include/asm-ppc/pgtable.h

   1 #ifdef __KERNEL__
   2 #ifndef _PPC_PGTABLE_H
   3 #define _PPC_PGTABLE_H
   4
   5 #include <asm-generic/4level-fixup.h>
   6
   7 #include <linux/config.h>
   8
   9 #ifndef __ASSEMBLY__
  10 #include <linux/sched.h>
  11 #include <linux/threads.h>
  12 #include <asm/processor.h>              /* For TASK_SIZE */
  13 #include <asm/mmu.h>
  14 #include <asm/page.h>
  15 struct mm_struct;
  16
  17 extern unsigned long va_to_phys(unsigned long address);
  18 extern pte_t *va_to_pte(unsigned long address);
  19 extern unsigned long ioremap_bot, ioremap_base;
  20 #endif /* __ASSEMBLY__ */
  21
  22 /*
  23  * The PowerPC MMU uses a hash table containing PTEs, together with
  24  * a set of 16 segment registers (on 32-bit implementations), to define
  25  * the virtual to physical address mapping.
  26  *
  27  * We use the hash table as an extended TLB, i.e. a cache of currently
  28  * active mappings.  We maintain a two-level page table tree, much
  29  * like that used by the i386, for the sake of the Linux memory
  30  * management code.  Low-level assembler code in hashtable.S
  31  * (procedure hash_page) is responsible for extracting ptes from the
  32  * tree and putting them into the hash table when necessary, and
  33  * updating the accessed and modified bits in the page table tree.
  34  */
  35
  36 /*
  37  * The PowerPC MPC8xx uses a TLB with hardware assisted, software tablewalk.
  38  * We also use the two level tables, but we can put the real bits in them
  39  * needed for the TLB and tablewalk.  These definitions require Mx_CTR.PPM = 0,
  40  * Mx_CTR.PPCS = 0, and MD_CTR.TWAM = 1.  The level 2 descriptor has
  41  * additional page protection (when Mx_CTR.PPCS = 1) that allows TLB hit
  42  * based upon user/super access.  The TLB does not have accessed nor write
  43  * protect.  We assume that if the TLB get loaded with an entry it is
  44  * accessed, and overload the changed bit for write protect.  We use
  45  * two bits in the software pte that are supposed to be set to zero in
  46  * the TLB entry (24 and 25) for these indicators.  Although the level 1
  47  * descriptor contains the guarded and writethrough/copyback bits, we can
  48  * set these at the page level since they get copied from the Mx_TWC
  49  * register when the TLB entry is loaded.  We will use bit 27 for guard, since
  50  * that is where it exists in the MD_TWC, and bit 26 for writethrough.
  51  * These will get masked from the level 2 descriptor at TLB load time, and
  52  * copied to the MD_TWC before it gets loaded.
  53  * Large page sizes added.  We currently support two sizes, 4K and 8M.
  54  * This also allows a TLB hander optimization because we can directly
  55  * load the PMD into MD_TWC.  The 8M pages are only used for kernel
  56  * mapping of well known areas.  The PMD (PGD) entries contain control
  57  * flags in addition to the address, so care must be taken that the
  58  * software no longer assumes these are only pointers.
  59  */
  60
  61 /*
  62  * At present, all PowerPC 400-class processors share a similar TLB
  63  * architecture. The instruction and data sides share a unified,
  64  * 64-entry, fully-associative TLB which is maintained totally under
  65  * software control. In addition, the instruction side has a
  66  * hardware-managed, 4-entry, fully-associative TLB which serves as a
  67  * first level to the shared TLB. These two TLBs are known as the UTLB
  68  * and ITLB, respectively (see "mmu.h" for definitions).
  69  */
  70
  71 /*
  72  * The normal case is that PTEs are 32-bits and we have a 1-page
  73  * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages.  -- paulus
  74  *
  75  * For any >32-bit physical address platform, we can use the following
  76  * two level page table layout where the pgdir is 8KB and the MS 13 bits
  77  * are an index to the second level table.  The combined pgdir/pmd first
  78  * level has 2048 entries and the second level has 512 64-bit PTE entries.
  79  * -Matt
  80  */
  81 /* PMD_SHIFT determines the size of the area mapped by the PTE pages */
  82 #define PMD_SHIFT       (PAGE_SHIFT + PTE_SHIFT)
  83 #define PMD_SIZE        (1UL << PMD_SHIFT)
  84 #define PMD_MASK        (~(PMD_SIZE-1))
  85
  86 /* PGDIR_SHIFT determines what a top-level page table entry can map */
  87 #define PGDIR_SHIFT     PMD_SHIFT
  88 #define PGDIR_SIZE      (1UL << PGDIR_SHIFT)
  89 #define PGDIR_MASK      (~(PGDIR_SIZE-1))
  90
  91 /*
  92  * entries per page directory level: our page-table tree is two-level, so
  93  * we don't really have any PMD directory.
