fork.c 59 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460
  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/fs.h>
  39. #include <linux/mm.h>
  40. #include <linux/vmacache.h>
  41. #include <linux/nsproxy.h>
  42. #include <linux/capability.h>
  43. #include <linux/cpu.h>
  44. #include <linux/cgroup.h>
  45. #include <linux/security.h>
  46. #include <linux/hugetlb.h>
  47. #include <linux/seccomp.h>
  48. #include <linux/swap.h>
  49. #include <linux/syscalls.h>
  50. #include <linux/jiffies.h>
  51. #include <linux/futex.h>
  52. #include <linux/compat.h>
  53. #include <linux/kthread.h>
  54. #include <linux/task_io_accounting_ops.h>
  55. #include <linux/rcupdate.h>
  56. #include <linux/ptrace.h>
  57. #include <linux/mount.h>
  58. #include <linux/audit.h>
  59. #include <linux/memcontrol.h>
  60. #include <linux/ftrace.h>
  61. #include <linux/proc_fs.h>
  62. #include <linux/profile.h>
  63. #include <linux/rmap.h>
  64. #include <linux/ksm.h>
  65. #include <linux/acct.h>
  66. #include <linux/userfaultfd_k.h>
  67. #include <linux/tsacct_kern.h>
  68. #include <linux/cn_proc.h>
  69. #include <linux/freezer.h>
  70. #include <linux/delayacct.h>
  71. #include <linux/taskstats_kern.h>
  72. #include <linux/random.h>
  73. #include <linux/tty.h>
  74. #include <linux/blkdev.h>
  75. #include <linux/fs_struct.h>
  76. #include <linux/magic.h>
  77. #include <linux/perf_event.h>
  78. #include <linux/posix-timers.h>
  79. #include <linux/user-return-notifier.h>
  80. #include <linux/oom.h>
  81. #include <linux/khugepaged.h>
  82. #include <linux/signalfd.h>
  83. #include <linux/uprobes.h>
  84. #include <linux/aio.h>
  85. #include <linux/compiler.h>
  86. #include <linux/sysctl.h>
  87. #include <linux/kcov.h>
  88. #include <linux/livepatch.h>
  89. #include <asm/pgtable.h>
  90. #include <asm/pgalloc.h>
  91. #include <linux/uaccess.h>
  92. #include <asm/mmu_context.h>
  93. #include <asm/cacheflush.h>
  94. #include <asm/tlbflush.h>
  95. #include <trace/events/sched.h>
  96. #define CREATE_TRACE_POINTS
  97. #include <trace/events/task.h>
  98. /*
  99. * Minimum number of threads to boot the kernel
  100. */
  101. #define MIN_THREADS 20
  102. /*
  103. * Maximum number of threads
  104. */
  105. #define MAX_THREADS FUTEX_TID_MASK
  106. /*
  107. * Protected counters by write_lock_irq(&tasklist_lock)
  108. */
  109. unsigned long total_forks; /* Handle normal Linux uptimes. */
  110. int nr_threads; /* The idle threads do not count.. */
  111. int max_threads; /* tunable limit on nr_threads */
  112. DEFINE_PER_CPU(unsigned long, process_counts) = 0;
  113. __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
  114. #ifdef CONFIG_PROVE_RCU
  115. int lockdep_tasklist_lock_is_held(void)
  116. {
  117. return lockdep_is_held(&tasklist_lock);
  118. }
  119. EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
  120. #endif /* #ifdef CONFIG_PROVE_RCU */
  121. int nr_processes(void)
  122. {
  123. int cpu;
  124. int total = 0;
  125. for_each_possible_cpu(cpu)
  126. total += per_cpu(process_counts, cpu);
  127. return total;
  128. }
  129. void __weak arch_release_task_struct(struct task_struct *tsk)
  130. {
  131. }
  132. #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
  133. static struct kmem_cache *task_struct_cachep;
  134. static inline struct task_struct *alloc_task_struct_node(int node)
  135. {
  136. return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
  137. }
  138. static inline void free_task_struct(struct task_struct *tsk)
  139. {
  140. kmem_cache_free(task_struct_cachep, tsk);
  141. }
  142. #endif
  143. void __weak arch_release_thread_stack(unsigned long *stack)
  144. {
  145. }
  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. local_irq_disable();
  179. for (i = 0; i < NR_CACHED_STACKS; i++) {
  180. struct vm_struct *s = this_cpu_read(cached_stacks[i]);
  181. if (!s)
  182. continue;
  183. this_cpu_write(cached_stacks[i], NULL);
  184. tsk->stack_vm_area = s;
  185. local_irq_enable();
  186. return s->addr;
  187. }
  188. local_irq_enable();
  189. stack = __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
  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. return stack;
  202. #else
  203. struct page *page = alloc_pages_node(node, THREADINFO_GFP,
  204. THREAD_SIZE_ORDER);
  205. return page ? page_address(page) : NULL;
  206. #endif
  207. }
  208. static inline void free_thread_stack(struct task_struct *tsk)
  209. {
  210. #ifdef CONFIG_VMAP_STACK
  211. if (task_stack_vm_area(tsk)) {
  212. unsigned long flags;
  213. int i;
  214. local_irq_save(flags);
  215. for (i = 0; i < NR_CACHED_STACKS; i++) {
  216. if (this_cpu_read(cached_stacks[i]))
  217. continue;
  218. this_cpu_write(cached_stacks[i], tsk->stack_vm_area);
  219. local_irq_restore(flags);
  220. return;
  221. }
  222. local_irq_restore(flags);
  223. vfree_atomic(tsk->stack);
  224. return;
  225. }
  226. #endif
  227. __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
  228. }
  229. # else
  230. static struct kmem_cache *thread_stack_cache;
  231. static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
  232. int node)
  233. {
  234. return kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
  235. }
  236. static void free_thread_stack(struct task_struct *tsk)
  237. {
  238. kmem_cache_free(thread_stack_cache, tsk->stack);
  239. }
  240. void thread_stack_cache_init(void)
  241. {
  242. thread_stack_cache = kmem_cache_create("thread_stack", THREAD_SIZE,
  243. THREAD_SIZE, 0, NULL);
  244. BUG_ON(thread_stack_cache == NULL);
  245. }
  246. # endif
  247. #endif
  248. /* SLAB cache for signal_struct structures (tsk->signal) */
  249. static struct kmem_cache *signal_cachep;
  250. /* SLAB cache for sighand_struct structures (tsk->sighand) */
  251. struct kmem_cache *sighand_cachep;
  252. /* SLAB cache for files_struct structures (tsk->files) */
  253. struct kmem_cache *files_cachep;
  254. /* SLAB cache for fs_struct structures (tsk->fs) */
  255. struct kmem_cache *fs_cachep;
  256. /* SLAB cache for vm_area_struct structures */
  257. struct kmem_cache *vm_area_cachep;
  258. /* SLAB cache for mm_struct structures (tsk->mm) */
  259. static struct kmem_cache *mm_cachep;
  260. static void account_kernel_stack(struct task_struct *tsk, int account)
  261. {
  262. void *stack = task_stack_page(tsk);
  263. struct vm_struct *vm = task_stack_vm_area(tsk);
  264. BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0);
  265. if (vm) {
  266. int i;
  267. BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE);
  268. for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
  269. mod_zone_page_state(page_zone(vm->pages[i]),
  270. NR_KERNEL_STACK_KB,
  271. PAGE_SIZE / 1024 * account);
  272. }
  273. /* All stack pages belong to the same memcg. */
  274. mod_memcg_page_state(vm->pages[0], MEMCG_KERNEL_STACK_KB,
  275. account * (THREAD_SIZE / 1024));
  276. } else {
  277. /*
  278. * All stack pages are in the same zone and belong to the
  279. * same memcg.
