1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _ASM_POWERPC_BOOK3S_32_PGTABLE_H
3 #define _ASM_POWERPC_BOOK3S_32_PGTABLE_H
5 #define __ARCH_USE_5LEVEL_HACK
6 #include <asm-generic/pgtable-nopmd.h>
8 #include <asm/book3s/32/hash.h>
10 /* And here we include common definitions */
12 #define _PAGE_KERNEL_RO 0
13 #define _PAGE_KERNEL_ROX (_PAGE_EXEC)
14 #define _PAGE_KERNEL_RW (_PAGE_DIRTY | _PAGE_RW)
15 #define _PAGE_KERNEL_RWX (_PAGE_DIRTY | _PAGE_RW | _PAGE_EXEC)
17 #define _PAGE_HPTEFLAGS _PAGE_HASHPTE
21 static inline bool pte_user(pte_t pte)
23 return pte_val(pte) & _PAGE_USER;
25 #endif /* __ASSEMBLY__ */
28 * Location of the PFN in the PTE. Most 32-bit platforms use the same
29 * as _PAGE_SHIFT here (ie, naturally aligned).
30 * Platform who don't just pre-define the value so we don't override it here.
32 #define PTE_RPN_SHIFT (PAGE_SHIFT)
35 * The mask covered by the RPN must be a ULL on 32-bit platforms with
38 #ifdef CONFIG_PTE_64BIT
39 #define PTE_RPN_MASK (~((1ULL << PTE_RPN_SHIFT) - 1))
41 #define PTE_RPN_MASK (~((1UL << PTE_RPN_SHIFT) - 1))
45 * _PAGE_CHG_MASK masks of bits that are to be preserved across
48 #define _PAGE_CHG_MASK (PTE_RPN_MASK | _PAGE_HASHPTE | _PAGE_DIRTY | \
49 _PAGE_ACCESSED | _PAGE_SPECIAL)
52 * We define 2 sets of base prot bits, one for basic pages (ie,
53 * cacheable kernel and user pages) and one for non cacheable
54 * pages. We always set _PAGE_COHERENT when SMP is enabled or
55 * the processor might need it for DMA coherency.
57 #define _PAGE_BASE_NC (_PAGE_PRESENT | _PAGE_ACCESSED)
58 #define _PAGE_BASE (_PAGE_BASE_NC | _PAGE_COHERENT)
61 * Permission masks used to generate the __P and __S table.
63 * Note:__pgprot is defined in arch/powerpc/include/asm/page.h
65 * Write permissions imply read permissions for now.
67 #define PAGE_NONE __pgprot(_PAGE_BASE)
68 #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW)
69 #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_RW | _PAGE_EXEC)
70 #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
71 #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
72 #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
73 #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
75 /* Permission masks used for kernel mappings */
76 #define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_KERNEL_RW)
77 #define PAGE_KERNEL_NC __pgprot(_PAGE_BASE_NC | _PAGE_KERNEL_RW | _PAGE_NO_CACHE)
78 #define PAGE_KERNEL_NCG __pgprot(_PAGE_BASE_NC | _PAGE_KERNEL_RW | \
79 _PAGE_NO_CACHE | _PAGE_GUARDED)
80 #define PAGE_KERNEL_X __pgprot(_PAGE_BASE | _PAGE_KERNEL_RWX)
81 #define PAGE_KERNEL_RO __pgprot(_PAGE_BASE | _PAGE_KERNEL_RO)
82 #define PAGE_KERNEL_ROX __pgprot(_PAGE_BASE | _PAGE_KERNEL_ROX)
85 * Protection used for kernel text. We want the debuggers to be able to
86 * set breakpoints anywhere, so don't write protect the kernel text
87 * on platforms where such control is possible.
