fork.c 44 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844
  1. /*
  2. * linux/kernel/fork.c
  3. *
  4. * Copyright (C) 1991, 1992 Linus Torvalds
  5. */
  6. /*
  7. * 'fork.c' contains the help-routines for the 'fork' system call
  8. * (see also entry.S and others).
  9. * Fork is rather simple, once you get the hang of it, but the memory
  10. * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
  11. */
  12. #include <linux/slab.h>
  13. #include <linux/init.h>
  14. #include <linux/unistd.h>
  15. #include <linux/module.h>
  16. #include <linux/vmalloc.h>
  17. #include <linux/completion.h>
  18. #include <linux/personality.h>
  19. #include <linux/mempolicy.h>
  20. #include <linux/sem.h>
  21. #include <linux/file.h>
  22. #include <linux/fdtable.h>
  23. #include <linux/iocontext.h>
  24. #include <linux/key.h>
  25. #include <linux/binfmts.h>
  26. #include <linux/mman.h>
  27. #include <linux/mmu_notifier.h>
  28. #include <linux/fs.h>
  29. #include <linux/nsproxy.h>
  30. #include <linux/capability.h>
  31. #include <linux/cpu.h>
  32. #include <linux/cgroup.h>
  33. #include <linux/security.h>
  34. #include <linux/hugetlb.h>
  35. #include <linux/seccomp.h>
  36. #include <linux/swap.h>
  37. #include <linux/syscalls.h>
  38. #include <linux/jiffies.h>
  39. #include <linux/futex.h>
  40. #include <linux/compat.h>
  41. #include <linux/kthread.h>
  42. #include <linux/task_io_accounting_ops.h>
  43. #include <linux/rcupdate.h>
  44. #include <linux/ptrace.h>
  45. #include <linux/mount.h>
  46. #include <linux/audit.h>
  47. #include <linux/memcontrol.h>
  48. #include <linux/ftrace.h>
  49. #include <linux/profile.h>
  50. #include <linux/rmap.h>
  51. #include <linux/ksm.h>
  52. #include <linux/acct.h>
  53. #include <linux/tsacct_kern.h>
  54. #include <linux/cn_proc.h>
  55. #include <linux/freezer.h>
  56. #include <linux/delayacct.h>
  57. #include <linux/taskstats_kern.h>
  58. #include <linux/random.h>
  59. #include <linux/tty.h>
  60. #include <linux/blkdev.h>
  61. #include <linux/fs_struct.h>
  62. #include <linux/magic.h>
  63. #include <linux/perf_event.h>
  64. #include <linux/posix-timers.h>
  65. #include <linux/user-return-notifier.h>
  66. #include <linux/oom.h>
  67. #include <linux/khugepaged.h>
  68. #include <linux/signalfd.h>
  69. #include <asm/pgtable.h>
  70. #include <asm/pgalloc.h>
  71. #include <asm/uaccess.h>
  72. #include <asm/mmu_context.h>
  73. #include <asm/cacheflush.h>
  74. #include <asm/tlbflush.h>
  75. #include <trace/events/sched.h>
  76. #define CREATE_TRACE_POINTS
  77. #include <trace/events/task.h>
  78. /*
  79. * Protected counters by write_lock_irq(&tasklist_lock)
  80. */
  81. unsigned long total_forks; /* Handle normal Linux uptimes. */
  82. int nr_threads; /* The idle threads do not count.. */
  83. int max_threads; /* tunable limit on nr_threads */
  84. DEFINE_PER_CPU(unsigned long, process_counts) = 0;
  85. __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
  86. #ifdef CONFIG_PROVE_RCU
  87. int lockdep_tasklist_lock_is_held(void)
  88. {
  89. return lockdep_is_held(&tasklist_lock);
  90. }
  91. EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
  92. #endif /* #ifdef CONFIG_PROVE_RCU */
  93. int nr_processes(void)
  94. {
  95. int cpu;
  96. int total = 0;
  97. for_each_possible_cpu(cpu)
  98. total += per_cpu(process_counts, cpu);
  99. return total;
  100. }
  101. #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
  102. # define alloc_task_struct_node(node) \
  103. kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node)
  104. # define free_task_struct(tsk) \
  105. kmem_cache_free(task_struct_cachep, (tsk))
  106. static struct kmem_cache *task_struct_cachep;
  107. #endif
  108. #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
  109. static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
  110. int node)
  111. {
  112. #ifdef CONFIG_DEBUG_STACK_USAGE
  113. gfp_t mask = GFP_KERNEL | __GFP_ZERO;
  114. #else
  115. gfp_t mask = GFP_KERNEL;
  116. #endif
  117. struct page *page = alloc_pages_node(node, mask, THREAD_SIZE_ORDER);
  118. return page ? page_address(page) : NULL;
  119. }
  120. static inline void free_thread_info(struct thread_info *ti)
  121. {
  122. free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
  123. }
  124. #endif
  125. /* SLAB cache for signal_struct structures (tsk->signal) */
  126. static struct kmem_cache *signal_cachep;
  127. /* SLAB cache for sighand_struct structures (tsk->sighand) */
  128. struct kmem_cache *sighand_cachep;
  129. /* SLAB cache for files_struct structures (tsk->files) */
  130. struct kmem_cache *files_cachep;
  131. /* SLAB cache for fs_struct structures (tsk->fs) */
  132. struct kmem_cache *fs_cachep;
  133. /* SLAB cache for vm_area_struct structures */
  134. struct kmem_cache *vm_area_cachep;
  135. /* SLAB cache for mm_struct structures (tsk->mm) */
  136. static struct kmem_cache *mm_cachep;
  137. static void account_kernel_stack(struct thread_info *ti, int account)
  138. {
  139. struct zone *zone = page_zone(virt_to_page(ti));
  140. mod_zone_page_state(zone, NR_KERNEL_STACK, account);
  141. }
  142. void free_task(struct task_struct *tsk)
  143. {
  144. account_kernel_stack(tsk->stack, -1);
  145. free_thread_info(tsk->stack);
  146. rt_mutex_debug_task_free(tsk);
  147. ftrace_graph_exit_task(tsk);
  148. put_seccomp_filter(tsk);
  149. free_task_struct(tsk);
  150. }
  151. EXPORT_SYMBOL(free_task);
  152. static inline void free_signal_struct(struct signal_struct *sig)
  153. {
  154. taskstats_tgid_free(sig);
  155. sched_autogroup_exit(sig);
  156. kmem_cache_free(signal_cachep, sig);
  157. }
  158. static inline void put_signal_struct(struct signal_struct *sig)
  159. {
  160. if (atomic_dec_and_test(&sig->sigcnt))
  161. free_signal_struct(sig);
  162. }
  163. void __put_task_struct(struct task_struct *tsk)
  164. {
  165. WARN_ON(!tsk->exit_state);
  166. WARN_ON(atomic_read(&tsk->usage));
  167. WARN_ON(tsk == current);
  168. security_task_free(tsk);
  169. exit_creds(tsk);
  170. delayacct_tsk_free(tsk);
  171. put_signal_struct(tsk->signal);
  172. if (!profile_handoff_task(tsk))
  173. free_task(tsk);
  174. }
  175. EXPORT_SYMBOL_GPL(__put_task_struct);
  176. /*
  177. * macro override instead of weak attribute alias, to workaround
  178. * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
  179. */
  180. #ifndef arch_task_cache_init
  181. #define arch_task_cache_init()
  182. #endif
  183. void __init fork_init(unsigned long mempages)
  184. {
  185. #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
  186. #ifndef ARCH_MIN_TASKALIGN
  187. #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
  188. #endif
  189. /* create a slab on which task_structs can be allocated */
  190. task_struct_cachep =
  191. kmem_cache_create("task_struct", sizeof(struct task_struct),
  192. ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
  193. #endif
  194. /* do the arch specific task caches init */
  195. arch_task_cache_init();
  196. /*
  197. * The default maximum number of threads is set to a safe
  198. * value: the thread structures can take up at most half
  199. * of memory.
