fork.c 63 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657
  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/sched/autogroup.h>
  14. #include <linux/sched/mm.h>
  15. #include <linux/sched/coredump.h>
  16. #include <linux/sched/user.h>
  17. #include <linux/sched/numa_balancing.h>
  18. #include <linux/sched/stat.h>
  19. #include <linux/sched/task.h>
  20. #include <linux/sched/task_stack.h>
  21. #include <linux/sched/cputime.h>
  22. #include <linux/rtmutex.h>
  23. #include <linux/init.h>
  24. #include <linux/unistd.h>
  25. #include <linux/module.h>
  26. #include <linux/vmalloc.h>
  27. #include <linux/completion.h>
  28. #include <linux/personality.h>
  29. #include <linux/mempolicy.h>
  30. #include <linux/sem.h>
  31. #include <linux/file.h>
  32. #include <linux/fdtable.h>
  33. #include <linux/iocontext.h>
  34. #include <linux/key.h>
  35. #include <linux/binfmts.h>
  36. #include <linux/mman.h>
  37. #include <linux/mmu_notifier.h>
  38. #include <linux/hmm.h>
  39. #include <linux/fs.h>
  40. #include <linux/mm.h>
  41. #include <linux/vmacache.h>
  42. #include <linux/nsproxy.h>
  43. #include <linux/capability.h>
  44. #include <linux/cpu.h>
  45. #include <linux/cgroup.h>
  46. #include <linux/security.h>
  47. #include <linux/hugetlb.h>
  48. #include <linux/seccomp.h>
  49. #include <linux/swap.h>
  50. #include <linux/syscalls.h>
  51. #include <linux/jiffies.h>
  52. #include <linux/futex.h>
  53. #include <linux/compat.h>
  54. #include <linux/kthread.h>
  55. #include <linux/task_io_accounting_ops.h>
  56. #include <linux/rcupdate.h>
  57. #include <linux/ptrace.h>
  58. #include <linux/mount.h>
  59. #include <linux/audit.h>
  60. #include <linux/memcontrol.h>
  61. #include <linux/ftrace.h>
  62. #include <linux/proc_fs.h>
  63. #include <linux/profile.h>
  64. #include <linux/rmap.h>
  65. #include <linux/ksm.h>
  66. #include <linux/acct.h>
  67. #include <linux/userfaultfd_k.h>
  68. #include <linux/tsacct_kern.h>
  69. #include <linux/cn_proc.h>
  70. #include <linux/freezer.h>
  71. #include <linux/delayacct.h>
  72. #include <linux/taskstats_kern.h>
  73. #include <linux/random.h>
  74. #include <linux/tty.h>
  75. #include <linux/blkdev.h>
  76. #include <linux/fs_struct.h>
  77. #include <linux/magic.h>
  78. #include <linux/sched/mm.h>
  79. #include <linux/perf_event.h>
  80. #include <linux/posix-timers.h>
  81. #include <linux/user-return-notifier.h>
  82. #include <linux/oom.h>
  83. #include <linux/khugepaged.h>
  84. #include <linux/signalfd.h>
  85. #include <linux/uprobes.h>
  86. #include <linux/aio.h>
  87. #include <linux/compiler.h>
  88. #include <linux/sysctl.h>
  89. #include <linux/kcov.h>
  90. #include <linux/livepatch.h>
  91. #include <linux/thread_info.h>
  92. #include <asm/pgtable.h>
  93. #include <asm/pgalloc.h>
  94. #include <linux/uaccess.h>
  95. #include <asm/mmu_context.h>
  96. #include <asm/cacheflush.h>
  97. #include <asm/tlbflush.h>
  98. #include <trace/events/sched.h>
  99. #define CREATE_TRACE_POINTS
  100. #include <trace/events/task.h>
  101. /*
  102. * Minimum number of threads to boot the kernel
  103. */
  104. #define MIN_THREADS 20
  105. /*
  106. * Maximum number of threads
  107. */
  108. #define MAX_THREADS FUTEX_TID_MASK
  109. /*
  110. * Protected counters by write_lock_irq(&tasklist_lock)
  111. */
  112. unsigned long total_forks; /* Handle normal Linux uptimes. */
  113. int nr_threads; /* The idle threads do not count.. */
  114. int max_threads; /* tunable limit on nr_threads */
  115. DEFINE_PER_CPU(unsigned long, process_counts) = 0;
  116. __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
  117. #ifdef CONFIG_PROVE_RCU
  118. int lockdep_tasklist_lock_is_held(void)
  119. {
  120. return lockdep_is_held(&tasklist_lock);
  121. }
  122. EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
  123. #endif /* #ifdef CONFIG_PROVE_RCU */
  124. int nr_processes(void)
  125. {
  126. int cpu;
  127. int total = 0;
  128. for_each_possible_cpu(cpu)
  129. total += per_cpu(process_counts, cpu);
  130. return total;
  131. }
  132. void __weak arch_release_task_struct(struct task_struct *tsk)
  133. {
  134. }
  135. #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
  136. static struct kmem_cache *task_struct_cachep;
  137. static inline struct task_struct *alloc_task_struct_node(int node)
  138. {
  139. return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
  140. }
  141. static inline void free_task_struct(struct task_struct *tsk)
  142. {
  143. kmem_cache_free(task_struct_cachep, tsk);
  144. }
  145. #endif
  146. #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
  147. /*
  148. * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
  149. * kmemcache based allocator.
  150. */
  151. # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
  152. #ifdef CONFIG_VMAP_STACK
  153. /*
  154. * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
  155. * flush. Try to minimize the number of calls by caching stacks.
  156. */
  157. #define NR_CACHED_STACKS 2
  158. static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]);
  159. static int free_vm_stack_cache(unsigned int cpu)
  160. {
  161. struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu);
  162. int i;
  163. for (i = 0; i < NR_CACHED_STACKS; i++) {
  164. struct vm_struct *vm_stack = cached_vm_stacks[i];
  165. if (!vm_stack)
  166. continue;
  167. vfree(vm_stack->addr);
  168. cached_vm_stacks[i] = NULL;
  169. }
  170. return 0;
  171. }
  172. #endif
  173. static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
  174. {
  175. #ifdef CONFIG_VMAP_STACK
  176. void *stack;
  177. int i;
  178. for (i = 0; i < NR_CACHED_STACKS; i++) {
  179. struct vm_struct *s;
  180. s = this_cpu_xchg(cached_stacks[i], NULL);
  181. if (!s)
  182. continue;
  183. /* Clear stale pointers from reused stack. */
  184. memset(s->addr, 0, THREAD_SIZE);
  185. tsk->stack_vm_area = s;
  186. tsk->stack = s->addr;
  187. return s->addr;
  188. }
  189. stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN,
  190. VMALLOC_START, VMALLOC_END,
  191. THREADINFO_GFP,
  192. PAGE_KERNEL,
  193. 0, node, __builtin_return_address(0));
  194. /*
  195. * We can't call find_vm_area() in interrupt context, and
  196. * free_thread_stack() can be called in interrupt context,
  197. * so cache the vm_struct.
  198. */
  199. if (stack) {
  200. tsk->stack_vm_area = find_vm_area(stack);
  201. tsk->stack = stack;
  202. }
  203. return stack;
  204. #else
  205. struct page *page = alloc_pages_node(node, THREADINFO_GFP,
  206. THREAD_SIZE_ORDER);
  207. if (likely(page)) {
  208. tsk->stack = page_address(page);
  209. return tsk->stack;
  210. }
  211. return NULL;
  212. #endif
  213. }
  214. static inline void free_thread_stack(struct task_struct *tsk)
  215. {
  216. #ifdef CONFIG_VMAP_STACK
  217. if (task_stack_vm_area(tsk)) {
  218. int i;
  219. for (i = 0; i < NR_CACHED_STACKS; i++) {
  220. if (this_cpu_cmpxchg(cached_stacks[i],
  221. NULL, tsk->stack_vm_area) != NULL)
  222. continue;
  223. return;
  224. }
  225. vfree_atomic(tsk->stack);
  226. return;
  227. }
  228. #endif
  229. __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
  230. }
  231. # else
  232. static struct kmem_cache *thread_stack_cache;
  233. static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
  234. int node)
  235. {
  236. unsigned long *stack;
  237. stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
  238. tsk->stack = stack;
  239. return stack;
  240. }
  241. static void free_thread_stack(struct task_struct *tsk)
  242. {
  243. kmem_cache_free(thread_stack_cache, tsk->stack);
  244. }
  245. void thread_stack_cache_init(void)
  246. {
  247. thread_stack_cache = kmem_cache_create_usercopy("thread_stack",
  248. THREAD_SIZE, THREAD_SIZE, 0, 0,
  249. THREAD_SIZE, NULL);
  250. BUG_ON(thread_stack_cache == NULL);
  251. }
  252. # endif
  253. #endif
  254. /* SLAB cache for signal_struct structures (tsk->signal) */
  255. static struct kmem_cache *signal_cachep;
  256. /* SLAB cache for sighand_struct structures (tsk->sighand) */
  257. struct kmem_cache *sighand_cachep;
  258. /* SLAB cache for files_struct structures (tsk->files) */
  259. struct kmem_cache *files_cachep;
  260. /* SLAB cache for fs_struct structures (tsk->fs) */
  261. struct kmem_cache *fs_cachep;
  262. /* SLAB cache for vm_area_struct structures */
  263. static struct kmem_cache *vm_area_cachep;
  264. /* SLAB cache for mm_struct structures (tsk->mm) */
  265. static struct kmem_cache *mm_cachep;
  266. struct vm_area_struct *vm_area_alloc(struct mm_struct *mm)
  267. {
  268. struct vm_area_struct *vma;
  269. vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
  270. if (vma)
  271. vma_init(vma, mm);
  272. return vma;
  273. }
  274. struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
  275. {
  276. struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
  277. if (new) {
  278. *new = *orig;
  279. INIT_LIST_HEAD(&new->anon_vma_chain);
  280. }
  281. return new;
  282. }
  283. void vm_area_free(struct vm_area_struct *vma)
  284. {
  285. kmem_cache_free(vm_area_cachep, vma);
  286. }
  287. static void account_kernel_stack(struct task_struct *tsk, int account)
  288. {
  289. void *stack = task_stack_page(tsk);
  290. struct vm_struct *vm = task_stack_vm_area(tsk);
  291. BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0);
  292. if (vm) {
  293. int i;
  294. BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE);
  295. for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
  296. mod_zone_page_state(page_zone(vm->pages[i]),
  297. NR_KERNEL_STACK_KB,
  298. PAGE_SIZE / 1024 * account);
  299. }
  300. /* All stack pages belong to the same memcg. */
  301. mod_memcg_page_state(vm->pages[0], MEMCG_KERNEL_STACK_KB,
  302. account * (THREAD_SIZE / 1024));
  303. } else {
  304. /*
  305. * All stack pages are in the same zone and belong to the
  306. * same memcg.
