]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - arch/powerpc/include/asm/book3s/32/pgtable.h
License cleanup: add SPDX GPL-2.0 license identifier to files with no license
[thirdparty/kernel/linux.git] / arch / powerpc / include / asm / book3s / 32 / pgtable.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
3dfcb315
AK
2#ifndef _ASM_POWERPC_BOOK3S_32_PGTABLE_H
3#define _ASM_POWERPC_BOOK3S_32_PGTABLE_H
4
9849a569 5#define __ARCH_USE_5LEVEL_HACK
3dfcb315
AK
6#include <asm-generic/pgtable-nopmd.h>
7
cbbb8683 8#include <asm/book3s/32/hash.h>
3dfcb315 9
cbbb8683
AK
10/* And here we include common definitions */
11#include <asm/pte-common.h>
3dfcb315 12
9b081e10
CL
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)
17
18#define PMD_CACHE_INDEX PMD_INDEX_SIZE
19
20#ifndef __ASSEMBLY__
21#define PTE_TABLE_SIZE (sizeof(pte_t) << PTE_INDEX_SIZE)
22#define PMD_TABLE_SIZE 0
23#define PUD_TABLE_SIZE 0
24#define PGD_TABLE_SIZE (sizeof(pgd_t) << PGD_INDEX_SIZE)
25#endif /* __ASSEMBLY__ */
26
27#define PTRS_PER_PTE (1 << PTE_INDEX_SIZE)
28#define PTRS_PER_PGD (1 << PGD_INDEX_SIZE)
29
3dfcb315
AK
30/*
31 * The normal case is that PTEs are 32-bits and we have a 1-page
32 * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages. -- paulus
33 *
34 * For any >32-bit physical address platform, we can use the following
35 * two level page table layout where the pgdir is 8KB and the MS 13 bits
36 * are an index to the second level table. The combined pgdir/pmd first
37 * level has 2048 entries and the second level has 512 64-bit PTE entries.
38 * -Matt
39 */
40/* PGDIR_SHIFT determines what a top-level page table entry can map */
9b081e10 41#define PGDIR_SHIFT (PAGE_SHIFT + PTE_INDEX_SIZE)
3dfcb315
AK
42#define PGDIR_SIZE (1UL << PGDIR_SHIFT)
43#define PGDIR_MASK (~(PGDIR_SIZE-1))
44
3dfcb315 45#define USER_PTRS_PER_PGD (TASK_SIZE / PGDIR_SIZE)
3dfcb315
AK
46/*
47 * This is the bottom of the PKMAP area with HIGHMEM or an arbitrary
48 * value (for now) on others, from where we can start layout kernel
49 * virtual space that goes below PKMAP and FIXMAP
50 */
51#ifdef CONFIG_HIGHMEM
52#define KVIRT_TOP PKMAP_BASE
53#else
54#define KVIRT_TOP (0xfe000000UL) /* for now, could be FIXMAP_BASE ? */
55#endif
56
57/*
58 * ioremap_bot starts at that address. Early ioremaps move down from there,
59 * until mem_init() at which point this becomes the top of the vmalloc
60 * and ioremap space
61 */
62#ifdef CONFIG_NOT_COHERENT_CACHE
63#define IOREMAP_TOP ((KVIRT_TOP - CONFIG_CONSISTENT_SIZE) & PAGE_MASK)
64#else
65#define IOREMAP_TOP KVIRT_TOP
66#endif
67
68/*
69 * Just any arbitrary offset to the start of the vmalloc VM area: the
70 * current 16MB value just means that there will be a 64MB "hole" after the
71 * physical memory until the kernel virtual memory starts. That means that
72 * any out-of-bounds memory accesses will hopefully be caught.
