xref: /linux/arch/powerpc/include/asm/nohash/32/pgtable.h (revision 5e0266f0e5f57617472d5aac4013f58a3ef264ac)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _ASM_POWERPC_NOHASH_32_PGTABLE_H
3 #define _ASM_POWERPC_NOHASH_32_PGTABLE_H
4 
5 #include <asm-generic/pgtable-nopmd.h>
6 
7 #ifndef __ASSEMBLY__
8 #include <linux/sched.h>
9 #include <linux/threads.h>
10 #include <asm/mmu.h>			/* For sub-arch specific PPC_PIN_SIZE */
11 
12 #ifdef CONFIG_44x
13 extern int icache_44x_need_flush;
14 #endif
15 
16 #endif /* __ASSEMBLY__ */
17 
18 #define PTE_INDEX_SIZE	PTE_SHIFT
19 #define PMD_INDEX_SIZE	0
20 #define PUD_INDEX_SIZE	0
21 #define PGD_INDEX_SIZE	(32 - PGDIR_SHIFT)
22 
23 #define PMD_CACHE_INDEX	PMD_INDEX_SIZE
24 #define PUD_CACHE_INDEX	PUD_INDEX_SIZE
25 
26 #ifndef __ASSEMBLY__
27 #define PTE_TABLE_SIZE	(sizeof(pte_t) << PTE_INDEX_SIZE)
28 #define PMD_TABLE_SIZE	0
29 #define PUD_TABLE_SIZE	0
30 #define PGD_TABLE_SIZE	(sizeof(pgd_t) << PGD_INDEX_SIZE)
31 
32 #define PMD_MASKED_BITS (PTE_TABLE_SIZE - 1)
33 #endif	/* __ASSEMBLY__ */
34 
35 #define PTRS_PER_PTE	(1 << PTE_INDEX_SIZE)
36 #define PTRS_PER_PGD	(1 << PGD_INDEX_SIZE)
37 
38 /*
39  * The normal case is that PTEs are 32-bits and we have a 1-page
40  * 1024-entry pgdir pointing to 1-page 1024-entry PTE pages.  -- paulus
41  *
42  * For any >32-bit physical address platform, we can use the following
43  * two level page table layout where the pgdir is 8KB and the MS 13 bits
44  * are an index to the second level table.  The combined pgdir/pmd first
45  * level has 2048 entries and the second level has 512 64-bit PTE entries.
46  * -Matt
47  */
48 /* PGDIR_SHIFT determines what a top-level page table entry can map */
49 #define PGDIR_SHIFT	(PAGE_SHIFT + PTE_INDEX_SIZE)
50 #define PGDIR_SIZE	(1UL << PGDIR_SHIFT)
51 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
52 
53 /* Bits to mask out from a PGD to get to the PUD page */
54 #define PGD_MASKED_BITS		0
55 
56 #define USER_PTRS_PER_PGD	(TASK_SIZE / PGDIR_SIZE)
57 
58 #define pte_ERROR(e) \
59 	pr_err("%s:%d: bad pte %llx.\n", __FILE__, __LINE__, \
60 		(unsigned long long)pte_val(e))
61 #define pgd_ERROR(e) \
62 	pr_err("%s:%d: bad pgd %08lx.\n", __FILE__, __LINE__, pgd_val(e))
63 
64 #ifndef __ASSEMBLY__
65 
66 int map_kernel_page(unsigned long va, phys_addr_t pa, pgprot_t prot);
67 void unmap_kernel_page(unsigned long va);
68 
69 #endif /* !__ASSEMBLY__ */
70 
71 
72 /*
73  * This is the bottom of the PKMAP area with HIGHMEM or an arbitrary
74  * value (for now) on others, from where we can start layout kernel
75  * virtual space that goes below PKMAP and FIXMAP
76  */
77 #include <asm/fixmap.h>
78 
79 /*
80  * ioremap_bot starts at that address. Early ioremaps move down from there,
81  * until mem_init() at which point this becomes the top of the vmalloc
82  * and ioremap space
83  */
84 #ifdef CONFIG_HIGHMEM
85 #define IOREMAP_TOP	PKMAP_BASE
86 #else
87 #define IOREMAP_TOP	FIXADDR_START
88 #endif
89 
90 /* PPC32 shares vmalloc area with ioremap */
91 #define IOREMAP_START	VMALLOC_START
92 #define IOREMAP_END	VMALLOC_END
93 
94 /*
95  * Just any arbitrary offset to the start of the vmalloc VM area: the
96  * current 16MB value just means that there will be a 64MB "hole" after the
97  * physical memory until the kernel virtual memory starts.  That means that
98  * any out-of-bounds memory accesses will hopefully be caught.
