arch/s390/include/asm/pgtable.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 /*
   3  *  S390 version
   4  *    Copyright IBM Corp. 1999, 2000
   5  *    Author(s): Hartmut Penner (hp@de.ibm.com)
   6  *               Ulrich Weigand (weigand@de.ibm.com)
   7  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   8  *
   9  *  Derived from "include/asm-i386/pgtable.h"
  10  */
  11
  12 #ifndef _ASM_S390_PGTABLE_H
  13 #define _ASM_S390_PGTABLE_H
  14
  15 #include <linux/sched.h>
  16 #include <linux/mm_types.h>
  17 #include <linux/page-flags.h>
  18 #include <linux/radix-tree.h>
  19 #include <linux/atomic.h>
  20 #include <asm/bug.h>
  21 #include <asm/page.h>
  22
  23 extern pgd_t swapper_pg_dir[];
  24 extern void paging_init(void);
  25
  26 enum {
  27         PG_DIRECT_MAP_4K = 0,
  28         PG_DIRECT_MAP_1M,
  29         PG_DIRECT_MAP_2G,
  30         PG_DIRECT_MAP_MAX
  31 };
  32
  33 extern atomic_long_t direct_pages_count[PG_DIRECT_MAP_MAX];
  34
  35 static inline void update_page_count(int level, long count)
  36 {
  37         if (IS_ENABLED(CONFIG_PROC_FS))
  38                 atomic_long_add(count, &direct_pages_count[level]);
  39 }
  40
  41 struct seq_file;
  42 void arch_report_meminfo(struct seq_file *m);
  43
  44 /*
  45  * The S390 doesn't have any external MMU info: the kernel page
  46  * tables contain all the necessary information.
  47  */
  48 #define update_mmu_cache(vma, address, ptep)     do { } while (0)
  49 #define update_mmu_cache_pmd(vma, address, ptep) do { } while (0)
  50
  51 /*
  52  * ZERO_PAGE is a global shared page that is always zero; used
  53  * for zero-mapped memory areas etc..
  54  */
  55
  56 extern unsigned long empty_zero_page;
  57 extern unsigned long zero_page_mask;
  58
  59 #define ZERO_PAGE(vaddr) \
  60         (virt_to_page((void *)(empty_zero_page + \
  61          (((unsigned long)(vaddr)) &zero_page_mask))))
  62 #define __HAVE_COLOR_ZERO_PAGE
  63
  64 /* TODO: s390 cannot support io_remap_pfn_range... */
  65
  66 #define FIRST_USER_ADDRESS  0UL
  67
  68 #define pte_ERROR(e) \
  69         printk("%s:%d: bad pte %p.\n", __FILE__, __LINE__, (void *) pte_val(e))
  70 #define pmd_ERROR(e) \
  71         printk("%s:%d: bad pmd %p.\n", __FILE__, __LINE__, (void *) pmd_val(e))
  72 #define pud_ERROR(e) \
  73         printk("%s:%d: bad pud %p.\n", __FILE__, __LINE__, (void *) pud_val(e))
  74 #define p4d_ERROR(e) \
  75         printk("%s:%d: bad p4d %p.\n", __FILE__, __LINE__, (void *) p4d_val(e))
  76 #define pgd_ERROR(e) \
  77         printk("%s:%d: bad pgd %p.\n", __FILE__, __LINE__, (void *) pgd_val(e))
  78
  79 /*
  80  * The vmalloc and module area will always be on the topmost area of the
  81  * kernel mapping. We reserve 128GB (64bit) for vmalloc and modules.
  82  * On 64 bit kernels we have a 2GB area at the top of the vmalloc area where
  83  * modules will reside. That makes sure that inter module branches always
  84  * happen without trampolines and in addition the placement within a 2GB frame
  85  * is branch prediction unit friendly.
  86  */
  87 extern unsigned long VMALLOC_START;
  88 extern unsigned long VMALLOC_END;
  89 extern struct page *vmemmap;
  90
  91 #define VMEM_MAX_PHYS ((unsigned long) vmemmap)
  92
  93 extern unsigned long MODULES_VADDR;
  94 extern unsigned long MODULES_END;
  95 #define MODULES_VADDR   MODULES_VADDR
  96 #define MODULES_END     MODULES_END
  97 #define MODULES_LEN     (1UL << 31)
  98
  99 static inline int is_module_addr(void *addr)
 100 {
 101         BUILD_BUG_ON(MODULES_LEN > (1UL << 31));
 102         if (addr < (void *)MODULES_VADDR)
 103                 return 0;
 104         if (addr > (void *)MODULES_END)
 105                 return 0;
 106         return 1;
 107 }
 108
 109 /*
 110  * A 64 bit pagetable entry of S390 has following format:
 111  * |                     PFRA                         |0IPC|  OS  |
 112  * 0000000000111111111122222222223333333333444444444455555555556666
 113  * 0123456789012345678901234567890123456789012345678901234567890123
 114  *
 115  * I Page-Invalid Bit:    Page is not available for address-translation
 116  * P Page-Protection Bit: Store access not possible for page
 117  * C Change-bit override: HW is not required to set change bit
 118  *
 119  * A 64 bit segmenttable entry of S390 has following format:
 120  * |        P-table origin                              |      TT
 121  * 0000000000111111111122222222223333333333444444444455555555556666
 122  * 0123456789012345678901234567890123456789012345678901234567890123
 123  *
 124  * I Segment-Invalid Bit:    Segment is not available for address-translation
 125  * C Common-Segment Bit:     Segment is not private (PoP 3-30)
 126  * P Page-Protection Bit: Store access not possible for page
 127  * TT Type 00
 128  *
 129  * A 64 bit region table entry of S390 has following format:
 130  * |        S-table origin                             |   TF  TTTL
 131  * 0000000000111111111122222222223333333333444444444455555555556666
 132  * 0123456789012345678901234567890123456789012345678901234567890123
 133  *
 134  * I Segment-Invalid Bit:    Segment is not available for address-translation
 135  * TT Type 01
 136  * TF
 137  * TL Table length
 138  *
 139  * The 64 bit regiontable origin of S390 has following format:
 140  * |      region table origon                          |       DTTL
 141  * 0000000000111111111122222222223333333333444444444455555555556666
 142  * 0123456789012345678901234567890123456789012345678901234567890123
 143  *
 144  * X Space-Switch event:
 145  * G Segment-Invalid Bit:
 146  * P Private-Space Bit:
 147  * S Storage-Alteration:
 148  * R Real space
 149  * TL Table-Length:
 150  *
 151  * A storage key has the following format:
 152  * | ACC |F|R|C|0|
 153  *  0   3 4 5 6 7
 154  * ACC: access key
 155  * F  : fetch protection bit
 156  * R  : referenced bit
 157  * C  : changed bit
 158  */
 159
 160 /* Hardware bits in the page table entry */
 161 #define _PAGE_NOEXEC    0x100           /* HW no-execute bit  */
 162 #define _PAGE_PROTECT   0x200           /* HW read-only bit  */
 163 #define _PAGE_INVALID   0x400           /* HW invalid bit    */
 164 #define _PAGE_LARGE     0x800           /* Bit to mark a large pte */
 165
 166 /* Software bits in the page table entry */
 167 #define _PAGE_PRESENT   0x001           /* SW pte present bit */
 168 #define _PAGE_YOUNG     0x004           /* SW pte young bit */
 169 #define _PAGE_DIRTY     0x008           /* SW pte dirty bit */
 170 #define _PAGE_READ      0x010           /* SW pte read bit */
 171 #define _PAGE_WRITE     0x020           /* SW pte write bit */
 172 #define _PAGE_SPECIAL   0x040           /* SW associated with special page */
 173 #define _PAGE_UNUSED    0x080           /* SW bit for pgste usage state */
 174 #define __HAVE_ARCH_PTE_SPECIAL
 175
 176 #ifdef CONFIG_MEM_SOFT_DIRTY
 177 #define _PAGE_SOFT_DIRTY 0x002          /* SW pte soft dirty bit */
 178 #else
 179 #define _PAGE_SOFT_DIRTY 0x000
 180 #endif
 181
 182 /* Set of bits not changed in pte_modify */
 183 #define _PAGE_CHG_MASK          (PAGE_MASK | _PAGE_SPECIAL | _PAGE_DIRTY | \
 184                                  _PAGE_YOUNG | _PAGE_SOFT_DIRTY)
 185
 186 /*
 187  * handle_pte_fault uses pte_present and pte_none to find out the pte type
 188  * WITHOUT holding the page table lock. The _PAGE_PRESENT bit is used to
 189  * distinguish present from not-present ptes. It is changed only with the page
 190  * table lock held.