  200. */
  201. max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
  202. /*
  203. * we need to allow at least 20 threads to boot a system
  204. */
  205. if (max_threads < 20)
  206. max_threads = 20;
  207. init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
  208. init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
  209. init_task.signal->rlim[RLIMIT_SIGPENDING] =
  210. init_task.signal->rlim[RLIMIT_NPROC];
  211. }
  212. int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
  213. struct task_struct *src)
  214. {
  215. *dst = *src;
  216. return 0;
  217. }
  218. static struct task_struct *dup_task_struct(struct task_struct *orig)
  219. {
  220. struct task_struct *tsk;
  221. struct thread_info *ti;
  222. unsigned long *stackend;
  223. int node = tsk_fork_get_node(orig);
  224. int err;
  225. prepare_to_copy(orig);
  226. tsk = alloc_task_struct_node(node);
  227. if (!tsk)
  228. return NULL;
  229. ti = alloc_thread_info_node(tsk, node);
  230. if (!ti) {
  231. free_task_struct(tsk);
  232. return NULL;
  233. }
  234. err = arch_dup_task_struct(tsk, orig);
  235. if (err)
  236. goto out;
  237. tsk->stack = ti;
  238. setup_thread_stack(tsk, orig);
  239. clear_user_return_notifier(tsk);
  240. clear_tsk_need_resched(tsk);
  241. stackend = end_of_stack(tsk);
  242. *stackend = STACK_END_MAGIC; /* for overflow detection */
  243. #ifdef CONFIG_CC_STACKPROTECTOR
  244. tsk->stack_canary = get_random_int();
  245. #endif
  246. /*
  247. * One for us, one for whoever does the "release_task()" (usually
  248. * parent)
  249. */
  250. atomic_set(&tsk->usage, 2);
  251. #ifdef CONFIG_BLK_DEV_IO_TRACE
  252. tsk->btrace_seq = 0;
  253. #endif
  254. tsk->splice_pipe = NULL;
  255. account_kernel_stack(ti, 1);
  256. return tsk;
  257. out:
  258. free_thread_info(ti);
  259. free_task_struct(tsk);
  260. return NULL;
  261. }
  262. #ifdef CONFIG_MMU
  263. static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
  264. {
  265. struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
  266. struct rb_node **rb_link, *rb_parent;
  267. int retval;
  268. unsigned long charge;
  269. struct mempolicy *pol;
  270. down_write(&oldmm->mmap_sem);
  271. flush_cache_dup_mm(oldmm);
  272. /*
  273. * Not linked in yet - no deadlock potential:
  274. */
  275. down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
  276. mm->locked_vm = 0;
  277. mm->mmap = NULL;
  278. mm->mmap_cache = NULL;
  279. mm->free_area_cache = oldmm->mmap_base;
  280. mm->cached_hole_size = ~0UL;
  281. mm->map_count = 0;
  282. cpumask_clear(mm_cpumask(mm));
  283. mm->mm_rb = RB_ROOT;
  284. rb_link = &mm->mm_rb.rb_node;
  285. rb_parent = NULL;
  286. pprev = &mm->mmap;
  287. retval = ksm_fork(mm, oldmm);
  288. if (retval)
  289. goto out;
  290. retval = khugepaged_fork(mm, oldmm);
  291. if (retval)
  292. goto out;
  293. prev = NULL;
  294. for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
  295. struct file *file;
  296. if (mpnt->vm_flags & VM_DONTCOPY) {
  297. long pages = vma_pages(mpnt);
  298. mm->total_vm -= pages;
  299. vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
  300. -pages);
  301. continue;
  302. }
  303. charge = 0;
  304. if (mpnt->vm_flags & VM_ACCOUNT) {
  305. unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
  306. if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
  307. goto fail_nomem;
  308. charge = len;
  309. }
  310. tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
  311. if (!tmp)
  312. goto fail_nomem;
  313. *tmp = *mpnt;
  314. INIT_LIST_HEAD(&tmp->anon_vma_chain);
  315. pol = mpol_dup(vma_policy(mpnt));
  316. retval = PTR_ERR(pol);
  317. if (IS_ERR(pol))
  318. goto fail_nomem_policy;
  319. vma_set_policy(tmp, pol);
  320. tmp->vm_mm = mm;
  321. if (anon_vma_fork(tmp, mpnt))
  322. goto fail_nomem_anon_vma_fork;
  323. tmp->vm_flags &= ~VM_LOCKED;
  324. tmp->vm_next = tmp->vm_prev = NULL;
  325. file = tmp->vm_file;
  326. if (file) {
  327. struct inode *inode = file->f_path.dentry->d_inode;
  328. struct address_space *mapping = file->f_mapping;
  329. get_file(file);
  330. if (tmp->vm_flags & VM_DENYWRITE)
  331. atomic_dec(&inode->i_writecount);
  332. mutex_lock(&mapping->i_mmap_mutex);
  333. if (tmp->vm_flags & VM_SHARED)
  334. mapping->i_mmap_writable++;
  335. flush_dcache_mmap_lock(mapping);
  336. /* insert tmp into the share list, just after mpnt */
  337. vma_prio_tree_add(tmp, mpnt);
  338. flush_dcache_mmap_unlock(mapping);
  339. mutex_unlock(&mapping->i_mmap_mutex);
  340. }
  341. /*
  342. * Clear hugetlb-related page reserves for children. This only
  343. * affects MAP_PRIVATE mappings. Faults generated by the child
  344. * are not guaranteed to succeed, even if read-only
  345. */
  346. if (is_vm_hugetlb_page(tmp))
  347. reset_vma_resv_huge_pages(tmp);
  348. /*
  349. * Link in the new vma and copy the page table entries.
  350. */
  351. *pprev = tmp;
  352. pprev = &tmp->vm_next;
  353. tmp->vm_prev = prev;
  354. prev = tmp;
  355. __vma_link_rb(mm, tmp, rb_link, rb_parent);
  356. rb_link = &tmp->vm_rb.rb_right;
  357. rb_parent = &tmp->vm_rb;
  358. mm->map_count++;
  359. retval = copy_page_range(mm, oldmm, mpnt);
  360. if (tmp->vm_ops && tmp->vm_ops->open)
  361. tmp->vm_ops->open(tmp);
  362. if (retval)
  363. goto out;
  364. }
  365. /* a new mm has just been created */
  366. arch_dup_mmap(oldmm, mm);
  367. retval = 0;
  368. out:
  369. up_write(&mm->mmap_sem);
  370. flush_tlb_mm(oldmm);
  371. up_write(&oldmm->mmap_sem);
  372. return retval;
  373. fail_nomem_anon_vma_fork:
  374. mpol_put(pol);
  375. fail_nomem_policy:
  376. kmem_cache_free(vm_area_cachep, tmp);
  377. fail_nomem:
  378. retval = -ENOMEM;
  379. vm_unacct_memory(charge);
  380. goto out;
  381. }
  382. static inline int mm_alloc_pgd(struct mm_struct *mm)
  383. {
  384. mm->pgd = pgd_alloc(mm);
  385. if (unlikely(!mm->pgd))
  386. return -ENOMEM;
  387. return 0;
  388. }
  389. static inline void mm_free_pgd(struct mm_struct *mm)
  390. {
  391. pgd_free(mm, mm->pgd);
  392. }
  393. #else
  394. #define dup_mmap(mm, oldmm) (0)
  395. #define mm_alloc_pgd(mm) (0)
  396. #define mm_free_pgd(mm)
  397. #endif /* CONFIG_MMU */
  398. __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
  399. #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
  400. #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
  401. static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
  402. static int __init coredump_filter_setup(char *s)
  403. {
  404. default_dump_filter =
  405. (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
  406. MMF_DUMP_FILTER_MASK;
  407. return 1;
  408. }
  409. __setup("coredump_filter=", coredump_filter_setup);
  410. #include <linux/init_task.h>
  411. static void mm_init_aio(struct mm_struct *mm)
  412. {
  413. #ifdef CONFIG_AIO
  414. spin_lock_init(&mm->ioctx_lock);
  415. INIT_HLIST_HEAD(&mm->ioctx_list);
  416. #endif
  417. }
  418. static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
  419. {
  420. atomic_set(&mm->mm_users, 1);
  421. atomic_set(&mm->mm_count, 1);
  422. init_rwsem(&mm->mmap_sem);
  423. INIT_LIST_HEAD(&mm->mmlist);
  424. mm->flags = (current->mm) ?
