SOFT_DIRTY));
874 }
875 
876 /*
877  * query functions pte_write/pte_dirty/pte_young only work if
878  * pte_present() is true. Undefined behaviour if not..
879  */
880 static inline int pte_write(pte_t pte)
881 {
882 	return (pte_val(pte) & _PAGE_WRITE) != 0;
883 }
884 
885 static inline int pte_dirty(pte_t pte)
886 {
887 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
888 }
889 
890 static inline int pte_young(pte_t pte)
891 {
892 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
893 }
894 
895 #define __HAVE_ARCH_PTE_UNUSED
896 static inline int pte_unused(pte_t pte)
897 {
898 	return pte_val(pte) & _PAGE_UNUSED;
899 }
900 
901 /*
902  * Extract the pgprot value from the given pte while at the same time making it
903  * usable for kernel address space mappings where fault driven dirty and
904  * young/old accounting is not supported, i.e _PAGE_PROTECT and _PAGE_INVALID
905  * must not be set.
906  */
907 static inline pgprot_t pte_pgprot(pte_t pte)
908 {
909 	unsigned long pte_flags = pte_val(pte) & _PAGE_CHG_MASK;
910 
911 	if (pte_write(pte))
912 		pte_flags |= pgprot_val(PAGE_KERNEL);
913 	else
914 		pte_flags |= pgprot_val(PAGE_KERNEL_RO);
915 	pte_flags |= pte_val(pte) & mio_wb_bit_mask;
916 
917 	return __pgprot(pte_flags);
918 }
919 
920 /*
921  * pgd/pmd/pte modification functions
922  */
923 
924 static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
925 {
926 	WRITE_ONCE(*pgdp, pgd);
927 }
928 
929 static inline void set_p4d(p4d_t *p4dp, p4d_t p4d)
930 {
931 	WRITE_ONCE(*p4dp, p4d);
932 }
933 
934 static inline void set_pud(pud_t *pudp, pud_t pud)
935 {
936 	WRITE_ONCE(*pudp, pud);
937 }
938 
939 static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
940 {
941 	WRITE_ONCE(*pmdp, pmd);
942 }
943 
944 static inline void set_pte(pte_t *ptep, pte_t pte)
945 {
946 	WRITE_ONCE(*ptep, pte);
947 }
948 
949 static inline void pgd_clear(pgd_t *pgd)
950 {
951 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
952 		set_pgd(pgd, __pgd(_REGION1_ENTRY_EMPTY));
953 }
954 
955 static inline void p4d_clear(p4d_t *p4d)
956 {
957 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
958 		set_p4d(p4d, __p4d(_REGION2_ENTRY_EMPTY));
959 }
960 
961 static inline void pud_clear(pud_t *pud)
962 {
963 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
964 		set_pud(pud, __pud(_REGION3_ENTRY_EMPTY));
965 }
966 
967 static inline void pmd_clear(pmd_t *pmdp)
968 {
969 	set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
970 }
971 
972 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
973 {
974 	set_pte(ptep, __pte(_PAGE_INVALID));
975 }
976 
977 /*
978  * The following pte modification functions only work if
979  * pte_present() is true. Undefined behaviour if not..