  307. */
  308. struct page *first_page = virt_to_page(stack);
  309. mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB,
  310. THREAD_SIZE / 1024 * account);
  311. mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB,
  312. account * (THREAD_SIZE / 1024));
  313. }
  314. }
  315. static void release_task_stack(struct task_struct *tsk)
  316. {
  317. if (WARN_ON(tsk->state != TASK_DEAD))
  318. return; /* Better to leak the stack than to free prematurely */
  319. account_kernel_stack(tsk, -1);
  320. free_thread_stack(tsk);
  321. tsk->stack = NULL;
  322. #ifdef CONFIG_VMAP_STACK
  323. tsk->stack_vm_area = NULL;
  324. #endif
  325. }
  326. #ifdef CONFIG_THREAD_INFO_IN_TASK
  327. void put_task_stack(struct task_struct *tsk)
  328. {
  329. if (atomic_dec_and_test(&tsk->stack_refcount))
  330. release_task_stack(tsk);
  331. }
  332. #endif
  333. void free_task(struct task_struct *tsk)
  334. {
  335. #ifndef CONFIG_THREAD_INFO_IN_TASK
  336. /*
  337. * The task is finally done with both the stack and thread_info,
  338. * so free both.
  339. */
  340. release_task_stack(tsk);
  341. #else
  342. /*
  343. * If the task had a separate stack allocation, it should be gone
  344. * by now.
  345. */
  346. WARN_ON_ONCE(atomic_read(&tsk->stack_refcount) != 0);
  347. #endif
  348. rt_mutex_debug_task_free(tsk);
  349. ftrace_graph_exit_task(tsk);
  350. put_seccomp_filter(tsk);
  351. arch_release_task_struct(tsk);
  352. if (tsk->flags & PF_KTHREAD)
  353. free_kthread_struct(tsk);
  354. free_task_struct(tsk);
  355. }
  356. EXPORT_SYMBOL(free_task);
  357. #ifdef CONFIG_MMU
  358. static __latent_entropy int dup_mmap(struct mm_struct *mm,
  359. struct mm_struct *oldmm)
  360. {
  361. struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
  362. struct rb_node **rb_link, *rb_parent;
  363. int retval;
  364. unsigned long charge;
  365. LIST_HEAD(uf);
  366. uprobe_start_dup_mmap();
  367. if (down_write_killable(&oldmm->mmap_sem)) {
  368. retval = -EINTR;
  369. goto fail_uprobe_end;
  370. }
  371. flush_cache_dup_mm(oldmm);
  372. uprobe_dup_mmap(oldmm, mm);
  373. /*
  374. * Not linked in yet - no deadlock potential:
  375. */
  376. down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
  377. /* No ordering required: file already has been exposed. */
  378. RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
  379. mm->total_vm = oldmm->total_vm;
  380. mm->data_vm = oldmm->data_vm;
  381. mm->exec_vm = oldmm->exec_vm;
  382. mm->stack_vm = oldmm->stack_vm;
  383. rb_link = &mm->mm_rb.rb_node;
  384. rb_parent = NULL;
  385. pprev = &mm->mmap;
  386. retval = ksm_fork(mm, oldmm);
  387. if (retval)
  388. goto out;
  389. retval = khugepaged_fork(mm, oldmm);
  390. if (retval)
  391. goto out;
  392. prev = NULL;
  393. for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
  394. struct file *file;
  395. if (mpnt->vm_flags & VM_DONTCOPY) {
  396. vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
  397. continue;
  398. }
  399. charge = 0;
  400. /*
  401. * Don't duplicate many vmas if we've been oom-killed (for
  402. * example)
  403. */
  404. if (fatal_signal_pending(current)) {
  405. retval = -EINTR;
  406. goto out;
  407. }
  408. if (mpnt->vm_flags & VM_ACCOUNT) {
  409. unsigned long len = vma_pages(mpnt);
  410. if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
  411. goto fail_nomem;
  412. charge = len;
  413. }
  414. tmp = vm_area_dup(mpnt);
  415. if (!tmp)
  416. goto fail_nomem;
  417. retval = vma_dup_policy(mpnt, tmp);
  418. if (retval)
  419. goto fail_nomem_policy;
  420. tmp->vm_mm = mm;
  421. retval = dup_userfaultfd(tmp, &uf);
  422. if (retval)
  423. goto fail_nomem_anon_vma_fork;
  424. if (tmp->vm_flags & VM_WIPEONFORK) {
  425. /* VM_WIPEONFORK gets a clean slate in the child. */
  426. tmp->anon_vma = NULL;
  427. if (anon_vma_prepare(tmp))
  428. goto fail_nomem_anon_vma_fork;
  429. } else if (anon_vma_fork(tmp, mpnt))
  430. goto fail_nomem_anon_vma_fork;
  431. tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT);
  432. tmp->vm_next = tmp->vm_prev = NULL;
  433. file = tmp->vm_file;
  434. if (file) {
  435. struct inode *inode = file_inode(file);
  436. struct address_space *mapping = file->f_mapping;
  437. get_file(file);
  438. if (tmp->vm_flags & VM_DENYWRITE)
  439. atomic_dec(&inode->i_writecount);
  440. i_mmap_lock_write(mapping);
  441. if (tmp->vm_flags & VM_SHARED)
  442. atomic_inc(&mapping->i_mmap_writable);
  443. flush_dcache_mmap_lock(mapping);
  444. /* insert tmp into the share list, just after mpnt */
  445. vma_interval_tree_insert_after(tmp, mpnt,
  446. &mapping->i_mmap);
  447. flush_dcache_mmap_unlock(mapping);
  448. i_mmap_unlock_write(mapping);
  449. }
  450. /*
  451. * Clear hugetlb-related page reserves for children. This only
  452. * affects MAP_PRIVATE mappings. Faults generated by the child
  453. * are not guaranteed to succeed, even if read-only
  454. */
  455. if (is_vm_hugetlb_page(tmp))
  456. reset_vma_resv_huge_pages(tmp);
  457. /*
  458. * Link in the new vma and copy the page table entries.
  459. */
  460. *pprev = tmp;
  461. pprev = &tmp->vm_next;
  462. tmp->vm_prev = prev;
  463. prev = tmp;
  464. __vma_link_rb(mm, tmp, rb_link, rb_parent);
  465. rb_link = &tmp->vm_rb.rb_right;
  466. rb_parent = &tmp->vm_rb;
  467. mm->map_count++;
  468. if (!(tmp->vm_flags & VM_WIPEONFORK))
  469. retval = copy_page_range(mm, oldmm, mpnt);
  470. if (tmp->vm_ops && tmp->vm_ops->open)
  471. tmp->vm_ops->open(tmp);
  472. if (retval)
  473. goto out;
  474. }
  475. /* a new mm has just been created */
  476. retval = arch_dup_mmap(oldmm, mm);
  477. out:
  478. up_write(&mm->mmap_sem);
  479. flush_tlb_mm(oldmm);
  480. up_write(&oldmm->mmap_sem);
  481. dup_userfaultfd_complete(&uf);
  482. fail_uprobe_end:
  483. uprobe_end_dup_mmap();
  484. return retval;
  485. fail_nomem_anon_vma_fork:
  486. mpol_put(vma_policy(tmp));
  487. fail_nomem_policy:
  488. vm_area_free(tmp);
  489. fail_nomem:
  490. retval = -ENOMEM;
  491. vm_unacct_memory(charge);
  492. goto out;
  493. }
  494. static inline int mm_alloc_pgd(struct mm_struct *mm)
  495. {
  496. mm->pgd = pgd_alloc(mm);
  497. if (unlikely(!mm->pgd))
  498. return -ENOMEM;
  499. return 0;
  500. }
  501. static inline void mm_free_pgd(struct mm_struct *mm)
  502. {
  503. pgd_free(mm, mm->pgd);
  504. }
  505. #else
  506. static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
  507. {
  508. down_write(&oldmm->mmap_sem);
  509. RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
  510. up_write(&oldmm->mmap_sem);
  511. return 0;
  512. }
  513. #define mm_alloc_pgd(mm) (0)
  514. #define mm_free_pgd(mm)
  515. #endif /* CONFIG_MMU */
  516. static void check_mm(struct mm_struct *mm)
  517. {
  518. int i;
  519. for (i = 0; i < NR_MM_COUNTERS; i++) {
  520. long x = atomic_long_read(&mm->rss_stat.count[i]);
  521. if (unlikely(x))
  522. printk(KERN_ALERT "BUG: Bad rss-counter state "
  523. "mm:%p idx:%d val:%ld\n", mm, i, x);
  524. }
  525. if (mm_pgtables_bytes(mm))
  526. pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
  527. mm_pgtables_bytes(mm));
  528. #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
  529. VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
  530. #endif
  531. }
  532. #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
  533. #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
  534. /*
  535. * Called when the last reference to the mm
  536. * is dropped: either by a lazy thread or by
  537. * mmput. Free the page directory and the mm.
  538. */
  539. void __mmdrop(struct mm_struct *mm)
  540. {
  541. BUG_ON(mm == &init_mm);
  542. WARN_ON_ONCE(mm == current->mm);
  543. WARN_ON_ONCE(mm == current->active_mm);
  544. mm_free_pgd(mm);
  545. destroy_context(mm);
  546. hmm_mm_destroy(mm);
  547. mmu_notifier_mm_destroy(mm);
  548. check_mm(mm);
  549. put_user_ns(mm->user_ns);
  550. free_mm(mm);
  551. }
  552. EXPORT_SYMBOL_GPL(__mmdrop);
  553. static void mmdrop_async_fn(struct work_struct *work)
  554. {
  555. struct mm_struct *mm;
  556. mm = container_of(work, struct mm_struct, async_put_work);
  557. __mmdrop(mm);
  558. }
  559. static void mmdrop_async(struct mm_struct *mm)
  560. {
  561. if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
  562. INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
  563. schedule_work(&mm->async_put_work);
  564. }
  565. }
  566. static inline void free_signal_struct(struct signal_struct *sig)
  567. {
  568. taskstats_tgid_free(sig);
  569. sched_autogroup_exit(sig);
  570. /*
  571. * __mmdrop is not safe to call from softirq context on x86 due to
  572. * pgd_dtor so postpone it to the async context
  573. */
  574. if (sig->oom_mm)
  575. mmdrop_async(sig->oom_mm);
  576. kmem_cache_free(signal_cachep, sig);
  577. }
  578. static inline void put_signal_struct(struct signal_struct *sig)
  579. {
  580. if (atomic_dec_and_test(&sig->sigcnt))
  581. free_signal_struct(sig);
  582. }
  583. void __put_task_struct(struct task_struct *tsk)
  584. {
  585. WARN_ON(!tsk->exit_state);
  586. WARN_ON(atomic_read(&tsk->usage));
  587. WARN_ON(tsk == current);
  588. cgroup_free(tsk);
  589. task_numa_free(tsk, true);
  590. security_task_free(tsk);
  591. exit_creds(tsk);
  592. delayacct_tsk_free(tsk);
  593. put_signal_struct(tsk->signal);
  594. if (!profile_handoff_task(tsk))
  595. free_task(tsk);
  596. }
  597. EXPORT_SYMBOL_GPL(__put_task_struct);
  598. void __init __weak arch_task_cache_init(void) { }
  599. /*
  600. * set_max_threads
  601. */
  602. static void set_max_threads(unsigned int max_threads_suggested)
  603. {
  604. u64 threads;
  605. /*
  606. * The number of threads shall be limited such that the thread
  607. * structures may only consume a small part of the available memory.