  280. */
  281. struct page *first_page = virt_to_page(stack);
  282. mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB,
  283. THREAD_SIZE / 1024 * account);
  284. mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB,
  285. account * (THREAD_SIZE / 1024));
  286. }
  287. }
  288. static void release_task_stack(struct task_struct *tsk)
  289. {
  290. if (WARN_ON(tsk->state != TASK_DEAD))
  291. return; /* Better to leak the stack than to free prematurely */
  292. account_kernel_stack(tsk, -1);
  293. arch_release_thread_stack(tsk->stack);
  294. free_thread_stack(tsk);
  295. tsk->stack = NULL;
  296. #ifdef CONFIG_VMAP_STACK
  297. tsk->stack_vm_area = NULL;
  298. #endif
  299. }
  300. #ifdef CONFIG_THREAD_INFO_IN_TASK
  301. void put_task_stack(struct task_struct *tsk)
  302. {
  303. if (atomic_dec_and_test(&tsk->stack_refcount))
  304. release_task_stack(tsk);
  305. }
  306. #endif
  307. void free_task(struct task_struct *tsk)
  308. {
  309. #ifndef CONFIG_THREAD_INFO_IN_TASK
  310. /*
  311. * The task is finally done with both the stack and thread_info,
  312. * so free both.
  313. */
  314. release_task_stack(tsk);
  315. #else
  316. /*
  317. * If the task had a separate stack allocation, it should be gone
  318. * by now.
  319. */
  320. WARN_ON_ONCE(atomic_read(&tsk->stack_refcount) != 0);
  321. #endif
  322. rt_mutex_debug_task_free(tsk);
  323. ftrace_graph_exit_task(tsk);
  324. put_seccomp_filter(tsk);
  325. arch_release_task_struct(tsk);
  326. if (tsk->flags & PF_KTHREAD)
  327. free_kthread_struct(tsk);
  328. free_task_struct(tsk);
  329. }
  330. EXPORT_SYMBOL(free_task);
  331. static inline void free_signal_struct(struct signal_struct *sig)
  332. {
  333. taskstats_tgid_free(sig);
  334. sched_autogroup_exit(sig);
  335. /*
  336. * __mmdrop is not safe to call from softirq context on x86 due to
  337. * pgd_dtor so postpone it to the async context
  338. */
  339. if (sig->oom_mm)
  340. mmdrop_async(sig->oom_mm);
  341. kmem_cache_free(signal_cachep, sig);
  342. }
  343. static inline void put_signal_struct(struct signal_struct *sig)
  344. {
  345. if (atomic_dec_and_test(&sig->sigcnt))
  346. free_signal_struct(sig);
  347. }
  348. void __put_task_struct(struct task_struct *tsk)
  349. {
  350. WARN_ON(!tsk->exit_state);
  351. WARN_ON(atomic_read(&tsk->usage));
  352. WARN_ON(tsk == current);
  353. cgroup_free(tsk);
  354. task_numa_free(tsk);
  355. security_task_free(tsk);
  356. exit_creds(tsk);
  357. delayacct_tsk_free(tsk);
  358. put_signal_struct(tsk->signal);
  359. if (!profile_handoff_task(tsk))
  360. free_task(tsk);
  361. }
  362. EXPORT_SYMBOL_GPL(__put_task_struct);
  363. void __init __weak arch_task_cache_init(void) { }
  364. /*
  365. * set_max_threads
  366. */
  367. static void set_max_threads(unsigned int max_threads_suggested)
  368. {
  369. u64 threads;
  370. /*
  371. * The number of threads shall be limited such that the thread
  372. * structures may only consume a small part of the available memory.
  373. */
  374. if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64)
  375. threads = MAX_THREADS;
  376. else
  377. threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE,
  378. (u64) THREAD_SIZE * 8UL);
  379. if (threads > max_threads_suggested)
  380. threads = max_threads_suggested;
  381. max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
  382. }
  383. #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
  384. /* Initialized by the architecture: */
  385. int arch_task_struct_size __read_mostly;
  386. #endif
  387. void __init fork_init(void)
  388. {
  389. int i;
  390. #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
  391. #ifndef ARCH_MIN_TASKALIGN
  392. #define ARCH_MIN_TASKALIGN 0
  393. #endif
  394. int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
  395. /* create a slab on which task_structs can be allocated */
  396. task_struct_cachep = kmem_cache_create("task_struct",
  397. arch_task_struct_size, align,
  398. SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL);
  399. #endif
  400. /* do the arch specific task caches init */
  401. arch_task_cache_init();
  402. set_max_threads(MAX_THREADS);
  403. init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
  404. init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
  405. init_task.signal->rlim[RLIMIT_SIGPENDING] =
  406. init_task.signal->rlim[RLIMIT_NPROC];
  407. for (i = 0; i < UCOUNT_COUNTS; i++) {
  408. init_user_ns.ucount_max[i] = max_threads/2;
  409. }
  410. #ifdef CONFIG_VMAP_STACK
  411. cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache",
  412. NULL, free_vm_stack_cache);
  413. #endif
  414. }
  415. int __weak arch_dup_task_struct(struct task_struct *dst,
  416. struct task_struct *src)
  417. {
  418. *dst = *src;
  419. return 0;
  420. }
  421. void set_task_stack_end_magic(struct task_struct *tsk)
  422. {
  423. unsigned long *stackend;
  424. stackend = end_of_stack(tsk);
  425. *stackend = STACK_END_MAGIC; /* for overflow detection */
  426. }
  427. static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
  428. {
  429. struct task_struct *tsk;
  430. unsigned long *stack;
  431. struct vm_struct *stack_vm_area;
  432. int err;
  433. if (node == NUMA_NO_NODE)
  434. node = tsk_fork_get_node(orig);
  435. tsk = alloc_task_struct_node(node);
  436. if (!tsk)
  437. return NULL;
  438. stack = alloc_thread_stack_node(tsk, node);
  439. if (!stack)
  440. goto free_tsk;
  441. stack_vm_area = task_stack_vm_area(tsk);
  442. err = arch_dup_task_struct(tsk, orig);
  443. /*
  444. * arch_dup_task_struct() clobbers the stack-related fields. Make
  445. * sure they're properly initialized before using any stack-related
  446. * functions again.
