xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision b8d312aa)
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/bug.h>
21 #include <asm/page.h>
22 
23 extern pgd_t swapper_pg_dir[];
24 extern void paging_init(void);
25 
26 enum {
27 	PG_DIRECT_MAP_4K = 0,
28 	PG_DIRECT_MAP_1M,
29 	PG_DIRECT_MAP_2G,
30 	PG_DIRECT_MAP_MAX
31 };
32 
33 extern atomic_long_t direct_pages_count[PG_DIRECT_MAP_MAX];
34 
35 static inline void update_page_count(int level, long count)
36 {
37 	if (IS_ENABLED(CONFIG_PROC_FS))
38 		atomic_long_add(count, &direct_pages_count[level]);
39 }
40 
41 struct seq_file;
42 void arch_report_meminfo(struct seq_file *m);
43 
44 /*
45  * The S390 doesn't have any external MMU info: the kernel page
46  * tables contain all the necessary information.
47  */
48 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
49 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
50 
51 /*
52  * ZERO_PAGE is a global shared page that is always zero; used
53  * for zero-mapped memory areas etc..
54  */
55 
56 extern unsigned long empty_zero_page;
57 extern unsigned long zero_page_mask;
58 
59 #define ZERO_PAGE(vaddr) \
60 	(virt_to_page((void *)(empty_zero_page + \
61 	 (((unsigned long)(vaddr)) &zero_page_mask))))
62 #define __HAVE_COLOR_ZERO_PAGE
63 
64 /* TODO: s390 cannot support io_remap_pfn_range... */
65 
66 #define FIRST_USER_ADDRESS  0UL
67 
68 #define pte_ERROR(e) \
69 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
70 #define pmd_ERROR(e) \
71 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
72 #define pud_ERROR(e) \
73 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
74 #define p4d_ERROR(e) \
75 	printk("%s:%d: bad p4d %p.\n", __FILE__, __LINE__, (void *) p4d_val(e))
76 #define pgd_ERROR(e) \
77 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
78 
79 /*
80  * The vmalloc and module area will always be on the topmost area of the
81  * kernel mapping. We reserve 128GB (64bit) for vmalloc and modules.
82  * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
83  * modules will reside. That makes sure that inter module branches always
84  * happen without trampolines and in addition the placement within a 2GB frame
85  * is branch prediction unit friendly.
86  */
87 extern unsigned long VMALLOC_START;
88 extern unsigned long VMALLOC_END;
89 extern struct page *vmemmap;
90 
91 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
92 
93 extern unsigned long MODULES_VADDR;
94 extern unsigned long MODULES_END;
95 #define MODULES_VADDR	MODULES_VADDR
96 #define MODULES_END	MODULES_END
97 #define MODULES_LEN	(1UL << 31)
98 
99 static inline int is_module_addr(void *addr)
100 {
101 	BUILD_BUG_ON(MODULES_LEN > (1UL << 31));
102 	if (addr < (void *)MODULES_VADDR)
103 		return 0;
104 	if (addr > (void *)MODULES_END)
105 		return 0;
106 	return 1;
107 }
108 
109 /*
110  * A 64 bit pagetable entry of S390 has following format:
111  * |			 PFRA			      |0IPC|  OS  |
112  * 0000000000111111111122222222223333333333444444444455555555556666
113  * 0123456789012345678901234567890123456789012345678901234567890123
114  *
115  * I Page-Invalid Bit:    Page is not available for address-translation
116  * P Page-Protection Bit: Store access not possible for page
117  * C Change-bit override: HW is not required to set change bit
118  *
119  * A 64 bit segmenttable entry of S390 has following format:
120  * |        P-table origin                              |      TT
121  * 0000000000111111111122222222223333333333444444444455555555556666
122  * 0123456789012345678901234567890123456789012345678901234567890123
123  *
124  * I Segment-Invalid Bit:    Segment is not available for address-translation
125  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
126  * P Page-Protection Bit: Store access not possible for page
127  * TT Type 00
128  *
129  * A 64 bit region table entry of S390 has following format:
130  * |        S-table origin                             |   TF  TTTL
131  * 0000000000111111111122222222223333333333444444444455555555556666
132  * 0123456789012345678901234567890123456789012345678901234567890123
133  *
134  * I Segment-Invalid Bit:    Segment is not available for address-translation
135  * TT Type 01
136  * TF
137  * TL Table length
138  *
139  * The 64 bit regiontable origin of S390 has following format:
140  * |      region table origon                          |       DTTL
141  * 0000000000111111111122222222223333333333444444444455555555556666
142  * 0123456789012345678901234567890123456789012345678901234567890123
143  *
144  * X Space-Switch event:
145  * G Segment-Invalid Bit:
146  * P Private-Space Bit:
147  * S Storage-Alteration:
148  * R Real space
149  * TL Table-Length:
150  *
151  * A storage key has the following format:
152  * | ACC |F|R|C|0|
153  *  0   3 4 5 6 7
154  * ACC: access key
155  * F  : fetch protection bit
156  * R  : referenced bit
157  * C  : changed bit
158  */
159 
160 /* Hardware bits in the page table entry */
161 #define _PAGE_NOEXEC	0x100		/* HW no-execute bit  */
162 #define _PAGE_PROTECT	0x200		/* HW read-only bit  */
163 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
164 #define _PAGE_LARGE	0x800		/* Bit to mark a large pte */
165 
166 /* Software bits in the page table entry */
167 #define _PAGE_PRESENT	0x001		/* SW pte present bit */
168 #define _PAGE_YOUNG	0x004		/* SW pte young bit */
169 #define _PAGE_DIRTY	0x008		/* SW pte dirty bit */
170 #define _PAGE_READ	0x010		/* SW pte read bit */
171 #define _PAGE_WRITE	0x020		/* SW pte write bit */
172 #define _PAGE_SPECIAL	0x040		/* SW associated with special page */
173 #define _PAGE_UNUSED	0x080		/* SW bit for pgste usage state */
174 
175 #ifdef CONFIG_MEM_SOFT_DIRTY
176 #define _PAGE_SOFT_DIRTY 0x002		/* SW pte soft dirty bit */
177 #else
178 #define _PAGE_SOFT_DIRTY 0x000
179 #endif
180 
181 /* Set of bits not changed in pte_modify */
182 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
183 				 _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
184 
185 /*
186  * handle_pte_fault uses pte_present and pte_none to find out the pte type
187  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
188  * distinguish present from not-present ptes. It is changed only with the page
189  * table lock held.
