]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Initialize MMU support. | |
3 | * | |
4 | * Copyright (C) 1998-2003 Hewlett-Packard Co | |
5 | * David Mosberger-Tang <davidm@hpl.hp.com> | |
6 | */ | |
1da177e4 LT |
7 | #include <linux/kernel.h> |
8 | #include <linux/init.h> | |
9 | ||
10 | #include <linux/bootmem.h> | |
11 | #include <linux/efi.h> | |
12 | #include <linux/elf.h> | |
98e4ae8a | 13 | #include <linux/memblock.h> |
1da177e4 LT |
14 | #include <linux/mm.h> |
15 | #include <linux/mmzone.h> | |
16 | #include <linux/module.h> | |
17 | #include <linux/personality.h> | |
18 | #include <linux/reboot.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/swap.h> | |
21 | #include <linux/proc_fs.h> | |
22 | #include <linux/bitops.h> | |
139b8304 | 23 | #include <linux/kexec.h> |
1da177e4 | 24 | |
1da177e4 | 25 | #include <asm/dma.h> |
1da177e4 LT |
26 | #include <asm/io.h> |
27 | #include <asm/machvec.h> | |
28 | #include <asm/numa.h> | |
29 | #include <asm/patch.h> | |
30 | #include <asm/pgalloc.h> | |
31 | #include <asm/sal.h> | |
32 | #include <asm/sections.h> | |
1da177e4 LT |
33 | #include <asm/tlb.h> |
34 | #include <asm/uaccess.h> | |
35 | #include <asm/unistd.h> | |
36 | #include <asm/mca.h> | |
37 | ||
1da177e4 LT |
38 | extern void ia64_tlb_init (void); |
39 | ||
40 | unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL; | |
41 | ||
42 | #ifdef CONFIG_VIRTUAL_MEM_MAP | |
126b3fcd TH |
43 | unsigned long VMALLOC_END = VMALLOC_END_INIT; |
44 | EXPORT_SYMBOL(VMALLOC_END); | |
1da177e4 LT |
45 | struct page *vmem_map; |
46 | EXPORT_SYMBOL(vmem_map); | |
47 | #endif | |
48 | ||
fde740e4 | 49 | struct page *zero_page_memmap_ptr; /* map entry for zero page */ |
1da177e4 LT |
50 | EXPORT_SYMBOL(zero_page_memmap_ptr); |
51 | ||
1da177e4 | 52 | void |
954ffcb3 | 53 | __ia64_sync_icache_dcache (pte_t pte) |
1da177e4 LT |
54 | { |
55 | unsigned long addr; | |
56 | struct page *page; | |
57 | ||
1da177e4 LT |
58 | page = pte_page(pte); |
59 | addr = (unsigned long) page_address(page); | |
60 | ||
61 | if (test_bit(PG_arch_1, &page->flags)) | |
62 | return; /* i-cache is already coherent with d-cache */ | |
63 | ||
273988fa | 64 | flush_icache_range(addr, addr + (PAGE_SIZE << compound_order(page))); |
1da177e4 LT |
65 | set_bit(PG_arch_1, &page->flags); /* mark page as clean */ |
66 | } | |
67 | ||
cde14bbf JB |
68 | /* |
69 | * Since DMA is i-cache coherent, any (complete) pages that were written via | |
70 | * DMA can be marked as "clean" so that lazy_mmu_prot_update() doesn't have to | |
71 | * flush them when they get mapped into an executable vm-area. | |
72 | */ | |
73 | void | |
74 | dma_mark_clean(void *addr, size_t size) | |
75 | { | |
76 | unsigned long pg_addr, end; | |
77 | ||
78 | pg_addr = PAGE_ALIGN((unsigned long) addr); | |
79 | end = (unsigned long) addr + size; | |
80 | while (pg_addr + PAGE_SIZE <= end) { | |
81 | struct page *page = virt_to_page(pg_addr); | |
82 | set_bit(PG_arch_1, &page->flags); | |
83 | pg_addr += PAGE_SIZE; | |
84 | } | |
85 | } | |
86 | ||
1da177e4 LT |
87 | inline void |
88 | ia64_set_rbs_bot (void) | |
89 | { | |
02b763b8 | 90 | unsigned long stack_size = rlimit_max(RLIMIT_STACK) & -16; |
1da177e4 LT |
91 | |
92 | if (stack_size > MAX_USER_STACK_SIZE) | |
93 | stack_size = MAX_USER_STACK_SIZE; | |
83d2cd3d | 94 | current->thread.rbs_bot = PAGE_ALIGN(current->mm->start_stack - stack_size); |
1da177e4 LT |
95 | } |
96 | ||
