xref: /openbmc/linux/arch/s390/include/asm/pgtable.h (revision 7b6d864b)
1 /*
2  *  S390 version
3  *    Copyright IBM Corp. 1999, 2000
4  *    Author(s): Hartmut Penner (hp@de.ibm.com)
5  *               Ulrich Weigand (weigand@de.ibm.com)
6  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
7  *
8  *  Derived from "include/asm-i386/pgtable.h"
9  */
10 
11 #ifndef _ASM_S390_PGTABLE_H
12 #define _ASM_S390_PGTABLE_H
13 
14 /*
15  * The Linux memory management assumes a three-level page table setup. For
16  * s390 31 bit we "fold" the mid level into the top-level page table, so
17  * that we physically have the same two-level page table as the s390 mmu
18  * expects in 31 bit mode. For s390 64 bit we use three of the five levels
19  * the hardware provides (region first and region second tables are not
20  * used).
21  *
22  * The "pgd_xxx()" functions are trivial for a folded two-level
23  * setup: the pgd is never bad, and a pmd always exists (as it's folded
24  * into the pgd entry)
25  *
26  * This file contains the functions and defines necessary to modify and use
27  * the S390 page table tree.
28  */
29 #ifndef __ASSEMBLY__
30 #include <linux/sched.h>
31 #include <linux/mm_types.h>
32 #include <linux/page-flags.h>
33 #include <asm/bug.h>
34 #include <asm/page.h>
35 
36 extern pgd_t swapper_pg_dir[] __attribute__ ((aligned (4096)));
37 extern void paging_init(void);
38 extern void vmem_map_init(void);
39 
40 /*
41  * The S390 doesn't have any external MMU info: the kernel page
42  * tables contain all the necessary information.
43  */
44 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
45 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
46 
47 /*
48  * ZERO_PAGE is a global shared page that is always zero; used
49  * for zero-mapped memory areas etc..
50  */
51 
52 extern unsigned long empty_zero_page;
53 extern unsigned long zero_page_mask;
54 
55 #define ZERO_PAGE(vaddr) \
56 	(virt_to_page((void *)(empty_zero_page + \
57 	 (((unsigned long)(vaddr)) &zero_page_mask))))
58 #define __HAVE_COLOR_ZERO_PAGE
59 
60 /* TODO: s390 cannot support io_remap_pfn_range... */
61 #endif /* !__ASSEMBLY__ */
62 
63 /*
64  * PMD_SHIFT determines the size of the area a second-level page
65  * table can map
66  * PGDIR_SHIFT determines what a third-level page table entry can map
67  */
68 #ifndef CONFIG_64BIT
69 # define PMD_SHIFT	20
70 # define PUD_SHIFT	20
71 # define PGDIR_SHIFT	20
72 #else /* CONFIG_64BIT */
73 # define PMD_SHIFT	20
74 # define PUD_SHIFT	31
75 # define PGDIR_SHIFT	42
76 #endif /* CONFIG_64BIT */
77 
78 #define PMD_SIZE        (1UL << PMD_SHIFT)
79 #define PMD_MASK        (~(PMD_SIZE-1))
80 #define PUD_SIZE	(1UL << PUD_SHIFT)
81 #define PUD_MASK	(~(PUD_SIZE-1))
82 #define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
83 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
84 
85 /*
86  * entries per page directory level: the S390 is two-level, so
87  * we don't really have any PMD directory physically.
88  * for S390 segment-table entries are combined to one PGD
89  * that leads to 1024 pte per pgd
90  */
91 #define PTRS_PER_PTE	256
92 #ifndef CONFIG_64BIT
93 #define PTRS_PER_PMD	1
94 #define PTRS_PER_PUD	1
95 #else /* CONFIG_64BIT */
96 #define PTRS_PER_PMD	2048
97 #define PTRS_PER_PUD	2048
98 #endif /* CONFIG_64BIT */
99 #define PTRS_PER_PGD	2048
100 
101 #define FIRST_USER_ADDRESS  0
102 
103 #define pte_ERROR(e) \
104 	printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
105 #define pmd_ERROR(e) \
106 	printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
107 #define pud_ERROR(e) \
108 	printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
109 #define pgd_ERROR(e) \
110 	printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
111 
112 #ifndef __ASSEMBLY__
113 /*
114  * The vmalloc and module area will always be on the topmost area of the kernel
115  * mapping. We reserve 96MB (31bit) / 128GB (64bit) for vmalloc and modules.
116  * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
117  * modules will reside. That makes sure that inter module branches always
118  * happen without trampolines and in addition the placement within a 2GB frame
119  * is branch prediction unit friendly.
120  */
121 extern unsigned long VMALLOC_START;
122 extern unsigned long VMALLOC_END;
123 extern struct page *vmemmap;
124 
125 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
126 
127 #ifdef CONFIG_64BIT
128 extern unsigned long MODULES_VADDR;
129 extern unsigned long MODULES_END;
130 #define MODULES_VADDR	MODULES_VADDR
131 #define MODULES_END	MODULES_END
132 #define MODULES_LEN	(1UL << 31)
133 #endif
134 
135 /*
136  * A 31 bit pagetable entry of S390 has following format:
137  *  |   PFRA          |    |  OS  |
138  * 0                   0IP0
139  * 00000000001111111111222222222233
140  * 01234567890123456789012345678901
141  *
142  * I Page-Invalid Bit:    Page is not available for address-translation
143  * P Page-Protection Bit: Store access not possible for page
144  *
145  * A 31 bit segmenttable entry of S390 has following format:
146  *  |   P-table origin      |  |PTL
147  * 0                         IC
148  * 00000000001111111111222222222233
149  * 01234567890123456789012345678901
150  *
151  * I Segment-Invalid Bit:    Segment is not available for address-translation
152  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
153  * PTL Page-Table-Length:    Page-table length (PTL+1*16 entries -> up to 256)
154  *
155  * The 31 bit segmenttable origin of S390 has following format:
156  *
157  *  |S-table origin   |     | STL |
158  * X                   **GPS
159  * 00000000001111111111222222222233
160  * 01234567890123456789012345678901
161  *
162  * X Space-Switch event:
163  * G Segment-Invalid Bit:     *
164  * P Private-Space Bit:       Segment is not private (PoP 3-30)
165  * S Storage-Alteration:
166  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
167  *
168  * A 64 bit pagetable entry of S390 has following format:
169  * |			 PFRA			      |0IPC|  OS  |
170  * 0000000000111111111122222222223333333333444444444455555555556666
171  * 0123456789012345678901234567890123456789012345678901234567890123
172  *
173  * I Page-Invalid Bit:    Page is not available for address-translation
174  * P Page-Protection Bit: Store access not possible for page
175  * C Change-bit override: HW is not required to set change bit
176  *
177  * A 64 bit segmenttable entry of S390 has following format:
178  * |        P-table origin                              |      TT
179  * 0000000000111111111122222222223333333333444444444455555555556666
180  * 0123456789012345678901234567890123456789012345678901234567890123
181  *
182  * I Segment-Invalid Bit:    Segment is not available for address-translation
183  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
184  * P Page-Protection Bit: Store access not possible for page
185  * TT Type 00
186  *
187  * A 64 bit region table entry of S390 has following format:
188  * |        S-table origin                             |   TF  TTTL
189  * 0000000000111111111122222222223333333333444444444455555555556666
190  * 0123456789012345678901234567890123456789012345678901234567890123
191  *
192  * I Segment-Invalid Bit:    Segment is not available for address-translation
193  * TT Type 01
194  * TF
195  * TL Table length
196  *
197  * The 64 bit regiontable origin of S390 has following format:
198  * |      region table origon                          |       DTTL
199  * 0000000000111111111122222222223333333333444444444455555555556666
200  * 0123456789012345678901234567890123456789012345678901234567890123
201  *
202  * X Space-Switch event:
203  * G Segment-Invalid Bit:
204  * P Private-Space Bit:
205  * S Storage-Alteration:
206  * R Real space
207  * TL Table-Length:
208  *
209  * A storage key has the following format:
210  * | ACC |F|R|C|0|
211  *  0   3 4 5 6 7
212  * ACC: access key
213  * F  : fetch protection bit
214  * R  : referenced bit
215  * C  : changed bit
216  */
217 
218 /* Hardware bits in the page table entry */
219 #define _PAGE_CO	0x100		/* HW Change-bit override */
220 #define _PAGE_RO	0x200		/* HW read-only bit  */
221 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
222 
223 /* Software bits in the page table entry */
224 #define _PAGE_SWT	0x001		/* SW pte type bit t */
225 #define _PAGE_SWX	0x002		/* SW pte type bit x */
226 #define _PAGE_SWC	0x004		/* SW pte changed bit */
227 #define _PAGE_SWR	0x008		/* SW pte referenced bit */
228 #define _PAGE_SWW	0x010		/* SW pte write bit */
229 #define _PAGE_SPECIAL	0x020		/* SW associated with special page */
230 #define __HAVE_ARCH_PTE_SPECIAL
231 
232 /* Set of bits not changed in pte_modify */
233 #define _PAGE_CHG_MASK		(PAGE_MASK | _PAGE_SPECIAL | _PAGE_CO | \
234 				 _PAGE_SWC | _PAGE_SWR)
235 
236 /* Six different types of pages. */
237 #define _PAGE_TYPE_EMPTY	0x400
238 #define _PAGE_TYPE_NONE		0x401
239 #define _PAGE_TYPE_SWAP		0x403
240 #define _PAGE_TYPE_FILE		0x601	/* bit 0x002 is used for offset !! */
241 #define _PAGE_TYPE_RO		0x200
242 #define _PAGE_TYPE_RW		0x000
243 
244 /*
245  * Only four types for huge pages, using the invalid bit and protection bit
246  * of a segment table entry.