 191  *
 192  * The following table gives the different possible bit combinations for
 193  * the pte hardware and software bits in the last 12 bits of a pte
 194  * (. unassigned bit, x don't care, t swap type):
 195  *
 196  *                              842100000000
 197  *                              000084210000
 198  *                              000000008421
 199  *                              .IR.uswrdy.p
 200  * empty                        .10.00000000
 201  * swap                         .11..ttttt.0
 202  * prot-none, clean, old        .11.xx0000.1
 203  * prot-none, clean, young      .11.xx0001.1
 204  * prot-none, dirty, old        .11.xx0010.1
 205  * prot-none, dirty, young      .11.xx0011.1
 206  * read-only, clean, old        .11.xx0100.1
 207  * read-only, clean, young      .01.xx0101.1
 208  * read-only, dirty, old        .11.xx0110.1
 209  * read-only, dirty, young      .01.xx0111.1
 210  * read-write, clean, old       .11.xx1100.1
 211  * read-write, clean, young     .01.xx1101.1
 212  * read-write, dirty, old       .10.xx1110.1
 213  * read-write, dirty, young     .00.xx1111.1
 214  * HW-bits: R read-only, I invalid
 215  * SW-bits: p present, y young, d dirty, r read, w write, s special,
 216  *          u unused, l large
 217  *
 218  * pte_none    is true for the bit pattern .10.00000000, pte == 0x400
 219  * pte_swap    is true for the bit pattern .11..ooooo.0, (pte & 0x201) == 0x200
 220  * pte_present is true for the bit pattern .xx.xxxxxx.1, (pte & 0x001) == 0x001
 221  */
 222
 223 /* Bits in the segment/region table address-space-control-element */
 224 #define _ASCE_ORIGIN            ~0xfffUL/* region/segment table origin      */
 225 #define _ASCE_PRIVATE_SPACE     0x100   /* private space control            */
 226 #define _ASCE_ALT_EVENT         0x80    /* storage alteration event control */
 227 #define _ASCE_SPACE_SWITCH      0x40    /* space switch event               */
 228 #define _ASCE_REAL_SPACE        0x20    /* real space control               */
 229 #define _ASCE_TYPE_MASK         0x0c    /* asce table type mask             */
 230 #define _ASCE_TYPE_REGION1      0x0c    /* region first table type          */
 231 #define _ASCE_TYPE_REGION2      0x08    /* region second table type         */
 232 #define _ASCE_TYPE_REGION3      0x04    /* region third table type          */
 233 #define _ASCE_TYPE_SEGMENT      0x00    /* segment table type               */
 234 #define _ASCE_TABLE_LENGTH      0x03    /* region table length              */
 235
 236 /* Bits in the region table entry */
 237 #define _REGION_ENTRY_ORIGIN    ~0xfffUL/* region/segment table origin      */
 238 #define _REGION_ENTRY_PROTECT   0x200   /* region protection bit            */
 239 #define _REGION_ENTRY_NOEXEC    0x100   /* region no-execute bit            */
 240 #define _REGION_ENTRY_OFFSET    0xc0    /* region table offset              */
 241 #define _REGION_ENTRY_INVALID   0x20    /* invalid region table entry       */
 242 #define _REGION_ENTRY_TYPE_MASK 0x0c    /* region/segment table type mask   */
 243 #define _REGION_ENTRY_TYPE_R1   0x0c    /* region first table type          */
 244 #define _REGION_ENTRY_TYPE_R2   0x08    /* region second table type         */
 245 #define _REGION_ENTRY_TYPE_R3   0x04    /* region third table type          */
 246 #define _REGION_ENTRY_LENGTH    0x03    /* region third length              */
 247
 248 #define _REGION1_ENTRY          (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_LENGTH)
 249 #define _REGION1_ENTRY_EMPTY    (_REGION_ENTRY_TYPE_R1 | _REGION_ENTRY_INVALID)
 250 #define _REGION2_ENTRY          (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_LENGTH)
 251 #define _REGION2_ENTRY_EMPTY    (_REGION_ENTRY_TYPE_R2 | _REGION_ENTRY_INVALID)
 252 #define _REGION3_ENTRY          (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_LENGTH)
 253 #define _REGION3_ENTRY_EMPTY    (_REGION_ENTRY_TYPE_R3 | _REGION_ENTRY_INVALID)
 254
 255 #define _REGION3_ENTRY_ORIGIN_LARGE ~0x7fffffffUL /* large page address      */
 256 #define _REGION3_ENTRY_DIRTY    0x2000  /* SW region dirty bit */
 257 #define _REGION3_ENTRY_YOUNG    0x1000  /* SW region young bit */
 258 #define _REGION3_ENTRY_LARGE    0x0400  /* RTTE-format control, large page  */
 259 #define _REGION3_ENTRY_READ     0x0002  /* SW region read bit */
 260 #define _REGION3_ENTRY_WRITE    0x0001  /* SW region write bit */
 261
 262 #ifdef CONFIG_MEM_SOFT_DIRTY
 263 #define _REGION3_ENTRY_SOFT_DIRTY 0x4000 /* SW region soft dirty bit */
 264 #else
 265 #define _REGION3_ENTRY_SOFT_DIRTY 0x0000 /* SW region soft dirty bit */
 266 #endif
 267
 268 #define _REGION_ENTRY_BITS       0xfffffffffffff22fUL
 269 #define _REGION_ENTRY_BITS_LARGE 0xffffffff8000fe2fUL
 270
 271 /* Bits in the segment table entry */
 272 #define _SEGMENT_ENTRY_BITS     0xfffffffffffffe33UL
 273 #define _SEGMENT_ENTRY_BITS_LARGE 0xfffffffffff0ff33UL
 274 #define _SEGMENT_ENTRY_ORIGIN_LARGE ~0xfffffUL /* large page address        */
 275 #define _SEGMENT_ENTRY_ORIGIN   ~0x7ffUL/* page table origin                */
 276 #define _SEGMENT_ENTRY_PROTECT  0x200   /* segment protection bit           */
 277 #define _SEGMENT_ENTRY_NOEXEC   0x100   /* segment no-execute bit           */
 278 #define _SEGMENT_ENTRY_INVALID  0x20    /* invalid segment table entry      */
 279
 280 #define _SEGMENT_ENTRY          (0)
 281 #define _SEGMENT_ENTRY_EMPTY    (_SEGMENT_ENTRY_INVALID)
 282
 283 #define _SEGMENT_ENTRY_DIRTY    0x2000  /* SW segment dirty bit */
 284 #define _SEGMENT_ENTRY_YOUNG    0x1000  /* SW segment young bit */
 285 #define _SEGMENT_ENTRY_LARGE    0x0400  /* STE-format control, large page */
 286 #define _SEGMENT_ENTRY_WRITE    0x0002  /* SW segment write bit */
 287 #define _SEGMENT_ENTRY_READ     0x0001  /* SW segment read bit */
 288
 289 #ifdef CONFIG_MEM_SOFT_DIRTY
 290 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x4000 /* SW segment soft dirty bit */
 291 #else
 292 #define _SEGMENT_ENTRY_SOFT_DIRTY 0x0000 /* SW segment soft dirty bit */
 293 #endif
 294
 295 #define _CRST_ENTRIES   2048    /* number of region/segment table entries */
 296 #define _PAGE_ENTRIES   256     /* number of page table entries */
 297
 298 #define _CRST_TABLE_SIZE (_CRST_ENTRIES * 8)
 299 #define _PAGE_TABLE_SIZE (_PAGE_ENTRIES * 8)
 300
 301 #define _REGION1_SHIFT  53
 302 #define _REGION2_SHIFT  42
 303 #define _REGION3_SHIFT  31
 304 #define _SEGMENT_SHIFT  20
 305
 306 #define _REGION1_INDEX  (0x7ffUL << _REGION1_SHIFT)
 307 #define _REGION2_INDEX  (0x7ffUL << _REGION2_SHIFT)
 308 #define _REGION3_INDEX  (0x7ffUL << _REGION3_SHIFT)
 309 #define _SEGMENT_INDEX  (0x7ffUL << _SEGMENT_SHIFT)
 310 #define _PAGE_INDEX     (0xffUL  << _PAGE_SHIFT)
 311
 312 #define _REGION1_SIZE   (1UL << _REGION1_SHIFT)
 313 #define _REGION2_SIZE   (1UL << _REGION2_SHIFT)
 314 #define _REGION3_SIZE   (1UL << _REGION3_SHIFT)
 315 #define _SEGMENT_SIZE   (1UL << _SEGMENT_SHIFT)
 316
 317 #define _REGION1_MASK   (~(_REGION1_SIZE - 1))
 318 #define _REGION2_MASK   (~(_REGION2_SIZE - 1))
 319 #define _REGION3_MASK   (~(_REGION3_SIZE - 1))
 320 #define _SEGMENT_MASK   (~(_SEGMENT_SIZE - 1))