  425. (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
  426. mm->core_state = NULL;
  427. mm->nr_ptes = 0;
  428. memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
  429. spin_lock_init(&mm->page_table_lock);
  430. mm->free_area_cache = TASK_UNMAPPED_BASE;
  431. mm->cached_hole_size = ~0UL;
  432. mm_init_aio(mm);
  433. mm_init_owner(mm, p);
  434. if (likely(!mm_alloc_pgd(mm))) {
  435. mm->def_flags = 0;
  436. mmu_notifier_mm_init(mm);
  437. return mm;
  438. }
  439. free_mm(mm);
  440. return NULL;
  441. }
  442. static void check_mm(struct mm_struct *mm)
  443. {
  444. int i;
  445. for (i = 0; i < NR_MM_COUNTERS; i++) {
  446. long x = atomic_long_read(&mm->rss_stat.count[i]);
  447. if (unlikely(x))
  448. printk(KERN_ALERT "BUG: Bad rss-counter state "
  449. "mm:%p idx:%d val:%ld\n", mm, i, x);
  450. }
  451. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  452. VM_BUG_ON(mm->pmd_huge_pte);
  453. #endif
  454. }
  455. /*
  456. * Allocate and initialize an mm_struct.
  457. */
  458. struct mm_struct *mm_alloc(void)
  459. {
  460. struct mm_struct *mm;
  461. mm = allocate_mm();
  462. if (!mm)
  463. return NULL;
  464. memset(mm, 0, sizeof(*mm));
  465. mm_init_cpumask(mm);
  466. return mm_init(mm, current);
  467. }
  468. /*
  469. * Called when the last reference to the mm
  470. * is dropped: either by a lazy thread or by
  471. * mmput. Free the page directory and the mm.
  472. */
  473. void __mmdrop(struct mm_struct *mm)
  474. {
  475. BUG_ON(mm == &init_mm);
  476. mm_free_pgd(mm);
  477. destroy_context(mm);
  478. mmu_notifier_mm_destroy(mm);
  479. check_mm(mm);
  480. free_mm(mm);
  481. }
  482. EXPORT_SYMBOL_GPL(__mmdrop);
  483. /*
  484. * Decrement the use count and release all resources for an mm.
  485. */
  486. void mmput(struct mm_struct *mm)
  487. {
  488. might_sleep();
  489. if (atomic_dec_and_test(&mm->mm_users)) {
  490. exit_aio(mm);
  491. ksm_exit(mm);
  492. khugepaged_exit(mm); /* must run before exit_mmap */
  493. exit_mmap(mm);
  494. set_mm_exe_file(mm, NULL);
  495. if (!list_empty(&mm->mmlist)) {
  496. spin_lock(&mmlist_lock);
  497. list_del(&mm->mmlist);
  498. spin_unlock(&mmlist_lock);
  499. }
  500. put_swap_token(mm);
  501. if (mm->binfmt)
  502. module_put(mm->binfmt->module);
  503. mmdrop(mm);
  504. }
  505. }
  506. EXPORT_SYMBOL_GPL(mmput);
  507. /*
  508. * We added or removed a vma mapping the executable. The vmas are only mapped
  509. * during exec and are not mapped with the mmap system call.
  510. * Callers must hold down_write() on the mm's mmap_sem for these
  511. */
  512. void added_exe_file_vma(struct mm_struct *mm)
  513. {
  514. mm->num_exe_file_vmas++;
  515. }
  516. void removed_exe_file_vma(struct mm_struct *mm)
  517. {
  518. mm->num_exe_file_vmas--;
  519. if ((mm->num_exe_file_vmas == 0) && mm->exe_file) {
  520. fput(mm->exe_file);
  521. mm->exe_file = NULL;
  522. }
  523. }
  524. void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
  525. {
  526. if (new_exe_file)
  527. get_file(new_exe_file);
  528. if (mm->exe_file)
  529. fput(mm->exe_file);
  530. mm->exe_file = new_exe_file;
  531. mm->num_exe_file_vmas = 0;
  532. }
  533. struct file *get_mm_exe_file(struct mm_struct *mm)
  534. {
  535. struct file *exe_file;
  536. /* We need mmap_sem to protect against races with removal of
  537. * VM_EXECUTABLE vmas */
  538. down_read(&mm->mmap_sem);
  539. exe_file = mm->exe_file;
  540. if (exe_file)
  541. get_file(exe_file);
  542. up_read(&mm->mmap_sem);
  543. return exe_file;
  544. }
  545. static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
  546. {
  547. /* It's safe to write the exe_file pointer without exe_file_lock because
  548. * this is called during fork when the task is not yet in /proc */
  549. newmm->exe_file = get_mm_exe_file(oldmm);
  550. }
  551. /**
  552. * get_task_mm - acquire a reference to the task's mm
  553. *
  554. * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
  555. * this kernel workthread has transiently adopted a user mm with use_mm,
  556. * to do its AIO) is not set and if so returns a reference to it, after
  557. * bumping up the use count. User must release the mm via mmput()
  558. * after use. Typically used by /proc and ptrace.
  559. */
  560. struct mm_struct *get_task_mm(struct task_struct *task)
  561. {
  562. struct mm_struct *mm;
  563. task_lock(task);
  564. mm = task->mm;
  565. if (mm) {
  566. if (task->flags & PF_KTHREAD)
  567. mm = NULL;
  568. else
  569. atomic_inc(&mm->mm_users);
  570. }
  571. task_unlock(task);
  572. return mm;
  573. }
  574. EXPORT_SYMBOL_GPL(get_task_mm);
  575. struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
  576. {
  577. struct mm_struct *mm;
  578. int err;
  579. err = mutex_lock_killable(&task->signal->cred_guard_mutex);
  580. if (err)
  581. return ERR_PTR(err);
  582. mm = get_task_mm(task);
  583. if (mm && mm != current->mm &&
  584. !ptrace_may_access(task, mode)) {
  585. mmput(mm);
  586. mm = ERR_PTR(-EACCES);
  587. }
  588. mutex_unlock(&task->signal->cred_guard_mutex);
  589. return mm;
  590. }
  591. static void complete_vfork_done(struct task_struct *tsk)
  592. {
  593. struct completion *vfork;
  594. task_lock(tsk);
  595. vfork = tsk->vfork_done;
  596. if (likely(vfork)) {
  597. tsk->vfork_done = NULL;
  598. complete(vfork);
  599. }
  600. task_unlock(tsk);
  601. }
  602. static int wait_for_vfork_done(struct task_struct *child,
  603. struct completion *vfork)
  604. {
  605. int killed;
  606. freezer_do_not_count();
  607. killed = wait_for_completion_killable(vfork);
  608. freezer_count();
  609. if (killed) {
  610. task_lock(child);
  611. child->vfork_done = NULL;
  612. task_unlock(child);
  613. }
  614. put_task_struct(child);
  615. return killed;
  616. }
  617. /* Please note the differences between mmput and mm_release.