  608. */
  609. if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64)
  610. threads = MAX_THREADS;
  611. else
  612. threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE,
  613. (u64) THREAD_SIZE * 8UL);
  614. if (threads > max_threads_suggested)
  615. threads = max_threads_suggested;
  616. max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
  617. }
  618. #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
  619. /* Initialized by the architecture: */
  620. int arch_task_struct_size __read_mostly;
  621. #endif
  622. static void task_struct_whitelist(unsigned long *offset, unsigned long *size)
  623. {
  624. /* Fetch thread_struct whitelist for the architecture. */
  625. arch_thread_struct_whitelist(offset, size);
  626. /*
  627. * Handle zero-sized whitelist or empty thread_struct, otherwise
  628. * adjust offset to position of thread_struct in task_struct.
  629. */
  630. if (unlikely(*size == 0))
  631. *offset = 0;
  632. else
  633. *offset += offsetof(struct task_struct, thread);
  634. }
  635. void __init fork_init(void)
  636. {
  637. int i;
  638. #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
  639. #ifndef ARCH_MIN_TASKALIGN
  640. #define ARCH_MIN_TASKALIGN 0
  641. #endif
  642. int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
  643. unsigned long useroffset, usersize;
  644. /* create a slab on which task_structs can be allocated */
  645. task_struct_whitelist(&useroffset, &usersize);
  646. task_struct_cachep = kmem_cache_create_usercopy("task_struct",
  647. arch_task_struct_size, align,
  648. SLAB_PANIC|SLAB_ACCOUNT,
  649. useroffset, usersize, NULL);
  650. #endif
  651. /* do the arch specific task caches init */
  652. arch_task_cache_init();
  653. set_max_threads(MAX_THREADS);
  654. init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
  655. init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
  656. init_task.signal->rlim[RLIMIT_SIGPENDING] =
  657. init_task.signal->rlim[RLIMIT_NPROC];
  658. for (i = 0; i < UCOUNT_COUNTS; i++) {
  659. init_user_ns.ucount_max[i] = max_threads/2;
  660. }
  661. #ifdef CONFIG_VMAP_STACK
  662. cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache",
  663. NULL, free_vm_stack_cache);
  664. #endif
  665. lockdep_init_task(&init_task);
  666. }
  667. int __weak arch_dup_task_struct(struct task_struct *dst,
  668. struct task_struct *src)
  669. {
  670. *dst = *src;
  671. return 0;
  672. }
  673. void set_task_stack_end_magic(struct task_struct *tsk)
  674. {
  675. unsigned long *stackend;
  676. stackend = end_of_stack(tsk);
  677. *stackend = STACK_END_MAGIC; /* for overflow detection */
  678. }
  679. static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
  680. {
  681. struct task_struct *tsk;
  682. unsigned long *stack;
  683. struct vm_struct *stack_vm_area;
  684. int err;
  685. if (node == NUMA_NO_NODE)
  686. node = tsk_fork_get_node(orig);
  687. tsk = alloc_task_struct_node(node);
  688. if (!tsk)
  689. return NULL;
  690. stack = alloc_thread_stack_node(tsk, node);
  691. if (!stack)
  692. goto free_tsk;
  693. stack_vm_area = task_stack_vm_area(tsk);
  694. err = arch_dup_task_struct(tsk, orig);
  695. /*
  696. * arch_dup_task_struct() clobbers the stack-related fields. Make
  697. * sure they're properly initialized before using any stack-related
  698. * functions again.
  699. */
  700. tsk->stack = stack;
  701. #ifdef CONFIG_VMAP_STACK
  702. tsk->stack_vm_area = stack_vm_area;
  703. #endif
  704. #ifdef CONFIG_THREAD_INFO_IN_TASK
  705. atomic_set(&tsk->stack_refcount, 1);
  706. #endif
  707. if (err)
  708. goto free_stack;
  709. #ifdef CONFIG_SECCOMP
  710. /*
  711. * We must handle setting up seccomp filters once we're under
  712. * the sighand lock in case orig has changed between now and
  713. * then. Until then, filter must be NULL to avoid messing up
  714. * the usage counts on the error path calling free_task.
  715. */
  716. tsk->seccomp.filter = NULL;
  717. #endif
  718. setup_thread_stack(tsk, orig);
  719. clear_user_return_notifier(tsk);
  720. clear_tsk_need_resched(tsk);
  721. set_task_stack_end_magic(tsk);
  722. #ifdef CONFIG_STACKPROTECTOR
  723. tsk->stack_canary = get_random_canary();
  724. #endif
  725. /*
  726. * One for us, one for whoever does the "release_task()" (usually
  727. * parent)
  728. */
  729. atomic_set(&tsk->usage, 2);
  730. #ifdef CONFIG_BLK_DEV_IO_TRACE
  731. tsk->btrace_seq = 0;
  732. #endif
  733. tsk->splice_pipe = NULL;
  734. tsk->task_frag.page = NULL;
  735. tsk->wake_q.next = NULL;
  736. account_kernel_stack(tsk, 1);
  737. kcov_task_init(tsk);
  738. #ifdef CONFIG_FAULT_INJECTION
  739. tsk->fail_nth = 0;
  740. #endif
  741. #ifdef CONFIG_BLK_CGROUP
  742. tsk->throttle_queue = NULL;
  743. tsk->use_memdelay = 0;
  744. #endif
  745. #ifdef CONFIG_MEMCG
  746. tsk->active_memcg = NULL;
  747. #endif
  748. return tsk;
  749. free_stack:
  750. free_thread_stack(tsk);
  751. free_tsk:
  752. free_task_struct(tsk);
  753. return NULL;
  754. }
  755. __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
  756. static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
  757. static int __init coredump_filter_setup(char *s)
  758. {
  759. default_dump_filter =
  760. (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
  761. MMF_DUMP_FILTER_MASK;
  762. return 1;
  763. }
  764. __setup("coredump_filter=", coredump_filter_setup);
  765. #include <linux/init_task.h>
  766. static void mm_init_aio(struct mm_struct *mm)
  767. {
  768. #ifdef CONFIG_AIO
  769. spin_lock_init(&mm->ioctx_lock);
  770. mm->ioctx_table = NULL;
  771. #endif
  772. }
  773. static __always_inline void mm_clear_owner(struct mm_struct *mm,
  774. struct task_struct *p)
  775. {
  776. #ifdef CONFIG_MEMCG
  777. if (mm->owner == p)
  778. WRITE_ONCE(mm->owner, NULL);
  779. #endif
  780. }
  781. static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
  782. {
  783. #ifdef CONFIG_MEMCG
  784. mm->owner = p;
  785. #endif
  786. }
  787. static void mm_init_uprobes_state(struct mm_struct *mm)
  788. {
  789. #ifdef CONFIG_UPROBES
  790. mm->uprobes_state.xol_area = NULL;
  791. #endif
  792. }
  793. static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
  794. struct user_namespace *user_ns)
  795. {
  796. mm->mmap = NULL;
  797. mm->mm_rb = RB_ROOT;
  798. mm->vmacache_seqnum = 0;
  799. atomic_set(&mm->mm_users, 1);
  800. atomic_set(&mm->mm_count, 1);
  801. init_rwsem(&mm->mmap_sem);
  802. INIT_LIST_HEAD(&mm->mmlist);
  803. mm->core_state = NULL;
  804. mm_pgtables_bytes_init(mm);
  805. mm->map_count = 0;
  806. mm->locked_vm = 0;
  807. mm->pinned_vm = 0;
  808. memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
  809. spin_lock_init(&mm->page_table_lock);
  810. spin_lock_init(&mm->arg_lock);
  811. mm_init_cpumask(mm);
  812. mm_init_aio(mm);
  813. mm_init_owner(mm, p);
  814. RCU_INIT_POINTER(mm->exe_file, NULL);
  815. mmu_notifier_mm_init(mm);
  816. hmm_mm_init(mm);
  817. init_tlb_flush_pending(mm);
  818. #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
  819. mm->pmd_huge_pte = NULL;
  820. #endif
  821. mm_init_uprobes_state(mm);
  822. if (current->mm) {
  823. mm->flags = current->mm->flags & MMF_INIT_MASK;
  824. mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
  825. } else {
  826. mm->flags = default_dump_filter;
  827. mm->def_flags = 0;
  828. }
  829. if (mm_alloc_pgd(mm))
  830. goto fail_nopgd;
  831. if (init_new_context(p, mm))
  832. goto fail_nocontext;
  833. mm->user_ns = get_user_ns(user_ns);
  834. return mm;
  835. fail_nocontext:
  836. mm_free_pgd(mm);
  837. fail_nopgd:
  838. free_mm(mm);
  839. return NULL;
  840. }
  841. /*
  842. * Allocate and initialize an mm_struct.
  843. */
  844. struct mm_struct *mm_alloc(void)
  845. {
  846. struct mm_struct *mm;
  847. mm = allocate_mm();
  848. if (!mm)
  849. return NULL;
  850. memset(mm, 0, sizeof(*mm));
  851. return mm_init(mm, current, current_user_ns());
  852. }
  853. static inline void __mmput(struct mm_struct *mm)
  854. {
  855. VM_BUG_ON(atomic_read(&mm->mm_users));
  856. uprobe_clear_state(mm);
  857. exit_aio(mm);
  858. ksm_exit(mm);
  859. khugepaged_exit(mm); /* must run before exit_mmap */
  860. exit_mmap(mm);
  861. mm_put_huge_zero_page(mm);
  862. set_mm_exe_file(mm, NULL);
  863. if (!list_empty(&mm->mmlist)) {
  864. spin_lock(&mmlist_lock);
  865. list_del(&mm->mmlist);
  866. spin_unlock(&mmlist_lock);
  867. }
  868. if (mm->binfmt)
  869. module_put(mm->binfmt->module);
  870. mmdrop(mm);
  871. }
  872. /*
  873. * Decrement the use count and release all resources for an mm.