97 | /* | |
98 | * This performs some platform-dependent address space initialization. | |
99 | * On IA-64, we want to setup the VM area for the register backing | |
100 | * store (which grows upwards) and install the gateway page which is | |
101 | * used for signal trampolines, etc. | |
102 | */ | |
103 | void | |
104 | ia64_init_addr_space (void) | |
105 | { | |
106 | struct vm_area_struct *vma; | |
107 | ||
108 | ia64_set_rbs_bot(); | |
109 | ||
110 | /* | |
111 | * If we're out of memory and kmem_cache_alloc() returns NULL, we simply ignore | |
112 | * the problem. When the process attempts to write to the register backing store | |
113 | * for the first time, it will get a SEGFAULT in this case. | |
114 | */ | |
c3762229 | 115 | vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 | 116 | if (vma) { |
5beb4930 | 117 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
1da177e4 LT |
118 | vma->vm_mm = current->mm; |
119 | vma->vm_start = current->thread.rbs_bot & PAGE_MASK; | |
120 | vma->vm_end = vma->vm_start + PAGE_SIZE; | |
46dea3d0 | 121 | vma->vm_flags = VM_DATA_DEFAULT_FLAGS|VM_GROWSUP|VM_ACCOUNT; |
3ed75eb8 | 122 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
1da177e4 LT |
123 | down_write(¤t->mm->mmap_sem); |
124 | if (insert_vm_struct(current->mm, vma)) { | |
125 | up_write(¤t->mm->mmap_sem); | |
126 | kmem_cache_free(vm_area_cachep, vma); | |
127 | return; | |
128 | } | |
129 | up_write(¤t->mm->mmap_sem); | |
130 | } | |
131 | ||
132 | /* map NaT-page at address zero to speed up speculative dereferencing of NULL: */ | |
133 | if (!(current->personality & MMAP_PAGE_ZERO)) { | |
c3762229 | 134 | vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 | 135 | if (vma) { |
5beb4930 | 136 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
1da177e4 LT |
137 | vma->vm_mm = current->mm; |
138 | vma->vm_end = PAGE_SIZE; | |
139 | vma->vm_page_prot = __pgprot(pgprot_val(PAGE_READONLY) | _PAGE_MA_NAT); | |
314e51b9 KK |
140 | vma->vm_flags = VM_READ | VM_MAYREAD | VM_IO | |
141 | VM_DONTEXPAND | VM_DONTDUMP; | |
1da177e4 LT |
142 | down_write(¤t->mm->mmap_sem); |
143 | if (insert_vm_struct(current->mm, vma)) { | |
144 | up_write(¤t->mm->mmap_sem); | |
145 | kmem_cache_free(vm_area_cachep, vma); | |
146 | return; | |
147 | } | |
148 | up_write(¤t->mm->mmap_sem); | |
149 | } | |
150 | } | |
151 | } | |
152 | ||
153 | void | |
154 | free_initmem (void) | |
155 | { | |
11199692 | 156 | free_reserved_area(ia64_imva(__init_begin), ia64_imva(__init_end), |
dbe67df4 | 157 | -1, "unused kernel"); |
1da177e4 LT |
158 | } |
159 | ||
dae28066 | 160 | void __init |
1da177e4 LT |
161 | free_initrd_mem (unsigned long start, unsigned long end) |
162 | { | |
1da177e4 LT |
163 | /* |
164 | * EFI uses 4KB pages while the kernel can use 4KB or bigger. | |
165 | * Thus EFI and the kernel may have different page sizes. It is | |
166 | * therefore possible to have the initrd share the same page as | |
167 | * the end of the kernel (given current setup). | |
168 | * | |
169 | * To avoid freeing/using the wrong page (kernel sized) we: | |
170 | * - align up the beginning of initrd | |
171 | * - align down the end of initrd | |
172 | * | |
173 | * | | | |
174 | * |=============| a000 | |
175 | * | | | |
176 | * | | | |
177 | * | | 9000 | |
178 | * |/////////////| | |
179 | * |/////////////| | |
180 | * |=============| 8000 | |
181 | * |///INITRD////| | |
182 | * |/////////////| | |