 321
 322 #define PMD_SHIFT       _SEGMENT_SHIFT
 323 #define PUD_SHIFT       _REGION3_SHIFT
 324 #define P4D_SHIFT       _REGION2_SHIFT
 325 #define PGDIR_SHIFT     _REGION1_SHIFT
 326
 327 #define PMD_SIZE        _SEGMENT_SIZE
 328 #define PUD_SIZE        _REGION3_SIZE
 329 #define P4D_SIZE        _REGION2_SIZE
 330 #define PGDIR_SIZE      _REGION1_SIZE
 331
 332 #define PMD_MASK        _SEGMENT_MASK
 333 #define PUD_MASK        _REGION3_MASK
 334 #define P4D_MASK        _REGION2_MASK
 335 #define PGDIR_MASK      _REGION1_MASK
 336
 337 #define PTRS_PER_PTE    _PAGE_ENTRIES
 338 #define PTRS_PER_PMD    _CRST_ENTRIES
 339 #define PTRS_PER_PUD    _CRST_ENTRIES
 340 #define PTRS_PER_P4D    _CRST_ENTRIES
 341 #define PTRS_PER_PGD    _CRST_ENTRIES
 342
 343 /*
 344  * Segment table and region3 table entry encoding
 345  * (R = read-only, I = invalid, y = young bit):
 346  *                              dy..R...I...wr
 347  * prot-none, clean, old        00..1...1...00
 348  * prot-none, clean, young      01..1...1...00
 349  * prot-none, dirty, old        10..1...1...00
 350  * prot-none, dirty, young      11..1...1...00
 351  * read-only, clean, old        00..1...1...01
 352  * read-only, clean, young      01..1...0...01
 353  * read-only, dirty, old        10..1...1...01
 354  * read-only, dirty, young      11..1...0...01
 355  * read-write, clean, old       00..1...1...11
 356  * read-write, clean, young     01..1...0...11
 357  * read-write, dirty, old       10..0...1...11
 358  * read-write, dirty, young     11..0...0...11
 359  * The segment table origin is used to distinguish empty (origin==0) from
 360  * read-write, old segment table entries (origin!=0)
 361  * HW-bits: R read-only, I invalid
 362  * SW-bits: y young, d dirty, r read, w write
 363  */
 364
 365 /* Page status table bits for virtualization */
 366 #define PGSTE_ACC_BITS  0xf000000000000000UL
 367 #define PGSTE_FP_BIT    0x0800000000000000UL
 368 #define PGSTE_PCL_BIT   0x0080000000000000UL
 369 #define PGSTE_HR_BIT    0x0040000000000000UL
 370 #define PGSTE_HC_BIT    0x0020000000000000UL
 371 #define PGSTE_GR_BIT    0x0004000000000000UL
 372 #define PGSTE_GC_BIT    0x0002000000000000UL
 373 #define PGSTE_UC_BIT    0x0000800000000000UL    /* user dirty (migration) */
 374 #define PGSTE_IN_BIT    0x0000400000000000UL    /* IPTE notify bit */
 375 #define PGSTE_VSIE_BIT  0x0000200000000000UL    /* ref'd in a shadow table */
 376
 377 /* Guest Page State used for virtualization */
 378 #define _PGSTE_GPS_ZERO                 0x0000000080000000UL
 379 #define _PGSTE_GPS_NODAT                0x0000000040000000UL
 380 #define _PGSTE_GPS_USAGE_MASK           0x0000000003000000UL
 381 #define _PGSTE_GPS_USAGE_STABLE         0x0000000000000000UL
 382 #define _PGSTE_GPS_USAGE_UNUSED         0x0000000001000000UL
 383 #define _PGSTE_GPS_USAGE_POT_VOLATILE   0x0000000002000000UL
 384 #define _PGSTE_GPS_USAGE_VOLATILE       _PGSTE_GPS_USAGE_MASK
 385
 386 /*
 387  * A user page table pointer has the space-switch-event bit, the
 388  * private-space-control bit and the storage-alteration-event-control
 389  * bit set. A kernel page table pointer doesn't need them.
 390  */
 391 #define _ASCE_USER_BITS         (_ASCE_SPACE_SWITCH | _ASCE_PRIVATE_SPACE | \
 392                                  _ASCE_ALT_EVENT)
 393
 394 /*
 395  * Page protection definitions.
 396  */
 397 #define PAGE_NONE       __pgprot(_PAGE_PRESENT | _PAGE_INVALID | _PAGE_PROTECT)
 398 #define PAGE_RO         __pgprot(_PAGE_PRESENT | _PAGE_READ | \
 399                                  _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
 400 #define PAGE_RX         __pgprot(_PAGE_PRESENT | _PAGE_READ | \
 401                                  _PAGE_INVALID | _PAGE_PROTECT)
 402 #define PAGE_RW         __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
 403                                  _PAGE_NOEXEC  | _PAGE_INVALID | _PAGE_PROTECT)
 404 #define PAGE_RWX        __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
 405                                  _PAGE_INVALID | _PAGE_PROTECT)
 406
 407 #define PAGE_SHARED     __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
 408                                  _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
 409 #define PAGE_KERNEL     __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
 410                                  _PAGE_YOUNG | _PAGE_DIRTY | _PAGE_NOEXEC)
 411 #define PAGE_KERNEL_RO  __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_YOUNG | \
 412                                  _PAGE_PROTECT | _PAGE_NOEXEC)
 413 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | \
 414                                   _PAGE_YOUNG | _PAGE_DIRTY)
 415
 416 /*
 417  * On s390 the page table entry has an invalid bit and a read-only bit.
 418  * Read permission implies execute permission and write permission
 419  * implies read permission.
 420  */
 421          /*xwr*/
 422 #define __P000  PAGE_NONE
 423 #define __P001  PAGE_RO
 424 #define __P010  PAGE_RO
 425 #define __P011  PAGE_RO
 426 #define __P100  PAGE_RX
 427 #define __P101  PAGE_RX
 428 #define __P110  PAGE_RX
 429 #define __P111  PAGE_RX
 430
 431 #define __S000  PAGE_NONE
 432 #define __S001  PAGE_RO
 433 #define __S010  PAGE_RW
 434 #define __S011  PAGE_RW
 435 #define __S100  PAGE_RX
 436 #define __S101  PAGE_RX
 437 #define __S110  PAGE_RWX
 438 #define __S111  PAGE_RWX
 439
 440 /*
 441  * Segment entry (large page) protection definitions.
 442  */
 443 #define SEGMENT_NONE    __pgprot(_SEGMENT_ENTRY_INVALID | \
 444                                  _SEGMENT_ENTRY_PROTECT)
 445 #define SEGMENT_RO      __pgprot(_SEGMENT_ENTRY_PROTECT | \
 446                                  _SEGMENT_ENTRY_READ | \
 447                                  _SEGMENT_ENTRY_NOEXEC)
 448 #define SEGMENT_RX      __pgprot(_SEGMENT_ENTRY_PROTECT | \
 449                                  _SEGMENT_ENTRY_READ)
 450 #define SEGMENT_RW      __pgprot(_SEGMENT_ENTRY_READ | \
 451                                  _SEGMENT_ENTRY_WRITE | \
 452                                  _SEGMENT_ENTRY_NOEXEC)
 453 #define SEGMENT_RWX     __pgprot(_SEGMENT_ENTRY_READ | \
 454                                  _SEGMENT_ENTRY_WRITE)
 455 #define SEGMENT_KERNEL  __pgprot(_SEGMENT_ENTRY |       \
 456                                  _SEGMENT_ENTRY_LARGE | \
 457                                  _SEGMENT_ENTRY_READ |  \
 458                                  _SEGMENT_ENTRY_WRITE | \
 459                                  _SEGMENT_ENTRY_YOUNG | \
 460                                  _SEGMENT_ENTRY_DIRTY | \
 461                                  _SEGMENT_ENTRY_NOEXEC)
 462 #define SEGMENT_KERNEL_RO __pgprot(_SEGMENT_ENTRY |     \
 463                                  _SEGMENT_ENTRY_LARGE | \
 464                                  _SEGMENT_ENTRY_READ |  \
 465                                  _SEGMENT_ENTRY_YOUNG | \
 466                                  _SEGMENT_ENTRY_PROTECT | \
 467                                  _SEGMENT_ENTRY_NOEXEC)
 468
 469 /*
 470  * Region3 entry (large page) protection definitions.
