init.c 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738
  1. /*
  2. * Initialize MMU support.
  3. *
  4. * Copyright (C) 1998-2003 Hewlett-Packard Co
  5. * David Mosberger-Tang <davidm@hpl.hp.com>
  6. */
  7. #include <linux/kernel.h>
  8. #include <linux/init.h>
  9. #include <linux/bootmem.h>
  10. #include <linux/efi.h>
  11. #include <linux/elf.h>
  12. #include <linux/mm.h>
  13. #include <linux/mmzone.h>
  14. #include <linux/module.h>
  15. #include <linux/personality.h>
  16. #include <linux/reboot.h>
  17. #include <linux/slab.h>
  18. #include <linux/swap.h>
  19. #include <linux/proc_fs.h>
  20. #include <linux/bitops.h>
  21. #include <linux/kexec.h>
  22. #include <asm/a.out.h>
  23. #include <asm/dma.h>
  24. #include <asm/ia32.h>
  25. #include <asm/io.h>
  26. #include <asm/machvec.h>
  27. #include <asm/numa.h>
  28. #include <asm/patch.h>
  29. #include <asm/pgalloc.h>
  30. #include <asm/sal.h>
  31. #include <asm/sections.h>
  32. #include <asm/system.h>
  33. #include <asm/tlb.h>
  34. #include <asm/uaccess.h>
  35. #include <asm/unistd.h>
  36. #include <asm/mca.h>
  37. DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
  38. extern void ia64_tlb_init (void);
  39. unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
  40. #ifdef CONFIG_VIRTUAL_MEM_MAP
  41. unsigned long vmalloc_end = VMALLOC_END_INIT;
  42. EXPORT_SYMBOL(vmalloc_end);
  43. struct page *vmem_map;
  44. EXPORT_SYMBOL(vmem_map);
  45. #endif
  46. struct page *zero_page_memmap_ptr; /* map entry for zero page */
  47. EXPORT_SYMBOL(zero_page_memmap_ptr);
  48. void
  49. __ia64_sync_icache_dcache (pte_t pte)
  50. {
  51. unsigned long addr;
  52. struct page *page;
  53. unsigned long order;
  54. page = pte_page(pte);
  55. addr = (unsigned long) page_address(page);
  56. if (test_bit(PG_arch_1, &page->flags))
  57. return; /* i-cache is already coherent with d-cache */
  58. if (PageCompound(page)) {
  59. order = compound_order(page);
  60. flush_icache_range(addr, addr + (1UL << order << PAGE_SHIFT));
  61. }
  62. else
  63. flush_icache_range(addr, addr + PAGE_SIZE);
  64. set_bit(PG_arch_1, &page->flags); /* mark page as clean */
  65. }
  66. /*
  67. * Since DMA is i-cache coherent, any (complete) pages that were written via
  68. * DMA can be marked as "clean" so that lazy_mmu_prot_update() doesn't have to
  69. * flush them when they get mapped into an executable vm-area.
  70. */
  71. void
  72. dma_mark_clean(void *addr, size_t size)
  73. {
  74. unsigned long pg_addr, end;
  75. pg_addr = PAGE_ALIGN((unsigned long) addr);
  76. end = (unsigned long) addr + size;
  77. while (pg_addr + PAGE_SIZE <= end) {
  78. struct page *page = virt_to_page(pg_addr);
  79. set_bit(PG_arch_1, &page->flags);
  80. pg_addr += PAGE_SIZE;
  81. }
  82. }
  83. inline void
  84. ia64_set_rbs_bot (void)
  85. {
  86. unsigned long stack_size = current->signal->rlim[RLIMIT_STACK].rlim_max & -16;
  87. if (stack_size > MAX_USER_STACK_SIZE)
  88. stack_size = MAX_USER_STACK_SIZE;
  89. current->thread.rbs_bot = PAGE_ALIGN(current->mm->start_stack - stack_size);
  90. }
  91. /*
  92. * This performs some platform-dependent address space initialization.
  93. * On IA-64, we want to setup the VM area for the register backing
  94. * store (which grows upwards) and install the gateway page which is
  95. * used for signal trampolines, etc.
