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 */
11 #include <asm/pte-common.h>
13 #define PTE_INDEX_SIZE PTE_SHIFT
14 #define PMD_INDEX_SIZE 0
15 #define PUD_INDEX_SIZE 0
16 #define PGD_INDEX_SIZE (32 - PGDIR_SHIFT)
18 #define PMD_CACHE_INDEX PMD_INDEX_SIZE
19 #define PUD_CACHE_INDEX PUD_INDEX_SIZE
22 #define PTE_TABLE_SIZE (sizeof(pte_t) << PTE_INDEX_SIZE)
23 #define PMD_TABLE_SIZE 0
24 #define PUD_TABLE_SIZE 0
25 #define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
26 #endif /* __ASSEMBLY__ */
28 #define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
29 #define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
32 * The normal case is that PTEs are 32-bits and we have a 1-page
33 * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages. -- paulus
35 * For any >32-bit physical address platform, we can use the following
36 * two level page table layout where the pgdir is 8KB and the MS 13 bits
37 * are an index to the second level table. The combined pgdir/pmd first
38 * level has 2048 entries and the second level has 512 64-bit PTE entries.
41 /* PGDIR_SHIFT determines what a top-level page table entry can map */
42 #define PGDIR_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
43 #define PGDIR_SIZE (1UL << PGDIR_SHIFT)
44 #define PGDIR_MASK (~(PGDIR_SIZE-1))
46 #define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE)
48 * This is the bottom of the PKMAP area with HIGHMEM or an arbitrary
49 * value (for now) on others, from where we can start layout kernel
50 * virtual space that goes below PKMAP and FIXMAP
53 #define KVIRT_TOP PKMAP_BASE
55 #define KVIRT_TOP (0xfe000000UL) /* for now, could be FIXMAP_BASE ? */
59 * ioremap_bot starts at that address. Early ioremaps move down from there,
60 * until mem_init() at which point this becomes the top of the vmalloc
63 #ifdef CONFIG_NOT_COHERENT_CACHE
64 #define IOREMAP_TOP ((KVIRT_TOP - CONFIG_CONSISTENT_SIZE) & PAGE_MASK)
66 #define IOREMAP_TOP KVIRT_TOP
70 * Just any arbitrary offset to the start of the vmalloc VM area: the
71 * current 16MB value just means that there will be a 64MB "hole" after the
72 * physical memory until the kernel virtual memory starts. That means that
73 * any out-of-bounds memory accesses will hopefully be caught.
74 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
75 * area for the same reason. ;)
77 * We no longer map larger than phys RAM with the BATs so we don't have
78 * to worry about the VMALLOC_OFFSET causing problems. We do have to worry
79 * about clashes between our early calls to ioremap() that start growing down
80 * from ioremap_base being run into the VM area allocations (growing upwards
81 * from VMALLOC_START). For this reason we have ioremap_bot to check when
82 * we actually run into our mappings setup in the early boot with the VM
83 * system. This really does become a problem for machines with good amounts
86 #define VMALLOC_OFFSET (0x1000000) /* 16M */
87 #define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
88 #define VMALLOC_END ioremap_bot
91 #include <linux/sched.h>
92 #include <linux/threads.h>
94 extern unsigned long ioremap_bot;
96 /* Bits to mask out from a PGD to get to the PUD page */
97 #define PGD_MASKED_BITS 0
99 #define pte_ERROR(e) \
100 pr_err("%s:%d: bad pte %llx.\n", __FILE__, __LINE__, \
101 (unsigned long long)pte_val(e))
102 #define pgd_ERROR(e) \
103 pr_err("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
105 * Bits in a linux-style PTE. These match the bits in the
106 * (hardware-defined) PowerPC PTE as closely as possible.