190  *
191  * The following table gives the different possible bit combinations for
192  * the pte hardware and software bits in the last 12 bits of a pte
193  * (. unassigned bit, x don't care, t swap type):
194  *
195  *				842100000000
196  *				000084210000
197  *				000000008421
198  *				.IR.uswrdy.p
199  * empty			.10.00000000
200  * swap				.11..ttttt.0
201  * prot-none, clean, old	.11.xx0000.1
202  * prot-none, clean, young	.11.xx0001.1
203  * prot-none, dirty, old	.11.xx0010.1
204  * prot-none, dirty, young	.11.xx0011.1
205  * read-only, clean, old	.11.xx0100.1
206  * read-only, clean, young	.01.xx0101.1
207  * read-only, dirty, old	.11.xx0110.1
208  * read-only, dirty, young	.01.xx0111.1
209  * read-write, clean, old	.11.xx1100.1
210  * read-write, clean, young	.01.xx1101.1
211  * read-write, dirty, old	.10.xx1110.1
212  * read-write, dirty, young	.00.xx1111.1
213  * HW-bits: R read-only, I invalid
214  * SW-bits: p present, y young, d dirty, r read, w write, s special,
215  *	    u unused, l large
216  *
217  * pte_none    is true for the bit pattern .10.00000000, pte == 0x400
218  * pte_swap    is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
219  * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
220  */
221 
222 /* Bits in the segment/region table address-space-control-element */
223 #define _ASCE_ORIGIN		~0xfffUL/* region/segment table origin	    */
224 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
225 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
226 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
227 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
228 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
229 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
230 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
231 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
232 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
233 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
234 
235 /* Bits in the region table entry */
236 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
237 #define _REGION_ENTRY_PROTECT	0x200	/* region protection bit	    */
238 #define _REGION_ENTRY_NOEXEC	0x100	/* region no-execute bit	    */
239 #define _REGION_ENTRY_OFFSET	0xc0	/* region table offset		    */
240 #define _REGION_ENTRY_INVALID	0x20	/* invalid region table entry	    */
241 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region table type mask	    */
242 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
243 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
244 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
245 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
246 
247 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
248 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
249 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
250 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
251 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
252 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
253 
254 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address	     */
255 #define _REGION3_ENTRY_DIRTY	0x2000	/* SW region dirty bit */
256 #define _REGION3_ENTRY_YOUNG	0x1000	/* SW region young bit */
257 #define _REGION3_ENTRY_LARGE	0x0400	/* RTTE-format control, large page  */
258 #define _REGION3_ENTRY_READ	0x0002	/* SW region read bit */
259 #define _REGION3_ENTRY_WRITE	0x0001	/* SW region write bit */
260 
261 #ifdef CONFIG_MEM_SOFT_DIRTY
262 #define _REGION3_ENTRY_SOFT_DIRTY 0x4000 /* SW region soft dirty bit */
263 #else
264 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
265 #endif
266 
267 #define _REGION_ENTRY_BITS	 0xfffffffffffff22fUL
268 #define _REGION_ENTRY_BITS_LARGE 0xffffffff8000fe2fUL
269 
270 /* Bits in the segment table entry */
271 #define _SEGMENT_ENTRY_BITS			0xfffffffffffffe33UL
272 #define _SEGMENT_ENTRY_BITS_LARGE		0xfffffffffff0ff33UL
273 #define _SEGMENT_ENTRY_HARDWARE_BITS		0xfffffffffffffe30UL
274 #define _SEGMENT_ENTRY_HARDWARE_BITS_LARGE	0xfffffffffff00730UL
275 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
276 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* page table origin		    */
277 #define _SEGMENT_ENTRY_PROTECT	0x200	/* segment protection bit	    */
278 #define _SEGMENT_ENTRY_NOEXEC	0x100	/* segment no-execute bit	    */
279 #define _SEGMENT_ENTRY_INVALID	0x20	/* invalid segment table entry	    */
280 #define _SEGMENT_ENTRY_TYPE_MASK 0x0c	/* segment table type mask	    */
281 
282 #define _SEGMENT_ENTRY		(0)
283 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INVALID)
284 
285 #define _SEGMENT_ENTRY_DIRTY	0x2000	/* SW segment dirty bit */
286 #define _SEGMENT_ENTRY_YOUNG	0x1000	/* SW segment young bit */
287 #define _SEGMENT_ENTRY_LARGE	0x0400	/* STE-format control, large page */
288 #define _SEGMENT_ENTRY_WRITE	0x0002	/* SW segment write bit */
289 #define _SEGMENT_ENTRY_READ	0x0001	/* SW segment read bit */
290 
291 #ifdef CONFIG_MEM_SOFT_DIRTY
292 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x4000 /* SW segment soft dirty bit */
293 #else
294 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
295 #endif
296 
297 #define _CRST_ENTRIES	2048	/* number of region/segment table entries */
298 #define _PAGE_ENTRIES	256	/* number of page table entries	*/
299 
300 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
301 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
302 
303 #define _REGION1_SHIFT	53
304 #define _REGION2_SHIFT	42
305 #define _REGION3_SHIFT	31
306 #define _SEGMENT_SHIFT	20
307 
308 #define _REGION1_INDEX	(0x7ffUL << _REGION1_SHIFT)
309 #define _REGION2_INDEX	(0x7ffUL << _REGION2_SHIFT)
310 #define _REGION3_INDEX	(0x7ffUL << _REGION3_SHIFT)
311 #define _SEGMENT_INDEX	(0x7ffUL << _SEGMENT_SHIFT)
312 #define _PAGE_INDEX	(0xffUL  << _PAGE_SHIFT)
313 
314 #define _REGION1_SIZE	(1UL << _REGION1_SHIFT)
315 #define _REGION2_SIZE	(1UL << _REGION2_SHIFT)
316 #define _REGION3_SIZE	(1UL << _REGION3_SHIFT)
317 #define _SEGMENT_SIZE	(1UL << _SEGMENT_SHIFT)
318 
319 #define _REGION1_MASK	(~(_REGION1_SIZE - 1))
320 #define _REGION2_MASK	(~(_REGION2_SIZE - 1))
321 #define _REGION3_MASK	(~(_REGION3_SIZE - 1))