  96. */
  97. void
  98. ia64_init_addr_space (void)
  99. {
  100. struct vm_area_struct *vma;
  101. ia64_set_rbs_bot();
  102. /*
  103. * If we're out of memory and kmem_cache_alloc() returns NULL, we simply ignore
  104. * the problem. When the process attempts to write to the register backing store
  105. * for the first time, it will get a SEGFAULT in this case.
  106. */
  107. vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
  108. if (vma) {
  109. vma->vm_mm = current->mm;
  110. vma->vm_start = current->thread.rbs_bot & PAGE_MASK;
  111. vma->vm_end = vma->vm_start + PAGE_SIZE;
  112. vma->vm_flags = VM_DATA_DEFAULT_FLAGS|VM_GROWSUP|VM_ACCOUNT;
  113. vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
  114. down_write(&current->mm->mmap_sem);
  115. if (insert_vm_struct(current->mm, vma)) {
  116. up_write(&current->mm->mmap_sem);
  117. kmem_cache_free(vm_area_cachep, vma);
  118. return;
  119. }
  120. up_write(&current->mm->mmap_sem);
  121. }
  122. /* map NaT-page at address zero to speed up speculative dereferencing of NULL: */
  123. if (!(current->personality & MMAP_PAGE_ZERO)) {
  124. vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
  125. if (vma) {
  126. vma->vm_mm = current->mm;
  127. vma->vm_end = PAGE_SIZE;
  128. vma->vm_page_prot = __pgprot(pgprot_val(PAGE_READONLY) | _PAGE_MA_NAT);
  129. vma->vm_flags = VM_READ | VM_MAYREAD | VM_IO | VM_RESERVED;
  130. down_write(&current->mm->mmap_sem);
  131. if (insert_vm_struct(current->mm, vma)) {
  132. up_write(&current->mm->mmap_sem);
  133. kmem_cache_free(vm_area_cachep, vma);
  134. return;
  135. }
  136. up_write(&current->mm->mmap_sem);
  137. }
  138. }
  139. }
  140. void
  141. free_initmem (void)
  142. {
  143. unsigned long addr, eaddr;
  144. addr = (unsigned long) ia64_imva(__init_begin);
  145. eaddr = (unsigned long) ia64_imva(__init_end);
  146. while (addr < eaddr) {
  147. ClearPageReserved(virt_to_page(addr));
  148. init_page_count(virt_to_page(addr));
  149. free_page(addr);
  150. ++totalram_pages;
  151. addr += PAGE_SIZE;
  152. }
  153. printk(KERN_INFO "Freeing unused kernel memory: %ldkB freed\n",
  154. (__init_end - __init_begin) >> 10);
  155. }
  156. void __init
  157. free_initrd_mem (unsigned long start, unsigned long end)
  158. {
  159. struct page *page;
  160. /*
  161. * EFI uses 4KB pages while the kernel can use 4KB or bigger.
  162. * Thus EFI and the kernel may have different page sizes. It is
  163. * therefore possible to have the initrd share the same page as
  164. * the end of the kernel (given current setup).
  165. *
  166. * To avoid freeing/using the wrong page (kernel sized) we:
  167. * - align up the beginning of initrd
  168. * - align down the end of initrd
  169. *
  170. * | |
  171. * |=============| a000
  172. * | |
  173. * | |
  174. * | | 9000
  175. * |/////////////|
  176. * |/////////////|
  177. * |=============| 8000
  178. * |///INITRD////|
  179. * |/////////////|
  180. * |/////////////| 7000
  181. * | |
  182. * |KKKKKKKKKKKKK|
  183. * |=============| 6000
  184. * |KKKKKKKKKKKKK|
  185. * |KKKKKKKKKKKKK|
  186. * K=kernel using 8KB pages
  187. *
  188. * In this example, we must free page 8000 ONLY. So we must align up
  189. * initrd_start and keep initrd_end as is.