  874. */
  875. void mmput(struct mm_struct *mm)
  876. {
  877. might_sleep();
  878. if (atomic_dec_and_test(&mm->mm_users))
  879. __mmput(mm);
  880. }
  881. EXPORT_SYMBOL_GPL(mmput);
  882. #ifdef CONFIG_MMU
  883. static void mmput_async_fn(struct work_struct *work)
  884. {
  885. struct mm_struct *mm = container_of(work, struct mm_struct,
  886. async_put_work);
  887. __mmput(mm);
  888. }
  889. void mmput_async(struct mm_struct *mm)
  890. {
  891. if (atomic_dec_and_test(&mm->mm_users)) {
  892. INIT_WORK(&mm->async_put_work, mmput_async_fn);
  893. schedule_work(&mm->async_put_work);
  894. }
  895. }
  896. #endif
  897. /**
  898. * set_mm_exe_file - change a reference to the mm's executable file
  899. *
  900. * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
  901. *
  902. * Main users are mmput() and sys_execve(). Callers prevent concurrent
  903. * invocations: in mmput() nobody alive left, in execve task is single
  904. * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
  905. * mm->exe_file, but does so without using set_mm_exe_file() in order
  906. * to do avoid the need for any locks.
  907. */
  908. void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
  909. {
  910. struct file *old_exe_file;
  911. /*
  912. * It is safe to dereference the exe_file without RCU as
  913. * this function is only called if nobody else can access
  914. * this mm -- see comment above for justification.
  915. */
  916. old_exe_file = rcu_dereference_raw(mm->exe_file);
  917. if (new_exe_file)
  918. get_file(new_exe_file);
  919. rcu_assign_pointer(mm->exe_file, new_exe_file);
  920. if (old_exe_file)
  921. fput(old_exe_file);
  922. }
  923. /**
  924. * get_mm_exe_file - acquire a reference to the mm's executable file
  925. *
  926. * Returns %NULL if mm has no associated executable file.
  927. * User must release file via fput().
  928. */
  929. struct file *get_mm_exe_file(struct mm_struct *mm)
  930. {
  931. struct file *exe_file;
  932. rcu_read_lock();
  933. exe_file = rcu_dereference(mm->exe_file);
  934. if (exe_file && !get_file_rcu(exe_file))
  935. exe_file = NULL;
  936. rcu_read_unlock();
  937. return exe_file;
  938. }
  939. EXPORT_SYMBOL(get_mm_exe_file);
  940. /**
  941. * get_task_exe_file - acquire a reference to the task's executable file
  942. *
  943. * Returns %NULL if task's mm (if any) has no associated executable file or
  944. * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
  945. * User must release file via fput().
  946. */
  947. struct file *get_task_exe_file(struct task_struct *task)
  948. {
  949. struct file *exe_file = NULL;
  950. struct mm_struct *mm;
  951. task_lock(task);
  952. mm = task->mm;
  953. if (mm) {
  954. if (!(task->flags & PF_KTHREAD))
  955. exe_file = get_mm_exe_file(mm);
  956. }
  957. task_unlock(task);
  958. return exe_file;
  959. }
  960. EXPORT_SYMBOL(get_task_exe_file);
  961. /**
  962. * get_task_mm - acquire a reference to the task's mm
  963. *
  964. * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
  965. * this kernel workthread has transiently adopted a user mm with use_mm,
  966. * to do its AIO) is not set and if so returns a reference to it, after
  967. * bumping up the use count. User must release the mm via mmput()
  968. * after use. Typically used by /proc and ptrace.
  969. */
  970. struct mm_struct *get_task_mm(struct task_struct *task)
  971. {
  972. struct mm_struct *mm;
  973. task_lock(task);
  974. mm = task->mm;
  975. if (mm) {
  976. if (task->flags & PF_KTHREAD)
  977. mm = NULL;
  978. else
  979. mmget(mm);
  980. }
  981. task_unlock(task);
  982. return mm;
  983. }
  984. EXPORT_SYMBOL_GPL(get_task_mm);
  985. struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
  986. {
  987. struct mm_struct *mm;
  988. int err;
  989. err = mutex_lock_killable(&task->signal->cred_guard_mutex);
  990. if (err)
  991. return ERR_PTR(err);
  992. mm = get_task_mm(task);
  993. if (mm && mm != current->mm &&
  994. !ptrace_may_access(task, mode)) {
  995. mmput(mm);
  996. mm = ERR_PTR(-EACCES);
  997. }
  998. mutex_unlock(&task->signal->cred_guard_mutex);
  999. return mm;
  1000. }
  1001. static void complete_vfork_done(struct task_struct *tsk)
  1002. {
  1003. struct completion *vfork;
  1004. task_lock(tsk);
  1005. vfork = tsk->vfork_done;
  1006. if (likely(vfork)) {
  1007. tsk->vfork_done = NULL;
  1008. complete(vfork);
  1009. }
  1010. task_unlock(tsk);
  1011. }
  1012. static int wait_for_vfork_done(struct task_struct *child,
  1013. struct completion *vfork)
  1014. {
  1015. int killed;
  1016. freezer_do_not_count();
  1017. killed = wait_for_completion_killable(vfork);
  1018. freezer_count();
  1019. if (killed) {
  1020. task_lock(child);
  1021. child->vfork_done = NULL;
  1022. task_unlock(child);
  1023. }
  1024. put_task_struct(child);
  1025. return killed;
  1026. }
  1027. /* Please note the differences between mmput and mm_release.
  1028. * mmput is called whenever we stop holding onto a mm_struct,
  1029. * error success whatever.
  1030. *
  1031. * mm_release is called after a mm_struct has been removed
  1032. * from the current process.
  1033. *
  1034. * This difference is important for error handling, when we
  1035. * only half set up a mm_struct for a new process and need to restore
  1036. * the old one. Because we mmput the new mm_struct before
  1037. * restoring the old one. . .
  1038. * Eric Biederman 10 January 1998
  1039. */
  1040. static void mm_release(struct task_struct *tsk, struct mm_struct *mm)
  1041. {
  1042. uprobe_free_utask(tsk);
  1043. /* Get rid of any cached register state */
  1044. deactivate_mm(tsk, mm);
  1045. /*
  1046. * Signal userspace if we're not exiting with a core dump
  1047. * because we want to leave the value intact for debugging
  1048. * purposes.
  1049. */
  1050. if (tsk->clear_child_tid) {
  1051. if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
  1052. atomic_read(&mm->mm_users) > 1) {
  1053. /*
  1054. * We don't check the error code - if userspace has
  1055. * not set up a proper pointer then tough luck.
  1056. */
  1057. put_user(0, tsk->clear_child_tid);
  1058. do_futex(tsk->clear_child_tid, FUTEX_WAKE,
  1059. 1, NULL, NULL, 0, 0);
  1060. }
  1061. tsk->clear_child_tid = NULL;
  1062. }
  1063. /*
  1064. * All done, finally we can wake up parent and return this mm to him.
  1065. * Also kthread_stop() uses this completion for synchronization.
  1066. */
  1067. if (tsk->vfork_done)
  1068. complete_vfork_done(tsk);
  1069. }
  1070. void exit_mm_release(struct task_struct *tsk, struct mm_struct *mm)
  1071. {
  1072. futex_exit_release(tsk);
  1073. mm_release(tsk, mm);
  1074. }
  1075. void exec_mm_release(struct task_struct *tsk, struct mm_struct *mm)
  1076. {
  1077. futex_exec_release(tsk);
  1078. mm_release(tsk, mm);
  1079. }
  1080. /*
  1081. * Allocate a new mm structure and copy contents from the
  1082. * mm structure of the passed in task structure.
  1083. */
  1084. static struct mm_struct *dup_mm(struct task_struct *tsk)
  1085. {
  1086. struct mm_struct *mm, *oldmm = current->mm;
  1087. int err;
  1088. mm = allocate_mm();
  1089. if (!mm)
  1090. goto fail_nomem;
  1091. memcpy(mm, oldmm, sizeof(*mm));
  1092. if (!mm_init(mm, tsk, mm->user_ns))
  1093. goto fail_nomem;
  1094. err = dup_mmap(mm, oldmm);
  1095. if (err)
  1096. goto free_pt;
  1097. mm->hiwater_rss = get_mm_rss(mm);
  1098. mm->hiwater_vm = mm->total_vm;
  1099. if (mm->binfmt && !try_module_get(mm->binfmt->module))
  1100. goto free_pt;
  1101. return mm;
  1102. free_pt:
  1103. /* don't put binfmt in mmput, we haven't got module yet */
  1104. mm->binfmt = NULL;
  1105. mm_init_owner(mm, NULL);
  1106. mmput(mm);
  1107. fail_nomem:
  1108. return NULL;
  1109. }
  1110. static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
  1111. {
  1112. struct mm_struct *mm, *oldmm;
  1113. int retval;
  1114. tsk->min_flt = tsk->maj_flt = 0;
  1115. tsk->nvcsw = tsk->nivcsw = 0;
  1116. #ifdef CONFIG_DETECT_HUNG_TASK
  1117. tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
  1118. tsk->last_switch_time = 0;
  1119. #endif
  1120. tsk->mm = NULL;
  1121. tsk->active_mm = NULL;
  1122. /*
  1123. * Are we cloning a kernel thread?
  1124. *
  1125. * We need to steal a active VM for that..
  1126. */
  1127. oldmm = current->mm;
  1128. if (!oldmm)
  1129. return 0;
  1130. /* initialize the new vmacache entries */
  1131. vmacache_flush(tsk);
  1132. if (clone_flags & CLONE_VM) {
  1133. mmget(oldmm);
  1134. mm = oldmm;
  1135. goto good_mm;
  1136. }
  1137. retval = -ENOMEM;
  1138. mm = dup_mm(tsk);
  1139. if (!mm)
  1140. goto fail_nomem;
  1141. good_mm:
  1142. tsk->mm = mm;
  1143. tsk->active_mm = mm;
  1144. return 0;
  1145. fail_nomem:
  1146. return retval;
  1147. }
  1148. static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
  1149. {
  1150. struct fs_struct *fs = current->fs;
  1151. if (clone_flags & CLONE_FS) {
  1152. /* tsk->fs is already what we want */
  1153. spin_lock(&fs->lock);
  1154. if (fs->in_exec) {
  1155. spin_unlock(&fs->lock);
  1156. return -EAGAIN;
  1157. }
  1158. fs->users++;
  1159. spin_unlock(&fs->lock);
  1160. return 0;
  1161. }
  1162. tsk->fs = copy_fs_struct(fs);
  1163. if (!tsk->fs)
  1164. return -ENOMEM;
  1165. return 0;
  1166. }
  1167. static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
  1168. {
  1169. struct files_struct *oldf, *newf;
  1170. int error = 0;
  1171. /*
  1172. * A background process may not have any files ...