  94  */
  95 #define PTRS_PER_PTE    (1 << PTE_SHIFT)
  96 #define PTRS_PER_PMD    1
  97 #define PTRS_PER_PGD    (1 << (32 - PGDIR_SHIFT))
  98
  99 #define USER_PTRS_PER_PGD       (TASK_SIZE / PGDIR_SIZE)
 100 #define FIRST_USER_ADDRESS      0
 101
 102 #define USER_PGD_PTRS (PAGE_OFFSET >> PGDIR_SHIFT)
 103 #define KERNEL_PGD_PTRS (PTRS_PER_PGD-USER_PGD_PTRS)
 104
 105 #define pte_ERROR(e) \
 106         printk("%s:%d: bad pte "PTE_FMT".\n", __FILE__, __LINE__, pte_val(e))
 107 #define pmd_ERROR(e) \
 108         printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
 109 #define pgd_ERROR(e) \
 110         printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
 111
 112 /*
 113  * Just any arbitrary offset to the start of the vmalloc VM area: the
 114  * current 64MB value just means that there will be a 64MB "hole" after the
 115  * physical memory until the kernel virtual memory starts.  That means that
 116  * any out-of-bounds memory accesses will hopefully be caught.
 117  * The vmalloc() routines leaves a hole of 4kB between each vmalloced
 118  * area for the same reason. ;)
 119  *
 120  * We no longer map larger than phys RAM with the BATs so we don't have
 121  * to worry about the VMALLOC_OFFSET causing problems.  We do have to worry
 122  * about clashes between our early calls to ioremap() that start growing down
 123  * from ioremap_base being run into the VM area allocations (growing upwards
 124  * from VMALLOC_START).  For this reason we have ioremap_bot to check when
 125  * we actually run into our mappings setup in the early boot with the VM
 126  * system.  This really does become a problem for machines with good amounts
 127  * of RAM.  -- Cort
 128  */
 129 #define VMALLOC_OFFSET (0x1000000) /* 16M */
 130 #ifdef CONFIG_44x
 131 #include <asm/ibm44x.h>
 132 #define VMALLOC_START (((_ALIGN((long)high_memory, PPC44x_PIN_SIZE) + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
 133 #else
 134 #define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
 135 #endif
 136 #define VMALLOC_END     ioremap_bot
 137
 138 /*
 139  * Bits in a linux-style PTE.  These match the bits in the
 140  * (hardware-defined) PowerPC PTE as closely as possible.
 141  */
 142
 143 #if defined(CONFIG_40x)
 144
 145 /* There are several potential gotchas here.  The 40x hardware TLBLO
 146    field looks like this:
 147
 148    0  1  2  3  4  ... 18 19 20 21 22 23 24 25 26 27 28 29 30 31
 149    RPN.....................  0  0 EX WR ZSEL.......  W  I  M  G
 150
 151    Where possible we make the Linux PTE bits match up with this
 152
 153    - bits 20 and 21 must be cleared, because we use 4k pages (40x can
 154      support down to 1k pages), this is done in the TLBMiss exception
 155      handler.
 156    - We use only zones 0 (for kernel pages) and 1 (for user pages)
 157      of the 16 available.  Bit 24-26 of the TLB are cleared in the TLB
 158      miss handler.  Bit 27 is PAGE_USER, thus selecting the correct
 159      zone.
 160    - PRESENT *must* be in the bottom two bits because swap cache
 161      entries use the top 30 bits.  Because 40x doesn't support SMP
 162      anyway, M is irrelevant so we borrow it for PAGE_PRESENT.  Bit 30
 163      is cleared in the TLB miss handler before the TLB entry is loaded.
 164    - All other bits of the PTE are loaded into TLBLO without
 165      modification, leaving us only the bits 20, 21, 24, 25, 26, 30 for
 166      software PTE bits.  We actually use use bits 21, 24, 25, and
 167      30 respectively for the software bits: ACCESSED, DIRTY, RW, and
 168      PRESENT.
 169 */
 170
 171 /* Definitions for 40x embedded chips. */
 172 #define _PAGE_GUARDED   0x001   /* G: page is guarded from prefetch */
 173 #define _PAGE_FILE      0x001   /* when !present: nonlinear file mapping */
 174 #define _PAGE_PRESENT   0x002   /* software: PTE contains a translation */
 175 #define _PAGE_NO_CACHE  0x004   /* I: caching is inhibited */
 176 #define _PAGE_WRITETHRU 0x008   /* W: caching is write-through */
 177 #define _PAGE_USER      0x010   /* matches one of the zone permission bits */
 178 #define _PAGE_RW        0x040   /* software: Writes permitted */
 179 #define _PAGE_DIRTY     0x080   /* software: dirty page */
 180 #define _PAGE_HWWRITE   0x100   /* hardware: Dirty & RW, set in exception */
 181 #define _PAGE_HWEXEC    0x200   /* hardware: EX permission */
 182 #define _PAGE_ACCESSED  0x400   /* software: R: page referenced */
 183
 184 #define _PMD_PRESENT    0x400   /* PMD points to page of PTEs */
 185 #define _PMD_BAD        0x802
 186 #define _PMD_SIZE       0x0e0   /* size field, != 0 for large-page PMD entry */
 187 #define _PMD_SIZE_4M    0x0c0
 188 #define _PMD_SIZE_16M   0x0e0
 189 #define PMD_PAGE_SIZE(pmdval)   (1024 << (((pmdval) & _PMD_SIZE) >> 4))
 190
 191 #elif defined(CONFIG_44x)
 192 /*
 193  * Definitions for PPC440
 194  *
 195  * Because of the 3 word TLB entries to support 36-bit addressing,
 196  * the attribute are difficult to map in such a fashion that they
 197  * are easily loaded during exception processing.  I decided to
 198  * organize the entry so the ERPN is the only portion in the
 199  * upper word of the PTE and the attribute bits below are packed
 200  * in as sensibly as they can be in the area below a 4KB page size
 201  * oriented RPN.  This at least makes it easy to load the RPN and
 202  * ERPN fields in the TLB. -Matt
 203  *
 204  * Note that these bits preclude future use of a page size
 205  * less than 4KB.
