init.c 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741
  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/memblock.h>
  13. #include <linux/mm.h>
  14. #include <linux/mmzone.h>
  15. #include <linux/module.h>
  16. #include <linux/personality.h>
  17. #include <linux/reboot.h>
  18. #include <linux/slab.h>
  19. #include <linux/swap.h>
  20. #include <linux/proc_fs.h>
  21. #include <linux/bitops.h>
  22. #include <linux/kexec.h>
  23. #include <asm/dma.h>
  24. #include <asm/io.h>
  25. #include <asm/machvec.h>
  26. #include <asm/numa.h>
  27. #include <asm/patch.h>
  28. #include <asm/pgalloc.h>
  29. #include <asm/sal.h>
  30. #include <asm/sections.h>
  31. #include <asm/tlb.h>
  32. #include <asm/uaccess.h>
  33. #include <asm/unistd.h>
  34. #include <asm/mca.h>
  35. #include <asm/paravirt.h>
  36. extern void ia64_tlb_init (void);
  37. unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
  38. #ifdef CONFIG_VIRTUAL_MEM_MAP
  39. unsigned long VMALLOC_END = VMALLOC_END_INIT;
  40. EXPORT_SYMBOL(VMALLOC_END);
  41. struct page *vmem_map;
  42. EXPORT_SYMBOL(vmem_map);
  43. #endif
  44. struct page *zero_page_memmap_ptr; /* map entry for zero page */
  45. EXPORT_SYMBOL(zero_page_memmap_ptr);
  46. void
  47. __ia64_sync_icache_dcache (pte_t pte)
  48. {
  49. unsigned long addr;
  50. struct page *page;
  51. page = pte_page(pte);
  52. addr = (unsigned long) page_address(page);
  53. if (test_bit(PG_arch_1, &page->flags))
  54. return; /* i-cache is already coherent with d-cache */
  55. flush_icache_range(addr, addr + (PAGE_SIZE << compound_order(page)));
  56. set_bit(PG_arch_1, &page->flags); /* mark page as clean */
  57. }
  58. /*
  59. * Since DMA is i-cache coherent, any (complete) pages that were written via
  60. * DMA can be marked as "clean" so that lazy_mmu_prot_update() doesn't have to
  61. * flush them when they get mapped into an executable vm-area.
  62. */
  63. void
  64. dma_mark_clean(void *addr, size_t size)
  65. {
  66. unsigned long pg_addr, end;
  67. pg_addr = PAGE_ALIGN((unsigned long) addr);
  68. end = (unsigned long) addr + size;
  69. while (pg_addr + PAGE_SIZE <= end) {
  70. struct page *page = virt_to_page(pg_addr);
  71. set_bit(PG_arch_1, &page->flags);
  72. pg_addr += PAGE_SIZE;
  73. }
  74. }
  75. inline void
  76. ia64_set_rbs_bot (void)
  77. {
  78. unsigned long stack_size = rlimit_max(RLIMIT_STACK) & -16;
  79. if (stack_size > MAX_USER_STACK_SIZE)
  80. stack_size = MAX_USER_STACK_SIZE;
  81. current->thread.rbs_bot = PAGE_ALIGN(current->mm->start_stack - stack_size);
  82. }
  83. /*
  84. * This performs some platform-dependent address space initialization.
  85. * On IA-64, we want to setup the VM area for the register backing
  86. * store (which grows upwards) and install the gateway page which is
  87. * used for signal trampolines, etc.
  88. */
  89. void
  90. ia64_init_addr_space (void)
  91. {
  92. struct vm_area_struct *vma;
  93. ia64_set_rbs_bot();
  94. /*
  95. * If we're out of memory and kmem_cache_alloc() returns NULL, we simply ignore
  96. * the problem. When the process attempts to write to the register backing store
  97. * for the first time, it will get a SEGFAULT in this case.