89 #if defined(CONFIG_KGDB) || defined(CONFIG_XMON) || defined(CONFIG_BDI_SWITCH) ||\
90 defined(CONFIG_KPROBES) || defined(CONFIG_DYNAMIC_FTRACE)
91 #define PAGE_KERNEL_TEXT PAGE_KERNEL_X
93 #define PAGE_KERNEL_TEXT PAGE_KERNEL_ROX
96 /* Make modules code happy. We don't set RO yet */
97 #define PAGE_KERNEL_EXEC PAGE_KERNEL_X
99 /* Advertise special mapping type for AGP */
100 #define PAGE_AGP (PAGE_KERNEL_NC)
101 #define HAVE_PAGE_AGP
103 #define PTE_INDEX_SIZE PTE_SHIFT
104 #define PMD_INDEX_SIZE 0
105 #define PUD_INDEX_SIZE 0
106 #define PGD_INDEX_SIZE (32 - PGDIR_SHIFT)
108 #define PMD_CACHE_INDEX PMD_INDEX_SIZE
109 #define PUD_CACHE_INDEX PUD_INDEX_SIZE
112 #define PTE_TABLE_SIZE (sizeof(pte_t) << PTE_INDEX_SIZE)
113 #define PMD_TABLE_SIZE 0
114 #define PUD_TABLE_SIZE 0
115 #define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
116 #endif /* __ASSEMBLY__ */
118 #define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
119 #define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
122 * The normal case is that PTEs are 32-bits and we have a 1-page
123 * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages. -- paulus
125 * For any >32-bit physical address platform, we can use the following
126 * two level page table layout where the pgdir is 8KB and the MS 13 bits
127 * are an index to the second level table. The combined pgdir/pmd first
128 * level has 2048 entries and the second level has 512 64-bit PTE entries.
131 /* PGDIR_SHIFT determines what a top-level page table entry can map */
132 #define PGDIR_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
133 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
134 #define PGDIR_MASK (~(PGDIR_SIZE-1))
136 #define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE)
140 int map_kernel_page(unsigned long va, phys_addr_t pa, pgprot_t prot);
142 #endif /* !__ASSEMBLY__ */
145 * This is the bottom of the PKMAP area with HIGHMEM or an arbitrary
146 * value (for now) on others, from where we can start layout kernel
147 * virtual space that goes below PKMAP and FIXMAP
149 #include <asm/fixmap.h>
151 #ifdef CONFIG_HIGHMEM
152 #define KVIRT_TOP PKMAP_BASE
154 #define KVIRT_TOP FIXADDR_START
158 * ioremap_bot starts at that address. Early ioremaps move down from there,
159 * until mem_init() at which point this becomes the top of the vmalloc
162 #ifdef CONFIG_NOT_COHERENT_CACHE
163 #define IOREMAP_TOP ((KVIRT_TOP - CONFIG_CONSISTENT_SIZE) & PAGE_MASK)
165 #define IOREMAP_TOP KVIRT_TOP
169 * Just any arbitrary offset to the start of the vmalloc VM area: the
170 * current 16MB value just means that there will be a 64MB "hole" after the
171 * physical memory until the kernel virtual memory starts. That means that
172 * any out-of-bounds memory accesses will hopefully be caught.
173 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
174 * area for the same reason. ;)
176 * We no longer map larger than phys RAM with the BATs so we don't have
177 * to worry about the VMALLOC_OFFSET causing problems. We do have to worry
178 * about clashes between our early calls to ioremap() that start growing down
179 * from ioremap_base being run into the VM area allocations (growing upwards
180 * from VMALLOC_START). For this reason we have ioremap_bot to check when
181 * we actually run into our mappings setup in the early boot with the VM
182 * system. This really does become a problem for machines with good amounts
185 #define VMALLOC_OFFSET (0x1000000) /* 16M */
188 * With CONFIG_STRICT_KERNEL_RWX, kernel segments are set NX. But when modules
189 * are used, NX cannot be set on VMALLOC space. So vmalloc VM space and linear
190 * memory shall not share segments.
192 #if defined(CONFIG_STRICT_KERNEL_RWX) && defined(CONFIG_MODULES)
193 #define VMALLOC_START ((_ALIGN((long)high_memory, 256L << 20) + VMALLOC_OFFSET) & \
194 ~(VMALLOC_OFFSET - 1))
196 #define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
198 #define VMALLOC_END ioremap_bot
201 #include <linux/sched.h>
202 #include <linux/threads.h>
204 extern unsigned long ioremap_bot;
206 /* Bits to mask out from a PGD to get to the PUD page */
207 #define PGD_MASKED_BITS 0
209 #define pte_ERROR(e) \
210 pr_err("%s:%d: bad pte %llx.\n", __FILE__, __LINE__, \
211 (unsigned long long)pte_val(e))
212 #define pgd_ERROR(e) \
213 pr_err("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
215 * Bits in a linux-style PTE. These match the bits in the
216 * (hardware-defined) PowerPC PTE as closely as possible.
