fork.c 43 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763
  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/smp_lock.h>
  16. #include <linux/module.h>
  17. #include <linux/vmalloc.h>
  18. #include <linux/completion.h>
  19. #include <linux/mnt_namespace.h>
  20. #include <linux/personality.h>
  21. #include <linux/mempolicy.h>
  22. #include <linux/sem.h>
  23. #include <linux/file.h>
  24. #include <linux/key.h>
  25. #include <linux/binfmts.h>
  26. #include <linux/mman.h>
  27. #include <linux/fs.h>
  28. #include <linux/nsproxy.h>
  29. #include <linux/capability.h>
  30. #include <linux/cpu.h>
  31. #include <linux/cpuset.h>
  32. #include <linux/security.h>
  33. #include <linux/swap.h>
  34. #include <linux/syscalls.h>
  35. #include <linux/jiffies.h>
  36. #include <linux/futex.h>
  37. #include <linux/task_io_accounting_ops.h>
  38. #include <linux/rcupdate.h>
  39. #include <linux/ptrace.h>
  40. #include <linux/mount.h>
  41. #include <linux/audit.h>
  42. #include <linux/profile.h>
  43. #include <linux/rmap.h>
  44. #include <linux/acct.h>
  45. #include <linux/tsacct_kern.h>
  46. #include <linux/cn_proc.h>
  47. #include <linux/delayacct.h>
  48. #include <linux/taskstats_kern.h>
  49. #include <linux/random.h>
  50. #include <asm/pgtable.h>
  51. #include <asm/pgalloc.h>
  52. #include <asm/uaccess.h>
  53. #include <asm/mmu_context.h>
  54. #include <asm/cacheflush.h>
  55. #include <asm/tlbflush.h>
  56. /*
  57. * Protected counters by write_lock_irq(&tasklist_lock)
  58. */
  59. unsigned long total_forks; /* Handle normal Linux uptimes. */
  60. int nr_threads; /* The idle threads do not count.. */
  61. int max_threads; /* tunable limit on nr_threads */
  62. DEFINE_PER_CPU(unsigned long, process_counts) = 0;
  63. __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
  64. int nr_processes(void)
  65. {
  66. int cpu;
  67. int total = 0;
  68. for_each_online_cpu(cpu)
  69. total += per_cpu(process_counts, cpu);
  70. return total;
  71. }
  72. #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
  73. # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
  74. # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
  75. static struct kmem_cache *task_struct_cachep;
  76. #endif
  77. /* SLAB cache for signal_struct structures (tsk->signal) */
  78. static struct kmem_cache *signal_cachep;
  79. /* SLAB cache for sighand_struct structures (tsk->sighand) */
  80. struct kmem_cache *sighand_cachep;
  81. /* SLAB cache for files_struct structures (tsk->files) */
  82. struct kmem_cache *files_cachep;
  83. /* SLAB cache for fs_struct structures (tsk->fs) */
  84. struct kmem_cache *fs_cachep;
  85. /* SLAB cache for vm_area_struct structures */
  86. struct kmem_cache *vm_area_cachep;
  87. /* SLAB cache for mm_struct structures (tsk->mm) */
  88. static struct kmem_cache *mm_cachep;
  89. void free_task(struct task_struct *tsk)
  90. {
  91. free_thread_info(tsk->thread_info);
  92. rt_mutex_debug_task_free(tsk);
  93. free_task_struct(tsk);
  94. }
  95. EXPORT_SYMBOL(free_task);
  96. void __put_task_struct(struct task_struct *tsk)
  97. {
  98. WARN_ON(!(tsk->exit_state & (EXIT_DEAD | EXIT_ZOMBIE)));
  99. WARN_ON(atomic_read(&tsk->usage));
  100. WARN_ON(tsk == current);
  101. security_task_free(tsk);
  102. free_uid(tsk->user);
  103. put_group_info(tsk->group_info);
  104. delayacct_tsk_free(tsk);
  105. if (!profile_handoff_task(tsk))
  106. free_task(tsk);
  107. }
  108. void __init fork_init(unsigned long mempages)
  109. {
  110. #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
  111. #ifndef ARCH_MIN_TASKALIGN
  112. #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
  113. #endif
  114. /* create a slab on which task_structs can be allocated */
  115. task_struct_cachep =
  116. kmem_cache_create("task_struct", sizeof(struct task_struct),
  117. ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL, NULL);
  118. #endif
  119. /*
  120. * The default maximum number of threads is set to a safe
  121. * value: the thread structures can take up at most half
  122. * of memory.
  123. */
  124. max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
  125. /*
  126. * we need to allow at least 20 threads to boot a system
  127. */
  128. if(max_threads < 20)
  129. max_threads = 20;
  130. init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
  131. init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
  132. init_task.signal->rlim[RLIMIT_SIGPENDING] =
  133. init_task.signal->rlim[RLIMIT_NPROC];
  134. }
  135. static struct task_struct *dup_task_struct(struct task_struct *orig)
  136. {
  137. struct task_struct *tsk;
  138. struct thread_info *ti;
  139. prepare_to_copy(orig);
  140. tsk = alloc_task_struct();
  141. if (!tsk)
  142. return NULL;
  143. ti = alloc_thread_info(tsk);
  144. if (!ti) {
  145. free_task_struct(tsk);
  146. return NULL;
  147. }
  148. *tsk = *orig;
  149. tsk->thread_info = ti;
  150. setup_thread_stack(tsk, orig);
  151. #ifdef CONFIG_CC_STACKPROTECTOR
  152. tsk->stack_canary = get_random_int();
  153. #endif
  154. /* One for us, one for whoever does the "release_task()" (usually parent) */
  155. atomic_set(&tsk->usage,2);
  156. atomic_set(&tsk->fs_excl, 0);
  157. #ifdef CONFIG_BLK_DEV_IO_TRACE
  158. tsk->btrace_seq = 0;
  159. #endif
  160. tsk->splice_pipe = NULL;
  161. return tsk;
  162. }
  163. #ifdef CONFIG_MMU
  164. static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
  165. {
  166. struct vm_area_struct *mpnt, *tmp, **pprev;
  167. struct rb_node **rb_link, *rb_parent;
  168. int retval;
  169. unsigned long charge;
  170. struct mempolicy *pol;
  171. down_write(&oldmm->mmap_sem);
  172. flush_cache_mm(oldmm);
  173. /*
  174. * Not linked in yet - no deadlock potential:
  175. */
  176. down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
  177. mm->locked_vm = 0;
  178. mm->mmap = NULL;
  179. mm->mmap_cache = NULL;
  180. mm->free_area_cache = oldmm->mmap_base;
  181. mm->cached_hole_size = ~0UL;
  182. mm->map_count = 0;
  183. cpus_clear(mm->cpu_vm_mask);
  184. mm->mm_rb = RB_ROOT;
  185. rb_link = &mm->mm_rb.rb_node;
  186. rb_parent = NULL;
  187. pprev = &mm->mmap;
  188. for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
  189. struct file *file;
  190. if (mpnt->vm_flags & VM_DONTCOPY) {
  191. long pages = vma_pages(mpnt);
  192. mm->total_vm -= pages;
  193. vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
  194. -pages);
  195. continue;
  196. }
  197. charge = 0;
  198. if (mpnt->vm_flags & VM_ACCOUNT) {
  199. unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
  200. if (security_vm_enough_memory(len))
  201. goto fail_nomem;
  202. charge = len;
  203. }
  204. tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
  205. if (!tmp)
  206. goto fail_nomem;
  207. *tmp = *mpnt;
  208. pol = mpol_copy(vma_policy(mpnt));
  209. retval = PTR_ERR(pol);
  210. if (IS_ERR(pol))
  211. goto fail_nomem_policy;
  212. vma_set_policy(tmp, pol);
  213. tmp->vm_flags &= ~VM_LOCKED;
  214. tmp->vm_mm = mm;
  215. tmp->vm_next = NULL;
  216. anon_vma_link(tmp);
  217. file = tmp->vm_file;
  218. if (file) {
  219. struct inode *inode = file->f_path.dentry->d_inode;
  220. get_file(file);
  221. if (tmp->vm_flags & VM_DENYWRITE)
  222. atomic_dec(&inode->i_writecount);
  223. /* insert tmp into the share list, just after mpnt */
  224. spin_lock(&file->f_mapping->i_mmap_lock);
  225. tmp->vm_truncate_count = mpnt->vm_truncate_count;
  226. flush_dcache_mmap_lock(file->f_mapping);
  227. vma_prio_tree_add(tmp, mpnt);
  228. flush_dcache_mmap_unlock(file->f_mapping);
  229. spin_unlock(&file->f_mapping->i_mmap_lock);
  230. }
  231. /*
  232. * Link in the new vma and copy the page table entries.