  190. */
  191. start = PAGE_ALIGN(start);
  192. end = end & PAGE_MASK;
  193. if (start < end)
  194. printk(KERN_INFO "Freeing initrd memory: %ldkB freed\n", (end - start) >> 10);
  195. for (; start < end; start += PAGE_SIZE) {
  196. if (!virt_addr_valid(start))
  197. continue;
  198. page = virt_to_page(start);
  199. ClearPageReserved(page);
  200. init_page_count(page);
  201. free_page(start);
  202. ++totalram_pages;
  203. }
  204. }
  205. /*
  206. * This installs a clean page in the kernel's page table.
  207. */
  208. static struct page * __init
  209. put_kernel_page (struct page *page, unsigned long address, pgprot_t pgprot)
  210. {
  211. pgd_t *pgd;
  212. pud_t *pud;
  213. pmd_t *pmd;
  214. pte_t *pte;
  215. if (!PageReserved(page))
  216. printk(KERN_ERR "put_kernel_page: page at 0x%p not in reserved memory\n",
  217. page_address(page));
  218. pgd = pgd_offset_k(address); /* note: this is NOT pgd_offset()! */
  219. {
  220. pud = pud_alloc(&init_mm, pgd, address);
  221. if (!pud)
  222. goto out;
  223. pmd = pmd_alloc(&init_mm, pud, address);
  224. if (!pmd)
  225. goto out;
  226. pte = pte_alloc_kernel(pmd, address);
  227. if (!pte)
  228. goto out;
  229. if (!pte_none(*pte))
  230. goto out;
  231. set_pte(pte, mk_pte(page, pgprot));
  232. }
  233. out:
  234. /* no need for flush_tlb */
  235. return page;
  236. }
  237. static void __init
  238. setup_gate (void)
  239. {
  240. struct page *page;
  241. /*
  242. * Map the gate page twice: once read-only to export the ELF
  243. * headers etc. and once execute-only page to enable
  244. * privilege-promotion via "epc":
  245. */
  246. page = virt_to_page(ia64_imva(__start_gate_section));
  247. put_kernel_page(page, GATE_ADDR, PAGE_READONLY);
  248. #ifdef HAVE_BUGGY_SEGREL
  249. page = virt_to_page(ia64_imva(__start_gate_section + PAGE_SIZE));
  250. put_kernel_page(page, GATE_ADDR + PAGE_SIZE, PAGE_GATE);
  251. #else
  252. put_kernel_page(page, GATE_ADDR + PERCPU_PAGE_SIZE, PAGE_GATE);
  253. /* Fill in the holes (if any) with read-only zero pages: */
  254. {
  255. unsigned long addr;
  256. for (addr = GATE_ADDR + PAGE_SIZE;
  257. addr < GATE_ADDR + PERCPU_PAGE_SIZE;
  258. addr += PAGE_SIZE)
  259. {
  260. put_kernel_page(ZERO_PAGE(0), addr,
  261. PAGE_READONLY);
  262. put_kernel_page(ZERO_PAGE(0), addr + PERCPU_PAGE_SIZE,
  263. PAGE_READONLY);
  264. }
  265. }
  266. #endif
  267. ia64_patch_gate();
  268. }
  269. void __devinit
  270. ia64_mmu_init (void *my_cpu_data)
  271. {
  272. unsigned long pta, impl_va_bits;
  273. extern void __devinit tlb_init (void);
  274. #ifdef CONFIG_DISABLE_VHPT
  275. # define VHPT_ENABLE_BIT 0
  276. #else
  277. # define VHPT_ENABLE_BIT 1
  278. #endif
  279. /*
  280. * Check if the virtually mapped linear page table (VMLPT) overlaps with a mapped
  281. * address space. The IA-64 architecture guarantees that at least 50 bits of
  282. * virtual address space are implemented but if we pick a large enough page size
  283. * (e.g., 64KB), the mapped address space is big enough that it will overlap with
  284. * VMLPT. I assume that once we run on machines big enough to warrant 64KB pages,
  285. * IMPL_VA_MSB will be significantly bigger, so this is unlikely to become a
  286. * problem in practice. Alternatively, we could truncate the top of the mapped
  287. * address space to not permit mappings that would overlap with the VMLPT.
