1 /* SPDX-License-Identifier: GPL-2.0 */
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)
9 * Derived from "include/asm-i386/pgtable.h"
12 #ifndef _ASM_S390_PGTABLE_H
13 #define _ASM_S390_PGTABLE_H
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>
24 extern pgd_t swapper_pg_dir[];
25 extern void paging_init(void);
34 extern atomic_long_t direct_pages_count[PG_DIRECT_MAP_MAX];
36 static inline void update_page_count(int level, long count)
38 if (IS_ENABLED(CONFIG_PROC_FS))
39 atomic_long_add(count, &direct_pages_count[level]);
43 void arch_report_meminfo(struct seq_file *m);
46 * The S390 doesn't have any external MMU info: the kernel page
47 * tables contain all the necessary information.
49 #define update_mmu_cache(vma, address, ptep) do { } while (0)
50 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
53 * ZERO_PAGE is a global shared page that is always zero; used
54 * for zero-mapped memory areas etc..
57 extern unsigned long empty_zero_page;
58 extern unsigned long zero_page_mask;
60 #define ZERO_PAGE(vaddr) \
61 (virt_to_page((void *)(empty_zero_page + \
62 (((unsigned long)(vaddr)) &zero_page_mask))))
63 #define __HAVE_COLOR_ZERO_PAGE
65 /* TODO: s390 cannot support io_remap_pfn_range... */
67 #define FIRST_USER_ADDRESS 0UL
69 #define pte_ERROR(e) \
70 printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
71 #define pmd_ERROR(e) \
72 printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
73 #define pud_ERROR(e) \
74 printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
75 #define p4d_ERROR(e) \
76 printk("%s:%d: bad p4d %p.\n", __FILE__, __LINE__, (void *) p4d_val(e))
77 #define pgd_ERROR(e) \
78 printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
81 * The vmalloc and module area will always be on the topmost area of the
82 * kernel mapping. We reserve 128GB (64bit) for vmalloc and modules.
83 * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
84 * modules will reside. That makes sure that inter module branches always
85 * happen without trampolines and in addition the placement within a 2GB frame
86 * is branch prediction unit friendly.
88 extern unsigned long VMALLOC_START;
89 extern unsigned long VMALLOC_END;
90 #define VMALLOC_DEFAULT_SIZE ((128UL << 30) - MODULES_LEN)
91 extern struct page *vmemmap;
93 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
95 extern unsigned long MODULES_VADDR;
96 extern unsigned long MODULES_END;
97 #define MODULES_VADDR MODULES_VADDR
98 #define MODULES_END MODULES_END
99 #define MODULES_LEN (1UL << 31)
101 static inline int is_module_addr(void *addr)
103 BUILD_BUG_ON(MODULES_LEN > (1UL << 31));
104 if (addr < (void *)MODULES_VADDR)
106 if (addr > (void *)MODULES_END)
112 * A 64 bit pagetable entry of S390 has following format:
114 * 0000000000111111111122222222223333333333444444444455555555556666
115 * 0123456789012345678901234567890123456789012345678901234567890123
117 * I Page-Invalid Bit: Page is not available for address-translation
118 * P Page-Protection Bit: Store access not possible for page
119 * C Change-bit override: HW is not required to set change bit
121 * A 64 bit segmenttable entry of S390 has following format:
122 * | P-table origin | TT
123 * 0000000000111111111122222222223333333333444444444455555555556666
124 * 0123456789012345678901234567890123456789012345678901234567890123
126 * I Segment-Invalid Bit: Segment is not available for address-translation
127 * C Common-Segment Bit: Segment is not private (PoP 3-30)
128 * P Page-Protection Bit: Store access not possible for page
131 * A 64 bit region table entry of S390 has following format:
132 * | S-table origin | TF TTTL
133 * 0000000000111111111122222222223333333333444444444455555555556666
134 * 0123456789012345678901234567890123456789012345678901234567890123
136 * I Segment-Invalid Bit: Segment is not available for address-translation
141 * The 64 bit regiontable origin of S390 has following format:
142 * | region table origon | DTTL
143 * 0000000000111111111122222222223333333333444444444455555555556666
144 * 0123456789012345678901234567890123456789012345678901234567890123
146 * X Space-Switch event:
147 * G Segment-Invalid Bit:
148 * P Private-Space Bit:
149 * S Storage-Alteration:
153 * A storage key has the following format:
157 * F : fetch protection bit
162 /* Hardware bits in the page table entry */
163 #define _PAGE_NOEXEC 0x100 /* HW no-execute bit */
164 #define _PAGE_PROTECT 0x200 /* HW read-only bit */
165 #define _PAGE_INVALID 0x400 /* HW invalid bit */
166 #define _PAGE_LARGE 0x800 /* Bit to mark a large pte */
168 /* Software bits in the page table entry */
169 #define _PAGE_PRESENT 0x001 /* SW pte present bit */
170 #define _PAGE_YOUNG 0x004 /* SW pte young bit */
171 #define _PAGE_DIRTY 0x008 /* SW pte dirty bit */
172 #define _PAGE_READ 0x010 /* SW pte read bit */
173 #define _PAGE_WRITE 0x020 /* SW pte write bit */
