]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/asm-powerpc/pgtable-ppc64.h
9b0f51ccad05de2310924d11ff2d4d15a0ebab6b
[mirror_ubuntu-jammy-kernel.git] / include / asm-powerpc / pgtable-ppc64.h
1 #ifndef _ASM_POWERPC_PGTABLE_PPC64_H_
2 #define _ASM_POWERPC_PGTABLE_PPC64_H_
3 /*
4 * This file contains the functions and defines necessary to modify and use
5 * the ppc64 hashed page table.
6 */
7
8 #ifndef __ASSEMBLY__
9 #include <linux/stddef.h>
10 #include <asm/processor.h> /* For TASK_SIZE */
11 #include <asm/mmu.h>
12 #include <asm/page.h>
13 #include <asm/tlbflush.h>
14 struct mm_struct;
15 #endif /* __ASSEMBLY__ */
16
17 #ifdef CONFIG_PPC_64K_PAGES
18 #include <asm/pgtable-64k.h>
19 #else
20 #include <asm/pgtable-4k.h>
21 #endif
22
23 #define FIRST_USER_ADDRESS 0
24
25 /*
26 * Size of EA range mapped by our pagetables.
27 */
28 #define PGTABLE_EADDR_SIZE (PTE_INDEX_SIZE + PMD_INDEX_SIZE + \
29 PUD_INDEX_SIZE + PGD_INDEX_SIZE + PAGE_SHIFT)
30 #define PGTABLE_RANGE (ASM_CONST(1) << PGTABLE_EADDR_SIZE)
31
32 #if TASK_SIZE_USER64 > PGTABLE_RANGE
33 #error TASK_SIZE_USER64 exceeds pagetable range
34 #endif
35
36 #if TASK_SIZE_USER64 > (1UL << (USER_ESID_BITS + SID_SHIFT))
37 #error TASK_SIZE_USER64 exceeds user VSID range
38 #endif
39
40
41 /*
42 * Define the address range of the vmalloc VM area.
43 */
44 #define VMALLOC_START ASM_CONST(0xD000000000000000)
45 #define VMALLOC_SIZE (PGTABLE_RANGE >> 1)
46 #define VMALLOC_END (VMALLOC_START + VMALLOC_SIZE)
47
48 /*
49 * Define the address ranges for MMIO and IO space :
50 *
51 * ISA_IO_BASE = VMALLOC_END, 64K reserved area
52 * PHB_IO_BASE = ISA_IO_BASE + 64K to ISA_IO_BASE + 2G, PHB IO spaces
53 * IOREMAP_BASE = ISA_IO_BASE + 2G to VMALLOC_START + PGTABLE_RANGE
54 */
55 #define FULL_IO_SIZE 0x80000000ul
56 #define ISA_IO_BASE (VMALLOC_END)
57 #define ISA_IO_END (VMALLOC_END + 0x10000ul)
58 #define PHB_IO_BASE (ISA_IO_END)
59 #define PHB_IO_END (VMALLOC_END + FULL_IO_SIZE)
60 #define IOREMAP_BASE (PHB_IO_END)
61 #define IOREMAP_END (VMALLOC_START + PGTABLE_RANGE)
62
63 /*
64 * Region IDs
65 */
66 #define REGION_SHIFT 60UL
67 #define REGION_MASK (0xfUL << REGION_SHIFT)
68 #define REGION_ID(ea) (((unsigned long)(ea)) >> REGION_SHIFT)
69
70 #define VMALLOC_REGION_ID (REGION_ID(VMALLOC_START))
71 #define KERNEL_REGION_ID (REGION_ID(PAGE_OFFSET))
72 #define USER_REGION_ID (0UL)
73
74 /*
75 * Common bits in a linux-style PTE. These match the bits in the
76 * (hardware-defined) PowerPC PTE as closely as possible. Additional
77 * bits may be defined in pgtable-*.h
78 */
79 #define _PAGE_PRESENT 0x0001 /* software: pte contains a translation */
80 #define _PAGE_USER 0x0002 /* matches one of the PP bits */
81 #define _PAGE_FILE 0x0002 /* (!present only) software: pte holds file offset */
82 #define _PAGE_EXEC 0x0004 /* No execute on POWER4 and newer (we invert) */
83 #define _PAGE_GUARDED 0x0008
84 #define _PAGE_COHERENT 0x0010 /* M: enforce memory coherence (SMP systems) */
85 #define _PAGE_NO_CACHE 0x0020 /* I: cache inhibit */