  618. * mmput is called whenever we stop holding onto a mm_struct,
  619. * error success whatever.
  620. *
  621. * mm_release is called after a mm_struct has been removed
  622. * from the current process.
  623. *
  624. * This difference is important for error handling, when we
  625. * only half set up a mm_struct for a new process and need to restore
  626. * the old one. Because we mmput the new mm_struct before
  627. * restoring the old one. . .
  628. * Eric Biederman 10 January 1998
  629. */
  630. void mm_release(struct task_struct *tsk, struct mm_struct *mm)
  631. {
  632. /* Get rid of any futexes when releasing the mm */
  633. #ifdef CONFIG_FUTEX
  634. if (unlikely(tsk->robust_list)) {
  635. exit_robust_list(tsk);
  636. tsk->robust_list = NULL;
  637. }
  638. #ifdef CONFIG_COMPAT
  639. if (unlikely(tsk->compat_robust_list)) {
  640. compat_exit_robust_list(tsk);
  641. tsk->compat_robust_list = NULL;
  642. }
  643. #endif
  644. if (unlikely(!list_empty(&tsk->pi_state_list)))
  645. exit_pi_state_list(tsk);
  646. #endif
  647. /* Get rid of any cached register state */
  648. deactivate_mm(tsk, mm);
  649. if (tsk->vfork_done)
  650. complete_vfork_done(tsk);
  651. /*
  652. * If we're exiting normally, clear a user-space tid field if
  653. * requested. We leave this alone when dying by signal, to leave
  654. * the value intact in a core dump, and to save the unnecessary
  655. * trouble, say, a killed vfork parent shouldn't touch this mm.
  656. * Userland only wants this done for a sys_exit.
  657. */
  658. if (tsk->clear_child_tid) {
  659. if (!(tsk->flags & PF_SIGNALED) &&
  660. atomic_read(&mm->mm_users) > 1) {
  661. /*
  662. * We don't check the error code - if userspace has
  663. * not set up a proper pointer then tough luck.
  664. */
  665. put_user(0, tsk->clear_child_tid);
  666. sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
  667. 1, NULL, NULL, 0);
  668. }
  669. tsk->clear_child_tid = NULL;
  670. }
  671. }
  672. /*
  673. * Allocate a new mm structure and copy contents from the
  674. * mm structure of the passed in task structure.
  675. */
  676. struct mm_struct *dup_mm(struct task_struct *tsk)
  677. {
  678. struct mm_struct *mm, *oldmm = current->mm;
  679. int err;
  680. if (!oldmm)
  681. return NULL;
  682. mm = allocate_mm();
  683. if (!mm)
  684. goto fail_nomem;
  685. memcpy(mm, oldmm, sizeof(*mm));
  686. mm_init_cpumask(mm);
  687. /* Initializing for Swap token stuff */
  688. mm->token_priority = 0;
  689. mm->last_interval = 0;
  690. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  691. mm->pmd_huge_pte = NULL;
  692. #endif
  693. if (!mm_init(mm, tsk))
  694. goto fail_nomem;
  695. if (init_new_context(tsk, mm))
  696. goto fail_nocontext;
  697. dup_mm_exe_file(oldmm, mm);
  698. err = dup_mmap(mm, oldmm);
  699. if (err)
  700. goto free_pt;
  701. mm->hiwater_rss = get_mm_rss(mm);
  702. mm->hiwater_vm = mm->total_vm;
  703. if (mm->binfmt && !try_module_get(mm->binfmt->module))
  704. goto free_pt;
  705. return mm;
  706. free_pt:
  707. /* don't put binfmt in mmput, we haven't got module yet */
  708. mm->binfmt = NULL;
  709. mmput(mm);
  710. fail_nomem:
  711. return NULL;
  712. fail_nocontext:
  713. /*
  714. * If init_new_context() failed, we cannot use mmput() to free the mm
  715. * because it calls destroy_context()
  716. */
  717. mm_free_pgd(mm);
  718. free_mm(mm);
  719. return NULL;
  720. }
  721. static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
  722. {
  723. struct mm_struct *mm, *oldmm;
  724. int retval;
  725. tsk->min_flt = tsk->maj_flt = 0;
  726. tsk->nvcsw = tsk->nivcsw = 0;
  727. #ifdef CONFIG_DETECT_HUNG_TASK
  728. tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
  729. #endif
  730. tsk->mm = NULL;
  731. tsk->active_mm = NULL;
  732. /*
  733. * Are we cloning a kernel thread?
  734. *
  735. * We need to steal a active VM for that..
  736. */
  737. oldmm = current->mm;
  738. if (!oldmm)
  739. return 0;
  740. if (clone_flags & CLONE_VM) {
  741. atomic_inc(&oldmm->mm_users);
  742. mm = oldmm;
  743. goto good_mm;
  744. }
  745. retval = -ENOMEM;
  746. mm = dup_mm(tsk);
  747. if (!mm)
  748. goto fail_nomem;
  749. good_mm:
  750. /* Initializing for Swap token stuff */
  751. mm->token_priority = 0;
  752. mm->last_interval = 0;
  753. tsk->mm = mm;
  754. tsk->active_mm = mm;
  755. return 0;
  756. fail_nomem:
  757. return retval;
  758. }
  759. static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
  760. {
  761. struct fs_struct *fs = current->fs;
  762. if (clone_flags & CLONE_FS) {
  763. /* tsk->fs is already what we want */
  764. spin_lock(&fs->lock);
  765. if (fs->in_exec) {
  766. spin_unlock(&fs->lock);
  767. return -EAGAIN;
  768. }
  769. fs->users++;
  770. spin_unlock(&fs->lock);
  771. return 0;
  772. }
  773. tsk->fs = copy_fs_struct(fs);
  774. if (!tsk->fs)
  775. return -ENOMEM;
  776. return 0;
  777. }
  778. static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
  779. {
  780. struct files_struct *oldf, *newf;
  781. int error = 0;
  782. /*
  783. * A background process may not have any files ...
  784. */
  785. oldf = current->files;
  786. if (!oldf)
  787. goto out;
  788. if (clone_flags & CLONE_FILES) {
  789. atomic_inc(&oldf->count);
  790. goto out;
  791. }
  792. newf = dup_fd(oldf, &error);
  793. if (!newf)
  794. goto out;
  795. tsk->files = newf;
  796. error = 0;
  797. out:
  798. return error;
  799. }
  800. static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
  801. {
  802. #ifdef CONFIG_BLOCK
  803. struct io_context *ioc = current->io_context;
  804. struct io_context *new_ioc;
  805. if (!ioc)
  806. return 0;
  807. /*
  808. * Share io context with parent, if CLONE_IO is set
  809. */
  810. if (clone_flags & CLONE_IO) {
  811. tsk->io_context = ioc_task_link(ioc);
  812. if (unlikely(!tsk->io_context))
  813. return -ENOMEM;
  814. } else if (ioprio_valid(ioc->ioprio)) {
  815. new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
  816. if (unlikely(!new_ioc))
  817. return -ENOMEM;
  818. new_ioc->ioprio = ioc->ioprio;
  819. put_io_context(new_ioc);
  820. }
  821. #endif
  822. return 0;
  823. }
  824. static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
  825. {
  826. struct sighand_struct *sig;
  827. if (clone_flags & CLONE_SIGHAND) {
  828. atomic_inc(&current->sighand->count);
  829. return 0;
  830. }
  831. sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
  832. rcu_assign_pointer(tsk->sighand, sig);
  833. if (!sig)
  834. return -ENOMEM;
  835. atomic_set(&sig->count, 1);
  836. memcpy(sig->action, current->sighand->action, sizeof(sig->action));
  837. return 0;
  838. }
  839. void __cleanup_sighand(struct sighand_struct *sighand)
  840. {
  841. if (atomic_dec_and_test(&sighand->count)) {
  842. signalfd_cleanup(sighand);
  843. kmem_cache_free(sighand_cachep, sighand);
  844. }
  845. }
  846. /*
  847. * Initialize POSIX timer handling for a thread group.