 471  */
 472
 473 #define REGION3_KERNEL  __pgprot(_REGION_ENTRY_TYPE_R3 | \
 474                                  _REGION3_ENTRY_LARGE |  \
 475                                  _REGION3_ENTRY_READ |   \
 476                                  _REGION3_ENTRY_WRITE |  \
 477                                  _REGION3_ENTRY_YOUNG |  \
 478                                  _REGION3_ENTRY_DIRTY | \
 479                                  _REGION_ENTRY_NOEXEC)
 480 #define REGION3_KERNEL_RO __pgprot(_REGION_ENTRY_TYPE_R3 | \
 481                                    _REGION3_ENTRY_LARGE |  \
 482                                    _REGION3_ENTRY_READ |   \
 483                                    _REGION3_ENTRY_YOUNG |  \
 484                                    _REGION_ENTRY_PROTECT | \
 485                                    _REGION_ENTRY_NOEXEC)
 486
 487 static inline bool mm_p4d_folded(struct mm_struct *mm)
 488 {
 489         return mm->context.asce_limit <= _REGION1_SIZE;
 490 }
 491 #define mm_p4d_folded(mm) mm_p4d_folded(mm)
 492
 493 static inline bool mm_pud_folded(struct mm_struct *mm)
 494 {
 495         return mm->context.asce_limit <= _REGION2_SIZE;
 496 }
 497 #define mm_pud_folded(mm) mm_pud_folded(mm)
 498
 499 static inline bool mm_pmd_folded(struct mm_struct *mm)
 500 {
 501         return mm->context.asce_limit <= _REGION3_SIZE;
 502 }
 503 #define mm_pmd_folded(mm) mm_pmd_folded(mm)
 504
 505 static inline int mm_has_pgste(struct mm_struct *mm)
 506 {
 507 #ifdef CONFIG_PGSTE
 508         if (unlikely(mm->context.has_pgste))
 509                 return 1;
 510 #endif
 511         return 0;
 512 }
 513
 514 static inline int mm_alloc_pgste(struct mm_struct *mm)
 515 {
 516 #ifdef CONFIG_PGSTE
 517         if (unlikely(mm->context.alloc_pgste))
 518                 return 1;
 519 #endif
 520         return 0;
 521 }
 522
 523 /*
 524  * In the case that a guest uses storage keys
 525  * faults should no longer be backed by zero pages
 526  */
 527 #define mm_forbids_zeropage mm_has_pgste
 528 static inline int mm_use_skey(struct mm_struct *mm)
 529 {
 530 #ifdef CONFIG_PGSTE
 531         if (mm->context.use_skey)
 532                 return 1;
 533 #endif
 534         return 0;
 535 }
 536
 537 static inline void csp(unsigned int *ptr, unsigned int old, unsigned int new)
 538 {
 539         register unsigned long reg2 asm("2") = old;
 540         register unsigned long reg3 asm("3") = new;
 541         unsigned long address = (unsigned long)ptr | 1;
 542
 543         asm volatile(
 544                 "       csp     %0,%3"
 545                 : "+d" (reg2), "+m" (*ptr)
 546                 : "d" (reg3), "d" (address)
 547                 : "cc");
 548 }
 549
 550 static inline void cspg(unsigned long *ptr, unsigned long old, unsigned long new)
 551 {
 552         register unsigned long reg2 asm("2") = old;
 553         register unsigned long reg3 asm("3") = new;
 554         unsigned long address = (unsigned long)ptr | 1;
 555
 556         asm volatile(
 557                 "       .insn   rre,0xb98a0000,%0,%3"
 558                 : "+d" (reg2), "+m" (*ptr)
 559                 : "d" (reg3), "d" (address)
 560                 : "cc");
 561 }
 562
 563 #define CRDTE_DTT_PAGE          0x00UL
 564 #define CRDTE_DTT_SEGMENT       0x10UL
 565 #define CRDTE_DTT_REGION3       0x14UL
 566 #define CRDTE_DTT_REGION2       0x18UL
 567 #define CRDTE_DTT_REGION1       0x1cUL
 568
 569 static inline void crdte(unsigned long old, unsigned long new,
 570                          unsigned long table, unsigned long dtt,
 571                          unsigned long address, unsigned long asce)
 572 {
 573         register unsigned long reg2 asm("2") = old;
 574         register unsigned long reg3 asm("3") = new;
 575         register unsigned long reg4 asm("4") = table | dtt;
 576         register unsigned long reg5 asm("5") = address;
 577
 578         asm volatile(".insn rrf,0xb98f0000,%0,%2,%4,0"
 579                      : "+d" (reg2)
 580                      : "d" (reg3), "d" (reg4), "d" (reg5), "a" (asce)
 581                      : "memory", "cc");
 582 }
 583
 584 /*
 585  * pgd/p4d/pud/pmd/pte query functions
 586  */
 587 static inline int pgd_folded(pgd_t pgd)
 588 {
 589         return (pgd_val(pgd) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R1;
 590 }
 591
 592 static inline int pgd_present(pgd_t pgd)
 593 {
 594         if (pgd_folded(pgd))
 595                 return 1;
 596         return (pgd_val(pgd) & _REGION_ENTRY_ORIGIN) != 0UL;
 597 }
 598
 599 static inline int pgd_none(pgd_t pgd)
 600 {
 601         if (pgd_folded(pgd))
 602                 return 0;
 603         return (pgd_val(pgd) & _REGION_ENTRY_INVALID) != 0UL;
 604 }
 605
 606 static inline int pgd_bad(pgd_t pgd)
 607 {
 608         /*
 609          * With dynamic page table levels the pgd can be a region table
 610          * entry or a segment table entry. Check for the bit that are
 611          * invalid for either table entry.
 612          */
 613         unsigned long mask =
 614                 ~_SEGMENT_ENTRY_ORIGIN & ~_REGION_ENTRY_INVALID &
 615                 ~_REGION_ENTRY_TYPE_MASK & ~_REGION_ENTRY_LENGTH;
 616         return (pgd_val(pgd) & mask) != 0;
 617 }
 618
 619 static inline int p4d_folded(p4d_t p4d)
 620 {
 621         return (p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2;
 622 }
 623
 624 static inline int p4d_present(p4d_t p4d)
 625 {
 626         if (p4d_folded(p4d))
 627                 return 1;
 628         return (p4d_val(p4d) & _REGION_ENTRY_ORIGIN) != 0UL;
 629 }
 630
 631 static inline int p4d_none(p4d_t p4d)
 632 {
 633         if (p4d_folded(p4d))
 634                 return 0;
 635         return p4d_val(p4d) == _REGION2_ENTRY_EMPTY;
 636 }
 637
 638 static inline unsigned long p4d_pfn(p4d_t p4d)
 639 {
 640         unsigned long origin_mask;
 641
 642         origin_mask = _REGION_ENTRY_ORIGIN;
 643         return (p4d_val(p4d) & origin_mask) >> PAGE_SHIFT;
 644 }
 645
 646 static inline int pud_folded(pud_t pud)
 647 {
 648         return (pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3;
 649 }
 650
 651 static inline int pud_present(pud_t pud)
 652 {
 653         if (pud_folded(pud))
 654                 return 1;
 655         return (pud_val(pud) & _REGION_ENTRY_ORIGIN) != 0UL;
 656 }
 657
 658 static inline int pud_none(pud_t pud)
 659 {
 660         if (pud_folded(pud))
 661                 return 0;
 662         return pud_val(pud) == _REGION3_ENTRY_EMPTY;
 663 }
 664
 665 static inline int pud_large(pud_t pud)
 666 {
 667         if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) != _REGION_ENTRY_TYPE_R3)
 668                 return 0;
 669         return !!(pud_val(pud) & _REGION3_ENTRY_LARGE);
 670 }
 671
 672 static inline unsigned long pud_pfn(pud_t pud)
 673 {
 674         unsigned long origin_mask;
 675
 676         origin_mask = _REGION_ENTRY_ORIGIN;
 677         if (pud_large(pud))
 678                 origin_mask = _REGION3_ENTRY_ORIGIN_LARGE;
 679         return (pud_val(pud) & origin_mask) >> PAGE_SHIFT;
 680 }
 681
 682 static inline int pmd_large(pmd_t pmd)
 683 {
 684         return (pmd_val(pmd) & _SEGMENT_ENTRY_LARGE) != 0;
 685 }
 686
 687 static inline int pmd_bad(pmd_t pmd)
 688 {
 689         if (pmd_large(pmd))
 690                 return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS_LARGE) != 0;
 691         return (pmd_val(pmd) & ~_SEGMENT_ENTRY_BITS) != 0;
 692 }
 693
 694 static inline int pud_bad(pud_t pud)
 695 {
 696         if ((pud_val(pud) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R3)
 697                 return pmd_bad(__pmd(pud_val(pud)));
 698         if (pud_large(pud))
 699                 return (pud_val(pud) & ~_REGION_ENTRY_BITS_LARGE) != 0;
 700         return (pud_val(pud) & ~_REGION_ENTRY_BITS) != 0;
 701 }
 702
 703 static inline int p4d_bad(p4d_t p4d)
 704 {
 705         if ((p4d_val(p4d) & _REGION_ENTRY_TYPE_MASK) < _REGION_ENTRY_TYPE_R2)
 706                 return pud_bad(__pud(p4d_val(p4d)));
 707         return (p4d_val(p4d) & ~_REGION_ENTRY_BITS) != 0;
 708 }
 709
 710 static inline int pmd_present(pmd_t pmd)
 711 {
 712         return pmd_val(pmd) != _SEGMENT_ENTRY_EMPTY;
 713 }
 714
 715 static inline int pmd_none(pmd_t pmd)
 716 {
 717         return pmd_val(pmd) == _SEGMENT_ENTRY_EMPTY;
 718 }
 719
 720 static inline unsigned long pmd_pfn(pmd_t pmd)
 721 {
 722         unsigned long origin_mask;
 723
 724         origin_mask = _SEGMENT_ENTRY_ORIGIN;
 725         if (pmd_large(pmd))
 726                 origin_mask = _SEGMENT_ENTRY_ORIGIN_LARGE;
 727         return (pmd_val(pmd) & origin_mask) >> PAGE_SHIFT;
 728 }
 729
 730 #define pmd_write pmd_write
 731 static inline int pmd_write(pmd_t pmd)
 732 {
 733         return (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE) != 0;
 734 }
 735
 736 static inline int pmd_dirty(pmd_t pmd)
 737 {
 738         int dirty = 1;
 739         if (pmd_large(pmd))
 740                 dirty = (pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY) != 0;