109 #define pte_clear(mm, addr, ptep) \
110 do { pte_update(ptep, ~_PAGE_HASHPTE, 0); } while (0)
112 #define pmd_none(pmd) (!pmd_val(pmd))
113 #define pmd_bad(pmd) (pmd_val(pmd) & _PMD_BAD)
114 #define pmd_present(pmd) (pmd_val(pmd) & _PMD_PRESENT_MASK)
115 static inline void pmd_clear(pmd_t *pmdp)
122 * When flushing the tlb entry for a page, we also need to flush the hash
123 * table entry. flush_hash_pages is assembler (for speed) in hashtable.S.
125 extern int flush_hash_pages(unsigned context, unsigned long va,
126 unsigned long pmdval, int count);
128 /* Add an HPTE to the hash table */
129 extern void add_hash_page(unsigned context, unsigned long va,
130 unsigned long pmdval);
132 /* Flush an entry from the TLB/hash table */
133 extern void flush_hash_entry(struct mm_struct *mm, pte_t *ptep,
134 unsigned long address);
137 * PTE updates. This function is called whenever an existing
138 * valid PTE is updated. This does -not- include set_pte_at()
139 * which nowadays only sets a new PTE.
141 * Depending on the type of MMU, we may need to use atomic updates
142 * and the PTE may be either 32 or 64 bit wide. In the later case,
143 * when using atomic updates, only the low part of the PTE is
144 * accessed atomically.
146 * In addition, on 44x, we also maintain a global flag indicating
147 * that an executable user mapping was modified, which is needed
148 * to properly flush the virtually tagged instruction cache of
149 * those implementations.
151 #ifndef CONFIG_PTE_64BIT
152 static inline unsigned long pte_update(pte_t *p,
156 unsigned long old, tmp;
158 __asm__ __volatile__("\
164 : "=&r" (old), "=&r" (tmp), "=m" (*p)
165 : "r" (p), "r" (clr), "r" (set), "m" (*p)
170 #else /* CONFIG_PTE_64BIT */
171 static inline unsigned long long pte_update(pte_t *p,
175 unsigned long long old;
178 __asm__ __volatile__("\
185 : "=&r" (old), "=&r" (tmp), "=m" (*p)
186 : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p)
191 #endif /* CONFIG_PTE_64BIT */
194 * 2.6 calls this without flushing the TLB entry; this is wrong
195 * for our hash-based implementation, we fix that up here.
197 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
198 static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep)
201 old = pte_update(ptep, _PAGE_ACCESSED, 0);
202 if (old & _PAGE_HASHPTE) {
203 unsigned long ptephys = __pa(ptep) & PAGE_MASK;
204 flush_hash_pages(context, addr, ptephys, 1);
206 return (old & _PAGE_ACCESSED) != 0;
208 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
209 __ptep_test_and_clear_young((__vma)->vm_mm->context.id, __addr, __ptep)
211 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
212 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
215 return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0));
218 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
219 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
222 pte_update(ptep, (_PAGE_RW | _PAGE_HWWRITE), _PAGE_RO);
224 static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
225 unsigned long addr, pte_t *ptep)
227 ptep_set_wrprotect(mm, addr, ptep);
231 static inline void __ptep_set_access_flags(struct vm_area_struct *vma,
232 pte_t *ptep, pte_t entry,
233 unsigned long address,
236 unsigned long set = pte_val(entry) &
237 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
238 unsigned long clr = ~pte_val(entry) & _PAGE_RO;
240 pte_update(ptep, clr, set);
242 flush_tlb_page(vma, address);
245 #define __HAVE_ARCH_PTE_SAME
246 #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0)
249 * Note that on Book E processors, the pmd contains the kernel virtual
250 * (lowmem) address of the pte page. The physical address is less useful
251 * because everything runs with translation enabled (even the TLB miss
252 * handler). On everything else the pmd contains the physical address
253 * of the pte page. -- paulus
256 #define pmd_page_vaddr(pmd) \
257 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
258 #define pmd_page(pmd) \
259 pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
261 #define pmd_page_vaddr(pmd) \
262 ((unsigned long) (pmd_val(pmd) & PAGE_MASK))
263 #define pmd_page(pmd) \
264 pfn_to_page((__pa(pmd_val(pmd)) >> PAGE_SHIFT))
267 /* to find an entry in a kernel page-table-directory */
268 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
270 /* to find an entry in a page-table-directory */
271 #define pgd_index(address) ((address) >> PGDIR_SHIFT)
272 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
274 /* Find an entry in the third-level page table.. */
275 #define pte_index(address) \
276 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
277 #define pte_offset_kernel(dir, addr) \
278 ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr))
279 #define pte_offset_map(dir, addr) \
280 ((pte_t *) kmap_atomic(pmd_page(*(dir))) + pte_index(addr))
281 #define pte_unmap(pte) kunmap_atomic(pte)
284 * Encode and decode a swap entry.