174 #define _PAGE_SPECIAL 0x040 /* SW associated with special page */
175 #define _PAGE_UNUSED 0x080 /* SW bit for pgste usage state */
177 #ifdef CONFIG_MEM_SOFT_DIRTY
178 #define _PAGE_SOFT_DIRTY 0x002 /* SW pte soft dirty bit */
180 #define _PAGE_SOFT_DIRTY 0x000
183 /* Set of bits not changed in pte_modify */
184 #define _PAGE_CHG_MASK (PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
185 _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
188 * handle_pte_fault uses pte_present and pte_none to find out the pte type
189 * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
190 * distinguish present from not-present ptes. It is changed only with the page
193 * The following table gives the different possible bit combinations for
194 * the pte hardware and software bits in the last 12 bits of a pte
195 * (. unassigned bit, x don't care, t swap type):
203 * prot-none, clean, old .11.xx0000.1
204 * prot-none, clean, young .11.xx0001.1
205 * prot-none, dirty, old .11.xx0010.1
206 * prot-none, dirty, young .11.xx0011.1
207 * read-only, clean, old .11.xx0100.1
208 * read-only, clean, young .01.xx0101.1
209 * read-only, dirty, old .11.xx0110.1
210 * read-only, dirty, young .01.xx0111.1
211 * read-write, clean, old .11.xx1100.1
212 * read-write, clean, young .01.xx1101.1
213 * read-write, dirty, old .10.xx1110.1
214 * read-write, dirty, young .00.xx1111.1
215 * HW-bits: R read-only, I invalid
216 * SW-bits: p present, y young, d dirty, r read, w write, s special,
219 * pte_none is true for the bit pattern .10.00000000, pte == 0x400
220 * pte_swap is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
221 * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
224 /* Bits in the segment/region table address-space-control-element */
225 #define _ASCE_ORIGIN ~0xfffUL/* region/segment table origin */
226 #define _ASCE_PRIVATE_SPACE 0x100 /* private space control */
227 #define _ASCE_ALT_EVENT 0x80 /* storage alteration event control */
228 #define _ASCE_SPACE_SWITCH 0x40 /* space switch event */
229 #define _ASCE_REAL_SPACE 0x20 /* real space control */
230 #define _ASCE_TYPE_MASK 0x0c /* asce table type mask */
231 #define _ASCE_TYPE_REGION1 0x0c /* region first table type */
232 #define _ASCE_TYPE_REGION2 0x08 /* region second table type */
233 #define _ASCE_TYPE_REGION3 0x04 /* region third table type */
234 #define _ASCE_TYPE_SEGMENT 0x00 /* segment table type */
235 #define _ASCE_TABLE_LENGTH 0x03 /* region table length */
237 /* Bits in the region table entry */
238 #define _REGION_ENTRY_ORIGIN ~0xfffUL/* region/segment table origin */
239 #define _REGION_ENTRY_PROTECT 0x200 /* region protection bit */
240 #define _REGION_ENTRY_NOEXEC 0x100 /* region no-execute bit */
241 #define _REGION_ENTRY_OFFSET 0xc0 /* region table offset */
242 #define _REGION_ENTRY_INVALID 0x20 /* invalid region table entry */
243 #define _REGION_ENTRY_TYPE_MASK 0x0c /* region table type mask */
244 #define _REGION_ENTRY_TYPE_R1 0x0c /* region first table type */
245 #define _REGION_ENTRY_TYPE_R2 0x08 /* region second table type */
246 #define _REGION_ENTRY_TYPE_R3 0x04 /* region third table type */
247 #define _REGION_ENTRY_LENGTH 0x03 /* region third length */
249 #define _REGION1_ENTRY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
250 #define _REGION1_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
251 #define _REGION2_ENTRY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
252 #define _REGION2_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
253 #define _REGION3_ENTRY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
254 #define _REGION3_ENTRY_EMPTY (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
256 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address */
257 #define _REGION3_ENTRY_DIRTY 0x2000 /* SW region dirty bit */
258 #define _REGION3_ENTRY_YOUNG 0x1000 /* SW region young bit */
259 #define _REGION3_ENTRY_LARGE 0x0400 /* RTTE-format control, large page */
260 #define _REGION3_ENTRY_READ 0x0002 /* SW region read bit */
261 #define _REGION3_ENTRY_WRITE 0x0001 /* SW region write bit */
263 #ifdef CONFIG_MEM_SOFT_DIRTY
264 #define _REGION3_ENTRY_SOFT_DIRTY 0x4000 /* SW region soft dirty bit */
266 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
269 #define _REGION_ENTRY_BITS 0xfffffffffffff22fUL
271 /* Bits in the segment table entry */
272 #define _SEGMENT_ENTRY_BITS 0xfffffffffffffe33UL
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 */
282 #define _SEGMENT_ENTRY (0)
283 #define _SEGMENT_ENTRY_EMPTY (_SEGMENT_ENTRY_INVALID)
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 */
291 #ifdef CONFIG_MEM_SOFT_DIRTY
292 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x4000 /* SW segment soft dirty bit */
294 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
297 #define _CRST_ENTRIES 2048 /* number of region/segment table entries */