  233. */
  234. *pprev = tmp;
  235. pprev = &tmp->vm_next;
  236. __vma_link_rb(mm, tmp, rb_link, rb_parent);
  237. rb_link = &tmp->vm_rb.rb_right;
  238. rb_parent = &tmp->vm_rb;
  239. mm->map_count++;
  240. retval = copy_page_range(mm, oldmm, mpnt);
  241. if (tmp->vm_ops && tmp->vm_ops->open)
  242. tmp->vm_ops->open(tmp);
  243. if (retval)
  244. goto out;
  245. }
  246. retval = 0;
  247. out:
  248. up_write(&mm->mmap_sem);
  249. flush_tlb_mm(oldmm);
  250. up_write(&oldmm->mmap_sem);
  251. return retval;
  252. fail_nomem_policy:
  253. kmem_cache_free(vm_area_cachep, tmp);
  254. fail_nomem:
  255. retval = -ENOMEM;
  256. vm_unacct_memory(charge);
  257. goto out;
  258. }
  259. static inline int mm_alloc_pgd(struct mm_struct * mm)
  260. {
  261. mm->pgd = pgd_alloc(mm);
  262. if (unlikely(!mm->pgd))
  263. return -ENOMEM;
  264. return 0;
  265. }
  266. static inline void mm_free_pgd(struct mm_struct * mm)
  267. {
  268. pgd_free(mm->pgd);
  269. }
  270. #else
  271. #define dup_mmap(mm, oldmm) (0)
  272. #define mm_alloc_pgd(mm) (0)
  273. #define mm_free_pgd(mm)
  274. #endif /* CONFIG_MMU */
  275. __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
  276. #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
  277. #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
  278. #include <linux/init_task.h>
  279. static struct mm_struct * mm_init(struct mm_struct * mm)
  280. {
  281. atomic_set(&mm->mm_users, 1);
  282. atomic_set(&mm->mm_count, 1);
  283. init_rwsem(&mm->mmap_sem);
  284. INIT_LIST_HEAD(&mm->mmlist);
  285. mm->core_waiters = 0;
  286. mm->nr_ptes = 0;
  287. set_mm_counter(mm, file_rss, 0);
  288. set_mm_counter(mm, anon_rss, 0);
  289. spin_lock_init(&mm->page_table_lock);
  290. rwlock_init(&mm->ioctx_list_lock);
  291. mm->ioctx_list = NULL;
  292. mm->free_area_cache = TASK_UNMAPPED_BASE;
  293. mm->cached_hole_size = ~0UL;
  294. if (likely(!mm_alloc_pgd(mm))) {
  295. mm->def_flags = 0;
  296. return mm;
  297. }
  298. free_mm(mm);
  299. return NULL;
  300. }
  301. /*
  302. * Allocate and initialize an mm_struct.
  303. */
  304. struct mm_struct * mm_alloc(void)
  305. {
  306. struct mm_struct * mm;
  307. mm = allocate_mm();
  308. if (mm) {
  309. memset(mm, 0, sizeof(*mm));
  310. mm = mm_init(mm);
  311. }
  312. return mm;
  313. }
  314. /*
  315. * Called when the last reference to the mm
  316. * is dropped: either by a lazy thread or by
  317. * mmput. Free the page directory and the mm.
  318. */
  319. void fastcall __mmdrop(struct mm_struct *mm)
  320. {
  321. BUG_ON(mm == &init_mm);
  322. mm_free_pgd(mm);
  323. destroy_context(mm);
  324. free_mm(mm);
  325. }
  326. /*
  327. * Decrement the use count and release all resources for an mm.
  328. */
  329. void mmput(struct mm_struct *mm)
  330. {
  331. might_sleep();
  332. if (atomic_dec_and_test(&mm->mm_users)) {
  333. exit_aio(mm);
  334. exit_mmap(mm);
  335. if (!list_empty(&mm->mmlist)) {
  336. spin_lock(&mmlist_lock);
  337. list_del(&mm->mmlist);
  338. spin_unlock(&mmlist_lock);
  339. }
  340. put_swap_token(mm);
  341. mmdrop(mm);
  342. }
  343. }
  344. EXPORT_SYMBOL_GPL(mmput);
  345. /**
  346. * get_task_mm - acquire a reference to the task's mm
  347. *
  348. * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning
  349. * this kernel workthread has transiently adopted a user mm with use_mm,
  350. * to do its AIO) is not set and if so returns a reference to it, after
  351. * bumping up the use count. User must release the mm via mmput()
  352. * after use. Typically used by /proc and ptrace.
  353. */
  354. struct mm_struct *get_task_mm(struct task_struct *task)
  355. {
  356. struct mm_struct *mm;
  357. task_lock(task);
  358. mm = task->mm;
  359. if (mm) {
  360. if (task->flags & PF_BORROWED_MM)
  361. mm = NULL;
  362. else
  363. atomic_inc(&mm->mm_users);
  364. }
  365. task_unlock(task);
  366. return mm;
  367. }
  368. EXPORT_SYMBOL_GPL(get_task_mm);
  369. /* Please note the differences between mmput and mm_release.
  370. * mmput is called whenever we stop holding onto a mm_struct,
  371. * error success whatever.
  372. *
  373. * mm_release is called after a mm_struct has been removed
  374. * from the current process.
  375. *
  376. * This difference is important for error handling, when we
  377. * only half set up a mm_struct for a new process and need to restore
  378. * the old one. Because we mmput the new mm_struct before
  379. * restoring the old one. . .
  380. * Eric Biederman 10 January 1998
  381. */
  382. void mm_release(struct task_struct *tsk, struct mm_struct *mm)
  383. {
  384. struct completion *vfork_done = tsk->vfork_done;
  385. /* Get rid of any cached register state */
  386. deactivate_mm(tsk, mm);
  387. /* notify parent sleeping on vfork() */
  388. if (vfork_done) {
  389. tsk->vfork_done = NULL;
  390. complete(vfork_done);
  391. }
  392. /*
  393. * If we're exiting normally, clear a user-space tid field if
  394. * requested. We leave this alone when dying by signal, to leave
  395. * the value intact in a core dump, and to save the unnecessary
  396. * trouble otherwise. Userland only wants this done for a sys_exit.
