Commit | Line | Data |
---|---|---|
14cf11af PM |
1 | /* |
2 | * PowerPC version | |
3 | * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org) | |
4 | * | |
5 | * Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au) | |
6 | * and Cort Dougan (PReP) (cort@cs.nmt.edu) | |
7 | * Copyright (C) 1996 Paul Mackerras | |
14cf11af PM |
8 | * |
9 | * Derived from "arch/i386/mm/init.c" | |
10 | * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds | |
11 | * | |
12 | * Dave Engebretsen <engebret@us.ibm.com> | |
13 | * Rework for PPC64 port. | |
14 | * | |
15 | * This program is free software; you can redistribute it and/or | |
16 | * modify it under the terms of the GNU General Public License | |
17 | * as published by the Free Software Foundation; either version | |
18 | * 2 of the License, or (at your option) any later version. | |
19 | * | |
20 | */ | |
21 | ||
cec08e7a BH |
22 | #undef DEBUG |
23 | ||
14cf11af PM |
24 | #include <linux/signal.h> |
25 | #include <linux/sched.h> | |
26 | #include <linux/kernel.h> | |
27 | #include <linux/errno.h> | |
28 | #include <linux/string.h> | |
29 | #include <linux/types.h> | |
30 | #include <linux/mman.h> | |
31 | #include <linux/mm.h> | |
32 | #include <linux/swap.h> | |
33 | #include <linux/stddef.h> | |
34 | #include <linux/vmalloc.h> | |
35 | #include <linux/init.h> | |
36 | #include <linux/delay.h> | |
37 | #include <linux/bootmem.h> | |
38 | #include <linux/highmem.h> | |
39 | #include <linux/idr.h> | |
40 | #include <linux/nodemask.h> | |
41 | #include <linux/module.h> | |
c9cf5528 | 42 | #include <linux/poison.h> |
95f72d1e | 43 | #include <linux/memblock.h> |
a4fe3ce7 | 44 | #include <linux/hugetlb.h> |
5a0e3ad6 | 45 | #include <linux/slab.h> |
14cf11af PM |
46 | |
47 | #include <asm/pgalloc.h> | |
48 | #include <asm/page.h> | |
49 | #include <asm/prom.h> | |
14cf11af PM |
50 | #include <asm/rtas.h> |
51 | #include <asm/io.h> | |
52 | #include <asm/mmu_context.h> | |
53 | #include <asm/pgtable.h> | |
54 | #include <asm/mmu.h> | |
55 | #include <asm/uaccess.h> | |
56 | #include <asm/smp.h> | |
57 | #include <asm/machdep.h> | |
58 | #include <asm/tlb.h> | |
59 | #include <asm/eeh.h> | |
60 | #include <asm/processor.h> | |
61 | #include <asm/mmzone.h> | |
62 | #include <asm/cputable.h> | |
14cf11af PM |
63 | #include <asm/sections.h> |
64 | #include <asm/system.h> | |
65 | #include <asm/iommu.h> | |
66 | #include <asm/abs_addr.h> | |
67 | #include <asm/vdso.h> | |
800fc3ee DG |
68 | |
69 | #include "mmu_decl.h" | |
14cf11af | 70 | |
94491685 | 71 | #ifdef CONFIG_PPC_STD_MMU_64 |
14cf11af PM |
72 | #if PGTABLE_RANGE > USER_VSID_RANGE |
73 | #warning Limited user VSID range means pagetable space is wasted | |
74 | #endif | |
75 | ||
76 | #if (TASK_SIZE_USER64 < PGTABLE_RANGE) && (TASK_SIZE_USER64 < USER_VSID_RANGE) | |
77 | #warning TASK_SIZE is smaller than it needs to be. | |
78 | #endif | |
94491685 | 79 | #endif /* CONFIG_PPC_STD_MMU_64 */ |
14cf11af | 80 | |
37dd2bad KG |
81 | phys_addr_t memstart_addr = ~0; |
82 | phys_addr_t kernstart_addr; | |
d7917ba7 | 83 | |
14cf11af PM |
84 | void free_initmem(void) |
85 | { | |
86 | unsigned long addr; | |
87 | ||
88 | addr = (unsigned long)__init_begin; | |