247  */
248 #define _HPAGE_TYPE_EMPTY	0x020	/* _SEGMENT_ENTRY_INV */
249 #define _HPAGE_TYPE_NONE	0x220
250 #define _HPAGE_TYPE_RO		0x200	/* _SEGMENT_ENTRY_RO  */
251 #define _HPAGE_TYPE_RW		0x000
252 
253 /*
254  * PTE type bits are rather complicated. handle_pte_fault uses pte_present,
255  * pte_none and pte_file to find out the pte type WITHOUT holding the page
256  * table lock. ptep_clear_flush on the other hand uses ptep_clear_flush to
257  * invalidate a given pte. ipte sets the hw invalid bit and clears all tlbs
258  * for the page. The page table entry is set to _PAGE_TYPE_EMPTY afterwards.
259  * This change is done while holding the lock, but the intermediate step
260  * of a previously valid pte with the hw invalid bit set can be observed by
261  * handle_pte_fault. That makes it necessary that all valid pte types with
262  * the hw invalid bit set must be distinguishable from the four pte types
263  * empty, none, swap and file.
264  *
265  *			irxt  ipte  irxt
266  * _PAGE_TYPE_EMPTY	1000   ->   1000
267  * _PAGE_TYPE_NONE	1001   ->   1001
268  * _PAGE_TYPE_SWAP	1011   ->   1011
269  * _PAGE_TYPE_FILE	11?1   ->   11?1
270  * _PAGE_TYPE_RO	0100   ->   1100
271  * _PAGE_TYPE_RW	0000   ->   1000
272  *
273  * pte_none is true for bits combinations 1000, 1010, 1100, 1110
274  * pte_present is true for bits combinations 0000, 0010, 0100, 0110, 1001
275  * pte_file is true for bits combinations 1101, 1111
276  * swap pte is 1011 and 0001, 0011, 0101, 0111 are invalid.
277  */
278 
279 #ifndef CONFIG_64BIT
280 
281 /* Bits in the segment table address-space-control-element */
282 #define _ASCE_SPACE_SWITCH	0x80000000UL	/* space switch event	    */
283 #define _ASCE_ORIGIN_MASK	0x7ffff000UL	/* segment table origin	    */
284 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
285 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
286 #define _ASCE_TABLE_LENGTH	0x7f	/* 128 x 64 entries = 8k	    */
287 
288 /* Bits in the segment table entry */
289 #define _SEGMENT_ENTRY_ORIGIN	0x7fffffc0UL	/* page table origin	    */
290 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
291 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
292 #define _SEGMENT_ENTRY_COMMON	0x10	/* common segment bit		    */
293 #define _SEGMENT_ENTRY_PTL	0x0f	/* page table length		    */
294 
295 #define _SEGMENT_ENTRY		(_SEGMENT_ENTRY_PTL)
296 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
297 
298 /* Page status table bits for virtualization */
299 #define PGSTE_ACC_BITS	0xf0000000UL
300 #define PGSTE_FP_BIT	0x08000000UL
301 #define PGSTE_PCL_BIT	0x00800000UL
302 #define PGSTE_HR_BIT	0x00400000UL
303 #define PGSTE_HC_BIT	0x00200000UL
304 #define PGSTE_GR_BIT	0x00040000UL
305 #define PGSTE_GC_BIT	0x00020000UL
306 #define PGSTE_UR_BIT	0x00008000UL
307 #define PGSTE_UC_BIT	0x00004000UL	/* user dirty (migration) */
308 #define PGSTE_IN_BIT	0x00002000UL	/* IPTE notify bit */
309 
310 #else /* CONFIG_64BIT */
311 
312 /* Bits in the segment/region table address-space-control-element */
313 #define _ASCE_ORIGIN		~0xfffUL/* segment table origin		    */
314 #define _ASCE_PRIVATE_SPACE	0x100	/* private space control	    */
315 #define _ASCE_ALT_EVENT		0x80	/* storage alteration event control */
316 #define _ASCE_SPACE_SWITCH	0x40	/* space switch event		    */
317 #define _ASCE_REAL_SPACE	0x20	/* real space control		    */
318 #define _ASCE_TYPE_MASK		0x0c	/* asce table type mask		    */
319 #define _ASCE_TYPE_REGION1	0x0c	/* region first table type	    */
320 #define _ASCE_TYPE_REGION2	0x08	/* region second table type	    */
321 #define _ASCE_TYPE_REGION3	0x04	/* region third table type	    */
322 #define _ASCE_TYPE_SEGMENT	0x00	/* segment table type		    */
323 #define _ASCE_TABLE_LENGTH	0x03	/* region table length		    */
324 
325 /* Bits in the region table entry */
326 #define _REGION_ENTRY_ORIGIN	~0xfffUL/* region/segment table origin	    */
327 #define _REGION_ENTRY_RO	0x200	/* region protection bit	    */
328 #define _REGION_ENTRY_INV	0x20	/* invalid region table entry	    */
329 #define _REGION_ENTRY_TYPE_MASK	0x0c	/* region/segment table type mask   */
330 #define _REGION_ENTRY_TYPE_R1	0x0c	/* region first table type	    */
331 #define _REGION_ENTRY_TYPE_R2	0x08	/* region second table type	    */
332 #define _REGION_ENTRY_TYPE_R3	0x04	/* region third table type	    */
333 #define _REGION_ENTRY_LENGTH	0x03	/* region third length		    */
334 
335 #define _REGION1_ENTRY		(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
336 #define _REGION1_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INV)
337 #define _REGION2_ENTRY		(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
338 #define _REGION2_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INV)
339 #define _REGION3_ENTRY		(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
340 #define _REGION3_ENTRY_EMPTY	(_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INV)
341 
342 #define _REGION3_ENTRY_LARGE	0x400	/* RTTE-format control, large page  */
343 #define _REGION3_ENTRY_RO	0x200	/* page protection bit		    */
344 #define _REGION3_ENTRY_CO	0x100	/* change-recording override	    */
345 
346 /* Bits in the segment table entry */
347 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address	    */
348 #define _SEGMENT_ENTRY_ORIGIN	~0x7ffUL/* segment table origin		    */
349 #define _SEGMENT_ENTRY_RO	0x200	/* page protection bit		    */