 206  *
 207  *
 208  * PPC 440 core has following TLB attribute fields;
 209  *
 210  *   TLB1:
 211  *   0  1  2  3  4  ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
 212  *   RPN.................................  -  -  -  -  -  - ERPN.......
 213  *
 214  *   TLB2:
 215  *   0  1  2  3  4  ... 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
 216  *   -  -  -  -  -    - U0 U1 U2 U3 W  I  M  G  E   - UX UW UR SX SW SR
 217  *
 218  * There are some constrains and options, to decide mapping software bits
 219  * into TLB entry.
 220  *
 221  *   - PRESENT *must* be in the bottom three bits because swap cache
 222  *     entries use the top 29 bits for TLB2.
 223  *
 224  *   - FILE *must* be in the bottom three bits because swap cache
 225  *     entries use the top 29 bits for TLB2.
 226  *
 227  *   - CACHE COHERENT bit (M) has no effect on PPC440 core, because it
 228  *     doesn't support SMP. So we can use this as software bit, like
 229  *     DIRTY.
 230  *
 231  * With the PPC 44x Linux implementation, the 0-11th LSBs of the PTE are used
 232  * for memory protection related functions (see PTE structure in
 233  * include/asm-ppc/mmu.h).  The _PAGE_XXX definitions in this file map to the
 234  * above bits.  Note that the bit values are CPU specific, not architecture
 235  * specific.
 236  *
 237  * The kernel PTE entry holds an arch-dependent swp_entry structure under
 238  * certain situations. In other words, in such situations some portion of
 239  * the PTE bits are used as a swp_entry. In the PPC implementation, the
 240  * 3-24th LSB are shared with swp_entry, however the 0-2nd three LSB still
 241  * hold protection values. That means the three protection bits are
 242  * reserved for both PTE and SWAP entry at the most significant three
 243  * LSBs.
 244  *
 245  * There are three protection bits available for SWAP entry:
 246  *      _PAGE_PRESENT
 247  *      _PAGE_FILE
 248  *      _PAGE_HASHPTE (if HW has)
 249  *
 250  * So those three bits have to be inside of 0-2nd LSB of PTE.
 251  *
 252  */
 253
 254 #define _PAGE_PRESENT   0x00000001              /* S: PTE valid */
 255 #define _PAGE_RW        0x00000002              /* S: Write permission */
 256 #define _PAGE_FILE      0x00000004              /* S: nonlinear file mapping */
 257 #define _PAGE_ACCESSED  0x00000008              /* S: Page referenced */
 258 #define _PAGE_HWWRITE   0x00000010              /* H: Dirty & RW */
 259 #define _PAGE_HWEXEC    0x00000020              /* H: Execute permission */
 260 #define _PAGE_USER      0x00000040              /* S: User page */
 261 #define _PAGE_ENDIAN    0x00000080              /* H: E bit */
 262 #define _PAGE_GUARDED   0x00000100              /* H: G bit */
 263 #define _PAGE_DIRTY     0x00000200              /* S: Page dirty */
 264 #define _PAGE_NO_CACHE  0x00000400              /* H: I bit */
 265 #define _PAGE_WRITETHRU 0x00000800              /* H: W bit */
 266
 267 /* TODO: Add large page lowmem mapping support */
 268 #define _PMD_PRESENT    0
 269 #define _PMD_PRESENT_MASK (PAGE_MASK)
 270 #define _PMD_BAD        (~PAGE_MASK)
 271
 272 /* ERPN in a PTE never gets cleared, ignore it */
 273 #define _PTE_NONE_MASK  0xffffffff00000000ULL
 274
 275 #elif defined(CONFIG_FSL_BOOKE)
 276 /*
 277    MMU Assist Register 3:
 278
 279    32 33 34 35 36  ... 50 51 52 53 54 55 56 57 58 59 60 61 62 63
 280    RPN......................  0  0 U0 U1 U2 U3 UX SX UW SW UR SR
 281
 282    - PRESENT *must* be in the bottom three bits because swap cache
 283      entries use the top 29 bits.
 284
 285    - FILE *must* be in the bottom three bits because swap cache
 286      entries use the top 29 bits.