322 #define _SEGMENT_MASK	(~(_SEGMENT_SIZE - 1))
323 
324 #define PMD_SHIFT	_SEGMENT_SHIFT
325 #define PUD_SHIFT	_REGION3_SHIFT
326 #define P4D_SHIFT	_REGION2_SHIFT
327 #define PGDIR_SHIFT	_REGION1_SHIFT
328 
329 #define PMD_SIZE	_SEGMENT_SIZE
330 #define PUD_SIZE	_REGION3_SIZE
331 #define P4D_SIZE	_REGION2_SIZE
332 #define PGDIR_SIZE	_REGION1_SIZE
333 
334 #define PMD_MASK	_SEGMENT_MASK
335 #define PUD_MASK	_REGION3_MASK
336 #define P4D_MASK	_REGION2_MASK
337 #define PGDIR_MASK	_REGION1_MASK
338 
339 #define PTRS_PER_PTE	_PAGE_ENTRIES
340 #define PTRS_PER_PMD	_CRST_ENTRIES
341 #define PTRS_PER_PUD	_CRST_ENTRIES
342 #define PTRS_PER_P4D	_CRST_ENTRIES
343 #define PTRS_PER_PGD	_CRST_ENTRIES
344 
345 #define MAX_PTRS_PER_P4D	PTRS_PER_P4D
346 
347 /*
348  * Segment table and region3 table entry encoding
349  * (R = read-only, I = invalid, y = young bit):
350  *				dy..R...I...wr
351  * prot-none, clean, old	00..1...1...00
352  * prot-none, clean, young	01..1...1...00
353  * prot-none, dirty, old	10..1...1...00
354  * prot-none, dirty, young	11..1...1...00
355  * read-only, clean, old	00..1...1...01
356  * read-only, clean, young	01..1...0...01
357  * read-only, dirty, old	10..1...1...01
358  * read-only, dirty, young	11..1...0...01
359  * read-write, clean, old	00..1...1...11
360  * read-write, clean, young	01..1...0...11
361  * read-write, dirty, old	10..0...1...11
362  * read-write, dirty, young	11..0...0...11
363  * The segment table origin is used to distinguish empty (origin==0) from
364  * read-write, old segment table entries (origin!=0)
365  * HW-bits: R read-only, I invalid
366  * SW-bits: y young, d dirty, r read, w write
367  */
368 
369 /* Page status table bits for virtualization */
370 #define PGSTE_ACC_BITS	0xf000000000000000UL
371 #define PGSTE_FP_BIT	0x0800000000000000UL
372 #define PGSTE_PCL_BIT	0x0080000000000000UL
373 #define PGSTE_HR_BIT	0x0040000000000000UL
374 #define PGSTE_HC_BIT	0x0020000000000000UL
375 #define PGSTE_GR_BIT	0x0004000000000000UL
376 #define PGSTE_GC_BIT	0x0002000000000000UL
377 #define PGSTE_UC_BIT	0x0000800000000000UL	/* user dirty (migration) */
378 #define PGSTE_IN_BIT	0x0000400000000000UL	/* IPTE notify bit */
379 #define PGSTE_VSIE_BIT	0x0000200000000000UL	/* ref'd in a shadow table */
380 
381 /* Guest Page State used for virtualization */
382 #define _PGSTE_GPS_ZERO			0x0000000080000000UL
383 #define _PGSTE_GPS_NODAT		0x0000000040000000UL
384 #define _PGSTE_GPS_USAGE_MASK		0x0000000003000000UL
385 #define _PGSTE_GPS_USAGE_STABLE		0x0000000000000000UL
386 #define _PGSTE_GPS_USAGE_UNUSED		0x0000000001000000UL
387 #define _PGSTE_GPS_USAGE_POT_VOLATILE	0x0000000002000000UL
388 #define _PGSTE_GPS_USAGE_VOLATILE	_PGSTE_GPS_USAGE_MASK
389 
390 /*
391  * A user page table pointer has the space-switch-event bit, the
392  * private-space-control bit and the storage-alteration-event-control
393  * bit set. A kernel page table pointer doesn't need them.
394  */
395 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
396 				 _ASCE_ALT_EVENT)
397 
398 /*
399  * Page protection definitions.
400  */
401 #define PAGE_NONE	__pgprot(_PAGE_PRESENT | _PAGE_INVALID | _PAGE_PROTECT)
402 #define PAGE_RO		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
403 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
404 #define PAGE_RX		__pgprot(_PAGE_PRESENT | _PAGE_READ | \
405 				 _PAGE_INVALID | _PAGE_PROTECT)
406 #define PAGE_RW		__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
407 				 _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
408 #define PAGE_RWX	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
409 				 _PAGE_INVALID | _PAGE_PROTECT)
410 
411 #define PAGE_SHARED	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
412 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
413 #define PAGE_KERNEL	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
414 				 _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
415 #define PAGE_KERNEL_RO	__pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
416 				 _PAGE_PROTECT | _PAGE_NOEXEC)
417 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
418 				  _PAGE_YOUNG |	_PAGE_DIRTY)
419 
420 /*
421  * On s390 the page table entry has an invalid bit and a read-only bit.
422  * Read permission implies execute permission and write permission
423  * implies read permission.
424  */
425          /*xwr*/
426 #define __P000	PAGE_NONE
427 #define __P001	PAGE_RO
428 #define __P010	PAGE_RO
429 #define __P011	PAGE_RO
430 #define __P100	PAGE_RX
431 #define __P101	PAGE_RX
432 #define __P110	PAGE_RX
433 #define __P111	PAGE_RX
434 
435 #define __S000	PAGE_NONE
436 #define __S001	PAGE_RO
437 #define __S010	PAGE_RW
438 #define __S011	PAGE_RW
439 #define __S100	PAGE_RX
440 #define __S101	PAGE_RX
441 #define __S110	PAGE_RWX
442 #define __S111	PAGE_RWX
443 
444 /*
445  * Segment entry (large page) protection definitions.
446  */
447 #define SEGMENT_NONE	__pgprot(_SEGMENT_ENTRY_INVALID | \
448 				 _SEGMENT_ENTRY_PROTECT)
449 #define SEGMENT_RO	__pgprot(_SEGMENT_ENTRY_PROTECT | \
450 				 _SEGMENT_ENTRY_READ | \
451 				 _SEGMENT_ENTRY_NOEXEC)
452 #define SEGMENT_RX	__pgprot(_SEGMENT_ENTRY_PROTECT | \
453 				 _SEGMENT_ENTRY_READ)
454 #define SEGMENT_RW	__pgprot(_SEGMENT_ENTRY_READ | \
455 				 _SEGMENT_ENTRY_WRITE | \
456 				 _SEGMENT_ENTRY_NOEXEC)
457 #define SEGMENT_RWX	__pgprot(_SEGMENT_ENTRY_READ | \
458 				 _SEGMENT_ENTRY_WRITE)
459 #define SEGMENT_KERNEL	__pgprot(_SEGMENT_ENTRY |	\
460 				 _SEGMENT_ENTRY_LARGE |	\
461 				 _SEGMENT_ENTRY_READ |	\
462 				 _SEGMENT_ENTRY_WRITE | \
463 				 _SEGMENT_ENTRY_YOUNG | \
464 				 _SEGMENT_ENTRY_DIRTY | \
465 				 _SEGMENT_ENTRY_NOEXEC)
466 #define SEGMENT_KERNEL_RO __pgprot(_SEGMENT_ENTRY |	\
467 				 _SEGMENT_ENTRY_LARGE |	\
468 				 _SEGMENT_ENTRY_READ |	\
469 				 _SEGMENT_ENTRY_YOUNG |	\
470 				 _SEGMENT_ENTRY_PROTECT | \
471 				 _SEGMENT_ENTRY_NOEXEC)
472 #define SEGMENT_KERNEL_EXEC __pgprot(_SEGMENT_ENTRY |	\
473 				 _SEGMENT_ENTRY_LARGE |	\
474 				 _SEGMENT_ENTRY_READ |	\
475 				 _SEGMENT_ENTRY_WRITE | \
476 				 _SEGMENT_ENTRY_YOUNG |	\
477 				 _SEGMENT_ENTRY_DIRTY)
478 
479 /*
480  * Region3 entry (large page) protection definitions.
