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> |
d9b2b2a2 | 43 | #include <linux/lmb.h> |
14cf11af PM |
44 | |
45 | #include <asm/pgalloc.h> | |
46 | #include <asm/page.h> | |
47 | #include <asm/prom.h> | |
14cf11af PM |
48 | #include <asm/rtas.h> |
49 | #include <asm/io.h> | |
50 | #include <asm/mmu_context.h> | |
51 | #include <asm/pgtable.h> | |
52 | #include <asm/mmu.h> | |
53 | #include <asm/uaccess.h> | |
54 | #include <asm/smp.h> | |
55 | #include <asm/machdep.h> | |
56 | #include <asm/tlb.h> | |
57 | #include <asm/eeh.h> | |
58 | #include <asm/processor.h> | |
59 | #include <asm/mmzone.h> | |
60 | #include <asm/cputable.h> | |
14cf11af PM |
61 | #include <asm/sections.h> |
62 | #include <asm/system.h> | |
63 | #include <asm/iommu.h> | |
64 | #include <asm/abs_addr.h> | |
65 | #include <asm/vdso.h> | |
800fc3ee DG |
66 | |
67 | #include "mmu_decl.h" | |
14cf11af | 68 | |
94491685 | 69 | #ifdef CONFIG_PPC_STD_MMU_64 |
14cf11af PM |
70 | #if PGTABLE_RANGE > USER_VSID_RANGE |
71 | #warning Limited user VSID range means pagetable space is wasted | |
72 | #endif | |
73 | ||
74 | #if (TASK_SIZE_USER64 < PGTABLE_RANGE) && (TASK_SIZE_USER64 < USER_VSID_RANGE) | |
75 | #warning TASK_SIZE is smaller than it needs to be. | |
76 | #endif | |
94491685 | 77 | #endif /* CONFIG_PPC_STD_MMU_64 */ |
14cf11af | 78 | |
37dd2bad KG |
79 | phys_addr_t memstart_addr = ~0; |
80 | phys_addr_t kernstart_addr; | |
d7917ba7 | 81 | |
14cf11af PM |
82 | void free_initmem(void) |
83 | { | |
84 | unsigned long addr; | |
85 | ||
86 | addr = (unsigned long)__init_begin; | |
87 | for (; addr < (unsigned long)__init_end; addr += PAGE_SIZE) { | |
c9cf5528 | 88 | memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE); |
14cf11af | 89 | ClearPageReserved(virt_to_page(addr)); |
7835e98b | 90 | init_page_count(virt_to_page(addr)); |
14cf11af PM |
91 | free_page(addr); |
92 | totalram_pages++; | |
93 | } | |
94 | printk ("Freeing unused kernel memory: %luk freed\n", | |
95 | ((unsigned long)__init_end - (unsigned long)__init_begin) >> 10); | |
96 | } | |
97 | ||
98 | #ifdef CONFIG_BLK_DEV_INITRD | |
99 | void free_initrd_mem(unsigned long start, unsigned long end) | |
100 | { | |
101 | if (start < end) | |
102 | printk ("Freeing initrd memory: %ldk freed\n", (end - start) >> 10); | |
103 | for (; start < end; start += PAGE_SIZE) { | |
104 | ClearPageReserved(virt_to_page(start)); | |
7835e98b | 105 | init_page_count(virt_to_page(start)); |
14cf11af PM |
106 | free_page(start); |
107 | totalram_pages++; | |
108 | } | |
109 | } | |
110 | #endif | |
111 | ||
df174e3b | 112 | #ifdef CONFIG_PROC_KCORE |
14cf11af PM |
113 | static struct kcore_list kcore_vmem; |
114 | ||
115 | static int __init setup_kcore(void) | |
116 | { | |
117 | int i; | |
118 | ||
119 | for (i=0; i < lmb.memory.cnt; i++) { | |
120 | unsigned long base, size; | |
121 | struct kcore_list *kcore_mem; | |
122 | ||
123 | base = lmb.memory.region[i].base; | |
124 | size = lmb.memory.region[i].size; | |
125 | ||
126 | /* GFP_ATOMIC to avoid might_sleep warnings during boot */ | |
127 | kcore_mem = kmalloc(sizeof(struct kcore_list), GFP_ATOMIC); | |
128 | if (!kcore_mem) | |