  1173. */
  1174. oldf = current->files;
  1175. if (!oldf)
  1176. goto out;
  1177. if (clone_flags & CLONE_FILES) {
  1178. atomic_inc(&oldf->count);
  1179. goto out;
  1180. }
  1181. newf = dup_fd(oldf, &error);
  1182. if (!newf)
  1183. goto out;
  1184. tsk->files = newf;
  1185. error = 0;
  1186. out:
  1187. return error;
  1188. }
  1189. static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
  1190. {
  1191. #ifdef CONFIG_BLOCK
  1192. struct io_context *ioc = current->io_context;
  1193. struct io_context *new_ioc;
  1194. if (!ioc)
  1195. return 0;
  1196. /*
  1197. * Share io context with parent, if CLONE_IO is set
  1198. */
  1199. if (clone_flags & CLONE_IO) {
  1200. ioc_task_link(ioc);
  1201. tsk->io_context = ioc;
  1202. } else if (ioprio_valid(ioc->ioprio)) {
  1203. new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
  1204. if (unlikely(!new_ioc))
  1205. return -ENOMEM;
  1206. new_ioc->ioprio = ioc->ioprio;
  1207. put_io_context(new_ioc);
  1208. }
  1209. #endif
  1210. return 0;
  1211. }
  1212. static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
  1213. {
  1214. struct sighand_struct *sig;
  1215. if (clone_flags & CLONE_SIGHAND) {
  1216. atomic_inc(&current->sighand->count);
  1217. return 0;
  1218. }
  1219. sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
  1220. rcu_assign_pointer(tsk->sighand, sig);
  1221. if (!sig)
  1222. return -ENOMEM;
  1223. atomic_set(&sig->count, 1);
  1224. spin_lock_irq(&current->sighand->siglock);
  1225. memcpy(sig->action, current->sighand->action, sizeof(sig->action));
  1226. spin_unlock_irq(&current->sighand->siglock);
  1227. return 0;
  1228. }
  1229. void __cleanup_sighand(struct sighand_struct *sighand)
  1230. {
  1231. if (atomic_dec_and_test(&sighand->count)) {
  1232. signalfd_cleanup(sighand);
  1233. /*
  1234. * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it
  1235. * without an RCU grace period, see __lock_task_sighand().
  1236. */
  1237. kmem_cache_free(sighand_cachep, sighand);
  1238. }
  1239. }
  1240. #ifdef CONFIG_POSIX_TIMERS
  1241. /*
  1242. * Initialize POSIX timer handling for a thread group.
  1243. */
  1244. static void posix_cpu_timers_init_group(struct signal_struct *sig)
  1245. {
  1246. unsigned long cpu_limit;
  1247. cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
  1248. if (cpu_limit != RLIM_INFINITY) {
  1249. sig->cputime_expires.prof_exp = cpu_limit * NSEC_PER_SEC;
  1250. sig->cputimer.running = true;
  1251. }
  1252. /* The timer lists. */
  1253. INIT_LIST_HEAD(&sig->cpu_timers[0]);
  1254. INIT_LIST_HEAD(&sig->cpu_timers[1]);
  1255. INIT_LIST_HEAD(&sig->cpu_timers[2]);
  1256. }
  1257. #else
  1258. static inline void posix_cpu_timers_init_group(struct signal_struct *sig) { }
  1259. #endif
  1260. static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
  1261. {
  1262. struct signal_struct *sig;
  1263. if (clone_flags & CLONE_THREAD)
  1264. return 0;
  1265. sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
  1266. tsk->signal = sig;
  1267. if (!sig)
  1268. return -ENOMEM;
  1269. sig->nr_threads = 1;
  1270. atomic_set(&sig->live, 1);
  1271. atomic_set(&sig->sigcnt, 1);
  1272. /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
  1273. sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
  1274. tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
  1275. init_waitqueue_head(&sig->wait_chldexit);
  1276. sig->curr_target = tsk;
  1277. init_sigpending(&sig->shared_pending);
  1278. INIT_HLIST_HEAD(&sig->multiprocess);
  1279. seqlock_init(&sig->stats_lock);
  1280. prev_cputime_init(&sig->prev_cputime);
  1281. #ifdef CONFIG_POSIX_TIMERS
  1282. INIT_LIST_HEAD(&sig->posix_timers);
  1283. hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
  1284. sig->real_timer.function = it_real_fn;
  1285. #endif
  1286. task_lock(current->group_leader);
  1287. memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
  1288. task_unlock(current->group_leader);
  1289. posix_cpu_timers_init_group(sig);
  1290. tty_audit_fork(sig);
  1291. sched_autogroup_fork(sig);
  1292. sig->oom_score_adj = current->signal->oom_score_adj;
  1293. sig->oom_score_adj_min = current->signal->oom_score_adj_min;
  1294. mutex_init(&sig->cred_guard_mutex);
  1295. return 0;
  1296. }
  1297. static void copy_seccomp(struct task_struct *p)
  1298. {
  1299. #ifdef CONFIG_SECCOMP
  1300. /*
  1301. * Must be called with sighand->lock held, which is common to
  1302. * all threads in the group. Holding cred_guard_mutex is not
  1303. * needed because this new task is not yet running and cannot
  1304. * be racing exec.
  1305. */
  1306. assert_spin_locked(&current->sighand->siglock);
  1307. /* Ref-count the new filter user, and assign it. */
  1308. get_seccomp_filter(current);
  1309. p->seccomp = current->seccomp;
  1310. /*
  1311. * Explicitly enable no_new_privs here in case it got set
  1312. * between the task_struct being duplicated and holding the
  1313. * sighand lock. The seccomp state and nnp must be in sync.
  1314. */
  1315. if (task_no_new_privs(current))
  1316. task_set_no_new_privs(p);
  1317. /*
  1318. * If the parent gained a seccomp mode after copying thread
  1319. * flags and between before we held the sighand lock, we have
  1320. * to manually enable the seccomp thread flag here.
  1321. */
  1322. if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
  1323. set_tsk_thread_flag(p, TIF_SECCOMP);
  1324. #endif
  1325. }
  1326. SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
  1327. {
  1328. current->clear_child_tid = tidptr;
  1329. return task_pid_vnr(current);
  1330. }
  1331. static void rt_mutex_init_task(struct task_struct *p)
  1332. {
  1333. raw_spin_lock_init(&p->pi_lock);
  1334. #ifdef CONFIG_RT_MUTEXES
  1335. p->pi_waiters = RB_ROOT_CACHED;
  1336. p->pi_top_task = NULL;
  1337. p->pi_blocked_on = NULL;
  1338. #endif
  1339. }
  1340. #ifdef CONFIG_POSIX_TIMERS
  1341. /*
  1342. * Initialize POSIX timer handling for a single task.
  1343. */
  1344. static void posix_cpu_timers_init(struct task_struct *tsk)
  1345. {
  1346. tsk->cputime_expires.prof_exp = 0;
  1347. tsk->cputime_expires.virt_exp = 0;
  1348. tsk->cputime_expires.sched_exp = 0;
  1349. INIT_LIST_HEAD(&tsk->cpu_timers[0]);
  1350. INIT_LIST_HEAD(&tsk->cpu_timers[1]);
  1351. INIT_LIST_HEAD(&tsk->cpu_timers[2]);
  1352. }
  1353. #else
  1354. static inline void posix_cpu_timers_init(struct task_struct *tsk) { }
  1355. #endif
  1356. static inline void init_task_pid_links(struct task_struct *task)
  1357. {
  1358. enum pid_type type;
  1359. for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
  1360. INIT_HLIST_NODE(&task->pid_links[type]);
  1361. }
  1362. }
  1363. static inline void
  1364. init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
  1365. {
  1366. if (type == PIDTYPE_PID)
  1367. task->thread_pid = pid;
  1368. else
  1369. task->signal->pids[type] = pid;
  1370. }
  1371. static inline void rcu_copy_process(struct task_struct *p)
  1372. {
  1373. #ifdef CONFIG_PREEMPT_RCU
  1374. p->rcu_read_lock_nesting = 0;
  1375. p->rcu_read_unlock_special.s = 0;
  1376. p->rcu_blocked_node = NULL;
  1377. INIT_LIST_HEAD(&p->rcu_node_entry);
  1378. #endif /* #ifdef CONFIG_PREEMPT_RCU */
  1379. #ifdef CONFIG_TASKS_RCU
  1380. p->rcu_tasks_holdout = false;
  1381. INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
  1382. p->rcu_tasks_idle_cpu = -1;
  1383. #endif /* #ifdef CONFIG_TASKS_RCU */
  1384. }
  1385. static void __delayed_free_task(struct rcu_head *rhp)
  1386. {
  1387. struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
  1388. free_task(tsk);
  1389. }
  1390. static __always_inline void delayed_free_task(struct task_struct *tsk)
  1391. {
  1392. if (IS_ENABLED(CONFIG_MEMCG))
  1393. call_rcu(&tsk->rcu, __delayed_free_task);
  1394. else
  1395. free_task(tsk);
  1396. }
  1397. static void copy_oom_score_adj(u64 clone_flags, struct task_struct *tsk)
  1398. {
  1399. /* Skip if kernel thread */
  1400. if (!tsk->mm)
  1401. return;
  1402. /* Skip if spawning a thread or using vfork */
  1403. if ((clone_flags & (CLONE_VM | CLONE_THREAD | CLONE_VFORK)) != CLONE_VM)
  1404. return;
  1405. /* We need to synchronize with __set_oom_adj */
  1406. mutex_lock(&oom_adj_mutex);
  1407. set_bit(MMF_MULTIPROCESS, &tsk->mm->flags);
  1408. /* Update the values in case they were changed after copy_signal */
  1409. tsk->signal->oom_score_adj = current->signal->oom_score_adj;
  1410. tsk->signal->oom_score_adj_min = current->signal->oom_score_adj_min;
  1411. mutex_unlock(&oom_adj_mutex);
  1412. }
  1413. /*
  1414. * This creates a new process as a copy of the old one,
  1415. * but does not actually start it yet.
  1416. *
  1417. * It copies the registers, and all the appropriate
  1418. * parts of the process environment (as per the clone
  1419. * flags). The actual kick-off is left to the caller.
  1420. */
  1421. static __latent_entropy struct task_struct *copy_process(
  1422. unsigned long clone_flags,
  1423. unsigned long stack_start,
  1424. unsigned long stack_size,
  1425. int __user *child_tidptr,
  1426. struct pid *pid,
  1427. int trace,
  1428. unsigned long tls,
  1429. int node)
  1430. {
  1431. int retval;
  1432. struct task_struct *p;
  1433. struct multiprocess_signals delayed;
  1434. /*
  1435. * Don't allow sharing the root directory with processes in a different
  1436. * namespace
  1437. */
  1438. if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
  1439. return ERR_PTR(-EINVAL);
  1440. if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
  1441. return ERR_PTR(-EINVAL);
  1442. /*
  1443. * Thread groups must share signals as well, and detached threads
  1444. * can only be started up within the thread group.
  1445. */
  1446. if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
  1447. return ERR_PTR(-EINVAL);
  1448. /*
  1449. * Shared signal handlers imply shared VM. By way of the above,
  1450. * thread groups also imply shared VM. Blocking this case allows
  1451. * for various simplifications in other code.
  1452. */
  1453. if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
  1454. return ERR_PTR(-EINVAL);
  1455. /*
  1456. * Siblings of global init remain as zombies on exit since they are
  1457. * not reaped by their parent (swapper). To solve this and to avoid
  1458. * multi-rooted process trees, prevent global and container-inits
  1459. * from creating siblings.