 287 */
 288
 289 /* Definitions for FSL Book-E Cores */
 290 #define _PAGE_PRESENT   0x00001 /* S: PTE contains a translation */
 291 #define _PAGE_USER      0x00002 /* S: User page (maps to UR) */
 292 #define _PAGE_FILE      0x00002 /* S: when !present: nonlinear file mapping */
 293 #define _PAGE_ACCESSED  0x00004 /* S: Page referenced */
 294 #define _PAGE_HWWRITE   0x00008 /* H: Dirty & RW, set in exception */
 295 #define _PAGE_RW        0x00010 /* S: Write permission */
 296 #define _PAGE_HWEXEC    0x00020 /* H: UX permission */
 297
 298 #define _PAGE_ENDIAN    0x00040 /* H: E bit */
 299 #define _PAGE_GUARDED   0x00080 /* H: G bit */
 300 #define _PAGE_COHERENT  0x00100 /* H: M bit */
 301 #define _PAGE_NO_CACHE  0x00200 /* H: I bit */
 302 #define _PAGE_WRITETHRU 0x00400 /* H: W bit */
 303
 304 #ifdef CONFIG_PTE_64BIT
 305 #define _PAGE_DIRTY     0x08000 /* S: Page dirty */
 306
 307 /* ERPN in a PTE never gets cleared, ignore it */
 308 #define _PTE_NONE_MASK  0xffffffffffff0000ULL
 309 #else
 310 #define _PAGE_DIRTY     0x00800 /* S: Page dirty */
 311 #endif
 312
 313 #define _PMD_PRESENT    0
 314 #define _PMD_PRESENT_MASK (PAGE_MASK)
 315 #define _PMD_BAD        (~PAGE_MASK)
 316
 317 #elif defined(CONFIG_8xx)
 318 /* Definitions for 8xx embedded chips. */
 319 #define _PAGE_PRESENT   0x0001  /* Page is valid */
 320 #define _PAGE_FILE      0x0002  /* when !present: nonlinear file mapping */
 321 #define _PAGE_NO_CACHE  0x0002  /* I: cache inhibit */
 322 #define _PAGE_SHARED    0x0004  /* No ASID (context) compare */
 323
 324 /* These five software bits must be masked out when the entry is loaded
 325  * into the TLB.
 326  */
 327 #define _PAGE_EXEC      0x0008  /* software: i-cache coherency required */
 328 #define _PAGE_GUARDED   0x0010  /* software: guarded access */
 329 #define _PAGE_DIRTY     0x0020  /* software: page changed */
 330 #define _PAGE_RW        0x0040  /* software: user write access allowed */
 331 #define _PAGE_ACCESSED  0x0080  /* software: page referenced */
 332
 333 /* Setting any bits in the nibble with the follow two controls will
 334  * require a TLB exception handler change.  It is assumed unused bits
 335  * are always zero.
 336  */
 337 #define _PAGE_HWWRITE   0x0100  /* h/w write enable: never set in Linux PTE */
 338 #define _PAGE_USER      0x0800  /* One of the PP bits, the other is USER&~RW */
 339
 340 #define _PMD_PRESENT    0x0001
 341 #define _PMD_BAD        0x0ff0
 342 #define _PMD_PAGE_MASK  0x000c
 343 #define _PMD_PAGE_8M    0x000c
 344
 345 /*
 346  * The 8xx TLB miss handler allegedly sets _PAGE_ACCESSED in the PTE
 347  * for an address even if _PAGE_PRESENT is not set, as a performance
 348  * optimization.  This is a bug if you ever want to use swap unless
 349  * _PAGE_ACCESSED is 2, which it isn't, or unless you have 8xx-specific
 350  * definitions for __swp_entry etc. below, which would be gross.
 351  *  -- paulus
 352  */
 353 #define _PTE_NONE_MASK _PAGE_ACCESSED
 354
 355 #else /* CONFIG_6xx */
 356 /* Definitions for 60x, 740/750, etc. */
 357 #define _PAGE_PRESENT   0x001   /* software: pte contains a translation */
 358 #define _PAGE_HASHPTE   0x002   /* hash_page has made an HPTE for this pte */
 359 #define _PAGE_FILE      0x004   /* when !present: nonlinear file mapping */
 360 #define _PAGE_USER      0x004   /* usermode access allowed */
 361 #define _PAGE_GUARDED   0x008   /* G: prohibit speculative access */
 362 #define _PAGE_COHERENT  0x010   /* M: enforce memory coherence (SMP systems) */
 363 #define _PAGE_NO_CACHE  0x020   /* I: cache inhibit */
 364 #define _PAGE_WRITETHRU 0x040   /* W: cache write-through */
 365 #define _PAGE_DIRTY     0x080   /* C: page changed */
 366 #define _PAGE_ACCESSED  0x100   /* R: page referenced */
 367 #define _PAGE_EXEC      0x200   /* software: i-cache coherency required */
 368 #define _PAGE_RW        0x400   /* software: user write access allowed */
 369
 370 #define _PTE_NONE_MASK  _PAGE_HASHPTE
 371
 372 #define _PMD_PRESENT    0
 373 #define _PMD_PRESENT_MASK (PAGE_MASK)
 374 #define _PMD_BAD        (~PAGE_MASK)
 375 #endif
 376
 377 /*
 378  * Some bits are only used on some cpu families...