481  */
482 
483 #define REGION3_KERNEL	__pgprot(_REGION_ENTRY_TYPE_R3 | \
484 				 _REGION3_ENTRY_LARGE |	 \
485 				 _REGION3_ENTRY_READ |	 \
486 				 _REGION3_ENTRY_WRITE |	 \
487 				 _REGION3_ENTRY_YOUNG |	 \
488 				 _REGION3_ENTRY_DIRTY | \
489 				 _REGION_ENTRY_NOEXEC)
490 #define REGION3_KERNEL_RO __pgprot(_REGION_ENTRY_TYPE_R3 | \
491 				   _REGION3_ENTRY_LARGE |  \
492 				   _REGION3_ENTRY_READ |   \
493 				   _REGION3_ENTRY_YOUNG |  \
494 				   _REGION_ENTRY_PROTECT | \
495 				   _REGION_ENTRY_NOEXEC)
496 
497 static inline bool mm_p4d_folded(struct mm_struct *mm)
498 {
499 	return mm->context.asce_limit <= _REGION1_SIZE;
500 }
501 #define mm_p4d_folded(mm) mm_p4d_folded(mm)
502 
503 static inline bool mm_pud_folded(struct mm_struct *mm)
504 {
505 	return mm->context.asce_limit <= _REGION2_SIZE;
506 }
507 #define mm_pud_folded(mm) mm_pud_folded(mm)
508 
509 static inline bool mm_pmd_folded(struct mm_struct *mm)
510 {
511 	return mm->context.asce_limit <= _REGION3_SIZE;
512 }
513 #define mm_pmd_folded(mm) mm_pmd_folded(mm)
514 
515 static inline int mm_has_pgste(struct mm_struct *mm)
516 {
517 #ifdef CONFIG_PGSTE
518 	if (unlikely(mm->context.has_pgste))
519 		return 1;
520 #endif
521 	return 0;
522 }
523 
524 static inline int mm_alloc_pgste(struct mm_struct *mm)
525 {
526 #ifdef CONFIG_PGSTE
527 	if (unlikely(mm->context.alloc_pgste))
528 		return 1;
529 #endif
530 	return 0;
531 }
532 
533 /*
534  * In the case that a guest uses storage keys
535  * faults should no longer be backed by zero pages
536  */
537 #define mm_forbids_zeropage mm_has_pgste
538 static inline int mm_uses_skeys(struct mm_struct *mm)
539 {
540 #ifdef CONFIG_PGSTE
541 	if (mm->context.uses_skeys)
542 		return 1;
543 #endif
544 	return 0;
545 }
546 
547 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
548 {
549 	register unsigned long reg2 asm("2") = old;
550 	register unsigned long reg3 asm("3") = new;
551 	unsigned long address = (unsigned long)ptr | 1;
552 
553 	asm volatile(
554 		"	csp	%0,%3"
555 		: "+d" (reg2), "+m" (*ptr)
556 		: "d" (reg3), "d" (address)
557 		: "cc");
558 }
559 
560 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
561 {
562 	register unsigned long reg2 asm("2") = old;
563 	register unsigned long reg3 asm("3") = new;
564 	unsigned long address = (unsigned long)ptr | 1;
565 
566 	asm volatile(
567 		"	.insn	rre,0xb98a0000,%0,%3"
568 		: "+d" (reg2), "+m" (*ptr)
569 		: "d" (reg3), "d" (address)
570 		: "cc");
571 }
572 
573 #define CRDTE_DTT_PAGE		0x00UL
574 #define CRDTE_DTT_SEGMENT	0x10UL
575 #define CRDTE_DTT_REGION3	0x14UL
576 #define CRDTE_DTT_REGION2	0x18UL
577 #define CRDTE_DTT_REGION1	0x1cUL
578 
579 static inline void crdte(unsigned long old, unsigned long new,
580 			 unsigned long table, unsigned long dtt,
581 			 unsigned long address, unsigned long asce)
582 {
583 	register unsigned long reg2 asm("2") = old;
584 	register unsigned long reg3 asm("3") = new;
585 	register unsigned long reg4 asm("4") = table | dtt;
586 	register unsigned long reg5 asm("5") = address;
587 
588 	asm volatile(".insn rrf,0xb98f0000,%0,%2,%4,0"
589 		     : "+d" (reg2)
590 		     : "d" (reg3), "d" (reg4), "d" (reg5), "a" (asce)
591 		     : "memory", "cc");
592 }
593 
594 /*
595  * pgd/p4d/pud/pmd/pte query functions
596  */
597 static inline int pgd_folded(pgd_t pgd)
598 {
599 	return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
600 }
601 
602 static inline int pgd_present(pgd_t pgd)
603 {
604 	if (pgd_folded(pgd))
605 		return 1;
606 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
607 }
608 
609 static inline int pgd_none(pgd_t pgd)
610 {
611 	if (pgd_folded(pgd))
612 		return 0;
613 	return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
614 }
615 
616 static inline int pgd_bad(pgd_t pgd)
617 {
618 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1)
619 		return 0;
620 	return (pgd_val(pgd) & ~_REGION_ENTRY_BITS) != 0;
621 }
622 
623 static inline unsigned long pgd_pfn(pgd_t pgd)
624 {
625 	unsigned long origin_mask;
626 
627 	origin_mask = _REGION_ENTRY_ORIGIN;
628 	return (pgd_val(pgd) & origin_mask) >> PAGE_SHIFT;
629 }
630 
631 static inline int p4d_folded(p4d_t p4d)
632 {
633 	return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
634 }
635 
636 static inline int p4d_present(p4d_t p4d)
637 {
638 	if (p4d_folded(p4d))
639 		return 1;
640 	return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
641 }
642 
643 static inline int p4d_none(p4d_t p4d)
644 {
645 	if (p4d_folded(p4d))
646 		return 0;
647 	return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
648 }
649 
650 static inline unsigned long p4d_pfn(p4d_t p4d)
651 {
652 	unsigned long origin_mask;
653 
654 	origin_mask = _REGION_ENTRY_ORIGIN;
655 	return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
656 }
657 
658 static inline int pud_folded(pud_t pud)
659 {
660 	return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
661 }
662 
663 static inline int pud_present(pud_t pud)
664 {
665 	if (pud_folded(pud))
666 		return 1;
667 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
668 }
669 
670 static inline int pud_none(pud_t pud)
671 {
672 	if (pud_folded(pud))
673 		return 0;
674 	return pud_val(pud) == _REGION3_ENTRY_EMPTY;
675 }
676 
677 static inline int pud_large(pud_t pud)
678 {
679 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
680 		return 0;
681 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
682 }
683 
684 static inline unsigned long pud_pfn(pud_t pud)
685 {
686 	unsigned long origin_mask;
687 
688 	origin_mask = _REGION_ENTRY_ORIGIN;
689 	if (pud_large(pud))
690 		origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
691 	return (pud_val(pud) & origin_mask) >> PAGE_SHIFT;
692 }
693 