  397. */
  398. if (tsk->clear_child_tid
  399. && !(tsk->flags & PF_SIGNALED)
  400. && atomic_read(&mm->mm_users) > 1) {
  401. u32 __user * tidptr = tsk->clear_child_tid;
  402. tsk->clear_child_tid = NULL;
  403. /*
  404. * We don't check the error code - if userspace has
  405. * not set up a proper pointer then tough luck.
  406. */
  407. put_user(0, tidptr);
  408. sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0);
  409. }
  410. }
  411. /*
  412. * Allocate a new mm structure and copy contents from the
  413. * mm structure of the passed in task structure.
  414. */
  415. static struct mm_struct *dup_mm(struct task_struct *tsk)
  416. {
  417. struct mm_struct *mm, *oldmm = current->mm;
  418. int err;
  419. if (!oldmm)
  420. return NULL;
  421. mm = allocate_mm();
  422. if (!mm)
  423. goto fail_nomem;
  424. memcpy(mm, oldmm, sizeof(*mm));
  425. /* Initializing for Swap token stuff */
  426. mm->token_priority = 0;
  427. mm->last_interval = 0;
  428. if (!mm_init(mm))
  429. goto fail_nomem;
  430. if (init_new_context(tsk, mm))
  431. goto fail_nocontext;
  432. err = dup_mmap(mm, oldmm);
  433. if (err)
  434. goto free_pt;
  435. mm->hiwater_rss = get_mm_rss(mm);
  436. mm->hiwater_vm = mm->total_vm;
  437. return mm;
  438. free_pt:
  439. mmput(mm);
  440. fail_nomem:
  441. return NULL;
  442. fail_nocontext:
  443. /*
  444. * If init_new_context() failed, we cannot use mmput() to free the mm
  445. * because it calls destroy_context()
  446. */
  447. mm_free_pgd(mm);
  448. free_mm(mm);
  449. return NULL;
  450. }
  451. static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
  452. {
  453. struct mm_struct * mm, *oldmm;
  454. int retval;
  455. tsk->min_flt = tsk->maj_flt = 0;
  456. tsk->nvcsw = tsk->nivcsw = 0;
  457. tsk->mm = NULL;
  458. tsk->active_mm = NULL;
  459. /*
  460. * Are we cloning a kernel thread?
  461. *
  462. * We need to steal a active VM for that..
  463. */
  464. oldmm = current->mm;
  465. if (!oldmm)
  466. return 0;
  467. if (clone_flags & CLONE_VM) {
  468. atomic_inc(&oldmm->mm_users);
  469. mm = oldmm;
  470. goto good_mm;
  471. }
  472. retval = -ENOMEM;
  473. mm = dup_mm(tsk);
  474. if (!mm)
  475. goto fail_nomem;
  476. good_mm:
  477. /* Initializing for Swap token stuff */
  478. mm->token_priority = 0;
  479. mm->last_interval = 0;
  480. tsk->mm = mm;
  481. tsk->active_mm = mm;
  482. return 0;
  483. fail_nomem:
  484. return retval;
  485. }
  486. static inline struct fs_struct *__copy_fs_struct(struct fs_struct *old)
  487. {
  488. struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL);
  489. /* We don't need to lock fs - think why ;-) */
  490. if (fs) {
  491. atomic_set(&fs->count, 1);
  492. rwlock_init(&fs->lock);
  493. fs->umask = old->umask;
  494. read_lock(&old->lock);
  495. fs->rootmnt = mntget(old->rootmnt);
  496. fs->root = dget(old->root);
  497. fs->pwdmnt = mntget(old->pwdmnt);
  498. fs->pwd = dget(old->pwd);
  499. if (old->altroot) {
  500. fs->altrootmnt = mntget(old->altrootmnt);
  501. fs->altroot = dget(old->altroot);
  502. } else {
  503. fs->altrootmnt = NULL;
  504. fs->altroot = NULL;
  505. }
  506. read_unlock(&old->lock);
  507. }
  508. return fs;
  509. }
  510. struct fs_struct *copy_fs_struct(struct fs_struct *old)
  511. {
  512. return __copy_fs_struct(old);
  513. }
  514. EXPORT_SYMBOL_GPL(copy_fs_struct);
  515. static inline int copy_fs(unsigned long clone_flags, struct task_struct * tsk)
  516. {
  517. if (clone_flags & CLONE_FS) {
  518. atomic_inc(&current->fs->count);
  519. return 0;
  520. }
  521. tsk->fs = __copy_fs_struct(current->fs);
  522. if (!tsk->fs)
  523. return -ENOMEM;
  524. return 0;
  525. }
  526. static int count_open_files(struct fdtable *fdt)
  527. {
  528. int size = fdt->max_fdset;
  529. int i;
  530. /* Find the last open fd */
  531. for (i = size/(8*sizeof(long)); i > 0; ) {
  532. if (fdt->open_fds->fds_bits[--i])
  533. break;
  534. }
  535. i = (i+1) * 8 * sizeof(long);
  536. return i;
  537. }
  538. static struct files_struct *alloc_files(void)
  539. {
  540. struct files_struct *newf;
  541. struct fdtable *fdt;
  542. newf = kmem_cache_alloc(files_cachep, GFP_KERNEL);
  543. if (!newf)
  544. goto out;
  545. atomic_set(&newf->count, 1);
  546. spin_lock_init(&newf->file_lock);
  547. newf->next_fd = 0;
  548. fdt = &newf->fdtab;
  549. fdt->max_fds = NR_OPEN_DEFAULT;
  550. fdt->max_fdset = EMBEDDED_FD_SET_SIZE;
  551. fdt->close_on_exec = (fd_set *)&newf->close_on_exec_init;
  552. fdt->open_fds = (fd_set *)&newf->open_fds_init;
  553. fdt->fd = &newf->fd_array[0];
  554. INIT_RCU_HEAD(&fdt->rcu);
  555. fdt->free_files = NULL;
  556. fdt->next = NULL;
  557. rcu_assign_pointer(newf->fdt, fdt);
  558. out:
  559. return newf;
  560. }
  561. /*
  562. * Allocate a new files structure and copy contents from the
  563. * passed in files structure.
  564. * errorp will be valid only when the returned files_struct is NULL.
  565. */
  566. static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
  567. {
  568. struct files_struct *newf;
  569. struct file **old_fds, **new_fds;
  570. int open_files, size, i, expand;
  571. struct fdtable *old_fdt, *new_fdt;
  572. *errorp = -ENOMEM;
  573. newf = alloc_files();
  574. if (!newf)
  575. goto out;
  576. spin_lock(&oldf->file_lock);
  577. old_fdt = files_fdtable(oldf);
  578. new_fdt = files_fdtable(newf);
  579. size = old_fdt->max_fdset;
  580. open_files = count_open_files(old_fdt);
  581. expand = 0;
  582. /*
  583. * Check whether we need to allocate a larger fd array or fd set.
  584. * Note: we're not a clone task, so the open count won't change.
  585. */
  586. if (open_files > new_fdt->max_fdset) {
  587. new_fdt->max_fdset = 0;
  588. expand = 1;
  589. }
  590. if (open_files > new_fdt->max_fds) {
  591. new_fdt->max_fds = 0;
  592. expand = 1;
  593. }
  594. /* if the old fdset gets grown now, we'll only copy up to "size" fds */
  595. if (expand) {
  596. spin_unlock(&oldf->file_lock);
  597. spin_lock(&newf->file_lock);
  598. *errorp = expand_files(newf, open_files-1);
  599. spin_unlock(&newf->file_lock);
  600. if (*errorp < 0)
  601. goto out_release;
  602. new_fdt = files_fdtable(newf);
  603. /*
  604. * Reacquire the oldf lock and a pointer to its fd table
  605. * who knows it may have a new bigger fd table. We need
  606. * the latest pointer.