183 | * |/////////////| 7000 | |
184 | * | | | |
185 | * |KKKKKKKKKKKKK| | |
186 | * |=============| 6000 | |
187 | * |KKKKKKKKKKKKK| | |
188 | * |KKKKKKKKKKKKK| | |
189 | * K=kernel using 8KB pages | |
190 | * | |
191 | * In this example, we must free page 8000 ONLY. So we must align up | |
192 | * initrd_start and keep initrd_end as is. | |
193 | */ | |
194 | start = PAGE_ALIGN(start); | |
195 | end = end & PAGE_MASK; | |
196 | ||
197 | if (start < end) | |
198 | printk(KERN_INFO "Freeing initrd memory: %ldkB freed\n", (end - start) >> 10); | |
199 | ||
200 | for (; start < end; start += PAGE_SIZE) { | |
201 | if (!virt_addr_valid(start)) | |
202 | continue; | |
66f62594 | 203 | free_reserved_page(virt_to_page(start)); |
1da177e4 LT |
204 | } |
205 | } | |
206 | ||
207 | /* | |
208 | * This installs a clean page in the kernel's page table. | |
209 | */ | |
dae28066 | 210 | static struct page * __init |
1da177e4 LT |
211 | put_kernel_page (struct page *page, unsigned long address, pgprot_t pgprot) |
212 | { | |
213 | pgd_t *pgd; | |
214 | pud_t *pud; | |
215 | pmd_t *pmd; | |
216 | pte_t *pte; | |
217 | ||
218 | if (!PageReserved(page)) | |
219 | printk(KERN_ERR "put_kernel_page: page at 0x%p not in reserved memory\n", | |
220 | page_address(page)); | |
221 | ||
222 | pgd = pgd_offset_k(address); /* note: this is NOT pgd_offset()! */ | |
223 | ||
1da177e4 LT |
224 | { |
225 | pud = pud_alloc(&init_mm, pgd, address); | |
226 | if (!pud) | |
227 | goto out; | |
1da177e4 LT |
228 | pmd = pmd_alloc(&init_mm, pud, address); |
229 | if (!pmd) | |
230 | goto out; | |
872fec16 | 231 | pte = pte_alloc_kernel(pmd, address); |
1da177e4 LT |
232 | if (!pte) |
233 | goto out; | |
872fec16 | 234 | if (!pte_none(*pte)) |
1da177e4 | 235 | goto out; |
1da177e4 | 236 | set_pte(pte, mk_pte(page, pgprot)); |
1da177e4 | 237 | } |
872fec16 | 238 | out: |
1da177e4 LT |
239 | /* no need for flush_tlb */ |
240 | return page; | |
241 | } | |
242 | ||
914a4ea4 | 243 | static void __init |
1da177e4 LT |
244 | setup_gate (void) |
245 | { | |
246 | struct page *page; | |
247 | ||
248 | /* | |
ad597bd5 DMT |
249 | * Map the gate page twice: once read-only to export the ELF |
250 | * headers etc. and once execute-only page to enable | |
251 | * privilege-promotion via "epc": | |
1da177e4 | 252 | */ |
e55645ec | 253 | page = virt_to_page(ia64_imva(__start_gate_section)); |
1da177e4 LT |
254 | put_kernel_page(page, GATE_ADDR, PAGE_READONLY); |
255 | #ifdef HAVE_BUGGY_SEGREL | |
e55645ec | 256 | page = virt_to_page(ia64_imva(__start_gate_section + PAGE_SIZE)); |
1da177e4 LT |
257 | put_kernel_page(page, GATE_ADDR + PAGE_SIZE, PAGE_GATE); |
258 | #else | |
259 | put_kernel_page(page, GATE_ADDR + PERCPU_PAGE_SIZE, PAGE_GATE); | |
ad597bd5 DMT |
260 | /* Fill in the holes (if any) with read-only zero pages: */ |
261 | { | |
262 | unsigned long addr; | |
263 | ||
264 | for (addr = GATE_ADDR + PAGE_SIZE; | |
265 | addr < GATE_ADDR + PERCPU_PAGE_SIZE; | |
266 | addr += PAGE_SIZE) | |
267 | { | |
268 | put_kernel_page(ZERO_PAGE(0), addr, | |
269 | PAGE_READONLY); | |
270 | put_kernel_page(ZERO_PAGE(0), addr + PERCPU_PAGE_SIZE, | |
271 | PAGE_READONLY); | |
272 | } | |
273 | } | |
1da177e4 LT |
274 | #endif |
275 | ia64_patch_gate(); | |
276 | } | |
277 | ||
a6c19dfe AL |
278 | static struct vm_area_struct gate_vma; |
279 | ||
280 | static int __init gate_vma_init(void) | |
281 | { | |
282 | gate_vma.vm_mm = NULL; | |
