Commit | Line | Data |
---|---|---|
a95d0e16 MS |
1 | /* |
2 | * Copyright (C) 2007-2008 Michal Simek <monstr@monstr.eu> | |
3 | * Copyright (C) 2006 Atmark Techno, Inc. | |
4 | * | |
5 | * This file is subject to the terms and conditions of the GNU General Public | |
6 | * License. See the file "COPYING" in the main directory of this archive | |
7 | * for more details. | |
8 | */ | |
9 | ||
10 | #include <linux/bootmem.h> | |
11 | #include <linux/init.h> | |
12 | #include <linux/kernel.h> | |
95f72d1e | 13 | #include <linux/memblock.h> |
a95d0e16 MS |
14 | #include <linux/mm.h> /* mem_init */ |
15 | #include <linux/initrd.h> | |
16 | #include <linux/pagemap.h> | |
17 | #include <linux/pfn.h> | |
5a0e3ad6 | 18 | #include <linux/slab.h> |
a95d0e16 | 19 | #include <linux/swap.h> |
66421a64 | 20 | #include <linux/export.h> |
a95d0e16 MS |
21 | |
22 | #include <asm/page.h> | |
23 | #include <asm/mmu_context.h> | |
24 | #include <asm/pgalloc.h> | |
25 | #include <asm/sections.h> | |
26 | #include <asm/tlb.h> | |
41938761 | 27 | #include <asm/fixmap.h> |
a95d0e16 | 28 | |
79bf3a13 MS |
29 | /* Use for MMU and noMMU because of PCI generic code */ |
30 | int mem_init_done; | |
31 | ||
4dc60832 | 32 | #ifndef CONFIG_MMU |
a95d0e16 | 33 | unsigned int __page_offset; |
5af7fa68 | 34 | EXPORT_SYMBOL(__page_offset); |
a95d0e16 | 35 | |
4dc60832 | 36 | #else |
4dc60832 MS |
37 | static int init_bootmem_done; |
38 | #endif /* CONFIG_MMU */ | |
39 | ||
a95d0e16 MS |
40 | char *klimit = _end; |
41 | ||
42 | /* | |
43 | * Initialize the bootmem system and give it all the memory we | |
44 | * have available. | |
45 | */ | |
4dc60832 | 46 | unsigned long memory_start; |
fd6ed51f | 47 | EXPORT_SYMBOL(memory_start); |
4dc60832 | 48 | unsigned long memory_size; |
ee4bcdf1 | 49 | EXPORT_SYMBOL(memory_size); |
83a92529 | 50 | unsigned long lowmem_size; |
a95d0e16 | 51 | |
2f2f371f MS |
52 | #ifdef CONFIG_HIGHMEM |
53 | pte_t *kmap_pte; | |
54 | EXPORT_SYMBOL(kmap_pte); | |
55 | pgprot_t kmap_prot; | |
56 | EXPORT_SYMBOL(kmap_prot); | |
57 | ||
58 | static inline pte_t *virt_to_kpte(unsigned long vaddr) | |
59 | { | |
60 | return pte_offset_kernel(pmd_offset(pgd_offset_k(vaddr), | |
61 | vaddr), vaddr); | |
62 | } | |
63 | ||
64 | static void __init highmem_init(void) | |
65 | { | |
66 | pr_debug("%x\n", (u32)PKMAP_BASE); | |
67 | map_page(PKMAP_BASE, 0, 0); /* XXX gross */ | |
68 | pkmap_page_table = virt_to_kpte(PKMAP_BASE); | |
69 | ||
70 | kmap_pte = virt_to_kpte(__fix_to_virt(FIX_KMAP_BEGIN)); | |
71 | kmap_prot = PAGE_KERNEL; | |
72 | } | |
73 | ||
74 | static unsigned long highmem_setup(void) | |
75 | { | |
76 | unsigned long pfn; | |
77 | unsigned long reservedpages = 0; | |
78 | ||
79 | for (pfn = max_low_pfn; pfn < max_pfn; ++pfn) { | |
80 | struct page *page = pfn_to_page(pfn); | |
81 | ||
82 | /* FIXME not sure about */ | |
83 | if (memblock_is_reserved(pfn << PAGE_SHIFT)) | |
84 | continue; | |
85 | ClearPageReserved(page); | |
86 | init_page_count(page); | |
87 | __free_page(page); | |
88 | totalhigh_pages++; | |
89 | reservedpages++; | |
90 | } | |
91 | totalram_pages += totalhigh_pages; | |