350 #define _SEGMENT_ENTRY_INV	0x20	/* invalid segment table entry	    */
351 
352 #define _SEGMENT_ENTRY		(0)
353 #define _SEGMENT_ENTRY_EMPTY	(_SEGMENT_ENTRY_INV)
354 
355 #define _SEGMENT_ENTRY_LARGE	0x400	/* STE-format control, large page   */
356 #define _SEGMENT_ENTRY_CO	0x100	/* change-recording override   */
357 #define _SEGMENT_ENTRY_SPLIT_BIT 0	/* THP splitting bit number */
358 #define _SEGMENT_ENTRY_SPLIT	(1UL << _SEGMENT_ENTRY_SPLIT_BIT)
359 
360 /* Set of bits not changed in pmd_modify */
361 #define _SEGMENT_CHG_MASK	(_SEGMENT_ENTRY_ORIGIN | _SEGMENT_ENTRY_LARGE \
362 				 | _SEGMENT_ENTRY_SPLIT | _SEGMENT_ENTRY_CO)
363 
364 /* Page status table bits for virtualization */
365 #define PGSTE_ACC_BITS	0xf000000000000000UL
366 #define PGSTE_FP_BIT	0x0800000000000000UL
367 #define PGSTE_PCL_BIT	0x0080000000000000UL
368 #define PGSTE_HR_BIT	0x0040000000000000UL
369 #define PGSTE_HC_BIT	0x0020000000000000UL
370 #define PGSTE_GR_BIT	0x0004000000000000UL
371 #define PGSTE_GC_BIT	0x0002000000000000UL
372 #define PGSTE_UR_BIT	0x0000800000000000UL
373 #define PGSTE_UC_BIT	0x0000400000000000UL	/* user dirty (migration) */
374 #define PGSTE_IN_BIT	0x0000200000000000UL	/* IPTE notify bit */
375 
376 #endif /* CONFIG_64BIT */
377 
378 /*
379  * A user page table pointer has the space-switch-event bit, the
380  * private-space-control bit and the storage-alteration-event-control
381  * bit set. A kernel page table pointer doesn't need them.
382  */
383 #define _ASCE_USER_BITS		(_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
384 				 _ASCE_ALT_EVENT)
385 
386 /*
387  * Page protection definitions.
388  */
389 #define PAGE_NONE	__pgprot(_PAGE_TYPE_NONE)
390 #define PAGE_RO		__pgprot(_PAGE_TYPE_RO)
391 #define PAGE_RW		__pgprot(_PAGE_TYPE_RO | _PAGE_SWW)
392 #define PAGE_RWC	__pgprot(_PAGE_TYPE_RW | _PAGE_SWW | _PAGE_SWC)
393 
394 #define PAGE_KERNEL	PAGE_RWC
395 #define PAGE_SHARED	PAGE_KERNEL
396 #define PAGE_COPY	PAGE_RO
397 
398 /*
399  * On s390 the page table entry has an invalid bit and a read-only bit.
400  * Read permission implies execute permission and write permission
401  * implies read permission.
402  */
403          /*xwr*/
404 #define __P000	PAGE_NONE
405 #define __P001	PAGE_RO
406 #define __P010	PAGE_RO
407 #define __P011	PAGE_RO
408 #define __P100	PAGE_RO
409 #define __P101	PAGE_RO
410 #define __P110	PAGE_RO
411 #define __P111	PAGE_RO
412 
413 #define __S000	PAGE_NONE
414 #define __S001	PAGE_RO
415 #define __S010	PAGE_RW
416 #define __S011	PAGE_RW
417 #define __S100	PAGE_RO
418 #define __S101	PAGE_RO
419 #define __S110	PAGE_RW
420 #define __S111	PAGE_RW
421 
422 /*
423  * Segment entry (large page) protection definitions.
424  */
425 #define SEGMENT_NONE	__pgprot(_HPAGE_TYPE_NONE)
426 #define SEGMENT_RO	__pgprot(_HPAGE_TYPE_RO)
427 #define SEGMENT_RW	__pgprot(_HPAGE_TYPE_RW)
428 
429 static inline int mm_exclusive(struct mm_struct *mm)
430 {
431 	return likely(mm == current->active_mm &&
432 		      atomic_read(&mm->context.attach_count) <= 1);
433 }
434 
435 static inline int mm_has_pgste(struct mm_struct *mm)
436 {
437 #ifdef CONFIG_PGSTE
438 	if (unlikely(mm->context.has_pgste))
439 		return 1;
440 #endif
441 	return 0;
442 }
443 /*
444  * pgd/pmd/pte query functions
445  */
446 #ifndef CONFIG_64BIT
447 
448 static inline int pgd_present(pgd_t pgd) { return 1; }
449 static inline int pgd_none(pgd_t pgd)    { return 0; }
450 static inline int pgd_bad(pgd_t pgd)     { return 0; }
451 
452 static inline int pud_present(pud_t pud) { return 1; }
453 static inline int pud_none(pud_t pud)	 { return 0; }
454 static inline int pud_large(pud_t pud)	 { return 0; }
455 static inline int pud_bad(pud_t pud)	 { return 0; }
456 
457 #else /* CONFIG_64BIT */
458 
459 static inline int pgd_present(pgd_t pgd)
460 {
461 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
462 		return 1;
463 	return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
464 }
465 
466 static inline int pgd_none(pgd_t pgd)
467 {
468 	if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
469 		return 0;
470 	return (pgd_val(pgd) & _REGION_ENTRY_INV) != 0UL;
471 }
472 
473 static inline int pgd_bad(pgd_t pgd)
474 {
475 	/*
476 	 * With dynamic page table levels the pgd can be a region table
477 	 * entry or a segment table entry. Check for the bit that are
478 	 * invalid for either table entry.
479 	 */
480 	unsigned long mask =
481 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
482 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
483 	return (pgd_val(pgd) & mask) != 0;
484 }
485 
486 static inline int pud_present(pud_t pud)
487 {
488 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
489 		return 1;
490 	return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
491 }
492 
493 static inline int pud_none(pud_t pud)
494 {
495 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
496 		return 0;
497 	return (pud_val(pud) & _REGION_ENTRY_INV) != 0UL;
498 }
499 
500 static inline int pud_large(pud_t pud)
501 {
502 	if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
503 		return 0;
504 	return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
505 }
506 
507 static inline int pud_bad(pud_t pud)
508 {
509 	/*
510 	 * With dynamic page table levels the pud can be a region table
511 	 * entry or a segment table entry. Check for the bit that are
512 	 * invalid for either table entry.