e48b1b45 | 129 | panic("%s: kmalloc failed\n", __func__); |
14cf11af PM |
130 | |
131 | kclist_add(kcore_mem, __va(base), size); | |
132 | } | |
133 | ||
134 | kclist_add(&kcore_vmem, (void *)VMALLOC_START, VMALLOC_END-VMALLOC_START); | |
135 | ||
136 | return 0; | |
137 | } | |
138 | module_init(setup_kcore); | |
df174e3b | 139 | #endif |
14cf11af | 140 | |
51cc5068 | 141 | static void pgd_ctor(void *addr) |
14cf11af | 142 | { |
51cc5068 AD |
143 | memset(addr, 0, PGD_TABLE_SIZE); |
144 | } | |
145 | ||
146 | static void pmd_ctor(void *addr) | |
147 | { | |
148 | memset(addr, 0, PMD_TABLE_SIZE); | |
14cf11af PM |
149 | } |
150 | ||
87655ff2 | 151 | static const unsigned int pgtable_cache_size[2] = { |
517e2263 | 152 | PGD_TABLE_SIZE, PMD_TABLE_SIZE |
14cf11af PM |
153 | }; |
154 | static const char *pgtable_cache_name[ARRAY_SIZE(pgtable_cache_size)] = { | |
517e2263 HD |
155 | #ifdef CONFIG_PPC_64K_PAGES |
156 | "pgd_cache", "pmd_cache", | |
157 | #else | |
158 | "pgd_cache", "pud_pmd_cache", | |
3c726f8d | 159 | #endif /* CONFIG_PPC_64K_PAGES */ |
517e2263 | 160 | }; |
14cf11af | 161 | |
f10a04c0 | 162 | #ifdef CONFIG_HUGETLB_PAGE |
0d9ea754 JT |
163 | /* Hugepages need an extra cache per hugepagesize, initialized in |
164 | * hugetlbpage.c. We can't put into the tables above, because HPAGE_SHIFT | |
165 | * is not compile time constant. */ | |
166 | struct kmem_cache *pgtable_cache[ARRAY_SIZE(pgtable_cache_size)+MMU_PAGE_COUNT]; | |
f10a04c0 | 167 | #else |
e18b890b | 168 | struct kmem_cache *pgtable_cache[ARRAY_SIZE(pgtable_cache_size)]; |
f10a04c0 | 169 | #endif |
14cf11af PM |
170 | |
171 | void pgtable_cache_init(void) | |
172 | { | |
51cc5068 AD |
173 | pgtable_cache[0] = kmem_cache_create(pgtable_cache_name[0], PGD_TABLE_SIZE, PGD_TABLE_SIZE, SLAB_PANIC, pgd_ctor); |
174 | pgtable_cache[1] = kmem_cache_create(pgtable_cache_name[1], PMD_TABLE_SIZE, PMD_TABLE_SIZE, SLAB_PANIC, pmd_ctor); | |
14cf11af | 175 | } |
d29eff7b AW |
176 | |
177 | #ifdef CONFIG_SPARSEMEM_VMEMMAP | |
178 | /* | |
179 | * Given an address within the vmemmap, determine the pfn of the page that | |
180 | * represents the start of the section it is within. Note that we have to | |
181 | * do this by hand as the proffered address may not be correctly aligned. | |
182 | * Subtraction of non-aligned pointers produces undefined results. | |
183 | */ | |
09de9ff8 | 184 | static unsigned long __meminit vmemmap_section_start(unsigned long page) |
d29eff7b AW |
185 | { |
186 | unsigned long offset = page - ((unsigned long)(vmemmap)); | |
187 | ||
188 | /* Return the pfn of the start of the section. */ | |
189 | return (offset / sizeof(struct page)) & PAGE_SECTION_MASK; | |
190 | } | |
191 | ||
192 | /* | |
193 | * Check if this vmemmap page is already initialised. If any section | |
194 | * which overlaps this vmemmap page is initialised then this page is | |
195 | * initialised already. | |
196 | */ | |
09de9ff8 | 197 | static int __meminit vmemmap_populated(unsigned long start, int page_size) |
d29eff7b AW |
198 | { |
199 | unsigned long end = start + page_size; | |
200 | ||
201 | for (; start < end; start += (PAGES_PER_SECTION * sizeof(struct page))) | |