92 | printk(KERN_INFO "High memory: %luk\n", | |
93 | totalhigh_pages << (PAGE_SHIFT-10)); | |
94 | ||
95 | return reservedpages; | |
96 | } | |
97 | #endif /* CONFIG_HIGHMEM */ | |
98 | ||
a95d0e16 MS |
99 | /* |
100 | * paging_init() sets up the page tables - in fact we've already done this. | |
101 | */ | |
102 | static void __init paging_init(void) | |
103 | { | |
a95d0e16 | 104 | unsigned long zones_size[MAX_NR_ZONES]; |
41938761 MS |
105 | #ifdef CONFIG_MMU |
106 | int idx; | |
107 | ||
108 | /* Setup fixmaps */ | |
109 | for (idx = 0; idx < __end_of_fixed_addresses; idx++) | |
110 | clear_fixmap(idx); | |
111 | #endif | |
a95d0e16 | 112 | |
5af90438 SM |
113 | /* Clean every zones */ |
114 | memset(zones_size, 0, sizeof(zones_size)); | |
115 | ||
2f2f371f MS |
116 | #ifdef CONFIG_HIGHMEM |
117 | highmem_init(); | |
118 | ||
119 | zones_size[ZONE_DMA] = max_low_pfn; | |
120 | zones_size[ZONE_HIGHMEM] = max_pfn; | |
121 | #else | |
83a92529 | 122 | zones_size[ZONE_DMA] = max_pfn; |
2f2f371f | 123 | #endif |
a95d0e16 | 124 | |
baab8a82 MS |
125 | /* We don't have holes in memory map */ |
126 | free_area_init_nodes(zones_size); | |
a95d0e16 MS |
127 | } |
128 | ||
129 | void __init setup_memory(void) | |
130 | { | |
a95d0e16 | 131 | unsigned long map_size; |
76bfcc81 BH |
132 | struct memblock_region *reg; |
133 | ||
4dc60832 | 134 | #ifndef CONFIG_MMU |
a95d0e16 MS |
135 | u32 kernel_align_start, kernel_align_size; |
136 | ||
137 | /* Find main memory where is the kernel */ | |
76bfcc81 BH |
138 | for_each_memblock(memory, reg) { |
139 | memory_start = (u32)reg->base; | |
83a92529 | 140 | lowmem_size = reg->size; |
a95d0e16 | 141 | if ((memory_start <= (u32)_text) && |
83a92529 MS |
142 | ((u32)_text <= (memory_start + lowmem_size - 1))) { |
143 | memory_size = lowmem_size; | |
a95d0e16 | 144 | PAGE_OFFSET = memory_start; |
83a92529 | 145 | printk(KERN_INFO "%s: Main mem: 0x%x, " |
db6e3f91 | 146 | "size 0x%08x\n", __func__, (u32) memory_start, |
83a92529 | 147 | (u32) memory_size); |
a95d0e16 MS |
148 | break; |
149 | } | |
150 | } | |
151 | ||
83a92529 MS |
152 | if (!memory_start || !memory_size) { |
153 | panic("%s: Missing memory setting 0x%08x, size=0x%08x\n", | |
154 | __func__, (u32) memory_start, (u32) memory_size); | |
a95d0e16 MS |
155 | } |
156 | ||
157 | /* reservation of region where is the kernel */ | |
158 | kernel_align_start = PAGE_DOWN((u32)_text); | |
159 | /* ALIGN can be remove because _end in vmlinux.lds.S is align */ | |
160 | kernel_align_size = PAGE_UP((u32)klimit) - kernel_align_start; | |
83a92529 | 161 | printk(KERN_INFO "%s: kernel addr:0x%08x-0x%08x size=0x%08x\n", |
a95d0e16 MS |
162 | __func__, kernel_align_start, kernel_align_start |
163 | + kernel_align_size, kernel_align_size); | |
83a92529 | 164 | memblock_reserve(kernel_align_start, kernel_align_size); |
4dc60832 | 165 | #endif |
a95d0e16 MS |
166 | /* |
167 | * Kernel: | |
168 | * start: base phys address of kernel - page align | |
169 | * end: base phys address of kernel - page align | |
170 | * | |
171 | * min_low_pfn - the first page (mm/bootmem.c - node_boot_start) | |