  607. */
  608. spin_lock(&oldf->file_lock);
  609. old_fdt = files_fdtable(oldf);
  610. }
  611. old_fds = old_fdt->fd;
  612. new_fds = new_fdt->fd;
  613. memcpy(new_fdt->open_fds->fds_bits, old_fdt->open_fds->fds_bits, open_files/8);
  614. memcpy(new_fdt->close_on_exec->fds_bits, old_fdt->close_on_exec->fds_bits, open_files/8);
  615. for (i = open_files; i != 0; i--) {
  616. struct file *f = *old_fds++;
  617. if (f) {
  618. get_file(f);
  619. } else {
  620. /*
  621. * The fd may be claimed in the fd bitmap but not yet
  622. * instantiated in the files array if a sibling thread
  623. * is partway through open(). So make sure that this
  624. * fd is available to the new process.
  625. */
  626. FD_CLR(open_files - i, new_fdt->open_fds);
  627. }
  628. rcu_assign_pointer(*new_fds++, f);
  629. }
  630. spin_unlock(&oldf->file_lock);
  631. /* compute the remainder to be cleared */
  632. size = (new_fdt->max_fds - open_files) * sizeof(struct file *);
  633. /* This is long word aligned thus could use a optimized version */
  634. memset(new_fds, 0, size);
  635. if (new_fdt->max_fdset > open_files) {
  636. int left = (new_fdt->max_fdset-open_files)/8;
  637. int start = open_files / (8 * sizeof(unsigned long));
  638. memset(&new_fdt->open_fds->fds_bits[start], 0, left);
  639. memset(&new_fdt->close_on_exec->fds_bits[start], 0, left);
  640. }
  641. out:
  642. return newf;
  643. out_release:
  644. free_fdset (new_fdt->close_on_exec, new_fdt->max_fdset);
  645. free_fdset (new_fdt->open_fds, new_fdt->max_fdset);
  646. free_fd_array(new_fdt->fd, new_fdt->max_fds);
  647. kmem_cache_free(files_cachep, newf);
  648. return NULL;
  649. }
  650. static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
  651. {
  652. struct files_struct *oldf, *newf;
  653. int error = 0;
  654. /*
  655. * A background process may not have any files ...
  656. */
  657. oldf = current->files;
  658. if (!oldf)
  659. goto out;
  660. if (clone_flags & CLONE_FILES) {
  661. atomic_inc(&oldf->count);
  662. goto out;
  663. }
  664. /*
  665. * Note: we may be using current for both targets (See exec.c)
  666. * This works because we cache current->files (old) as oldf. Don't
  667. * break this.
  668. */
  669. tsk->files = NULL;
  670. newf = dup_fd(oldf, &error);
  671. if (!newf)
  672. goto out;
  673. tsk->files = newf;
  674. error = 0;
  675. out:
  676. return error;
  677. }
  678. /*
  679. * Helper to unshare the files of the current task.
  680. * We don't want to expose copy_files internals to
  681. * the exec layer of the kernel.
  682. */
  683. int unshare_files(void)
  684. {
  685. struct files_struct *files = current->files;
  686. int rc;
  687. BUG_ON(!files);
  688. /* This can race but the race causes us to copy when we don't
  689. need to and drop the copy */
  690. if(atomic_read(&files->count) == 1)
  691. {
  692. atomic_inc(&files->count);
  693. return 0;
  694. }
  695. rc = copy_files(0, current);
  696. if(rc)
  697. current->files = files;
  698. return rc;
  699. }
  700. EXPORT_SYMBOL(unshare_files);
  701. static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk)
  702. {
  703. struct sighand_struct *sig;
  704. if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) {
  705. atomic_inc(&current->sighand->count);
  706. return 0;
  707. }
  708. sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
  709. rcu_assign_pointer(tsk->sighand, sig);
  710. if (!sig)
  711. return -ENOMEM;
  712. atomic_set(&sig->count, 1);
  713. memcpy(sig->action, current->sighand->action, sizeof(sig->action));
  714. return 0;
  715. }
  716. void __cleanup_sighand(struct sighand_struct *sighand)
  717. {
  718. if (atomic_dec_and_test(&sighand->count))
  719. kmem_cache_free(sighand_cachep, sighand);
  720. }
  721. static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk)
  722. {
  723. struct signal_struct *sig;
  724. int ret;
  725. if (clone_flags & CLONE_THREAD) {
  726. atomic_inc(&current->signal->count);
  727. atomic_inc(&current->signal->live);
  728. return 0;
  729. }
  730. sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
  731. tsk->signal = sig;
  732. if (!sig)
  733. return -ENOMEM;
  734. ret = copy_thread_group_keys(tsk);
  735. if (ret < 0) {
  736. kmem_cache_free(signal_cachep, sig);
  737. return ret;
  738. }
  739. atomic_set(&sig->count, 1);
  740. atomic_set(&sig->live, 1);
  741. init_waitqueue_head(&sig->wait_chldexit);
  742. sig->flags = 0;
  743. sig->group_exit_code = 0;
  744. sig->group_exit_task = NULL;
  745. sig->group_stop_count = 0;
  746. sig->curr_target = NULL;
  747. init_sigpending(&sig->shared_pending);
  748. INIT_LIST_HEAD(&sig->posix_timers);
  749. hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_REL);
  750. sig->it_real_incr.tv64 = 0;
  751. sig->real_timer.function = it_real_fn;
  752. sig->tsk = tsk;
  753. sig->it_virt_expires = cputime_zero;
  754. sig->it_virt_incr = cputime_zero;
  755. sig->it_prof_expires = cputime_zero;
  756. sig->it_prof_incr = cputime_zero;
  757. sig->leader = 0; /* session leadership doesn't inherit */
  758. sig->tty_old_pgrp = 0;
  759. sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
  760. sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
  761. sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
  762. sig->sched_time = 0;
  763. INIT_LIST_HEAD(&sig->cpu_timers[0]);
  764. INIT_LIST_HEAD(&sig->cpu_timers[1]);
  765. INIT_LIST_HEAD(&sig->cpu_timers[2]);
  766. taskstats_tgid_init(sig);
  767. task_lock(current->group_leader);
  768. memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
  769. task_unlock(current->group_leader);
  770. if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
  771. /*
  772. * New sole thread in the process gets an expiry time
  773. * of the whole CPU time limit.
  774. */
  775. tsk->it_prof_expires =
  776. secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur);
  777. }
  778. acct_init_pacct(&sig->pacct);
  779. return 0;
  780. }
  781. void __cleanup_signal(struct signal_struct *sig)
  782. {
  783. exit_thread_group_keys(sig);
  784. kmem_cache_free(signal_cachep, sig);
  785. }
  786. static inline void cleanup_signal(struct task_struct *tsk)
  787. {
  788. struct signal_struct *sig = tsk->signal;
  789. atomic_dec(&sig->live);
  790. if (atomic_dec_and_test(&sig->count))
  791. __cleanup_signal(sig);
  792. }
  793. static inline void copy_flags(unsigned long clone_flags, struct task_struct *p)
  794. {
  795. unsigned long new_flags = p->flags;
  796. new_flags &= ~(PF_SUPERPRIV | PF_NOFREEZE);
  797. new_flags |= PF_FORKNOEXEC;
  798. if (!(clone_flags & CLONE_PTRACE))
  799. p->ptrace = 0;
  800. p->flags = new_flags;
  801. }
  802. asmlinkage long sys_set_tid_address(int __user *tidptr)
  803. {
  804. current->clear_child_tid = tidptr;
  805. return current->pid;
  806. }
  807. static inline void rt_mutex_init_task(struct task_struct *p)
  808. {
  809. #ifdef CONFIG_RT_MUTEXES
  810. spin_lock_init(&p->pi_lock);
  811. plist_head_init(&p->pi_waiters, &p->pi_lock);
  812. p->pi_blocked_on = NULL;
  813. #endif
  814. }
  815. /*
  816. * This creates a new process as a copy of the old one,
  817. * but does not actually start it yet.