283 | gate_vma.vm_start = FIXADDR_USER_START; | |
284 | gate_vma.vm_end = FIXADDR_USER_END; | |
285 | gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC; | |
286 | gate_vma.vm_page_prot = __P101; | |
287 | ||
288 | return 0; | |
289 | } | |
290 | __initcall(gate_vma_init); | |
291 | ||
292 | struct vm_area_struct *get_gate_vma(struct mm_struct *mm) | |
293 | { | |
294 | return &gate_vma; | |
295 | } | |
296 | ||
297 | int in_gate_area_no_mm(unsigned long addr) | |
298 | { | |
299 | if ((addr >= FIXADDR_USER_START) && (addr < FIXADDR_USER_END)) | |
300 | return 1; | |
301 | return 0; | |
302 | } | |
303 | ||
304 | int in_gate_area(struct mm_struct *mm, unsigned long addr) | |
305 | { | |
306 | return in_gate_area_no_mm(addr); | |
307 | } | |
308 | ||
5b5e76e9 | 309 | void ia64_mmu_init(void *my_cpu_data) |
1da177e4 | 310 | { |
00b65985 | 311 | unsigned long pta, impl_va_bits; |
5b5e76e9 | 312 | extern void tlb_init(void); |
1da177e4 LT |
313 | |
314 | #ifdef CONFIG_DISABLE_VHPT | |
315 | # define VHPT_ENABLE_BIT 0 | |
316 | #else | |
317 | # define VHPT_ENABLE_BIT 1 | |
318 | #endif | |
319 | ||
1da177e4 LT |
320 | /* |
321 | * Check if the virtually mapped linear page table (VMLPT) overlaps with a mapped | |
322 | * address space. The IA-64 architecture guarantees that at least 50 bits of | |
323 | * virtual address space are implemented but if we pick a large enough page size | |
324 | * (e.g., 64KB), the mapped address space is big enough that it will overlap with | |
325 | * VMLPT. I assume that once we run on machines big enough to warrant 64KB pages, | |
326 | * IMPL_VA_MSB will be significantly bigger, so this is unlikely to become a | |
327 | * problem in practice. Alternatively, we could truncate the top of the mapped | |
328 | * address space to not permit mappings that would overlap with the VMLPT. | |
329 | * --davidm 00/12/06 | |
330 | */ | |
331 | # define pte_bits 3 | |
332 | # define mapped_space_bits (3*(PAGE_SHIFT - pte_bits) + PAGE_SHIFT) | |
333 | /* | |
334 | * The virtual page table has to cover the entire implemented address space within | |
335 | * a region even though not all of this space may be mappable. The reason for | |
336 | * this is that the Access bit and Dirty bit fault handlers perform | |
337 | * non-speculative accesses to the virtual page table, so the address range of the | |
338 | * virtual page table itself needs to be covered by virtual page table. | |
339 | */ | |
340 | # define vmlpt_bits (impl_va_bits - PAGE_SHIFT + pte_bits) | |
341 | # define POW2(n) (1ULL << (n)) | |
342 | ||
343 | impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61))); | |
344 | ||
345 | if (impl_va_bits < 51 || impl_va_bits > 61) | |
346 | panic("CPU has bogus IMPL_VA_MSB value of %lu!\n", impl_va_bits - 1); | |
6cf07a8c PC |
347 | /* |
348 | * mapped_space_bits - PAGE_SHIFT is the total number of ptes we need, | |
349 | * which must fit into "vmlpt_bits - pte_bits" slots. Second half of | |
350 | * the test makes sure that our mapped space doesn't overlap the | |
351 | * unimplemented hole in the middle of the region. | |
352 | */ | |
353 | if ((mapped_space_bits - PAGE_SHIFT > vmlpt_bits - pte_bits) || | |
354 | (mapped_space_bits > impl_va_bits - 1)) | |
355 | panic("Cannot build a big enough virtual-linear page table" | |
356 | " to cover mapped address space.\n" | |
357 | " Try using a smaller page size.\n"); | |
358 | ||
1da177e4 LT |
359 | |