  98. */
  99. vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
  100. if (vma) {
  101. INIT_LIST_HEAD(&vma->anon_vma_chain);
  102. vma->vm_mm = current->mm;
  103. vma->vm_start = current->thread.rbs_bot & PAGE_MASK;
  104. vma->vm_end = vma->vm_start + PAGE_SIZE;
  105. vma->vm_flags = VM_DATA_DEFAULT_FLAGS|VM_GROWSUP|VM_ACCOUNT;
  106. vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
  107. down_write(&current->mm->mmap_sem);
  108. if (insert_vm_struct(current->mm, vma)) {
  109. up_write(&current->mm->mmap_sem);
  110. kmem_cache_free(vm_area_cachep, vma);
  111. return;
  112. }
  113. up_write(&current->mm->mmap_sem);
  114. }
  115. /* map NaT-page at address zero to speed up speculative dereferencing of NULL: */
  116. if (!(current->personality & MMAP_PAGE_ZERO)) {
  117. vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
  118. if (vma) {
  119. INIT_LIST_HEAD(&vma->anon_vma_chain);
  120. vma->vm_mm = current->mm;
  121. vma->vm_end = PAGE_SIZE;
  122. vma->vm_page_prot = __pgprot(pgprot_val(PAGE_READONLY) | _PAGE_MA_NAT);
  123. vma->vm_flags = VM_READ | VM_MAYREAD | VM_IO |
  124. VM_DONTEXPAND | VM_DONTDUMP;
  125. down_write(&current->mm->mmap_sem);
  126. if (insert_vm_struct(current->mm, vma)) {
  127. up_write(&current->mm->mmap_sem);
  128. kmem_cache_free(vm_area_cachep, vma);
  129. return;
  130. }
  131. up_write(&current->mm->mmap_sem);
  132. }
  133. }
  134. }
  135. void
  136. free_initmem (void)
  137. {
  138. free_reserved_area(ia64_imva(__init_begin), ia64_imva(__init_end),
  139. -1, "unused kernel");
  140. }
  141. void __init
  142. free_initrd_mem (unsigned long start, unsigned long end)
  143. {
  144. /*
  145. * EFI uses 4KB pages while the kernel can use 4KB or bigger.
  146. * Thus EFI and the kernel may have different page sizes. It is
  147. * therefore possible to have the initrd share the same page as
  148. * the end of the kernel (given current setup).
  149. *
  150. * To avoid freeing/using the wrong page (kernel sized) we:
  151. * - align up the beginning of initrd
  152. * - align down the end of initrd
  153. *
  154. * | |
  155. * |=============| a000
  156. * | |
  157. * | |
  158. * | | 9000
  159. * |/////////////|
  160. * |/////////////|
  161. * |=============| 8000
  162. * |///INITRD////|
  163. * |/////////////|
  164. * |/////////////| 7000
  165. * | |
  166. * |KKKKKKKKKKKKK|
  167. * |=============| 6000
  168. * |KKKKKKKKKKKKK|
  169. * |KKKKKKKKKKKKK|
  170. * K=kernel using 8KB pages
  171. *
  172. * In this example, we must free page 8000 ONLY. So we must align up
  173. * initrd_start and keep initrd_end as is.
  174. */
  175. start = PAGE_ALIGN(start);
  176. end = end & PAGE_MASK;
  177. if (start < end)
  178. printk(KERN_INFO "Freeing initrd memory: %ldkB freed\n", (end - start) >> 10);
  179. for (; start < end; start += PAGE_SIZE) {
  180. if (!virt_addr_valid(start))
  181. continue;
  182. free_reserved_page(virt_to_page(start));
  183. }
  184. }
  185. /*
  186. * This installs a clean page in the kernel's page table.