86 #define _PAGE_WRITETHRU 0x0040 /* W: cache write-through */
87 #define _PAGE_DIRTY 0x0080 /* C: page changed */
88 #define _PAGE_ACCESSED 0x0100 /* R: page referenced */
89 #define _PAGE_RW 0x0200 /* software: user write access allowed */
90 #define _PAGE_HASHPTE 0x0400 /* software: pte has an associated HPTE */
91 #define _PAGE_BUSY 0x0800 /* software: PTE & hash are busy */
92
93 #define _PAGE_BASE (_PAGE_PRESENT | _PAGE_ACCESSED | _PAGE_COHERENT)
94
95 #define _PAGE_WRENABLE (_PAGE_RW | _PAGE_DIRTY)
96
97 /* __pgprot defined in asm-powerpc/page.h */
98 #define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED)
99
100 #define PAGE_SHARED __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER)
101 #define PAGE_SHARED_X __pgprot(_PAGE_BASE | _PAGE_RW | _PAGE_USER | _PAGE_EXEC)
102 #define PAGE_COPY __pgprot(_PAGE_BASE | _PAGE_USER)
103 #define PAGE_COPY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
104 #define PAGE_READONLY __pgprot(_PAGE_BASE | _PAGE_USER)
105 #define PAGE_READONLY_X __pgprot(_PAGE_BASE | _PAGE_USER | _PAGE_EXEC)
106 #define PAGE_KERNEL __pgprot(_PAGE_BASE | _PAGE_WRENABLE)
107 #define PAGE_KERNEL_CI __pgprot(_PAGE_PRESENT | _PAGE_ACCESSED | \
108 _PAGE_WRENABLE | _PAGE_NO_CACHE | _PAGE_GUARDED)
109 #define PAGE_KERNEL_EXEC __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_EXEC)
110
111 #define PAGE_AGP __pgprot(_PAGE_BASE | _PAGE_WRENABLE | _PAGE_NO_CACHE)
112 #define HAVE_PAGE_AGP
113
114 /* PTEIDX nibble */
115 #define _PTEIDX_SECONDARY 0x8
116 #define _PTEIDX_GROUP_IX 0x7
117
118
119 /*
120 * POWER4 and newer have per page execute protection, older chips can only
121 * do this on a segment (256MB) basis.
122 *
123 * Also, write permissions imply read permissions.
124 * This is the closest we can get..
125 *
126 * Note due to the way vm flags are laid out, the bits are XWR
127 */
128 #define __P000 PAGE_NONE
129 #define __P001 PAGE_READONLY
130 #define __P010 PAGE_COPY
131 #define __P011 PAGE_COPY
132 #define __P100 PAGE_READONLY_X
133 #define __P101 PAGE_READONLY_X
134 #define __P110 PAGE_COPY_X
135 #define __P111 PAGE_COPY_X
136
137 #define __S000 PAGE_NONE
138 #define __S001 PAGE_READONLY
139 #define __S010 PAGE_SHARED
140 #define __S011 PAGE_SHARED
141 #define __S100 PAGE_READONLY_X
142 #define __S101 PAGE_READONLY_X
143 #define __S110 PAGE_SHARED_X
144 #define __S111 PAGE_SHARED_X
145
146 #ifndef __ASSEMBLY__
147
148 /*
149 * ZERO_PAGE is a global shared page that is always zero: used
150 * for zero-mapped memory areas etc..
151 */
152 extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
153 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
154 #endif /* __ASSEMBLY__ */
155
156 #ifdef CONFIG_HUGETLB_PAGE
157
158 #define HAVE_ARCH_UNMAPPED_AREA
159 #define HAVE_ARCH_UNMAPPED_AREA_TOPDOWN
160
161 #endif
162
163 #ifndef __ASSEMBLY__
164
165 /*
166 * Conversion functions: convert a page and protection to a page entry,
167 * and a page entry and page directory to the page they refer to.