  818. *
  819. * It copies the registers, and all the appropriate
  820. * parts of the process environment (as per the clone
  821. * flags). The actual kick-off is left to the caller.
  822. */
  823. static struct task_struct *copy_process(unsigned long clone_flags,
  824. unsigned long stack_start,
  825. struct pt_regs *regs,
  826. unsigned long stack_size,
  827. int __user *parent_tidptr,
  828. int __user *child_tidptr,
  829. int pid)
  830. {
  831. int retval;
  832. struct task_struct *p = NULL;
  833. if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
  834. return ERR_PTR(-EINVAL);
  835. /*
  836. * Thread groups must share signals as well, and detached threads
  837. * can only be started up within the thread group.
  838. */
  839. if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
  840. return ERR_PTR(-EINVAL);
  841. /*
  842. * Shared signal handlers imply shared VM. By way of the above,
  843. * thread groups also imply shared VM. Blocking this case allows
  844. * for various simplifications in other code.
  845. */
  846. if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
  847. return ERR_PTR(-EINVAL);
  848. retval = security_task_create(clone_flags);
  849. if (retval)
  850. goto fork_out;
  851. retval = -ENOMEM;
  852. p = dup_task_struct(current);
  853. if (!p)
  854. goto fork_out;
  855. rt_mutex_init_task(p);
  856. #ifdef CONFIG_TRACE_IRQFLAGS
  857. DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
  858. DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
  859. #endif
  860. retval = -EAGAIN;
  861. if (atomic_read(&p->user->processes) >=
  862. p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
  863. if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
  864. p->user != &root_user)
  865. goto bad_fork_free;
  866. }
  867. atomic_inc(&p->user->__count);
  868. atomic_inc(&p->user->processes);
  869. get_group_info(p->group_info);
  870. /*
  871. * If multiple threads are within copy_process(), then this check
  872. * triggers too late. This doesn't hurt, the check is only there
  873. * to stop root fork bombs.
  874. */
  875. if (nr_threads >= max_threads)
  876. goto bad_fork_cleanup_count;
  877. if (!try_module_get(task_thread_info(p)->exec_domain->module))
  878. goto bad_fork_cleanup_count;
  879. if (p->binfmt && !try_module_get(p->binfmt->module))
  880. goto bad_fork_cleanup_put_domain;
  881. p->did_exec = 0;
  882. delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
  883. copy_flags(clone_flags, p);
  884. p->pid = pid;
  885. retval = -EFAULT;
  886. if (clone_flags & CLONE_PARENT_SETTID)
  887. if (put_user(p->pid, parent_tidptr))
  888. goto bad_fork_cleanup_delays_binfmt;
  889. INIT_LIST_HEAD(&p->children);
  890. INIT_LIST_HEAD(&p->sibling);
  891. p->vfork_done = NULL;
  892. spin_lock_init(&p->alloc_lock);
  893. clear_tsk_thread_flag(p, TIF_SIGPENDING);
  894. init_sigpending(&p->pending);
  895. p->utime = cputime_zero;
  896. p->stime = cputime_zero;
  897. p->sched_time = 0;
  898. p->rchar = 0; /* I/O counter: bytes read */
  899. p->wchar = 0; /* I/O counter: bytes written */
  900. p->syscr = 0; /* I/O counter: read syscalls */
  901. p->syscw = 0; /* I/O counter: write syscalls */
  902. task_io_accounting_init(p);
  903. acct_clear_integrals(p);
  904. p->it_virt_expires = cputime_zero;
  905. p->it_prof_expires = cputime_zero;
  906. p->it_sched_expires = 0;
  907. INIT_LIST_HEAD(&p->cpu_timers[0]);
  908. INIT_LIST_HEAD(&p->cpu_timers[1]);
  909. INIT_LIST_HEAD(&p->cpu_timers[2]);
  910. p->lock_depth = -1; /* -1 = no lock */
  911. do_posix_clock_monotonic_gettime(&p->start_time);
  912. p->security = NULL;
  913. p->io_context = NULL;
  914. p->io_wait = NULL;
  915. p->audit_context = NULL;
  916. cpuset_fork(p);
  917. #ifdef CONFIG_NUMA
  918. p->mempolicy = mpol_copy(p->mempolicy);
  919. if (IS_ERR(p->mempolicy)) {
  920. retval = PTR_ERR(p->mempolicy);
  921. p->mempolicy = NULL;
  922. goto bad_fork_cleanup_cpuset;
  923. }
  924. mpol_fix_fork_child_flag(p);
  925. #endif
  926. #ifdef CONFIG_TRACE_IRQFLAGS
  927. p->irq_events = 0;
  928. #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
  929. p->hardirqs_enabled = 1;
  930. #else
  931. p->hardirqs_enabled = 0;
  932. #endif
  933. p->hardirq_enable_ip = 0;
  934. p->hardirq_enable_event = 0;
  935. p->hardirq_disable_ip = _THIS_IP_;
  936. p->hardirq_disable_event = 0;
  937. p->softirqs_enabled = 1;
  938. p->softirq_enable_ip = _THIS_IP_;
  939. p->softirq_enable_event = 0;
  940. p->softirq_disable_ip = 0;
  941. p->softirq_disable_event = 0;
  942. p->hardirq_context = 0;
  943. p->softirq_context = 0;
  944. #endif
  945. #ifdef CONFIG_LOCKDEP
  946. p->lockdep_depth = 0; /* no locks held yet */
  947. p->curr_chain_key = 0;
  948. p->lockdep_recursion = 0;
  949. #endif
  950. #ifdef CONFIG_DEBUG_MUTEXES
  951. p->blocked_on = NULL; /* not blocked yet */
  952. #endif
  953. p->tgid = p->pid;
  954. if (clone_flags & CLONE_THREAD)
  955. p->tgid = current->tgid;
  956. if ((retval = security_task_alloc(p)))
  957. goto bad_fork_cleanup_policy;
  958. if ((retval = audit_alloc(p)))
  959. goto bad_fork_cleanup_security;
  960. /* copy all the process information */
  961. if ((retval = copy_semundo(clone_flags, p)))
  962. goto bad_fork_cleanup_audit;
  963. if ((retval = copy_files(clone_flags, p)))
  964. goto bad_fork_cleanup_semundo;
  965. if ((retval = copy_fs(clone_flags, p)))
  966. goto bad_fork_cleanup_files;
  967. if ((retval = copy_sighand(clone_flags, p)))
  968. goto bad_fork_cleanup_fs;
  969. if ((retval = copy_signal(clone_flags, p)))
  970. goto bad_fork_cleanup_sighand;
  971. if ((retval = copy_mm(clone_flags, p)))
  972. goto bad_fork_cleanup_signal;
  973. if ((retval = copy_keys(clone_flags, p)))
  974. goto bad_fork_cleanup_mm;
  975. if ((retval = copy_namespaces(clone_flags, p)))
  976. goto bad_fork_cleanup_keys;
  977. retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs);
  978. if (retval)
  979. goto bad_fork_cleanup_namespaces;
  980. p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
  981. /*
  982. * Clear TID on mm_release()?