513 	 */
514 	unsigned long mask =
515 		~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INV &
516 		~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
517 	return (pud_val(pud) & mask) != 0;
518 }
519 
520 #endif /* CONFIG_64BIT */
521 
522 static inline int pmd_present(pmd_t pmd)
523 {
524 	unsigned long mask = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO;
525 	return (pmd_val(pmd) & mask) == _HPAGE_TYPE_NONE ||
526 	       !(pmd_val(pmd) & _SEGMENT_ENTRY_INV);
527 }
528 
529 static inline int pmd_none(pmd_t pmd)
530 {
531 	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) &&
532 	       !(pmd_val(pmd) & _SEGMENT_ENTRY_RO);
533 }
534 
535 static inline int pmd_large(pmd_t pmd)
536 {
537 #ifdef CONFIG_64BIT
538 	return !!(pmd_val(pmd) & _SEGMENT_ENTRY_LARGE);
539 #else
540 	return 0;
541 #endif
542 }
543 
544 static inline int pmd_bad(pmd_t pmd)
545 {
546 	unsigned long mask = ~_SEGMENT_ENTRY_ORIGIN & ~_SEGMENT_ENTRY_INV;
547 	return (pmd_val(pmd) & mask) != _SEGMENT_ENTRY;
548 }
549 
550 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
551 extern void pmdp_splitting_flush(struct vm_area_struct *vma,
552 				 unsigned long addr, pmd_t *pmdp);
553 
554 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
555 extern int pmdp_set_access_flags(struct vm_area_struct *vma,
556 				 unsigned long address, pmd_t *pmdp,
557 				 pmd_t entry, int dirty);
558 
559 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
560 extern int pmdp_clear_flush_young(struct vm_area_struct *vma,
561 				  unsigned long address, pmd_t *pmdp);
562 
563 #define __HAVE_ARCH_PMD_WRITE
564 static inline int pmd_write(pmd_t pmd)
565 {
566 	return (pmd_val(pmd) & _SEGMENT_ENTRY_RO) == 0;
567 }
568 
569 static inline int pmd_young(pmd_t pmd)
570 {
571 	return 0;
572 }
573 
574 static inline int pte_none(pte_t pte)
575 {
576 	return (pte_val(pte) & _PAGE_INVALID) && !(pte_val(pte) & _PAGE_SWT);
577 }
578 
579 static inline int pte_present(pte_t pte)
580 {
581 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT | _PAGE_SWX;
582 	return (pte_val(pte) & mask) == _PAGE_TYPE_NONE ||
583 		(!(pte_val(pte) & _PAGE_INVALID) &&
584 		 !(pte_val(pte) & _PAGE_SWT));
585 }
586 
587 static inline int pte_file(pte_t pte)
588 {
589 	unsigned long mask = _PAGE_RO | _PAGE_INVALID | _PAGE_SWT;
590 	return (pte_val(pte) & mask) == _PAGE_TYPE_FILE;
591 }
592 
593 static inline int pte_special(pte_t pte)
594 {
595 	return (pte_val(pte) & _PAGE_SPECIAL);
596 }
597 
598 #define __HAVE_ARCH_PTE_SAME
599 static inline int pte_same(pte_t a, pte_t b)
600 {
601 	return pte_val(a) == pte_val(b);
602 }
603 
604 static inline pgste_t pgste_get_lock(pte_t *ptep)
605 {
606 	unsigned long new = 0;
607 #ifdef CONFIG_PGSTE
608 	unsigned long old;
609 
610 	preempt_disable();
611 	asm(
612 		"	lg	%0,%2\n"
613 		"0:	lgr	%1,%0\n"
614 		"	nihh	%0,0xff7f\n"	/* clear PCL bit in old */
615 		"	oihh	%1,0x0080\n"	/* set PCL bit in new */
616 		"	csg	%0,%1,%2\n"
617 		"	jl	0b\n"
618 		: "=&d" (old), "=&d" (new), "=Q" (ptep[PTRS_PER_PTE])
619 		: "Q" (ptep[PTRS_PER_PTE]) : "cc", "memory");
620 #endif
621 	return __pgste(new);
622 }
623 
624 static inline void pgste_set_unlock(pte_t *ptep, pgste_t pgste)
625 {
626 #ifdef CONFIG_PGSTE
627 	asm(
628 		"	nihh	%1,0xff7f\n"	/* clear PCL bit */
629 		"	stg	%1,%0\n"
630 		: "=Q" (ptep[PTRS_PER_PTE])
631 		: "d" (pgste_val(pgste)), "Q" (ptep[PTRS_PER_PTE])
632 		: "cc", "memory");
633 	preempt_enable();
634 #endif
635 }
636 
637 static inline void pgste_set(pte_t *ptep, pgste_t pgste)
638 {
639 #ifdef CONFIG_PGSTE
640 	*(pgste_t *)(ptep + PTRS_PER_PTE) = pgste;
641 #endif
642 }
643 
644 static inline pgste_t pgste_update_all(pte_t *ptep, pgste_t pgste)
645 {
646 #ifdef CONFIG_PGSTE
647 	unsigned long address, bits;
648 	unsigned char skey;
649 
650 	if (pte_val(*ptep) & _PAGE_INVALID)
651 		return pgste;
652 	address = pte_val(*ptep) & PAGE_MASK;
653 	skey = page_get_storage_key(address);
654 	bits = skey & (_PAGE_CHANGED | _PAGE_REFERENCED);
655 	/* Clear page changed & referenced bit in the storage key */
656 	if (bits & _PAGE_CHANGED)
657 		page_set_storage_key(address, skey ^ bits, 0);
658 	else if (bits)
659 		page_reset_referenced(address);
660 	/* Transfer page changed & referenced bit to guest bits in pgste */
661 	pgste_val(pgste) |= bits << 48;		/* GR bit & GC bit */
662 	/* Get host changed & referenced bits from pgste */
663 	bits |= (pgste_val(pgste) & (PGSTE_HR_BIT | PGSTE_HC_BIT)) >> 52;
664 	/* Transfer page changed & referenced bit to kvm user bits */
665 	pgste_val(pgste) |= bits << 45;		/* PGSTE_UR_BIT & PGSTE_UC_BIT */
666 	/* Clear relevant host bits in pgste. */
667 	pgste_val(pgste) &= ~(PGSTE_HR_BIT | PGSTE_HC_BIT);
668 	pgste_val(pgste) &= ~(PGSTE_ACC_BITS | PGSTE_FP_BIT);
669 	/* Copy page access key and fetch protection bit to pgste */
670 	pgste_val(pgste) |=
671 		(unsigned long) (skey & (_PAGE_ACC_BITS | _PAGE_FP_BIT)) << 56;
672 	/* Transfer referenced bit to pte */
673 	pte_val(*ptep) |= (bits & _PAGE_REFERENCED) << 1;
674 #endif
675 	return pgste;
676 
677 }
678 
679 static inline pgste_t pgste_update_young(pte_t *ptep, pgste_t pgste)
680 {
681 #ifdef CONFIG_PGSTE
682 	int young;
683 
684 	if (pte_val(*ptep) & _PAGE_INVALID)
685 		return pgste;
686 	/* Get referenced bit from storage key */
687 	young = page_reset_referenced(pte_val(*ptep) & PAGE_MASK);
688 	if (young)
689 		pgste_val(pgste) |= PGSTE_GR_BIT;
690 	/* Get host referenced bit from pgste */
691 	if (pgste_val(pgste) & PGSTE_HR_BIT) {
692 		pgste_val(pgste) &= ~PGSTE_HR_BIT;
693 		young = 1;
694 	}
695 	/* Transfer referenced bit to kvm user bits and pte */
696 	if (young) {
697 		pgste_val(pgste) |= PGSTE_UR_BIT;
698 		pte_val(*ptep) |= _PAGE_SWR;
699 	}
700 #endif
701 	return pgste;
702 }
703 
704 static inline void pgste_set_key(pte_t *ptep, pgste_t pgste, pte_t entry)
705 {
706 #ifdef CONFIG_PGSTE
707 	unsigned long address;
708 	unsigned long nkey;
709 
710 	if (pte_val(entry) & _PAGE_INVALID)
711 		return;
712 	VM_BUG_ON(!(pte_val(*ptep) & _PAGE_INVALID));
713 	address = pte_val(entry) & PAGE_MASK;
714 	/*
715 	 * Set page access key and fetch protection bit from pgste.