980  */
981 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
982 {
983 	pte = clear_pte_bit(pte, __pgprot(~_PAGE_CHG_MASK));
984 	pte = set_pte_bit(pte, newprot);
985 	/*
986 	 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
987 	 * has the invalid bit set, clear it again for readable, young pages
988 	 */
989 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
990 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
991 	/*
992 	 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
993 	 * protection bit set, clear it again for writable, dirty pages
994 	 */
995 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
996 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
997 	return pte;
998 }
999 
1000 static inline pte_t pte_wrprotect(pte_t pte)
1001 {
1002 	pte = clear_pte_bit(pte, __pgprot(_PAGE_WRITE));
1003 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1004 }
1005 
1006 static inline pte_t pte_mkwrite_novma(pte_t pte)
1007 {
1008 	pte = set_pte_bit(pte, __pgprot(_PAGE_WRITE));
1009 	if (pte_val(pte) & _PAGE_DIRTY)
1010 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1011 	return pte;
1012 }
1013 
1014 static inline pte_t pte_mkclean(pte_t pte)
1015 {
1016 	pte = clear_pte_bit(pte, __pgprot(_PAGE_DIRTY));
1017 	return set_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1018 }
1019 
1020 static inline pte_t pte_mkdirty(pte_t pte)
1021 {
1022 	pte = set_pte_bit(pte, __pgprot(_PAGE_DIRTY | _PAGE_SOFT_DIRTY));
1023 	if (pte_val(pte) & _PAGE_WRITE)
1024 		pte = clear_pte_bit(pte, __pgprot(_PAGE_PROTECT));
1025 	return pte;
1026 }
1027 
1028 static inline pte_t pte_mkold(pte_t pte)
1029 {
1030 	pte = clear_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1031 	return set_pte_bit(pte, __pgprot(_PAGE_INVALID));
1032 }
1033 
1034 static inline pte_t pte_mkyoung(pte_t pte)
1035 {
1036 	pte = set_pte_bit(pte, __pgprot(_PAGE_YOUNG));
1037 	if (pte_val(pte) & _PAGE_READ)
1038 		pte = clear_pte_bit(pte, __pgprot(_PAGE_INVALID));
1039 	return pte;
1040 }
1041 
1042 static inline pte_t pte_mkspecial(pte_t pte)
1043 {
1044 	return set_pte_bit(pte, __pgprot(_PAGE_SPECIAL));
1045 }
1046 
1047 #ifdef CONFIG_HUGETLB_PAGE
1048 static inline pte_t pte_mkhuge(pte_t pte)
1049 {
1050 	return set_pte_bit(pte, __pgprot(_PAGE_LARGE));
1051 }
1052 #endif
1053 
1054 #define IPTE_GLOBAL	0
1055 #define	IPTE_LOCAL	1
1056 
1057 #define IPTE_NODAT	0x400
1058 #define IPTE_GUEST_ASCE	0x800
1059 
1060 static __always_inline void __ptep_rdp(unsigned long addr, pte_t *ptep,
1061 				       unsigned long opt, unsigned long asce,
1062 				       int local)
1063 {
1064 	unsigned long pto;
1065 
1066 	pto = __pa(ptep) & ~(PTRS_PER_PTE * sizeof(pte_t) - 1);
1067 	asm volatile(".insn rrf,0xb98b0000,%[r1],%[r2],%[asce],%[m4]"
1068 		     : "+m" (*ptep)
1069 		     : [r1] "a" (pto), [r2] "a" ((addr & PAGE_MASK) | opt),
1070 		       [asce] "a" (asce), [m4] "i" (local));
1071 }
1072 
1073 static __always_inline void __ptep_ipte(unsigned long address, pte_t *ptep,
1074 					unsigned long opt, unsigned long asce,
1075 					int local)
1076 {
1077 	unsigned long pto = __pa(ptep);
1078 
1079 	if (__builtin_constant_p(opt) && opt == 0) {
1080 		/* Invalidation + TLB flush for the pte */
1081 		asm volatile(
1082 			"	ipte	%[r1],%[r2],0,%[m4]"
1083 			: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
1084 			  [m4] "i" (local));
1085 		return;
1086 	}
1087 
1088 	/* Invalidate ptes with options + TLB flush of the ptes */
1089 	opt = opt | (asce & _ASCE_ORIGIN);
1090 	asm volatile(
1091 		"	ipte	%[r1],%[r2],%[r3],%[m4]"
1092 		: [r2] "+a" (address), [r3] "+a" (opt)
1093 		: [r1] "a" (pto), [m4] "i" (local) : "memory");
1094 }
1095 
1096 static __always_inline void __ptep_ipte_range(unsigned long address, int nr,
1097 					      pte_t *ptep, int local)
1098 {
1099 	unsigned long pto = __pa(ptep);
1100 
1101 	/* Invalidate a range of ptes + TLB flush of the ptes */
1102 	do {
1103 		asm volatile(
1104 			"	ipte %[r1],%[r2],%[r3],%[m4]"
1105 			: [r2] "+a" (address), [r3] "+a" (nr)
1106 			: [r1] "a" (pto), [m4] "i" (local) : "memory");
1107 	} while (nr != 255);
1108 }
1109 
1110 /*
1111  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1112  * both clear the TLB for the unmapped pte. The reason is that
1113  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1114  * to modify an active pte. The sequence is
1115  *   1) ptep_get_and_clear
1116  *   2) set_pte_at
1117  *   3) flush_tlb_range
1118  * On s390 the tlb needs to get flushed with the modification of the pte
1119  * if the pte is active. The only way how this can be implemented is to
1120  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1121  * is a nop.