172 | * max_low_pfn | |
173 | * max_mapnr - the first unused page (mm/bootmem.c - node_low_pfn) | |
174 | * num_physpages - number of all pages | |
175 | */ | |
176 | ||
177 | /* memory start is from the kernel end (aligned) to higher addr */ | |
178 | min_low_pfn = memory_start >> PAGE_SHIFT; /* minimum for allocation */ | |
179 | /* RAM is assumed contiguous */ | |
180 | num_physpages = max_mapnr = memory_size >> PAGE_SHIFT; | |
83a92529 MS |
181 | max_low_pfn = ((u64)memory_start + (u64)lowmem_size) >> PAGE_SHIFT; |
182 | max_pfn = ((u64)memory_start + (u64)memory_size) >> PAGE_SHIFT; | |
a95d0e16 MS |
183 | |
184 | printk(KERN_INFO "%s: max_mapnr: %#lx\n", __func__, max_mapnr); | |
185 | printk(KERN_INFO "%s: min_low_pfn: %#lx\n", __func__, min_low_pfn); | |
186 | printk(KERN_INFO "%s: max_low_pfn: %#lx\n", __func__, max_low_pfn); | |
83a92529 | 187 | printk(KERN_INFO "%s: max_pfn: %#lx\n", __func__, max_pfn); |
a95d0e16 MS |
188 | |
189 | /* | |
190 | * Find an area to use for the bootmem bitmap. | |
191 | * We look for the first area which is at least | |
192 | * 128kB in length (128kB is enough for a bitmap | |
193 | * for 4GB of memory, using 4kB pages), plus 1 page | |
194 | * (in case the address isn't page-aligned). | |
195 | */ | |
e0581667 | 196 | map_size = init_bootmem_node(NODE_DATA(0), |
8f37b6c9 | 197 | PFN_UP(TOPHYS((u32)klimit)), min_low_pfn, max_low_pfn); |
95f72d1e | 198 | memblock_reserve(PFN_UP(TOPHYS((u32)klimit)) << PAGE_SHIFT, map_size); |
a95d0e16 | 199 | |
baab8a82 MS |
200 | /* Add active regions with valid PFNs */ |
201 | for_each_memblock(memory, reg) { | |
202 | unsigned long start_pfn, end_pfn; | |
203 | ||
204 | start_pfn = memblock_region_memory_base_pfn(reg); | |
205 | end_pfn = memblock_region_memory_end_pfn(reg); | |
206 | memblock_set_node(start_pfn << PAGE_SHIFT, | |
207 | (end_pfn - start_pfn) << PAGE_SHIFT, 0); | |
208 | } | |
209 | ||
a95d0e16 | 210 | /* free bootmem is whole main memory */ |
baab8a82 | 211 | free_bootmem_with_active_regions(0, max_low_pfn); |
a95d0e16 MS |
212 | |
213 | /* reserve allocate blocks */ | |
76bfcc81 | 214 | for_each_memblock(reserved, reg) { |
83a92529 MS |
215 | unsigned long top = reg->base + reg->size - 1; |
216 | ||
217 | pr_debug("reserved - 0x%08x-0x%08x, %lx, %lx\n", | |
218 | (u32) reg->base, (u32) reg->size, top, | |
219 | memory_start + lowmem_size - 1); | |
220 | ||
221 | if (top <= (memory_start + lowmem_size - 1)) { | |
222 | reserve_bootmem(reg->base, reg->size, BOOTMEM_DEFAULT); | |
223 | } else if (reg->base < (memory_start + lowmem_size - 1)) { | |
224 | unsigned long trunc_size = memory_start + lowmem_size - | |
225 | reg->base; | |
226 | reserve_bootmem(reg->base, trunc_size, BOOTMEM_DEFAULT); | |
227 | } | |
a95d0e16 | 228 | } |
83a92529 | 229 | |
baab8a82 MS |
230 | /* XXX need to clip this if using highmem? */ |
231 | sparse_memory_present_with_active_regions(0); | |
232 | ||
4dc60832 MS |
233 | #ifdef CONFIG_MMU |
234 | init_bootmem_done = 1; | |
235 | #endif | |
a95d0e16 MS |
236 | paging_init(); |
237 | } | |
238 | ||
239 | void free_init_pages(char *what, unsigned long begin, unsigned long end) | |