  288. * --davidm 00/12/06
  289. */
  290. # define pte_bits 3
  291. # define mapped_space_bits (3*(PAGE_SHIFT - pte_bits) + PAGE_SHIFT)
  292. /*
  293. * The virtual page table has to cover the entire implemented address space within
  294. * a region even though not all of this space may be mappable. The reason for
  295. * this is that the Access bit and Dirty bit fault handlers perform
  296. * non-speculative accesses to the virtual page table, so the address range of the
  297. * virtual page table itself needs to be covered by virtual page table.
  298. */
  299. # define vmlpt_bits (impl_va_bits - PAGE_SHIFT + pte_bits)
  300. # define POW2(n) (1ULL << (n))
  301. impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61)));
  302. if (impl_va_bits < 51 || impl_va_bits > 61)
  303. panic("CPU has bogus IMPL_VA_MSB value of %lu!\n", impl_va_bits - 1);
  304. /*
  305. * mapped_space_bits - PAGE_SHIFT is the total number of ptes we need,
  306. * which must fit into "vmlpt_bits - pte_bits" slots. Second half of
  307. * the test makes sure that our mapped space doesn't overlap the
  308. * unimplemented hole in the middle of the region.
  309. */
  310. if ((mapped_space_bits - PAGE_SHIFT > vmlpt_bits - pte_bits) ||
  311. (mapped_space_bits > impl_va_bits - 1))
  312. panic("Cannot build a big enough virtual-linear page table"
  313. " to cover mapped address space.\n"
  314. " Try using a smaller page size.\n");
  315. /* place the VMLPT at the end of each page-table mapped region: */
  316. pta = POW2(61) - POW2(vmlpt_bits);
  317. /*
  318. * Set the (virtually mapped linear) page table address. Bit
  319. * 8 selects between the short and long format, bits 2-7 the
  320. * size of the table, and bit 0 whether the VHPT walker is
  321. * enabled.
  322. */
  323. ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | VHPT_ENABLE_BIT);
  324. ia64_tlb_init();
  325. #ifdef CONFIG_HUGETLB_PAGE
  326. ia64_set_rr(HPAGE_REGION_BASE, HPAGE_SHIFT << 2);
  327. ia64_srlz_d();
  328. #endif
  329. }
  330. #ifdef CONFIG_VIRTUAL_MEM_MAP
  331. int vmemmap_find_next_valid_pfn(int node, int i)
  332. {
  333. unsigned long end_address, hole_next_pfn;
  334. unsigned long stop_address;
  335. pg_data_t *pgdat = NODE_DATA(node);
  336. end_address = (unsigned long) &vmem_map[pgdat->node_start_pfn + i];
  337. end_address = PAGE_ALIGN(end_address);
  338. stop_address = (unsigned long) &vmem_map[
  339. pgdat->node_start_pfn + pgdat->node_spanned_pages];
  340. do {
  341. pgd_t *pgd;
  342. pud_t *pud;
  343. pmd_t *pmd;
  344. pte_t *pte;
  345. pgd = pgd_offset_k(end_address);
  346. if (pgd_none(*pgd)) {
  347. end_address += PGDIR_SIZE;
  348. continue;
  349. }
  350. pud = pud_offset(pgd, end_address);
  351. if (pud_none(*pud)) {
  352. end_address += PUD_SIZE;
  353. continue;
  354. }
  355. pmd = pmd_offset(pud, end_address);
  356. if (pmd_none(*pmd)) {
  357. end_address += PMD_SIZE;
  358. continue;
  359. }
  360. pte = pte_offset_kernel(pmd, end_address);
  361. retry_pte:
  362. if (pte_none(*pte)) {
  363. end_address += PAGE_SIZE;
  364. pte++;
  365. if ((end_address < stop_address) &&
  366. (end_address != ALIGN(end_address, 1UL << PMD_SHIFT)))
  367. goto retry_pte;
  368. continue;
  369. }
  370. /* Found next valid vmem_map page */
  371. break;
  372. } while (end_address < stop_address);
  373. end_address = min(end_address, stop_address);