219 #define pte_clear(mm, addr, ptep) \
220 do { pte_update(ptep, ~_PAGE_HASHPTE, 0); } while (0)
222 #define pmd_none(pmd) (!pmd_val(pmd))
223 #define pmd_bad(pmd) (pmd_val(pmd) & _PMD_BAD)
224 #define pmd_present(pmd) (pmd_val(pmd) & _PMD_PRESENT_MASK)
225 static inline void pmd_clear(pmd_t *pmdp)
232 * When flushing the tlb entry for a page, we also need to flush the hash
233 * table entry. flush_hash_pages is assembler (for speed) in hashtable.S.
235 extern int flush_hash_pages(unsigned context, unsigned long va,
236 unsigned long pmdval, int count);
238 /* Add an HPTE to the hash table */
239 extern void add_hash_page(unsigned context, unsigned long va,
240 unsigned long pmdval);
242 /* Flush an entry from the TLB/hash table */
243 extern void flush_hash_entry(struct mm_struct *mm, pte_t *ptep,
244 unsigned long address);
247 * PTE updates. This function is called whenever an existing
248 * valid PTE is updated. This does -not- include set_pte_at()
249 * which nowadays only sets a new PTE.
251 * Depending on the type of MMU, we may need to use atomic updates
252 * and the PTE may be either 32 or 64 bit wide. In the later case,
253 * when using atomic updates, only the low part of the PTE is
254 * accessed atomically.
256 * In addition, on 44x, we also maintain a global flag indicating
257 * that an executable user mapping was modified, which is needed
258 * to properly flush the virtually tagged instruction cache of
259 * those implementations.
261 #ifndef CONFIG_PTE_64BIT
262 static inline unsigned long pte_update(pte_t *p,
266 unsigned long old, tmp;
268 __asm__ __volatile__("\
274 : "=&r" (old), "=&r" (tmp), "=m" (*p)
275 : "r" (p), "r" (clr), "r" (set), "m" (*p)
280 #else /* CONFIG_PTE_64BIT */
281 static inline unsigned long long pte_update(pte_t *p,
285 unsigned long long old;
288 __asm__ __volatile__("\
295 : "=&r" (old), "=&r" (tmp), "=m" (*p)
296 : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p)
301 #endif /* CONFIG_PTE_64BIT */
304 * 2.6 calls this without flushing the TLB entry; this is wrong
305 * for our hash-based implementation, we fix that up here.
307 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
308 static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep)
311 old = pte_update(ptep, _PAGE_ACCESSED, 0);
312 if (old & _PAGE_HASHPTE) {
313 unsigned long ptephys = __pa(ptep) & PAGE_MASK;
314 flush_hash_pages(context, addr, ptephys, 1);
316 return (old & _PAGE_ACCESSED) != 0;
318 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
319 __ptep_test_and_clear_young((__vma)->vm_mm->context.id, __addr, __ptep)
321 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
322 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
325 return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0));
328 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
329 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
332 pte_update(ptep, _PAGE_RW, 0);
335 static inline void __ptep_set_access_flags(struct vm_area_struct *vma,
336 pte_t *ptep, pte_t entry,
337 unsigned long address,
340 unsigned long set = pte_val(entry) &
341 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
343 pte_update(ptep, 0, set);
345 flush_tlb_page(vma, address);
348 #define __HAVE_ARCH_PTE_SAME
349 #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0)
351 #define pmd_page_vaddr(pmd) \
352 ((unsigned long)__va(pmd_val(pmd) & ~(PTE_TABLE_SIZE - 1)))
353 #define pmd_page(pmd) \
354 pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
356 /* to find an entry in a kernel page-table-directory */
357 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
359 /* to find an entry in a page-table-directory */
360 #define pgd_index(address) ((address) >> PGDIR_SHIFT)
361 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
363 /* Find an entry in the third-level page table.. */
364 #define pte_index(address) \
365 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
366 #define pte_offset_kernel(dir, addr) \
367 ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr))
368 #define pte_offset_map(dir, addr) \
369 ((pte_t *)(kmap_atomic(pmd_page(*(dir))) + \
370 (pmd_page_vaddr(*(dir)) & ~PAGE_MASK)) + pte_index(addr))
371 #define pte_unmap(pte) kunmap_atomic(pte)
374 * Encode and decode a swap entry.