  447. */
  448. tsk->stack = stack;
  449. #ifdef CONFIG_VMAP_STACK
  450. tsk->stack_vm_area = stack_vm_area;
  451. #endif
  452. #ifdef CONFIG_THREAD_INFO_IN_TASK
  453. atomic_set(&tsk->stack_refcount, 1);
  454. #endif
  455. if (err)
  456. goto free_stack;
  457. #ifdef CONFIG_SECCOMP
  458. /*
  459. * We must handle setting up seccomp filters once we're under
  460. * the sighand lock in case orig has changed between now and
  461. * then. Until then, filter must be NULL to avoid messing up
  462. * the usage counts on the error path calling free_task.
  463. */
  464. tsk->seccomp.filter = NULL;
  465. #endif
  466. setup_thread_stack(tsk, orig);
  467. clear_user_return_notifier(tsk);
  468. clear_tsk_need_resched(tsk);
  469. set_task_stack_end_magic(tsk);
  470. #ifdef CONFIG_CC_STACKPROTECTOR
  471. tsk->stack_canary = get_random_long();
  472. #endif
  473. /*
  474. * One for us, one for whoever does the "release_task()" (usually
  475. * parent)
  476. */
  477. atomic_set(&tsk->usage, 2);
  478. #ifdef CONFIG_BLK_DEV_IO_TRACE
  479. tsk->btrace_seq = 0;
  480. #endif
  481. tsk->splice_pipe = NULL;
  482. tsk->task_frag.page = NULL;
  483. tsk->wake_q.next = NULL;
  484. account_kernel_stack(tsk, 1);
  485. kcov_task_init(tsk);
  486. return tsk;
  487. free_stack:
  488. free_thread_stack(tsk);
  489. free_tsk:
  490. free_task_struct(tsk);
  491. return NULL;
  492. }
  493. #ifdef CONFIG_MMU
  494. static __latent_entropy int dup_mmap(struct mm_struct *mm,
  495. struct mm_struct *oldmm)
  496. {
  497. struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
  498. struct rb_node **rb_link, *rb_parent;
  499. int retval;
  500. unsigned long charge;
  501. LIST_HEAD(uf);
  502. uprobe_start_dup_mmap();
  503. if (down_write_killable(&oldmm->mmap_sem)) {
  504. retval = -EINTR;
  505. goto fail_uprobe_end;
  506. }
  507. flush_cache_dup_mm(oldmm);
  508. uprobe_dup_mmap(oldmm, mm);
  509. /*
  510. * Not linked in yet - no deadlock potential:
  511. */
  512. down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
  513. /* No ordering required: file already has been exposed. */
  514. RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
  515. mm->total_vm = oldmm->total_vm;
  516. mm->data_vm = oldmm->data_vm;
  517. mm->exec_vm = oldmm->exec_vm;
  518. mm->stack_vm = oldmm->stack_vm;
  519. rb_link = &mm->mm_rb.rb_node;
  520. rb_parent = NULL;
  521. pprev = &mm->mmap;
  522. retval = ksm_fork(mm, oldmm);
  523. if (retval)
  524. goto out;
  525. retval = khugepaged_fork(mm, oldmm);
  526. if (retval)
  527. goto out;
  528. prev = NULL;
  529. for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
  530. struct file *file;
  531. if (mpnt->vm_flags & VM_DONTCOPY) {
  532. vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
  533. continue;
  534. }
  535. charge = 0;
  536. if (mpnt->vm_flags & VM_ACCOUNT) {
  537. unsigned long len = vma_pages(mpnt);
  538. if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
  539. goto fail_nomem;
  540. charge = len;
  541. }
  542. tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
  543. if (!tmp)
  544. goto fail_nomem;
  545. *tmp = *mpnt;
  546. INIT_LIST_HEAD(&tmp->anon_vma_chain);
  547. retval = vma_dup_policy(mpnt, tmp);
  548. if (retval)
  549. goto fail_nomem_policy;
  550. tmp->vm_mm = mm;
  551. retval = dup_userfaultfd(tmp, &uf);
  552. if (retval)
  553. goto fail_nomem_anon_vma_fork;
  554. if (anon_vma_fork(tmp, mpnt))
  555. goto fail_nomem_anon_vma_fork;
  556. tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT);
  557. tmp->vm_next = tmp->vm_prev = NULL;
  558. file = tmp->vm_file;
  559. if (file) {
  560. struct inode *inode = file_inode(file);
  561. struct address_space *mapping = file->f_mapping;
  562. get_file(file);
  563. if (tmp->vm_flags & VM_DENYWRITE)
  564. atomic_dec(&inode->i_writecount);
  565. i_mmap_lock_write(mapping);
  566. if (tmp->vm_flags & VM_SHARED)
  567. atomic_inc(&mapping->i_mmap_writable);
  568. flush_dcache_mmap_lock(mapping);
  569. /* insert tmp into the share list, just after mpnt */
  570. vma_interval_tree_insert_after(tmp, mpnt,
  571. &mapping->i_mmap);
  572. flush_dcache_mmap_unlock(mapping);
  573. i_mmap_unlock_write(mapping);
  574. }
  575. /*
  576. * Clear hugetlb-related page reserves for children. This only
  577. * affects MAP_PRIVATE mappings. Faults generated by the child
  578. * are not guaranteed to succeed, even if read-only
  579. */
  580. if (is_vm_hugetlb_page(tmp))
  581. reset_vma_resv_huge_pages(tmp);
  582. /*
  583. * Link in the new vma and copy the page table entries.