  848. */
  849. static void posix_cpu_timers_init_group(struct signal_struct *sig)
  850. {
  851. unsigned long cpu_limit;
  852. /* Thread group counters. */
  853. thread_group_cputime_init(sig);
  854. cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
  855. if (cpu_limit != RLIM_INFINITY) {
  856. sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
  857. sig->cputimer.running = 1;
  858. }
  859. /* The timer lists. */
  860. INIT_LIST_HEAD(&sig->cpu_timers[0]);
  861. INIT_LIST_HEAD(&sig->cpu_timers[1]);
  862. INIT_LIST_HEAD(&sig->cpu_timers[2]);
  863. }
  864. static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
  865. {
  866. struct signal_struct *sig;
  867. if (clone_flags & CLONE_THREAD)
  868. return 0;
  869. sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
  870. tsk->signal = sig;
  871. if (!sig)
  872. return -ENOMEM;
  873. sig->nr_threads = 1;
  874. atomic_set(&sig->live, 1);
  875. atomic_set(&sig->sigcnt, 1);
  876. init_waitqueue_head(&sig->wait_chldexit);
  877. if (clone_flags & CLONE_NEWPID)
  878. sig->flags |= SIGNAL_UNKILLABLE;
  879. sig->curr_target = tsk;
  880. init_sigpending(&sig->shared_pending);
  881. INIT_LIST_HEAD(&sig->posix_timers);
  882. hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
  883. sig->real_timer.function = it_real_fn;
  884. task_lock(current->group_leader);
  885. memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
  886. task_unlock(current->group_leader);
  887. posix_cpu_timers_init_group(sig);
  888. tty_audit_fork(sig);
  889. sched_autogroup_fork(sig);
  890. #ifdef CONFIG_CGROUPS
  891. init_rwsem(&sig->group_rwsem);
  892. #endif
  893. sig->oom_adj = current->signal->oom_adj;
  894. sig->oom_score_adj = current->signal->oom_score_adj;
  895. sig->oom_score_adj_min = current->signal->oom_score_adj_min;
  896. sig->has_child_subreaper = current->signal->has_child_subreaper ||
  897. current->signal->is_child_subreaper;
  898. mutex_init(&sig->cred_guard_mutex);
  899. return 0;
  900. }
  901. static void copy_flags(unsigned long clone_flags, struct task_struct *p)
  902. {
  903. unsigned long new_flags = p->flags;
  904. new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
  905. new_flags |= PF_FORKNOEXEC;
  906. p->flags = new_flags;
  907. }
  908. SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
  909. {
  910. current->clear_child_tid = tidptr;
  911. return task_pid_vnr(current);
  912. }
  913. static void rt_mutex_init_task(struct task_struct *p)
  914. {
  915. raw_spin_lock_init(&p->pi_lock);
  916. #ifdef CONFIG_RT_MUTEXES
  917. plist_head_init(&p->pi_waiters);
  918. p->pi_blocked_on = NULL;
  919. #endif
  920. }
  921. #ifdef CONFIG_MM_OWNER
  922. void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
  923. {
  924. mm->owner = p;
  925. }
  926. #endif /* CONFIG_MM_OWNER */
  927. /*
  928. * Initialize POSIX timer handling for a single task.
  929. */
  930. static void posix_cpu_timers_init(struct task_struct *tsk)
  931. {
  932. tsk->cputime_expires.prof_exp = 0;
  933. tsk->cputime_expires.virt_exp = 0;
  934. tsk->cputime_expires.sched_exp = 0;
  935. INIT_LIST_HEAD(&tsk->cpu_timers[0]);
  936. INIT_LIST_HEAD(&tsk->cpu_timers[1]);
  937. INIT_LIST_HEAD(&tsk->cpu_timers[2]);
  938. }
  939. /*
  940. * This creates a new process as a copy of the old one,
  941. * but does not actually start it yet.
  942. *
  943. * It copies the registers, and all the appropriate
  944. * parts of the process environment (as per the clone
  945. * flags). The actual kick-off is left to the caller.
  946. */
  947. static struct task_struct *copy_process(unsigned long clone_flags,
  948. unsigned long stack_start,
  949. struct pt_regs *regs,
  950. unsigned long stack_size,
  951. int __user *child_tidptr,
  952. struct pid *pid,
  953. int trace)
  954. {
  955. int retval;
  956. struct task_struct *p;
  957. int cgroup_callbacks_done = 0;
  958. if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
  959. return ERR_PTR(-EINVAL);
  960. /*
  961. * Thread groups must share signals as well, and detached threads
  962. * can only be started up within the thread group.
  963. */
  964. if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
  965. return ERR_PTR(-EINVAL);
  966. /*
  967. * Shared signal handlers imply shared VM. By way of the above,
  968. * thread groups also imply shared VM. Blocking this case allows
  969. * for various simplifications in other code.
  970. */
  971. if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
  972. return ERR_PTR(-EINVAL);
  973. /*
  974. * Siblings of global init remain as zombies on exit since they are
  975. * not reaped by their parent (swapper). To solve this and to avoid
  976. * multi-rooted process trees, prevent global and container-inits
  977. * from creating siblings.
  978. */
  979. if ((clone_flags & CLONE_PARENT) &&
  980. current->signal->flags & SIGNAL_UNKILLABLE)
  981. return ERR_PTR(-EINVAL);
  982. retval = security_task_create(clone_flags);
  983. if (retval)
  984. goto fork_out;
  985. retval = -ENOMEM;
  986. p = dup_task_struct(current);
  987. if (!p)
  988. goto fork_out;
  989. ftrace_graph_init_task(p);
  990. get_seccomp_filter(p);
  991. rt_mutex_init_task(p);
  992. #ifdef CONFIG_PROVE_LOCKING
  993. DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
  994. DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
  995. #endif
  996. retval = -EAGAIN;
  997. if (atomic_read(&p->real_cred->user->processes) >=
  998. task_rlimit(p, RLIMIT_NPROC)) {
  999. if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
  1000. p->real_cred->user != INIT_USER)
  1001. goto bad_fork_free;
  1002. }
  1003. current->flags &= ~PF_NPROC_EXCEEDED;
  1004. retval = copy_creds(p, clone_flags);
  1005. if (retval < 0)
  1006. goto bad_fork_free;
  1007. /*
  1008. * If multiple threads are within copy_process(), then this check
  1009. * triggers too late. This doesn't hurt, the check is only there
  1010. * to stop root fork bombs.