240 | { | |
241 | unsigned long addr; | |
242 | ||
243 | for (addr = begin; addr < end; addr += PAGE_SIZE) { | |
244 | ClearPageReserved(virt_to_page(addr)); | |
245 | init_page_count(virt_to_page(addr)); | |
a95d0e16 MS |
246 | free_page(addr); |
247 | totalram_pages++; | |
248 | } | |
249 | printk(KERN_INFO "Freeing %s: %ldk freed\n", what, (end - begin) >> 10); | |
250 | } | |
251 | ||
252 | #ifdef CONFIG_BLK_DEV_INITRD | |
253 | void free_initrd_mem(unsigned long start, unsigned long end) | |
254 | { | |
255 | int pages = 0; | |
256 | for (; start < end; start += PAGE_SIZE) { | |
257 | ClearPageReserved(virt_to_page(start)); | |
258 | init_page_count(virt_to_page(start)); | |
259 | free_page(start); | |
260 | totalram_pages++; | |
261 | pages++; | |
262 | } | |
d6f61770 LS |
263 | printk(KERN_NOTICE "Freeing initrd memory: %dk freed\n", |
264 | (int)(pages * (PAGE_SIZE / 1024))); | |
a95d0e16 MS |
265 | } |
266 | #endif | |
267 | ||
268 | void free_initmem(void) | |
269 | { | |
270 | free_init_pages("unused kernel memory", | |
271 | (unsigned long)(&__init_begin), | |
272 | (unsigned long)(&__init_end)); | |
273 | } | |
274 | ||
a95d0e16 MS |
275 | void __init mem_init(void) |
276 | { | |
83299799 MS |
277 | pg_data_t *pgdat; |
278 | unsigned long reservedpages = 0, codesize, initsize, datasize, bsssize; | |
279 | ||
83a92529 MS |
280 | high_memory = (void *)__va(memory_start + lowmem_size - 1); |
281 | ||
a95d0e16 MS |
282 | /* this will put all memory onto the freelists */ |
283 | totalram_pages += free_all_bootmem(); | |
284 | ||
83299799 MS |
285 | for_each_online_pgdat(pgdat) { |
286 | unsigned long i; | |
287 | struct page *page; | |
288 | ||
289 | for (i = 0; i < pgdat->node_spanned_pages; i++) { | |
290 | if (!pfn_valid(pgdat->node_start_pfn + i)) | |
291 | continue; | |
292 | page = pgdat_page_nr(pgdat, i); | |
293 | if (PageReserved(page)) | |
294 | reservedpages++; | |
295 | } | |
296 | } | |
297 | ||
2f2f371f MS |
298 | #ifdef CONFIG_HIGHMEM |
299 | reservedpages -= highmem_setup(); | |
300 | #endif | |
301 | ||
83299799 MS |
302 | codesize = (unsigned long)&_sdata - (unsigned long)&_stext; |
303 | datasize = (unsigned long)&_edata - (unsigned long)&_sdata; | |
304 | initsize = (unsigned long)&__init_end - (unsigned long)&__init_begin; | |
305 | bsssize = (unsigned long)&__bss_stop - (unsigned long)&__bss_start; | |
306 | ||
307 | pr_info("Memory: %luk/%luk available (%luk kernel code, " | |
308 | "%luk reserved, %luk data, %luk bss, %luk init)\n", | |
309 | nr_free_pages() << (PAGE_SHIFT-10), | |
310 | num_physpages << (PAGE_SHIFT-10), | |
311 | codesize >> 10, | |
312 | reservedpages << (PAGE_SHIFT-10), | |
313 | datasize >> 10, | |
314 | bsssize >> 10, | |
315 | initsize >> 10); | |
316 | ||
317 | #ifdef CONFIG_MMU | |
318 | pr_info("Kernel virtual memory layout:\n"); | |
319 | pr_info(" * 0x%08lx..0x%08lx : fixmap\n", FIXADDR_START, FIXADDR_TOP); | |
2f2f371f MS |
320 | #ifdef CONFIG_HIGHMEM |
321 | pr_info(" * 0x%08lx..0x%08lx : highmem PTEs\n", | |
322 | PKMAP_BASE, PKMAP_ADDR(LAST_PKMAP)); | |
323 | #endif /* CONFIG_HIGHMEM */ | |