  584. */
  585. *pprev = tmp;
  586. pprev = &tmp->vm_next;
  587. tmp->vm_prev = prev;
  588. prev = tmp;
  589. __vma_link_rb(mm, tmp, rb_link, rb_parent);
  590. rb_link = &tmp->vm_rb.rb_right;
  591. rb_parent = &tmp->vm_rb;
  592. mm->map_count++;
  593. retval = copy_page_range(mm, oldmm, mpnt);
  594. if (tmp->vm_ops && tmp->vm_ops->open)
  595. tmp->vm_ops->open(tmp);
  596. if (retval)
  597. goto out;
  598. }
  599. /* a new mm has just been created */
  600. arch_dup_mmap(oldmm, mm);
  601. retval = 0;
  602. out:
  603. up_write(&mm->mmap_sem);
  604. flush_tlb_mm(oldmm);
  605. up_write(&oldmm->mmap_sem);
  606. dup_userfaultfd_complete(&uf);
  607. fail_uprobe_end:
  608. uprobe_end_dup_mmap();
  609. return retval;
  610. fail_nomem_anon_vma_fork:
  611. mpol_put(vma_policy(tmp));
  612. fail_nomem_policy:
  613. kmem_cache_free(vm_area_cachep, tmp);
  614. fail_nomem:
  615. retval = -ENOMEM;
  616. vm_unacct_memory(charge);
  617. goto out;
  618. }
  619. static inline int mm_alloc_pgd(struct mm_struct *mm)
  620. {
  621. mm->pgd = pgd_alloc(mm);
  622. if (unlikely(!mm->pgd))
  623. return -ENOMEM;
  624. return 0;
  625. }
  626. static inline void mm_free_pgd(struct mm_struct *mm)
  627. {
  628. pgd_free(mm, mm->pgd);
  629. }
  630. #else
  631. static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
  632. {
  633. down_write(&oldmm->mmap_sem);
  634. RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
  635. up_write(&oldmm->mmap_sem);
  636. return 0;
  637. }
  638. #define mm_alloc_pgd(mm) (0)
  639. #define mm_free_pgd(mm)
  640. #endif /* CONFIG_MMU */
  641. __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
  642. #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
  643. #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
  644. static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
  645. static int __init coredump_filter_setup(char *s)
  646. {
  647. default_dump_filter =
  648. (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
  649. MMF_DUMP_FILTER_MASK;
  650. return 1;
  651. }
  652. __setup("coredump_filter=", coredump_filter_setup);
  653. #include <linux/init_task.h>
  654. static void mm_init_aio(struct mm_struct *mm)
  655. {
  656. #ifdef CONFIG_AIO
  657. spin_lock_init(&mm->ioctx_lock);
  658. mm->ioctx_table = NULL;
  659. #endif
  660. }
  661. static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
  662. {
  663. #ifdef CONFIG_MEMCG
  664. mm->owner = p;
  665. #endif
  666. }
  667. static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
  668. struct user_namespace *user_ns)
  669. {
  670. mm->mmap = NULL;
  671. mm->mm_rb = RB_ROOT;
  672. mm->vmacache_seqnum = 0;
  673. atomic_set(&mm->mm_users, 1);
  674. atomic_set(&mm->mm_count, 1);
  675. init_rwsem(&mm->mmap_sem);
  676. INIT_LIST_HEAD(&mm->mmlist);
  677. mm->core_state = NULL;
  678. atomic_long_set(&mm->nr_ptes, 0);
  679. mm_nr_pmds_init(mm);
  680. mm->map_count = 0;
  681. mm->locked_vm = 0;
  682. mm->pinned_vm = 0;
  683. memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
  684. spin_lock_init(&mm->page_table_lock);
  685. mm_init_cpumask(mm);
  686. mm_init_aio(mm);
  687. mm_init_owner(mm, p);
  688. mmu_notifier_mm_init(mm);
  689. clear_tlb_flush_pending(mm);
  690. #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
  691. mm->pmd_huge_pte = NULL;
  692. #endif
  693. if (current->mm) {
  694. mm->flags = current->mm->flags & MMF_INIT_MASK;
  695. mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
  696. } else {
  697. mm->flags = default_dump_filter;
  698. mm->def_flags = 0;
  699. }
  700. if (mm_alloc_pgd(mm))
  701. goto fail_nopgd;
  702. if (init_new_context(p, mm))
  703. goto fail_nocontext;
  704. mm->user_ns = get_user_ns(user_ns);
  705. return mm;
  706. fail_nocontext:
  707. mm_free_pgd(mm);
  708. fail_nopgd:
  709. free_mm(mm);
  710. return NULL;
  711. }
  712. static void check_mm(struct mm_struct *mm)
  713. {
  714. int i;
  715. for (i = 0; i < NR_MM_COUNTERS; i++) {
  716. long x = atomic_long_read(&mm->rss_stat.count[i]);
  717. if (unlikely(x))
  718. printk(KERN_ALERT "BUG: Bad rss-counter state "
  719. "mm:%p idx:%d val:%ld\n", mm, i, x);
  720. }
  721. if (atomic_long_read(&mm->nr_ptes))
  722. pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n",
  723. atomic_long_read(&mm->nr_ptes));
  724. if (mm_nr_pmds(mm))
  725. pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n",
  726. mm_nr_pmds(mm));
  727. #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
  728. VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
  729. #endif
  730. }
  731. /*
  732. * Allocate and initialize an mm_struct.
  733. */
  734. struct mm_struct *mm_alloc(void)
  735. {
  736. struct mm_struct *mm;
  737. mm = allocate_mm();
  738. if (!mm)
  739. return NULL;
  740. memset(mm, 0, sizeof(*mm));
  741. return mm_init(mm, current, current_user_ns());
  742. }
  743. /*
  744. * Called when the last reference to the mm
  745. * is dropped: either by a lazy thread or by
  746. * mmput. Free the page directory and the mm.
  747. */
  748. void __mmdrop(struct mm_struct *mm)
  749. {
  750. BUG_ON(mm == &init_mm);
  751. mm_free_pgd(mm);
  752. destroy_context(mm);
  753. mmu_notifier_mm_destroy(mm);
  754. check_mm(mm);
  755. put_user_ns(mm->user_ns);
  756. free_mm(mm);
  757. }
  758. EXPORT_SYMBOL_GPL(__mmdrop);
  759. static inline void __mmput(struct mm_struct *mm)
  760. {
  761. VM_BUG_ON(atomic_read(&mm->mm_users));
  762. uprobe_clear_state(mm);
  763. exit_aio(mm);
  764. ksm_exit(mm);
  765. khugepaged_exit(mm); /* must run before exit_mmap */
  766. exit_mmap(mm);
  767. mm_put_huge_zero_page(mm);
  768. set_mm_exe_file(mm, NULL);
  769. if (!list_empty(&mm->mmlist)) {
  770. spin_lock(&mmlist_lock);
  771. list_del(&mm->mmlist);
  772. spin_unlock(&mmlist_lock);
  773. }
  774. if (mm->binfmt)
  775. module_put(mm->binfmt->module);
  776. set_bit(MMF_OOM_SKIP, &mm->flags);
  777. mmdrop(mm);
  778. }
  779. /*
  780. * Decrement the use count and release all resources for an mm.
  781. */
  782. void mmput(struct mm_struct *mm)
  783. {
  784. might_sleep();
  785. if (atomic_dec_and_test(&mm->mm_users))
  786. __mmput(mm);
  787. }
  788. EXPORT_SYMBOL_GPL(mmput);
  789. #ifdef CONFIG_MMU
  790. static void mmput_async_fn(struct work_struct *work)
  791. {
  792. struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
  793. __mmput(mm);
  794. }
  795. void mmput_async(struct mm_struct *mm)
  796. {
  797. if (atomic_dec_and_test(&mm->mm_users)) {
  798. INIT_WORK(&mm->async_put_work, mmput_async_fn);
  799. schedule_work(&mm->async_put_work);
  800. }
  801. }
  802. #endif
  803. /**
  804. * set_mm_exe_file - change a reference to the mm's executable file
  805. *
  806. * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
  807. *
  808. * Main users are mmput() and sys_execve(). Callers prevent concurrent
  809. * invocations: in mmput() nobody alive left, in execve task is single
  810. * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
  811. * mm->exe_file, but does so without using set_mm_exe_file() in order
  812. * to do avoid the need for any locks.
  813. */
  814. void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
  815. {
  816. struct file *old_exe_file;
  817. /*
  818. * It is safe to dereference the exe_file without RCU as
  819. * this function is only called if nobody else can access
  820. * this mm -- see comment above for justification.