360 | /* place the VMLPT at the end of each page-table mapped region: */ | |
361 | pta = POW2(61) - POW2(vmlpt_bits); | |
362 | ||
1da177e4 LT |
363 | /* |
364 | * Set the (virtually mapped linear) page table address. Bit | |
365 | * 8 selects between the short and long format, bits 2-7 the | |
366 | * size of the table, and bit 0 whether the VHPT walker is | |
367 | * enabled. | |
368 | */ | |
369 | ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | VHPT_ENABLE_BIT); | |
370 | ||
371 | ia64_tlb_init(); | |
372 | ||
373 | #ifdef CONFIG_HUGETLB_PAGE | |
374 | ia64_set_rr(HPAGE_REGION_BASE, HPAGE_SHIFT << 2); | |
375 | ia64_srlz_d(); | |
376 | #endif | |
377 | } | |
378 | ||
379 | #ifdef CONFIG_VIRTUAL_MEM_MAP | |
e44e41d0 BP |
380 | int vmemmap_find_next_valid_pfn(int node, int i) |
381 | { | |
382 | unsigned long end_address, hole_next_pfn; | |
383 | unsigned long stop_address; | |
384 | pg_data_t *pgdat = NODE_DATA(node); | |
385 | ||
386 | end_address = (unsigned long) &vmem_map[pgdat->node_start_pfn + i]; | |
387 | end_address = PAGE_ALIGN(end_address); | |
6408068e | 388 | stop_address = (unsigned long) &vmem_map[pgdat_end_pfn(pgdat)]; |
e44e41d0 BP |
389 | |
390 | do { | |
391 | pgd_t *pgd; | |
392 | pud_t *pud; | |
393 | pmd_t *pmd; | |
394 | pte_t *pte; | |
395 | ||
396 | pgd = pgd_offset_k(end_address); | |
397 | if (pgd_none(*pgd)) { | |
398 | end_address += PGDIR_SIZE; | |
399 | continue; | |
400 | } | |
401 | ||
402 | pud = pud_offset(pgd, end_address); | |
403 | if (pud_none(*pud)) { | |
404 | end_address += PUD_SIZE; | |
405 | continue; | |
406 | } | |
407 | ||
408 | pmd = pmd_offset(pud, end_address); | |
409 | if (pmd_none(*pmd)) { | |
410 | end_address += PMD_SIZE; | |
411 | continue; | |
412 | } | |
413 | ||
414 | pte = pte_offset_kernel(pmd, end_address); | |
415 | retry_pte: | |
416 | if (pte_none(*pte)) { | |
417 | end_address += PAGE_SIZE; | |
418 | pte++; | |
419 | if ((end_address < stop_address) && | |
420 | (end_address != ALIGN(end_address, 1UL << PMD_SHIFT))) | |
421 | goto retry_pte; | |
422 | continue; | |
423 | } | |
424 | /* Found next valid vmem_map page */ | |
425 | break; | |
426 | } while (end_address < stop_address); | |
427 | ||
428 | end_address = min(end_address, stop_address); | |
429 | end_address = end_address - (unsigned long) vmem_map + sizeof(struct page) - 1; | |
430 | hole_next_pfn = end_address / sizeof(struct page); | |
431 | return hole_next_pfn - pgdat->node_start_pfn; | |
432 | } | |
1da177e4 | 433 | |
e088a4ad | 434 | int __init create_mem_map_page_table(u64 start, u64 end, void *arg) |
1da177e4 LT |
435 | { |
436 | unsigned long address, start_page, end_page; | |
437 | struct page *map_start, *map_end; | |
438 | int node; | |
439 | pgd_t *pgd; | |
440 | pud_t *pud; | |
441 | pmd_t *pmd; | |
442 | pte_t *pte; | |
443 | ||
444 | map_start = vmem_map + (__pa(start) >> PAGE_SHIFT); | |
445 | map_end = vmem_map + (__pa(end) >> PAGE_SHIFT); | |
446 | ||
447 | start_page = (unsigned long) map_start & PAGE_MASK; | |
448 | end_page = PAGE_ALIGN((unsigned long) map_end); | |
449 | node = paddr_to_nid(__pa(start)); | |
450 | ||
451 | for (address = start_page; address < end_page; address += PAGE_SIZE) { | |
452 | pgd = pgd_offset_k(address); | |
453 | if (pgd_none(*pgd)) | |
454 | pgd_populate(&init_mm, pgd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)); | |
455 | pud = pud_offset(pgd, address); | |
456 | ||
457 | if (pud_none(*pud)) | |