83299799 MS |
324 | pr_info(" * 0x%08lx..0x%08lx : early ioremap\n", |
325 | ioremap_bot, ioremap_base); | |
326 | pr_info(" * 0x%08lx..0x%08lx : vmalloc & ioremap\n", | |
327 | (unsigned long)VMALLOC_START, VMALLOC_END); | |
328 | #endif | |
4dc60832 | 329 | mem_init_done = 1; |
a95d0e16 MS |
330 | } |
331 | ||
4dc60832 | 332 | #ifndef CONFIG_MMU |
79bf3a13 MS |
333 | int page_is_ram(unsigned long pfn) |
334 | { | |
335 | return __range_ok(pfn, 0); | |
336 | } | |
4dc60832 MS |
337 | #else |
338 | int page_is_ram(unsigned long pfn) | |
339 | { | |
340 | return pfn < max_low_pfn; | |
341 | } | |
342 | ||
343 | /* | |
344 | * Check for command-line options that affect what MMU_init will do. | |
345 | */ | |
346 | static void mm_cmdline_setup(void) | |
347 | { | |
348 | unsigned long maxmem = 0; | |
349 | char *p = cmd_line; | |
350 | ||
351 | /* Look for mem= option on command line */ | |
352 | p = strstr(cmd_line, "mem="); | |
353 | if (p) { | |
354 | p += 4; | |
355 | maxmem = memparse(p, &p); | |
356 | if (maxmem && memory_size > maxmem) { | |
357 | memory_size = maxmem; | |
da5ab11c | 358 | memblock.memory.regions[0].size = memory_size; |
4dc60832 MS |
359 | } |
360 | } | |
361 | } | |
362 | ||
363 | /* | |
364 | * MMU_init_hw does the chip-specific initialization of the MMU hardware. | |
365 | */ | |
366 | static void __init mmu_init_hw(void) | |
367 | { | |
368 | /* | |
369 | * The Zone Protection Register (ZPR) defines how protection will | |
370 | * be applied to every page which is a member of a given zone. At | |
371 | * present, we utilize only two of the zones. | |
372 | * The zone index bits (of ZSEL) in the PTE are used for software | |
373 | * indicators, except the LSB. For user access, zone 1 is used, | |
374 | * for kernel access, zone 0 is used. We set all but zone 1 | |
375 | * to zero, allowing only kernel access as indicated in the PTE. | |
376 | * For zone 1, we set a 01 binary (a value of 10 will not work) | |
377 | * to allow user access as indicated in the PTE. This also allows | |
378 | * kernel access as indicated in the PTE. | |
379 | */ | |
380 | __asm__ __volatile__ ("ori r11, r0, 0x10000000;" \ | |
381 | "mts rzpr, r11;" | |
382 | : : : "r11"); | |
383 | } | |
384 | ||
385 | /* | |
386 | * MMU_init sets up the basic memory mappings for the kernel, | |
387 | * including both RAM and possibly some I/O regions, | |
388 | * and sets up the page tables and the MMU hardware ready to go. | |
389 | */ | |
390 | ||
391 | /* called from head.S */ | |
392 | asmlinkage void __init mmu_init(void) | |
393 | { | |
394 | unsigned int kstart, ksize; | |
395 | ||
95f72d1e | 396 | if (!memblock.reserved.cnt) { |
4dc60832 MS |
397 | printk(KERN_EMERG "Error memory count\n"); |
398 | machine_restart(NULL); | |
399 | } | |
400 | ||
da5ab11c | 401 | if ((u32) memblock.memory.regions[0].size < 0x1000000) { |
4dc60832 MS |
402 | printk(KERN_EMERG "Memory must be greater than 16MB\n"); |
403 | machine_restart(NULL); | |
404 | } | |
405 | /* Find main memory where the kernel is */ | |
da5ab11c | 406 | memory_start = (u32) memblock.memory.regions[0].base; |
83a92529 MS |