 379  */
 380 #ifndef _PAGE_HASHPTE
 381 #define _PAGE_HASHPTE   0
 382 #endif
 383 #ifndef _PTE_NONE_MASK
 384 #define _PTE_NONE_MASK 0
 385 #endif
 386 #ifndef _PAGE_SHARED
 387 #define _PAGE_SHARED    0
 388 #endif
 389 #ifndef _PAGE_HWWRITE
 390 #define _PAGE_HWWRITE   0
 391 #endif
 392 #ifndef _PAGE_HWEXEC
 393 #define _PAGE_HWEXEC    0
 394 #endif
 395 #ifndef _PAGE_EXEC
 396 #define _PAGE_EXEC      0
 397 #endif
 398 #ifndef _PMD_PRESENT_MASK
 399 #define _PMD_PRESENT_MASK       _PMD_PRESENT
 400 #endif
 401 #ifndef _PMD_SIZE
 402 #define _PMD_SIZE       0
 403 #define PMD_PAGE_SIZE(pmd)      bad_call_to_PMD_PAGE_SIZE()
 404 #endif
 405
 406 #define _PAGE_CHG_MASK  (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY)
 407
 408 /*
 409  * Note: the _PAGE_COHERENT bit automatically gets set in the hardware
 410  * PTE if CONFIG_SMP is defined (hash_page does this); there is no need
 411  * to have it in the Linux PTE, and in fact the bit could be reused for
 412  * another purpose.  -- paulus.
 413  */
 414
 415 #ifdef CONFIG_44x
 416 #define _PAGE_BASE      (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_GUARDED)
 417 #else
 418 #define _PAGE_BASE      (_PAGE_PRESENT | _PAGE_ACCESSED)
 419 #endif
 420 #define _PAGE_WRENABLE  (_PAGE_RW | _PAGE_DIRTY | _PAGE_HWWRITE)
 421 #define _PAGE_KERNEL    (_PAGE_BASE | _PAGE_SHARED | _PAGE_WRENABLE)
 422
 423 #ifdef CONFIG_PPC_STD_MMU
 424 /* On standard PPC MMU, no user access implies kernel read/write access,
 425  * so to write-protect kernel memory we must turn on user access */
 426 #define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED | _PAGE_USER)
 427 #else
 428 #define _PAGE_KERNEL_RO (_PAGE_BASE | _PAGE_SHARED)
 429 #endif
 430
 431 #define _PAGE_IO        (_PAGE_KERNEL | _PAGE_NO_CACHE | _PAGE_GUARDED)
 432 #define _PAGE_RAM       (_PAGE_KERNEL | _PAGE_HWEXEC)
 433
 434 #if defined(CONFIG_KGDB) || defined(CONFIG_XMON) || defined(CONFIG_BDI_SWITCH)
 435 /* We want the debuggers to be able to set breakpoints anywhere, so
 436  * don't write protect the kernel text */
 437 #define _PAGE_RAM_TEXT  _PAGE_RAM
 438 #else
 439 #define _PAGE_RAM_TEXT  (_PAGE_KERNEL_RO | _PAGE_HWEXEC)
 440 #endif
 441
 442 #define PAGE_NONE       __pgprot(_PAGE_BASE)
 443 #define PAGE_READONLY   __pgprot(_PAGE_BASE | _PAGE_USER)
 444 #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
 445 #define PAGE_SHARED     __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW)
 446 #define PAGE_SHARED_X   __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW | _PAGE_EXEC)
 447 #define PAGE_COPY       __pgprot(_PAGE_BASE | _PAGE_USER)
 448 #define PAGE_COPY_X     __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
 449
 450 #define PAGE_KERNEL             __pgprot(_PAGE_RAM)
 451 #define PAGE_KERNEL_NOCACHE     __pgprot(_PAGE_IO)
 452
 453 /*
 454  * The PowerPC can only do execute protection on a segment (256MB) basis,
 455  * not on a page basis.  So we consider execute permission the same as read.
 456  * Also, write permissions imply read permissions.
 457  * This is the closest we can get..
 458  */
 459 #define __P000  PAGE_NONE
 460 #define __P001  PAGE_READONLY_X
 461 #define __P010  PAGE_COPY
 462 #define __P011  PAGE_COPY_X
 463 #define __P100  PAGE_READONLY
 464 #define __P101  PAGE_READONLY_X
 465 #define __P110  PAGE_COPY
 466 #define __P111  PAGE_COPY_X
 467
 468 #define __S000  PAGE_NONE
 469 #define __S001  PAGE_READONLY_X
 470 #define __S010  PAGE_SHARED
 471 #define __S011  PAGE_SHARED_X
 472 #define __S100  PAGE_READONLY
 473 #define __S101  PAGE_READONLY_X
 474 #define __S110  PAGE_SHARED
 475 #define __S111  PAGE_SHARED_X
 476
 477 #ifndef __ASSEMBLY__
 478 /* Make sure we get a link error if PMD_PAGE_SIZE is ever called on a
 479  * kernel without large page PMD support */
 480 extern unsigned long bad_call_to_PMD_PAGE_SIZE(void);
 481
 482 /*
 483  * Conversions between PTE values and page frame numbers.
 484  */
 485
 486 /* in some case we want to additionaly adjust where the pfn is in the pte to
 487  * allow room for more flags */
 488 #if defined(CONFIG_FSL_BOOKE) && defined(CONFIG_PTE_64BIT)
 489 #define PFN_SHIFT_OFFSET        (PAGE_SHIFT + 8)
 490 #else
 491 #define PFN_SHIFT_OFFSET        (PAGE_SHIFT)
 492 #endif
 493
 494 #define pte_pfn(x)              (pte_val(x) >> PFN_SHIFT_OFFSET)
 495 #define pte_page(x)             pfn_to_page(pte_pfn(x))
 496
 497 #define pfn_pte(pfn, prot)      __pte(((pte_basic_t)(pfn) << PFN_SHIFT_OFFSET) |\
 498                                         pgprot_val(prot))
 499 #define mk_pte(page, prot)      pfn_pte(page_to_pfn(page), prot)
 500
 501 /*
 502  * ZERO_PAGE is a global shared page that is always zero: used
 503  * for zero-mapped memory areas etc..