89 | for (; addr < (unsigned long)__init_end; addr += PAGE_SIZE) { | |
c9cf5528 | 90 | memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE); |
14cf11af | 91 | ClearPageReserved(virt_to_page(addr)); |
7835e98b | 92 | init_page_count(virt_to_page(addr)); |
14cf11af PM |
93 | free_page(addr); |
94 | totalram_pages++; | |
95 | } | |
96 | printk ("Freeing unused kernel memory: %luk freed\n", | |
97 | ((unsigned long)__init_end - (unsigned long)__init_begin) >> 10); | |
98 | } | |
99 | ||
100 | #ifdef CONFIG_BLK_DEV_INITRD | |
101 | void free_initrd_mem(unsigned long start, unsigned long end) | |
102 | { | |
103 | if (start < end) | |
104 | printk ("Freeing initrd memory: %ldk freed\n", (end - start) >> 10); | |
105 | for (; start < end; start += PAGE_SIZE) { | |
106 | ClearPageReserved(virt_to_page(start)); | |
7835e98b | 107 | init_page_count(virt_to_page(start)); |
14cf11af PM |
108 | free_page(start); |
109 | totalram_pages++; | |
110 | } | |
111 | } | |
112 | #endif | |
113 | ||
51cc5068 | 114 | static void pgd_ctor(void *addr) |
14cf11af | 115 | { |
51cc5068 AD |
116 | memset(addr, 0, PGD_TABLE_SIZE); |
117 | } | |
118 | ||
119 | static void pmd_ctor(void *addr) | |
120 | { | |
121 | memset(addr, 0, PMD_TABLE_SIZE); | |
14cf11af PM |
122 | } |
123 | ||
a0668cdc DG |
124 | struct kmem_cache *pgtable_cache[MAX_PGTABLE_INDEX_SIZE]; |
125 | ||
126 | /* | |
127 | * Create a kmem_cache() for pagetables. This is not used for PTE | |
128 | * pages - they're linked to struct page, come from the normal free | |
129 | * pages pool and have a different entry size (see real_pte_t) to | |
130 | * everything else. Caches created by this function are used for all | |
131 | * the higher level pagetables, and for hugepage pagetables. | |
132 | */ | |
133 | void pgtable_cache_add(unsigned shift, void (*ctor)(void *)) | |
134 | { | |
135 | char *name; | |
136 | unsigned long table_size = sizeof(void *) << shift; | |
137 | unsigned long align = table_size; | |
138 | ||
139 | /* When batching pgtable pointers for RCU freeing, we store | |
140 | * the index size in the low bits. Table alignment must be | |
a4fe3ce7 DG |
141 | * big enough to fit it. |
142 | * | |
143 | * Likewise, hugeapge pagetable pointers contain a (different) | |
144 | * shift value in the low bits. All tables must be aligned so | |
145 | * as to leave enough 0 bits in the address to contain it. */ | |
146 | unsigned long minalign = max(MAX_PGTABLE_INDEX_SIZE + 1, | |
147 | HUGEPD_SHIFT_MASK + 1); | |
a0668cdc DG |
148 | struct kmem_cache *new; |
149 | ||
150 | /* It would be nice if this was a BUILD_BUG_ON(), but at the | |
151 | * moment, gcc doesn't seem to recognize is_power_of_2 as a | |
152 | * constant expression, so so much for that. */ | |
153 | BUG_ON(!is_power_of_2(minalign)); | |
154 | BUG_ON((shift < 1) || (shift > MAX_PGTABLE_INDEX_SIZE)); | |
155 | ||
156 | if (PGT_CACHE(shift)) | |
157 | return; /* Already have a cache of this size */ | |
158 | ||
159 | align = max_t(unsigned long, align, minalign); | |
160 | name = kasprintf(GFP_KERNEL, "pgtable-2^%d", shift); | |
161 | new = kmem_cache_create(name, table_size, align, 0, ctor); | |
162 | PGT_CACHE(shift) = new; | |
163 | ||