  187. */
  188. static struct page * __init
  189. put_kernel_page (struct page *page, unsigned long address, pgprot_t pgprot)
  190. {
  191. pgd_t *pgd;
  192. pud_t *pud;
  193. pmd_t *pmd;
  194. pte_t *pte;
  195. if (!PageReserved(page))
  196. printk(KERN_ERR "put_kernel_page: page at 0x%p not in reserved memory\n",
  197. page_address(page));
  198. pgd = pgd_offset_k(address); /* note: this is NOT pgd_offset()! */
  199. {
  200. pud = pud_alloc(&init_mm, pgd, address);
  201. if (!pud)
  202. goto out;
  203. pmd = pmd_alloc(&init_mm, pud, address);
  204. if (!pmd)
  205. goto out;
  206. pte = pte_alloc_kernel(pmd, address);
  207. if (!pte)
  208. goto out;
  209. if (!pte_none(*pte))
  210. goto out;
  211. set_pte(pte, mk_pte(page, pgprot));
  212. }
  213. out:
  214. /* no need for flush_tlb */
  215. return page;
  216. }
  217. static void __init
  218. setup_gate (void)
  219. {
  220. void *gate_section;
  221. struct page *page;
  222. /*
  223. * Map the gate page twice: once read-only to export the ELF
  224. * headers etc. and once execute-only page to enable
  225. * privilege-promotion via "epc":
  226. */
  227. gate_section = paravirt_get_gate_section();
  228. page = virt_to_page(ia64_imva(gate_section));
  229. put_kernel_page(page, GATE_ADDR, PAGE_READONLY);
  230. #ifdef HAVE_BUGGY_SEGREL
  231. page = virt_to_page(ia64_imva(gate_section + PAGE_SIZE));
  232. put_kernel_page(page, GATE_ADDR + PAGE_SIZE, PAGE_GATE);
  233. #else
  234. put_kernel_page(page, GATE_ADDR + PERCPU_PAGE_SIZE, PAGE_GATE);
  235. /* Fill in the holes (if any) with read-only zero pages: */
  236. {
  237. unsigned long addr;
  238. for (addr = GATE_ADDR + PAGE_SIZE;
  239. addr < GATE_ADDR + PERCPU_PAGE_SIZE;
  240. addr += PAGE_SIZE)
  241. {
  242. put_kernel_page(ZERO_PAGE(0), addr,
  243. PAGE_READONLY);
  244. put_kernel_page(ZERO_PAGE(0), addr + PERCPU_PAGE_SIZE,
  245. PAGE_READONLY);
  246. }
  247. }
  248. #endif
  249. ia64_patch_gate();
  250. }
  251. static struct vm_area_struct gate_vma;
  252. static int __init gate_vma_init(void)
  253. {
  254. gate_vma.vm_mm = NULL;
  255. gate_vma.vm_start = FIXADDR_USER_START;
  256. gate_vma.vm_end = FIXADDR_USER_END;
  257. gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC;
  258. gate_vma.vm_page_prot = __P101;
  259. return 0;
  260. }
  261. __initcall(gate_vma_init);
  262. struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
  263. {
  264. return &gate_vma;
  265. }
  266. int in_gate_area_no_mm(unsigned long addr)
  267. {
  268. if ((addr >= FIXADDR_USER_START) && (addr < FIXADDR_USER_END))
  269. return 1;
  270. return 0;
  271. }
  272. int in_gate_area(struct mm_struct *mm, unsigned long addr)
  273. {
  274. return in_gate_area_no_mm(addr);
  275. }
  276. void ia64_mmu_init(void *my_cpu_data)
  277. {
  278. unsigned long pta, impl_va_bits;
  279. extern void tlb_init(void);
  280. #ifdef CONFIG_DISABLE_VHPT
  281. # define VHPT_ENABLE_BIT 0
  282. #else
  283. # define VHPT_ENABLE_BIT 1
  284. #endif
  285. /*
  286. * Check if the virtually mapped linear page table (VMLPT) overlaps with a mapped
  287. * address space. The IA-64 architecture guarantees that at least 50 bits of
  288. * virtual address space are implemented but if we pick a large enough page size
  289. * (e.g., 64KB), the mapped address space is big enough that it will overlap with
  290. * VMLPT. I assume that once we run on machines big enough to warrant 64KB pages,
  291. * IMPL_VA_MSB will be significantly bigger, so this is unlikely to become a
  292. * problem in practice. Alternatively, we could truncate the top of the mapped
  293. * address space to not permit mappings that would overlap with the VMLPT.