  1460. */
  1461. if ((clone_flags & CLONE_PARENT) &&
  1462. current->signal->flags & SIGNAL_UNKILLABLE)
  1463. return ERR_PTR(-EINVAL);
  1464. /*
  1465. * If the new process will be in a different pid or user namespace
  1466. * do not allow it to share a thread group with the forking task.
  1467. */
  1468. if (clone_flags & CLONE_THREAD) {
  1469. if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
  1470. (task_active_pid_ns(current) !=
  1471. current->nsproxy->pid_ns_for_children))
  1472. return ERR_PTR(-EINVAL);
  1473. }
  1474. /*
  1475. * Force any signals received before this point to be delivered
  1476. * before the fork happens. Collect up signals sent to multiple
  1477. * processes that happen during the fork and delay them so that
  1478. * they appear to happen after the fork.
  1479. */
  1480. sigemptyset(&delayed.signal);
  1481. INIT_HLIST_NODE(&delayed.node);
  1482. spin_lock_irq(&current->sighand->siglock);
  1483. if (!(clone_flags & CLONE_THREAD))
  1484. hlist_add_head(&delayed.node, &current->signal->multiprocess);
  1485. recalc_sigpending();
  1486. spin_unlock_irq(&current->sighand->siglock);
  1487. retval = -ERESTARTNOINTR;
  1488. if (signal_pending(current))
  1489. goto fork_out;
  1490. retval = -ENOMEM;
  1491. p = dup_task_struct(current, node);
  1492. if (!p)
  1493. goto fork_out;
  1494. /*
  1495. * This _must_ happen before we call free_task(), i.e. before we jump
  1496. * to any of the bad_fork_* labels. This is to avoid freeing
  1497. * p->set_child_tid which is (ab)used as a kthread's data pointer for
  1498. * kernel threads (PF_KTHREAD).
  1499. */
  1500. p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
  1501. /*
  1502. * Clear TID on mm_release()?
  1503. */
  1504. p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
  1505. ftrace_graph_init_task(p);
  1506. rt_mutex_init_task(p);
  1507. #ifdef CONFIG_PROVE_LOCKING
  1508. DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
  1509. DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
  1510. #endif
  1511. retval = -EAGAIN;
  1512. if (atomic_read(&p->real_cred->user->processes) >=
  1513. task_rlimit(p, RLIMIT_NPROC)) {
  1514. if (p->real_cred->user != INIT_USER &&
  1515. !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
  1516. goto bad_fork_free;
  1517. }
  1518. current->flags &= ~PF_NPROC_EXCEEDED;
  1519. retval = copy_creds(p, clone_flags);
  1520. if (retval < 0)
  1521. goto bad_fork_free;
  1522. /*
  1523. * If multiple threads are within copy_process(), then this check
  1524. * triggers too late. This doesn't hurt, the check is only there
  1525. * to stop root fork bombs.
  1526. */
  1527. retval = -EAGAIN;
  1528. if (nr_threads >= max_threads)
  1529. goto bad_fork_cleanup_count;
  1530. delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
  1531. p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE);
  1532. p->flags |= PF_FORKNOEXEC;
  1533. INIT_LIST_HEAD(&p->children);
  1534. INIT_LIST_HEAD(&p->sibling);
  1535. rcu_copy_process(p);
  1536. p->vfork_done = NULL;
  1537. spin_lock_init(&p->alloc_lock);
  1538. init_sigpending(&p->pending);
  1539. p->utime = p->stime = p->gtime = 0;
  1540. #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
  1541. p->utimescaled = p->stimescaled = 0;
  1542. #endif
  1543. prev_cputime_init(&p->prev_cputime);
  1544. #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
  1545. seqcount_init(&p->vtime.seqcount);
  1546. p->vtime.starttime = 0;
  1547. p->vtime.state = VTIME_INACTIVE;
  1548. #endif
  1549. #if defined(SPLIT_RSS_COUNTING)
  1550. memset(&p->rss_stat, 0, sizeof(p->rss_stat));
  1551. #endif
  1552. p->default_timer_slack_ns = current->timer_slack_ns;
  1553. task_io_accounting_init(&p->ioac);
  1554. acct_clear_integrals(p);
  1555. posix_cpu_timers_init(p);
  1556. p->io_context = NULL;
  1557. audit_set_context(p, NULL);
  1558. cgroup_fork(p);
  1559. #ifdef CONFIG_NUMA
  1560. p->mempolicy = mpol_dup(p->mempolicy);
  1561. if (IS_ERR(p->mempolicy)) {
  1562. retval = PTR_ERR(p->mempolicy);
  1563. p->mempolicy = NULL;
  1564. goto bad_fork_cleanup_threadgroup_lock;
  1565. }
  1566. #endif
  1567. #ifdef CONFIG_CPUSETS
  1568. p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
  1569. p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
  1570. seqcount_init(&p->mems_allowed_seq);
  1571. #endif
  1572. #ifdef CONFIG_TRACE_IRQFLAGS
  1573. p->irq_events = 0;
  1574. p->hardirqs_enabled = 0;
  1575. p->hardirq_enable_ip = 0;
  1576. p->hardirq_enable_event = 0;
  1577. p->hardirq_disable_ip = _THIS_IP_;
  1578. p->hardirq_disable_event = 0;
  1579. p->softirqs_enabled = 1;
  1580. p->softirq_enable_ip = _THIS_IP_;
  1581. p->softirq_enable_event = 0;
  1582. p->softirq_disable_ip = 0;
  1583. p->softirq_disable_event = 0;
  1584. p->hardirq_context = 0;
  1585. p->softirq_context = 0;
  1586. #endif
  1587. p->pagefault_disabled = 0;
  1588. #ifdef CONFIG_LOCKDEP
  1589. p->lockdep_depth = 0; /* no locks held yet */
  1590. p->curr_chain_key = 0;
  1591. p->lockdep_recursion = 0;
  1592. lockdep_init_task(p);
  1593. #endif
  1594. #ifdef CONFIG_DEBUG_MUTEXES
  1595. p->blocked_on = NULL; /* not blocked yet */
  1596. #endif
  1597. #ifdef CONFIG_BCACHE
  1598. p->sequential_io = 0;
  1599. p->sequential_io_avg = 0;
  1600. #endif
  1601. /* Perform scheduler related setup. Assign this task to a CPU. */
  1602. retval = sched_fork(clone_flags, p);
  1603. if (retval)
  1604. goto bad_fork_cleanup_policy;
  1605. retval = perf_event_init_task(p);
  1606. if (retval)
  1607. goto bad_fork_cleanup_policy;
  1608. retval = audit_alloc(p);
  1609. if (retval)
  1610. goto bad_fork_cleanup_perf;
  1611. /* copy all the process information */
  1612. shm_init_task(p);
  1613. retval = security_task_alloc(p, clone_flags);
  1614. if (retval)
  1615. goto bad_fork_cleanup_audit;
  1616. retval = copy_semundo(clone_flags, p);
  1617. if (retval)
  1618. goto bad_fork_cleanup_security;
  1619. retval = copy_files(clone_flags, p);
  1620. if (retval)
  1621. goto bad_fork_cleanup_semundo;
  1622. retval = copy_fs(clone_flags, p);
  1623. if (retval)
  1624. goto bad_fork_cleanup_files;
  1625. retval = copy_sighand(clone_flags, p);
  1626. if (retval)
  1627. goto bad_fork_cleanup_fs;
  1628. retval = copy_signal(clone_flags, p);
  1629. if (retval)
  1630. goto bad_fork_cleanup_sighand;
  1631. retval = copy_mm(clone_flags, p);
  1632. if (retval)
  1633. goto bad_fork_cleanup_signal;
  1634. retval = copy_namespaces(clone_flags, p);
  1635. if (retval)
  1636. goto bad_fork_cleanup_mm;
  1637. retval = copy_io(clone_flags, p);
  1638. if (retval)
  1639. goto bad_fork_cleanup_namespaces;
  1640. retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls);
  1641. if (retval)
  1642. goto bad_fork_cleanup_io;
  1643. if (pid != &init_struct_pid) {
  1644. pid = alloc_pid(p->nsproxy->pid_ns_for_children);
  1645. if (IS_ERR(pid)) {
  1646. retval = PTR_ERR(pid);
  1647. goto bad_fork_cleanup_thread;
  1648. }
  1649. }
  1650. #ifdef CONFIG_BLOCK
  1651. p->plug = NULL;
  1652. #endif
  1653. futex_init_task(p);
  1654. /*
  1655. * sigaltstack should be cleared when sharing the same VM
  1656. */
  1657. if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
  1658. sas_ss_reset(p);
  1659. /*
  1660. * Syscall tracing and stepping should be turned off in the
  1661. * child regardless of CLONE_PTRACE.
  1662. */
  1663. user_disable_single_step(p);
  1664. clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
  1665. #ifdef TIF_SYSCALL_EMU
  1666. clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
  1667. #endif
  1668. clear_all_latency_tracing(p);
  1669. /* ok, now we should be set up.. */
  1670. p->pid = pid_nr(pid);
  1671. if (clone_flags & CLONE_THREAD) {
  1672. p->group_leader = current->group_leader;
  1673. p->tgid = current->tgid;
  1674. } else {
  1675. p->group_leader = p;
  1676. p->tgid = p->pid;
  1677. }
  1678. p->nr_dirtied = 0;
  1679. p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
  1680. p->dirty_paused_when = 0;
  1681. p->pdeath_signal = 0;
  1682. INIT_LIST_HEAD(&p->thread_group);
  1683. p->task_works = NULL;
  1684. cgroup_threadgroup_change_begin(current);
  1685. /*
  1686. * Ensure that the cgroup subsystem policies allow the new process to be
  1687. * forked. It should be noted the the new process's css_set can be changed
  1688. * between here and cgroup_post_fork() if an organisation operation is in
  1689. * progress.
  1690. */
  1691. retval = cgroup_can_fork(p);
  1692. if (retval)
  1693. goto bad_fork_free_pid;
  1694. /*
  1695. * From this point on we must avoid any synchronous user-space
  1696. * communication until we take the tasklist-lock. In particular, we do
  1697. * not want user-space to be able to predict the process start-time by
  1698. * stalling fork(2) after we recorded the start_time but before it is
  1699. * visible to the system.
  1700. */
  1701. p->start_time = ktime_get_ns();
  1702. p->real_start_time = ktime_get_boot_ns();
  1703. /*
  1704. * Make it visible to the rest of the system, but dont wake it up yet.
  1705. * Need tasklist lock for parent etc handling!