  983. */
  984. p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
  985. p->robust_list = NULL;
  986. #ifdef CONFIG_COMPAT
  987. p->compat_robust_list = NULL;
  988. #endif
  989. INIT_LIST_HEAD(&p->pi_state_list);
  990. p->pi_state_cache = NULL;
  991. /*
  992. * sigaltstack should be cleared when sharing the same VM
  993. */
  994. if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
  995. p->sas_ss_sp = p->sas_ss_size = 0;
  996. /*
  997. * Syscall tracing should be turned off in the child regardless
  998. * of CLONE_PTRACE.
  999. */
  1000. clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
  1001. #ifdef TIF_SYSCALL_EMU
  1002. clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
  1003. #endif
  1004. /* Our parent execution domain becomes current domain
  1005. These must match for thread signalling to apply */
  1006. p->parent_exec_id = p->self_exec_id;
  1007. /* ok, now we should be set up.. */
  1008. p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
  1009. p->pdeath_signal = 0;
  1010. p->exit_state = 0;
  1011. /*
  1012. * Ok, make it visible to the rest of the system.
  1013. * We dont wake it up yet.
  1014. */
  1015. p->group_leader = p;
  1016. INIT_LIST_HEAD(&p->thread_group);
  1017. INIT_LIST_HEAD(&p->ptrace_children);
  1018. INIT_LIST_HEAD(&p->ptrace_list);
  1019. /* Perform scheduler related setup. Assign this task to a CPU. */
  1020. sched_fork(p, clone_flags);
  1021. /* Need tasklist lock for parent etc handling! */
  1022. write_lock_irq(&tasklist_lock);
  1023. /* for sys_ioprio_set(IOPRIO_WHO_PGRP) */
  1024. p->ioprio = current->ioprio;
  1025. /*
  1026. * The task hasn't been attached yet, so its cpus_allowed mask will
  1027. * not be changed, nor will its assigned CPU.
  1028. *
  1029. * The cpus_allowed mask of the parent may have changed after it was
  1030. * copied first time - so re-copy it here, then check the child's CPU
  1031. * to ensure it is on a valid CPU (and if not, just force it back to
  1032. * parent's CPU). This avoids alot of nasty races.
  1033. */
  1034. p->cpus_allowed = current->cpus_allowed;
  1035. if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) ||
  1036. !cpu_online(task_cpu(p))))
  1037. set_task_cpu(p, smp_processor_id());
  1038. /* CLONE_PARENT re-uses the old parent */
  1039. if (clone_flags & (CLONE_PARENT|CLONE_THREAD))
  1040. p->real_parent = current->real_parent;
  1041. else
  1042. p->real_parent = current;
  1043. p->parent = p->real_parent;
  1044. spin_lock(&current->sighand->siglock);
  1045. /*
  1046. * Process group and session signals need to be delivered to just the
  1047. * parent before the fork or both the parent and the child after the
  1048. * fork. Restart if a signal comes in before we add the new process to
  1049. * it's process group.
  1050. * A fatal signal pending means that current will exit, so the new
  1051. * thread can't slip out of an OOM kill (or normal SIGKILL).
  1052. */
  1053. recalc_sigpending();
  1054. if (signal_pending(current)) {
  1055. spin_unlock(&current->sighand->siglock);
  1056. write_unlock_irq(&tasklist_lock);
  1057. retval = -ERESTARTNOINTR;
  1058. goto bad_fork_cleanup_namespaces;
  1059. }
  1060. if (clone_flags & CLONE_THREAD) {
  1061. p->group_leader = current->group_leader;
  1062. list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
  1063. if (!cputime_eq(current->signal->it_virt_expires,
  1064. cputime_zero) ||
  1065. !cputime_eq(current->signal->it_prof_expires,
  1066. cputime_zero) ||
  1067. current->signal->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY ||
  1068. !list_empty(&current->signal->cpu_timers[0]) ||
  1069. !list_empty(&current->signal->cpu_timers[1]) ||
  1070. !list_empty(&current->signal->cpu_timers[2])) {
  1071. /*
  1072. * Have child wake up on its first tick to check
  1073. * for process CPU timers.
  1074. */
  1075. p->it_prof_expires = jiffies_to_cputime(1);
  1076. }
  1077. }
  1078. if (likely(p->pid)) {
  1079. add_parent(p);
  1080. if (unlikely(p->ptrace & PT_PTRACED))
  1081. __ptrace_link(p, current->parent);
  1082. if (thread_group_leader(p)) {
  1083. p->signal->tty = current->signal->tty;
  1084. p->signal->pgrp = process_group(current);
  1085. set_signal_session(p->signal, process_session(current));
  1086. attach_pid(p, PIDTYPE_PGID, process_group(p));
  1087. attach_pid(p, PIDTYPE_SID, process_session(p));
  1088. list_add_tail_rcu(&p->tasks, &init_task.tasks);
  1089. __get_cpu_var(process_counts)++;
  1090. }
  1091. attach_pid(p, PIDTYPE_PID, p->pid);
  1092. nr_threads++;
  1093. }
  1094. total_forks++;
  1095. spin_unlock(&current->sighand->siglock);
  1096. write_unlock_irq(&tasklist_lock);
  1097. proc_fork_connector(p);
  1098. return p;
  1099. bad_fork_cleanup_namespaces:
  1100. exit_task_namespaces(p);
  1101. bad_fork_cleanup_keys:
  1102. exit_keys(p);
  1103. bad_fork_cleanup_mm:
  1104. if (p->mm)
  1105. mmput(p->mm);
  1106. bad_fork_cleanup_signal:
  1107. cleanup_signal(p);
  1108. bad_fork_cleanup_sighand:
  1109. __cleanup_sighand(p->sighand);
  1110. bad_fork_cleanup_fs:
  1111. exit_fs(p); /* blocking */
  1112. bad_fork_cleanup_files:
  1113. exit_files(p); /* blocking */
  1114. bad_fork_cleanup_semundo:
  1115. exit_sem(p);
  1116. bad_fork_cleanup_audit:
  1117. audit_free(p);
  1118. bad_fork_cleanup_security:
  1119. security_task_free(p);
  1120. bad_fork_cleanup_policy:
  1121. #ifdef CONFIG_NUMA
  1122. mpol_free(p->mempolicy);
  1123. bad_fork_cleanup_cpuset:
  1124. #endif
  1125. cpuset_exit(p);
  1126. bad_fork_cleanup_delays_binfmt:
  1127. delayacct_tsk_free(p);
  1128. if (p->binfmt)
  1129. module_put(p->binfmt->module);
  1130. bad_fork_cleanup_put_domain:
  1131. module_put(task_thread_info(p)->exec_domain->module);
  1132. bad_fork_cleanup_count:
  1133. put_group_info(p->group_info);
  1134. atomic_dec(&p->user->processes);
  1135. free_uid(p->user);
  1136. bad_fork_free:
  1137. free_task(p);
  1138. fork_out:
  1139. return ERR_PTR(retval);
  1140. }
  1141. noinline struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
  1142. {
  1143. memset(regs, 0, sizeof(struct pt_regs));
  1144. return regs;
  1145. }
  1146. struct task_struct * __devinit fork_idle(int cpu)
  1147. {
  1148. struct task_struct *task;
  1149. struct pt_regs regs;
  1150. task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL, NULL, 0);
  1151. if (!IS_ERR(task))
  1152. init_idle(task, cpu);
  1153. return task;
  1154. }
  1155. static inline int fork_traceflag (unsigned clone_flags)
  1156. {
  1157. if (clone_flags & CLONE_UNTRACED)
  1158. return 0;
  1159. else if (clone_flags & CLONE_VFORK) {
  1160. if (current->ptrace & PT_TRACE_VFORK)
  1161. return PTRACE_EVENT_VFORK;
  1162. } else if ((clone_flags & CSIGNAL) != SIGCHLD) {
  1163. if (current->ptrace & PT_TRACE_CLONE)
  1164. return PTRACE_EVENT_CLONE;
  1165. } else if (current->ptrace & PT_TRACE_FORK)
  1166. return PTRACE_EVENT_FORK;
  1167. return 0;
  1168. }
  1169. /*
  1170. * Ok, this is the main fork-routine.