  294. * --davidm 00/12/06
  295. */
  296. # define pte_bits 3
  297. # define mapped_space_bits (3*(PAGE_SHIFT - pte_bits) + PAGE_SHIFT)
  298. /*
  299. * The virtual page table has to cover the entire implemented address space within
  300. * a region even though not all of this space may be mappable. The reason for
  301. * this is that the Access bit and Dirty bit fault handlers perform
  302. * non-speculative accesses to the virtual page table, so the address range of the
  303. * virtual page table itself needs to be covered by virtual page table.
  304. */
  305. # define vmlpt_bits (impl_va_bits - PAGE_SHIFT + pte_bits)
  306. # define POW2(n) (1ULL << (n))
  307. impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61)));
  308. if (impl_va_bits < 51 || impl_va_bits > 61)
  309. panic("CPU has bogus IMPL_VA_MSB value of %lu!\n", impl_va_bits - 1);
  310. /*
  311. * mapped_space_bits - PAGE_SHIFT is the total number of ptes we need,
  312. * which must fit into "vmlpt_bits - pte_bits" slots. Second half of
  313. * the test makes sure that our mapped space doesn't overlap the
  314. * unimplemented hole in the middle of the region.
  315. */
  316. if ((mapped_space_bits - PAGE_SHIFT > vmlpt_bits - pte_bits) ||
  317. (mapped_space_bits > impl_va_bits - 1))
  318. panic("Cannot build a big enough virtual-linear page table"
  319. " to cover mapped address space.\n"
  320. " Try using a smaller page size.\n");
  321. /* place the VMLPT at the end of each page-table mapped region: */
  322. pta = POW2(61) - POW2(vmlpt_bits);
  323. /*
  324. * Set the (virtually mapped linear) page table address. Bit
  325. * 8 selects between the short and long format, bits 2-7 the
  326. * size of the table, and bit 0 whether the VHPT walker is
  327. * enabled.
  328. */
  329. ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | VHPT_ENABLE_BIT);
  330. ia64_tlb_init();
  331. #ifdef CONFIG_HUGETLB_PAGE
  332. ia64_set_rr(HPAGE_REGION_BASE, HPAGE_SHIFT << 2);
  333. ia64_srlz_d();
  334. #endif
  335. }
  336. #ifdef CONFIG_VIRTUAL_MEM_MAP
  337. int vmemmap_find_next_valid_pfn(int node, int i)
  338. {
  339. unsigned long end_address, hole_next_pfn;
  340. unsigned long stop_address;
  341. pg_data_t *pgdat = NODE_DATA(node);
  342. end_address = (unsigned long) &vmem_map[pgdat->node_start_pfn + i];
  343. end_address = PAGE_ALIGN(end_address);
  344. stop_address = (unsigned long) &vmem_map[pgdat_end_pfn(pgdat)];
  345. do {
  346. pgd_t *pgd;
  347. pud_t *pud;
  348. pmd_t *pmd;
  349. pte_t *pte;
  350. pgd = pgd_offset_k(end_address);
  351. if (pgd_none(*pgd)) {
  352. end_address += PGDIR_SIZE;
  353. continue;
  354. }
  355. pud = pud_offset(pgd, end_address);
  356. if (pud_none(*pud)) {
  357. end_address += PUD_SIZE;
  358. continue;
  359. }
  360. pmd = pmd_offset(pud, end_address);
  361. if (pmd_none(*pmd)) {
  362. end_address += PMD_SIZE;