298 #define _PAGE_ENTRIES 256 /* number of page table entries */
300 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
301 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
303 #define _REGION1_SHIFT 53
304 #define _REGION2_SHIFT 42
305 #define _REGION3_SHIFT 31
306 #define _SEGMENT_SHIFT 20
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)
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)
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))
324 #define PMD_SHIFT _SEGMENT_SHIFT
325 #define PUD_SHIFT _REGION3_SHIFT
326 #define P4D_SHIFT _REGION2_SHIFT
327 #define PGDIR_SHIFT _REGION1_SHIFT
329 #define PMD_SIZE _SEGMENT_SIZE
330 #define PUD_SIZE _REGION3_SIZE
331 #define P4D_SIZE _REGION2_SIZE
332 #define PGDIR_SIZE _REGION1_SIZE
334 #define PMD_MASK _SEGMENT_MASK
335 #define PUD_MASK _REGION3_MASK
336 #define P4D_MASK _REGION2_MASK
337 #define PGDIR_MASK _REGION1_MASK
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
345 #define MAX_PTRS_PER_P4D PTRS_PER_P4D
348 * Segment table and region3 table entry encoding
349 * (R = read-only, I = invalid, y = young bit):
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
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 */
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
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.
395 #define _ASCE_USER_BITS (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
399 * Page protection definitions.
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)
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)
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.
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
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
445 * Segment entry (large page) protection definitions.
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 | \
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)
480 * Region3 entry (large page) protection definitions.
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)
497 static inline bool mm_p4d_folded(struct mm_struct *mm)
499 return mm->context.asce_limit <= _REGION1_SIZE;
501 #define mm_p4d_folded(mm) mm_p4d_folded(mm)
503 static inline bool mm_pud_folded(struct mm_struct *mm)
505 return mm->context.asce_limit <= _REGION2_SIZE;
507 #define mm_pud_folded(mm) mm_pud_folded(mm)
509 static inline bool mm_pmd_folded(struct mm_struct *mm)
511 return mm->context.asce_limit <= _REGION3_SIZE;
513 #define mm_pmd_folded(mm) mm_pmd_folded(mm)
515 static inline int mm_has_pgste(struct mm_struct *mm)
518 if (unlikely(mm->context.has_pgste))
524 static inline int mm_is_protected(struct mm_struct *mm)
527 if (unlikely(atomic_read(&mm->context.is_protected)))
533 static inline int mm_alloc_pgste(struct mm_struct *mm)
536 if (unlikely(mm->context.alloc_pgste))
543 * In the case that a guest uses storage keys
544 * faults should no longer be backed by zero pages
546 #define mm_forbids_zeropage mm_has_pgste
547 static inline int mm_uses_skeys(struct mm_struct *mm)
550 if (mm->context.uses_skeys)
556 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
558 register unsigned long reg2 asm("2") = old;
559 register unsigned long reg3 asm("3") = new;
560 unsigned long address = (unsigned long)ptr | 1;
564 : "+d" (reg2), "+m" (*ptr)
565 : "d" (reg3), "d" (address)
569 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
571 register unsigned long reg2 asm("2") = old;
572 register unsigned long reg3 asm("3") = new;
573 unsigned long address = (unsigned long)ptr | 1;
576 " .insn rre,0xb98a0000,%0,%3"
577 : "+d" (reg2), "+m" (*ptr)
578 : "d" (reg3), "d" (address)
582 #define CRDTE_DTT_PAGE 0x00UL
583 #define CRDTE_DTT_SEGMENT 0x10UL
584 #define CRDTE_DTT_REGION3 0x14UL
585 #define CRDTE_DTT_REGION2 0x18UL
586 #define CRDTE_DTT_REGION1 0x1cUL
588 static inline void crdte(unsigned long old, unsigned long new,
589 unsigned long table, unsigned long dtt,
590 unsigned long address, unsigned long asce)
592 register unsigned long reg2 asm("2") = old;
593 register unsigned long reg3 asm("3") = new;
594 register unsigned long reg4 asm("4") = table | dtt;
595 register unsigned long reg5 asm("5") = address;
597 asm volatile(".insn rrf,0xb98f0000,%0,%2,%4,0"
599 : "d" (reg3), "d" (reg4), "d" (reg5), "a" (asce)
604 * pgd/p4d/pud/pmd/pte query functions
606 static inline int pgd_folded(pgd_t pgd)
608 return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
611 static inline int pgd_present(pgd_t pgd)
615 return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
618 static inline int pgd_none(pgd_t pgd)
622 return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
625 static inline int pgd_bad(pgd_t pgd)
627 if ((pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1)
629 return (pgd_val(pgd) & ~_REGION_ENTRY_BITS) != 0;
632 static inline unsigned long pgd_pfn(pgd_t pgd)
634 unsigned long origin_mask;
636 origin_mask = _REGION_ENTRY_ORIGIN;