694 static inline int pmd_large(pmd_t pmd)
695 {
696 	return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
697 }
698 
699 static inline int pmd_bad(pmd_t pmd)
700 {
701 	if ((pmd_val(pmd) & _SEGMENT_ENTRY_TYPE_MASK) > 0)
702 		return 1;
703 	if (pmd_large(pmd))
704 		return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
705 	return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
706 }
707 
708 static inline int pud_bad(pud_t pud)
709 {
710 	unsigned long type = pud_val(pud) & _REGION_ENTRY_TYPE_MASK;
711 
712 	if (type > _REGION_ENTRY_TYPE_R3)
713 		return 1;
714 	if (type < _REGION_ENTRY_TYPE_R3)
715 		return 0;
716 	if (pud_large(pud))
717 		return (pud_val(pud) & ~_REGION_ENTRY_BITS_LARGE) != 0;
718 	return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
719 }
720 
721 static inline int p4d_bad(p4d_t p4d)
722 {
723 	unsigned long type = p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK;
724 
725 	if (type > _REGION_ENTRY_TYPE_R2)
726 		return 1;
727 	if (type < _REGION_ENTRY_TYPE_R2)
728 		return 0;
729 	return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
730 }
731 
732 static inline int pmd_present(pmd_t pmd)
733 {
734 	return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
735 }
736 
737 static inline int pmd_none(pmd_t pmd)
738 {
739 	return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
740 }
741 
742 static inline unsigned long pmd_pfn(pmd_t pmd)
743 {
744 	unsigned long origin_mask;
745 
746 	origin_mask = _SEGMENT_ENTRY_ORIGIN;
747 	if (pmd_large(pmd))
748 		origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
749 	return (pmd_val(pmd) & origin_mask) >> PAGE_SHIFT;
750 }
751 
752 #define pmd_write pmd_write
753 static inline int pmd_write(pmd_t pmd)
754 {
755 	return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
756 }
757 
758 static inline int pmd_dirty(pmd_t pmd)
759 {
760 	int dirty = 1;
761 	if (pmd_large(pmd))
762 		dirty = (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
763 	return dirty;
764 }
765 
766 static inline int pmd_young(pmd_t pmd)
767 {
768 	int young = 1;
769 	if (pmd_large(pmd))
770 		young = (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
771 	return young;
772 }
773 
774 static inline int pte_present(pte_t pte)
775 {
776 	/* Bit pattern: (pte & 0x001) == 0x001 */
777 	return (pte_val(pte) & _PAGE_PRESENT) != 0;
778 }
779 
780 static inline int pte_none(pte_t pte)
781 {
782 	/* Bit pattern: pte == 0x400 */
783 	return pte_val(pte) == _PAGE_INVALID;
784 }
785 
786 static inline int pte_swap(pte_t pte)
787 {
788 	/* Bit pattern: (pte & 0x201) == 0x200 */
789 	return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
790 		== _PAGE_PROTECT;
791 }
792 
793 static inline int pte_special(pte_t pte)
794 {
795 	return (pte_val(pte) & _PAGE_SPECIAL);
796 }
797 
798 #define __HAVE_ARCH_PTE_SAME
799 static inline int pte_same(pte_t a, pte_t b)
800 {
801 	return pte_val(a) == pte_val(b);
802 }
803 
804 #ifdef CONFIG_NUMA_BALANCING
805 static inline int pte_protnone(pte_t pte)
806 {
807 	return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
808 }
809 
810 static inline int pmd_protnone(pmd_t pmd)
811 {
812 	/* pmd_large(pmd) implies pmd_present(pmd) */
813 	return pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
814 }
815 #endif
816 
817 static inline int pte_soft_dirty(pte_t pte)
818 {
819 	return pte_val(pte) & _PAGE_SOFT_DIRTY;
820 }
821 #define pte_swp_soft_dirty pte_soft_dirty
822 
823 static inline pte_t pte_mksoft_dirty(pte_t pte)
824 {
825 	pte_val(pte) |= _PAGE_SOFT_DIRTY;
826 	return pte;
827 }
828 #define pte_swp_mksoft_dirty pte_mksoft_dirty
829 
830 static inline pte_t pte_clear_soft_dirty(pte_t pte)
831 {
832 	pte_val(pte) &= ~_PAGE_SOFT_DIRTY;
833 	return pte;
834 }
835 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
836 
837 static inline int pmd_soft_dirty(pmd_t pmd)
838 {
839 	return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
840 }
841 
842 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
843 {
844 	pmd_val(pmd) |= _SEGMENT_ENTRY_SOFT_DIRTY;
845 	return pmd;
846 }
847 
848 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
849 {
850 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_SOFT_DIRTY;
851 	return pmd;
852 }
853 
854 /*
855  * query functions pte_write/pte_dirty/pte_young only work if
856  * pte_present() is true. Undefined behaviour if not..
857  */
858 static inline int pte_write(pte_t pte)
859 {
860 	return (pte_val(pte) & _PAGE_WRITE) != 0;
861 }
862 
863 static inline int pte_dirty(pte_t pte)
864 {
865 	return (pte_val(pte) & _PAGE_DIRTY) != 0;
866 }
867 
868 static inline int pte_young(pte_t pte)
869 {
870 	return (pte_val(pte) & _PAGE_YOUNG) != 0;
871 }
872 
873 #define __HAVE_ARCH_PTE_UNUSED
874 static inline int pte_unused(pte_t pte)
875 {
876 	return pte_val(pte) & _PAGE_UNUSED;
877 }
878 
879 /*
880  * pgd/pmd/pte modification functions
881  */
882 
883 static inline void pgd_clear(pgd_t *pgd)
884 {
885 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
886 		pgd_val(*pgd) = _REGION1_ENTRY_EMPTY;
887 }
888 
889 static inline void p4d_clear(p4d_t *p4d)
890 {
891 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
892 		p4d_val(*p4d) = _REGION2_ENTRY_EMPTY;
893 }
894 
895 static inline void pud_clear(pud_t *pud)
896 {
897 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
898 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
899 }
900 
901 static inline void pmd_clear(pmd_t *pmdp)
902 {
903 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
904 }
905 
906 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
907 {
908 	pte_val(*ptep) = _PAGE_INVALID;
909 }
910 
911 /*
912  * The following pte modification functions only work if
913  * pte_present() is true. Undefined behaviour if not..