  363. continue;
  364. }
  365. pte = pte_offset_kernel(pmd, end_address);
  366. retry_pte:
  367. if (pte_none(*pte)) {
  368. end_address += PAGE_SIZE;
  369. pte++;
  370. if ((end_address < stop_address) &&
  371. (end_address != ALIGN(end_address, 1UL << PMD_SHIFT)))
  372. goto retry_pte;
  373. continue;
  374. }
  375. /* Found next valid vmem_map page */
  376. break;
  377. } while (end_address < stop_address);
  378. end_address = min(end_address, stop_address);
  379. end_address = end_address - (unsigned long) vmem_map + sizeof(struct page) - 1;
  380. hole_next_pfn = end_address / sizeof(struct page);
  381. return hole_next_pfn - pgdat->node_start_pfn;
  382. }
  383. int __init create_mem_map_page_table(u64 start, u64 end, void *arg)
  384. {
  385. unsigned long address, start_page, end_page;
  386. struct page *map_start, *map_end;
  387. int node;
  388. pgd_t *pgd;
  389. pud_t *pud;
  390. pmd_t *pmd;
  391. pte_t *pte;
  392. map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
  393. map_end = vmem_map + (__pa(end) >> PAGE_SHIFT);
  394. start_page = (unsigned long) map_start & PAGE_MASK;
  395. end_page = PAGE_ALIGN((unsigned long) map_end);
  396. node = paddr_to_nid(__pa(start));
  397. for (address = start_page; address < end_page; address += PAGE_SIZE) {
  398. pgd = pgd_offset_k(address);
  399. if (pgd_none(*pgd))
  400. pgd_populate(&init_mm, pgd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
  401. pud = pud_offset(pgd, address);
  402. if (pud_none(*pud))
  403. pud_populate(&init_mm, pud, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
  404. pmd = pmd_offset(pud, address);
  405. if (pmd_none(*pmd))
  406. pmd_populate_kernel(&init_mm, pmd, alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE));
  407. pte = pte_offset_kernel(pmd, address);
  408. if (pte_none(*pte))
  409. set_pte(pte, pfn_pte(__pa(alloc_bootmem_pages_node(NODE_DATA(node), PAGE_SIZE)) >> PAGE_SHIFT,
  410. PAGE_KERNEL));
  411. }
  412. return 0;
  413. }
  414. struct memmap_init_callback_data {
  415. struct page *start;
  416. struct page *end;
  417. int nid;
  418. unsigned long zone;
  419. };
  420. static int __meminit
  421. virtual_memmap_init(u64 start, u64 end, void *arg)
  422. {
  423. struct memmap_init_callback_data *args;
  424. struct page *map_start, *map_end;
  425. args = (struct memmap_init_callback_data *) arg;
  426. map_start = vmem_map + (__pa(start) >> PAGE_SHIFT);
  427. map_end = vmem_map + (__pa(end) >> PAGE_SHIFT);
  428. if (map_start < args->start)
  429. map_start = args->start;
  430. if (map_end > args->end)
  431. map_end = args->end;
  432. /*
  433. * We have to initialize "out of bounds" struct page elements that fit completely
  434. * on the same pages that were allocated for the "in bounds" elements because they
  435. * may be referenced later (and found to be "reserved").