73 * The vmalloc() routines leaves a hole of 4kB between each vmalloced
74 * area for the same reason. ;)
75 *
76 * We no longer map larger than phys RAM with the BATs so we don't have
77 * to worry about the VMALLOC_OFFSET causing problems. We do have to worry
78 * about clashes between our early calls to ioremap() that start growing down
79 * from ioremap_base being run into the VM area allocations (growing upwards
80 * from VMALLOC_START). For this reason we have ioremap_bot to check when
81 * we actually run into our mappings setup in the early boot with the VM
82 * system. This really does become a problem for machines with good amounts
83 * of RAM. -- Cort
84 */
85#define VMALLOC_OFFSET (0x1000000) /* 16M */
86#ifdef PPC_PIN_SIZE
87#define VMALLOC_START (((_ALIGN((long)high_memory, PPC_PIN_SIZE) + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
88#else
89#define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
90#endif
91#define VMALLOC_END ioremap_bot
92
cbbb8683
AK
93#ifndef __ASSEMBLY__
94#include <linux/sched.h>
95#include <linux/threads.h>
96#include <asm/io.h> /* For sub-arch specific PPC_PIN_SIZE */
97
98extern unsigned long ioremap_bot;
99
9b081e10
CL
100/* Bits to mask out from a PGD to get to the PUD page */
101#define PGD_MASKED_BITS 0
cbbb8683
AK
102
103#define pte_ERROR(e) \
104 pr_err("%s:%d: bad pte %llx.\n", __FILE__, __LINE__, \
105 (unsigned long long)pte_val(e))
106#define pgd_ERROR(e) \
107 pr_err("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
3dfcb315
AK
108/*
109 * Bits in a linux-style PTE. These match the bits in the
110 * (hardware-defined) PowerPC PTE as closely as possible.
111 */
112
3dfcb315
AK
113#define pte_clear(mm, addr, ptep) \
114 do { pte_update(ptep, ~_PAGE_HASHPTE, 0); } while (0)
115
116#define pmd_none(pmd) (!pmd_val(pmd))
117#define pmd_bad(pmd) (pmd_val(pmd) & _PMD_BAD)
118#define pmd_present(pmd) (pmd_val(pmd) & _PMD_PRESENT_MASK)
f281b5d5
AK
119static inline void pmd_clear(pmd_t *pmdp)
120{
121 *pmdp = __pmd(0);
122}
123
3dfcb315
AK
124
125/*
126 * When flushing the tlb entry for a page, we also need to flush the hash
127 * table entry. flush_hash_pages is assembler (for speed) in hashtable.S.
128 */
129extern int flush_hash_pages(unsigned context, unsigned long va,
130 unsigned long pmdval, int count);
131
132/* Add an HPTE to the hash table */
133extern void add_hash_page(unsigned context, unsigned long va,
134 unsigned long pmdval);
135
136/* Flush an entry from the TLB/hash table */
137extern void flush_hash_entry(struct mm_struct *mm, pte_t *ptep,
138 unsigned long address);
139
140/*
141 * PTE updates. This function is called whenever an existing
142 * valid PTE is updated. This does -not- include set_pte_at()
143 * which nowadays only sets a new PTE.
144 *
145 * Depending on the type of MMU, we may need to use atomic updates
146 * and the PTE may be either 32 or 64 bit wide. In the later case,
147 * when using atomic updates, only the low part of the PTE is
148 * accessed atomically.
149 *
150 * In addition, on 44x, we also maintain a global flag indicating
151 * that an executable user mapping was modified, which is needed
152 * to properly flush the virtually tagged instruction cache of
153 * those implementations.
154 */
155#ifndef CONFIG_PTE_64BIT
156static inline unsigned long pte_update(pte_t *p,
157 unsigned long clr,
158 unsigned long set)
159{
3dfcb315
AK
160 unsigned long old, tmp;
161
162 __asm__ __volatile__("\
1631: lwarx %0,0,%3\n\
164 andc %1,%0,%4\n\
165 or %1,%1,%5\n"
166 PPC405_ERR77(0,%3)
167" stwcx. %1,0,%3\n\
168 bne- 1b"
169 : "=&r" (old), "=&r" (tmp), "=m" (*p)
170 : "r" (p), "r" (clr), "r" (set), "m" (*p)
171 : "cc" );
cbbb8683 172
3dfcb315
AK
173 return old;
174}
175#else /* CONFIG_PTE_64BIT */
176static inline unsigned long long pte_update(pte_t *p,
177 unsigned long clr,
178 unsigned long set)
179{
3dfcb315
AK
180 unsigned long long old;
181 unsigned long tmp;
182
183 __asm__ __volatile__("\
1841: lwarx %L0,0,%4\n\
185 lwzx %0,0,%3\n\
186 andc %1,%L0,%5\n\
187 or %1,%1,%6\n"
188 PPC405_ERR77(0,%3)
189" stwcx. %1,0,%4\n\
190 bne- 1b"
191 : "=&r" (old), "=&r" (tmp), "=m" (*p)
192 : "r" (p), "r" ((unsigned long)(p) + 4), "r" (clr), "r" (set), "m" (*p)
193 : "cc" );
cbbb8683 194
3dfcb315
AK
195 return old;
196}
197#endif /* CONFIG_PTE_64BIT */
198
199/*
200 * 2.6 calls this without flushing the TLB entry; this is wrong
201 * for our hash-based implementation, we fix that up here.