99  * The vmalloc() routines leaves a hole of 4kB between each vmalloced
100  * area for the same reason. ;)
101  *
102  * We no longer map larger than phys RAM with the BATs so we don't have
103  * to worry about the VMALLOC_OFFSET causing problems.  We do have to worry
104  * about clashes between our early calls to ioremap() that start growing down
105  * from IOREMAP_TOP being run into the VM area allocations (growing upwards
106  * from VMALLOC_START).  For this reason we have ioremap_bot to check when
107  * we actually run into our mappings setup in the early boot with the VM
108  * system.  This really does become a problem for machines with good amounts
109  * of RAM.  -- Cort
110  */
111 #define VMALLOC_OFFSET (0x1000000) /* 16M */
112 #ifdef PPC_PIN_SIZE
113 #define VMALLOC_START (((ALIGN((long)high_memory, PPC_PIN_SIZE) + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
114 #else
115 #define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
116 #endif
117 
118 #ifdef CONFIG_KASAN_VMALLOC
119 #define VMALLOC_END	ALIGN_DOWN(ioremap_bot, PAGE_SIZE << KASAN_SHADOW_SCALE_SHIFT)
120 #else
121 #define VMALLOC_END	ioremap_bot
122 #endif
123 
124 /*
125  * Bits in a linux-style PTE.  These match the bits in the
126  * (hardware-defined) PowerPC PTE as closely as possible.
127  */
128 
129 #if defined(CONFIG_40x)
130 #include <asm/nohash/32/pte-40x.h>
131 #elif defined(CONFIG_44x)
132 #include <asm/nohash/32/pte-44x.h>
133 #elif defined(CONFIG_PPC_85xx) && defined(CONFIG_PTE_64BIT)
134 #include <asm/nohash/pte-e500.h>
135 #elif defined(CONFIG_PPC_85xx)
136 #include <asm/nohash/32/pte-85xx.h>
137 #elif defined(CONFIG_PPC_8xx)
138 #include <asm/nohash/32/pte-8xx.h>
139 #endif
140 
141 /*
142  * Location of the PFN in the PTE. Most 32-bit platforms use the same
143  * as _PAGE_SHIFT here (ie, naturally aligned).
144  * Platform who don't just pre-define the value so we don't override it here.
145  */
146 #ifndef PTE_RPN_SHIFT
147 #define PTE_RPN_SHIFT	(PAGE_SHIFT)
148 #endif
149 
150 /*
151  * The mask covered by the RPN must be a ULL on 32-bit platforms with
152  * 64-bit PTEs.
153  */
154 #if defined(CONFIG_PPC32) && defined(CONFIG_PTE_64BIT)
155 #define PTE_RPN_MASK	(~((1ULL << PTE_RPN_SHIFT) - 1))
156 #define MAX_POSSIBLE_PHYSMEM_BITS 36
157 #else
158 #define PTE_RPN_MASK	(~((1UL << PTE_RPN_SHIFT) - 1))
159 #define MAX_POSSIBLE_PHYSMEM_BITS 32
160 #endif
161 
162 /*
163  * _PAGE_CHG_MASK masks of bits that are to be preserved across
164  * pgprot changes.