375 * Note that the bits we use in a PTE for representing a swap entry
376 * must not include the _PAGE_PRESENT bit or the _PAGE_HASHPTE bit (if used).
379 #define __swp_type(entry) ((entry).val & 0x1f)
380 #define __swp_offset(entry) ((entry).val >> 5)
381 #define __swp_entry(type, offset) ((swp_entry_t) { (type) | ((offset) << 5) })
382 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> 3 })
383 #define __swp_entry_to_pte(x) ((pte_t) { (x).val << 3 })
385 /* Generic accessors to PTE bits */
386 static inline int pte_write(pte_t pte) { return !!(pte_val(pte) & _PAGE_RW);}
387 static inline int pte_read(pte_t pte) { return 1; }
388 static inline int pte_dirty(pte_t pte) { return !!(pte_val(pte) & _PAGE_DIRTY); }
389 static inline int pte_young(pte_t pte) { return !!(pte_val(pte) & _PAGE_ACCESSED); }
390 static inline int pte_special(pte_t pte) { return !!(pte_val(pte) & _PAGE_SPECIAL); }
391 static inline int pte_none(pte_t pte) { return (pte_val(pte) & ~_PTE_NONE_MASK) == 0; }
392 static inline bool pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC; }
394 static inline int pte_present(pte_t pte)
396 return pte_val(pte) & _PAGE_PRESENT;
399 static inline bool pte_hw_valid(pte_t pte)
401 return pte_val(pte) & _PAGE_PRESENT;
404 static inline bool pte_hashpte(pte_t pte)
406 return !!(pte_val(pte) & _PAGE_HASHPTE);
409 static inline bool pte_ci(pte_t pte)
411 return !!(pte_val(pte) & _PAGE_NO_CACHE);
415 * We only find page table entry in the last level
416 * Hence no need for other accessors
418 #define pte_access_permitted pte_access_permitted
419 static inline bool pte_access_permitted(pte_t pte, bool write)
422 * A read-only access is controlled by _PAGE_USER bit.
423 * We have _PAGE_READ set for WRITE and EXECUTE
425 if (!pte_present(pte) || !pte_user(pte) || !pte_read(pte))
428 if (write && !pte_write(pte))
434 /* Conversion functions: convert a page and protection to a page entry,
435 * and a page entry and page directory to the page they refer to.
437 * Even if PTEs can be unsigned long long, a PFN is always an unsigned
440 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
442 return __pte(((pte_basic_t)(pfn) << PTE_RPN_SHIFT) |
446 static inline unsigned long pte_pfn(pte_t pte)
448 return pte_val(pte) >> PTE_RPN_SHIFT;
451 /* Generic modifiers for PTE bits */
452 static inline pte_t pte_wrprotect(pte_t pte)
454 return __pte(pte_val(pte) & ~_PAGE_RW);
457 static inline pte_t pte_exprotect(pte_t pte)
459 return __pte(pte_val(pte) & ~_PAGE_EXEC);
462 static inline pte_t pte_mkclean(pte_t pte)
464 return __pte(pte_val(pte) & ~_PAGE_DIRTY);
467 static inline pte_t pte_mkold(pte_t pte)
469 return __pte(pte_val(pte) & ~_PAGE_ACCESSED);
472 static inline pte_t pte_mkexec(pte_t pte)
474 return __pte(pte_val(pte) | _PAGE_EXEC);
477 static inline pte_t pte_mkpte(pte_t pte)
482 static inline pte_t pte_mkwrite(pte_t pte)
484 return __pte(pte_val(pte) | _PAGE_RW);
487 static inline pte_t pte_mkdirty(pte_t pte)
489 return __pte(pte_val(pte) | _PAGE_DIRTY);
492 static inline pte_t pte_mkyoung(pte_t pte)
494 return __pte(pte_val(pte) | _PAGE_ACCESSED);
497 static inline pte_t pte_mkspecial(pte_t pte)
499 return __pte(pte_val(pte) | _PAGE_SPECIAL);
502 static inline pte_t pte_mkhuge(pte_t pte)
507 static inline pte_t pte_mkprivileged(pte_t pte)
509 return __pte(pte_val(pte) & ~_PAGE_USER);
512 static inline pte_t pte_mkuser(pte_t pte)
514 return __pte(pte_val(pte) | _PAGE_USER);
517 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