1122  */
1123 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1124 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1125 
1126 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1127 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1128 					    unsigned long addr, pte_t *ptep)
1129 {
1130 	pte_t pte = *ptep;
1131 
1132 	pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1133 	return pte_young(pte);
1134 }
1135 
1136 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1137 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1138 					 unsigned long address, pte_t *ptep)
1139 {
1140 	return ptep_test_and_clear_young(vma, address, ptep);
1141 }
1142 
1143 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1144 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1145 				       unsigned long addr, pte_t *ptep)
1146 {
1147 	pte_t res;
1148 
1149 	res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1150 	/* At this point the reference through the mapping is still present */
1151 	if (mm_is_protected(mm) && pte_present(res))
1152 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1153 	return res;
1154 }
1155 
1156 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1157 pte_t ptep_modify_prot_start(struct vm_area_struct *, unsigned long, pte_t *);
1158 void ptep_modify_prot_commit(struct vm_area_struct *, unsigned long,
1159 			     pte_t *, pte_t, pte_t);
1160 
1161 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1162 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1163 				     unsigned long addr, pte_t *ptep)
1164 {
1165 	pte_t res;
1166 
1167 	res = ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1168 	/* At this point the reference through the mapping is still present */
1169 	if (mm_is_protected(vma->vm_mm) && pte_present(res))
1170 		uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1171 	return res;
1172 }
1173 
1174 /*
1175  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1176  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1177  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1178  * cannot be accessed while the batched unmap is running. In this case
1179  * full==1 and a simple pte_clear is enough. See tlb.h.
1180  */
1181 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1182 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1183 					    unsigned long addr,
1184 					    pte_t *ptep, int full)
1185 {
1186 	pte_t res;
1187 
1188 	if (full) {
1189 		res = *ptep;
1190 		set_pte(ptep, __pte(_PAGE_INVALID));
1191 	} else {
1192 		res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1193 	}
1194 	/* Nothing to do */
1195 	if (!mm_is_protected(mm) || !pte_present(res))
1196 		return res;
1197 	/*
1198 	 * At this point the reference through the mapping is still present.
1199 	 * The notifier should have destroyed all protected vCPUs at this
1200 	 * point, so the destroy should be successful.
1201 	 */
1202 	if (full && !uv_destroy_owned_page(pte_val(res) & PAGE_MASK))
1203 		return res;
1204 	/*
1205 	 * If something went wrong and the page could not be destroyed, or
1206 	 * if this is not a mm teardown, the slower export is used as
1207 	 * fallback instead.
1208 	 */
1209 	uv_convert_owned_from_secure(pte_val(res) & PAGE_MASK);
1210 	return res;
1211 }
1212 
1213 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1214 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1215 				      unsigned long addr, pte_t *ptep)
1216 {
1217 	pte_t pte = *ptep;
1218 
1219 	if (pte_write(pte))
1220 		ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1221 }
1222 
1223 /*
1224  * Check if PTEs only differ in _PAGE_PROTECT HW bit, but also allow SW PTE
1225  * bits in the comparison. Those might change e.g. because of dirty and young
1226  * tracking.
1227  */
1228 static inline int pte_allow_rdp(pte_t old, pte_t new)
1229 {
1230 	/*
1231 	 * Only allow changes from RO to RW
1232 	 */
1233 	if (!(pte_val(old) & _PAGE_PROTECT) || pte_val(new) & _PAGE_PROTECT)
1234 		return 0;
1235 
1236 	return (pte_val(old) & _PAGE_RDP_MASK) == (pte_val(new) & _PAGE_RDP_MASK);
1237 }
1238 
1239 static inline void flush_tlb_fix_spurious_fault(struct vm_area_struct *vma,
1240 						unsigned long address,
1241 						pte_t *ptep)
1242 {
1243 	/*
1244 	 * RDP might not have propagated the PTE protection reset to all CPUs,
1245 	 * so there could be spurious TLB protection faults.