 504  */
 505 extern unsigned long empty_zero_page[1024];
 506 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
 507
 508 #endif /* __ASSEMBLY__ */
 509
 510 #define pte_none(pte)           ((pte_val(pte) & ~_PTE_NONE_MASK) == 0)
 511 #define pte_present(pte)        (pte_val(pte) & _PAGE_PRESENT)
 512 #define pte_clear(mm,addr,ptep) do { set_pte_at((mm), (addr), (ptep), __pte(0)); } while (0)
 513
 514 #define pmd_none(pmd)           (!pmd_val(pmd))
 515 #define pmd_bad(pmd)            (pmd_val(pmd) & _PMD_BAD)
 516 #define pmd_present(pmd)        (pmd_val(pmd) & _PMD_PRESENT_MASK)
 517 #define pmd_clear(pmdp)         do { pmd_val(*(pmdp)) = 0; } while (0)
 518
 519 #ifndef __ASSEMBLY__
 520 /*
 521  * The "pgd_xxx()" functions here are trivial for a folded two-level
 522  * setup: the pgd is never bad, and a pmd always exists (as it's folded
 523  * into the pgd entry)
 524  */
 525 static inline int pgd_none(pgd_t pgd)           { return 0; }
 526 static inline int pgd_bad(pgd_t pgd)            { return 0; }
 527 static inline int pgd_present(pgd_t pgd)        { return 1; }
 528 #define pgd_clear(xp)                           do { } while (0)
 529
 530 #define pgd_page(pgd) \
 531         ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK))
 532
 533 /*
 534  * The following only work if pte_present() is true.
 535  * Undefined behaviour if not..
 536  */
 537 static inline int pte_read(pte_t pte)           { return pte_val(pte) & _PAGE_USER; }
 538 static inline int pte_write(pte_t pte)          { return pte_val(pte) & _PAGE_RW; }
 539 static inline int pte_exec(pte_t pte)           { return pte_val(pte) & _PAGE_EXEC; }
 540 static inline int pte_dirty(pte_t pte)          { return pte_val(pte) & _PAGE_DIRTY; }
 541 static inline int pte_young(pte_t pte)          { return pte_val(pte) & _PAGE_ACCESSED; }
 542 static inline int pte_file(pte_t pte)           { return pte_val(pte) & _PAGE_FILE; }
 543
 544 static inline void pte_uncache(pte_t pte)       { pte_val(pte) |= _PAGE_NO_CACHE; }
 545 static inline void pte_cache(pte_t pte)         { pte_val(pte) &= ~_PAGE_NO_CACHE; }
 546
 547 static inline pte_t pte_rdprotect(pte_t pte) {
 548         pte_val(pte) &= ~_PAGE_USER; return pte; }
 549 static inline pte_t pte_wrprotect(pte_t pte) {
 550         pte_val(pte) &= ~(_PAGE_RW | _PAGE_HWWRITE); return pte; }
 551 static inline pte_t pte_exprotect(pte_t pte) {
 552         pte_val(pte) &= ~_PAGE_EXEC; return pte; }
 553 static inline pte_t pte_mkclean(pte_t pte) {
 554         pte_val(pte) &= ~(_PAGE_DIRTY | _PAGE_HWWRITE); return pte; }
 555 static inline pte_t pte_mkold(pte_t pte) {
 556         pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
 557
 558 static inline pte_t pte_mkread(pte_t pte) {
 559         pte_val(pte) |= _PAGE_USER; return pte; }
 560 static inline pte_t pte_mkexec(pte_t pte) {
 561         pte_val(pte) |= _PAGE_USER | _PAGE_EXEC; return pte; }
 562 static inline pte_t pte_mkwrite(pte_t pte) {
 563         pte_val(pte) |= _PAGE_RW; return pte; }
 564 static inline pte_t pte_mkdirty(pte_t pte) {
 565         pte_val(pte) |= _PAGE_DIRTY; return pte; }
 566 static inline pte_t pte_mkyoung(pte_t pte) {
 567         pte_val(pte) |= _PAGE_ACCESSED; return pte; }
 568
 569 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 570 {
 571         pte_val(pte) = (pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot);
 572         return pte;
 573 }
 574
 575 /*
 576  * When flushing the tlb entry for a page, we also need to flush the hash
 577  * table entry.  flush_hash_pages is assembler (for speed) in hashtable.S.
 578  */
 579 extern int flush_hash_pages(unsigned context, unsigned long va,
 580                             unsigned long pmdval, int count);
 581
 582 /* Add an HPTE to the hash table */
 583 extern void add_hash_page(unsigned context, unsigned long va,
 584                           unsigned long pmdval);
 585
 586 /*
 587  * Atomic PTE updates.