164 | pr_debug("Allocated pgtable cache for order %d\n", shift); | |
165 | } | |
166 | ||
14cf11af PM |
167 | |
168 | void pgtable_cache_init(void) | |
169 | { | |
a0668cdc DG |
170 | pgtable_cache_add(PGD_INDEX_SIZE, pgd_ctor); |
171 | pgtable_cache_add(PMD_INDEX_SIZE, pmd_ctor); | |
172 | if (!PGT_CACHE(PGD_INDEX_SIZE) || !PGT_CACHE(PMD_INDEX_SIZE)) | |
173 | panic("Couldn't allocate pgtable caches"); | |
174 | ||
175 | /* In all current configs, when the PUD index exists it's the | |
176 | * same size as either the pgd or pmd index. Verify that the | |
177 | * initialization above has also created a PUD cache. This | |
178 | * will need re-examiniation if we add new possibilities for | |
179 | * the pagetable layout. */ | |
180 | BUG_ON(PUD_INDEX_SIZE && !PGT_CACHE(PUD_INDEX_SIZE)); | |
14cf11af | 181 | } |
d29eff7b AW |
182 | |
183 | #ifdef CONFIG_SPARSEMEM_VMEMMAP | |
184 | /* | |
185 | * Given an address within the vmemmap, determine the pfn of the page that | |
186 | * represents the start of the section it is within. Note that we have to | |
187 | * do this by hand as the proffered address may not be correctly aligned. | |
188 | * Subtraction of non-aligned pointers produces undefined results. | |
189 | */ | |
09de9ff8 | 190 | static unsigned long __meminit vmemmap_section_start(unsigned long page) |
d29eff7b AW |
191 | { |
192 | unsigned long offset = page - ((unsigned long)(vmemmap)); | |
193 | ||
194 | /* Return the pfn of the start of the section. */ | |
195 | return (offset / sizeof(struct page)) & PAGE_SECTION_MASK; | |
196 | } | |
197 | ||
198 | /* | |
199 | * Check if this vmemmap page is already initialised. If any section | |
200 | * which overlaps this vmemmap page is initialised then this page is | |
201 | * initialised already. | |
202 | */ | |
09de9ff8 | 203 | static int __meminit vmemmap_populated(unsigned long start, int page_size) |
d29eff7b AW |
204 | { |
205 | unsigned long end = start + page_size; | |
206 | ||
207 | for (; start < end; start += (PAGES_PER_SECTION * sizeof(struct page))) | |
208 | if (pfn_valid(vmemmap_section_start(start))) | |
209 | return 1; | |
210 | ||
211 | return 0; | |
212 | } | |
213 | ||
32a74949 BH |
214 | /* On hash-based CPUs, the vmemmap is bolted in the hash table. |
215 | * | |
216 | * On Book3E CPUs, the vmemmap is currently mapped in the top half of | |
217 | * the vmalloc space using normal page tables, though the size of | |
218 | * pages encoded in the PTEs can be different | |
219 | */ | |
220 | ||
221 | #ifdef CONFIG_PPC_BOOK3E | |
222 | static void __meminit vmemmap_create_mapping(unsigned long start, | |
223 | unsigned long page_size, | |
224 | unsigned long phys) | |
225 | { | |
226 | /* Create a PTE encoding without page size */ | |
227 | unsigned long i, flags = _PAGE_PRESENT | _PAGE_ACCESSED | | |
228 | _PAGE_KERNEL_RW; | |
229 | ||
230 | /* PTEs only contain page size encodings up to 32M */ | |
231 | BUG_ON(mmu_psize_defs[mmu_vmemmap_psize].enc > 0xf); | |
232 | ||
233 | /* Encode the size in the PTE */ | |
234 | flags |= mmu_psize_defs[mmu_vmemmap_psize].enc << 8; | |
235 | ||
236 | /* For each PTE for that area, map things. Note that we don't | |
237 | * increment phys because all PTEs are of the large size and | |