1246 	 * NOTE: This will also be called when a racing pagetable update on
1247 	 * another thread already installed the correct PTE. Both cases cannot
1248 	 * really be distinguished.
1249 	 * Therefore, only do the local TLB flush when RDP can be used, and the
1250 	 * PTE does not have _PAGE_PROTECT set, to avoid unnecessary overhead.
1251 	 * A local RDP can be used to do the flush.
1252 	 */
1253 	if (MACHINE_HAS_RDP && !(pte_val(*ptep) & _PAGE_PROTECT))
1254 		__ptep_rdp(address, ptep, 0, 0, 1);
1255 }
1256 #define flush_tlb_fix_spurious_fault flush_tlb_fix_spurious_fault
1257 
1258 void ptep_reset_dat_prot(struct mm_struct *mm, unsigned long addr, pte_t *ptep,
1259 			 pte_t new);
1260 
1261 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1262 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1263 					unsigned long addr, pte_t *ptep,
1264 					pte_t entry, int dirty)
1265 {
1266 	if (pte_same(*ptep, entry))
1267 		return 0;
1268 	if (MACHINE_HAS_RDP && !mm_has_pgste(vma->vm_mm) && pte_allow_rdp(*ptep, entry))
1269 		ptep_reset_dat_prot(vma->vm_mm, addr, ptep, entry);
1270 	else
1271 		ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1272 	return 1;
1273 }
1274 
1275 /*
1276  * Additional functions to handle KVM guest page tables
1277  */
1278 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1279 		     pte_t *ptep, pte_t entry);
1280 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1281 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1282 		 pte_t *ptep, unsigned long bits);
1283 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1284 		    pte_t *ptep, int prot, unsigned long bit);
1285 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1286 		     pte_t *ptep , int reset);
1287 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1288 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1289 		    pte_t *sptep, pte_t *tptep, pte_t pte);
1290 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1291 
1292 bool ptep_test_and_clear_uc(struct mm_struct *mm, unsigned long address,
1293 			    pte_t *ptep);
1294 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1295 			  unsigned char key, bool nq);
1296 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1297 			       unsigned char key, unsigned char *oldkey,
1298 			       bool nq, bool mr, bool mc);
1299 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1300 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1301 			  unsigned char *key);
1302 
1303 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1304 				unsigned long bits, unsigned long value);
1305 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1306 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1307 			unsigned long *oldpte, unsigned long *oldpgste);
1308 void gmap_pmdp_csp(struct mm_struct *mm, unsigned long vmaddr);
1309 void gmap_pmdp_invalidate(struct mm_struct *mm, unsigned long vmaddr);
1310 void gmap_pmdp_idte_local(struct mm_struct *mm, unsigned long vmaddr);
1311 void gmap_pmdp_idte_global(struct mm_struct *mm, unsigned long vmaddr);
1312 
1313 #define pgprot_writecombine	pgprot_writecombine
1314 pgprot_t pgprot_writecombine(pgprot_t prot);
1315 
1316 #define pgprot_writethrough	pgprot_writethrough
1317 pgprot_t pgprot_writethrough(pgprot_t prot);
1318 
1319 #define PFN_PTE_SHIFT		PAGE_SHIFT
1320 
1321 /*
1322  * Set multiple PTEs to consecutive pages with a single call.  All PTEs
1323  * are within the same folio, PMD and VMA.
1324  */
1325 static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
1326 			      pte_t *ptep, pte_t entry, unsigned int nr)
1327 {
1328 	if (pte_present(entry))
1329 		entry = clear_pte_bit(entry, __pgprot(_PAGE_UNUSED));
1330 	if (mm_has_pgste(mm)) {
1331 		for (;;) {
1332 			ptep_set_pte_at(mm, addr, ptep, entry);
1333 			if (--nr == 0)
1334 				break;
1335 			ptep++;
1336 			entry = __pte(pte_val(entry) + PAGE_SIZE);
1337 			addr += PAGE_SIZE;
1338 		}
1339 	} else {
1340 		for (;;) {
1341 			set_pte(ptep, entry);
1342 			if (--nr == 0)
1343 				break;
1344 			ptep++;
1345 			entry = __pte(pte_val(entry) + PAGE_SIZE);
1346 		}
1347 	}
1348 }
1349 #define set_ptes set_ptes
1350 
1351 /*
1352  * Conversion functions: convert a page and protection to a page entry,
1353  * and a page entry and page directory to the page they refer to.