  1171. *
  1172. * It copies the process, and if successful kick-starts
  1173. * it and waits for it to finish using the VM if required.
  1174. */
  1175. long do_fork(unsigned long clone_flags,
  1176. unsigned long stack_start,
  1177. struct pt_regs *regs,
  1178. unsigned long stack_size,
  1179. int __user *parent_tidptr,
  1180. int __user *child_tidptr)
  1181. {
  1182. struct task_struct *p;
  1183. int trace = 0;
  1184. struct pid *pid = alloc_pid();
  1185. long nr;
  1186. if (!pid)
  1187. return -EAGAIN;
  1188. nr = pid->nr;
  1189. if (unlikely(current->ptrace)) {
  1190. trace = fork_traceflag (clone_flags);
  1191. if (trace)
  1192. clone_flags |= CLONE_PTRACE;
  1193. }
  1194. p = copy_process(clone_flags, stack_start, regs, stack_size, parent_tidptr, child_tidptr, nr);
  1195. /*
  1196. * Do this prior waking up the new thread - the thread pointer
  1197. * might get invalid after that point, if the thread exits quickly.
  1198. */
  1199. if (!IS_ERR(p)) {
  1200. struct completion vfork;
  1201. if (clone_flags & CLONE_VFORK) {
  1202. p->vfork_done = &vfork;
  1203. init_completion(&vfork);
  1204. }
  1205. if ((p->ptrace & PT_PTRACED) || (clone_flags & CLONE_STOPPED)) {
  1206. /*
  1207. * We'll start up with an immediate SIGSTOP.
  1208. */
  1209. sigaddset(&p->pending.signal, SIGSTOP);
  1210. set_tsk_thread_flag(p, TIF_SIGPENDING);
  1211. }
  1212. if (!(clone_flags & CLONE_STOPPED))
  1213. wake_up_new_task(p, clone_flags);
  1214. else
  1215. p->state = TASK_STOPPED;
  1216. if (unlikely (trace)) {
  1217. current->ptrace_message = nr;
  1218. ptrace_notify ((trace << 8) | SIGTRAP);
  1219. }
  1220. if (clone_flags & CLONE_VFORK) {
  1221. wait_for_completion(&vfork);
  1222. if (unlikely (current->ptrace & PT_TRACE_VFORK_DONE)) {
  1223. current->ptrace_message = nr;
  1224. ptrace_notify ((PTRACE_EVENT_VFORK_DONE << 8) | SIGTRAP);
  1225. }
  1226. }
  1227. } else {
  1228. free_pid(pid);
  1229. nr = PTR_ERR(p);
  1230. }
  1231. return nr;
  1232. }
  1233. #ifndef ARCH_MIN_MMSTRUCT_ALIGN
  1234. #define ARCH_MIN_MMSTRUCT_ALIGN 0
  1235. #endif
  1236. static void sighand_ctor(void *data, struct kmem_cache *cachep, unsigned long flags)
  1237. {
  1238. struct sighand_struct *sighand = data;
  1239. if ((flags & (SLAB_CTOR_VERIFY | SLAB_CTOR_CONSTRUCTOR)) ==
  1240. SLAB_CTOR_CONSTRUCTOR)
  1241. spin_lock_init(&sighand->siglock);
  1242. }
  1243. void __init proc_caches_init(void)
  1244. {
  1245. sighand_cachep = kmem_cache_create("sighand_cache",
  1246. sizeof(struct sighand_struct), 0,
  1247. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU,
  1248. sighand_ctor, NULL);
  1249. signal_cachep = kmem_cache_create("signal_cache",
  1250. sizeof(struct signal_struct), 0,
  1251. SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
  1252. files_cachep = kmem_cache_create("files_cache",
  1253. sizeof(struct files_struct), 0,
  1254. SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
  1255. fs_cachep = kmem_cache_create("fs_cache",
  1256. sizeof(struct fs_struct), 0,
  1257. SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
  1258. vm_area_cachep = kmem_cache_create("vm_area_struct",
  1259. sizeof(struct vm_area_struct), 0,
  1260. SLAB_PANIC, NULL, NULL);
  1261. mm_cachep = kmem_cache_create("mm_struct",
  1262. sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
  1263. SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
  1264. }
  1265. /*
  1266. * Check constraints on flags passed to the unshare system call and
  1267. * force unsharing of additional process context as appropriate.
  1268. */
  1269. static inline void check_unshare_flags(unsigned long *flags_ptr)
  1270. {
  1271. /*
  1272. * If unsharing a thread from a thread group, must also
  1273. * unshare vm.
  1274. */
  1275. if (*flags_ptr & CLONE_THREAD)
  1276. *flags_ptr |= CLONE_VM;
  1277. /*
  1278. * If unsharing vm, must also unshare signal handlers.
  1279. */
  1280. if (*flags_ptr & CLONE_VM)
  1281. *flags_ptr |= CLONE_SIGHAND;
  1282. /*
  1283. * If unsharing signal handlers and the task was created
  1284. * using CLONE_THREAD, then must unshare the thread
  1285. */
  1286. if ((*flags_ptr & CLONE_SIGHAND) &&
  1287. (atomic_read(&current->signal->count) > 1))
  1288. *flags_ptr |= CLONE_THREAD;
  1289. /*
  1290. * If unsharing namespace, must also unshare filesystem information.