165  */
166 #define _PAGE_CHG_MASK	(PTE_RPN_MASK | _PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_SPECIAL)
167 
168 #ifndef __ASSEMBLY__
169 
170 #define pte_clear(mm, addr, ptep) \
171 	do { pte_update(mm, addr, ptep, ~0, 0, 0); } while (0)
172 
173 #ifndef pte_mkwrite
174 static inline pte_t pte_mkwrite(pte_t pte)
175 {
176 	return __pte(pte_val(pte) | _PAGE_RW);
177 }
178 #endif
179 
180 static inline pte_t pte_mkdirty(pte_t pte)
181 {
182 	return __pte(pte_val(pte) | _PAGE_DIRTY);
183 }
184 
185 static inline pte_t pte_mkyoung(pte_t pte)
186 {
187 	return __pte(pte_val(pte) | _PAGE_ACCESSED);
188 }
189 
190 #ifndef pte_wrprotect
191 static inline pte_t pte_wrprotect(pte_t pte)
192 {
193 	return __pte(pte_val(pte) & ~_PAGE_RW);
194 }
195 #endif
196 
197 #ifndef pte_mkexec
198 static inline pte_t pte_mkexec(pte_t pte)
199 {
200 	return __pte(pte_val(pte) | _PAGE_EXEC);
201 }
202 #endif
203 
204 #define pmd_none(pmd)		(!pmd_val(pmd))
205 #define	pmd_bad(pmd)		(pmd_val(pmd) & _PMD_BAD)
206 #define	pmd_present(pmd)	(pmd_val(pmd) & _PMD_PRESENT_MASK)
207 static inline void pmd_clear(pmd_t *pmdp)
208 {
209 	*pmdp = __pmd(0);
210 }
211 
212 /*
213  * PTE updates. This function is called whenever an existing
214  * valid PTE is updated. This does -not- include set_pte_at()
215  * which nowadays only sets a new PTE.
216  *
217  * Depending on the type of MMU, we may need to use atomic updates
218  * and the PTE may be either 32 or 64 bit wide. In the later case,
219  * when using atomic updates, only the low part of the PTE is
220  * accessed atomically.
221  *
222  * In addition, on 44x, we also maintain a global flag indicating
223  * that an executable user mapping was modified, which is needed
224  * to properly flush the virtually tagged instruction cache of
225  * those implementations.
226  *
227  * On the 8xx, the page tables are a bit special. For 16k pages, we have
228  * 4 identical entries. For 512k pages, we have 128 entries as if it was
229  * 4k pages, but they are flagged as 512k pages for the hardware.
230  * For other page sizes, we have a single entry in the table.
231  */
232 #ifdef CONFIG_PPC_8xx
233 static pmd_t *pmd_off(struct mm_struct *mm, unsigned long addr);
234 static int hugepd_ok(hugepd_t hpd);
235 
236 static int number_of_cells_per_pte(pmd_t *pmd, pte_basic_t val, int huge)
237 {
238 	if (!huge)
239 		return PAGE_SIZE / SZ_4K;
240 	else if (hugepd_ok(*((hugepd_t *)pmd)))
241 		return 1;
242 	else if (IS_ENABLED(CONFIG_PPC_4K_PAGES) && !(val & _PAGE_HUGE))
243 		return SZ_16K / SZ_4K;
244 	else
245 		return SZ_512K / SZ_4K;
246 }
247 
248 static inline pte_basic_t pte_update(struct mm_struct *mm, unsigned long addr, pte_t *p,
249 				     unsigned long clr, unsigned long set, int huge)
250 {
251 	pte_basic_t *entry = (pte_basic_t *)p;
252 	pte_basic_t old = pte_val(*p);
253 	pte_basic_t new = (old & ~(pte_basic_t)clr) | set;
254 	int num, i;
255 	pmd_t *pmd = pmd_off(mm, addr);
256 
257 	num = number_of_cells_per_pte(pmd, new, huge);
258 
259 	for (i = 0; i < num; i += PAGE_SIZE / SZ_4K, new += PAGE_SIZE) {
260 		*entry++ = new;
261 		if (IS_ENABLED(CONFIG_PPC_16K_PAGES) && num != 1) {
262 			*entry++ = new;
263 			*entry++ = new;
264 			*entry++ = new;
265 		}
266 	}
267 
268 	return old;
269 }
270 
271 #ifdef CONFIG_PPC_16K_PAGES
272 #define ptep_get ptep_get
273 static inline pte_t ptep_get(pte_t *ptep)
274 {
275 	pte_basic_t val = READ_ONCE(ptep->pte);
276 	pte_t pte = {val, val, val, val};
277 
278 	return pte;
279 }
280 #endif /* CONFIG_PPC_16K_PAGES */
281 
282 #else
283 static inline pte_basic_t pte_update(struct mm_struct *mm, unsigned long addr, pte_t *p,