637 return (pgd_val(pgd) & origin_mask) >> PAGE_SHIFT;
640 static inline int p4d_folded(p4d_t p4d)
642 return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
645 static inline int p4d_present(p4d_t p4d)
649 return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
652 static inline int p4d_none(p4d_t p4d)
656 return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
659 static inline unsigned long p4d_pfn(p4d_t p4d)
661 unsigned long origin_mask;
663 origin_mask = _REGION_ENTRY_ORIGIN;
664 return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
667 static inline int pud_folded(pud_t pud)
669 return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
672 static inline int pud_present(pud_t pud)
676 return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
679 static inline int pud_none(pud_t pud)
683 return pud_val(pud) == _REGION3_ENTRY_EMPTY;
686 #define pud_leaf pud_large
687 static inline int pud_large(pud_t pud)
689 if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
691 return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
694 static inline unsigned long pud_pfn(pud_t pud)
696 unsigned long origin_mask;
698 origin_mask = _REGION_ENTRY_ORIGIN;
700 origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
701 return (pud_val(pud) & origin_mask) >> PAGE_SHIFT;
704 #define pmd_leaf pmd_large
705 static inline int pmd_large(pmd_t pmd)
707 return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
710 static inline int pmd_bad(pmd_t pmd)
712 if ((pmd_val(pmd) & _SEGMENT_ENTRY_TYPE_MASK) > 0 || pmd_large(pmd))
714 return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
717 static inline int pud_bad(pud_t pud)
719 unsigned long type = pud_val(pud) & _REGION_ENTRY_TYPE_MASK;
721 if (type > _REGION_ENTRY_TYPE_R3 || pud_large(pud))
723 if (type < _REGION_ENTRY_TYPE_R3)
725 return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
728 static inline int p4d_bad(p4d_t p4d)
730 unsigned long type = p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK;
732 if (type > _REGION_ENTRY_TYPE_R2)
734 if (type < _REGION_ENTRY_TYPE_R2)
736 return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
739 static inline int pmd_present(pmd_t pmd)
741 return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
744 static inline int pmd_none(pmd_t pmd)
746 return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
749 static inline unsigned long pmd_pfn(pmd_t pmd)
751 unsigned long origin_mask;
753 origin_mask = _SEGMENT_ENTRY_ORIGIN;
755 origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
756 return (pmd_val(pmd) & origin_mask) >> PAGE_SHIFT;
759 #define pmd_write pmd_write
760 static inline int pmd_write(pmd_t pmd)
762 return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
765 #define pud_write pud_write
766 static inline int pud_write(pud_t pud)
768 return (pud_val(pud) & _REGION3_ENTRY_WRITE) != 0;
771 static inline int pmd_dirty(pmd_t pmd)
773 return (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
776 static inline int pmd_young(pmd_t pmd)
778 return (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
781 static inline int pte_present(pte_t pte)
783 /* Bit pattern: (pte & 0x001) == 0x001 */
784 return (pte_val(pte) & _PAGE_PRESENT) != 0;
787 static inline int pte_none(pte_t pte)
789 /* Bit pattern: pte == 0x400 */
790 return pte_val(pte) == _PAGE_INVALID;
793 static inline int pte_swap(pte_t pte)
795 /* Bit pattern: (pte & 0x201) == 0x200 */
796 return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
800 static inline int pte_special(pte_t pte)
802 return (pte_val(pte) & _PAGE_SPECIAL);
805 #define __HAVE_ARCH_PTE_SAME
806 static inline int pte_same(pte_t a, pte_t b)
808 return pte_val(a) == pte_val(b);
811 #ifdef CONFIG_NUMA_BALANCING
812 static inline int pte_protnone(pte_t pte)
814 return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
817 static inline int pmd_protnone(pmd_t pmd)
819 /* pmd_large(pmd) implies pmd_present(pmd) */
820 return pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
824 static inline int pte_soft_dirty(pte_t pte)
826 return pte_val(pte) & _PAGE_SOFT_DIRTY;
828 #define pte_swp_soft_dirty pte_soft_dirty
830 static inline pte_t pte_mksoft_dirty(pte_t pte)
832 pte_val(pte) |= _PAGE_SOFT_DIRTY;
835 #define pte_swp_mksoft_dirty pte_mksoft_dirty
837 static inline pte_t pte_clear_soft_dirty(pte_t pte)
839 pte_val(pte) &= ~_PAGE_SOFT_DIRTY;
842 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
844 static inline int pmd_soft_dirty(pmd_t pmd)
846 return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
849 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
851 pmd_val(pmd) |= _SEGMENT_ENTRY_SOFT_DIRTY;
855 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
857 pmd_val(pmd) &= ~_SEGMENT_ENTRY_SOFT_DIRTY;
862 * query functions pte_write/pte_dirty/pte_young only work if
863 * pte_present() is true. Undefined behaviour if not..