 741         return dirty;
 742 }
 743
 744 static inline int pmd_young(pmd_t pmd)
 745 {
 746         int young = 1;
 747         if (pmd_large(pmd))
 748                 young = (pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG) != 0;
 749         return young;
 750 }
 751
 752 static inline int pte_present(pte_t pte)
 753 {
 754         /* Bit pattern: (pte & 0x001) == 0x001 */
 755         return (pte_val(pte) & _PAGE_PRESENT) != 0;
 756 }
 757
 758 static inline int pte_none(pte_t pte)
 759 {
 760         /* Bit pattern: pte == 0x400 */
 761         return pte_val(pte) == _PAGE_INVALID;
 762 }
 763
 764 static inline int pte_swap(pte_t pte)
 765 {
 766         /* Bit pattern: (pte & 0x201) == 0x200 */
 767         return (pte_val(pte) & (_PAGE_PROTECT | _PAGE_PRESENT))
 768                 == _PAGE_PROTECT;
 769 }
 770
 771 static inline int pte_special(pte_t pte)
 772 {
 773         return (pte_val(pte) & _PAGE_SPECIAL);
 774 }
 775
 776 #define __HAVE_ARCH_PTE_SAME
 777 static inline int pte_same(pte_t a, pte_t b)
 778 {
 779         return pte_val(a) == pte_val(b);
 780 }
 781
 782 #ifdef CONFIG_NUMA_BALANCING
 783 static inline int pte_protnone(pte_t pte)
 784 {
 785         return pte_present(pte) && !(pte_val(pte) & _PAGE_READ);
 786 }
 787
 788 static inline int pmd_protnone(pmd_t pmd)
 789 {
 790         /* pmd_large(pmd) implies pmd_present(pmd) */
 791         return pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_READ);
 792 }
 793 #endif
 794
 795 static inline int pte_soft_dirty(pte_t pte)
 796 {
 797         return pte_val(pte) & _PAGE_SOFT_DIRTY;
 798 }
 799 #define pte_swp_soft_dirty pte_soft_dirty
 800
 801 static inline pte_t pte_mksoft_dirty(pte_t pte)
 802 {
 803         pte_val(pte) |= _PAGE_SOFT_DIRTY;
 804         return pte;
 805 }
 806 #define pte_swp_mksoft_dirty pte_mksoft_dirty
 807
 808 static inline pte_t pte_clear_soft_dirty(pte_t pte)
 809 {
 810         pte_val(pte) &= ~_PAGE_SOFT_DIRTY;
 811         return pte;
 812 }
 813 #define pte_swp_clear_soft_dirty pte_clear_soft_dirty
 814
 815 static inline int pmd_soft_dirty(pmd_t pmd)
 816 {
 817         return pmd_val(pmd) & _SEGMENT_ENTRY_SOFT_DIRTY;
 818 }
 819
 820 static inline pmd_t pmd_mksoft_dirty(pmd_t pmd)
 821 {
 822         pmd_val(pmd) |= _SEGMENT_ENTRY_SOFT_DIRTY;
 823         return pmd;
 824 }
 825
 826 static inline pmd_t pmd_clear_soft_dirty(pmd_t pmd)
 827 {
 828         pmd_val(pmd) &= ~_SEGMENT_ENTRY_SOFT_DIRTY;
 829         return pmd;
 830 }
 831
 832 /*
 833  * query functions pte_write/pte_dirty/pte_young only work if
 834  * pte_present() is true. Undefined behaviour if not..
 835  */
 836 static inline int pte_write(pte_t pte)
 837 {
 838         return (pte_val(pte) & _PAGE_WRITE) != 0;
 839 }
 840
 841 static inline int pte_dirty(pte_t pte)
 842 {
 843         return (pte_val(pte) & _PAGE_DIRTY) != 0;
 844 }
 845
 846 static inline int pte_young(pte_t pte)
 847 {
 848         return (pte_val(pte) & _PAGE_YOUNG) != 0;
 849 }
 850
 851 #define __HAVE_ARCH_PTE_UNUSED
 852 static inline int pte_unused(pte_t pte)
 853 {
 854         return pte_val(pte) & _PAGE_UNUSED;
 855 }
 856
 857 /*
 858  * pgd/pmd/pte modification functions
 859  */
 860
 861 static inline void pgd_clear(pgd_t *pgd)
 862 {
 863         if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
 864                 pgd_val(*pgd) = _REGION1_ENTRY_EMPTY;
 865 }
 866
 867 static inline void p4d_clear(p4d_t *p4d)
 868 {
 869         if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
 870                 p4d_val(*p4d) = _REGION2_ENTRY_EMPTY;
 871 }
 872
 873 static inline void pud_clear(pud_t *pud)
 874 {
 875         if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
 876                 pud_val(*pud) = _REGION3_ENTRY_EMPTY;
 877 }
 878
 879 static inline void pmd_clear(pmd_t *pmdp)
 880 {
 881         pmd_val(*pmdp) = _SEGMENT_ENTRY_EMPTY;
 882 }
 883
 884 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
 885 {
 886         pte_val(*ptep) = _PAGE_INVALID;
 887 }
 888
 889 /*
 890  * The following pte modification functions only work if
 891  * pte_present() is true. Undefined behaviour if not..
 892  */
 893 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 894 {
 895         pte_val(pte) &= _PAGE_CHG_MASK;
 896         pte_val(pte) |= pgprot_val(newprot);
 897         /*
 898          * newprot for PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX
 899          * has the invalid bit set, clear it again for readable, young pages
 900          */
 901         if ((pte_val(pte) & _PAGE_YOUNG) && (pte_val(pte) & _PAGE_READ))
 902                 pte_val(pte) &= ~_PAGE_INVALID;
 903         /*
 904          * newprot for PAGE_RO, PAGE_RX, PAGE_RW and PAGE_RWX has the page
 905          * protection bit set, clear it again for writable, dirty pages
 906          */
 907         if ((pte_val(pte) & _PAGE_DIRTY) && (pte_val(pte) & _PAGE_WRITE))
 908                 pte_val(pte) &= ~_PAGE_PROTECT;
 909         return pte;
 910 }
 911
 912 static inline pte_t pte_wrprotect(pte_t pte)
 913 {
 914         pte_val(pte) &= ~_PAGE_WRITE;
 915         pte_val(pte) |= _PAGE_PROTECT;
 916         return pte;
 917 }
 918
 919 static inline pte_t pte_mkwrite(pte_t pte)
 920 {
 921         pte_val(pte) |= _PAGE_WRITE;
 922         if (pte_val(pte) & _PAGE_DIRTY)
 923                 pte_val(pte) &= ~_PAGE_PROTECT;
 924         return pte;
 925 }
 926
 927 static inline pte_t pte_mkclean(pte_t pte)
 928 {
 929         pte_val(pte) &= ~_PAGE_DIRTY;
 930         pte_val(pte) |= _PAGE_PROTECT;
 931         return pte;
 932 }
 933
 934 static inline pte_t pte_mkdirty(pte_t pte)
 935 {
 936         pte_val(pte) |= _PAGE_DIRTY | _PAGE_SOFT_DIRTY;
 937         if (pte_val(pte) & _PAGE_WRITE)
 938                 pte_val(pte) &= ~_PAGE_PROTECT;
 939         return pte;
 940 }
 941
 942 static inline pte_t pte_mkold(pte_t pte)
 943 {
 944         pte_val(pte) &= ~_PAGE_YOUNG;
 945         pte_val(pte) |= _PAGE_INVALID;
 946         return pte;
 947 }
 948
 949 static inline pte_t pte_mkyoung(pte_t pte)
 950 {
 951         pte_val(pte) |= _PAGE_YOUNG;
 952         if (pte_val(pte) & _PAGE_READ)
 953                 pte_val(pte) &= ~_PAGE_INVALID;
 954         return pte;
 955 }
 956
 957 static inline pte_t pte_mkspecial(pte_t pte)
 958 {
 959         pte_val(pte) |= _PAGE_SPECIAL;
 960         return pte;
 961 }
 962
 963 #ifdef CONFIG_HUGETLB_PAGE
 964 static inline pte_t pte_mkhuge(pte_t pte)
 965 {
 966         pte_val(pte) |= _PAGE_LARGE;
 967         return pte;
 968 }
 969 #endif
 970
 971 #define IPTE_GLOBAL     0
 972 #define IPTE_LOCAL      1
 973
 974 #define IPTE_NODAT      0x400
 975 #define IPTE_GUEST_ASCE 0x800
 976
 977 static inline void __ptep_ipte(unsigned long address, pte_t *ptep,
 978                                unsigned long opt, unsigned long asce,
 979                                int local)
 980 {
 981         unsigned long pto = (unsigned long) ptep;
 982
 983         if (__builtin_constant_p(opt) && opt == 0) {
 984                 /* Invalidation + TLB flush for the pte */
 985                 asm volatile(
 986                         "       .insn   rrf,0xb2210000,%[r1],%[r2],0,%[m4]"
 987                         : "+m" (*ptep) : [r1] "a" (pto), [r2] "a" (address),
 988                           [m4] "i" (local));
 989                 return;
 990         }
 991
 992         /* Invalidate ptes with options + TLB flush of the ptes */
 993         opt = opt | (asce & _ASCE_ORIGIN);
 994         asm volatile(
 995                 "       .insn   rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
 996                 : [r2] "+a" (address), [r3] "+a" (opt)
 997                 : [r1] "a" (pto), [m4] "i" (local) : "memory");
 998 }
 999
1000 static inline void __ptep_ipte_range(unsigned long address, int nr,
1001                                      pte_t *ptep, int local)
1002 {
1003         unsigned long pto = (unsigned long) ptep;
1004
1005         /* Invalidate a range of ptes + TLB flush of the ptes */
1006         do {
1007                 asm volatile(
1008                         "       .insn rrf,0xb2210000,%[r1],%[r2],%[r3],%[m4]"
1009                         : [r2] "+a" (address), [r3] "+a" (nr)
1010                         : [r1] "a" (pto), [m4] "i" (local) : "memory");
1011         } while (nr != 255);
1012 }
1013
1014 /*
1015  * This is hard to understand. ptep_get_and_clear and ptep_clear_flush
1016  * both clear the TLB for the unmapped pte. The reason is that
1017  * ptep_get_and_clear is used in common code (e.g. change_pte_range)
1018  * to modify an active pte. The sequence is
1019  *   1) ptep_get_and_clear
1020  *   2) set_pte_at
1021  *   3) flush_tlb_range
1022  * On s390 the tlb needs to get flushed with the modification of the pte
1023  * if the pte is active. The only way how this can be implemented is to
1024  * have ptep_get_and_clear do the tlb flush. In exchange flush_tlb_range
1025  * is a nop.