  1706. */
  1707. write_lock_irq(&tasklist_lock);
  1708. /* CLONE_PARENT re-uses the old parent */
  1709. if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
  1710. p->real_parent = current->real_parent;
  1711. p->parent_exec_id = current->parent_exec_id;
  1712. if (clone_flags & CLONE_THREAD)
  1713. p->exit_signal = -1;
  1714. else
  1715. p->exit_signal = current->group_leader->exit_signal;
  1716. } else {
  1717. p->real_parent = current;
  1718. p->parent_exec_id = current->self_exec_id;
  1719. p->exit_signal = (clone_flags & CSIGNAL);
  1720. }
  1721. klp_copy_process(p);
  1722. spin_lock(&current->sighand->siglock);
  1723. /*
  1724. * Copy seccomp details explicitly here, in case they were changed
  1725. * before holding sighand lock.
  1726. */
  1727. copy_seccomp(p);
  1728. rseq_fork(p, clone_flags);
  1729. /* Don't start children in a dying pid namespace */
  1730. if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) {
  1731. retval = -ENOMEM;
  1732. goto bad_fork_cancel_cgroup;
  1733. }
  1734. /* Let kill terminate clone/fork in the middle */
  1735. if (fatal_signal_pending(current)) {
  1736. retval = -EINTR;
  1737. goto bad_fork_cancel_cgroup;
  1738. }
  1739. init_task_pid_links(p);
  1740. if (likely(p->pid)) {
  1741. ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
  1742. init_task_pid(p, PIDTYPE_PID, pid);
  1743. if (thread_group_leader(p)) {
  1744. init_task_pid(p, PIDTYPE_TGID, pid);
  1745. init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
  1746. init_task_pid(p, PIDTYPE_SID, task_session(current));
  1747. if (is_child_reaper(pid)) {
  1748. ns_of_pid(pid)->child_reaper = p;
  1749. p->signal->flags |= SIGNAL_UNKILLABLE;
  1750. }
  1751. p->signal->shared_pending.signal = delayed.signal;
  1752. p->signal->tty = tty_kref_get(current->signal->tty);
  1753. /*
  1754. * Inherit has_child_subreaper flag under the same
  1755. * tasklist_lock with adding child to the process tree
  1756. * for propagate_has_child_subreaper optimization.
  1757. */
  1758. p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper ||
  1759. p->real_parent->signal->is_child_subreaper;
  1760. list_add_tail(&p->sibling, &p->real_parent->children);
  1761. list_add_tail_rcu(&p->tasks, &init_task.tasks);
  1762. attach_pid(p, PIDTYPE_TGID);
  1763. attach_pid(p, PIDTYPE_PGID);
  1764. attach_pid(p, PIDTYPE_SID);
  1765. __this_cpu_inc(process_counts);
  1766. } else {
  1767. current->signal->nr_threads++;
  1768. atomic_inc(&current->signal->live);
  1769. atomic_inc(&current->signal->sigcnt);
  1770. task_join_group_stop(p);
  1771. list_add_tail_rcu(&p->thread_group,
  1772. &p->group_leader->thread_group);
  1773. list_add_tail_rcu(&p->thread_node,
  1774. &p->signal->thread_head);
  1775. }
  1776. attach_pid(p, PIDTYPE_PID);
  1777. nr_threads++;
  1778. }
  1779. total_forks++;
  1780. hlist_del_init(&delayed.node);
  1781. spin_unlock(&current->sighand->siglock);
  1782. syscall_tracepoint_update(p);
  1783. write_unlock_irq(&tasklist_lock);
  1784. proc_fork_connector(p);
  1785. cgroup_post_fork(p);
  1786. cgroup_threadgroup_change_end(current);
  1787. perf_event_fork(p);
  1788. trace_task_newtask(p, clone_flags);
  1789. uprobe_copy_process(p, clone_flags);
  1790. copy_oom_score_adj(clone_flags, p);
  1791. return p;
  1792. bad_fork_cancel_cgroup:
  1793. spin_unlock(&current->sighand->siglock);
  1794. write_unlock_irq(&tasklist_lock);
  1795. cgroup_cancel_fork(p);
  1796. bad_fork_free_pid:
  1797. cgroup_threadgroup_change_end(current);
  1798. if (pid != &init_struct_pid)
  1799. free_pid(pid);
  1800. bad_fork_cleanup_thread:
  1801. exit_thread(p);
  1802. bad_fork_cleanup_io:
  1803. if (p->io_context)
  1804. exit_io_context(p);
  1805. bad_fork_cleanup_namespaces:
  1806. exit_task_namespaces(p);
  1807. bad_fork_cleanup_mm:
  1808. if (p->mm) {
  1809. mm_clear_owner(p->mm, p);
  1810. mmput(p->mm);
  1811. }
  1812. bad_fork_cleanup_signal:
  1813. if (!(clone_flags & CLONE_THREAD))
  1814. free_signal_struct(p->signal);
  1815. bad_fork_cleanup_sighand:
  1816. __cleanup_sighand(p->sighand);
  1817. bad_fork_cleanup_fs:
  1818. exit_fs(p); /* blocking */
  1819. bad_fork_cleanup_files:
  1820. exit_files(p); /* blocking */
  1821. bad_fork_cleanup_semundo:
  1822. exit_sem(p);
  1823. bad_fork_cleanup_security:
  1824. security_task_free(p);
  1825. bad_fork_cleanup_audit:
  1826. audit_free(p);
  1827. bad_fork_cleanup_perf:
  1828. perf_event_free_task(p);
  1829. bad_fork_cleanup_policy:
  1830. lockdep_free_task(p);
  1831. #ifdef CONFIG_NUMA
  1832. mpol_put(p->mempolicy);
  1833. bad_fork_cleanup_threadgroup_lock:
  1834. #endif
  1835. delayacct_tsk_free(p);
  1836. bad_fork_cleanup_count:
  1837. atomic_dec(&p->cred->user->processes);
  1838. exit_creds(p);
  1839. bad_fork_free:
  1840. p->state = TASK_DEAD;
  1841. put_task_stack(p);
  1842. delayed_free_task(p);
  1843. fork_out:
  1844. spin_lock_irq(&current->sighand->siglock);
  1845. hlist_del_init(&delayed.node);
  1846. spin_unlock_irq(&current->sighand->siglock);
  1847. return ERR_PTR(retval);
  1848. }
  1849. static inline void init_idle_pids(struct task_struct *idle)
  1850. {
  1851. enum pid_type type;
  1852. for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
  1853. INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */
  1854. init_task_pid(idle, type, &init_struct_pid);
  1855. }
  1856. }
  1857. struct task_struct *fork_idle(int cpu)
  1858. {
  1859. struct task_struct *task;
  1860. task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0,
  1861. cpu_to_node(cpu));
  1862. if (!IS_ERR(task)) {
  1863. init_idle_pids(task);
  1864. init_idle(task, cpu);
  1865. }
  1866. return task;
  1867. }
  1868. /*
  1869. * Ok, this is the main fork-routine.
  1870. *
  1871. * It copies the process, and if successful kick-starts
  1872. * it and waits for it to finish using the VM if required.
  1873. */
  1874. long _do_fork(unsigned long clone_flags,
  1875. unsigned long stack_start,
  1876. unsigned long stack_size,
  1877. int __user *parent_tidptr,
  1878. int __user *child_tidptr,
  1879. unsigned long tls)
  1880. {
  1881. struct completion vfork;
  1882. struct pid *pid;
  1883. struct task_struct *p;
  1884. int trace = 0;
  1885. long nr;
  1886. /*
  1887. * Determine whether and which event to report to ptracer. When
  1888. * called from kernel_thread or CLONE_UNTRACED is explicitly
  1889. * requested, no event is reported; otherwise, report if the event
  1890. * for the type of forking is enabled.
  1891. */
  1892. if (!(clone_flags & CLONE_UNTRACED)) {
  1893. if (clone_flags & CLONE_VFORK)
  1894. trace = PTRACE_EVENT_VFORK;
  1895. else if ((clone_flags & CSIGNAL) != SIGCHLD)
  1896. trace = PTRACE_EVENT_CLONE;
  1897. else
  1898. trace = PTRACE_EVENT_FORK;
  1899. if (likely(!ptrace_event_enabled(current, trace)))
  1900. trace = 0;
  1901. }
  1902. p = copy_process(clone_flags, stack_start, stack_size,
  1903. child_tidptr, NULL, trace, tls, NUMA_NO_NODE);
  1904. add_latent_entropy();
  1905. if (IS_ERR(p))
  1906. return PTR_ERR(p);
  1907. /*
  1908. * Do this prior waking up the new thread - the thread pointer
  1909. * might get invalid after that point, if the thread exits quickly.
  1910. */
  1911. trace_sched_process_fork(current, p);
  1912. pid = get_task_pid(p, PIDTYPE_PID);
  1913. nr = pid_vnr(pid);
  1914. if (clone_flags & CLONE_PARENT_SETTID)
  1915. put_user(nr, parent_tidptr);
  1916. if (clone_flags & CLONE_VFORK) {
  1917. p->vfork_done = &vfork;
  1918. init_completion(&vfork);
  1919. get_task_struct(p);
  1920. }
  1921. wake_up_new_task(p);
  1922. /* forking complete and child started to run, tell ptracer */
  1923. if (unlikely(trace))
  1924. ptrace_event_pid(trace, pid);
  1925. if (clone_flags & CLONE_VFORK) {
  1926. if (!wait_for_vfork_done(p, &vfork))
  1927. ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
  1928. }
  1929. put_pid(pid);
  1930. return nr;
  1931. }
  1932. #ifndef CONFIG_HAVE_COPY_THREAD_TLS
  1933. /* For compatibility with architectures that call do_fork directly rather than
  1934. * using the syscall entry points below. */
  1935. long do_fork(unsigned long clone_flags,
  1936. unsigned long stack_start,
  1937. unsigned long stack_size,
  1938. int __user *parent_tidptr,
  1939. int __user *child_tidptr)
  1940. {
  1941. return _do_fork(clone_flags, stack_start, stack_size,
  1942. parent_tidptr, child_tidptr, 0);
  1943. }
  1944. #endif
  1945. /*
  1946. * Create a kernel thread.