202 */
203#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
204static inline int __ptep_test_and_clear_young(unsigned int context, unsigned long addr, pte_t *ptep)
205{
206 unsigned long old;
207 old = pte_update(ptep, _PAGE_ACCESSED, 0);
3dfcb315
AK
208 if (old & _PAGE_HASHPTE) {
209 unsigned long ptephys = __pa(ptep) & PAGE_MASK;
210 flush_hash_pages(context, addr, ptephys, 1);
211 }
3dfcb315
AK
212 return (old & _PAGE_ACCESSED) != 0;
213}
214#define ptep_test_and_clear_young(__vma, __addr, __ptep) \
215 __ptep_test_and_clear_young((__vma)->vm_mm->context.id, __addr, __ptep)
216
217#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
218static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
219 pte_t *ptep)
220{
221 return __pte(pte_update(ptep, ~_PAGE_HASHPTE, 0));
222}
223
224#define __HAVE_ARCH_PTEP_SET_WRPROTECT
225static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
226 pte_t *ptep)
227{
228 pte_update(ptep, (_PAGE_RW | _PAGE_HWWRITE), _PAGE_RO);
229}
230static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
231 unsigned long addr, pte_t *ptep)
232{
233 ptep_set_wrprotect(mm, addr, ptep);
234}
235
236
c6d1a767 237static inline void __ptep_set_access_flags(struct mm_struct *mm,
b3603e17
AK
238 pte_t *ptep, pte_t entry,
239 unsigned long address)
3dfcb315
AK
240{
241 unsigned long set = pte_val(entry) &
242 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
243 unsigned long clr = ~pte_val(entry) & _PAGE_RO;
244
245 pte_update(ptep, clr, set);
246}
247
248#define __HAVE_ARCH_PTE_SAME
249#define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HASHPTE) == 0)
250
251/*
252 * Note that on Book E processors, the pmd contains the kernel virtual
253 * (lowmem) address of the pte page. The physical address is less useful
254 * because everything runs with translation enabled (even the TLB miss
255 * handler). On everything else the pmd contains the physical address
256 * of the pte page. -- paulus
257 */
258#ifndef CONFIG_BOOKE
259#define pmd_page_vaddr(pmd) \
260 ((unsigned long) __va(pmd_val(pmd) & PAGE_MASK))
261#define pmd_page(pmd) \
262 pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
263#else
264#define pmd_page_vaddr(pmd) \
265 ((unsigned long) (pmd_val(pmd) & PAGE_MASK))
266#define pmd_page(pmd) \
267 pfn_to_page((__pa(pmd_val(pmd)) >> PAGE_SHIFT))
268#endif
269
270/* to find an entry in a kernel page-table-directory */
271#define pgd_offset_k(address) pgd_offset(&init_mm, address)
272
273/* to find an entry in a page-table-directory */
274#define pgd_index(address) ((address) >> PGDIR_SHIFT)
275#define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
276
277/* Find an entry in the third-level page table.. */
278#define pte_index(address) \
279 (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))
280#define pte_offset_kernel(dir, addr) \
281 ((pte_t *) pmd_page_vaddr(*(dir)) + pte_index(addr))
282#define pte_offset_map(dir, addr) \
283 ((pte_t *) kmap_atomic(pmd_page(*(dir))) + pte_index(addr))
284#define pte_unmap(pte) kunmap_atomic(pte)
285
286/*
287 * Encode and decode a swap entry.