1026  */
1027 pte_t ptep_xchg_direct(struct mm_struct *, unsigned long, pte_t *, pte_t);
1028 pte_t ptep_xchg_lazy(struct mm_struct *, unsigned long, pte_t *, pte_t);
1029
1030 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
1031 static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
1032                                             unsigned long addr, pte_t *ptep)
1033 {
1034         pte_t pte = *ptep;
1035
1036         pte = ptep_xchg_direct(vma->vm_mm, addr, ptep, pte_mkold(pte));
1037         return pte_young(pte);
1038 }
1039
1040 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
1041 static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
1042                                          unsigned long address, pte_t *ptep)
1043 {
1044         return ptep_test_and_clear_young(vma, address, ptep);
1045 }
1046
1047 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
1048 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
1049                                        unsigned long addr, pte_t *ptep)
1050 {
1051         return ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1052 }
1053
1054 #define __HAVE_ARCH_PTEP_MODIFY_PROT_TRANSACTION
1055 pte_t ptep_modify_prot_start(struct mm_struct *, unsigned long, pte_t *);
1056 void ptep_modify_prot_commit(struct mm_struct *, unsigned long, pte_t *, pte_t);
1057
1058 #define __HAVE_ARCH_PTEP_CLEAR_FLUSH
1059 static inline pte_t ptep_clear_flush(struct vm_area_struct *vma,
1060                                      unsigned long addr, pte_t *ptep)
1061 {
1062         return ptep_xchg_direct(vma->vm_mm, addr, ptep, __pte(_PAGE_INVALID));
1063 }
1064
1065 /*
1066  * The batched pte unmap code uses ptep_get_and_clear_full to clear the
1067  * ptes. Here an optimization is possible. tlb_gather_mmu flushes all
1068  * tlbs of an mm if it can guarantee that the ptes of the mm_struct
1069  * cannot be accessed while the batched unmap is running. In this case
1070  * full==1 and a simple pte_clear is enough. See tlb.h.
1071  */
1072 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
1073 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
1074                                             unsigned long addr,
1075                                             pte_t *ptep, int full)
1076 {
1077         if (full) {
1078                 pte_t pte = *ptep;
1079                 *ptep = __pte(_PAGE_INVALID);
1080                 return pte;
1081         }
1082         return ptep_xchg_lazy(mm, addr, ptep, __pte(_PAGE_INVALID));
1083 }
1084
1085 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
1086 static inline void ptep_set_wrprotect(struct mm_struct *mm,
1087                                       unsigned long addr, pte_t *ptep)
1088 {
1089         pte_t pte = *ptep;
1090
1091         if (pte_write(pte))
1092                 ptep_xchg_lazy(mm, addr, ptep, pte_wrprotect(pte));
1093 }
1094
1095 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
1096 static inline int ptep_set_access_flags(struct vm_area_struct *vma,
1097                                         unsigned long addr, pte_t *ptep,
1098                                         pte_t entry, int dirty)
1099 {
1100         if (pte_same(*ptep, entry))
1101                 return 0;
1102         ptep_xchg_direct(vma->vm_mm, addr, ptep, entry);
1103         return 1;
1104 }
1105
1106 /*
1107  * Additional functions to handle KVM guest page tables
1108  */
1109 void ptep_set_pte_at(struct mm_struct *mm, unsigned long addr,
1110                      pte_t *ptep, pte_t entry);
1111 void ptep_set_notify(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1112 void ptep_notify(struct mm_struct *mm, unsigned long addr,
1113                  pte_t *ptep, unsigned long bits);
1114 int ptep_force_prot(struct mm_struct *mm, unsigned long gaddr,
1115                     pte_t *ptep, int prot, unsigned long bit);
1116 void ptep_zap_unused(struct mm_struct *mm, unsigned long addr,
1117                      pte_t *ptep , int reset);
1118 void ptep_zap_key(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
1119 int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
1120                     pte_t *sptep, pte_t *tptep, pte_t pte);
1121 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep);
1122
1123 bool test_and_clear_guest_dirty(struct mm_struct *mm, unsigned long address);
1124 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1125                           unsigned char key, bool nq);
1126 int cond_set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1127                                unsigned char key, unsigned char *oldkey,
1128                                bool nq, bool mr, bool mc);
1129 int reset_guest_reference_bit(struct mm_struct *mm, unsigned long addr);
1130 int get_guest_storage_key(struct mm_struct *mm, unsigned long addr,
1131                           unsigned char *key);
1132
1133 int set_pgste_bits(struct mm_struct *mm, unsigned long addr,
1134                                 unsigned long bits, unsigned long value);
1135 int get_pgste(struct mm_struct *mm, unsigned long hva, unsigned long *pgstep);
1136 int pgste_perform_essa(struct mm_struct *mm, unsigned long hva, int orc,
1137                         unsigned long *oldpte, unsigned long *oldpgste);
1138
1139 /*
1140  * Certain architectures need to do special things when PTEs
1141  * within a page table are directly modified.  Thus, the following
1142  * hook is made available.
1143  */
1144 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
1145                               pte_t *ptep, pte_t entry)
1146 {
1147         if (pte_present(entry))
1148                 pte_val(entry) &= ~_PAGE_UNUSED;
1149         if (mm_has_pgste(mm))
1150                 ptep_set_pte_at(mm, addr, ptep, entry);
1151         else
1152                 *ptep = entry;
1153 }
1154
1155 /*
1156  * Conversion functions: convert a page and protection to a page entry,
1157  * and a page entry and page directory to the page they refer to.