285 * Note that the bits we use in a PTE for representing a swap entry
286 * must not include the _PAGE_PRESENT bit or the _PAGE_HASHPTE bit (if used).
289 #define __swp_type(entry) ((entry).val & 0x1f)
290 #define __swp_offset(entry) ((entry).val >> 5)
291 #define __swp_entry(type, offset) ((swp_entry_t) { (type) | ((offset) << 5) })
292 #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> 3 })
293 #define __swp_entry_to_pte(x) ((pte_t) { (x).val << 3 })
295 int map_kernel_page(unsigned long va, phys_addr_t pa, int flags);
297 /* Generic accessors to PTE bits */
298 static inline int pte_write(pte_t pte) { return !!(pte_val(pte) & _PAGE_RW);}
299 static inline int pte_read(pte_t pte) { return 1; }
300 static inline int pte_dirty(pte_t pte) { return !!(pte_val(pte) & _PAGE_DIRTY); }
301 static inline int pte_young(pte_t pte) { return !!(pte_val(pte) & _PAGE_ACCESSED); }
302 static inline int pte_special(pte_t pte) { return !!(pte_val(pte) & _PAGE_SPECIAL); }
303 static inline int pte_none(pte_t pte) { return (pte_val(pte) & ~_PTE_NONE_MASK) == 0; }
304 static inline pgprot_t pte_pgprot(pte_t pte) { return __pgprot(pte_val(pte) & PAGE_PROT_BITS); }
306 static inline int pte_present(pte_t pte)
308 return pte_val(pte) & _PAGE_PRESENT;
312 * We only find page table entry in the last level
313 * Hence no need for other accessors
315 #define pte_access_permitted pte_access_permitted
316 static inline bool pte_access_permitted(pte_t pte, bool write)
318 unsigned long pteval = pte_val(pte);
320 * A read-only access is controlled by _PAGE_USER bit.
321 * We have _PAGE_READ set for WRITE and EXECUTE
323 unsigned long need_pte_bits = _PAGE_PRESENT | _PAGE_USER;
326 need_pte_bits |= _PAGE_WRITE;
328 if ((pteval & need_pte_bits) != need_pte_bits)
334 /* Conversion functions: convert a page and protection to a page entry,
335 * and a page entry and page directory to the page they refer to.