519 return __pte((pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot));
524 /* This low level function performs the actual PTE insertion
525 * Setting the PTE depends on the MMU type and other factors. It's
526 * an horrible mess that I'm not going to try to clean up now but
527 * I'm keeping it in one place rather than spread around
529 static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
530 pte_t *ptep, pte_t pte, int percpu)
532 #if defined(CONFIG_SMP) && !defined(CONFIG_PTE_64BIT)
533 /* First case is 32-bit Hash MMU in SMP mode with 32-bit PTEs. We use the
534 * helper pte_update() which does an atomic update. We need to do that
535 * because a concurrent invalidation can clear _PAGE_HASHPTE. If it's a
536 * per-CPU PTE such as a kmap_atomic, we do a simple update preserving
537 * the hash bits instead (ie, same as the non-SMP case)
540 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
541 | (pte_val(pte) & ~_PAGE_HASHPTE));
543 pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte));
545 #elif defined(CONFIG_PTE_64BIT)
546 /* Second case is 32-bit with 64-bit PTE. In this case, we
547 * can just store as long as we do the two halves in the right order
548 * with a barrier in between. This is possible because we take care,
549 * in the hash code, to pre-invalidate if the PTE was already hashed,
550 * which synchronizes us with any concurrent invalidation.
551 * In the percpu case, we also fallback to the simple update preserving
555 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
556 | (pte_val(pte) & ~_PAGE_HASHPTE));
559 if (pte_val(*ptep) & _PAGE_HASHPTE)
560 flush_hash_entry(mm, ptep, addr);
561 __asm__ __volatile__("\
565 : "=m" (*ptep), "=m" (*((unsigned char *)ptep+4))
566 : "r" (pte) : "memory");
569 /* Third case is 32-bit hash table in UP mode, we need to preserve
570 * the _PAGE_HASHPTE bit since we may not have invalidated the previous
571 * translation in the hash yet (done in a subsequent flush_tlb_xxx())
572 * and see we need to keep track that this PTE needs invalidating
574 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
575 | (pte_val(pte) & ~_PAGE_HASHPTE));
580 * Macro to mark a page protection value as "uncacheable".
583 #define _PAGE_CACHE_CTL (_PAGE_COHERENT | _PAGE_GUARDED | _PAGE_NO_CACHE | \
586 #define pgprot_noncached pgprot_noncached
587 static inline pgprot_t pgprot_noncached(pgprot_t prot)
589 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
590 _PAGE_NO_CACHE | _PAGE_GUARDED);
593 #define pgprot_noncached_wc pgprot_noncached_wc
594 static inline pgprot_t pgprot_noncached_wc(pgprot_t prot)
596 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
600 #define pgprot_cached pgprot_cached
601 static inline pgprot_t pgprot_cached(pgprot_t prot)
603 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
607 #define pgprot_cached_wthru pgprot_cached_wthru
608 static inline pgprot_t pgprot_cached_wthru(pgprot_t prot)
610 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
611 _PAGE_COHERENT | _PAGE_WRITETHRU);
614 #define pgprot_cached_noncoherent pgprot_cached_noncoherent
615 static inline pgprot_t pgprot_cached_noncoherent(pgprot_t prot)
617 return __pgprot(pgprot_val(prot) & ~_PAGE_CACHE_CTL);
620 #define pgprot_writecombine pgprot_writecombine
621 static inline pgprot_t pgprot_writecombine(pgprot_t prot)
623 return pgprot_noncached_wc(prot);
626 #endif /* !__ASSEMBLY__ */
628 #endif /* _ASM_POWERPC_BOOK3S_32_PGTABLE_H */