  1947. */
  1948. pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
  1949. {
  1950. return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn,
  1951. (unsigned long)arg, NULL, NULL, 0);
  1952. }
  1953. #ifdef __ARCH_WANT_SYS_FORK
  1954. SYSCALL_DEFINE0(fork)
  1955. {
  1956. #ifdef CONFIG_MMU
  1957. return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0);
  1958. #else
  1959. /* can not support in nommu mode */
  1960. return -EINVAL;
  1961. #endif
  1962. }
  1963. #endif
  1964. #ifdef __ARCH_WANT_SYS_VFORK
  1965. SYSCALL_DEFINE0(vfork)
  1966. {
  1967. return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0,
  1968. 0, NULL, NULL, 0);
  1969. }
  1970. #endif
  1971. #ifdef __ARCH_WANT_SYS_CLONE
  1972. #ifdef CONFIG_CLONE_BACKWARDS
  1973. SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
  1974. int __user *, parent_tidptr,
  1975. unsigned long, tls,
  1976. int __user *, child_tidptr)
  1977. #elif defined(CONFIG_CLONE_BACKWARDS2)
  1978. SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
  1979. int __user *, parent_tidptr,
  1980. int __user *, child_tidptr,
  1981. unsigned long, tls)
  1982. #elif defined(CONFIG_CLONE_BACKWARDS3)
  1983. SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
  1984. int, stack_size,
  1985. int __user *, parent_tidptr,
  1986. int __user *, child_tidptr,
  1987. unsigned long, tls)
  1988. #else
  1989. SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
  1990. int __user *, parent_tidptr,
  1991. int __user *, child_tidptr,
  1992. unsigned long, tls)
  1993. #endif
  1994. {
  1995. return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls);
  1996. }
  1997. #endif
  1998. void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data)
  1999. {
  2000. struct task_struct *leader, *parent, *child;
  2001. int res;
  2002. read_lock(&tasklist_lock);
  2003. leader = top = top->group_leader;
  2004. down:
  2005. for_each_thread(leader, parent) {
  2006. list_for_each_entry(child, &parent->children, sibling) {
  2007. res = visitor(child, data);
  2008. if (res) {
  2009. if (res < 0)
  2010. goto out;
  2011. leader = child;
  2012. goto down;
  2013. }
  2014. up:
  2015. ;
  2016. }
  2017. }
  2018. if (leader != top) {
  2019. child = leader;
  2020. parent = child->real_parent;
  2021. leader = parent->group_leader;
  2022. goto up;
  2023. }
  2024. out:
  2025. read_unlock(&tasklist_lock);
  2026. }
  2027. #ifndef ARCH_MIN_MMSTRUCT_ALIGN
  2028. #define ARCH_MIN_MMSTRUCT_ALIGN 0
  2029. #endif
  2030. static void sighand_ctor(void *data)
  2031. {
  2032. struct sighand_struct *sighand = data;
  2033. spin_lock_init(&sighand->siglock);
  2034. init_waitqueue_head(&sighand->signalfd_wqh);
  2035. }
  2036. void __init proc_caches_init(void)
  2037. {
  2038. unsigned int mm_size;
  2039. sighand_cachep = kmem_cache_create("sighand_cache",
  2040. sizeof(struct sighand_struct), 0,
  2041. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU|
  2042. SLAB_ACCOUNT, sighand_ctor);
  2043. signal_cachep = kmem_cache_create("signal_cache",
  2044. sizeof(struct signal_struct), 0,
  2045. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
  2046. NULL);
  2047. files_cachep = kmem_cache_create("files_cache",
  2048. sizeof(struct files_struct), 0,
  2049. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
  2050. NULL);
  2051. fs_cachep = kmem_cache_create("fs_cache",
  2052. sizeof(struct fs_struct), 0,
  2053. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
  2054. NULL);
  2055. /*
  2056. * The mm_cpumask is located at the end of mm_struct, and is
  2057. * dynamically sized based on the maximum CPU number this system
  2058. * can have, taking hotplug into account (nr_cpu_ids).
  2059. */
  2060. mm_size = sizeof(struct mm_struct) + cpumask_size();
  2061. mm_cachep = kmem_cache_create_usercopy("mm_struct",
  2062. mm_size, ARCH_MIN_MMSTRUCT_ALIGN,
  2063. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
  2064. offsetof(struct mm_struct, saved_auxv),
  2065. sizeof_field(struct mm_struct, saved_auxv),
  2066. NULL);
  2067. vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
  2068. mmap_init();
  2069. nsproxy_cache_init();
  2070. }
  2071. /*
  2072. * Check constraints on flags passed to the unshare system call.
  2073. */
  2074. static int check_unshare_flags(unsigned long unshare_flags)
  2075. {
  2076. if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
  2077. CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
  2078. CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
  2079. CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
  2080. return -EINVAL;
  2081. /*
  2082. * Not implemented, but pretend it works if there is nothing
  2083. * to unshare. Note that unsharing the address space or the
  2084. * signal handlers also need to unshare the signal queues (aka
  2085. * CLONE_THREAD).
  2086. */
  2087. if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
  2088. if (!thread_group_empty(current))
  2089. return -EINVAL;
  2090. }
  2091. if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
  2092. if (atomic_read(&current->sighand->count) > 1)
  2093. return -EINVAL;
  2094. }
  2095. if (unshare_flags & CLONE_VM) {
  2096. if (!current_is_single_threaded())
  2097. return -EINVAL;
  2098. }
  2099. return 0;
  2100. }
  2101. /*
  2102. * Unshare the filesystem structure if it is being shared
  2103. */
  2104. static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
  2105. {
  2106. struct fs_struct *fs = current->fs;
  2107. if (!(unshare_flags & CLONE_FS) || !fs)
  2108. return 0;
  2109. /* don't need lock here; in the worst case we'll do useless copy */
  2110. if (fs->users == 1)
  2111. return 0;
  2112. *new_fsp = copy_fs_struct(fs);
  2113. if (!*new_fsp)
  2114. return -ENOMEM;
  2115. return 0;
  2116. }
  2117. /*
  2118. * Unshare file descriptor table if it is being shared
  2119. */
  2120. static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
  2121. {
  2122. struct files_struct *fd = current->files;
  2123. int error = 0;
  2124. if ((unshare_flags & CLONE_FILES) &&
  2125. (fd && atomic_read(&fd->count) > 1)) {
  2126. *new_fdp = dup_fd(fd, &error);
  2127. if (!*new_fdp)
  2128. return error;
  2129. }
  2130. return 0;
  2131. }
  2132. /*
  2133. * unshare allows a process to 'unshare' part of the process
  2134. * context which was originally shared using clone. copy_*
  2135. * functions used by do_fork() cannot be used here directly
  2136. * because they modify an inactive task_struct that is being
  2137. * constructed. Here we are modifying the current, active,
  2138. * task_struct.
  2139. */
  2140. int ksys_unshare(unsigned long unshare_flags)
  2141. {
  2142. struct fs_struct *fs, *new_fs = NULL;
  2143. struct files_struct *fd, *new_fd = NULL;
  2144. struct cred *new_cred = NULL;
  2145. struct nsproxy *new_nsproxy = NULL;
  2146. int do_sysvsem = 0;
  2147. int err;
  2148. /*
  2149. * If unsharing a user namespace must also unshare the thread group
  2150. * and unshare the filesystem root and working directories.
  2151. */
  2152. if (unshare_flags & CLONE_NEWUSER)
  2153. unshare_flags |= CLONE_THREAD | CLONE_FS;
  2154. /*
  2155. * If unsharing vm, must also unshare signal handlers.
  2156. */
  2157. if (unshare_flags & CLONE_VM)
  2158. unshare_flags |= CLONE_SIGHAND;
  2159. /*
  2160. * If unsharing a signal handlers, must also unshare the signal queues.
  2161. */
  2162. if (unshare_flags & CLONE_SIGHAND)
  2163. unshare_flags |= CLONE_THREAD;
  2164. /*
  2165. * If unsharing namespace, must also unshare filesystem information.
  2166. */
  2167. if (unshare_flags & CLONE_NEWNS)
  2168. unshare_flags |= CLONE_FS;
  2169. err = check_unshare_flags(unshare_flags);
  2170. if (err)
  2171. goto bad_unshare_out;
  2172. /*
  2173. * CLONE_NEWIPC must also detach from the undolist: after switching
  2174. * to a new ipc namespace, the semaphore arrays from the old
  2175. * namespace are unreachable.
  2176. */
  2177. if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
  2178. do_sysvsem = 1;
  2179. err = unshare_fs(unshare_flags, &new_fs);
  2180. if (err)
  2181. goto bad_unshare_out;
  2182. err = unshare_fd(unshare_flags, &new_fd);
  2183. if (err)
  2184. goto bad_unshare_cleanup_fs;
  2185. err = unshare_userns(unshare_flags, &new_cred);
  2186. if (err)
  2187. goto bad_unshare_cleanup_fd;
  2188. err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
  2189. new_cred, new_fs);
  2190. if (err)
  2191. goto bad_unshare_cleanup_cred;
  2192. if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
  2193. if (do_sysvsem) {
  2194. /*
  2195. * CLONE_SYSVSEM is equivalent to sys_exit().
  2196. */
  2197. exit_sem(current);
  2198. }
  2199. if (unshare_flags & CLONE_NEWIPC) {
  2200. /* Orphan segments in old ns (see sem above). */
  2201. exit_shm(current);
  2202. shm_init_task(current);
  2203. }
  2204. if (new_nsproxy)
  2205. switch_task_namespaces(current, new_nsproxy);
  2206. task_lock(current);
  2207. if (new_fs) {
  2208. fs = current->fs;
  2209. spin_lock(&fs->lock);
  2210. current->fs = new_fs;
  2211. if (--fs->users)
  2212. new_fs = NULL;
  2213. else
  2214. new_fs = fs;
  2215. spin_unlock(&fs->lock);
  2216. }
  2217. if (new_fd) {
  2218. fd = current->files;
  2219. current->files = new_fd;
  2220. new_fd = fd;
  2221. }
  2222. task_unlock(current);
  2223. if (new_cred) {
  2224. /* Install the new user namespace */
  2225. commit_creds(new_cred);
  2226. new_cred = NULL;
  2227. }
  2228. }
  2229. perf_event_namespaces(current);
  2230. bad_unshare_cleanup_cred:
  2231. if (new_cred)
  2232. put_cred(new_cred);
  2233. bad_unshare_cleanup_fd:
  2234. if (new_fd)
  2235. put_files_struct(new_fd);
  2236. bad_unshare_cleanup_fs:
  2237. if (new_fs)
  2238. free_fs_struct(new_fs);
  2239. bad_unshare_out:
  2240. return err;
  2241. }
  2242. SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
  2243. {
  2244. return ksys_unshare(unshare_flags);
  2245. }
  2246. /*
  2247. * Helper to unshare the files of the current task.
  2248. * We don't want to expose copy_files internals to
  2249. * the exec layer of the kernel.
  2250. */
  2251. int unshare_files(struct files_struct **displaced)
  2252. {
  2253. struct task_struct *task = current;
  2254. struct files_struct *copy = NULL;
  2255. int error;
  2256. error = unshare_fd(CLONE_FILES, &copy);
  2257. if (error || !copy) {
  2258. *displaced = NULL;
  2259. return error;
  2260. }
  2261. *displaced = task->files;
  2262. task_lock(task);
  2263. task->files = copy;
  2264. task_unlock(task);
  2265. return 0;
  2266. }
  2267. int sysctl_max_threads(struct ctl_table *table, int write,
  2268. void __user *buffer, size_t *lenp, loff_t *ppos)
  2269. {
  2270. struct ctl_table t;
  2271. int ret;
  2272. int threads = max_threads;
  2273. int min = 1;
  2274. int max = MAX_THREADS;
  2275. t = *table;
  2276. t.data = &threads;
  2277. t.extra1 = &min;
  2278. t.extra2 = &max;
  2279. ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
  2280. if (ret || !write)
  2281. return ret;
  2282. max_threads = threads;
  2283. return 0;
  2284. }