914  */
915 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
916 {
917 	pte_val(pte) &= _PAGE_CHG_MASK;
918 	pte_val(pte) |= pgprot_val(newprot);
919 	/*
920 	 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
921 	 * has the invalid bit set, clear it again for readable, young pages
922 	 */
923 	if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
924 		pte_val(pte) &= ~_PAGE_INVALID;
925 	/*
926 	 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
927 	 * protection bit set, clear it again for writable, dirty pages
928 	 */
929 	if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
930 		pte_val(pte) &= ~_PAGE_PROTECT;
931 	return pte;
932 }
933 
934 static inline pte_t pte_wrprotect(pte_t pte)
935 {
936 	pte_val(pte) &= ~_PAGE_WRITE;
937 	pte_val(pte) |= _PAGE_PROTECT;
938 	return pte;
939 }
940 
941 static inline pte_t pte_mkwrite(pte_t pte)
942 {
943 	pte_val(pte) |= _PAGE_WRITE;
944 	if (pte_val(pte) & _PAGE_DIRTY)
945 		pte_val(pte) &= ~_PAGE_PROTECT;
946 	return pte;
947 }
948 
949 static inline pte_t pte_mkclean(pte_t pte)
950 {
951 	pte_val(pte) &= ~_PAGE_DIRTY;
952 	pte_val(pte) |= _PAGE_PROTECT;
953 	return pte;
954 }
955 
956 static inline pte_t pte_mkdirty(pte_t pte)
957 {
958 	pte_val(pte) |= _PAGE_DIRTY | _PAGE_SOFT_DIRTY;
959 	if (pte_val(pte) & _PAGE_WRITE)
960 		pte_val(pte) &= ~_PAGE_PROTECT;
961 	return pte;
962 }
963 
964 static inline pte_t pte_mkold(pte_t pte)
965 {
966 	pte_val(pte) &= ~_PAGE_YOUNG;
967 	pte_val(pte) |= _PAGE_INVALID;
968 	return pte;
969 }
970 
971 static inline pte_t pte_mkyoung(pte_t pte)
972 {
973 	pte_val(pte) |= _PAGE_YOUNG;
974 	if (pte_val(pte) & _PAGE_READ)
975 		pte_val(pte) &= ~_PAGE_INVALID;
976 	return pte;
977 }
978 
979 static inline pte_t pte_mkspecial(pte_t pte)
980 {
981 	pte_val(pte) |= _PAGE_SPECIAL;
982 	return pte;
983 }
984 
985 #ifdef CONFIG_HUGETLB_PAGE
986 static inline pte_t pte_mkhuge(pte_t pte)
987 {
988 	pte_val(pte) |= _PAGE_LARGE;
989 	return pte;
990 }
991 #endif
992 
993 #define IPTE_GLOBAL	0
994 #define	IPTE_LOCAL	1
995 
996 #define IPTE_NODAT	0x400
997 #define IPTE_GUEST_ASCE	0x800
998 
999 static inline void __ptep_ipte(unsigned long address, pte_t *ptep,
1000 			       unsigned long opt, unsigned long asce,
1001 			       int local)
1002 {
1003 	unsigned long pto = (unsigned long) ptep;
1004 
1005 	if (__builtin_constant_p(opt) && opt == 0) {
1006 		/* Invalidation + TLB flush for the pte */
1007 		asm volatile(
1008 			"	.insn	rrf,0xb2210000,%[r1],%[r2],0,%[m4]"
1009 			: "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
1010 			  [m4] "i" (local));
1011 		return;
1012 	}
1013 
1014 	/* Invalidate ptes with options + TLB flush of the ptes */
1015 	opt = opt | (asce & _ASCE_ORIGIN);
1016 	asm volatile(
1017 		"	.insn	rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
1018 		: [r2] "+a" (address), [r3] "+a" (opt)
1019 		: [r1] "a" (pto), [m4] "i" (local) : "memory");
1020 }
1021 
1022 static inline void __ptep_ipte_range(unsigned long address, int nr,
1023 				     pte_t *ptep, int local)
1024 {
1025 	unsigned long pto = (unsigned long) ptep;
1026 
1027 	/* Invalidate a range of ptes + TLB flush of the ptes */
1028 	do {
1029 		asm volatile(
1030 			"       .insn rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
1031 			: [r2] "+a" (address), [r3] "+a" (nr)
1032 			: [r1] "a" (pto), [m4] "i" (local) : "memory");
1033 	} while (nr != 255);
1034 }
1035 
1036 /*
1037  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1038  * both clear the TLB for the unmapped pte. The reason is that
1039  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1040  * to modify an active pte. The sequence is
1041  *   1) ptep_get_and_clear
1042  *   2) set_pte_at
1043  *   3) flush_tlb_range
1044  * On s390 the tlb needs to get flushed with the modification of the pte
1045  * if the pte is active. The only way how this can be implemented is to
1046  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1047  * is a nop.
1048  */
1049 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1050 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1051 
1052 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1053 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1054 					    unsigned long addr, pte_t *ptep)
1055 {
1056 	pte_t pte = *ptep;
1057 
1058 	pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1059 	return pte_young(pte);
1060 }
1061 
1062 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1063 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1064 					 unsigned long address, pte_t *ptep)
1065 {
1066 	return ptep_test_and_clear_young(vma, address, ptep);
1067 }
1068 
1069 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1070 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1071 				       unsigned long addr, pte_t *ptep)
1072 {
1073 	return ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1074 }
1075 
1076 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1077 pte_t ptep_modify_prot_start(struct vm_area_struct *, unsigned long, pte_t *);
1078 void ptep_modify_prot_commit(struct vm_area_struct *, unsigned long,
1079 			     pte_t *, pte_t, pte_t);
1080 
1081 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1082 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1083 				     unsigned long addr, pte_t *ptep)
1084 {
1085 	return ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1086 }
1087 
1088 /*
1089  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1090  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1091  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1092  * cannot be accessed while the batched unmap is running. In this case
1093  * full==1 and a simple pte_clear is enough. See tlb.h.
1094  */
1095 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1096 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1097 					    unsigned long addr,
1098 					    pte_t *ptep, int full)
1099 {
1100 	if (full) {
1101 		pte_t pte = *ptep;
1102 		*ptep = __pte(_PAGE_INVALID);
1103 		return pte;
1104 	}
1105 	return ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1106 }
1107 
1108 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1109 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1110 				      unsigned long addr, pte_t *ptep)
1111 {
1112 	pte_t pte = *ptep;
1113 
1114 	if (pte_write(pte))
1115 		ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1116 }
1117 
1118 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1119 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1120 					unsigned long addr, pte_t *ptep,
1121 					pte_t entry, int dirty)
1122 {
1123 	if (pte_same(*ptep, entry))
1124 		return 0;
1125 	ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1126 	return 1;
1127 }
1128 
1129 /*
1130  * Additional functions to handle KVM guest page tables
1131  */
1132 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1133 		     pte_t *ptep, pte_t entry);
1134 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1135 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1136 		 pte_t *ptep, unsigned long bits);
1137 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1138 		    pte_t *ptep, int prot, unsigned long bit);
1139 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1140 		     pte_t *ptep , int reset);
1141 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1142 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1143 		    pte_t *sptep, pte_t *tptep, pte_t pte);
1144 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1145 
1146 bool ptep_test_and_clear_uc(struct mm_struct *mm, unsigned long address,
1147 			    pte_t *ptep);
1148 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1149 			  unsigned char key, bool nq);
1150 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1151 			       unsigned char key, unsigned char *oldkey,
1152 			       bool nq, bool mr, bool mc);
1153 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1154 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1155 			  unsigned char *key);
1156 
1157 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1158 				unsigned long bits, unsigned long value);
1159 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1160 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1161 			unsigned long *oldpte, unsigned long *oldpgste);
1162 void gmap_pmdp_csp(struct mm_struct *mm, unsigned long vmaddr);
1163 void gmap_pmdp_invalidate(struct mm_struct *mm, unsigned long vmaddr);
1164 void gmap_pmdp_idte_local(struct mm_struct *mm, unsigned long vmaddr);
1165 void gmap_pmdp_idte_global(struct mm_struct *mm, unsigned long vmaddr);
1166 
1167 /*
1168  * Certain architectures need to do special things when PTEs
1169  * within a page table are directly modified.  Thus, the following
1170  * hook is made available.
1171  */
1172 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
1173 			      pte_t *ptep, pte_t entry)
1174 {
1175 	if (!MACHINE_HAS_NX)
1176 		pte_val(entry) &= ~_PAGE_NOEXEC;
1177 	if (pte_present(entry))
1178 		pte_val(entry) &= ~_PAGE_UNUSED;
1179 	if (mm_has_pgste(mm))
1180 		ptep_set_pte_at(mm, addr, ptep, entry);
1181 	else
1182 		*ptep = entry;
1183 }
1184 
1185 /*
1186  * Conversion functions: convert a page and protection to a page entry,
1187  * and a page entry and page directory to the page they refer to.