458 | pud_populate(&init_mm, pud, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)); | |
459 | pmd = pmd_offset(pud, address); | |
460 | ||
461 | if (pmd_none(*pmd)) | |
462 | pmd_populate_kernel(&init_mm, pmd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)); | |
463 | pte = pte_offset_kernel(pmd, address); | |
464 | ||
465 | if (pte_none(*pte)) | |
466 | set_pte(pte, pfn_pte(__pa(alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)) >> PAGE_SHIFT, | |
467 | PAGE_KERNEL)); | |
468 | } | |
469 | return 0; | |
470 | } | |
471 | ||
472 | struct memmap_init_callback_data { | |
473 | struct page *start; | |
474 | struct page *end; | |
475 | int nid; | |
476 | unsigned long zone; | |
477 | }; | |
478 | ||
18b8befd | 479 | static int __meminit |
e088a4ad | 480 | virtual_memmap_init(u64 start, u64 end, void *arg) |
1da177e4 LT |
481 | { |
482 | struct memmap_init_callback_data *args; | |
483 | struct page *map_start, *map_end; | |
484 | ||
485 | args = (struct memmap_init_callback_data *) arg; | |
486 | map_start = vmem_map + (__pa(start) >> PAGE_SHIFT); | |
487 | map_end = vmem_map + (__pa(end) >> PAGE_SHIFT); | |
488 | ||
489 | if (map_start < args->start) | |
490 | map_start = args->start; | |
491 | if (map_end > args->end) | |
492 | map_end = args->end; | |
493 | ||
494 | /* | |
495 | * We have to initialize "out of bounds" struct page elements that fit completely | |
496 | * on the same pages that were allocated for the "in bounds" elements because they | |
497 | * may be referenced later (and found to be "reserved"). | |
498 | */ | |
499 | map_start -= ((unsigned long) map_start & (PAGE_SIZE - 1)) / sizeof(struct page); | |
500 | map_end += ((PAGE_ALIGN((unsigned long) map_end) - (unsigned long) map_end) | |
501 | / sizeof(struct page)); | |
502 | ||
503 | if (map_start < map_end) | |
504 | memmap_init_zone((unsigned long)(map_end - map_start), | |
a2f3aa02 DH |
505 | args->nid, args->zone, page_to_pfn(map_start), |
506 | MEMMAP_EARLY); | |
1da177e4 LT |
507 | return 0; |
508 | } | |
509 | ||
18b8befd | 510 | void __meminit |
1da177e4 LT |
511 | memmap_init (unsigned long size, int nid, unsigned long zone, |
512 | unsigned long start_pfn) | |
513 | { | |
514 | if (!vmem_map) | |
a2f3aa02 | 515 | memmap_init_zone(size, nid, zone, start_pfn, MEMMAP_EARLY); |
1da177e4 LT |
516 | else { |
517 | struct page *start; | |
518 | struct memmap_init_callback_data args; | |
519 | ||
520 | start = pfn_to_page(start_pfn); | |
521 | args.start = start; | |
522 | args.end = start + size; | |
523 | args.nid = nid; | |
524 | args.zone = zone; | |
525 | ||
526 | efi_memmap_walk(virtual_memmap_init, &args); | |
527 | } | |
528 | } | |
529 | ||
530 | int | |
531 | ia64_pfn_valid (unsigned long pfn) | |
532 | { | |
533 | char byte; | |
534 | struct page *pg = pfn_to_page(pfn); | |
535 | ||
536 | return (__get_user(byte, (char __user *) pg) == 0) | |
537 | && ((((u64)pg & PAGE_MASK) == (((u64)(pg + 1) - 1) & PAGE_MASK)) | |
538 | || (__get_user(byte, (char __user *) (pg + 1) - 1) == 0)); | |
539 | } | |
540 | EXPORT_SYMBOL(ia64_pfn_valid); | |
541 | ||
e088a4ad | 542 | int __init find_largest_hole(u64 start, u64 end, void *arg) |
1da177e4 LT |
543 | { |
544 | u64 *max_gap = arg; | |
545 | ||
546 | static u64 last_end = PAGE_OFFSET; | |
547 | ||
548 | /* NOTE: this algorithm assumes efi memmap table is ordered */ | |
549 | ||
550 | if (*max_gap < (start - last_end)) | |
551 | *max_gap = start - last_end; | |