168 *
169 * mk_pte takes a (struct page *) as input
170 */
171 #define mk_pte(page, pgprot) pfn_pte(page_to_pfn(page), (pgprot))
172
173 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t pgprot)
174 {
175 pte_t pte;
176
177
178 pte_val(pte) = (pfn << PTE_RPN_SHIFT) | pgprot_val(pgprot);
179 return pte;
180 }
181
182 #define pte_modify(_pte, newprot) \
183 (__pte((pte_val(_pte) & _PAGE_CHG_MASK) | pgprot_val(newprot)))
184
185 #define pte_none(pte) ((pte_val(pte) & ~_PAGE_HPTEFLAGS) == 0)
186 #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT)
187
188 /* pte_clear moved to later in this file */
189
190 #define pte_pfn(x) ((unsigned long)((pte_val(x)>>PTE_RPN_SHIFT)))
191 #define pte_page(x) pfn_to_page(pte_pfn(x))
192
193 #define PMD_BAD_BITS (PTE_TABLE_SIZE-1)
194 #define PUD_BAD_BITS (PMD_TABLE_SIZE-1)
195
196 #define pmd_set(pmdp, pmdval) (pmd_val(*(pmdp)) = (pmdval))
197 #define pmd_none(pmd) (!pmd_val(pmd))
198 #define pmd_bad(pmd) (!is_kernel_addr(pmd_val(pmd)) \
199 || (pmd_val(pmd) & PMD_BAD_BITS))
200 #define pmd_present(pmd) (pmd_val(pmd) != 0)
201 #define pmd_clear(pmdp) (pmd_val(*(pmdp)) = 0)
202 #define pmd_page_vaddr(pmd) (pmd_val(pmd) & ~PMD_MASKED_BITS)
203 #define pmd_page(pmd) virt_to_page(pmd_page_vaddr(pmd))
204
205 #define pud_set(pudp, pudval) (pud_val(*(pudp)) = (pudval))
206 #define pud_none(pud) (!pud_val(pud))
207 #define pud_bad(pud) (!is_kernel_addr(pud_val(pud)) \
208 || (pud_val(pud) & PUD_BAD_BITS))
209 #define pud_present(pud) (pud_val(pud) != 0)
210 #define pud_clear(pudp) (pud_val(*(pudp)) = 0)
211 #define pud_page_vaddr(pud) (pud_val(pud) & ~PUD_MASKED_BITS)
212 #define pud_page(pud) virt_to_page(pud_page_vaddr(pud))
213
214 #define pgd_set(pgdp, pudp) ({pgd_val(*(pgdp)) = (unsigned long)(pudp);})
215
216 /*
217 * Find an entry in a page-table-directory. We combine the address region
218 * (the high order N bits) and the pgd portion of the address.
219 */
220 /* to avoid overflow in free_pgtables we don't use PTRS_PER_PGD here */
221 #define pgd_index(address) (((address) >> (PGDIR_SHIFT)) & 0x1ff)
222
223 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index(address))
224
225 #define pmd_offset(pudp,addr) \
226 (((pmd_t *) pud_page_vaddr(*(pudp))) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)))
227
228 #define pte_offset_kernel(dir,addr) \
229 (((pte_t *) pmd_page_vaddr(*(dir))) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)))
230
231 #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr))
232 #define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr))
233 #define pte_unmap(pte) do { } while(0)
234 #define pte_unmap_nested(pte) do { } while(0)
235
236 /* to find an entry in a kernel page-table-directory */
237 /* This now only contains the vmalloc pages */
238 #define pgd_offset_k(address) pgd_offset(&init_mm, address)
239
240 /*
241 * The following only work if pte_present() is true.
242 * Undefined behaviour if not..
243 */
244 static inline int pte_read(pte_t pte) { return pte_val(pte) & _PAGE_USER;}
245 static inline int pte_write(pte_t pte) { return pte_val(pte) & _PAGE_RW;}
246 static inline int pte_exec(pte_t pte) { return pte_val(pte) & _PAGE_EXEC;}
247 static inline int pte_dirty(pte_t pte) { return pte_val(pte) & _PAGE_DIRTY;}
248 static inline int pte_young(pte_t pte) { return pte_val(pte) & _PAGE_ACCESSED;}
249 static inline int pte_file(pte_t pte) { return pte_val(pte) & _PAGE_FILE;}
250
251 static inline void pte_uncache(pte_t pte) { pte_val(pte) |= _PAGE_NO_CACHE; }
252 static inline void pte_cache(pte_t pte) { pte_val(pte) &= ~_PAGE_NO_CACHE; }
253
254 static inline pte_t pte_rdprotect(pte_t pte) {
255 pte_val(pte) &= ~_PAGE_USER; return pte; }
256 static inline pte_t pte_exprotect(pte_t pte) {
257 pte_val(pte) &= ~_PAGE_EXEC; return pte; }
258 static inline pte_t pte_wrprotect(pte_t pte) {
259 pte_val(pte) &= ~(_PAGE_RW); return pte; }
260 static inline pte_t pte_mkclean(pte_t pte) {
261 pte_val(pte) &= ~(_PAGE_DIRTY); return pte; }
262 static inline pte_t pte_mkold(pte_t pte) {