1188  */
1189 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1190 {
1191 	pte_t __pte;
1192 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1193 	return pte_mkyoung(__pte);
1194 }
1195 
1196 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1197 {
1198 	unsigned long physpage = page_to_phys(page);
1199 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1200 
1201 	if (pte_write(__pte) && PageDirty(page))
1202 		__pte = pte_mkdirty(__pte);
1203 	return __pte;
1204 }
1205 
1206 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1207 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1208 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1209 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1210 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1211 
1212 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1213 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1214 #define p4d_deref(pud) (p4d_val(pud) & _REGION_ENTRY_ORIGIN)
1215 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1216 
1217 /*
1218  * The pgd_offset function *always* adds the index for the top-level
1219  * region/segment table. This is done to get a sequence like the
1220  * following to work:
1221  *	pgdp = pgd_offset(current->mm, addr);
1222  *	pgd = READ_ONCE(*pgdp);
1223  *	p4dp = p4d_offset(&pgd, addr);
1224  *	...
1225  * The subsequent p4d_offset, pud_offset and pmd_offset functions
1226  * only add an index if they dereferenced the pointer.
1227  */
1228 static inline pgd_t *pgd_offset_raw(pgd_t *pgd, unsigned long address)
1229 {
1230 	unsigned long rste;
1231 	unsigned int shift;
1232 
1233 	/* Get the first entry of the top level table */
1234 	rste = pgd_val(*pgd);
1235 	/* Pick up the shift from the table type of the first entry */
1236 	shift = ((rste & _REGION_ENTRY_TYPE_MASK) >> 2) * 11 + 20;
1237 	return pgd + ((address >> shift) & (PTRS_PER_PGD - 1));
1238 }
1239 
1240 #define pgd_offset(mm, address) pgd_offset_raw(READ_ONCE((mm)->pgd), address)
1241 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1242 
1243 static inline p4d_t *p4d_offset(pgd_t *pgd, unsigned long address)
1244 {
1245 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R1)
1246 		return (p4d_t *) pgd_deref(*pgd) + p4d_index(address);
1247 	return (p4d_t *) pgd;
1248 }
1249 
1250 static inline pud_t *pud_offset(p4d_t *p4d, unsigned long address)
1251 {
1252 	if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R2)
1253 		return (pud_t *) p4d_deref(*p4d) + pud_index(address);
1254 	return (pud_t *) p4d;
1255 }
1256 
1257 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1258 {
1259 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R3)
1260 		return (pmd_t *) pud_deref(*pud) + pmd_index(address);
1261 	return (pmd_t *) pud;
1262 }
1263 
1264 static inline pte_t *pte_offset(pmd_t *pmd, unsigned long address)
1265 {
1266 	return (pte_t *) pmd_deref(*pmd) + pte_index(address);
1267 }
1268 
1269 #define pte_offset_kernel(pmd, address) pte_offset(pmd, address)
1270 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1271 #define pte_unmap(pte) do { } while (0)
1272 
1273 static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
1274 {
1275 	return end <= current->mm->context.asce_limit;
1276 }
1277 #define gup_fast_permitted gup_fast_permitted
1278 
1279 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1280 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1281 #define pte_page(x) pfn_to_page(pte_pfn(x))
1282 
1283 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1284 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1285 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
1286 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
1287 
1288 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1289 {
1290 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_WRITE;
1291 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1292 	return pmd;
1293 }
1294 
1295 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1296 {
1297 	pmd_val(pmd) |= _SEGMENT_ENTRY_WRITE;
1298 	if (pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1299 		return pmd;
1300 	pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1301 	return pmd;
1302 }
1303 
1304 static inline pmd_t pmd_mkclean(pmd_t pmd)
1305 {
1306 	if (pmd_large(pmd)) {
1307 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_DIRTY;
1308 		pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1309 	}
1310 	return pmd;
1311 }
1312 
1313 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1314 {
1315 	if (pmd_large(pmd)) {
1316 		pmd_val(pmd) |= _SEGMENT_ENTRY_DIRTY |
1317 				_SEGMENT_ENTRY_SOFT_DIRTY;
1318 		if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1319 			pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1320 	}
1321 	return pmd;
1322 }
1323 
1324 static inline pud_t pud_wrprotect(pud_t pud)
1325 {
1326 	pud_val(pud) &= ~_REGION3_ENTRY_WRITE;
1327 	pud_val(pud) |= _REGION_ENTRY_PROTECT;
1328 	return pud;
1329 }
1330 
1331 static inline pud_t pud_mkwrite(pud_t pud)
1332 {
1333 	pud_val(pud) |= _REGION3_ENTRY_WRITE;
1334 	if (pud_large(pud) && !(pud_val(pud) & _REGION3_ENTRY_DIRTY))
1335 		return pud;
1336 	pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1337 	return pud;
1338 }
1339 
1340 static inline pud_t pud_mkclean(pud_t pud)
1341 {
1342 	if (pud_large(pud)) {
1343 		pud_val(pud) &= ~_REGION3_ENTRY_DIRTY;
1344 		pud_val(pud) |= _REGION_ENTRY_PROTECT;
1345 	}
1346 	return pud;
1347 }
1348 
1349 static inline pud_t pud_mkdirty(pud_t pud)
1350 {
1351 	if (pud_large(pud)) {
1352 		pud_val(pud) |= _REGION3_ENTRY_DIRTY |
1353 				_REGION3_ENTRY_SOFT_DIRTY;
1354 		if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1355 			pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1356 	}
1357 	return pud;
1358 }
1359 
1360 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1361 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1362 {
1363 	/*
1364 	 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1365 	 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1366 	 */
1367 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1368 		return pgprot_val(SEGMENT_NONE);
1369 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1370 		return pgprot_val(SEGMENT_RO);
1371 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1372 		return pgprot_val(SEGMENT_RX);
1373 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1374 		return pgprot_val(SEGMENT_RW);
1375 	return pgprot_val(SEGMENT_RWX);
1376 }
1377 
1378 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1379 {
1380 	if (pmd_large(pmd)) {
1381 		pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1382 		if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1383 			pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1384 	}
1385 	return pmd;
1386 }
1387 
1388 static inline pmd_t pmd_mkold(pmd_t pmd)
1389 {
1390 	if (pmd_large(pmd)) {
1391 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1392 		pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1393 	}
1394 	return pmd;
1395 }
1396 
1397 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1398 {
1399 	if (pmd_large(pmd)) {
1400 		pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN_LARGE |
1401 			_SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_YOUNG |
1402 			_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_SOFT_DIRTY;
1403 		pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1404 		if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1405 			pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1406 		if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1407 			pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1408 		return pmd;
1409 	}
1410 	pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN;
1411 	pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1412 	return pmd;
1413 }
1414 
1415 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1416 {
1417 	pmd_t __pmd;
1418 	pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1419 	return __pmd;
1420 }
1421 
1422 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1423 
1424 static inline void __pmdp_csp(pmd_t *pmdp)
1425 {
1426 	csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1427 	    pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1428 }
1429 
1430 #define IDTE_GLOBAL	0
1431 #define IDTE_LOCAL	1
1432 
1433 #define IDTE_PTOA	0x0800
1434 #define IDTE_NODAT	0x1000
1435 #define IDTE_GUEST_ASCE	0x2000
1436 
1437 static inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1438 			       unsigned long opt, unsigned long asce,