  436. */
  437. map_start -= ((unsigned long) map_start & (PAGE_SIZE - 1)) / sizeof(struct page);
  438. map_end += ((PAGE_ALIGN((unsigned long) map_end) - (unsigned long) map_end)
  439. / sizeof(struct page));
  440. if (map_start < map_end)
  441. memmap_init_zone((unsigned long)(map_end - map_start),
  442. args->nid, args->zone, page_to_pfn(map_start),
  443. MEMMAP_EARLY);
  444. return 0;
  445. }
  446. void __meminit
  447. memmap_init (unsigned long size, int nid, unsigned long zone,
  448. unsigned long start_pfn)
  449. {
  450. if (!vmem_map)
  451. memmap_init_zone(size, nid, zone, start_pfn, MEMMAP_EARLY);
  452. else {
  453. struct page *start;
  454. struct memmap_init_callback_data args;
  455. start = pfn_to_page(start_pfn);
  456. args.start = start;
  457. args.end = start + size;
  458. args.nid = nid;
  459. args.zone = zone;
  460. efi_memmap_walk(virtual_memmap_init, &args);
  461. }
  462. }
  463. int
  464. ia64_pfn_valid (unsigned long pfn)
  465. {
  466. char byte;
  467. struct page *pg = pfn_to_page(pfn);
  468. return (__get_user(byte, (char __user *) pg) == 0)
  469. && ((((u64)pg & PAGE_MASK) == (((u64)(pg + 1) - 1) & PAGE_MASK))
  470. || (__get_user(byte, (char __user *) (pg + 1) - 1) == 0));
  471. }
  472. EXPORT_SYMBOL(ia64_pfn_valid);
  473. int __init find_largest_hole(u64 start, u64 end, void *arg)
  474. {
  475. u64 *max_gap = arg;
  476. static u64 last_end = PAGE_OFFSET;
  477. /* NOTE: this algorithm assumes efi memmap table is ordered */
  478. if (*max_gap < (start - last_end))
  479. *max_gap = start - last_end;
  480. last_end = end;
  481. return 0;
  482. }
  483. #endif /* CONFIG_VIRTUAL_MEM_MAP */
  484. int __init register_active_ranges(u64 start, u64 len, int nid)
  485. {
  486. u64 end = start + len;
  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. memblock_add_node(__pa(start), end - start, nid);
  495. return 0;
  496. }
  497. int
  498. find_max_min_low_pfn (u64 start, u64 end, void *arg)
  499. {
  500. unsigned long pfn_start, pfn_end;
  501. #ifdef CONFIG_FLATMEM
  502. pfn_start = (PAGE_ALIGN(__pa(start))) >> PAGE_SHIFT;
  503. pfn_end = (PAGE_ALIGN(__pa(end - 1))) >> PAGE_SHIFT;
  504. #else
  505. pfn_start = GRANULEROUNDDOWN(__pa(start)) >> PAGE_SHIFT;
  506. pfn_end = GRANULEROUNDUP(__pa(end - 1)) >> PAGE_SHIFT;
  507. #endif
  508. min_low_pfn = min(min_low_pfn, pfn_start);
  509. max_low_pfn = max(max_low_pfn, pfn_end);
  510. return 0;
  511. }
  512. /*
  513. * Boot command-line option "nolwsys" can be used to disable the use of any light-weight
  514. * system call handler. When this option is in effect, all fsyscalls will end up bubbling
  515. * down into the kernel and calling the normal (heavy-weight) syscall handler. This is
  516. * useful for performance testing, but conceivably could also come in handy for debugging
  517. * purposes.
  518. */
  519. static int nolwsys __initdata;
  520. static int __init
  521. nolwsys_setup (char *s)
  522. {
  523. nolwsys = 1;
  524. return 1;
  525. }
  526. __setup("nolwsys", nolwsys_setup);
  527. void __init
  528. mem_init (void)
  529. {
  530. int i;
  531. BUG_ON(PTRS_PER_PGD * sizeof(pgd_t) != PAGE_SIZE);
  532. BUG_ON(PTRS_PER_PMD * sizeof(pmd_t) != PAGE_SIZE);
  533. BUG_ON(PTRS_PER_PTE * sizeof(pte_t) != PAGE_SIZE);
  534. #ifdef CONFIG_PCI
  535. /*
  536. * This needs to be called _after_ the command line has been parsed but _before_
  537. * any drivers that may need the PCI DMA interface are initialized or bootmem has
  538. * been freed.
  539. */
  540. platform_dma_init();
  541. #endif
  542. #ifdef CONFIG_FLATMEM
  543. BUG_ON(!mem_map);
  544. #endif
  545. set_max_mapnr(max_low_pfn);
  546. high_memory = __va(max_low_pfn * PAGE_SIZE);
  547. free_all_bootmem();
  548. mem_init_print_info(NULL);
  549. /*
  550. * For fsyscall entrpoints with no light-weight handler, use the ordinary
  551. * (heavy-weight) handler, but mark it by setting bit 0, so the fsyscall entry
  552. * code can tell them apart.