1354  */
1355 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1356 {
1357 	pte_t __pte;
1358 
1359 	__pte = __pte(physpage | pgprot_val(pgprot));
1360 	if (!MACHINE_HAS_NX)
1361 		__pte = clear_pte_bit(__pte, __pgprot(_PAGE_NOEXEC));
1362 	return pte_mkyoung(__pte);
1363 }
1364 
1365 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1366 {
1367 	unsigned long physpage = page_to_phys(page);
1368 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1369 
1370 	if (pte_write(__pte) && PageDirty(page))
1371 		__pte = pte_mkdirty(__pte);
1372 	return __pte;
1373 }
1374 
1375 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1376 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1377 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1378 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1379 
1380 #define p4d_deref(pud) ((unsigned long)__va(p4d_val(pud) & _REGION_ENTRY_ORIGIN))
1381 #define pgd_deref(pgd) ((unsigned long)__va(pgd_val(pgd) & _REGION_ENTRY_ORIGIN))
1382 
1383 static inline unsigned long pmd_deref(pmd_t pmd)
1384 {
1385 	unsigned long origin_mask;
1386 
1387 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
1388 	if (pmd_leaf(pmd))
1389 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
1390 	return (unsigned long)__va(pmd_val(pmd) & origin_mask);
1391 }
1392 
1393 static inline unsigned long pmd_pfn(pmd_t pmd)
1394 {
1395 	return __pa(pmd_deref(pmd)) >> PAGE_SHIFT;
1396 }
1397 
1398 static inline unsigned long pud_deref(pud_t pud)
1399 {
1400 	unsigned long origin_mask;
1401 
1402 	origin_mask = _REGION_ENTRY_ORIGIN;
1403 	if (pud_leaf(pud))
1404 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
1405 	return (unsigned long)__va(pud_val(pud) & origin_mask);
1406 }
1407 
1408 static inline unsigned long pud_pfn(pud_t pud)
1409 {
1410 	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
1411 }
1412 
1413 /*
1414  * The pgd_offset function *always* adds the index for the top-level
1415  * region/segment table. This is done to get a sequence like the
1416  * following to work:
1417  *	pgdp = pgd_offset(current->mm, addr);
1418  *	pgd = READ_ONCE(*pgdp);
1419  *	p4dp = p4d_offset(&pgd, addr);
1420  *	...
1421  * The subsequent p4d_offset, pud_offset and pmd_offset functions
1422  * only add an index if they dereferenced the pointer.
1423  */
1424 static inline pgd_t *pgd_offset_raw(pgd_t *pgd, unsigned long address)
1425 {
1426 	unsigned long rste;
1427 	unsigned int shift;
1428 
1429 	/* Get the first entry of the top level table */
1430 	rste = pgd_val(*pgd);
1431 	/* Pick up the shift from the table type of the first entry */
1432 	shift = ((rste & _REGION_ENTRY_TYPE_MASK) >> 2) * 11 + 20;
1433 	return pgd + ((address >> shift) & (PTRS_PER_PGD - 1));
1434 }
1435 
1436 #define pgd_offset(mm, address) pgd_offset_raw(READ_ONCE((mm)->pgd), address)
1437 
1438 static inline p4d_t *p4d_offset_lockless(pgd_t *pgdp, pgd_t pgd, unsigned long address)
1439 {
1440 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R1)
1441 		return (p4d_t *) pgd_deref(pgd) + p4d_index(address);
1442 	return (p4d_t *) pgdp;
1443 }
1444 #define p4d_offset_lockless p4d_offset_lockless
1445 
1446 static inline p4d_t *p4d_offset(pgd_t *pgdp, unsigned long address)
1447 {
1448 	return p4d_offset_lockless(pgdp, *pgdp, address);
1449 }
1450 
1451 static inline pud_t *pud_offset_lockless(p4d_t *p4dp, p4d_t p4d, unsigned long address)
1452 {
1453 	if ((p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R2)
1454 		return (pud_t *) p4d_deref(p4d) + pud_index(address);
1455 	return (pud_t *) p4dp;
1456 }
1457 #define pud_offset_lockless pud_offset_lockless
1458 
1459 static inline pud_t *pud_offset(p4d_t *p4dp, unsigned long address)
1460 {
1461 	return pud_offset_lockless(p4dp, *p4dp, address);
1462 }
1463 #define pud_offset pud_offset
1464 
1465 static inline pmd_t *pmd_offset_lockless(pud_t *pudp, pud_t pud, unsigned long address)
1466 {
1467 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R3)
1468 		return (pmd_t *) pud_deref(pud) + pmd_index(address);
1469 	return (pmd_t *) pudp;
1470 }
1471 #define pmd_offset_lockless pmd_offset_lockless
1472 
1473 static inline pmd_t *pmd_offset(pud_t *pudp, unsigned long address)
1474 {
1475 	return pmd_offset_lockless(pudp, *pudp, address);
1476 }
1477 #define pmd_offset pmd_offset
1478 
1479 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
1480 {
1481 	return (unsigned long) pmd_deref(pmd);
1482 }
1483 
1484 static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
1485 {