407 | lowmem_size = memory_size = (u32) memblock.memory.regions[0].size; |
408 | ||
409 | if (lowmem_size > CONFIG_LOWMEM_SIZE) { | |
410 | lowmem_size = CONFIG_LOWMEM_SIZE; | |
2f2f371f | 411 | #ifndef CONFIG_HIGHMEM |
83a92529 | 412 | memory_size = lowmem_size; |
2f2f371f | 413 | #endif |
83a92529 | 414 | } |
4dc60832 MS |
415 | |
416 | mm_cmdline_setup(); /* FIXME parse args from command line - not used */ | |
417 | ||
418 | /* | |
419 | * Map out the kernel text/data/bss from the available physical | |
420 | * memory. | |
421 | */ | |
422 | kstart = __pa(CONFIG_KERNEL_START); /* kernel start */ | |
423 | /* kernel size */ | |
424 | ksize = PAGE_ALIGN(((u32)_end - (u32)CONFIG_KERNEL_START)); | |
95f72d1e | 425 | memblock_reserve(kstart, ksize); |
4dc60832 MS |
426 | |
427 | #if defined(CONFIG_BLK_DEV_INITRD) | |
428 | /* Remove the init RAM disk from the available memory. */ | |
429 | /* if (initrd_start) { | |
430 | mem_pieces_remove(&phys_avail, __pa(initrd_start), | |
431 | initrd_end - initrd_start, 1); | |
432 | }*/ | |
433 | #endif /* CONFIG_BLK_DEV_INITRD */ | |
434 | ||
435 | /* Initialize the MMU hardware */ | |
436 | mmu_init_hw(); | |
437 | ||
438 | /* Map in all of RAM starting at CONFIG_KERNEL_START */ | |
439 | mapin_ram(); | |
440 | ||
41938761 | 441 | /* Extend vmalloc and ioremap area as big as possible */ |
2f2f371f MS |
442 | #ifdef CONFIG_HIGHMEM |
443 | ioremap_base = ioremap_bot = PKMAP_BASE; | |
444 | #else | |
41938761 | 445 | ioremap_base = ioremap_bot = FIXADDR_START; |
2f2f371f | 446 | #endif |
41938761 | 447 | |
4dc60832 MS |
448 | /* Initialize the context management stuff */ |
449 | mmu_context_init(); | |
83a92529 MS |
450 | |
451 | /* Shortly after that, the entire linear mapping will be available */ | |
452 | /* This will also cause that unflatten device tree will be allocated | |
453 | * inside 768MB limit */ | |
454 | memblock_set_current_limit(memory_start + lowmem_size - 1); | |
4dc60832 MS |
455 | } |
456 | ||
457 | /* This is only called until mem_init is done. */ | |
458 | void __init *early_get_page(void) | |
459 | { | |
460 | void *p; | |
461 | if (init_bootmem_done) { | |
462 | p = alloc_bootmem_pages(PAGE_SIZE); | |
463 | } else { | |
464 | /* | |
465 | * Mem start + 32MB -> here is limit | |
466 | * because of mem mapping from head.S | |
467 | */ | |
95f72d1e | 468 | p = __va(memblock_alloc_base(PAGE_SIZE, PAGE_SIZE, |
4dc60832 MS |
469 | memory_start + 0x2000000)); |
470 | } | |
471 | return p; | |
472 | } | |
a84642a3 | 473 | |
79bf3a13 MS |
474 | #endif /* CONFIG_MMU */ |
475 | ||
a84642a3 MS |
476 | void * __init_refok alloc_maybe_bootmem(size_t size, gfp_t mask) |
477 | { | |
478 | if (mem_init_done) | |
479 | return kmalloc(size, mask); | |
480 | else | |
481 | return alloc_bootmem(size); | |
482 | } | |
483 | ||
484 | void * __init_refok zalloc_maybe_bootmem(size_t size, gfp_t mask) | |
485 | { | |
486 | void *p; | |
487 | ||
488 | if (mem_init_done) | |
489 | p = kzalloc(size, mask); | |
490 | else { | |
491 | p = alloc_bootmem(size); | |
492 | if (p) | |
493 | memset(p, 0, size); | |
494 | } | |
495 | return p; | |
496 | } |