716 	 * The guest C/R information is still in the PGSTE, set real
717 	 * key C/R to 0.
718 	 */
719 	nkey = (pgste_val(pgste) & (PGSTE_ACC_BITS | PGSTE_FP_BIT)) >> 56;
720 	page_set_storage_key(address, nkey, 0);
721 #endif
722 }
723 
724 static inline void pgste_set_pte(pte_t *ptep, pte_t entry)
725 {
726 	if (!MACHINE_HAS_ESOP && (pte_val(entry) & _PAGE_SWW)) {
727 		/*
728 		 * Without enhanced suppression-on-protection force
729 		 * the dirty bit on for all writable ptes.
730 		 */
731 		pte_val(entry) |= _PAGE_SWC;
732 		pte_val(entry) &= ~_PAGE_RO;
733 	}
734 	*ptep = entry;
735 }
736 
737 /**
738  * struct gmap_struct - guest address space
739  * @mm: pointer to the parent mm_struct
740  * @table: pointer to the page directory
741  * @asce: address space control element for gmap page table
742  * @crst_list: list of all crst tables used in the guest address space
743  */
744 struct gmap {
745 	struct list_head list;
746 	struct mm_struct *mm;
747 	unsigned long *table;
748 	unsigned long asce;
749 	void *private;
750 	struct list_head crst_list;
751 };
752 
753 /**
754  * struct gmap_rmap - reverse mapping for segment table entries
755  * @gmap: pointer to the gmap_struct
756  * @entry: pointer to a segment table entry
757  * @vmaddr: virtual address in the guest address space
758  */
759 struct gmap_rmap {
760 	struct list_head list;
761 	struct gmap *gmap;
762 	unsigned long *entry;
763 	unsigned long vmaddr;
764 };
765 
766 /**
767  * struct gmap_pgtable - gmap information attached to a page table
768  * @vmaddr: address of the 1MB segment in the process virtual memory
769  * @mapper: list of segment table entries mapping a page table
770  */
771 struct gmap_pgtable {
772 	unsigned long vmaddr;
773 	struct list_head mapper;
774 };
775 
776 /**
777  * struct gmap_notifier - notify function block for page invalidation
778  * @notifier_call: address of callback function
779  */
780 struct gmap_notifier {
781 	struct list_head list;
782 	void (*notifier_call)(struct gmap *gmap, unsigned long address);
783 };
784 
785 struct gmap *gmap_alloc(struct mm_struct *mm);
786 void gmap_free(struct gmap *gmap);
787 void gmap_enable(struct gmap *gmap);
788 void gmap_disable(struct gmap *gmap);
789 int gmap_map_segment(struct gmap *gmap, unsigned long from,
790 		     unsigned long to, unsigned long len);
791 int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len);
792 unsigned long __gmap_translate(unsigned long address, struct gmap *);
793 unsigned long gmap_translate(unsigned long address, struct gmap *);
794 unsigned long __gmap_fault(unsigned long address, struct gmap *);
795 unsigned long gmap_fault(unsigned long address, struct gmap *);
796 void gmap_discard(unsigned long from, unsigned long to, struct gmap *);
797 
798 void gmap_register_ipte_notifier(struct gmap_notifier *);
799 void gmap_unregister_ipte_notifier(struct gmap_notifier *);
800 int gmap_ipte_notify(struct gmap *, unsigned long start, unsigned long len);
801 void gmap_do_ipte_notify(struct mm_struct *, unsigned long addr, pte_t *);
802 
803 static inline pgste_t pgste_ipte_notify(struct mm_struct *mm,
804 					unsigned long addr,
805 					pte_t *ptep, pgste_t pgste)
806 {
807 #ifdef CONFIG_PGSTE
808 	if (pgste_val(pgste) & PGSTE_IN_BIT) {
809 		pgste_val(pgste) &= ~PGSTE_IN_BIT;
810 		gmap_do_ipte_notify(mm, addr, ptep);
811 	}
812 #endif
813 	return pgste;
814 }
815 
816 /*
817  * Certain architectures need to do special things when PTEs
818  * within a page table are directly modified.  Thus, the following
819  * hook is made available.
820  */
821 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
822 			      pte_t *ptep, pte_t entry)
823 {
824 	pgste_t pgste;
825 
826 	if (mm_has_pgste(mm)) {
827 		pgste = pgste_get_lock(ptep);
828 		pgste_set_key(ptep, pgste, entry);
829 		pgste_set_pte(ptep, entry);
830 		pgste_set_unlock(ptep, pgste);
831 	} else {
832 		if (!(pte_val(entry) & _PAGE_INVALID) && MACHINE_HAS_EDAT1)
833 			pte_val(entry) |= _PAGE_CO;
834 		*ptep = entry;
835 	}
836 }
837 
838 /*
839  * query functions pte_write/pte_dirty/pte_young only work if
840  * pte_present() is true. Undefined behaviour if not..
841  */
842 static inline int pte_write(pte_t pte)
843 {
844 	return (pte_val(pte) & _PAGE_SWW) != 0;
845 }
846 
847 static inline int pte_dirty(pte_t pte)
848 {
849 	return (pte_val(pte) & _PAGE_SWC) != 0;
850 }
851 
852 static inline int pte_young(pte_t pte)
853 {
854 #ifdef CONFIG_PGSTE
855 	if (pte_val(pte) & _PAGE_SWR)
856 		return 1;
857 #endif
858 	return 0;
859 }
860 
861 /*
862  * pgd/pmd/pte modification functions
863  */
864 
865 static inline void pgd_clear(pgd_t *pgd)
866 {
867 #ifdef CONFIG_64BIT
868 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
869 		pgd_val(*pgd) = _REGION2_ENTRY_EMPTY;
870 #endif
871 }
872 
873 static inline void pud_clear(pud_t *pud)
874 {
875 #ifdef CONFIG_64BIT
876 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
877 		pud_val(*pud) = _REGION3_ENTRY_EMPTY;
878 #endif
879 }
880 
881 static inline void pmd_clear(pmd_t *pmdp)
882 {
883 	pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
884 }
885 
886 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
887 {
888 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
889 }
890 
891 /*
892  * The following pte modification functions only work if
893  * pte_present() is true. Undefined behaviour if not..
894  */
895 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
896 {
897 	pte_val(pte) &= _PAGE_CHG_MASK;
898 	pte_val(pte) |= pgprot_val(newprot);
899 	if ((pte_val(pte) & _PAGE_SWC) && (pte_val(pte) & _PAGE_SWW))
900 		pte_val(pte) &= ~_PAGE_RO;
901 	return pte;
902 }
903 
904 static inline pte_t pte_wrprotect(pte_t pte)
905 {
906 	pte_val(pte) &= ~_PAGE_SWW;
907 	/* Do not clobber _PAGE_TYPE_NONE pages!  */
908 	if (!(pte_val(pte) & _PAGE_INVALID))
909 		pte_val(pte) |= _PAGE_RO;
910 	return pte;
911 }
912 
913 static inline pte_t pte_mkwrite(pte_t pte)
914 {
915 	pte_val(pte) |= _PAGE_SWW;
916 	if (pte_val(pte) & _PAGE_SWC)
917 		pte_val(pte) &= ~_PAGE_RO;
918 	return pte;
919 }
920 
921 static inline pte_t pte_mkclean(pte_t pte)
922 {
923 	pte_val(pte) &= ~_PAGE_SWC;
924 	/* Do not clobber _PAGE_TYPE_NONE pages!  */
925 	if (!(pte_val(pte) & _PAGE_INVALID))
926 		pte_val(pte) |= _PAGE_RO;
927 	return pte;
928 }
929 
930 static inline pte_t pte_mkdirty(pte_t pte)
931 {
932 	pte_val(pte) |= _PAGE_SWC;
933 	if (pte_val(pte) & _PAGE_SWW)
934 		pte_val(pte) &= ~_PAGE_RO;
935 	return pte;
936 }
937 
938 static inline pte_t pte_mkold(pte_t pte)
939 {
940 #ifdef CONFIG_PGSTE
941 	pte_val(pte) &= ~_PAGE_SWR;
942 #endif
943 	return pte;
944 }
945 
946 static inline pte_t pte_mkyoung(pte_t pte)
947 {
948 	return pte;
949 }
950 
951 static inline pte_t pte_mkspecial(pte_t pte)
952 {
953 	pte_val(pte) |= _PAGE_SPECIAL;
954 	return pte;
955 }
956 
957 #ifdef CONFIG_HUGETLB_PAGE
958 static inline pte_t pte_mkhuge(pte_t pte)
959 {
960 	pte_val(pte) |= (_SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_CO);
961 	return pte;
962 }
963 #endif
964 
965 /*
966  * Get (and clear) the user dirty bit for a pte.