552 | last_end = end; | |
553 | return 0; | |
554 | } | |
05e0caad | 555 | |
139b8304 BP |
556 | #endif /* CONFIG_VIRTUAL_MEM_MAP */ |
557 | ||
e088a4ad | 558 | int __init register_active_ranges(u64 start, u64 len, int nid) |
05e0caad | 559 | { |
98075d24 | 560 | u64 end = start + len; |
139b8304 | 561 | |
139b8304 BP |
562 | #ifdef CONFIG_KEXEC |
563 | if (start > crashk_res.start && start < crashk_res.end) | |
564 | start = crashk_res.end; | |
565 | if (end > crashk_res.start && end < crashk_res.end) | |
566 | end = crashk_res.start; | |
567 | #endif | |
568 | ||
569 | if (start < end) | |
98e4ae8a | 570 | memblock_add_node(__pa(start), end - start, nid); |
05e0caad MG |
571 | return 0; |
572 | } | |
1da177e4 | 573 | |
a3f5c338 | 574 | int |
e088a4ad | 575 | find_max_min_low_pfn (u64 start, u64 end, void *arg) |
a3f5c338 ZN |
576 | { |
577 | unsigned long pfn_start, pfn_end; | |
578 | #ifdef CONFIG_FLATMEM | |
579 | pfn_start = (PAGE_ALIGN(__pa(start))) >> PAGE_SHIFT; | |
580 | pfn_end = (PAGE_ALIGN(__pa(end - 1))) >> PAGE_SHIFT; | |
581 | #else | |
582 | pfn_start = GRANULEROUNDDOWN(__pa(start)) >> PAGE_SHIFT; | |
583 | pfn_end = GRANULEROUNDUP(__pa(end - 1)) >> PAGE_SHIFT; | |
584 | #endif | |
585 | min_low_pfn = min(min_low_pfn, pfn_start); | |
586 | max_low_pfn = max(max_low_pfn, pfn_end); | |
587 | return 0; | |
588 | } | |
589 | ||
1da177e4 LT |
590 | /* |
591 | * Boot command-line option "nolwsys" can be used to disable the use of any light-weight | |
592 | * system call handler. When this option is in effect, all fsyscalls will end up bubbling | |
593 | * down into the kernel and calling the normal (heavy-weight) syscall handler. This is | |
594 | * useful for performance testing, but conceivably could also come in handy for debugging | |
595 | * purposes. | |
596 | */ | |
597 | ||
03906ea0 | 598 | static int nolwsys __initdata; |
1da177e4 LT |
599 | |
600 | static int __init | |
601 | nolwsys_setup (char *s) | |
602 | { | |
603 | nolwsys = 1; | |
604 | return 1; | |
605 | } | |
606 | ||
607 | __setup("nolwsys", nolwsys_setup); | |
608 | ||
dae28066 | 609 | void __init |
1da177e4 LT |
610 | mem_init (void) |
611 | { | |
1da177e4 | 612 | int i; |
1da177e4 | 613 | |
fde740e4 RH |
614 | BUG_ON(PTRS_PER_PGD * sizeof(pgd_t) != PAGE_SIZE); |
615 | BUG_ON(PTRS_PER_PMD * sizeof(pmd_t) != PAGE_SIZE); | |
616 | BUG_ON(PTRS_PER_PTE * sizeof(pte_t) != PAGE_SIZE); | |
617 | ||
1da177e4 LT |
618 | #ifdef CONFIG_PCI |
619 | /* | |
620 | * This needs to be called _after_ the command line has been parsed but _before_ | |
621 | * any drivers that may need the PCI DMA interface are initialized or bootmem has | |
622 | * been freed. | |
623 | */ | |
624 | platform_dma_init(); | |
625 | #endif | |
626 | ||
2d4b1fa2 | 627 | #ifdef CONFIG_FLATMEM |
80a03e29 | 628 | BUG_ON(!mem_map); |
1da177e4 LT |
629 | #endif |
630 | ||
b57b63a2 | 631 | set_max_mapnr(max_low_pfn); |
1da177e4 | 632 | high_memory = __va(max_low_pfn * PAGE_SIZE); |
b57b63a2 | 633 | free_all_bootmem(); |
de4bcddc | 634 | mem_init_print_info(NULL); |
1da177e4 LT |
635 | |
636 | /* | |
637 | * For fsyscall entrpoints with no light-weight handler, use the ordinary | |
638 | * (heavy-weight) handler, but mark it by setting bit 0, so the fsyscall entry | |
639 | * code can tell them apart. | |
640 | */ | |
641 | for (i = 0; i < NR_syscalls; ++i) { | |
e55645ec | 642 | extern unsigned long fsyscall_table[NR_syscalls]; |