  821. */
  822. old_exe_file = rcu_dereference_raw(mm->exe_file);
  823. if (new_exe_file)
  824. get_file(new_exe_file);
  825. rcu_assign_pointer(mm->exe_file, new_exe_file);
  826. if (old_exe_file)
  827. fput(old_exe_file);
  828. }
  829. /**
  830. * get_mm_exe_file - acquire a reference to the mm's executable file
  831. *
  832. * Returns %NULL if mm has no associated executable file.
  833. * User must release file via fput().
  834. */
  835. struct file *get_mm_exe_file(struct mm_struct *mm)
  836. {
  837. struct file *exe_file;
  838. rcu_read_lock();
  839. exe_file = rcu_dereference(mm->exe_file);
  840. if (exe_file && !get_file_rcu(exe_file))
  841. exe_file = NULL;
  842. rcu_read_unlock();
  843. return exe_file;
  844. }
  845. EXPORT_SYMBOL(get_mm_exe_file);
  846. /**
  847. * get_task_exe_file - acquire a reference to the task's executable file
  848. *
  849. * Returns %NULL if task's mm (if any) has no associated executable file or
  850. * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
  851. * User must release file via fput().
  852. */
  853. struct file *get_task_exe_file(struct task_struct *task)
  854. {
  855. struct file *exe_file = NULL;
  856. struct mm_struct *mm;
  857. task_lock(task);
  858. mm = task->mm;
  859. if (mm) {
  860. if (!(task->flags & PF_KTHREAD))
  861. exe_file = get_mm_exe_file(mm);
  862. }
  863. task_unlock(task);
  864. return exe_file;
  865. }
  866. EXPORT_SYMBOL(get_task_exe_file);
  867. /**
  868. * get_task_mm - acquire a reference to the task's mm
  869. *
  870. * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
  871. * this kernel workthread has transiently adopted a user mm with use_mm,
  872. * to do its AIO) is not set and if so returns a reference to it, after
  873. * bumping up the use count. User must release the mm via mmput()
  874. * after use. Typically used by /proc and ptrace.
  875. */
  876. struct mm_struct *get_task_mm(struct task_struct *task)
  877. {
  878. struct mm_struct *mm;
  879. task_lock(task);
  880. mm = task->mm;
  881. if (mm) {
  882. if (task->flags & PF_KTHREAD)
  883. mm = NULL;
  884. else
  885. mmget(mm);
  886. }
  887. task_unlock(task);
  888. return mm;
  889. }
  890. EXPORT_SYMBOL_GPL(get_task_mm);
  891. struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
  892. {
  893. struct mm_struct *mm;
  894. int err;
  895. err = mutex_lock_killable(&task->signal->cred_guard_mutex);
  896. if (err)
  897. return ERR_PTR(err);
  898. mm = get_task_mm(task);
  899. if (mm && mm != current->mm &&
  900. !ptrace_may_access(task, mode)) {
  901. mmput(mm);
  902. mm = ERR_PTR(-EACCES);
  903. }
  904. mutex_unlock(&task->signal->cred_guard_mutex);
  905. return mm;
  906. }
  907. static void complete_vfork_done(struct task_struct *tsk)
  908. {
  909. struct completion *vfork;
  910. task_lock(tsk);
  911. vfork = tsk->vfork_done;
  912. if (likely(vfork)) {
  913. tsk->vfork_done = NULL;
  914. complete(vfork);
  915. }
  916. task_unlock(tsk);
  917. }
  918. static int wait_for_vfork_done(struct task_struct *child,
  919. struct completion *vfork)
  920. {
  921. int killed;
  922. freezer_do_not_count();
  923. killed = wait_for_completion_killable(vfork);
  924. freezer_count();
  925. if (killed) {
  926. task_lock(child);
  927. child->vfork_done = NULL;
  928. task_unlock(child);
  929. }
  930. put_task_struct(child);
  931. return killed;
  932. }
  933. /* Please note the differences between mmput and mm_release.
  934. * mmput is called whenever we stop holding onto a mm_struct,
  935. * error success whatever.
  936. *
  937. * mm_release is called after a mm_struct has been removed
  938. * from the current process.
  939. *
  940. * This difference is important for error handling, when we
  941. * only half set up a mm_struct for a new process and need to restore
  942. * the old one. Because we mmput the new mm_struct before
  943. * restoring the old one. . .
  944. * Eric Biederman 10 January 1998
  945. */
  946. void mm_release(struct task_struct *tsk, struct mm_struct *mm)
  947. {
  948. /* Get rid of any futexes when releasing the mm */
  949. #ifdef CONFIG_FUTEX
  950. if (unlikely(tsk->robust_list)) {
  951. exit_robust_list(tsk);
  952. tsk->robust_list = NULL;
  953. }
  954. #ifdef CONFIG_COMPAT
  955. if (unlikely(tsk->compat_robust_list)) {
  956. compat_exit_robust_list(tsk);
  957. tsk->compat_robust_list = NULL;
  958. }
  959. #endif
  960. if (unlikely(!list_empty(&tsk->pi_state_list)))
  961. exit_pi_state_list(tsk);
  962. #endif
  963. uprobe_free_utask(tsk);
  964. /* Get rid of any cached register state */
  965. deactivate_mm(tsk, mm);
  966. /*
  967. * Signal userspace if we're not exiting with a core dump
  968. * because we want to leave the value intact for debugging
  969. * purposes.
  970. */
  971. if (tsk->clear_child_tid) {
  972. if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
  973. atomic_read(&mm->mm_users) > 1) {
  974. /*
  975. * We don't check the error code - if userspace has
  976. * not set up a proper pointer then tough luck.
  977. */
  978. put_user(0, tsk->clear_child_tid);
  979. sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
  980. 1, NULL, NULL, 0);
  981. }
  982. tsk->clear_child_tid = NULL;
  983. }
  984. /*
  985. * All done, finally we can wake up parent and return this mm to him.
  986. * Also kthread_stop() uses this completion for synchronization.
  987. */
  988. if (tsk->vfork_done)
  989. complete_vfork_done(tsk);
  990. }
  991. /*
  992. * Allocate a new mm structure and copy contents from the
  993. * mm structure of the passed in task structure.
  994. */
  995. static struct mm_struct *dup_mm(struct task_struct *tsk)
  996. {
  997. struct mm_struct *mm, *oldmm = current->mm;
  998. int err;
  999. mm = allocate_mm();
  1000. if (!mm)
  1001. goto fail_nomem;
  1002. memcpy(mm, oldmm, sizeof(*mm));
  1003. if (!mm_init(mm, tsk, mm->user_ns))
  1004. goto fail_nomem;
  1005. err = dup_mmap(mm, oldmm);
  1006. if (err)
  1007. goto free_pt;
  1008. mm->hiwater_rss = get_mm_rss(mm);
  1009. mm->hiwater_vm = mm->total_vm;
  1010. if (mm->binfmt && !try_module_get(mm->binfmt->module))
  1011. goto free_pt;
  1012. return mm;
  1013. free_pt:
  1014. /* don't put binfmt in mmput, we haven't got module yet */
  1015. mm->binfmt = NULL;
  1016. mmput(mm);
  1017. fail_nomem:
  1018. return NULL;
  1019. }
  1020. static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
  1021. {
  1022. struct mm_struct *mm, *oldmm;
  1023. int retval;
  1024. tsk->min_flt = tsk->maj_flt = 0;
  1025. tsk->nvcsw = tsk->nivcsw = 0;
  1026. #ifdef CONFIG_DETECT_HUNG_TASK
  1027. tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
  1028. #endif
  1029. tsk->mm = NULL;
  1030. tsk->active_mm = NULL;
  1031. /*
  1032. * Are we cloning a kernel thread?