238 | * thus must have the low bits clear | |
239 | */ | |
240 | for (i = 0; i < page_size; i += PAGE_SIZE) | |
241 | BUG_ON(map_kernel_page(start + i, phys, flags)); | |
242 | } | |
243 | #else /* CONFIG_PPC_BOOK3E */ | |
244 | static void __meminit vmemmap_create_mapping(unsigned long start, | |
245 | unsigned long page_size, | |
246 | unsigned long phys) | |
247 | { | |
248 | int mapped = htab_bolt_mapping(start, start + page_size, phys, | |
249 | PAGE_KERNEL, mmu_vmemmap_psize, | |
250 | mmu_kernel_ssize); | |
251 | BUG_ON(mapped < 0); | |
252 | } | |
253 | #endif /* CONFIG_PPC_BOOK3E */ | |
254 | ||
91eea67c MN |
255 | struct vmemmap_backing *vmemmap_list; |
256 | ||
257 | static __meminit struct vmemmap_backing * vmemmap_list_alloc(int node) | |
258 | { | |
259 | static struct vmemmap_backing *next; | |
260 | static int num_left; | |
261 | ||
262 | /* allocate a page when required and hand out chunks */ | |
263 | if (!next || !num_left) { | |
264 | next = vmemmap_alloc_block(PAGE_SIZE, node); | |
265 | if (unlikely(!next)) { | |
266 | WARN_ON(1); | |
267 | return NULL; | |
268 | } | |
269 | num_left = PAGE_SIZE / sizeof(struct vmemmap_backing); | |
270 | } | |
271 | ||
272 | num_left--; | |
273 | ||
274 | return next++; | |
275 | } | |
276 | ||
277 | static __meminit void vmemmap_list_populate(unsigned long phys, | |
278 | unsigned long start, | |
279 | int node) | |
280 | { | |
281 | struct vmemmap_backing *vmem_back; | |
282 | ||
283 | vmem_back = vmemmap_list_alloc(node); | |
284 | if (unlikely(!vmem_back)) { | |
285 | WARN_ON(1); | |
286 | return; | |
287 | } | |
288 | ||
289 | vmem_back->phys = phys; | |
290 | vmem_back->virt_addr = start; | |
291 | vmem_back->list = vmemmap_list; | |
292 | ||
293 | vmemmap_list = vmem_back; | |
294 | } | |
295 | ||
d29eff7b | 296 | int __meminit vmemmap_populate(struct page *start_page, |
cec08e7a | 297 | unsigned long nr_pages, int node) |
d29eff7b | 298 | { |
d29eff7b AW |
299 | unsigned long start = (unsigned long)start_page; |
300 | unsigned long end = (unsigned long)(start_page + nr_pages); | |
cec08e7a | 301 | unsigned long page_size = 1 << mmu_psize_defs[mmu_vmemmap_psize].shift; |
d29eff7b | 302 | |
d29eff7b AW |
303 | /* Align to the page size of the linear mapping. */ |
304 | start = _ALIGN_DOWN(start, page_size); | |
305 | ||
32a74949 BH |
306 | pr_debug("vmemmap_populate page %p, %ld pages, node %d\n", |
307 | start_page, nr_pages, node); | |
308 | pr_debug(" -> map %lx..%lx\n", start, end); | |
309 | ||
d29eff7b | 310 | for (; start < end; start += page_size) { |
d29eff7b AW |
311 | void *p; |
312 | ||
313 | if (vmemmap_populated(start, page_size)) | |
314 | continue; | |
315 | ||
316 | p = vmemmap_alloc_block(page_size, node); | |
317 | if (!p) | |
318 | return -ENOMEM; | |
319 | ||
91eea67c MN |
320 | vmemmap_list_populate(__pa(p), start, node); |
321 | ||
32a74949 BH |
322 | pr_debug(" * %016lx..%016lx allocated at %p\n", |
323 | start, start + page_size, p); | |
d29eff7b | 324 | |
32a74949 | 325 | vmemmap_create_mapping(start, page_size, __pa(p)); |
d29eff7b AW |
326 | } |
327 | ||
328 | return 0; | |
329 | } | |
cec08e7a | 330 | #endif /* CONFIG_SPARSEMEM_VMEMMAP */ |