1486 	return end <= current->mm->context.asce_limit;
1487 }
1488 #define gup_fast_permitted gup_fast_permitted
1489 
1490 #define pfn_pte(pfn, pgprot)	mk_pte_phys(((pfn) << PAGE_SHIFT), (pgprot))
1491 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1492 #define pte_page(x) pfn_to_page(pte_pfn(x))
1493 
1494 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1495 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1496 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
1497 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
1498 
1499 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1500 {
1501 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1502 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1503 }
1504 
1505 static inline pmd_t pmd_mkwrite_novma(pmd_t pmd)
1506 {
1507 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_WRITE));
1508 	if (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY)
1509 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1510 	return pmd;
1511 }
1512 
1513 static inline pmd_t pmd_mkclean(pmd_t pmd)
1514 {
1515 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY));
1516 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1517 }
1518 
1519 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1520 {
1521 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_SOFT_DIRTY));
1522 	if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1523 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1524 	return pmd;
1525 }
1526 
1527 static inline pud_t pud_wrprotect(pud_t pud)
1528 {
1529 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1530 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1531 }
1532 
1533 static inline pud_t pud_mkwrite(pud_t pud)
1534 {
1535 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_WRITE));
1536 	if (pud_val(pud) & _REGION3_ENTRY_DIRTY)
1537 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1538 	return pud;
1539 }
1540 
1541 static inline pud_t pud_mkclean(pud_t pud)
1542 {
1543 	pud = clear_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY));
1544 	return set_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1545 }
1546 
1547 static inline pud_t pud_mkdirty(pud_t pud)
1548 {
1549 	pud = set_pud_bit(pud, __pgprot(_REGION3_ENTRY_DIRTY | _REGION3_ENTRY_SOFT_DIRTY));
1550 	if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1551 		pud = clear_pud_bit(pud, __pgprot(_REGION_ENTRY_PROTECT));
1552 	return pud;
1553 }
1554 
1555 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1556 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1557 {
1558 	/*
1559 	 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1560 	 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1561 	 */
1562 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1563 		return pgprot_val(SEGMENT_NONE);
1564 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1565 		return pgprot_val(SEGMENT_RO);
1566 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1567 		return pgprot_val(SEGMENT_RX);
1568 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1569 		return pgprot_val(SEGMENT_RW);
1570 	return pgprot_val(SEGMENT_RWX);
1571 }
1572 
1573 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1574 {
1575 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1576 	if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1577 		pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1578 	return pmd;
1579 }
1580 
1581 static inline pmd_t pmd_mkold(pmd_t pmd)
1582 {
1583 	pmd = clear_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1584 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1585 }
1586 
1587 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1588 {
1589 	unsigned long mask;
1590 
1591 	mask  = _SEGMENT_ENTRY_ORIGIN_LARGE;
1592 	mask |= _SEGMENT_ENTRY_DIRTY;
1593 	mask |= _SEGMENT_ENTRY_YOUNG;
1594 	mask |=	_SEGMENT_ENTRY_LARGE;
1595 	mask |= _SEGMENT_ENTRY_SOFT_DIRTY;
1596 	pmd = __pmd(pmd_val(pmd) & mask);
1597 	pmd = set_pmd_bit(pmd, __pgprot(massage_pgprot_pmd(newprot)));
1598 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1599 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1600 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1601 		pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_INVALID));
1602 	return pmd;
1603 }
1604 
1605 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1606 {
1607 	return __pmd(physpage + massage_pgprot_pmd(pgprot));
1608 }
1609 
1610 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1611 