  1033. *
  1034. * We need to steal a active VM for that..
  1035. */
  1036. oldmm = current->mm;
  1037. if (!oldmm)
  1038. return 0;
  1039. /* initialize the new vmacache entries */
  1040. vmacache_flush(tsk);
  1041. if (clone_flags & CLONE_VM) {
  1042. mmget(oldmm);
  1043. mm = oldmm;
  1044. goto good_mm;
  1045. }
  1046. retval = -ENOMEM;
  1047. mm = dup_mm(tsk);
  1048. if (!mm)
  1049. goto fail_nomem;
  1050. good_mm:
  1051. tsk->mm = mm;
  1052. tsk->active_mm = mm;
  1053. return 0;
  1054. fail_nomem:
  1055. return retval;
  1056. }
  1057. static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
  1058. {
  1059. struct fs_struct *fs = current->fs;
  1060. if (clone_flags & CLONE_FS) {
  1061. /* tsk->fs is already what we want */
  1062. spin_lock(&fs->lock);
  1063. if (fs->in_exec) {
  1064. spin_unlock(&fs->lock);
  1065. return -EAGAIN;
  1066. }
  1067. fs->users++;
  1068. spin_unlock(&fs->lock);
  1069. return 0;
  1070. }
  1071. tsk->fs = copy_fs_struct(fs);
  1072. if (!tsk->fs)
  1073. return -ENOMEM;
  1074. return 0;
  1075. }
  1076. static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
  1077. {
  1078. struct files_struct *oldf, *newf;
  1079. int error = 0;
  1080. /*
  1081. * A background process may not have any files ...
  1082. */
  1083. oldf = current->files;
  1084. if (!oldf)
  1085. goto out;
  1086. if (clone_flags & CLONE_FILES) {
  1087. atomic_inc(&oldf->count);
  1088. goto out;
  1089. }
  1090. newf = dup_fd(oldf, &error);
  1091. if (!newf)
  1092. goto out;
  1093. tsk->files = newf;
  1094. error = 0;
  1095. out:
  1096. return error;
  1097. }
  1098. static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
  1099. {
  1100. #ifdef CONFIG_BLOCK
  1101. struct io_context *ioc = current->io_context;
  1102. struct io_context *new_ioc;
  1103. if (!ioc)
  1104. return 0;
  1105. /*
  1106. * Share io context with parent, if CLONE_IO is set
  1107. */
  1108. if (clone_flags & CLONE_IO) {
  1109. ioc_task_link(ioc);
  1110. tsk->io_context = ioc;
  1111. } else if (ioprio_valid(ioc->ioprio)) {
  1112. new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
  1113. if (unlikely(!new_ioc))
  1114. return -ENOMEM;
  1115. new_ioc->ioprio = ioc->ioprio;
  1116. put_io_context(new_ioc);
  1117. }
  1118. #endif
  1119. return 0;
  1120. }
  1121. static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
  1122. {
  1123. struct sighand_struct *sig;
  1124. if (clone_flags & CLONE_SIGHAND) {
  1125. atomic_inc(&current->sighand->count);
  1126. return 0;
  1127. }
  1128. sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
  1129. rcu_assign_pointer(tsk->sighand, sig);
  1130. if (!sig)
  1131. return -ENOMEM;
  1132. atomic_set(&sig->count, 1);
  1133. memcpy(sig->action, current->sighand->action, sizeof(sig->action));
  1134. return 0;
  1135. }
  1136. void __cleanup_sighand(struct sighand_struct *sighand)
  1137. {
  1138. if (atomic_dec_and_test(&sighand->count)) {
  1139. signalfd_cleanup(sighand);
  1140. /*
  1141. * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it
  1142. * without an RCU grace period, see __lock_task_sighand().
  1143. */
  1144. kmem_cache_free(sighand_cachep, sighand);
  1145. }
  1146. }
  1147. #ifdef CONFIG_POSIX_TIMERS
  1148. /*
  1149. * Initialize POSIX timer handling for a thread group.
  1150. */
  1151. static void posix_cpu_timers_init_group(struct signal_struct *sig)
  1152. {
  1153. unsigned long cpu_limit;
  1154. cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
  1155. if (cpu_limit != RLIM_INFINITY) {
  1156. sig->cputime_expires.prof_exp = cpu_limit * NSEC_PER_SEC;
  1157. sig->cputimer.running = true;
  1158. }
  1159. /* The timer lists. */
  1160. INIT_LIST_HEAD(&sig->cpu_timers[0]);
  1161. INIT_LIST_HEAD(&sig->cpu_timers[1]);
  1162. INIT_LIST_HEAD(&sig->cpu_timers[2]);
  1163. }
  1164. #else
  1165. static inline void posix_cpu_timers_init_group(struct signal_struct *sig) { }
  1166. #endif
  1167. static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
  1168. {
  1169. struct signal_struct *sig;
  1170. if (clone_flags & CLONE_THREAD)
  1171. return 0;
  1172. sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
  1173. tsk->signal = sig;
  1174. if (!sig)
  1175. return -ENOMEM;
  1176. sig->nr_threads = 1;
  1177. atomic_set(&sig->live, 1);
  1178. atomic_set(&sig->sigcnt, 1);
  1179. /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
  1180. sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
  1181. tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
  1182. init_waitqueue_head(&sig->wait_chldexit);
  1183. sig->curr_target = tsk;
  1184. init_sigpending(&sig->shared_pending);
  1185. seqlock_init(&sig->stats_lock);
  1186. prev_cputime_init(&sig->prev_cputime);
  1187. #ifdef CONFIG_POSIX_TIMERS
  1188. INIT_LIST_HEAD(&sig->posix_timers);
  1189. hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
  1190. sig->real_timer.function = it_real_fn;
  1191. #endif
  1192. task_lock(current->group_leader);
  1193. memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
  1194. task_unlock(current->group_leader);
  1195. posix_cpu_timers_init_group(sig);
  1196. tty_audit_fork(sig);
  1197. sched_autogroup_fork(sig);
  1198. sig->oom_score_adj = current->signal->oom_score_adj;
  1199. sig->oom_score_adj_min = current->signal->oom_score_adj_min;
  1200. mutex_init(&sig->cred_guard_mutex);
  1201. return 0;
  1202. }
  1203. static void copy_seccomp(struct task_struct *p)
  1204. {
  1205. #ifdef CONFIG_SECCOMP
  1206. /*
  1207. * Must be called with sighand->lock held, which is common to
  1208. * all threads in the group. Holding cred_guard_mutex is not
  1209. * needed because this new task is not yet running and cannot
  1210. * be racing exec.