967  */
968 static inline int ptep_test_and_clear_user_dirty(struct mm_struct *mm,
969 						 pte_t *ptep)
970 {
971 	pgste_t pgste;
972 	int dirty = 0;
973 
974 	if (mm_has_pgste(mm)) {
975 		pgste = pgste_get_lock(ptep);
976 		pgste = pgste_update_all(ptep, pgste);
977 		dirty = !!(pgste_val(pgste) & PGSTE_UC_BIT);
978 		pgste_val(pgste) &= ~PGSTE_UC_BIT;
979 		pgste_set_unlock(ptep, pgste);
980 		return dirty;
981 	}
982 	return dirty;
983 }
984 
985 /*
986  * Get (and clear) the user referenced bit for a pte.
987  */
988 static inline int ptep_test_and_clear_user_young(struct mm_struct *mm,
989 						 pte_t *ptep)
990 {
991 	pgste_t pgste;
992 	int young = 0;
993 
994 	if (mm_has_pgste(mm)) {
995 		pgste = pgste_get_lock(ptep);
996 		pgste = pgste_update_young(ptep, pgste);
997 		young = !!(pgste_val(pgste) & PGSTE_UR_BIT);
998 		pgste_val(pgste) &= ~PGSTE_UR_BIT;
999 		pgste_set_unlock(ptep, pgste);
1000 	}
1001 	return young;
1002 }
1003 
1004 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1005 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1006 					    unsigned long addr, pte_t *ptep)
1007 {
1008 	pgste_t pgste;
1009 	pte_t pte;
1010 
1011 	if (mm_has_pgste(vma->vm_mm)) {
1012 		pgste = pgste_get_lock(ptep);
1013 		pgste = pgste_update_young(ptep, pgste);
1014 		pte = *ptep;
1015 		*ptep = pte_mkold(pte);
1016 		pgste_set_unlock(ptep, pgste);
1017 		return pte_young(pte);
1018 	}
1019 	return 0;
1020 }
1021 
1022 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1023 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1024 					 unsigned long address, pte_t *ptep)
1025 {
1026 	/* No need to flush TLB
1027 	 * On s390 reference bits are in storage key and never in TLB
1028 	 * With virtualization we handle the reference bit, without we
1029 	 * we can simply return */
1030 	return ptep_test_and_clear_young(vma, address, ptep);
1031 }
1032 
1033 static inline void __ptep_ipte(unsigned long address, pte_t *ptep)
1034 {
1035 	if (!(pte_val(*ptep) & _PAGE_INVALID)) {
1036 #ifndef CONFIG_64BIT
1037 		/* pto must point to the start of the segment table */
1038 		pte_t *pto = (pte_t *) (((unsigned long) ptep) & 0x7ffffc00);
1039 #else
1040 		/* ipte in zarch mode can do the math */
1041 		pte_t *pto = ptep;
1042 #endif
1043 		asm volatile(
1044 			"	ipte	%2,%3"
1045 			: "=m" (*ptep) : "m" (*ptep),
1046 			  "a" (pto), "a" (address));
1047 	}
1048 }
1049 
1050 /*
1051  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1052  * both clear the TLB for the unmapped pte. The reason is that
1053  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1054  * to modify an active pte. The sequence is
1055  *   1) ptep_get_and_clear
1056  *   2) set_pte_at
1057  *   3) flush_tlb_range
1058  * On s390 the tlb needs to get flushed with the modification of the pte
1059  * if the pte is active. The only way how this can be implemented is to
1060  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1061  * is a nop.
1062  */
1063 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1064 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1065 				       unsigned long address, pte_t *ptep)
1066 {
1067 	pgste_t pgste;
1068 	pte_t pte;
1069 
1070 	mm->context.flush_mm = 1;
1071 	if (mm_has_pgste(mm)) {
1072 		pgste = pgste_get_lock(ptep);
1073 		pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1074 	}
1075 
1076 	pte = *ptep;
1077 	if (!mm_exclusive(mm))
1078 		__ptep_ipte(address, ptep);
1079 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1080 
1081 	if (mm_has_pgste(mm)) {
1082 		pgste = pgste_update_all(&pte, pgste);
1083 		pgste_set_unlock(ptep, pgste);
1084 	}
1085 	return pte;
1086 }
1087 
1088 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1089 static inline pte_t ptep_modify_prot_start(struct mm_struct *mm,
1090 					   unsigned long address,
1091 					   pte_t *ptep)
1092 {
1093 	pgste_t pgste;
1094 	pte_t pte;
1095 
1096 	mm->context.flush_mm = 1;
1097 	if (mm_has_pgste(mm)) {
1098 		pgste = pgste_get_lock(ptep);
1099 		pgste_ipte_notify(mm, address, ptep, pgste);
1100 	}
1101 
1102 	pte = *ptep;
1103 	if (!mm_exclusive(mm))
1104 		__ptep_ipte(address, ptep);
1105 
1106 	if (mm_has_pgste(mm)) {
1107 		pgste = pgste_update_all(&pte, pgste);
1108 		pgste_set(ptep, pgste);
1109 	}
1110 	return pte;
1111 }
1112 
1113 static inline void ptep_modify_prot_commit(struct mm_struct *mm,
1114 					   unsigned long address,
1115 					   pte_t *ptep, pte_t pte)
1116 {
1117 	pgste_t pgste;
1118 
1119 	if (mm_has_pgste(mm)) {
1120 		pgste = *(pgste_t *)(ptep + PTRS_PER_PTE);
1121 		pgste_set_key(ptep, pgste, pte);
1122 		pgste_set_pte(ptep, pte);
1123 		pgste_set_unlock(ptep, pgste);
1124 	} else
1125 		*ptep = pte;
1126 }
1127 
1128 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1129 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1130 				     unsigned long address, pte_t *ptep)
1131 {
1132 	pgste_t pgste;
1133 	pte_t pte;
1134 
1135 	if (mm_has_pgste(vma->vm_mm)) {
1136 		pgste = pgste_get_lock(ptep);
1137 		pgste = pgste_ipte_notify(vma->vm_mm, address, ptep, pgste);
1138 	}
1139 
1140 	pte = *ptep;
1141 	__ptep_ipte(address, ptep);
1142 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1143 
1144 	if (mm_has_pgste(vma->vm_mm)) {
1145 		pgste = pgste_update_all(&pte, pgste);
1146 		pgste_set_unlock(ptep, pgste);
1147 	}
1148 	return pte;
1149 }
1150 
1151 /*
1152  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1153  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1154  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1155  * cannot be accessed while the batched unmap is running. In this case
1156  * full==1 and a simple pte_clear is enough. See tlb.h.