263 pte_val(pte) &= ~_PAGE_ACCESSED; return pte; }
264 static inline pte_t pte_mkread(pte_t pte) {
265 pte_val(pte) |= _PAGE_USER; return pte; }
266 static inline pte_t pte_mkexec(pte_t pte) {
267 pte_val(pte) |= _PAGE_USER | _PAGE_EXEC; return pte; }
268 static inline pte_t pte_mkwrite(pte_t pte) {
269 pte_val(pte) |= _PAGE_RW; return pte; }
270 static inline pte_t pte_mkdirty(pte_t pte) {
271 pte_val(pte) |= _PAGE_DIRTY; return pte; }
272 static inline pte_t pte_mkyoung(pte_t pte) {
273 pte_val(pte) |= _PAGE_ACCESSED; return pte; }
274 static inline pte_t pte_mkhuge(pte_t pte) {
275 return pte; }
276
277 /* Atomic PTE updates */
278 static inline unsigned long pte_update(struct mm_struct *mm,
279 unsigned long addr,
280 pte_t *ptep, unsigned long clr,
281 int huge)
282 {
283 unsigned long old, tmp;
284
285 __asm__ __volatile__(
286 "1: ldarx %0,0,%3 # pte_update\n\
287 andi. %1,%0,%6\n\
288 bne- 1b \n\
289 andc %1,%0,%4 \n\
290 stdcx. %1,0,%3 \n\
291 bne- 1b"
292 : "=&r" (old), "=&r" (tmp), "=m" (*ptep)
293 : "r" (ptep), "r" (clr), "m" (*ptep), "i" (_PAGE_BUSY)
294 : "cc" );
295
296 if (old & _PAGE_HASHPTE)
297 hpte_need_flush(mm, addr, ptep, old, huge);
298 return old;
299 }
300
301 static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
302 unsigned long addr, pte_t *ptep)
303 {
304 unsigned long old;
305
306 if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
307 return 0;
308 old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0);
309 return (old & _PAGE_ACCESSED) != 0;
310 }
311 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
312 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
313 ({ \
314 int __r; \
315 __r = __ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep); \
316 __r; \
317 })
318
319 /*
320 * On RW/DIRTY bit transitions we can avoid flushing the hpte. For the
321 * moment we always flush but we need to fix hpte_update and test if the
322 * optimisation is worth it.
323 */
324 static inline int __ptep_test_and_clear_dirty(struct mm_struct *mm,
325 unsigned long addr, pte_t *ptep)
326 {
327 unsigned long old;
328
329 if ((pte_val(*ptep) & _PAGE_DIRTY) == 0)
330 return 0;
331 old = pte_update(mm, addr, ptep, _PAGE_DIRTY, 0);
332 return (old & _PAGE_DIRTY) != 0;
333 }
334 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_DIRTY
335 #define ptep_test_and_clear_dirty(__vma, __addr, __ptep) \
336 ({ \
337 int __r; \
338 __r = __ptep_test_and_clear_dirty((__vma)->vm_mm, __addr, __ptep); \
339 __r; \
340 })
341
342 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
343 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
344 pte_t *ptep)
345 {
346 unsigned long old;
347
348 if ((pte_val(*ptep) & _PAGE_RW) == 0)
349 return;
350 old = pte_update(mm, addr, ptep, _PAGE_RW, 0);
351 }
352
353 /*
354 * We currently remove entries from the hashtable regardless of whether
355 * the entry was young or dirty. The generic routines only flush if the
356 * entry was young or dirty which is not good enough.
357 *
358 * We should be more intelligent about this but for the moment we override
359 * these functions and force a tlb flush unconditionally
360 */
361 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
362 #define ptep_clear_flush_young(__vma, __address, __ptep) \
363 ({ \
364 int __young = __ptep_test_and_clear_young((__vma)->vm_mm, __address, \
365 __ptep); \
366 __young; \
367 })
368
369 #define __HAVE_ARCH_PTEP_CLEAR_DIRTY_FLUSH
370 #define ptep_clear_flush_dirty(__vma, __address, __ptep) \
371 ({ \
372 int __dirty = __ptep_test_and_clear_dirty((__vma)->vm_mm, __address, \
373 __ptep); \
374 __dirty; \
375 })
376
377 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
378 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
379 unsigned long addr, pte_t *ptep)
380 {
381 unsigned long old = pte_update(mm, addr, ptep, ~0UL, 0);
382 return __pte(old);
383 }
384
385 static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
386 pte_t * ptep)
387 {
388 pte_update(mm, addr, ptep, ~0UL, 0);
389 }
390
391 /*
392 * set_pte stores a linux PTE into the linux page table.