1439 			       int local)
1440 {
1441 	unsigned long sto;
1442 
1443 	sto = (unsigned long) pmdp - pmd_index(addr) * sizeof(pmd_t);
1444 	if (__builtin_constant_p(opt) && opt == 0) {
1445 		/* flush without guest asce */
1446 		asm volatile(
1447 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1448 			: "+m" (*pmdp)
1449 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1450 			  [m4] "i" (local)
1451 			: "cc" );
1452 	} else {
1453 		/* flush with guest asce */
1454 		asm volatile(
1455 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1456 			: "+m" (*pmdp)
1457 			: [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1458 			  [r3] "a" (asce), [m4] "i" (local)
1459 			: "cc" );
1460 	}
1461 }
1462 
1463 static inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1464 			       unsigned long opt, unsigned long asce,
1465 			       int local)
1466 {
1467 	unsigned long r3o;
1468 
1469 	r3o = (unsigned long) pudp - pud_index(addr) * sizeof(pud_t);
1470 	r3o |= _ASCE_TYPE_REGION3;
1471 	if (__builtin_constant_p(opt) && opt == 0) {
1472 		/* flush without guest asce */
1473 		asm volatile(
1474 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1475 			: "+m" (*pudp)
1476 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1477 			  [m4] "i" (local)
1478 			: "cc");
1479 	} else {
1480 		/* flush with guest asce */
1481 		asm volatile(
1482 			"	.insn	rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1483 			: "+m" (*pudp)
1484 			: [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1485 			  [r3] "a" (asce), [m4] "i" (local)
1486 			: "cc" );
1487 	}
1488 }
1489 
1490 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1491 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1492 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1493 
1494 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1495 
1496 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1497 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1498 				pgtable_t pgtable);
1499 
1500 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1501 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1502 
1503 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1504 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1505 					unsigned long addr, pmd_t *pmdp,
1506 					pmd_t entry, int dirty)
1507 {
1508 	VM_BUG_ON(addr & ~HPAGE_MASK);
1509 
1510 	entry = pmd_mkyoung(entry);
1511 	if (dirty)
1512 		entry = pmd_mkdirty(entry);
1513 	if (pmd_val(*pmdp) == pmd_val(entry))
1514 		return 0;
1515 	pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1516 	return 1;
1517 }
1518 
1519 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1520 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1521 					    unsigned long addr, pmd_t *pmdp)
1522 {
1523 	pmd_t pmd = *pmdp;
1524 
1525 	pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1526 	return pmd_young(pmd);
1527 }
1528 
1529 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1530 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1531 					 unsigned long addr, pmd_t *pmdp)
1532 {
1533 	VM_BUG_ON(addr & ~HPAGE_MASK);
1534 	return pmdp_test_and_clear_young(vma, addr, pmdp);
1535 }
1536 
1537 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1538 			      pmd_t *pmdp, pmd_t entry)
1539 {
1540 	if (!MACHINE_HAS_NX)
1541 		pmd_val(entry) &= ~_SEGMENT_ENTRY_NOEXEC;
1542 	*pmdp = entry;
1543 }
1544 
1545 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1546 {
1547 	pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1548 	pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1549 	pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1550 	return pmd;
1551 }
1552 
1553 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1554 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1555 					    unsigned long addr, pmd_t *pmdp)
1556 {
1557 	return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1558 }
1559 
1560 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1561 static inline pmd_t pmdp_huge_get_and_clear_full(struct mm_struct *mm,
1562 						 unsigned long addr,
1563 						 pmd_t *pmdp, int full)
1564 {
1565 	if (full) {
1566 		pmd_t pmd = *pmdp;
1567 		*pmdp = __pmd(_SEGMENT_ENTRY_EMPTY);
1568 		return pmd;
1569 	}
1570 	return pmdp_xchg_lazy(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1571 }
1572 
1573 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1574 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1575 					  unsigned long addr, pmd_t *pmdp)
1576 {
1577 	return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1578 }
1579 
1580 #define __HAVE_ARCH_PMDP_INVALIDATE
1581 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1582 				   unsigned long addr, pmd_t *pmdp)
1583 {
1584 	pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1585 
1586 	return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1587 }
1588 
1589 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1590 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1591 				      unsigned long addr, pmd_t *pmdp)
1592 {
1593 	pmd_t pmd = *pmdp;
1594 
1595 	if (pmd_write(pmd))
1596 		pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1597 }
1598 
1599 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1600 					unsigned long address,
1601 					pmd_t *pmdp)
1602 {
1603 	return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1604 }
1605 #define pmdp_collapse_flush pmdp_collapse_flush
1606 
1607 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1608 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1609 
1610 static inline int pmd_trans_huge(pmd_t pmd)
1611 {
1612 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1613 }
1614 
1615 #define has_transparent_hugepage has_transparent_hugepage
1616 static inline int has_transparent_hugepage(void)
1617 {
1618 	return MACHINE_HAS_EDAT1 ? 1 : 0;
1619 }
1620 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1621 
1622 /*
1623  * 64 bit swap entry format:
1624  * A page-table entry has some bits we have to treat in a special way.
1625  * Bits 52 and bit 55 have to be zero, otherwise a specification
1626  * exception will occur instead of a page translation exception. The
1627  * specification exception has the bad habit not to store necessary
1628  * information in the lowcore.
1629  * Bits 54 and 63 are used to indicate the page type.
1630  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1631  * This leaves the bits 0-51 and bits 56-62 to store type and offset.
1632  * We use the 5 bits from 57-61 for the type and the 52 bits from 0-51
1633  * for the offset.
1634  * |			  offset			|01100|type |00|
1635  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1636  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1637  */
1638 
1639 #define __SWP_OFFSET_MASK	((1UL << 52) - 1)
1640 #define __SWP_OFFSET_SHIFT	12
1641 #define __SWP_TYPE_MASK		((1UL << 5) - 1)
1642 #define __SWP_TYPE_SHIFT	2
1643 
1644 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1645 {
1646 	pte_t pte;
1647 
1648 	pte_val(pte) = _PAGE_INVALID | _PAGE_PROTECT;
1649 	pte_val(pte) |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1650 	pte_val(pte) |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1651 	return pte;
1652 }
1653 
1654 static inline unsigned long __swp_type(swp_entry_t entry)
1655 {
1656 	return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1657 }
1658 
1659 static inline unsigned long __swp_offset(swp_entry_t entry)
1660 {
1661 	return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1662 }
1663 
1664 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1665 {
1666 	return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1667 }
1668 
1669 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1670 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1671 
1672 #define kern_addr_valid(addr)   (1)
1673 
1674 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1675 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1676 extern int s390_enable_sie(void);
1677 extern int s390_enable_skey(void);
1678 extern void s390_reset_cmma(struct mm_struct *mm);
1679 
1680 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1681 #define HAVE_ARCH_UNMAPPED_AREA
1682 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1683 
1684 /*
1685  * No page table caches to initialise
1686  */
1687 static inline void pgtable_cache_init(void) { }
1688 static inline void check_pgt_cache(void) { }
1689 
1690 #include <asm-generic/pgtable.h>
1691 
1692 #endif /* _S390_PAGE_H */
1693