  1291. */
  1292. if (*flags_ptr & CLONE_NEWNS)
  1293. *flags_ptr |= CLONE_FS;
  1294. }
  1295. /*
  1296. * Unsharing of tasks created with CLONE_THREAD is not supported yet
  1297. */
  1298. static int unshare_thread(unsigned long unshare_flags)
  1299. {
  1300. if (unshare_flags & CLONE_THREAD)
  1301. return -EINVAL;
  1302. return 0;
  1303. }
  1304. /*
  1305. * Unshare the filesystem structure if it is being shared
  1306. */
  1307. static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
  1308. {
  1309. struct fs_struct *fs = current->fs;
  1310. if ((unshare_flags & CLONE_FS) &&
  1311. (fs && atomic_read(&fs->count) > 1)) {
  1312. *new_fsp = __copy_fs_struct(current->fs);
  1313. if (!*new_fsp)
  1314. return -ENOMEM;
  1315. }
  1316. return 0;
  1317. }
  1318. /*
  1319. * Unshare the mnt_namespace structure if it is being shared
  1320. */
  1321. static int unshare_mnt_namespace(unsigned long unshare_flags,
  1322. struct mnt_namespace **new_nsp, struct fs_struct *new_fs)
  1323. {
  1324. struct mnt_namespace *ns = current->nsproxy->mnt_ns;
  1325. if ((unshare_flags & CLONE_NEWNS) && ns) {
  1326. if (!capable(CAP_SYS_ADMIN))
  1327. return -EPERM;
  1328. *new_nsp = dup_mnt_ns(current, new_fs ? new_fs : current->fs);
  1329. if (!*new_nsp)
  1330. return -ENOMEM;
  1331. }
  1332. return 0;
  1333. }
  1334. /*
  1335. * Unsharing of sighand is not supported yet
  1336. */
  1337. static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
  1338. {
  1339. struct sighand_struct *sigh = current->sighand;
  1340. if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
  1341. return -EINVAL;
  1342. else
  1343. return 0;
  1344. }
  1345. /*
  1346. * Unshare vm if it is being shared
  1347. */
  1348. static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
  1349. {
  1350. struct mm_struct *mm = current->mm;
  1351. if ((unshare_flags & CLONE_VM) &&
  1352. (mm && atomic_read(&mm->mm_users) > 1)) {
  1353. return -EINVAL;
  1354. }
  1355. return 0;
  1356. }
  1357. /*
  1358. * Unshare file descriptor table if it is being shared
  1359. */
  1360. static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
  1361. {
  1362. struct files_struct *fd = current->files;
  1363. int error = 0;
  1364. if ((unshare_flags & CLONE_FILES) &&
  1365. (fd && atomic_read(&fd->count) > 1)) {
  1366. *new_fdp = dup_fd(fd, &error);
  1367. if (!*new_fdp)
  1368. return error;
  1369. }
  1370. return 0;
  1371. }
  1372. /*
  1373. * Unsharing of semundo for tasks created with CLONE_SYSVSEM is not
  1374. * supported yet
  1375. */
  1376. static int unshare_semundo(unsigned long unshare_flags, struct sem_undo_list **new_ulistp)
  1377. {
  1378. if (unshare_flags & CLONE_SYSVSEM)
  1379. return -EINVAL;
  1380. return 0;
  1381. }
  1382. #ifndef CONFIG_IPC_NS
  1383. static inline int unshare_ipcs(unsigned long flags, struct ipc_namespace **ns)
  1384. {
  1385. if (flags & CLONE_NEWIPC)
  1386. return -EINVAL;
  1387. return 0;
  1388. }
  1389. #endif
  1390. /*
  1391. * unshare allows a process to 'unshare' part of the process
  1392. * context which was originally shared using clone. copy_*
  1393. * functions used by do_fork() cannot be used here directly
  1394. * because they modify an inactive task_struct that is being
  1395. * constructed. Here we are modifying the current, active,
  1396. * task_struct.
  1397. */
  1398. asmlinkage long sys_unshare(unsigned long unshare_flags)
  1399. {
  1400. int err = 0;
  1401. struct fs_struct *fs, *new_fs = NULL;
  1402. struct mnt_namespace *ns, *new_ns = NULL;
  1403. struct sighand_struct *new_sigh = NULL;
  1404. struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
  1405. struct files_struct *fd, *new_fd = NULL;
  1406. struct sem_undo_list *new_ulist = NULL;
  1407. struct nsproxy *new_nsproxy = NULL, *old_nsproxy = NULL;
  1408. struct uts_namespace *uts, *new_uts = NULL;
  1409. struct ipc_namespace *ipc, *new_ipc = NULL;
  1410. check_unshare_flags(&unshare_flags);
  1411. /* Return -EINVAL for all unsupported flags */
  1412. err = -EINVAL;
  1413. if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
  1414. CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
  1415. CLONE_NEWUTS|CLONE_NEWIPC))
  1416. goto bad_unshare_out;
  1417. if ((err = unshare_thread(unshare_flags)))
  1418. goto bad_unshare_out;
  1419. if ((err = unshare_fs(unshare_flags, &new_fs)))
  1420. goto bad_unshare_cleanup_thread;
  1421. if ((err = unshare_mnt_namespace(unshare_flags, &new_ns, new_fs)))
  1422. goto bad_unshare_cleanup_fs;
  1423. if ((err = unshare_sighand(unshare_flags, &new_sigh)))
  1424. goto bad_unshare_cleanup_ns;
  1425. if ((err = unshare_vm(unshare_flags, &new_mm)))
  1426. goto bad_unshare_cleanup_sigh;
  1427. if ((err = unshare_fd(unshare_flags, &new_fd)))
  1428. goto bad_unshare_cleanup_vm;
  1429. if ((err = unshare_semundo(unshare_flags, &new_ulist)))
  1430. goto bad_unshare_cleanup_fd;
  1431. if ((err = unshare_utsname(unshare_flags, &new_uts)))
  1432. goto bad_unshare_cleanup_semundo;
  1433. if ((err = unshare_ipcs(unshare_flags, &new_ipc)))
  1434. goto bad_unshare_cleanup_uts;
  1435. if (new_ns || new_uts || new_ipc) {
  1436. old_nsproxy = current->nsproxy;
  1437. new_nsproxy = dup_namespaces(old_nsproxy);
  1438. if (!new_nsproxy) {
  1439. err = -ENOMEM;
  1440. goto bad_unshare_cleanup_ipc;
  1441. }
  1442. }
  1443. if (new_fs || new_ns || new_mm || new_fd || new_ulist ||
  1444. new_uts || new_ipc) {
  1445. task_lock(current);
  1446. if (new_nsproxy) {
  1447. current->nsproxy = new_nsproxy;
  1448. new_nsproxy = old_nsproxy;
  1449. }
  1450. if (new_fs) {
  1451. fs = current->fs;
  1452. current->fs = new_fs;
  1453. new_fs = fs;
  1454. }
  1455. if (new_ns) {
  1456. ns = current->nsproxy->mnt_ns;
  1457. current->nsproxy->mnt_ns = new_ns;
  1458. new_ns = ns;
  1459. }
  1460. if (new_mm) {
  1461. mm = current->mm;
  1462. active_mm = current->active_mm;
  1463. current->mm = new_mm;
  1464. current->active_mm = new_mm;
  1465. activate_mm(active_mm, new_mm);
  1466. new_mm = mm;
  1467. }
  1468. if (new_fd) {
  1469. fd = current->files;
  1470. current->files = new_fd;
  1471. new_fd = fd;
  1472. }
  1473. if (new_uts) {
  1474. uts = current->nsproxy->uts_ns;
  1475. current->nsproxy->uts_ns = new_uts;
  1476. new_uts = uts;
  1477. }
  1478. if (new_ipc) {
  1479. ipc = current->nsproxy->ipc_ns;
  1480. current->nsproxy->ipc_ns = new_ipc;
  1481. new_ipc = ipc;
  1482. }
  1483. task_unlock(current);
  1484. }
  1485. if (new_nsproxy)
  1486. put_nsproxy(new_nsproxy);
  1487. bad_unshare_cleanup_ipc:
  1488. if (new_ipc)
  1489. put_ipc_ns(new_ipc);
  1490. bad_unshare_cleanup_uts:
  1491. if (new_uts)
  1492. put_uts_ns(new_uts);
  1493. bad_unshare_cleanup_semundo:
  1494. bad_unshare_cleanup_fd:
  1495. if (new_fd)
  1496. put_files_struct(new_fd);
  1497. bad_unshare_cleanup_vm:
  1498. if (new_mm)
  1499. mmput(new_mm);
  1500. bad_unshare_cleanup_sigh:
  1501. if (new_sigh)
  1502. if (atomic_dec_and_test(&new_sigh->count))
  1503. kmem_cache_free(sighand_cachep, new_sigh);
  1504. bad_unshare_cleanup_ns:
  1505. if (new_ns)
  1506. put_mnt_ns(new_ns);
  1507. bad_unshare_cleanup_fs:
  1508. if (new_fs)
  1509. put_fs_struct(new_fs);
  1510. bad_unshare_cleanup_thread:
  1511. bad_unshare_out:
  1512. return err;
  1513. }