1157  */
1158 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1159 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1160 					    unsigned long address,
1161 					    pte_t *ptep, int full)
1162 {
1163 	pgste_t pgste;
1164 	pte_t pte;
1165 
1166 	if (mm_has_pgste(mm)) {
1167 		pgste = pgste_get_lock(ptep);
1168 		if (!full)
1169 			pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1170 	}
1171 
1172 	pte = *ptep;
1173 	if (!full)
1174 		__ptep_ipte(address, ptep);
1175 	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
1176 
1177 	if (mm_has_pgste(mm)) {
1178 		pgste = pgste_update_all(&pte, pgste);
1179 		pgste_set_unlock(ptep, pgste);
1180 	}
1181 	return pte;
1182 }
1183 
1184 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1185 static inline pte_t ptep_set_wrprotect(struct mm_struct *mm,
1186 				       unsigned long address, pte_t *ptep)
1187 {
1188 	pgste_t pgste;
1189 	pte_t pte = *ptep;
1190 
1191 	if (pte_write(pte)) {
1192 		mm->context.flush_mm = 1;
1193 		if (mm_has_pgste(mm)) {
1194 			pgste = pgste_get_lock(ptep);
1195 			pgste = pgste_ipte_notify(mm, address, ptep, pgste);
1196 		}
1197 
1198 		if (!mm_exclusive(mm))
1199 			__ptep_ipte(address, ptep);
1200 		pte = pte_wrprotect(pte);
1201 
1202 		if (mm_has_pgste(mm)) {
1203 			pgste_set_pte(ptep, pte);
1204 			pgste_set_unlock(ptep, pgste);
1205 		} else
1206 			*ptep = pte;
1207 	}
1208 	return pte;
1209 }
1210 
1211 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1212 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1213 					unsigned long address, pte_t *ptep,
1214 					pte_t entry, int dirty)
1215 {
1216 	pgste_t pgste;
1217 
1218 	if (pte_same(*ptep, entry))
1219 		return 0;
1220 	if (mm_has_pgste(vma->vm_mm)) {
1221 		pgste = pgste_get_lock(ptep);
1222 		pgste = pgste_ipte_notify(vma->vm_mm, address, ptep, pgste);
1223 	}
1224 
1225 	__ptep_ipte(address, ptep);
1226 
1227 	if (mm_has_pgste(vma->vm_mm)) {
1228 		pgste_set_pte(ptep, entry);
1229 		pgste_set_unlock(ptep, pgste);
1230 	} else
1231 		*ptep = entry;
1232 	return 1;
1233 }
1234 
1235 /*
1236  * Conversion functions: convert a page and protection to a page entry,
1237  * and a page entry and page directory to the page they refer to.
1238  */
1239 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1240 {
1241 	pte_t __pte;
1242 	pte_val(__pte) = physpage + pgprot_val(pgprot);
1243 	return __pte;
1244 }
1245 
1246 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1247 {
1248 	unsigned long physpage = page_to_phys(page);
1249 	pte_t __pte = mk_pte_phys(physpage, pgprot);
1250 
1251 	if ((pte_val(__pte) & _PAGE_SWW) && PageDirty(page)) {
1252 		pte_val(__pte) |= _PAGE_SWC;
1253 		pte_val(__pte) &= ~_PAGE_RO;
1254 	}
1255 	return __pte;
1256 }
1257 
1258 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1259 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1260 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1261 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1262 
1263 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1264 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1265 
1266 #ifndef CONFIG_64BIT
1267 
1268 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1269 #define pud_deref(pmd) ({ BUG(); 0UL; })
1270 #define pgd_deref(pmd) ({ BUG(); 0UL; })
1271 
1272 #define pud_offset(pgd, address) ((pud_t *) pgd)
1273 #define pmd_offset(pud, address) ((pmd_t *) pud + pmd_index(address))
1274 
1275 #else /* CONFIG_64BIT */
1276 
1277 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1278 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1279 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1280 
1281 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
1282 {
1283 	pud_t *pud = (pud_t *) pgd;
1284 	if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1285 		pud = (pud_t *) pgd_deref(*pgd);
1286 	return pud  + pud_index(address);
1287 }
1288 
1289 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1290 {
1291 	pmd_t *pmd = (pmd_t *) pud;
1292 	if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1293 		pmd = (pmd_t *) pud_deref(*pud);
1294 	return pmd + pmd_index(address);
1295 }
1296 
1297 #endif /* CONFIG_64BIT */
1298 
1299 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1300 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1301 #define pte_page(x) pfn_to_page(pte_pfn(x))
1302 
1303 #define pmd_page(pmd) pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
1304 
1305 /* Find an entry in the lowest level page table.. */
1306 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1307 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1308 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1309 #define pte_unmap(pte) do { } while (0)
1310 
1311 static inline void __pmd_idte(unsigned long address, pmd_t *pmdp)
1312 {
1313 	unsigned long sto = (unsigned long) pmdp -
1314 			    pmd_index(address) * sizeof(pmd_t);
1315 
1316 	if (!(pmd_val(*pmdp) & _SEGMENT_ENTRY_INV)) {
1317 		asm volatile(
1318 			"	.insn	rrf,0xb98e0000,%2,%3,0,0"
1319 			: "=m" (*pmdp)
1320 			: "m" (*pmdp), "a" (sto),
1321 			  "a" ((address & HPAGE_MASK))
1322 			: "cc"
1323 		);
1324 	}
1325 }
1326 
1327 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1328 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1329 {
1330 	/*
1331 	 * pgprot is PAGE_NONE, PAGE_RO, or PAGE_RW (see __Pxxx / __Sxxx)
1332 	 * Convert to segment table entry format.