1158  */
1159 static inline pte_t mk_pte_phys(unsigned long physpage, pgprot_t pgprot)
1160 {
1161         pte_t __pte;
1162         pte_val(__pte) = physpage + pgprot_val(pgprot);
1163         if (!MACHINE_HAS_NX)
1164                 pte_val(__pte) &= ~_PAGE_NOEXEC;
1165         return pte_mkyoung(__pte);
1166 }
1167
1168 static inline pte_t mk_pte(struct page *page, pgprot_t pgprot)
1169 {
1170         unsigned long physpage = page_to_phys(page);
1171         pte_t __pte = mk_pte_phys(physpage, pgprot);
1172
1173         if (pte_write(__pte) && PageDirty(page))
1174                 __pte = pte_mkdirty(__pte);
1175         return __pte;
1176 }
1177
1178 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD-1))
1179 #define p4d_index(address) (((address) >> P4D_SHIFT) & (PTRS_PER_P4D-1))
1180 #define pud_index(address) (((address) >> PUD_SHIFT) & (PTRS_PER_PUD-1))
1181 #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1))
1182 #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1))
1183
1184 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
1185 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
1186
1187 #define pmd_deref(pmd) (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN)
1188 #define pud_deref(pud) (pud_val(pud) & _REGION_ENTRY_ORIGIN)
1189 #define p4d_deref(pud) (p4d_val(pud) & _REGION_ENTRY_ORIGIN)
1190 #define pgd_deref(pgd) (pgd_val(pgd) & _REGION_ENTRY_ORIGIN)
1191
1192 static inline p4d_t *p4d_offset(pgd_t *pgd, unsigned long address)
1193 {
1194         p4d_t *p4d = (p4d_t *) pgd;
1195
1196         if ((pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R1)
1197                 p4d = (p4d_t *) pgd_deref(*pgd);
1198         return p4d + p4d_index(address);
1199 }
1200
1201 static inline pud_t *pud_offset(p4d_t *p4d, unsigned long address)
1202 {
1203         pud_t *pud = (pud_t *) p4d;
1204
1205         if ((p4d_val(*p4d) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R2)
1206                 pud = (pud_t *) p4d_deref(*p4d);
1207         return pud + pud_index(address);
1208 }
1209
1210 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
1211 {
1212         pmd_t *pmd = (pmd_t *) pud;
1213
1214         if ((pud_val(*pud) & _REGION_ENTRY_TYPE_MASK) == _REGION_ENTRY_TYPE_R3)
1215                 pmd = (pmd_t *) pud_deref(*pud);
1216         return pmd + pmd_index(address);
1217 }
1218
1219 #define pfn_pte(pfn,pgprot) mk_pte_phys(__pa((pfn) << PAGE_SHIFT),(pgprot))
1220 #define pte_pfn(x) (pte_val(x) >> PAGE_SHIFT)
1221 #define pte_page(x) pfn_to_page(pte_pfn(x))
1222
1223 #define pmd_page(pmd) pfn_to_page(pmd_pfn(pmd))
1224 #define pud_page(pud) pfn_to_page(pud_pfn(pud))
1225 #define p4d_page(pud) pfn_to_page(p4d_pfn(p4d))
1226
1227 /* Find an entry in the lowest level page table.. */
1228 #define pte_offset(pmd, addr) ((pte_t *) pmd_deref(*(pmd)) + pte_index(addr))
1229 #define pte_offset_kernel(pmd, address) pte_offset(pmd,address)
1230 #define pte_offset_map(pmd, address) pte_offset_kernel(pmd, address)
1231 #define pte_unmap(pte) do { } while (0)
1232
1233 static inline pmd_t pmd_wrprotect(pmd_t pmd)
1234 {
1235         pmd_val(pmd) &= ~_SEGMENT_ENTRY_WRITE;
1236         pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1237         return pmd;
1238 }
1239
1240 static inline pmd_t pmd_mkwrite(pmd_t pmd)
1241 {
1242         pmd_val(pmd) |= _SEGMENT_ENTRY_WRITE;
1243         if (pmd_large(pmd) && !(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1244                 return pmd;
1245         pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1246         return pmd;
1247 }
1248
1249 static inline pmd_t pmd_mkclean(pmd_t pmd)
1250 {
1251         if (pmd_large(pmd)) {
1252                 pmd_val(pmd) &= ~_SEGMENT_ENTRY_DIRTY;
1253                 pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1254         }
1255         return pmd;
1256 }
1257
1258 static inline pmd_t pmd_mkdirty(pmd_t pmd)
1259 {
1260         if (pmd_large(pmd)) {
1261                 pmd_val(pmd) |= _SEGMENT_ENTRY_DIRTY |
1262                                 _SEGMENT_ENTRY_SOFT_DIRTY;
1263                 if (pmd_val(pmd) & _SEGMENT_ENTRY_WRITE)
1264                         pmd_val(pmd) &= ~_SEGMENT_ENTRY_PROTECT;
1265         }
1266         return pmd;
1267 }
1268
1269 static inline pud_t pud_wrprotect(pud_t pud)
1270 {
1271         pud_val(pud) &= ~_REGION3_ENTRY_WRITE;
1272         pud_val(pud) |= _REGION_ENTRY_PROTECT;
1273         return pud;
1274 }
1275
1276 static inline pud_t pud_mkwrite(pud_t pud)
1277 {
1278         pud_val(pud) |= _REGION3_ENTRY_WRITE;
1279         if (pud_large(pud) && !(pud_val(pud) & _REGION3_ENTRY_DIRTY))
1280                 return pud;
1281         pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1282         return pud;
1283 }
1284
1285 static inline pud_t pud_mkclean(pud_t pud)
1286 {
1287         if (pud_large(pud)) {
1288                 pud_val(pud) &= ~_REGION3_ENTRY_DIRTY;
1289                 pud_val(pud) |= _REGION_ENTRY_PROTECT;
1290         }
1291         return pud;
1292 }
1293
1294 static inline pud_t pud_mkdirty(pud_t pud)
1295 {
1296         if (pud_large(pud)) {
1297                 pud_val(pud) |= _REGION3_ENTRY_DIRTY |
1298                                 _REGION3_ENTRY_SOFT_DIRTY;
1299                 if (pud_val(pud) & _REGION3_ENTRY_WRITE)
1300                         pud_val(pud) &= ~_REGION_ENTRY_PROTECT;
1301         }
1302         return pud;
1303 }
1304
1305 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLB_PAGE)
1306 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
1307 {
1308         /*
1309          * pgprot is PAGE_NONE, PAGE_RO, PAGE_RX, PAGE_RW or PAGE_RWX
1310          * (see __Pxxx / __Sxxx). Convert to segment table entry format.
1311          */
1312         if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
1313                 return pgprot_val(SEGMENT_NONE);
1314         if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
1315                 return pgprot_val(SEGMENT_RO);
1316         if (pgprot_val(pgprot) == pgprot_val(PAGE_RX))
1317                 return pgprot_val(SEGMENT_RX);
1318         if (pgprot_val(pgprot) == pgprot_val(PAGE_RW))
1319                 return pgprot_val(SEGMENT_RW);
1320         return pgprot_val(SEGMENT_RWX);
1321 }
1322
1323 static inline pmd_t pmd_mkyoung(pmd_t pmd)
1324 {
1325         if (pmd_large(pmd)) {
1326                 pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1327                 if (pmd_val(pmd) & _SEGMENT_ENTRY_READ)
1328                         pmd_val(pmd) &= ~_SEGMENT_ENTRY_INVALID;
1329         }
1330         return pmd;
1331 }
1332
1333 static inline pmd_t pmd_mkold(pmd_t pmd)
1334 {
1335         if (pmd_large(pmd)) {
1336                 pmd_val(pmd) &= ~_SEGMENT_ENTRY_YOUNG;
1337                 pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1338         }
1339         return pmd;
1340 }
1341
1342 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
1343 {
1344         if (pmd_large(pmd)) {
1345                 pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN_LARGE |
1346                         _SEGMENT_ENTRY_DIRTY | _SEGMENT_ENTRY_YOUNG |
1347                         _SEGMENT_ENTRY_LARGE | _SEGMENT_ENTRY_SOFT_DIRTY;
1348                 pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1349                 if (!(pmd_val(pmd) & _SEGMENT_ENTRY_DIRTY))
1350                         pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1351                 if (!(pmd_val(pmd) & _SEGMENT_ENTRY_YOUNG))
1352                         pmd_val(pmd) |= _SEGMENT_ENTRY_INVALID;
1353                 return pmd;
1354         }
1355         pmd_val(pmd) &= _SEGMENT_ENTRY_ORIGIN;
1356         pmd_val(pmd) |= massage_pgprot_pmd(newprot);
1357         return pmd;
1358 }
1359
1360 static inline pmd_t mk_pmd_phys(unsigned long physpage, pgprot_t pgprot)
1361 {
1362         pmd_t __pmd;
1363         pmd_val(__pmd) = physpage + massage_pgprot_pmd(pgprot);
1364         return __pmd;
1365 }
1366
1367 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLB_PAGE */
1368
1369 static inline void __pmdp_csp(pmd_t *pmdp)
1370 {
1371         csp((unsigned int *)pmdp + 1, pmd_val(*pmdp),
1372             pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1373 }
1374
1375 #define IDTE_GLOBAL     0
1376 #define IDTE_LOCAL      1
1377
1378 #define IDTE_PTOA       0x0800
1379 #define IDTE_NODAT      0x1000
1380 #define IDTE_GUEST_ASCE 0x2000
1381
1382 static inline void __pmdp_idte(unsigned long addr, pmd_t *pmdp,
1383                                unsigned long opt, unsigned long asce,
1384                                int local)
1385 {
1386         unsigned long sto;
1387
1388         sto = (unsigned long) pmdp - pmd_index(addr) * sizeof(pmd_t);
1389         if (__builtin_constant_p(opt) && opt == 0) {
1390                 /* flush without guest asce */
1391                 asm volatile(
1392                         "       .insn   rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1393                         : "+m" (*pmdp)
1394                         : [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK)),
1395                           [m4] "i" (local)
1396                         : "cc" );
1397         } else {
1398                 /* flush with guest asce */
1399                 asm volatile(
1400                         "       .insn   rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1401                         : "+m" (*pmdp)