288 * Note that the bits we use in a PTE for representing a swap entry
289 * must not include the _PAGE_PRESENT bit or the _PAGE_HASHPTE bit (if used).
290 * -- paulus
291 */
292#define __swp_type(entry) ((entry).val & 0x1f)
293#define __swp_offset(entry) ((entry).val >> 5)
294#define __swp_entry(type, offset) ((swp_entry_t) { (type) | ((offset) << 5) })
295#define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) >> 3 })
296#define __swp_entry_to_pte(x) ((pte_t) { (x).val << 3 })
297
4386c096
CL
298int map_kernel_page(unsigned long va, phys_addr_t pa, int flags);
299
1ca72129
AK
300/* Generic accessors to PTE bits */
301static inline int pte_write(pte_t pte) { return !!(pte_val(pte) & _PAGE_RW);}
ca8afd40 302static inline int pte_read(pte_t pte) { return 1; }
1ca72129
AK
303static inline int pte_dirty(pte_t pte) { return !!(pte_val(pte) & _PAGE_DIRTY); }
304static inline int pte_young(pte_t pte) { return !!(pte_val(pte) & _PAGE_ACCESSED); }
305static inline int pte_special(pte_t pte) { return !!(pte_val(pte) & _PAGE_SPECIAL); }
306static inline int pte_none(pte_t pte) { return (pte_val(pte) & ~_PTE_NONE_MASK) == 0; }
307static inline pgprot_t pte_pgprot(pte_t pte) { return __pgprot(pte_val(pte) & PAGE_PROT_BITS); }
308
309static inline int pte_present(pte_t pte)
310{
311 return pte_val(pte) & _PAGE_PRESENT;
312}
313
314/* Conversion functions: convert a page and protection to a page entry,
315 * and a page entry and page directory to the page they refer to.
316 *
317 * Even if PTEs can be unsigned long long, a PFN is always an unsigned
318 * long for now.
319 */
320static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
321{
322 return __pte(((pte_basic_t)(pfn) << PTE_RPN_SHIFT) |
323 pgprot_val(pgprot));
324}
325
326static inline unsigned long pte_pfn(pte_t pte)
327{
328 return pte_val(pte) >> PTE_RPN_SHIFT;
329}
330
331/* Generic modifiers for PTE bits */
332static inline pte_t pte_wrprotect(pte_t pte)
333{
334 return __pte(pte_val(pte) & ~_PAGE_RW);
335}
336
337static inline pte_t pte_mkclean(pte_t pte)
338{
339 return __pte(pte_val(pte) & ~_PAGE_DIRTY);
340}
341
342static inline pte_t pte_mkold(pte_t pte)
343{
344 return __pte(pte_val(pte) & ~_PAGE_ACCESSED);
345}
346
347static inline pte_t pte_mkwrite(pte_t pte)
348{
349 return __pte(pte_val(pte) | _PAGE_RW);
350}
351
352static inline pte_t pte_mkdirty(pte_t pte)
353{
354 return __pte(pte_val(pte) | _PAGE_DIRTY);
355}
356
357static inline pte_t pte_mkyoung(pte_t pte)
358{
359 return __pte(pte_val(pte) | _PAGE_ACCESSED);
360}
361
362static inline pte_t pte_mkspecial(pte_t pte)
363{
364 return __pte(pte_val(pte) | _PAGE_SPECIAL);
365}
366
367static inline pte_t pte_mkhuge(pte_t pte)
368{
369 return pte;
370}
371
372static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
373{
374 return __pte((pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot));
375}
376
377
378
379/* This low level function performs the actual PTE insertion
380 * Setting the PTE depends on the MMU type and other factors. It's
381 * an horrible mess that I'm not going to try to clean up now but
382 * I'm keeping it in one place rather than spread around
383 */
384static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
385 pte_t *ptep, pte_t pte, int percpu)
386{
387#if defined(CONFIG_PPC_STD_MMU_32) && defined(CONFIG_SMP) && !defined(CONFIG_PTE_64BIT)
388 /* First case is 32-bit Hash MMU in SMP mode with 32-bit PTEs. We use the
389 * helper pte_update() which does an atomic update. We need to do that
390 * because a concurrent invalidation can clear _PAGE_HASHPTE. If it's a
391 * per-CPU PTE such as a kmap_atomic, we do a simple update preserving
392 * the hash bits instead (ie, same as the non-SMP case)
393 */
394 if (percpu)
395 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
396 | (pte_val(pte) & ~_PAGE_HASHPTE));
397 else
398 pte_update(ptep, ~_PAGE_HASHPTE, pte_val(pte));
399
400#elif defined(CONFIG_PPC32) && defined(CONFIG_PTE_64BIT)
401 /* Second case is 32-bit with 64-bit PTE. In this case, we
402 * can just store as long as we do the two halves in the right order
403 * with a barrier in between. This is possible because we take care,
404 * in the hash code, to pre-invalidate if the PTE was already hashed,
405 * which synchronizes us with any concurrent invalidation.