284 				     unsigned long clr, unsigned long set, int huge)
285 {
286 	pte_basic_t old = pte_val(*p);
287 	pte_basic_t new = (old & ~(pte_basic_t)clr) | set;
288 
289 	*p = __pte(new);
290 
291 #ifdef CONFIG_44x
292 	if ((old & _PAGE_USER) && (old & _PAGE_EXEC))
293 		icache_44x_need_flush = 1;
294 #endif
295 	return old;
296 }
297 #endif
298 
299 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
300 static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
301 					      unsigned long addr, pte_t *ptep)
302 {
303 	unsigned long old;
304 	old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0, 0);
305 	return (old & _PAGE_ACCESSED) != 0;
306 }
307 #define ptep_test_and_clear_young(__vma, __addr, __ptep) \
308 	__ptep_test_and_clear_young((__vma)->vm_mm, __addr, __ptep)
309 
310 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
311 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
312 				       pte_t *ptep)
313 {
314 	return __pte(pte_update(mm, addr, ptep, ~0, 0, 0));
315 }
316 
317 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
318 #ifndef ptep_set_wrprotect
319 static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
320 				      pte_t *ptep)
321 {
322 	pte_update(mm, addr, ptep, _PAGE_RW, 0, 0);
323 }
324 #endif
325 
326 #ifndef __ptep_set_access_flags
327 static inline void __ptep_set_access_flags(struct vm_area_struct *vma,
328 					   pte_t *ptep, pte_t entry,
329 					   unsigned long address,
330 					   int psize)
331 {
332 	unsigned long set = pte_val(entry) &
333 			    (_PAGE_DIRTY | _PAGE_ACCESSED | _PAGE_RW | _PAGE_EXEC);
334 	int huge = psize > mmu_virtual_psize ? 1 : 0;
335 
336 	pte_update(vma->vm_mm, address, ptep, 0, set, huge);
337 
338 	flush_tlb_page(vma, address);
339 }
340 #endif
341 
342 static inline int pte_young(pte_t pte)
343 {
344 	return pte_val(pte) & _PAGE_ACCESSED;
345 }
346 
347 /*
348  * Note that on Book E processors, the pmd contains the kernel virtual
349  * (lowmem) address of the pte page.  The physical address is less useful
350  * because everything runs with translation enabled (even the TLB miss
351  * handler).  On everything else the pmd contains the physical address
352  * of the pte page.  -- paulus
353  */
354 #ifndef CONFIG_BOOKE
355 #define pmd_pfn(pmd)		(pmd_val(pmd) >> PAGE_SHIFT)
356 #else
357 #define pmd_page_vaddr(pmd)	\
358 	((unsigned long)(pmd_val(pmd) & ~(PTE_TABLE_SIZE - 1)))
359 #define pmd_pfn(pmd)		(__pa(pmd_val(pmd)) >> PAGE_SHIFT)
360 #endif
361 
362 #define pmd_page(pmd)		pfn_to_page(pmd_pfn(pmd))
363 
364 /*
365  * Encode/decode swap entries and swap PTEs. Swap PTEs are all PTEs that
366  * are !pte_none() && !pte_present().
367  *
368  * Format of swap PTEs (32bit PTEs):
369  *
370  *                         1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3
371  *   0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
372  *   <------------------ offset -------------------> < type -> E 0 0
373  *
374  * E is the exclusive marker that is not stored in swap entries.
375  *
376  * For 64bit PTEs, the offset is extended by 32bit.
377  */
378 #define __swp_type(entry)		((entry).val & 0x1f)
379 #define __swp_offset(entry)		((entry).val >> 5)
380 #define __swp_entry(type, offset)	((swp_entry_t) { ((type) & 0x1f) | ((offset) << 5) })
381 #define __pte_to_swp_entry(pte)		((swp_entry_t) { pte_val(pte) >> 3 })
382 #define __swp_entry_to_pte(x)		((pte_t) { (x).val << 3 })
383 
384 /* We borrow LSB 2 to store the exclusive marker in swap PTEs. */
385 #define _PAGE_SWP_EXCLUSIVE	0x000004
386 
387 #endif /* !__ASSEMBLY__ */
388 
389 #endif /* __ASM_POWERPC_NOHASH_32_PGTABLE_H */
390