1612 static inline void __pmdp_csp(pmd_t *pmdp)
1613 {
1614 	csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1615 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1616 }
1617 
1618 #define IDTE_GLOBAL	0
1619 #define IDTE_LOCAL	1
1620 
1621 #define IDTE_PTOA	0x0800
1622 #define IDTE_NODAT	0x1000
1623 #define IDTE_GUEST_ASCE	0x2000
1624 
1625 static __always_inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1626 					unsigned long opt, unsigned long asce,
1627 					int local)
1628 {
1629 	unsigned long sto;
1630 
1631 	sto = __pa(pmdp) - pmd_index(addr) * sizeof(pmd_t);
1632 	if (__builtin_constant_p(opt) && opt == 0) {
1633 		/* flush without guest asce */
1634 		asm volatile(
1635 			"	idte	%[r1],0,%[r2],%[m4]"
1636 			: "+m" (*pmdp)
1637 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1638 			  [m4] "i" (local)
1639 			: "cc" );
1640 	} else {
1641 		/* flush with guest asce */
1642 		asm volatile(
1643 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1644 			: "+m" (*pmdp)
1645 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1646 			  [r3] "a" (asce), [m4] "i" (local)
1647 			: "cc" );
1648 	}
1649 }
1650 
1651 static __always_inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1652 					unsigned long opt, unsigned long asce,
1653 					int local)
1654 {
1655 	unsigned long r3o;
1656 
1657 	r3o = __pa(pudp) - pud_index(addr) * sizeof(pud_t);
1658 	r3o |= _ASCE_TYPE_REGION3;
1659 	if (__builtin_constant_p(opt) && opt == 0) {
1660 		/* flush without guest asce */
1661 		asm volatile(
1662 			"	idte	%[r1],0,%[r2],%[m4]"
1663 			: "+m" (*pudp)
1664 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1665 			  [m4] "i" (local)
1666 			: "cc");
1667 	} else {
1668 		/* flush with guest asce */
1669 		asm volatile(
1670 			"	idte	%[r1],%[r3],%[r2],%[m4]"
1671 			: "+m" (*pudp)
1672 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1673 			  [r3] "a" (asce), [m4] "i" (local)
1674 			: "cc" );
1675 	}
1676 }
1677 
1678 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1679 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1680 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1681 
1682 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1683 
1684 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1685 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1686 				pgtable_t pgtable);
1687 
1688 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1689 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1690 
1691 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1692 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1693 					unsigned long addr, pmd_t *pmdp,
1694 					pmd_t entry, int dirty)
1695 {
1696 	VM_BUG_ON(addr & ~HPAGE_MASK);
1697 
1698 	entry = pmd_mkyoung(entry);
1699 	if (dirty)
1700 		entry = pmd_mkdirty(entry);
1701 	if (pmd_val(*pmdp) == pmd_val(entry))
1702 		return 0;
1703 	pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1704 	return 1;
1705 }
1706 
1707 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1708 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1709 					    unsigned long addr, pmd_t *pmdp)
1710 {
1711 	pmd_t pmd = *pmdp;
1712 
1713 	pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1714 	return pmd_young(pmd);
1715 }
1716 
1717 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1718 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1719 					 unsigned long addr, pmd_t *pmdp)
1720 {
1721 	VM_BUG_ON(addr & ~HPAGE_MASK);
1722 	return pmdp_test_and_clear_young(vma, addr, pmdp);
1723 }
1724 
1725 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1726 			      pmd_t *pmdp, pmd_t entry)
1727 {
1728 	if (!MACHINE_HAS_NX)
1729 		entry = clear_pmd_bit(entry, __pgprot(_SEGMENT_ENTRY_NOEXEC));
1730 	set_pmd(pmdp, entry);
1731 }
1732 
1733 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1734 {
1735 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_LARGE));
1736 	pmd = set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_YOUNG));
1737 	return set_pmd_bit(pmd, __pgprot(_SEGMENT_ENTRY_PROTECT));
1738 }
1739 
1740 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1741 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1742 					    unsigned long addr, pmd_t *pmdp)
1743 {
1744 	return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1745 }
1746 
1747 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1748 static inline pmd_t pmdp_huge_get_and_clear_full(struct vm_area_struct *vma,
1749 						 unsigned long addr,