1402                         : [r1] "a" (sto), [r2] "a" ((addr & HPAGE_MASK) | opt),
1403                           [r3] "a" (asce), [m4] "i" (local)
1404                         : "cc" );
1405         }
1406 }
1407
1408 static inline void __pudp_idte(unsigned long addr, pud_t *pudp,
1409                                unsigned long opt, unsigned long asce,
1410                                int local)
1411 {
1412         unsigned long r3o;
1413
1414         r3o = (unsigned long) pudp - pud_index(addr) * sizeof(pud_t);
1415         r3o |= _ASCE_TYPE_REGION3;
1416         if (__builtin_constant_p(opt) && opt == 0) {
1417                 /* flush without guest asce */
1418                 asm volatile(
1419                         "       .insn   rrf,0xb98e0000,%[r1],%[r2],0,%[m4]"
1420                         : "+m" (*pudp)
1421                         : [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK)),
1422                           [m4] "i" (local)
1423                         : "cc");
1424         } else {
1425                 /* flush with guest asce */
1426                 asm volatile(
1427                         "       .insn   rrf,0xb98e0000,%[r1],%[r2],%[r3],%[m4]"
1428                         : "+m" (*pudp)
1429                         : [r1] "a" (r3o), [r2] "a" ((addr & PUD_MASK) | opt),
1430                           [r3] "a" (asce), [m4] "i" (local)
1431                         : "cc" );
1432         }
1433 }
1434
1435 pmd_t pmdp_xchg_direct(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1436 pmd_t pmdp_xchg_lazy(struct mm_struct *, unsigned long, pmd_t *, pmd_t);
1437 pud_t pudp_xchg_direct(struct mm_struct *, unsigned long, pud_t *, pud_t);
1438
1439 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1440
1441 #define __HAVE_ARCH_PGTABLE_DEPOSIT
1442 void pgtable_trans_huge_deposit(struct mm_struct *mm, pmd_t *pmdp,
1443                                 pgtable_t pgtable);
1444
1445 #define __HAVE_ARCH_PGTABLE_WITHDRAW
1446 pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
1447
1448 #define  __HAVE_ARCH_PMDP_SET_ACCESS_FLAGS
1449 static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
1450                                         unsigned long addr, pmd_t *pmdp,
1451                                         pmd_t entry, int dirty)
1452 {
1453         VM_BUG_ON(addr & ~HPAGE_MASK);
1454
1455         entry = pmd_mkyoung(entry);
1456         if (dirty)
1457                 entry = pmd_mkdirty(entry);
1458         if (pmd_val(*pmdp) == pmd_val(entry))
1459                 return 0;
1460         pmdp_xchg_direct(vma->vm_mm, addr, pmdp, entry);
1461         return 1;
1462 }
1463
1464 #define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
1465 static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
1466                                             unsigned long addr, pmd_t *pmdp)
1467 {
1468         pmd_t pmd = *pmdp;
1469
1470         pmd = pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd_mkold(pmd));
1471         return pmd_young(pmd);
1472 }
1473
1474 #define __HAVE_ARCH_PMDP_CLEAR_YOUNG_FLUSH
1475 static inline int pmdp_clear_flush_young(struct vm_area_struct *vma,
1476                                          unsigned long addr, pmd_t *pmdp)
1477 {
1478         VM_BUG_ON(addr & ~HPAGE_MASK);
1479         return pmdp_test_and_clear_young(vma, addr, pmdp);
1480 }
1481
1482 static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
1483                               pmd_t *pmdp, pmd_t entry)
1484 {
1485         if (!MACHINE_HAS_NX)
1486                 pmd_val(entry) &= ~_SEGMENT_ENTRY_NOEXEC;
1487         *pmdp = entry;
1488 }
1489
1490 static inline pmd_t pmd_mkhuge(pmd_t pmd)
1491 {
1492         pmd_val(pmd) |= _SEGMENT_ENTRY_LARGE;
1493         pmd_val(pmd) |= _SEGMENT_ENTRY_YOUNG;
1494         pmd_val(pmd) |= _SEGMENT_ENTRY_PROTECT;
1495         return pmd;
1496 }
1497
1498 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
1499 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
1500                                             unsigned long addr, pmd_t *pmdp)
1501 {
1502         return pmdp_xchg_direct(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1503 }
1504
1505 #define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR_FULL
1506 static inline pmd_t pmdp_huge_get_and_clear_full(struct mm_struct *mm,
1507                                                  unsigned long addr,
1508                                                  pmd_t *pmdp, int full)
1509 {
1510         if (full) {
1511                 pmd_t pmd = *pmdp;
1512                 *pmdp = __pmd(_SEGMENT_ENTRY_EMPTY);
1513                 return pmd;
1514         }
1515         return pmdp_xchg_lazy(mm, addr, pmdp, __pmd(_SEGMENT_ENTRY_EMPTY));
1516 }
1517
1518 #define __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH
1519 static inline pmd_t pmdp_huge_clear_flush(struct vm_area_struct *vma,
1520                                           unsigned long addr, pmd_t *pmdp)
1521 {
1522         return pmdp_huge_get_and_clear(vma->vm_mm, addr, pmdp);
1523 }
1524
1525 #define __HAVE_ARCH_PMDP_INVALIDATE
1526 static inline void pmdp_invalidate(struct vm_area_struct *vma,
1527                                    unsigned long addr, pmd_t *pmdp)
1528 {
1529         pmd_t pmd = __pmd(pmd_val(*pmdp) | _SEGMENT_ENTRY_INVALID);
1530
1531         pmdp_xchg_direct(vma->vm_mm, addr, pmdp, pmd);
1532 }
1533
1534 #define __HAVE_ARCH_PMDP_SET_WRPROTECT
1535 static inline void pmdp_set_wrprotect(struct mm_struct *mm,
1536                                       unsigned long addr, pmd_t *pmdp)
1537 {
1538         pmd_t pmd = *pmdp;
1539
1540         if (pmd_write(pmd))
1541                 pmd = pmdp_xchg_lazy(mm, addr, pmdp, pmd_wrprotect(pmd));
1542 }
1543
1544 static inline pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
1545                                         unsigned long address,
1546                                         pmd_t *pmdp)
1547 {
1548         return pmdp_huge_get_and_clear(vma->vm_mm, address, pmdp);
1549 }
1550 #define pmdp_collapse_flush pmdp_collapse_flush
1551
1552 #define pfn_pmd(pfn, pgprot)    mk_pmd_phys(__pa((pfn) << PAGE_SHIFT), (pgprot))
1553 #define mk_pmd(page, pgprot)    pfn_pmd(page_to_pfn(page), (pgprot))
1554
1555 static inline int pmd_trans_huge(pmd_t pmd)
1556 {
1557         return pmd_val(pmd) & _SEGMENT_ENTRY_LARGE;
1558 }
1559
1560 #define has_transparent_hugepage has_transparent_hugepage
1561 static inline int has_transparent_hugepage(void)
1562 {
1563         return MACHINE_HAS_EDAT1 ? 1 : 0;
1564 }
1565 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
1566
1567 /*
1568  * 64 bit swap entry format:
1569  * A page-table entry has some bits we have to treat in a special way.
1570  * Bits 52 and bit 55 have to be zero, otherwise a specification
1571  * exception will occur instead of a page translation exception. The
1572  * specification exception has the bad habit not to store necessary
1573  * information in the lowcore.
1574  * Bits 54 and 63 are used to indicate the page type.
1575  * A swap pte is indicated by bit pattern (pte & 0x201) == 0x200
1576  * This leaves the bits 0-51 and bits 56-62 to store type and offset.
1577  * We use the 5 bits from 57-61 for the type and the 52 bits from 0-51
1578  * for the offset.
1579  * |                      offset                        |01100|type |00|
1580  * |0000000000111111111122222222223333333333444444444455|55555|55566|66|
1581  * |0123456789012345678901234567890123456789012345678901|23456|78901|23|
1582  */
1583
1584 #define __SWP_OFFSET_MASK       ((1UL << 52) - 1)
1585 #define __SWP_OFFSET_SHIFT      12
1586 #define __SWP_TYPE_MASK         ((1UL << 5) - 1)
1587 #define __SWP_TYPE_SHIFT        2
1588
1589 static inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
1590 {
1591         pte_t pte;
1592
1593         pte_val(pte) = _PAGE_INVALID | _PAGE_PROTECT;
1594         pte_val(pte) |= (offset & __SWP_OFFSET_MASK) << __SWP_OFFSET_SHIFT;
1595         pte_val(pte) |= (type & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT;
1596         return pte;
1597 }
1598
1599 static inline unsigned long __swp_type(swp_entry_t entry)
1600 {
1601         return (entry.val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK;
1602 }
1603
1604 static inline unsigned long __swp_offset(swp_entry_t entry)
1605 {
1606         return (entry.val >> __SWP_OFFSET_SHIFT) & __SWP_OFFSET_MASK;
1607 }
1608
1609 static inline swp_entry_t __swp_entry(unsigned long type, unsigned long offset)
1610 {
1611         return (swp_entry_t) { pte_val(mk_swap_pte(type, offset)) };
1612 }
1613
1614 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) })
1615 #define __swp_entry_to_pte(x)   ((pte_t) { (x).val })
1616
1617 #define kern_addr_valid(addr)   (1)
1618
1619 extern int vmem_add_mapping(unsigned long start, unsigned long size);
1620 extern int vmem_remove_mapping(unsigned long start, unsigned long size);
1621 extern int s390_enable_sie(void);
1622 extern int s390_enable_skey(void);
1623 extern void s390_reset_cmma(struct mm_struct *mm);
1624
1625 /* s390 has a private copy of get unmapped area to deal with cache synonyms */
1626 #define HAVE_ARCH_UNMAPPED_AREA
1627 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
1628
1629 /*
1630  * No page table caches to initialise
1631  */
1632 static inline void pgtable_cache_init(void) { }
1633 static inline void check_pgt_cache(void) { }
1634
1635 #include <asm-generic/pgtable.h>
1636
1637 #endif /* _S390_PAGE_H */