  1011. */
  1012. retval = -EAGAIN;
  1013. if (nr_threads >= max_threads)
  1014. goto bad_fork_cleanup_count;
  1015. if (!try_module_get(task_thread_info(p)->exec_domain->module))
  1016. goto bad_fork_cleanup_count;
  1017. p->did_exec = 0;
  1018. delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
  1019. copy_flags(clone_flags, p);
  1020. INIT_LIST_HEAD(&p->children);
  1021. INIT_LIST_HEAD(&p->sibling);
  1022. rcu_copy_process(p);
  1023. p->vfork_done = NULL;
  1024. spin_lock_init(&p->alloc_lock);
  1025. init_sigpending(&p->pending);
  1026. p->utime = p->stime = p->gtime = 0;
  1027. p->utimescaled = p->stimescaled = 0;
  1028. #ifndef CONFIG_VIRT_CPU_ACCOUNTING
  1029. p->prev_utime = p->prev_stime = 0;
  1030. #endif
  1031. #if defined(SPLIT_RSS_COUNTING)
  1032. memset(&p->rss_stat, 0, sizeof(p->rss_stat));
  1033. #endif
  1034. p->default_timer_slack_ns = current->timer_slack_ns;
  1035. task_io_accounting_init(&p->ioac);
  1036. acct_clear_integrals(p);
  1037. posix_cpu_timers_init(p);
  1038. do_posix_clock_monotonic_gettime(&p->start_time);
  1039. p->real_start_time = p->start_time;
  1040. monotonic_to_bootbased(&p->real_start_time);
  1041. p->io_context = NULL;
  1042. p->audit_context = NULL;
  1043. if (clone_flags & CLONE_THREAD)
  1044. threadgroup_change_begin(current);
  1045. cgroup_fork(p);
  1046. #ifdef CONFIG_NUMA
  1047. p->mempolicy = mpol_dup(p->mempolicy);
  1048. if (IS_ERR(p->mempolicy)) {
  1049. retval = PTR_ERR(p->mempolicy);
  1050. p->mempolicy = NULL;
  1051. goto bad_fork_cleanup_cgroup;
  1052. }
  1053. mpol_fix_fork_child_flag(p);
  1054. #endif
  1055. #ifdef CONFIG_CPUSETS
  1056. p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
  1057. p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
  1058. seqcount_init(&p->mems_allowed_seq);
  1059. #endif
  1060. #ifdef CONFIG_TRACE_IRQFLAGS
  1061. p->irq_events = 0;
  1062. #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
  1063. p->hardirqs_enabled = 1;
  1064. #else
  1065. p->hardirqs_enabled = 0;
  1066. #endif
  1067. p->hardirq_enable_ip = 0;
  1068. p->hardirq_enable_event = 0;
  1069. p->hardirq_disable_ip = _THIS_IP_;
  1070. p->hardirq_disable_event = 0;
  1071. p->softirqs_enabled = 1;
  1072. p->softirq_enable_ip = _THIS_IP_;
  1073. p->softirq_enable_event = 0;
  1074. p->softirq_disable_ip = 0;
  1075. p->softirq_disable_event = 0;
  1076. p->hardirq_context = 0;
  1077. p->softirq_context = 0;
  1078. #endif
  1079. #ifdef CONFIG_LOCKDEP
  1080. p->lockdep_depth = 0; /* no locks held yet */
  1081. p->curr_chain_key = 0;
  1082. p->lockdep_recursion = 0;
  1083. #endif
  1084. #ifdef CONFIG_DEBUG_MUTEXES
  1085. p->blocked_on = NULL; /* not blocked yet */
  1086. #endif
  1087. #ifdef CONFIG_CGROUP_MEM_RES_CTLR
  1088. p->memcg_batch.do_batch = 0;
  1089. p->memcg_batch.memcg = NULL;
  1090. #endif
  1091. /* Perform scheduler related setup. Assign this task to a CPU. */
  1092. sched_fork(p);
  1093. retval = perf_event_init_task(p);
  1094. if (retval)
  1095. goto bad_fork_cleanup_policy;
  1096. retval = audit_alloc(p);
  1097. if (retval)
  1098. goto bad_fork_cleanup_policy;
  1099. /* copy all the process information */
  1100. retval = copy_semundo(clone_flags, p);
  1101. if (retval)
  1102. goto bad_fork_cleanup_audit;
  1103. retval = copy_files(clone_flags, p);
  1104. if (retval)
  1105. goto bad_fork_cleanup_semundo;
  1106. retval = copy_fs(clone_flags, p);
  1107. if (retval)
  1108. goto bad_fork_cleanup_files;
  1109. retval = copy_sighand(clone_flags, p);
  1110. if (retval)
  1111. goto bad_fork_cleanup_fs;
  1112. retval = copy_signal(clone_flags, p);
  1113. if (retval)
  1114. goto bad_fork_cleanup_sighand;
  1115. retval = copy_mm(clone_flags, p);
  1116. if (retval)
  1117. goto bad_fork_cleanup_signal;
  1118. retval = copy_namespaces(clone_flags, p);
  1119. if (retval)
  1120. goto bad_fork_cleanup_mm;
  1121. retval = copy_io(clone_flags, p);
  1122. if (retval)
  1123. goto bad_fork_cleanup_namespaces;
  1124. retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
  1125. if (retval)
  1126. goto bad_fork_cleanup_io;
  1127. if (pid != &init_struct_pid) {
  1128. retval = -ENOMEM;
  1129. pid = alloc_pid(p->nsproxy->pid_ns);
  1130. if (!pid)
  1131. goto bad_fork_cleanup_io;
  1132. }
  1133. p->pid = pid_nr(pid);
  1134. p->tgid = p->pid;
  1135. if (clone_flags & CLONE_THREAD)
  1136. p->tgid = current->tgid;
  1137. p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
  1138. /*
  1139. * Clear TID on mm_release()?
  1140. */
  1141. p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
  1142. #ifdef CONFIG_BLOCK
  1143. p->plug = NULL;
  1144. #endif
  1145. #ifdef CONFIG_FUTEX
  1146. p->robust_list = NULL;
  1147. #ifdef CONFIG_COMPAT
  1148. p->compat_robust_list = NULL;
  1149. #endif
  1150. INIT_LIST_HEAD(&p->pi_state_list);
  1151. p->pi_state_cache = NULL;
  1152. #endif
  1153. /*
  1154. * sigaltstack should be cleared when sharing the same VM
  1155. */
  1156. if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
  1157. p->sas_ss_sp = p->sas_ss_size = 0;
  1158. /*
  1159. * Syscall tracing and stepping should be turned off in the
  1160. * child regardless of CLONE_PTRACE.
  1161. */
  1162. user_disable_single_step(p);
  1163. clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
  1164. #ifdef TIF_SYSCALL_EMU
  1165. clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
  1166. #endif
  1167. clear_all_latency_tracing(p);
  1168. /* ok, now we should be set up.. */
  1169. if (clone_flags & CLONE_THREAD)
  1170. p->exit_signal = -1;
  1171. else if (clone_flags & CLONE_PARENT)
  1172. p->exit_signal = current->group_leader->exit_signal;
  1173. else
  1174. p->exit_signal = (clone_flags & CSIGNAL);
  1175. p->pdeath_signal = 0;
  1176. p->exit_state = 0;
  1177. p->nr_dirtied = 0;
  1178. p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
  1179. p->dirty_paused_when = 0;
  1180. /*
  1181. * Ok, make it visible to the rest of the system.
  1182. * We dont wake it up yet.