406 * In the percpu case, we also fallback to the simple update preserving
407 * the hash bits
408 */
409 if (percpu) {
410 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
411 | (pte_val(pte) & ~_PAGE_HASHPTE));
412 return;
413 }
414 if (pte_val(*ptep) & _PAGE_HASHPTE)
415 flush_hash_entry(mm, ptep, addr);
416 __asm__ __volatile__("\
417 stw%U0%X0 %2,%0\n\
418 eieio\n\
419 stw%U0%X0 %L2,%1"
420 : "=m" (*ptep), "=m" (*((unsigned char *)ptep+4))
421 : "r" (pte) : "memory");
422
423#elif defined(CONFIG_PPC_STD_MMU_32)
424 /* Third case is 32-bit hash table in UP mode, we need to preserve
425 * the _PAGE_HASHPTE bit since we may not have invalidated the previous
426 * translation in the hash yet (done in a subsequent flush_tlb_xxx())
427 * and see we need to keep track that this PTE needs invalidating
428 */
429 *ptep = __pte((pte_val(*ptep) & _PAGE_HASHPTE)
430 | (pte_val(pte) & ~_PAGE_HASHPTE));
431
432#else
433#error "Not supported "
434#endif
435}
436
437/*
438 * Macro to mark a page protection value as "uncacheable".
439 */
440
441#define _PAGE_CACHE_CTL (_PAGE_COHERENT | _PAGE_GUARDED | _PAGE_NO_CACHE | \
442 _PAGE_WRITETHRU)
443
444#define pgprot_noncached pgprot_noncached
445static inline pgprot_t pgprot_noncached(pgprot_t prot)
446{
447 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
448 _PAGE_NO_CACHE | _PAGE_GUARDED);
449}
450
451#define pgprot_noncached_wc pgprot_noncached_wc
452static inline pgprot_t pgprot_noncached_wc(pgprot_t prot)
453{
454 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
455 _PAGE_NO_CACHE);
456}
457
458#define pgprot_cached pgprot_cached
459static inline pgprot_t pgprot_cached(pgprot_t prot)
460{
461 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
462 _PAGE_COHERENT);
463}
464
465#define pgprot_cached_wthru pgprot_cached_wthru
466static inline pgprot_t pgprot_cached_wthru(pgprot_t prot)
467{
468 return __pgprot((pgprot_val(prot) & ~_PAGE_CACHE_CTL) |
469 _PAGE_COHERENT | _PAGE_WRITETHRU);
470}
471
472#define pgprot_cached_noncoherent pgprot_cached_noncoherent
473static inline pgprot_t pgprot_cached_noncoherent(pgprot_t prot)
474{
475 return __pgprot(pgprot_val(prot) & ~_PAGE_CACHE_CTL);
476}
477
478#define pgprot_writecombine pgprot_writecombine
479static inline pgprot_t pgprot_writecombine(pgprot_t prot)
480{
481 return pgprot_noncached_wc(prot);
482}
483
3dfcb315
AK
484#endif /* !__ASSEMBLY__ */
485
486#endif /* _ASM_POWERPC_BOOK3S_32_PGTABLE_H */