  553. */
  554. for (i = 0; i < NR_syscalls; ++i) {
  555. extern unsigned long sys_call_table[NR_syscalls];
  556. unsigned long *fsyscall_table = paravirt_get_fsyscall_table();
  557. if (!fsyscall_table[i] || nolwsys)
  558. fsyscall_table[i] = sys_call_table[i] | 1;
  559. }
  560. setup_gate();
  561. }
  562. #ifdef CONFIG_MEMORY_HOTPLUG
  563. int arch_add_memory(int nid, u64 start, u64 size)
  564. {
  565. pg_data_t *pgdat;
  566. struct zone *zone;
  567. unsigned long start_pfn = start >> PAGE_SHIFT;
  568. unsigned long nr_pages = size >> PAGE_SHIFT;
  569. int ret;
  570. pgdat = NODE_DATA(nid);
  571. zone = pgdat->node_zones +
  572. zone_for_memory(nid, start, size, ZONE_NORMAL);
  573. ret = __add_pages(nid, zone, start_pfn, nr_pages);
  574. if (ret)
  575. printk("%s: Problem encountered in __add_pages() as ret=%d\n",
  576. __func__, ret);
  577. return ret;
  578. }
  579. #ifdef CONFIG_MEMORY_HOTREMOVE
  580. int arch_remove_memory(u64 start, u64 size)
  581. {
  582. unsigned long start_pfn = start >> PAGE_SHIFT;
  583. unsigned long nr_pages = size >> PAGE_SHIFT;
  584. struct zone *zone;
  585. int ret;
  586. zone = page_zone(pfn_to_page(start_pfn));
  587. ret = __remove_pages(zone, start_pfn, nr_pages);
  588. if (ret)
  589. pr_warn("%s: Problem encountered in __remove_pages() as"
  590. " ret=%d\n", __func__, ret);
  591. return ret;
  592. }
  593. #endif
  594. #endif
  595. /**
  596. * show_mem - give short summary of memory stats
  597. *
  598. * Shows a simple page count of reserved and used pages in the system.
  599. * For discontig machines, it does this on a per-pgdat basis.
  600. */
  601. void show_mem(unsigned int filter)
  602. {
  603. int total_reserved = 0;
  604. unsigned long total_present = 0;
  605. pg_data_t *pgdat;
  606. printk(KERN_INFO "Mem-info:\n");
  607. show_free_areas(filter);
  608. printk(KERN_INFO "Node memory in pages:\n");
  609. for_each_online_pgdat(pgdat) {
  610. unsigned long present;
  611. unsigned long flags;
  612. int reserved = 0;
  613. int nid = pgdat->node_id;
  614. int zoneid;
  615. if (skip_free_areas_node(filter, nid))
  616. continue;
  617. pgdat_resize_lock(pgdat, &flags);
  618. for (zoneid = 0; zoneid < MAX_NR_ZONES; zoneid++) {
  619. struct zone *zone = &pgdat->node_zones[zoneid];
  620. if (!populated_zone(zone))
  621. continue;
  622. reserved += zone->present_pages - zone->managed_pages;
  623. }
  624. present = pgdat->node_present_pages;
  625. pgdat_resize_unlock(pgdat, &flags);
  626. total_present += present;
  627. total_reserved += reserved;
  628. printk(KERN_INFO "Node %4d: RAM: %11ld, rsvd: %8d, ",
  629. nid, present, reserved);
  630. }
  631. printk(KERN_INFO "%ld pages of RAM\n", total_present);
  632. printk(KERN_INFO "%d reserved pages\n", total_reserved);
  633. printk(KERN_INFO "Total of %ld pages in page table cache\n",
  634. quicklist_total_size());
  635. printk(KERN_INFO "%ld free buffer pages\n", nr_free_buffer_pages());
  636. }