 588  *
 589  * pte_update clears and sets bit atomically, and returns
 590  * the old pte value.  In the 64-bit PTE case we lock around the
 591  * low PTE word since we expect ALL flag bits to be there
 592  */
 593 #ifndef CONFIG_PTE_64BIT
 594 static inline unsigned long pte_update(pte_t *p, unsigned long clr,
 595                                        unsigned long set)
 596 {
 597         unsigned long old, tmp;
 598
 599         __asm__ __volatile__("\
 600 1:      lwarx   %0,0,%3\n\
 601         andc    %1,%0,%4\n\
 602         or      %1,%1,%5\n"
 603         PPC405_ERR77(0,%3)
 604 "       stwcx.  %1,0,%3\n\
 605         bne-    1b"
 606         : "=&r" (old), "=&r" (tmp), "=m" (*p)
 607         : "r" (p), "r" (clr), "r" (set), "m" (*p)
 608         : "cc" );
 609         return old;
 610 }
 611 #else
 612 static inline unsigned long long pte_update(pte_t *p, unsigned long clr,
 613                                        unsigned long set)
 614 {
 615         unsigned long long old;
 616         unsigned long tmp;
 617
 618         __asm__ __volatile__("\
 619 1:      lwarx   %L0,0,%4\n\
 620         lwzx    %0,0,%3\n\
 621         andc    %1,%L0,%5\n\
 622         or      %1,%1,%6\n"
 623         PPC405_ERR77(0,%3)
 624 "       stwcx.  %1,0,%4\n\
 625         bne-    1b"
 626         : "=&r" (old), "=&r" (tmp), "=m" (*p)
 627         : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p)
 628         : "cc" );
 629         return old;
 630 }
 631 #endif
 632
 633 /*
 634  * set_pte stores a linux PTE into the linux page table.
 635  * On machines which use an MMU hash table we avoid changing the
 636  * _PAGE_HASHPTE bit.
 637  */
 638 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
 639                               pte_t *ptep, pte_t pte)
 640 {
 641 #if _PAGE_HASHPTE != 0
 642         pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte) & ~_PAGE_HASHPTE);
 643 #else
 644         *ptep = pte;
 645 #endif
 646 }
 647
 648 /*
 649  * 2.6 calles this without flushing the TLB entry, this is wrong
 650  * for our hash-based implementation, we fix that up here
 651  */
 652 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 653 static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep)
 654 {
 655         unsigned long old;
 656         old = pte_update(ptep, _PAGE_ACCESSED, 0);
 657 #if _PAGE_HASHPTE != 0
 658         if (old & _PAGE_HASHPTE) {
 659                 unsigned long ptephys = __pa(ptep) & PAGE_MASK;
 660                 flush_hash_pages(context, addr, ptephys, 1);
 661         }
 662 #endif
 663         return (old & _PAGE_ACCESSED) != 0;
 664 }
 665 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
 666         __ptep_test_and_clear_young((__vma)->vm_mm->context, __addr, __ptep)
 667
 668 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
 669 static inline int ptep_test_and_clear_dirty(struct vm_area_struct *vma,
 670                                             unsigned long addr, pte_t *ptep)
 671 {
 672         return (pte_update(ptep, (_PAGE_DIRTY | _PAGE_HWWRITE), 0) & _PAGE_DIRTY) != 0;
 673 }
 674
 675 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 676 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 677                                        pte_t *ptep)
 678 {
 679         return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0));
 680 }
 681
 682 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 683 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
 684                                       pte_t *ptep)
 685 {
 686         pte_update(ptep, (_PAGE_RW | _PAGE_HWWRITE), 0);
 687 }
 688
 689 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 690 static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty)
 691 {
 692         unsigned long bits = pte_val(entry) &
 693                 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW);
 694         pte_update(ptep, 0, bits);
 695 }
 696
 697 #define  ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
 698         do {                                                               \
 699                 __ptep_set_access_flags(__ptep, __entry, __dirty);         \
 700                 flush_tlb_page_nohash(__vma, __address);                   \
 701         } while(0)
 702
 703 /*
 704  * Macro to mark a page protection value as "uncacheable".