865 static inline int pte_write(pte_t pte)
867 return (pte_val(pte) & _PAGE_WRITE) != 0;
870 static inline int pte_dirty(pte_t pte)
872 return (pte_val(pte) & _PAGE_DIRTY) != 0;
875 static inline int pte_young(pte_t pte)
877 return (pte_val(pte) & _PAGE_YOUNG) != 0;
880 #define __HAVE_ARCH_PTE_UNUSED
881 static inline int pte_unused(pte_t pte)
883 return pte_val(pte) & _PAGE_UNUSED;
887 * pgd/pmd/pte modification functions
890 static inline void pgd_clear(pgd_t *pgd)
892 if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
893 pgd_val(*pgd) = _REGION1_ENTRY_EMPTY;
896 static inline void p4d_clear(p4d_t *p4d)
898 if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
899 p4d_val(*p4d) = _REGION2_ENTRY_EMPTY;
902 static inline void pud_clear(pud_t *pud)
904 if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
905 pud_val(*pud) = _REGION3_ENTRY_EMPTY;
908 static inline void pmd_clear(pmd_t *pmdp)
910 pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
913 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
915 pte_val(*ptep) = _PAGE_INVALID;
919 * The following pte modification functions only work if
920 * pte_present() is true. Undefined behaviour if not..
922 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
924 pte_val(pte) &= _PAGE_CHG_MASK;
925 pte_val(pte) |= pgprot_val(newprot);
927 * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
928 * has the invalid bit set, clear it again for readable, young pages
930 if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
931 pte_val(pte) &= ~_PAGE_INVALID;
933 * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
934 * protection bit set, clear it again for writable, dirty pages
936 if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
937 pte_val(pte) &= ~_PAGE_PROTECT;
941 static inline pte_t pte_wrprotect(pte_t pte)
943 pte_val(pte) &= ~_PAGE_WRITE;
944 pte_val(pte) |= _PAGE_PROTECT;
948 static inline pte_t pte_mkwrite(pte_t pte)
950 pte_val(pte) |= _PAGE_WRITE;
951 if (pte_val(pte) & _PAGE_DIRTY)
952 pte_val(pte) &= ~_PAGE_PROTECT;
956 static inline pte_t pte_mkclean(pte_t pte)
958 pte_val(pte) &= ~_PAGE_DIRTY;
959 pte_val(pte) |= _PAGE_PROTECT;
963 static inline pte_t pte_mkdirty(pte_t pte)
965 pte_val(pte) |= _PAGE_DIRTY | _PAGE_SOFT_DIRTY;
966 if (pte_val(pte) & _PAGE_WRITE)
967 pte_val(pte) &= ~_PAGE_PROTECT;
971 static inline pte_t pte_mkold(pte_t pte)
973 pte_val(pte) &= ~_PAGE_YOUNG;
974 pte_val(pte) |= _PAGE_INVALID;
978 static inline pte_t pte_mkyoung(pte_t pte)
980 pte_val(pte) |= _PAGE_YOUNG;
981 if (pte_val(pte) & _PAGE_READ)
982 pte_val(pte) &= ~_PAGE_INVALID;
986 static inline pte_t pte_mkspecial(pte_t pte)
988 pte_val(pte) |= _PAGE_SPECIAL;
992 #ifdef CONFIG_HUGETLB_PAGE
993 static inline pte_t pte_mkhuge(pte_t pte)
995 pte_val(pte) |= _PAGE_LARGE;
1000 #define IPTE_GLOBAL 0
1001 #define IPTE_LOCAL 1
1003 #define IPTE_NODAT 0x400
1004 #define IPTE_GUEST_ASCE 0x800
1006 static __always_inline void __ptep_ipte(unsigned long address, pte_t *ptep,
1007 unsigned long opt, unsigned long asce,
1010 unsigned long pto = (unsigned long) ptep;
1012 if (__builtin_constant_p(opt) && opt == 0) {
1013 /* Invalidation + TLB flush for the pte */
1015 " .insn rrf,0xb2210000,%[r1],%[r2],0,%[m4]"
1016 : "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
1021 /* Invalidate ptes with options + TLB flush of the ptes */
1022 opt = opt | (asce & _ASCE_ORIGIN);
1024 " .insn rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
1025 : [r2] "+a" (address), [r3] "+a" (opt)
1026 : [r1] "a" (pto), [m4] "i" (local) : "memory");
1029 static __always_inline void __ptep_ipte_range(unsigned long address, int nr,
1030 pte_t *ptep, int local)
1032 unsigned long pto = (unsigned long) ptep;
1034 /* Invalidate a range of ptes + TLB flush of the ptes */
1037 " .insn rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
1038 : [r2] "+a" (address), [r3] "+a" (nr)
1039 : [r1] "a" (pto), [m4] "i" (local) : "memory");
1040 } while (nr != 255);
1044 * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1045 * both clear the TLB for the unmapped pte. The reason is that
1046 * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1047 * to modify an active pte. The sequence is
1048 * 1) ptep_get_and_clear
1050 * 3) flush_tlb_range
1051 * On s390 the tlb needs to get flushed with the modification of the pte
1052 * if the pte is active. The only way how this can be implemented is to
1053 * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1056 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1057 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1059 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1060 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1061 unsigned long addr, pte_t *ptep)
1065 pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1066 return pte_young(pte);
1069 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1070 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1071 unsigned long address, pte_t *ptep)
1073 return ptep_test_and_clear_young(vma, address, ptep);
1076 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1077 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1078 unsigned long addr, pte_t *ptep)
1082 res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1083 if (mm_is_protected(mm) && pte_present(res))
1084 uv_convert_from_secure(pte_val(res) & PAGE_MASK);
1088 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1089 pte_t ptep_modify_prot_start(struct vm_area_struct *, unsigned long, pte_t *);