  1211. */
  1212. assert_spin_locked(&current->sighand->siglock);
  1213. /* Ref-count the new filter user, and assign it. */
  1214. get_seccomp_filter(current);
  1215. p->seccomp = current->seccomp;
  1216. /*
  1217. * Explicitly enable no_new_privs here in case it got set
  1218. * between the task_struct being duplicated and holding the
  1219. * sighand lock. The seccomp state and nnp must be in sync.
  1220. */
  1221. if (task_no_new_privs(current))
  1222. task_set_no_new_privs(p);
  1223. /*
  1224. * If the parent gained a seccomp mode after copying thread
  1225. * flags and between before we held the sighand lock, we have
  1226. * to manually enable the seccomp thread flag here.
  1227. */
  1228. if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
  1229. set_tsk_thread_flag(p, TIF_SECCOMP);
  1230. #endif
  1231. }
  1232. SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
  1233. {
  1234. current->clear_child_tid = tidptr;
  1235. return task_pid_vnr(current);
  1236. }
  1237. static void rt_mutex_init_task(struct task_struct *p)
  1238. {
  1239. raw_spin_lock_init(&p->pi_lock);
  1240. #ifdef CONFIG_RT_MUTEXES
  1241. p->pi_waiters = RB_ROOT;
  1242. p->pi_waiters_leftmost = NULL;
  1243. p->pi_top_task = NULL;
  1244. p->pi_blocked_on = NULL;
  1245. #endif
  1246. }
  1247. #ifdef CONFIG_POSIX_TIMERS
  1248. /*
  1249. * Initialize POSIX timer handling for a single task.
  1250. */
  1251. static void posix_cpu_timers_init(struct task_struct *tsk)
  1252. {
  1253. tsk->cputime_expires.prof_exp = 0;
  1254. tsk->cputime_expires.virt_exp = 0;
  1255. tsk->cputime_expires.sched_exp = 0;
  1256. INIT_LIST_HEAD(&tsk->cpu_timers[0]);
  1257. INIT_LIST_HEAD(&tsk->cpu_timers[1]);
  1258. INIT_LIST_HEAD(&tsk->cpu_timers[2]);
  1259. }
  1260. #else
  1261. static inline void posix_cpu_timers_init(struct task_struct *tsk) { }
  1262. #endif
  1263. static inline void
  1264. init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
  1265. {
  1266. task->pids[type].pid = pid;
  1267. }
  1268. static inline void rcu_copy_process(struct task_struct *p)
  1269. {
  1270. #ifdef CONFIG_PREEMPT_RCU
  1271. p->rcu_read_lock_nesting = 0;
  1272. p->rcu_read_unlock_special.s = 0;
  1273. p->rcu_blocked_node = NULL;
  1274. INIT_LIST_HEAD(&p->rcu_node_entry);
  1275. #endif /* #ifdef CONFIG_PREEMPT_RCU */
  1276. #ifdef CONFIG_TASKS_RCU
  1277. p->rcu_tasks_holdout = false;
  1278. INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
  1279. p->rcu_tasks_idle_cpu = -1;
  1280. #endif /* #ifdef CONFIG_TASKS_RCU */
  1281. }
  1282. /*
  1283. * This creates a new process as a copy of the old one,
  1284. * but does not actually start it yet.
  1285. *
  1286. * It copies the registers, and all the appropriate
  1287. * parts of the process environment (as per the clone
  1288. * flags). The actual kick-off is left to the caller.
  1289. */
  1290. static __latent_entropy struct task_struct *copy_process(
  1291. unsigned long clone_flags,
  1292. unsigned long stack_start,
  1293. unsigned long stack_size,
  1294. int __user *child_tidptr,
  1295. struct pid *pid,
  1296. int trace,
  1297. unsigned long tls,
  1298. int node)
  1299. {
  1300. int retval;
  1301. struct task_struct *p;
  1302. if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
  1303. return ERR_PTR(-EINVAL);
  1304. if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
  1305. return ERR_PTR(-EINVAL);
  1306. /*
  1307. * Thread groups must share signals as well, and detached threads
  1308. * can only be started up within the thread group.
  1309. */
  1310. if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
  1311. return ERR_PTR(-EINVAL);
  1312. /*
  1313. * Shared signal handlers imply shared VM. By way of the above,
  1314. * thread groups also imply shared VM. Blocking this case allows
  1315. * for various simplifications in other code.
  1316. */
  1317. if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
  1318. return ERR_PTR(-EINVAL);
  1319. /*
  1320. * Siblings of global init remain as zombies on exit since they are
  1321. * not reaped by their parent (swapper). To solve this and to avoid
  1322. * multi-rooted process trees, prevent global and container-inits
  1323. * from creating siblings.
  1324. */
  1325. if ((clone_flags & CLONE_PARENT) &&
  1326. current->signal->flags & SIGNAL_UNKILLABLE)
  1327. return ERR_PTR(-EINVAL);
  1328. /*
  1329. * If the new process will be in a different pid or user namespace
  1330. * do not allow it to share a thread group with the forking task.
  1331. */
  1332. if (clone_flags & CLONE_THREAD) {
  1333. if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
  1334. (task_active_pid_ns(current) !=
  1335. current->nsproxy->pid_ns_for_children))
  1336. return ERR_PTR(-EINVAL);
  1337. }
  1338. retval = security_task_create(clone_flags);
  1339. if (retval)
  1340. goto fork_out;
  1341. retval = -ENOMEM;
  1342. p = dup_task_struct(current, node);
  1343. if (!p)
  1344. goto fork_out;
  1345. /*
  1346. * This _must_ happen before we call free_task(), i.e. before we jump
  1347. * to any of the bad_fork_* labels. This is to avoid freeing
  1348. * p->set_child_tid which is (ab)used as a kthread's data pointer for
  1349. * kernel threads (PF_KTHREAD).
  1350. */
  1351. p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
  1352. /*
  1353. * Clear TID on mm_release()?
  1354. */
  1355. p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
  1356. ftrace_graph_init_task(p);
  1357. rt_mutex_init_task(p);
  1358. #ifdef CONFIG_PROVE_LOCKING
  1359. DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
  1360. DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
  1361. #endif
  1362. retval = -EAGAIN;
  1363. if (atomic_read(&p->real_cred->user->processes) >=
  1364. task_rlimit(p, RLIMIT_NPROC)) {
  1365. if (p->real_cred->user != INIT_USER &&
  1366. !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
  1367. goto bad_fork_free;
  1368. }
  1369. current->flags &= ~PF_NPROC_EXCEEDED;
  1370. retval = copy_creds(p, clone_flags);
  1371. if (retval < 0)
  1372. goto bad_fork_free;
  1373. /*
  1374. * If multiple threads are within copy_process(), then this check
  1375. * triggers too late. This doesn't hurt, the check is only there
  1376. * to stop root fork bombs.