1da177e4 LT |
643 | extern unsigned long sys_call_table[NR_syscalls]; |
644 | ||
645 | if (!fsyscall_table[i] || nolwsys) | |
646 | fsyscall_table[i] = sys_call_table[i] | 1; | |
647 | } | |
648 | setup_gate(); | |
1da177e4 | 649 | } |
1681b8e1 YG |
650 | |
651 | #ifdef CONFIG_MEMORY_HOTPLUG | |
bc02af93 | 652 | int arch_add_memory(int nid, u64 start, u64 size) |
1681b8e1 YG |
653 | { |
654 | pg_data_t *pgdat; | |
655 | struct zone *zone; | |
656 | unsigned long start_pfn = start >> PAGE_SHIFT; | |
657 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
658 | int ret; | |
659 | ||
bc02af93 | 660 | pgdat = NODE_DATA(nid); |
1681b8e1 | 661 | |
ed562ae6 WN |
662 | zone = pgdat->node_zones + |
663 | zone_for_memory(nid, start, size, ZONE_NORMAL); | |
c04fc586 | 664 | ret = __add_pages(nid, zone, start_pfn, nr_pages); |
1681b8e1 YG |
665 | |
666 | if (ret) | |
667 | printk("%s: Problem encountered in __add_pages() as ret=%d\n", | |
d4ed8084 | 668 | __func__, ret); |
1681b8e1 YG |
669 | |
670 | return ret; | |
671 | } | |
24d335ca WC |
672 | |
673 | #ifdef CONFIG_MEMORY_HOTREMOVE | |
674 | int arch_remove_memory(u64 start, u64 size) | |
675 | { | |
676 | unsigned long start_pfn = start >> PAGE_SHIFT; | |
677 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
678 | struct zone *zone; | |
679 | int ret; | |
680 | ||
681 | zone = page_zone(pfn_to_page(start_pfn)); | |
682 | ret = __remove_pages(zone, start_pfn, nr_pages); | |
683 | if (ret) | |
684 | pr_warn("%s: Problem encountered in __remove_pages() as" | |
685 | " ret=%d\n", __func__, ret); | |
686 | ||
687 | return ret; | |
688 | } | |
689 | #endif | |
1681b8e1 | 690 | #endif |
839052d2 | 691 | |
aec6a888 MG |
692 | /** |
693 | * show_mem - give short summary of memory stats | |
694 | * | |
695 | * Shows a simple page count of reserved and used pages in the system. | |
696 | * For discontig machines, it does this on a per-pgdat basis. | |
697 | */ | |
698 | void show_mem(unsigned int filter) | |
699 | { | |
700 | int total_reserved = 0; | |
701 | unsigned long total_present = 0; | |
702 | pg_data_t *pgdat; | |
703 | ||
704 | printk(KERN_INFO "Mem-info:\n"); | |
705 | show_free_areas(filter); | |
706 | printk(KERN_INFO "Node memory in pages:\n"); | |
707 | for_each_online_pgdat(pgdat) { | |
708 | unsigned long present; | |
709 | unsigned long flags; | |
710 | int reserved = 0; | |
711 | int nid = pgdat->node_id; | |
712 | int zoneid; | |
713 | ||
714 | if (skip_free_areas_node(filter, nid)) | |
715 | continue; | |
716 | pgdat_resize_lock(pgdat, &flags); | |
717 | ||
718 | for (zoneid = 0; zoneid < MAX_NR_ZONES; zoneid++) { | |
719 | struct zone *zone = &pgdat->node_zones[zoneid]; | |
720 | if (!populated_zone(zone)) | |
721 | continue; | |
722 | ||
723 | reserved += zone->present_pages - zone->managed_pages; | |
724 | } | |
725 | present = pgdat->node_present_pages; | |
726 | ||
727 | pgdat_resize_unlock(pgdat, &flags); | |
728 | total_present += present; | |
729 | total_reserved += reserved; | |
730 | printk(KERN_INFO "Node %4d: RAM: %11ld, rsvd: %8d, ", | |
731 | nid, present, reserved); | |
732 | } | |
733 | printk(KERN_INFO "%ld pages of RAM\n", total_present); | |
734 | printk(KERN_INFO "%d reserved pages\n", total_reserved); | |
735 | printk(KERN_INFO "Total of %ld pages in page table cache\n", | |
736 | quicklist_total_size()); | |
737 | printk(KERN_INFO "%ld free buffer pages\n", nr_free_buffer_pages()); | |
738 | } |