1333 	 */
1334 	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1335 		return pgprot_val(SEGMENT_NONE);
1336 	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1337 		return pgprot_val(SEGMENT_RO);
1338 	return pgprot_val(SEGMENT_RW);
1339 }
1340 
1341 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1342 {
1343 	pmd_val(pmd) &= _SEGMENT_CHG_MASK;
1344 	pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1345 	return pmd;
1346 }
1347 
1348 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1349 {
1350 	pmd_t __pmd;
1351 	pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1352 	return __pmd;
1353 }
1354 
1355 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1356 {
1357 	/* Do not clobber _HPAGE_TYPE_NONE pages! */
1358 	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_INV))
1359 		pmd_val(pmd) &= ~_SEGMENT_ENTRY_RO;
1360 	return pmd;
1361 }
1362 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1363 
1364 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1365 
1366 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1367 extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1368 				       pgtable_t pgtable);
1369 
1370 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1371 extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1372 
1373 static inline int pmd_trans_splitting(pmd_t pmd)
1374 {
1375 	return pmd_val(pmd) & _SEGMENT_ENTRY_SPLIT;
1376 }
1377 
1378 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1379 			      pmd_t *pmdp, pmd_t entry)
1380 {
1381 	if (!(pmd_val(entry) & _SEGMENT_ENTRY_INV) && MACHINE_HAS_EDAT1)
1382 		pmd_val(entry) |= _SEGMENT_ENTRY_CO;
1383 	*pmdp = entry;
1384 }
1385 
1386 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1387 {
1388 	pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1389 	return pmd;
1390 }
1391 
1392 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1393 {
1394 	pmd_val(pmd) |= _SEGMENT_ENTRY_RO;
1395 	return pmd;
1396 }
1397 
1398 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1399 {
1400 	/* No dirty bit in the segment table entry. */
1401 	return pmd;
1402 }
1403 
1404 static inline pmd_t pmd_mkold(pmd_t pmd)
1405 {
1406 	/* No referenced bit in the segment table entry. */
1407 	return pmd;
1408 }
1409 
1410 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1411 {
1412 	/* No referenced bit in the segment table entry. */
1413 	return pmd;
1414 }
1415 
1416 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1417 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1418 					    unsigned long address, pmd_t *pmdp)
1419 {
1420 	unsigned long pmd_addr = pmd_val(*pmdp) & HPAGE_MASK;
1421 	long tmp, rc;
1422 	int counter;
1423 
1424 	rc = 0;
1425 	if (MACHINE_HAS_RRBM) {
1426 		counter = PTRS_PER_PTE >> 6;
1427 		asm volatile(
1428 			"0:	.insn	rre,0xb9ae0000,%0,%3\n"	/* rrbm */
1429 			"	ogr	%1,%0\n"
1430 			"	la	%3,0(%4,%3)\n"
1431 			"	brct	%2,0b\n"
1432 			: "=&d" (tmp), "+&d" (rc), "+d" (counter),
1433 			  "+a" (pmd_addr)
1434 			: "a" (64 * 4096UL) : "cc");
1435 		rc = !!rc;
1436 	} else {
1437 		counter = PTRS_PER_PTE;
1438 		asm volatile(
1439 			"0:	rrbe	0,%2\n"
1440 			"	la	%2,0(%3,%2)\n"
1441 			"	brc	12,1f\n"
1442 			"	lhi	%0,1\n"
1443 			"1:	brct	%1,0b\n"
1444 			: "+d" (rc), "+d" (counter), "+a" (pmd_addr)
1445 			: "a" (4096UL) : "cc");
1446 	}
1447 	return rc;
1448 }
1449 
1450 #define __HAVE_ARCH_PMDP_GET_AND_CLEAR
1451 static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
1452 				       unsigned long address, pmd_t *pmdp)
1453 {
1454 	pmd_t pmd = *pmdp;
1455 
1456 	__pmd_idte(address, pmdp);
1457 	pmd_clear(pmdp);
1458 	return pmd;
1459 }
1460 
1461 #define __HAVE_ARCH_PMDP_CLEAR_FLUSH
1462 static inline pmd_t pmdp_clear_flush(struct vm_area_struct *vma,
1463 				     unsigned long address, pmd_t *pmdp)
1464 {
1465 	return pmdp_get_and_clear(vma->vm_mm, address, pmdp);
1466 }
1467 
1468 #define __HAVE_ARCH_PMDP_INVALIDATE
1469 static inline void pmdp_invalidate(struct vm_area_struct *vma,
1470 				   unsigned long address, pmd_t *pmdp)
1471 {
1472 	__pmd_idte(address, pmdp);
1473 }
1474 
1475 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1476 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1477 				      unsigned long address, pmd_t *pmdp)
1478 {
1479 	pmd_t pmd = *pmdp;
1480 
1481 	if (pmd_write(pmd)) {
1482 		__pmd_idte(address, pmdp);
1483 		set_pmd_at(mm, address, pmdp, pmd_wrprotect(pmd));
1484 	}
1485 }
1486 
1487 #define pfn_pmd(pfn, pgprot)	mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1488 #define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
1489 
1490 static inline int pmd_trans_huge(pmd_t pmd)
1491 {
1492 	return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1493 }
1494 
1495 static inline int has_transparent_hugepage(void)
1496 {
1497 	return MACHINE_HAS_HPAGE ? 1 : 0;
1498 }
1499 
1500 static inline unsigned long pmd_pfn(pmd_t pmd)
1501 {
1502 	return pmd_val(pmd) >> PAGE_SHIFT;
1503 }
1504 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1505 
1506 /*
1507  * 31 bit swap entry format:
1508  * A page-table entry has some bits we have to treat in a special way.
1509  * Bits 0, 20 and bit 23 have to be zero, otherwise an specification
1510  * exception will occur instead of a page translation exception. The
1511  * specifiation exception has the bad habit not to store necessary
1512  * information in the lowcore.
1513  * Bit 21 and bit 22 are the page invalid bit and the page protection
1514  * bit. We set both to indicate a swapped page.
1515  * Bit 30 and 31 are used to distinguish the different page types. For
1516  * a swapped page these bits need to be zero.
1517  * This leaves the bits 1-19 and bits 24-29 to store type and offset.
1518  * We use the 5 bits from 25-29 for the type and the 20 bits from 1-19
1519  * plus 24 for the offset.
1520  * 0|     offset        |0110|o|type |00|
1521  * 0 0000000001111111111 2222 2 22222 33
1522  * 0 1234567890123456789 0123 4 56789 01
1523  *
1524  * 64 bit swap entry format:
1525  * A page-table entry has some bits we have to treat in a special way.
1526  * Bits 52 and bit 55 have to be zero, otherwise an specification
1527  * exception will occur instead of a page translation exception. The
1528  * specifiation exception has the bad habit not to store necessary
1529  * information in the lowcore.
1530  * Bit 53 and bit 54 are the page invalid bit and the page protection
1531  * bit. We set both to indicate a swapped page.
1532  * Bit 62 and 63 are used to distinguish the different page types. For
1533  * a swapped page these bits need to be zero.
1534  * This leaves the bits 0-51 and bits 56-61 to store type and offset.
1535  * We use the 5 bits from 57-61 for the type and the 53 bits from 0-51
1536  * plus 56 for the offset.
1537  * |                      offset                        |0110|o|type |00|
1538  *  0000000000111111111122222222223333333333444444444455 5555 5 55566 66
1539  *  0123456789012345678901234567890123456789012345678901 2345 6 78901 23
1540  */
1541 #ifndef CONFIG_64BIT
1542 #define __SWP_OFFSET_MASK (~0UL >> 12)
1543 #else
1544 #define __SWP_OFFSET_MASK (~0UL >> 11)
1545 #endif
1546 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1547 {
1548 	pte_t pte;
1549 	offset &= __SWP_OFFSET_MASK;
1550 	pte_val(pte) = _PAGE_TYPE_SWAP | ((type & 0x1f) << 2) |
1551 		((offset & 1UL) << 7) | ((offset & ~1UL) << 11);
1552 	return pte;
1553 }
1554 
1555 #define __swp_type(entry)	(((entry).val >> 2) & 0x1f)
1556 #define __swp_offset(entry)	(((entry).val >> 11) | (((entry).val >> 7) & 1))
1557 #define __swp_entry(type,offset) ((swp_entry_t) { pte_val(mk_swap_pte((type),(offset))) })
1558 
1559 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
1560 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
1561 
1562 #ifndef CONFIG_64BIT
1563 # define PTE_FILE_MAX_BITS	26
1564 #else /* CONFIG_64BIT */
1565 # define PTE_FILE_MAX_BITS	59
1566 #endif /* CONFIG_64BIT */
1567 
1568 #define pte_to_pgoff(__pte) \
1569 	((((__pte).pte >> 12) << 7) + (((__pte).pte >> 1) & 0x7f))
1570 
1571 #define pgoff_to_pte(__off) \
1572 	((pte_t) { ((((__off) & 0x7f) << 1) + (((__off) >> 7) << 12)) \
1573 		   | _PAGE_TYPE_FILE })
1574 
1575 #endif /* !__ASSEMBLY__ */
1576 
1577 #define kern_addr_valid(addr)   (1)
1578 
1579 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1580 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1581 extern int s390_enable_sie(void);
1582 
1583 /*
1584  * No page table caches to initialise
1585  */
1586 static inline void pgtable_cache_init(void) { }
1587 static inline void check_pgt_cache(void) { }
1588 
1589 #include <asm-generic/pgtable.h>
1590 
1591 #endif /* _S390_PAGE_H */
1592