  374. end_address = end_address - (unsigned long) vmem_map + sizeof(struct page) - 1;
  375. hole_next_pfn = end_address / sizeof(struct page);
  376. return hole_next_pfn - pgdat->node_start_pfn;
  377. }
  378. int __init
  379. create_mem_map_page_table (u64 start, u64 end, void *arg)
  380. {
  381. unsigned long address, start_page, end_page;
  382. struct page *map_start, *map_end;
  383. int node;
  384. pgd_t *pgd;
  385. pud_t *pud;
  386. pmd_t *pmd;
  387. pte_t *pte;
  388. map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
  389. map_end = vmem_map + (__pa(end) >> PAGE_SHIFT);
  390. start_page = (unsigned long) map_start & PAGE_MASK;
  391. end_page = PAGE_ALIGN((unsigned long) map_end);
  392. node = paddr_to_nid(__pa(start));
  393. for (address = start_page; address < end_page; address += PAGE_SIZE) {
  394. pgd = pgd_offset_k(address);
  395. if (pgd_none(*pgd))
  396. pgd_populate(&init_mm, pgd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
  397. pud = pud_offset(pgd, address);
  398. if (pud_none(*pud))
  399. pud_populate(&init_mm, pud, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
  400. pmd = pmd_offset(pud, address);
  401. if (pmd_none(*pmd))
  402. pmd_populate_kernel(&init_mm, pmd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
  403. pte = pte_offset_kernel(pmd, address);
  404. if (pte_none(*pte))
  405. set_pte(pte, pfn_pte(__pa(alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)) >> PAGE_SHIFT,
  406. PAGE_KERNEL));
  407. }
  408. return 0;
  409. }
  410. struct memmap_init_callback_data {
  411. struct page *start;
  412. struct page *end;
  413. int nid;
  414. unsigned long zone;
  415. };
  416. static int __meminit
  417. virtual_memmap_init (u64 start, u64 end, void *arg)
  418. {
  419. struct memmap_init_callback_data *args;
  420. struct page *map_start, *map_end;
  421. args = (struct memmap_init_callback_data *) arg;
  422. map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
  423. map_end = vmem_map + (__pa(end) >> PAGE_SHIFT);
  424. if (map_start < args->start)
  425. map_start = args->start;
  426. if (map_end > args->end)
  427. map_end = args->end;
  428. /*
  429. * We have to initialize "out of bounds" struct page elements that fit completely
  430. * on the same pages that were allocated for the "in bounds" elements because they
  431. * may be referenced later (and found to be "reserved").
  432. */
  433. map_start -= ((unsigned long) map_start & (PAGE_SIZE - 1)) / sizeof(struct page);
  434. map_end += ((PAGE_ALIGN((unsigned long) map_end) - (unsigned long) map_end)
  435. / sizeof(struct page));
  436. if (map_start < map_end)
  437. memmap_init_zone((unsigned long)(map_end - map_start),
  438. args->nid, args->zone, page_to_pfn(map_start),
  439. MEMMAP_EARLY);
  440. return 0;
  441. }
  442. void __meminit
  443. memmap_init (unsigned long size, int nid, unsigned long zone,
  444. unsigned long start_pfn)
  445. {
  446. if (!vmem_map)
  447. memmap_init_zone(size, nid, zone, start_pfn, MEMMAP_EARLY);
  448. else {
  449. struct page *start;
  450. struct memmap_init_callback_data args;
  451. start = pfn_to_page(start_pfn);
  452. args.start = start;
  453. args.end = start + size;
  454. args.nid = nid;
  455. args.zone = zone;
  456. efi_memmap_walk(virtual_memmap_init, &args);
  457. }
  458. }
  459. int