337 * Even if PTEs can be unsigned long long, a PFN is always an unsigned
340 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
342 return __pte(((pte_basic_t)(pfn) << PTE_RPN_SHIFT) |
346 static inline unsigned long pte_pfn(pte_t pte)
348 return pte_val(pte) >> PTE_RPN_SHIFT;
351 /* Generic modifiers for PTE bits */
352 static inline pte_t pte_wrprotect(pte_t pte)
354 return __pte(pte_val(pte) & ~_PAGE_RW);
357 static inline pte_t pte_mkclean(pte_t pte)
359 return __pte(pte_val(pte) & ~_PAGE_DIRTY);
362 static inline pte_t pte_mkold(pte_t pte)
364 return __pte(pte_val(pte) & ~_PAGE_ACCESSED);
367 static inline pte_t pte_mkwrite(pte_t pte)
369 return __pte(pte_val(pte) | _PAGE_RW);
372 static inline pte_t pte_mkdirty(pte_t pte)
374 return __pte(pte_val(pte) | _PAGE_DIRTY);
377 static inline pte_t pte_mkyoung(pte_t pte)
379 return __pte(pte_val(pte) | _PAGE_ACCESSED);
382 static inline pte_t pte_mkspecial(pte_t pte)
384 return __pte(pte_val(pte) | _PAGE_SPECIAL);
387 static inline pte_t pte_mkhuge(pte_t pte)
392 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
394 return __pte((pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot));
399 /* This low level function performs the actual PTE insertion
400 * Setting the PTE depends on the MMU type and other factors. It's
401 * an horrible mess that I'm not going to try to clean up now but
402 * I'm keeping it in one place rather than spread around
404 static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
405 pte_t *ptep, pte_t pte, int percpu)
407 #if defined(CONFIG_PPC_STD_MMU_32) && defined(CONFIG_SMP) && !defined(CONFIG_PTE_64BIT)
408 /* First case is 32-bit Hash MMU in SMP mode with 32-bit PTEs. We use the
409 * helper pte_update() which does an atomic update. We need to do that
410 * because a concurrent invalidation can clear _PAGE_HASHPTE. If it's a
411 * per-CPU PTE such as a kmap_atomic, we do a simple update preserving
412 * the hash bits instead (ie, same as the non-SMP case)
415 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
416 | (pte_val(pte) & ~_PAGE_HASHPTE));
418 pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte));
420 #elif defined(CONFIG_PPC32) && defined(CONFIG_PTE_64BIT)
421 /* Second case is 32-bit with 64-bit PTE. In this case, we
422 * can just store as long as we do the two halves in the right order
423 * with a barrier in between. This is possible because we take care,
424 * in the hash code, to pre-invalidate if the PTE was already hashed,
425 * which synchronizes us with any concurrent invalidation.
426 * In the percpu case, we also fallback to the simple update preserving
430 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
431 | (pte_val(pte) & ~_PAGE_HASHPTE));
434 if (pte_val(*ptep) & _PAGE_HASHPTE)
435 flush_hash_entry(mm, ptep, addr);
436 __asm__ __volatile__("\
440 : "=m" (*ptep), "=m" (*((unsigned char *)ptep+4))
441 : "r" (pte) : "memory");
443 #elif defined(CONFIG_PPC_STD_MMU_32)
444 /* Third case is 32-bit hash table in UP mode, we need to preserve
445 * the _PAGE_HASHPTE bit since we may not have invalidated the previous
446 * translation in the hash yet (done in a subsequent flush_tlb_xxx())
447 * and see we need to keep track that this PTE needs invalidating
449 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
450 | (pte_val(pte) & ~_PAGE_HASHPTE));
453 #error "Not supported "
458 * Macro to mark a page protection value as "uncacheable".
461 #define _PAGE_CACHE_CTL (_PAGE_COHERENT | _PAGE_GUARDED | _PAGE_NO_CACHE | \
464 #define pgprot_noncached pgprot_noncached
465 static inline pgprot_t pgprot_noncached(pgprot_t prot)
467 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
468 _PAGE_NO_CACHE | _PAGE_GUARDED);
471 #define pgprot_noncached_wc pgprot_noncached_wc
472 static inline pgprot_t pgprot_noncached_wc(pgprot_t prot)
474 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
478 #define pgprot_cached pgprot_cached
479 static inline pgprot_t pgprot_cached(pgprot_t prot)
481 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
485 #define pgprot_cached_wthru pgprot_cached_wthru
486 static inline pgprot_t pgprot_cached_wthru(pgprot_t prot)
488 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
489 _PAGE_COHERENT | _PAGE_WRITETHRU);
492 #define pgprot_cached_noncoherent pgprot_cached_noncoherent
493 static inline pgprot_t pgprot_cached_noncoherent(pgprot_t prot)
495 return __pgprot(pgprot_val(prot) & ~_PAGE_CACHE_CTL);
498 #define pgprot_writecombine pgprot_writecombine
499 static inline pgprot_t pgprot_writecombine(pgprot_t prot)
501 return pgprot_noncached_wc(prot);
504 #endif /* !__ASSEMBLY__ */
506 #endif /* _ASM_POWERPC_BOOK3S_32_PGTABLE_H */