1090 void ptep_modify_prot_commit(struct vm_area_struct *, unsigned long,
1091 pte_t *, pte_t, pte_t);
1093 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1094 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1095 unsigned long addr, pte_t *ptep)
1099 res = ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1100 if (mm_is_protected(vma->vm_mm) && pte_present(res))
1101 uv_convert_from_secure(pte_val(res) & PAGE_MASK);
1106 * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1107 * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1108 * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1109 * cannot be accessed while the batched unmap is running. In this case
1110 * full==1 and a simple pte_clear is enough. See tlb.h.
1112 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1113 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1115 pte_t *ptep, int full)
1121 *ptep = __pte(_PAGE_INVALID);
1123 res = ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1125 if (mm_is_protected(mm) && pte_present(res))
1126 uv_convert_from_secure(pte_val(res) & PAGE_MASK);
1130 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1131 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1132 unsigned long addr, pte_t *ptep)
1137 ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1140 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1141 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1142 unsigned long addr, pte_t *ptep,
1143 pte_t entry, int dirty)
1145 if (pte_same(*ptep, entry))
1147 ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1152 * Additional functions to handle KVM guest page tables
1154 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1155 pte_t *ptep, pte_t entry);
1156 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1157 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1158 pte_t *ptep, unsigned long bits);
1159 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1160 pte_t *ptep, int prot, unsigned long bit);
1161 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1162 pte_t *ptep , int reset);
1163 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1164 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1165 pte_t *sptep, pte_t *tptep, pte_t pte);
1166 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1168 bool ptep_test_and_clear_uc(struct mm_struct *mm, unsigned long address,
1170 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1171 unsigned char key, bool nq);
1172 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1173 unsigned char key, unsigned char *oldkey,
1174 bool nq, bool mr, bool mc);
1175 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1176 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1177 unsigned char *key);
1179 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1180 unsigned long bits, unsigned long value);
1181 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1182 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1183 unsigned long *oldpte, unsigned long *oldpgste);
1184 void gmap_pmdp_csp(struct mm_struct *mm, unsigned long vmaddr);
1185 void gmap_pmdp_invalidate(struct mm_struct *mm, unsigned long vmaddr);
1186 void gmap_pmdp_idte_local(struct mm_struct *mm, unsigned long vmaddr);
1187 void gmap_pmdp_idte_global(struct mm_struct *mm, unsigned long vmaddr);
1189 #define pgprot_writecombine pgprot_writecombine
1190 pgprot_t pgprot_writecombine(pgprot_t prot);
1192 #define pgprot_writethrough pgprot_writethrough
1193 pgprot_t pgprot_writethrough(pgprot_t prot);
1196 * Certain architectures need to do special things when PTEs
1197 * within a page table are directly modified. Thus, the following
1198 * hook is made available.
1200 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
1201 pte_t *ptep, pte_t entry)
1203 if (pte_present(entry))
1204 pte_val(entry) &= ~_PAGE_UNUSED;
1205 if (mm_has_pgste(mm))
1206 ptep_set_pte_at(mm, addr, ptep, entry);
1212 * Conversion functions: convert a page and protection to a page entry,
1213 * and a page entry and page directory to the page they refer to.
1215 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1219 pte_val(__pte) = physpage | pgprot_val(pgprot);
1220 if (!MACHINE_HAS_NX)
1221 pte_val(__pte) &= ~_PAGE_NOEXEC;
1222 return pte_mkyoung(__pte);
1225 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1227 unsigned long physpage = page_to_phys(page);
1228 pte_t __pte = mk_pte_phys(physpage, pgprot);
1230 if (pte_write(__pte) && PageDirty(page))
1231 __pte = pte_mkdirty(__pte);
1235 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1236 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1237 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1238 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1240 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1241 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1242 #define p4d_deref(pud) (p4d_val(pud) & _REGION_ENTRY_ORIGIN)
1243 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1246 * The pgd_offset function *always* adds the index for the top-level
1247 * region/segment table. This is done to get a sequence like the
1248 * following to work:
1249 * pgdp = pgd_offset(current->mm, addr);
1250 * pgd = READ_ONCE(*pgdp);
1251 * p4dp = p4d_offset(&pgd, addr);
1253 * The subsequent p4d_offset, pud_offset and pmd_offset functions
1254 * only add an index if they dereferenced the pointer.