393 */
394 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
395 pte_t *ptep, pte_t pte)
396 {
397 if (pte_present(*ptep))
398 pte_clear(mm, addr, ptep);
399 pte = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
400 *ptep = pte;
401 }
402
403 /* Set the dirty and/or accessed bits atomically in a linux PTE, this
404 * function doesn't need to flush the hash entry
405 */
406 #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
407 static inline void __ptep_set_access_flags(pte_t *ptep, pte_t entry, int dirty)
408 {
409 unsigned long bits = pte_val(entry) &
410 (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
411 unsigned long old, tmp;
412
413 __asm__ __volatile__(
414 "1: ldarx %0,0,%4\n\
415 andi. %1,%0,%6\n\
416 bne- 1b \n\
417 or %0,%3,%0\n\
418 stdcx. %0,0,%4\n\
419 bne- 1b"
420 :"=&r" (old), "=&r" (tmp), "=m" (*ptep)
421 :"r" (bits), "r" (ptep), "m" (*ptep), "i" (_PAGE_BUSY)
422 :"cc");
423 }
424 #define ptep_set_access_flags(__vma, __address, __ptep, __entry, __dirty) \
425 do { \
426 __ptep_set_access_flags(__ptep, __entry, __dirty); \
427 flush_tlb_page_nohash(__vma, __address); \
428 } while(0)
429
430 /*
431 * Macro to mark a page protection value as "uncacheable".
432 */
433 #define pgprot_noncached(prot) (__pgprot(pgprot_val(prot) | _PAGE_NO_CACHE | _PAGE_GUARDED))
434
435 struct file;
436 extern pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
437 unsigned long size, pgprot_t vma_prot);
438 #define __HAVE_PHYS_MEM_ACCESS_PROT
439
440 #define __HAVE_ARCH_PTE_SAME
441 #define pte_same(A,B) (((pte_val(A) ^ pte_val(B)) & ~_PAGE_HPTEFLAGS) == 0)
442
443 #define pte_ERROR(e) \
444 printk("%s:%d: bad pte %08lx.\n", __FILE__, __LINE__, pte_val(e))
445 #define pmd_ERROR(e) \
446 printk("%s:%d: bad pmd %08lx.\n", __FILE__, __LINE__, pmd_val(e))
447 #define pgd_ERROR(e) \
448 printk("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
449
450 extern pgd_t swapper_pg_dir[];
451
452 extern void paging_init(void);
453
454 /* Encode and de-code a swap entry */
455 #define __swp_type(entry) (((entry).val >> 1) & 0x3f)
456 #define __swp_offset(entry) ((entry).val >> 8)
457 #define __swp_entry(type, offset) ((swp_entry_t){((type)<< 1)|((offset)<<8)})
458 #define __pte_to_swp_entry(pte) ((swp_entry_t){pte_val(pte) >> PTE_RPN_SHIFT})
459 #define __swp_entry_to_pte(x) ((pte_t) { (x).val << PTE_RPN_SHIFT })
460 #define pte_to_pgoff(pte) (pte_val(pte) >> PTE_RPN_SHIFT)
461 #define pgoff_to_pte(off) ((pte_t) {((off) << PTE_RPN_SHIFT)|_PAGE_FILE})
462 #define PTE_FILE_MAX_BITS (BITS_PER_LONG - PTE_RPN_SHIFT)
463
464 /*
465 * kern_addr_valid is intended to indicate whether an address is a valid
466 * kernel address. Most 32-bit archs define it as always true (like this)
467 * but most 64-bit archs actually perform a test. What should we do here?
468 * The only use is in fs/ncpfs/dir.c
469 */
470 #define kern_addr_valid(addr) (1)
471
472 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
473 remap_pfn_range(vma, vaddr, pfn, size, prot)
474
475 void pgtable_cache_init(void);
476
477 /*
478 * find_linux_pte returns the address of a linux pte for a given
479 * effective address and directory. If not found, it returns zero.
480 */static inline pte_t *find_linux_pte(pgd_t *pgdir, unsigned long ea)
481 {
482 pgd_t *pg;
483 pud_t *pu;
484 pmd_t *pm;
485 pte_t *pt = NULL;
486
487 pg = pgdir + pgd_index(ea);
488 if (!pgd_none(*pg)) {
489 pu = pud_offset(pg, ea);
490 if (!pud_none(*pu)) {
491 pm = pmd_offset(pu, ea);
492 if (pmd_present(*pm))
493 pt = pte_offset_kernel(pm, ea);
494 }
495 }
496 return pt;
497 }
498
499 #endif /* __ASSEMBLY__ */
500
501 #endif /* _ASM_POWERPC_PGTABLE_PPC64_H_ */