  460. ia64_pfn_valid (unsigned long pfn)
  461. {
  462. char byte;
  463. struct page *pg = pfn_to_page(pfn);
  464. return (__get_user(byte, (char __user *) pg) == 0)
  465. && ((((u64)pg & PAGE_MASK) == (((u64)(pg + 1) - 1) & PAGE_MASK))
  466. || (__get_user(byte, (char __user *) (pg + 1) - 1) == 0));
  467. }
  468. EXPORT_SYMBOL(ia64_pfn_valid);
  469. int __init
  470. find_largest_hole (u64 start, u64 end, void *arg)
  471. {
  472. u64 *max_gap = arg;
  473. static u64 last_end = PAGE_OFFSET;
  474. /* NOTE: this algorithm assumes efi memmap table is ordered */
  475. if (*max_gap < (start - last_end))
  476. *max_gap = start - last_end;
  477. last_end = end;
  478. return 0;
  479. }
  480. #endif /* CONFIG_VIRTUAL_MEM_MAP */
  481. int __init
  482. register_active_ranges(u64 start, u64 end, void *arg)
  483. {
  484. int nid = paddr_to_nid(__pa(start));
  485. if (nid < 0)
  486. nid = 0;
  487. #ifdef CONFIG_KEXEC
  488. if (start > crashk_res.start && start < crashk_res.end)
  489. start = crashk_res.end;
  490. if (end > crashk_res.start && end < crashk_res.end)
  491. end = crashk_res.start;
  492. #endif
  493. if (start < end)
  494. add_active_range(nid, __pa(start) >> PAGE_SHIFT,
  495. __pa(end) >> PAGE_SHIFT);
  496. return 0;
  497. }
  498. static int __init
  499. count_reserved_pages (u64 start, u64 end, void *arg)
  500. {
  501. unsigned long num_reserved = 0;
  502. unsigned long *count = arg;
  503. for (; start < end; start += PAGE_SIZE)
  504. if (PageReserved(virt_to_page(start)))
  505. ++num_reserved;
  506. *count += num_reserved;
  507. return 0;
  508. }
  509. int
  510. find_max_min_low_pfn (unsigned long start, unsigned long end, void *arg)
  511. {
  512. unsigned long pfn_start, pfn_end;
  513. #ifdef CONFIG_FLATMEM
  514. pfn_start = (PAGE_ALIGN(__pa(start))) >> PAGE_SHIFT;
  515. pfn_end = (PAGE_ALIGN(__pa(end - 1))) >> PAGE_SHIFT;
  516. #else
  517. pfn_start = GRANULEROUNDDOWN(__pa(start)) >> PAGE_SHIFT;
  518. pfn_end = GRANULEROUNDUP(__pa(end - 1)) >> PAGE_SHIFT;
  519. #endif
  520. min_low_pfn = min(min_low_pfn, pfn_start);
  521. max_low_pfn = max(max_low_pfn, pfn_end);
  522. return 0;
  523. }
  524. /*
  525. * Boot command-line option "nolwsys" can be used to disable the use of any light-weight
  526. * system call handler. When this option is in effect, all fsyscalls will end up bubbling
  527. * down into the kernel and calling the normal (heavy-weight) syscall handler. This is
  528. * useful for performance testing, but conceivably could also come in handy for debugging
  529. * purposes.
  530. */
  531. static int nolwsys __initdata;
  532. static int __init
  533. nolwsys_setup (char *s)
  534. {
  535. nolwsys = 1;
  536. return 1;
  537. }
  538. __setup("nolwsys", nolwsys_setup);
  539. void __init
  540. mem_init (void)
  541. {
  542. long reserved_pages, codesize, datasize, initsize;
  543. pg_data_t *pgdat;
  544. int i;
  545. static struct kcore_list kcore_mem, kcore_vmem, kcore_kernel;
  546. BUG_ON(PTRS_PER_PGD * sizeof(pgd_t) != PAGE_SIZE);
  547. BUG_ON(PTRS_PER_PMD * sizeof(pmd_t) != PAGE_SIZE);
  548. BUG_ON(PTRS_PER_PTE * sizeof(pte_t) != PAGE_SIZE);
  549. #ifdef CONFIG_PCI
  550. /*
  551. * This needs to be called _after_ the command line has been parsed but _before_
  552. * any drivers that may need the PCI DMA interface are initialized or bootmem has
  553. * been freed.