1256 static inline pgd_t *pgd_offset_raw(pgd_t *pgd, unsigned long address)
1261 /* Get the first entry of the top level table */
1262 rste = pgd_val(*pgd);
1263 /* Pick up the shift from the table type of the first entry */
1264 shift = ((rste & _REGION_ENTRY_TYPE_MASK) >> 2) * 11 + 20;
1265 return pgd + ((address >> shift) & (PTRS_PER_PGD - 1));
1268 #define pgd_offset(mm, address) pgd_offset_raw(READ_ONCE((mm)->pgd), address)
1270 static inline p4d_t *p4d_offset(pgd_t *pgd, unsigned long address)
1272 if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R1)
1273 return (p4d_t *) pgd_deref(*pgd) + p4d_index(address);
1274 return (p4d_t *) pgd;
1277 static inline pud_t *pud_offset(p4d_t *p4d, unsigned long address)
1279 if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R2)
1280 return (pud_t *) p4d_deref(*p4d) + pud_index(address);
1281 return (pud_t *) p4d;
1283 #define pud_offset pud_offset
1285 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1287 if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) >= _REGION_ENTRY_TYPE_R3)
1288 return (pmd_t *) pud_deref(*pud) + pmd_index(address);
1289 return (pmd_t *) pud;
1291 #define pmd_offset pmd_offset
1293 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
1295 return (unsigned long) pmd_deref(pmd);
1298 static inline bool gup_fast_permitted(unsigned long start, unsigned long end)
1300 return end <= current->mm->context.asce_limit;
1302 #define gup_fast_permitted gup_fast_permitted
1304 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1305 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1306 #define pte_page(x) pfn_to_page(pte_pfn(x))
1308 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1309 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1310 #define p4d_page(p4d) pfn_to_page(p4d_pfn(p4d))
1311 #define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd))
1313 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1315 pmd_val(pmd) &= ~_SEGMENT_ENTRY_WRITE;
1316 pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1320 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1322 pmd_val(pmd) |= _SEGMENT_ENTRY_WRITE;
1323 if (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY)
1324 pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1328 static inline pmd_t pmd_mkclean(pmd_t pmd)
1330 pmd_val(pmd) &= ~_SEGMENT_ENTRY_DIRTY;
1331 pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1335 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1337 pmd_val(pmd) |= _SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_SOFT_DIRTY;
1338 if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1339 pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1343 static inline pud_t pud_wrprotect(pud_t pud)
1345 pud_val(pud) &= ~_REGION3_ENTRY_WRITE;
1346 pud_val(pud) |= _REGION_ENTRY_PROTECT;
1350 static inline pud_t pud_mkwrite(pud_t pud)
1352 pud_val(pud) |= _REGION3_ENTRY_WRITE;
1353 if (pud_val(pud) & _REGION3_ENTRY_DIRTY)
1354 pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1358 static inline pud_t pud_mkclean(pud_t pud)
1360 pud_val(pud) &= ~_REGION3_ENTRY_DIRTY;
1361 pud_val(pud) |= _REGION_ENTRY_PROTECT;
1365 static inline pud_t pud_mkdirty(pud_t pud)
1367 pud_val(pud) |= _REGION3_ENTRY_DIRTY | _REGION3_ENTRY_SOFT_DIRTY;
1368 if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1369 pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1373 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1374 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1377 * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1378 * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1380 if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1381 return pgprot_val(SEGMENT_NONE);
1382 if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1383 return pgprot_val(SEGMENT_RO);
1384 if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1385 return pgprot_val(SEGMENT_RX);
1386 if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1387 return pgprot_val(SEGMENT_RW);
1388 return pgprot_val(SEGMENT_RWX);
1391 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1393 pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1394 if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1395 pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1399 static inline pmd_t pmd_mkold(pmd_t pmd)
1401 pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1402 pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1406 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1408 pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN_LARGE |
1409 _SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_YOUNG |
1410 _SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_SOFT_DIRTY;
1411 pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1412 if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1413 pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1414 if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1415 pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1419 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1422 pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1426 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1428 static inline void __pmdp_csp(pmd_t *pmdp)
1430 csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1431 pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1434 #define IDTE_GLOBAL 0
1435 #define IDTE_LOCAL 1
1437 #define IDTE_PTOA 0x0800
1438 #define IDTE_NODAT 0x1000
1439 #define IDTE_GUEST_ASCE 0x2000
1441 static __always_inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1442 unsigned long opt, unsigned long asce,
1447 sto = (unsigned long) pmdp - pmd_index(addr) * sizeof(pmd_t);
1448 if (__builtin_constant_p(opt) && opt == 0) {
1449 /* flush without guest asce */
1451 " .insn rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1453 : [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1457 /* flush with guest asce */
1459 " .insn rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1461 : [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1462 [r3] "a" (asce), [m4] "i" (local)
1467 static __always_inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1468 unsigned long opt, unsigned long asce,
1473 r3o = (unsigned long) pudp - pud_index(addr) * sizeof(pud_t);