1750 						 pmd_t *pmdp, int full)
1751 {
1752 	if (full) {
1753 		pmd_t pmd = *pmdp;
1754 		set_pmd(pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1755 		return pmd;
1756 	}
1757 	return pmdp_xchg_lazy(vma->vm_mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1758 }
1759 
1760 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1761 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1762 					  unsigned long addr, pmd_t *pmdp)
1763 {
1764 	return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1765 }
1766 
1767 #define __HAVE_ARCH_PMDP_INVALIDATE
1768 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1769 				   unsigned long addr, pmd_t *pmdp)
1770 {
1771 	pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1772 
1773 	return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1774 }
1775 
1776 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1777 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1778 				      unsigned long addr, pmd_t *pmdp)
1779 {
1780 	pmd_t pmd = *pmdp;
1781 
1782 	if (pmd_write(pmd))
1783 		pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1784 }
1785 
1786 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1787 					unsigned long address,
1788 					pmd_t *pmdp)
1789 {
1790 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1791 }
1792 #define pmdp_collapse_flush pmdp_collapse_flush
1793 
1794 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(((pfn) << PAGE_SHIFT), (pgprot))
1795 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1796 
1797 static inline int pmd_trans_huge(pmd_t pmd)
1798 {
1799 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1800 }
1801 
1802 #define has_transparent_hugepage has_transparent_hugepage
1803 static inline int has_transparent_hugepage(void)
1804 {
1805 	return MACHINE_HAS_EDAT1 ? 1 : 0;
1806 }
1807 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1808 
1809 /*
1810  * 64 bit swap entry format:
1811  * A page-table entry has some bits we have to treat in a special way.
1812  * Bits 54 and 63 are used to indicate the page type. Bit 53 marks the pte
1813  * as invalid.
1814  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1815  * |			  offset			|E11XX|type |S0|
1816  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1817  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1818  *
1819  * Bits 0-51 store the offset.
1820  * Bit 52 (E) is used to remember PG_anon_exclusive.
1821  * Bits 57-61 store the type.
1822  * Bit 62 (S) is used for softdirty tracking.
1823  * Bits 55 and 56 (X) are unused.
1824  */
1825 
1826 #define __SWP_OFFSET_MASK	((1UL << 52) - 1)
1827 #define __SWP_OFFSET_SHIFT	12
1828 #define __SWP_TYPE_MASK		((1UL << 5) - 1)
1829 #define __SWP_TYPE_SHIFT	2
1830 
1831 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1832 {
1833 	unsigned long pteval;
1834 
1835 	pteval = _PAGE_INVALID | _PAGE_PROTECT;
1836 	pteval |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1837 	pteval |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1838 	return __pte(pteval);
1839 }
1840 
1841 static inline unsigned long __swp_type(swp_entry_t entry)
1842 {
1843 	return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1844 }
1845 
1846 static inline unsigned long __swp_offset(swp_entry_t entry)
1847 {
1848 	return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1849 }
1850 
1851 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1852 {
1853 	return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1854 }
1855 
1856 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1857 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1858 
1859 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1860 extern void vmem_remove_mapping(unsigned long start, unsigned long size);
1861 extern int __vmem_map_4k_page(unsigned long addr, unsigned long phys, pgprot_t prot, bool alloc);
1862 extern int vmem_map_4k_page(unsigned long addr, unsigned long phys, pgprot_t prot);
1863 extern void vmem_unmap_4k_page(unsigned long addr);
1864 extern pte_t *vmem_get_alloc_pte(unsigned long addr, bool alloc);
1865 extern int s390_enable_sie(void);
1866 extern int s390_enable_skey(void);
1867 extern void s390_reset_cmma(struct mm_struct *mm);
1868 
1869 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1870 #define HAVE_ARCH_UNMAPPED_AREA
1871 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1872 
1873 #define pmd_pgtable(pmd) \
1874 	((pgtable_t)__va(pmd_val(pmd) & -sizeof(pte_t)*PTRS_PER_PTE))
1875 
1876 #endif /* _S390_PAGE_H */
1877