202 | if (pfn_valid(vmemmap_section_start(start))) | |
203 | return 1; | |
204 | ||
205 | return 0; | |
206 | } | |
207 | ||
32a74949 BH |
208 | /* On hash-based CPUs, the vmemmap is bolted in the hash table. |
209 | * | |
210 | * On Book3E CPUs, the vmemmap is currently mapped in the top half of | |
211 | * the vmalloc space using normal page tables, though the size of | |
212 | * pages encoded in the PTEs can be different | |
213 | */ | |
214 | ||
215 | #ifdef CONFIG_PPC_BOOK3E | |
216 | static void __meminit vmemmap_create_mapping(unsigned long start, | |
217 | unsigned long page_size, | |
218 | unsigned long phys) | |
219 | { | |
220 | /* Create a PTE encoding without page size */ | |
221 | unsigned long i, flags = _PAGE_PRESENT | _PAGE_ACCESSED | | |
222 | _PAGE_KERNEL_RW; | |
223 | ||
224 | /* PTEs only contain page size encodings up to 32M */ | |
225 | BUG_ON(mmu_psize_defs[mmu_vmemmap_psize].enc > 0xf); | |
226 | ||
227 | /* Encode the size in the PTE */ | |
228 | flags |= mmu_psize_defs[mmu_vmemmap_psize].enc << 8; | |
229 | ||
230 | /* For each PTE for that area, map things. Note that we don't | |
231 | * increment phys because all PTEs are of the large size and | |
232 | * thus must have the low bits clear | |
233 | */ | |
234 | for (i = 0; i < page_size; i += PAGE_SIZE) | |
235 | BUG_ON(map_kernel_page(start + i, phys, flags)); | |
236 | } | |
237 | #else /* CONFIG_PPC_BOOK3E */ | |
238 | static void __meminit vmemmap_create_mapping(unsigned long start, | |
239 | unsigned long page_size, | |
240 | unsigned long phys) | |
241 | { | |
242 | int mapped = htab_bolt_mapping(start, start + page_size, phys, | |
243 | PAGE_KERNEL, mmu_vmemmap_psize, | |
244 | mmu_kernel_ssize); | |
245 | BUG_ON(mapped < 0); | |
246 | } | |
247 | #endif /* CONFIG_PPC_BOOK3E */ | |
248 | ||
d29eff7b | 249 | int __meminit vmemmap_populate(struct page *start_page, |
cec08e7a | 250 | unsigned long nr_pages, int node) |
d29eff7b | 251 | { |
d29eff7b AW |
252 | unsigned long start = (unsigned long)start_page; |
253 | unsigned long end = (unsigned long)(start_page + nr_pages); | |
cec08e7a | 254 | unsigned long page_size = 1 << mmu_psize_defs[mmu_vmemmap_psize].shift; |
d29eff7b | 255 | |
d29eff7b AW |
256 | /* Align to the page size of the linear mapping. */ |
257 | start = _ALIGN_DOWN(start, page_size); | |
258 | ||
32a74949 BH |
259 | pr_debug("vmemmap_populate page %p, %ld pages, node %d\n", |
260 | start_page, nr_pages, node); | |
261 | pr_debug(" -> map %lx..%lx\n", start, end); | |
262 | ||
d29eff7b | 263 | for (; start < end; start += page_size) { |
d29eff7b AW |
264 | void *p; |
265 | ||
266 | if (vmemmap_populated(start, page_size)) | |
267 | continue; | |
268 | ||
269 | p = vmemmap_alloc_block(page_size, node); | |
270 | if (!p) | |
271 | return -ENOMEM; | |
272 | ||
32a74949 BH |
273 | pr_debug(" * %016lx..%016lx allocated at %p\n", |
274 | start, start + page_size, p); | |
d29eff7b | 275 | |
32a74949 | 276 | vmemmap_create_mapping(start, page_size, __pa(p)); |
d29eff7b AW |
277 | } |
278 | ||
279 | return 0; | |
280 | } | |
cec08e7a | 281 | #endif /* CONFIG_SPARSEMEM_VMEMMAP */ |