1474 r3o |= _ASCE_TYPE_REGION3;
1475 if (__builtin_constant_p(opt) && opt == 0) {
1476 /* flush without guest asce */
1478 " .insn rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1480 : [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1484 /* flush with guest asce */
1486 " .insn rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1488 : [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1489 [r3] "a" (asce), [m4] "i" (local)
1494 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1495 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1496 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1498 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1500 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1501 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1504 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1505 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1507 #define __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1508 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1509 unsigned long addr, pmd_t *pmdp,
1510 pmd_t entry, int dirty)
1512 VM_BUG_ON(addr & ~HPAGE_MASK);
1514 entry = pmd_mkyoung(entry);
1516 entry = pmd_mkdirty(entry);
1517 if (pmd_val(*pmdp) == pmd_val(entry))
1519 pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1523 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1524 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1525 unsigned long addr, pmd_t *pmdp)
1529 pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1530 return pmd_young(pmd);
1533 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1534 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1535 unsigned long addr, pmd_t *pmdp)
1537 VM_BUG_ON(addr & ~HPAGE_MASK);
1538 return pmdp_test_and_clear_young(vma, addr, pmdp);
1541 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1542 pmd_t *pmdp, pmd_t entry)
1544 if (!MACHINE_HAS_NX)
1545 pmd_val(entry) &= ~_SEGMENT_ENTRY_NOEXEC;
1549 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1551 pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1552 pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1553 pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1557 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1558 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1559 unsigned long addr, pmd_t *pmdp)
1561 return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1564 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1565 static inline pmd_t pmdp_huge_get_and_clear_full(struct vm_area_struct *vma,
1567 pmd_t *pmdp, int full)
1571 *pmdp = __pmd(_SEGMENT_ENTRY_EMPTY);
1574 return pmdp_xchg_lazy(vma->vm_mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1577 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1578 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1579 unsigned long addr, pmd_t *pmdp)
1581 return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1584 #define __HAVE_ARCH_PMDP_INVALIDATE
1585 static inline pmd_t pmdp_invalidate(struct vm_area_struct *vma,
1586 unsigned long addr, pmd_t *pmdp)
1588 pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1590 return pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1593 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1594 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1595 unsigned long addr, pmd_t *pmdp)
1600 pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1603 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1604 unsigned long address,
1607 return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1609 #define pmdp_collapse_flush pmdp_collapse_flush
1611 #define pfn_pmd(pfn, pgprot) mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1612 #define mk_pmd(page, pgprot) pfn_pmd(page_to_pfn(page), (pgprot))
1614 static inline int pmd_trans_huge(pmd_t pmd)
1616 return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1619 #define has_transparent_hugepage has_transparent_hugepage
1620 static inline int has_transparent_hugepage(void)
1622 return MACHINE_HAS_EDAT1 ? 1 : 0;
1624 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1627 * 64 bit swap entry format:
1628 * A page-table entry has some bits we have to treat in a special way.
1629 * Bits 52 and bit 55 have to be zero, otherwise a specification
1630 * exception will occur instead of a page translation exception. The
1631 * specification exception has the bad habit not to store necessary
1632 * information in the lowcore.
1633 * Bits 54 and 63 are used to indicate the page type.
1634 * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1635 * This leaves the bits 0-51 and bits 56-62 to store type and offset.
1636 * We use the 5 bits from 57-61 for the type and the 52 bits from 0-51
1638 * | offset |01100|type |00|
1639 * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1640 * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1643 #define __SWP_OFFSET_MASK ((1UL << 52) - 1)
1644 #define __SWP_OFFSET_SHIFT 12
1645 #define __SWP_TYPE_MASK ((1UL << 5) - 1)
1646 #define __SWP_TYPE_SHIFT 2
1648 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1652 pte_val(pte) = _PAGE_INVALID | _PAGE_PROTECT;
1653 pte_val(pte) |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1654 pte_val(pte) |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1658 static inline unsigned long __swp_type(swp_entry_t entry)
1660 return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1663 static inline unsigned long __swp_offset(swp_entry_t entry)
1665 return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1668 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1670 return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1673 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
1674 #define __swp_entry_to_pte(x) ((pte_t) { (x).val })
1676 #define kern_addr_valid(addr) (1)
1678 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1679 extern void vmem_remove_mapping(unsigned long start, unsigned long size);
1680 extern int s390_enable_sie(void);
1681 extern int s390_enable_skey(void);
1682 extern void s390_reset_cmma(struct mm_struct *mm);
1684 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1685 #define HAVE_ARCH_UNMAPPED_AREA
1686 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1688 #endif /* _S390_PAGE_H */