 705  */
 706 #define pgprot_noncached(prot)  (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
 707
 708 struct file;
 709 extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
 710                                      unsigned long size, pgprot_t vma_prot);
 711 #define __HAVE_PHYS_MEM_ACCESS_PROT
 712
 713 #define __HAVE_ARCH_PTE_SAME
 714 #define pte_same(A,B)   (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0)
 715
 716 /*
 717  * Note that on Book E processors, the pmd contains the kernel virtual
 718  * (lowmem) address of the pte page.  The physical address is less useful
 719  * because everything runs with translation enabled (even the TLB miss
 720  * handler).  On everything else the pmd contains the physical address
 721  * of the pte page.  -- paulus
 722  */
 723 #ifndef CONFIG_BOOKE
 724 #define pmd_page_kernel(pmd)    \
 725         ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
 726 #define pmd_page(pmd)           \
 727         (mem_map + (pmd_val(pmd) >> PAGE_SHIFT))
 728 #else
 729 #define pmd_page_kernel(pmd)    \
 730         ((unsigned long) (pmd_val(pmd) & PAGE_MASK))
 731 #define pmd_page(pmd)           \
 732         (mem_map + (__pa(pmd_val(pmd)) >> PAGE_SHIFT))
 733 #endif
 734
 735 /* to find an entry in a kernel page-table-directory */
 736 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
 737
 738 /* to find an entry in a page-table-directory */
 739 #define pgd_index(address)       ((address) >> PGDIR_SHIFT)
 740 #define pgd_offset(mm, address)  ((mm)->pgd + pgd_index(address))
 741
 742 /* Find an entry in the second-level page table.. */
 743 static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address)
 744 {
 745         return (pmd_t *) dir;
 746 }
 747
 748 /* Find an entry in the third-level page table.. */
 749 #define pte_index(address)              \
 750         (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
 751 #define pte_offset_kernel(dir, addr)    \
 752         ((pte_t *) pmd_page_kernel(*(dir)) + pte_index(addr))
 753 #define pte_offset_map(dir, addr)               \
 754         ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE0) + pte_index(addr))
 755 #define pte_offset_map_nested(dir, addr)        \
 756         ((pte_t *) kmap_atomic(pmd_page(*(dir)), KM_PTE1) + pte_index(addr))
 757
 758 #define pte_unmap(pte)          kunmap_atomic(pte, KM_PTE0)
 759 #define pte_unmap_nested(pte)   kunmap_atomic(pte, KM_PTE1)
 760
 761 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
 762
 763 extern void paging_init(void);
 764
 765 /*
 766  * Encode and decode a swap entry.
 767  * Note that the bits we use in a PTE for representing a swap entry
 768  * must not include the _PAGE_PRESENT bit, the _PAGE_FILE bit, or the
 769  *_PAGE_HASHPTE bit (if used).  -- paulus
 770  */
 771 #define __swp_type(entry)               ((entry).val & 0x1f)
 772 #define __swp_offset(entry)             ((entry).val >> 5)
 773 #define __swp_entry(type, offset)       ((swp_entry_t) { (type) | ((offset) << 5) })
 774 #define __pte_to_swp_entry(pte)         ((swp_entry_t) { pte_val(pte) >> 3 })
 775 #define __swp_entry_to_pte(x)           ((pte_t) { (x).val << 3 })
 776
 777 /* Encode and decode a nonlinear file mapping entry */
 778 #define PTE_FILE_MAX_BITS       29
 779 #define pte_to_pgoff(pte)       (pte_val(pte) >> 3)
 780 #define pgoff_to_pte(off)       ((pte_t) { ((off) << 3) | _PAGE_FILE })
 781
 782 /* CONFIG_APUS */
 783 /* For virtual address to physical address conversion */
 784 extern void cache_clear(__u32 addr, int length);
 785 extern void cache_push(__u32 addr, int length);
 786 extern int mm_end_of_chunk (unsigned long addr, int len);
 787 extern unsigned long iopa(unsigned long addr);
 788 extern unsigned long mm_ptov(unsigned long addr) __attribute_const__;
 789
 790 /* Values for nocacheflag and cmode */
 791 /* These are not used by the APUS kernel_map, but prevents
 792    compilation errors. */
 793 #define KERNELMAP_FULL_CACHING          0
 794 #define KERNELMAP_NOCACHE_SER           1
 795 #define KERNELMAP_NOCACHE_NONSER        2
 796 #define KERNELMAP_NO_COPYBACK           3
 797
 798 /*
 799  * Map some physical address range into the kernel address space.
 800  */
 801 extern unsigned long kernel_map(unsigned long paddr, unsigned long size,
 802                                 int nocacheflag, unsigned long *memavailp );
 803
 804 /*
 805  * Set cache mode of (kernel space) address range.
 806  */
 807 extern void kernel_set_cachemode (unsigned long address, unsigned long size,
 808                                  unsigned int cmode);
 809
 810 /* Needs to be defined here and not in linux/mm.h, as it is arch dependent */
 811 #define kern_addr_valid(addr)   (1)
 812
 813 #ifdef CONFIG_PHYS_64BIT
 814 extern int remap_pfn_range(struct vm_area_struct *vma, unsigned long from,
 815                         unsigned long paddr, unsigned long size, pgprot_t prot);
 816
 817 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
 818                                         unsigned long vaddr,
 819                                         unsigned long pfn,
 820                                         unsigned long size,
 821                                         pgprot_t prot)
 822 {
 823         phys_addr_t paddr64 = fixup_bigphys_addr(pfn << PAGE_SHIFT, size);
 824         return remap_pfn_range(vma, vaddr, paddr64 >> PAGE_SHIFT, size, prot);
 825 }
 826 #else
 827 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot)         \
 828                 remap_pfn_range(vma, vaddr, pfn, size, prot)
 829 #endif
 830
 831 #define MK_IOSPACE_PFN(space, pfn)      (pfn)
 832 #define GET_IOSPACE(pfn)                0
 833 #define GET_PFN(pfn)                    (pfn)
 834
 835 /*
 836  * No page table caches to initialise
 837  */
 838 #define pgtable_cache_init()    do { } while (0)
 839
 840 extern int get_pteptr(struct mm_struct *mm, unsigned long addr, pte_t **ptep);
 841
 842 #include <asm-generic/pgtable.h>
 843
 844 #endif /* !__ASSEMBLY__ */
 845
 846 #endif /* _PPC_PGTABLE_H */
 847 #endif /* __KERNEL__ */