  1183. */
  1184. p->group_leader = p;
  1185. INIT_LIST_HEAD(&p->thread_group);
  1186. /* Now that the task is set up, run cgroup callbacks if
  1187. * necessary. We need to run them before the task is visible
  1188. * on the tasklist. */
  1189. cgroup_fork_callbacks(p);
  1190. cgroup_callbacks_done = 1;
  1191. /* Need tasklist lock for parent etc handling! */
  1192. write_lock_irq(&tasklist_lock);
  1193. /* CLONE_PARENT re-uses the old parent */
  1194. if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
  1195. p->real_parent = current->real_parent;
  1196. p->parent_exec_id = current->parent_exec_id;
  1197. } else {
  1198. p->real_parent = current;
  1199. p->parent_exec_id = current->self_exec_id;
  1200. }
  1201. spin_lock(&current->sighand->siglock);
  1202. /*
  1203. * Process group and session signals need to be delivered to just the
  1204. * parent before the fork or both the parent and the child after the
  1205. * fork. Restart if a signal comes in before we add the new process to
  1206. * it's process group.
  1207. * A fatal signal pending means that current will exit, so the new
  1208. * thread can't slip out of an OOM kill (or normal SIGKILL).
  1209. */
  1210. recalc_sigpending();
  1211. if (signal_pending(current)) {
  1212. spin_unlock(&current->sighand->siglock);
  1213. write_unlock_irq(&tasklist_lock);
  1214. retval = -ERESTARTNOINTR;
  1215. goto bad_fork_free_pid;
  1216. }
  1217. if (clone_flags & CLONE_THREAD) {
  1218. current->signal->nr_threads++;
  1219. atomic_inc(&current->signal->live);
  1220. atomic_inc(&current->signal->sigcnt);
  1221. p->group_leader = current->group_leader;
  1222. list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
  1223. }
  1224. if (likely(p->pid)) {
  1225. ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
  1226. if (thread_group_leader(p)) {
  1227. if (is_child_reaper(pid))
  1228. p->nsproxy->pid_ns->child_reaper = p;
  1229. p->signal->leader_pid = pid;
  1230. p->signal->tty = tty_kref_get(current->signal->tty);
  1231. attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
  1232. attach_pid(p, PIDTYPE_SID, task_session(current));
  1233. list_add_tail(&p->sibling, &p->real_parent->children);
  1234. list_add_tail_rcu(&p->tasks, &init_task.tasks);
  1235. __this_cpu_inc(process_counts);
  1236. }
  1237. attach_pid(p, PIDTYPE_PID, pid);
  1238. nr_threads++;
  1239. }
  1240. total_forks++;
  1241. spin_unlock(&current->sighand->siglock);
  1242. write_unlock_irq(&tasklist_lock);
  1243. proc_fork_connector(p);
  1244. cgroup_post_fork(p);
  1245. if (clone_flags & CLONE_THREAD)
  1246. threadgroup_change_end(current);
  1247. perf_event_fork(p);
  1248. trace_task_newtask(p, clone_flags);
  1249. return p;
  1250. bad_fork_free_pid:
  1251. if (pid != &init_struct_pid)
  1252. free_pid(pid);
  1253. bad_fork_cleanup_io:
  1254. if (p->io_context)
  1255. exit_io_context(p);
  1256. bad_fork_cleanup_namespaces:
  1257. exit_task_namespaces(p);
  1258. bad_fork_cleanup_mm:
  1259. if (p->mm)
  1260. mmput(p->mm);
  1261. bad_fork_cleanup_signal:
  1262. if (!(clone_flags & CLONE_THREAD))
  1263. free_signal_struct(p->signal);
  1264. bad_fork_cleanup_sighand:
  1265. __cleanup_sighand(p->sighand);
  1266. bad_fork_cleanup_fs:
  1267. exit_fs(p); /* blocking */
  1268. bad_fork_cleanup_files:
  1269. exit_files(p); /* blocking */
  1270. bad_fork_cleanup_semundo:
  1271. exit_sem(p);
  1272. bad_fork_cleanup_audit:
  1273. audit_free(p);
  1274. bad_fork_cleanup_policy:
  1275. perf_event_free_task(p);
  1276. #ifdef CONFIG_NUMA
  1277. mpol_put(p->mempolicy);
  1278. bad_fork_cleanup_cgroup:
  1279. #endif
  1280. if (clone_flags & CLONE_THREAD)
  1281. threadgroup_change_end(current);
  1282. cgroup_exit(p, cgroup_callbacks_done);
  1283. delayacct_tsk_free(p);
  1284. module_put(task_thread_info(p)->exec_domain->module);
  1285. bad_fork_cleanup_count:
  1286. atomic_dec(&p->cred->user->processes);
  1287. exit_creds(p);
  1288. bad_fork_free:
  1289. free_task(p);
  1290. fork_out:
  1291. return ERR_PTR(retval);
  1292. }
  1293. noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
  1294. {
  1295. memset(regs, 0, sizeof(struct pt_regs));
  1296. return regs;
  1297. }
  1298. static inline void init_idle_pids(struct pid_link *links)
  1299. {
  1300. enum pid_type type;
  1301. for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
  1302. INIT_HLIST_NODE(&links[type].node); /* not really needed */
  1303. links[type].pid = &init_struct_pid;
  1304. }
  1305. }
  1306. struct task_struct * __cpuinit fork_idle(int cpu)
  1307. {
  1308. struct task_struct *task;
  1309. struct pt_regs regs;
  1310. task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL,
  1311. &init_struct_pid, 0);
  1312. if (!IS_ERR(task)) {
  1313. init_idle_pids(task->pids);
  1314. init_idle(task, cpu);
  1315. }
  1316. return task;
  1317. }
  1318. /*
  1319. * Ok, this is the main fork-routine.
  1320. *
  1321. * It copies the process, and if successful kick-starts
  1322. * it and waits for it to finish using the VM if required.
  1323. */
  1324. long do_fork(unsigned long clone_flags,
  1325. unsigned long stack_start,
  1326. struct pt_regs *regs,
  1327. unsigned long stack_size,
  1328. int __user *parent_tidptr,
  1329. int __user *child_tidptr)
  1330. {
  1331. struct task_struct *p;
  1332. int trace = 0;
  1333. long nr;
  1334. /*
  1335. * Do some preliminary argument and permissions checking before we
  1336. * actually start allocating stuff
  1337. */
  1338. if (clone_flags & CLONE_NEWUSER) {
  1339. if (clone_flags & CLONE_THREAD)
  1340. return -EINVAL;
  1341. /* hopefully this check will go away when userns support is
  1342. * complete
  1343. */
  1344. if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
  1345. !capable(CAP_SETGID))
  1346. return -EPERM;
  1347. }
  1348. /*
  1349. * Determine whether and which event to report to ptracer. When
  1350. * called from kernel_thread or CLONE_UNTRACED is explicitly
  1351. * requested, no event is reported; otherwise, report if the event
  1352. * for the type of forking is enabled.
  1353. */
  1354. if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) {
  1355. if (clone_flags & CLONE_VFORK)
  1356. trace = PTRACE_EVENT_VFORK;
  1357. else if ((clone_flags & CSIGNAL) != SIGCHLD)
  1358. trace = PTRACE_EVENT_CLONE;
  1359. else
  1360. trace = PTRACE_EVENT_FORK;
  1361. if (likely(!ptrace_event_enabled(current, trace)))
  1362. trace = 0;
  1363. }
  1364. p = copy_process(clone_flags, stack_start, regs, stack_size,
  1365. child_tidptr, NULL, trace);
  1366. /*
  1367. * Do this prior waking up the new thread - the thread pointer
  1368. * might get invalid after that point, if the thread exits quickly.