  554. */
  555. platform_dma_init();
  556. #endif
  557. #ifdef CONFIG_FLATMEM
  558. if (!mem_map)
  559. BUG();
  560. max_mapnr = max_low_pfn;
  561. #endif
  562. high_memory = __va(max_low_pfn * PAGE_SIZE);
  563. kclist_add(&kcore_mem, __va(0), max_low_pfn * PAGE_SIZE);
  564. kclist_add(&kcore_vmem, (void *)VMALLOC_START, VMALLOC_END-VMALLOC_START);
  565. kclist_add(&kcore_kernel, _stext, _end - _stext);
  566. for_each_online_pgdat(pgdat)
  567. if (pgdat->bdata->node_bootmem_map)
  568. totalram_pages += free_all_bootmem_node(pgdat);
  569. reserved_pages = 0;
  570. efi_memmap_walk(count_reserved_pages, &reserved_pages);
  571. codesize = (unsigned long) _etext - (unsigned long) _stext;
  572. datasize = (unsigned long) _edata - (unsigned long) _etext;
  573. initsize = (unsigned long) __init_end - (unsigned long) __init_begin;
  574. printk(KERN_INFO "Memory: %luk/%luk available (%luk code, %luk reserved, "
  575. "%luk data, %luk init)\n", (unsigned long) nr_free_pages() << (PAGE_SHIFT - 10),
  576. num_physpages << (PAGE_SHIFT - 10), codesize >> 10,
  577. reserved_pages << (PAGE_SHIFT - 10), datasize >> 10, initsize >> 10);
  578. /*
  579. * For fsyscall entrpoints with no light-weight handler, use the ordinary
  580. * (heavy-weight) handler, but mark it by setting bit 0, so the fsyscall entry
  581. * code can tell them apart.
  582. */
  583. for (i = 0; i < NR_syscalls; ++i) {
  584. extern unsigned long fsyscall_table[NR_syscalls];
  585. extern unsigned long sys_call_table[NR_syscalls];
  586. if (!fsyscall_table[i] || nolwsys)
  587. fsyscall_table[i] = sys_call_table[i] | 1;
  588. }
  589. setup_gate();
  590. #ifdef CONFIG_IA32_SUPPORT
  591. ia32_mem_init();
  592. #endif
  593. }
  594. #ifdef CONFIG_MEMORY_HOTPLUG
  595. void online_page(struct page *page)
  596. {
  597. ClearPageReserved(page);
  598. init_page_count(page);
  599. __free_page(page);
  600. totalram_pages++;
  601. num_physpages++;
  602. }
  603. int arch_add_memory(int nid, u64 start, u64 size)
  604. {
  605. pg_data_t *pgdat;
  606. struct zone *zone;
  607. unsigned long start_pfn = start >> PAGE_SHIFT;
  608. unsigned long nr_pages = size >> PAGE_SHIFT;
  609. int ret;
  610. pgdat = NODE_DATA(nid);
  611. zone = pgdat->node_zones + ZONE_NORMAL;
  612. ret = __add_pages(zone, start_pfn, nr_pages);
  613. if (ret)
  614. printk("%s: Problem encountered in __add_pages() as ret=%d\n",
  615. __FUNCTION__, ret);
  616. return ret;
  617. }
  618. #ifdef CONFIG_MEMORY_HOTREMOVE
  619. int remove_memory(u64 start, u64 size)
  620. {
  621. unsigned long start_pfn, end_pfn;
  622. unsigned long timeout = 120 * HZ;
  623. int ret;
  624. start_pfn = start >> PAGE_SHIFT;
  625. end_pfn = start_pfn + (size >> PAGE_SHIFT);
  626. ret = offline_pages(start_pfn, end_pfn, timeout);
  627. if (ret)
  628. goto out;
  629. /* we can free mem_map at this point */
  630. out:
  631. return ret;
  632. }
  633. EXPORT_SYMBOL_GPL(remove_memory);
  634. #endif /* CONFIG_MEMORY_HOTREMOVE */
  635. #endif