  1369. */
  1370. if (!IS_ERR(p)) {
  1371. struct completion vfork;
  1372. trace_sched_process_fork(current, p);
  1373. nr = task_pid_vnr(p);
  1374. if (clone_flags & CLONE_PARENT_SETTID)
  1375. put_user(nr, parent_tidptr);
  1376. if (clone_flags & CLONE_VFORK) {
  1377. p->vfork_done = &vfork;
  1378. init_completion(&vfork);
  1379. get_task_struct(p);
  1380. }
  1381. wake_up_new_task(p);
  1382. /* forking complete and child started to run, tell ptracer */
  1383. if (unlikely(trace))
  1384. ptrace_event(trace, nr);
  1385. if (clone_flags & CLONE_VFORK) {
  1386. if (!wait_for_vfork_done(p, &vfork))
  1387. ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
  1388. }
  1389. } else {
  1390. nr = PTR_ERR(p);
  1391. }
  1392. return nr;
  1393. }
  1394. #ifndef ARCH_MIN_MMSTRUCT_ALIGN
  1395. #define ARCH_MIN_MMSTRUCT_ALIGN 0
  1396. #endif
  1397. static void sighand_ctor(void *data)
  1398. {
  1399. struct sighand_struct *sighand = data;
  1400. spin_lock_init(&sighand->siglock);
  1401. init_waitqueue_head(&sighand->signalfd_wqh);
  1402. }
  1403. void __init proc_caches_init(void)
  1404. {
  1405. sighand_cachep = kmem_cache_create("sighand_cache",
  1406. sizeof(struct sighand_struct), 0,
  1407. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
  1408. SLAB_NOTRACK, sighand_ctor);
  1409. signal_cachep = kmem_cache_create("signal_cache",
  1410. sizeof(struct signal_struct), 0,
  1411. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
  1412. files_cachep = kmem_cache_create("files_cache",
  1413. sizeof(struct files_struct), 0,
  1414. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
  1415. fs_cachep = kmem_cache_create("fs_cache",
  1416. sizeof(struct fs_struct), 0,
  1417. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
  1418. /*
  1419. * FIXME! The "sizeof(struct mm_struct)" currently includes the
  1420. * whole struct cpumask for the OFFSTACK case. We could change
  1421. * this to *only* allocate as much of it as required by the
  1422. * maximum number of CPU's we can ever have. The cpumask_allocation
  1423. * is at the end of the structure, exactly for that reason.
  1424. */
  1425. mm_cachep = kmem_cache_create("mm_struct",
  1426. sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
  1427. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
  1428. vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
  1429. mmap_init();
  1430. nsproxy_cache_init();
  1431. }
  1432. /*
  1433. * Check constraints on flags passed to the unshare system call.
  1434. */
  1435. static int check_unshare_flags(unsigned long unshare_flags)
  1436. {
  1437. if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
  1438. CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
  1439. CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
  1440. return -EINVAL;
  1441. /*
  1442. * Not implemented, but pretend it works if there is nothing to
  1443. * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
  1444. * needs to unshare vm.
  1445. */
  1446. if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
  1447. /* FIXME: get_task_mm() increments ->mm_users */
  1448. if (atomic_read(&current->mm->mm_users) > 1)
  1449. return -EINVAL;
  1450. }
  1451. return 0;
  1452. }
  1453. /*
  1454. * Unshare the filesystem structure if it is being shared
  1455. */
  1456. static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
  1457. {
  1458. struct fs_struct *fs = current->fs;
  1459. if (!(unshare_flags & CLONE_FS) || !fs)
  1460. return 0;
  1461. /* don't need lock here; in the worst case we'll do useless copy */
  1462. if (fs->users == 1)
  1463. return 0;
  1464. *new_fsp = copy_fs_struct(fs);
  1465. if (!*new_fsp)
  1466. return -ENOMEM;
  1467. return 0;
  1468. }
  1469. /*
  1470. * Unshare file descriptor table if it is being shared
  1471. */
  1472. static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
  1473. {
  1474. struct files_struct *fd = current->files;
  1475. int error = 0;
  1476. if ((unshare_flags & CLONE_FILES) &&
  1477. (fd && atomic_read(&fd->count) > 1)) {
  1478. *new_fdp = dup_fd(fd, &error);
  1479. if (!*new_fdp)
  1480. return error;
  1481. }
  1482. return 0;
  1483. }
  1484. /*
  1485. * unshare allows a process to 'unshare' part of the process
  1486. * context which was originally shared using clone. copy_*
  1487. * functions used by do_fork() cannot be used here directly
  1488. * because they modify an inactive task_struct that is being
  1489. * constructed. Here we are modifying the current, active,
  1490. * task_struct.
  1491. */
  1492. SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
  1493. {
  1494. struct fs_struct *fs, *new_fs = NULL;
  1495. struct files_struct *fd, *new_fd = NULL;
  1496. struct nsproxy *new_nsproxy = NULL;
  1497. int do_sysvsem = 0;
  1498. int err;
  1499. err = check_unshare_flags(unshare_flags);
  1500. if (err)
  1501. goto bad_unshare_out;
  1502. /*
  1503. * If unsharing namespace, must also unshare filesystem information.
  1504. */
  1505. if (unshare_flags & CLONE_NEWNS)
  1506. unshare_flags |= CLONE_FS;
  1507. /*
  1508. * CLONE_NEWIPC must also detach from the undolist: after switching
  1509. * to a new ipc namespace, the semaphore arrays from the old
  1510. * namespace are unreachable.
  1511. */
  1512. if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
  1513. do_sysvsem = 1;
  1514. err = unshare_fs(unshare_flags, &new_fs);
  1515. if (err)
  1516. goto bad_unshare_out;
  1517. err = unshare_fd(unshare_flags, &new_fd);
  1518. if (err)
  1519. goto bad_unshare_cleanup_fs;
  1520. err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs);
  1521. if (err)
  1522. goto bad_unshare_cleanup_fd;
  1523. if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
  1524. if (do_sysvsem) {
  1525. /*
  1526. * CLONE_SYSVSEM is equivalent to sys_exit().
  1527. */
  1528. exit_sem(current);
  1529. }
  1530. if (new_nsproxy) {
  1531. switch_task_namespaces(current, new_nsproxy);
  1532. new_nsproxy = NULL;
  1533. }
  1534. task_lock(current);
  1535. if (new_fs) {
  1536. fs = current->fs;
  1537. spin_lock(&fs->lock);
  1538. current->fs = new_fs;
  1539. if (--fs->users)
  1540. new_fs = NULL;
  1541. else
  1542. new_fs = fs;
  1543. spin_unlock(&fs->lock);
  1544. }
  1545. if (new_fd) {
  1546. fd = current->files;
  1547. current->files = new_fd;
  1548. new_fd = fd;
  1549. }
  1550. task_unlock(current);
  1551. }
  1552. if (new_nsproxy)
  1553. put_nsproxy(new_nsproxy);
  1554. bad_unshare_cleanup_fd:
  1555. if (new_fd)
  1556. put_files_struct(new_fd);
  1557. bad_unshare_cleanup_fs:
  1558. if (new_fs)
  1559. free_fs_struct(new_fs);
  1560. bad_unshare_out:
  1561. return err;
  1562. }
  1563. /*
  1564. * Helper to unshare the files of the current task.
  1565. * We don't want to expose copy_files internals to
  1566. * the exec layer of the kernel.
  1567. */
  1568. int unshare_files(struct files_struct **displaced)
  1569. {
  1570. struct task_struct *task = current;
  1571. struct files_struct *copy = NULL;
  1572. int error;
  1573. error = unshare_fd(CLONE_FILES, &copy);
  1574. if (error || !copy) {
  1575. *displaced = NULL;
  1576. return error;
  1577. }
  1578. *displaced = task->files;
  1579. task_lock(task);
  1580. task->files = copy;
  1581. task_unlock(task);
  1582. return 0;
  1583. }