  1377. */
  1378. retval = -EAGAIN;
  1379. if (nr_threads >= max_threads)
  1380. goto bad_fork_cleanup_count;
  1381. delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
  1382. p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE);
  1383. p->flags |= PF_FORKNOEXEC;
  1384. INIT_LIST_HEAD(&p->children);
  1385. INIT_LIST_HEAD(&p->sibling);
  1386. rcu_copy_process(p);
  1387. p->vfork_done = NULL;
  1388. spin_lock_init(&p->alloc_lock);
  1389. init_sigpending(&p->pending);
  1390. p->utime = p->stime = p->gtime = 0;
  1391. #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
  1392. p->utimescaled = p->stimescaled = 0;
  1393. #endif
  1394. prev_cputime_init(&p->prev_cputime);
  1395. #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
  1396. seqcount_init(&p->vtime.seqcount);
  1397. p->vtime.starttime = 0;
  1398. p->vtime.state = VTIME_INACTIVE;
  1399. #endif
  1400. #if defined(SPLIT_RSS_COUNTING)
  1401. memset(&p->rss_stat, 0, sizeof(p->rss_stat));
  1402. #endif
  1403. p->default_timer_slack_ns = current->timer_slack_ns;
  1404. task_io_accounting_init(&p->ioac);
  1405. acct_clear_integrals(p);
  1406. posix_cpu_timers_init(p);
  1407. p->start_time = ktime_get_ns();
  1408. p->real_start_time = ktime_get_boot_ns();
  1409. p->io_context = NULL;
  1410. p->audit_context = NULL;
  1411. cgroup_fork(p);
  1412. #ifdef CONFIG_NUMA
  1413. p->mempolicy = mpol_dup(p->mempolicy);
  1414. if (IS_ERR(p->mempolicy)) {
  1415. retval = PTR_ERR(p->mempolicy);
  1416. p->mempolicy = NULL;
  1417. goto bad_fork_cleanup_threadgroup_lock;
  1418. }
  1419. #endif
  1420. #ifdef CONFIG_CPUSETS
  1421. p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
  1422. p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
  1423. seqcount_init(&p->mems_allowed_seq);
  1424. #endif
  1425. #ifdef CONFIG_TRACE_IRQFLAGS
  1426. p->irq_events = 0;
  1427. p->hardirqs_enabled = 0;
  1428. p->hardirq_enable_ip = 0;
  1429. p->hardirq_enable_event = 0;
  1430. p->hardirq_disable_ip = _THIS_IP_;
  1431. p->hardirq_disable_event = 0;
  1432. p->softirqs_enabled = 1;
  1433. p->softirq_enable_ip = _THIS_IP_;
  1434. p->softirq_enable_event = 0;
  1435. p->softirq_disable_ip = 0;
  1436. p->softirq_disable_event = 0;
  1437. p->hardirq_context = 0;
  1438. p->softirq_context = 0;
  1439. #endif
  1440. p->pagefault_disabled = 0;
  1441. #ifdef CONFIG_LOCKDEP
  1442. p->lockdep_depth = 0; /* no locks held yet */
  1443. p->curr_chain_key = 0;
  1444. p->lockdep_recursion = 0;
  1445. #endif
  1446. #ifdef CONFIG_DEBUG_MUTEXES
  1447. p->blocked_on = NULL; /* not blocked yet */
  1448. #endif
  1449. #ifdef CONFIG_BCACHE
  1450. p->sequential_io = 0;
  1451. p->sequential_io_avg = 0;
  1452. #endif
  1453. /* Perform scheduler related setup. Assign this task to a CPU. */
  1454. retval = sched_fork(clone_flags, p);
  1455. if (retval)
  1456. goto bad_fork_cleanup_policy;
  1457. retval = perf_event_init_task(p);
  1458. if (retval)
  1459. goto bad_fork_cleanup_policy;
  1460. retval = audit_alloc(p);
  1461. if (retval)
  1462. goto bad_fork_cleanup_perf;
  1463. /* copy all the process information */
  1464. shm_init_task(p);
  1465. retval = security_task_alloc(p, clone_flags);
  1466. if (retval)
  1467. goto bad_fork_cleanup_audit;
  1468. retval = copy_semundo(clone_flags, p);
  1469. if (retval)
  1470. goto bad_fork_cleanup_security;
  1471. retval = copy_files(clone_flags, p);
  1472. if (retval)
  1473. goto bad_fork_cleanup_semundo;
  1474. retval = copy_fs(clone_flags, p);
  1475. if (retval)
  1476. goto bad_fork_cleanup_files;
  1477. retval = copy_sighand(clone_flags, p);
  1478. if (retval)
  1479. goto bad_fork_cleanup_fs;
  1480. retval = copy_signal(clone_flags, p);
  1481. if (retval)
  1482. goto bad_fork_cleanup_sighand;
  1483. retval = copy_mm(clone_flags, p);
  1484. if (retval)
  1485. goto bad_fork_cleanup_signal;
  1486. retval = copy_namespaces(clone_flags, p);
  1487. if (retval)
  1488. goto bad_fork_cleanup_mm;
  1489. retval = copy_io(clone_flags, p);
  1490. if (retval)
  1491. goto bad_fork_cleanup_namespaces;
  1492. retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls);
  1493. if (retval)
  1494. goto bad_fork_cleanup_io;
  1495. if (pid != &init_struct_pid) {
  1496. pid = alloc_pid(p->nsproxy->pid_ns_for_children);
  1497. if (IS_ERR(pid)) {
  1498. retval = PTR_ERR(pid);
  1499. goto bad_fork_cleanup_thread;
  1500. }
  1501. }
  1502. #ifdef CONFIG_BLOCK
  1503. p->plug = NULL;
  1504. #endif
  1505. #ifdef CONFIG_FUTEX
  1506. p->robust_list = NULL;
  1507. #ifdef CONFIG_COMPAT
  1508. p->compat_robust_list = NULL;
  1509. #endif
  1510. INIT_LIST_HEAD(&p->pi_state_list);
  1511. p->pi_state_cache = NULL;
  1512. #endif
  1513. /*
  1514. * sigaltstack should be cleared when sharing the same VM
  1515. */
  1516. if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
  1517. sas_ss_reset(p);
  1518. /*
  1519. * Syscall tracing and stepping should be turned off in the
  1520. * child regardless of CLONE_PTRACE.
  1521. */
  1522. user_disable_single_step(p);
  1523. clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
  1524. #ifdef TIF_SYSCALL_EMU
  1525. clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
  1526. #endif
  1527. clear_all_latency_tracing(p);
  1528. /* ok, now we should be set up.. */
  1529. p->pid = pid_nr(pid);
  1530. if (clone_flags & CLONE_THREAD) {
  1531. p->exit_signal = -1;
  1532. p->group_leader = current->group_leader;
  1533. p->tgid = current->tgid;
  1534. } else {
  1535. if (clone_flags & CLONE_PARENT)
  1536. p->exit_signal = current->group_leader->exit_signal;
  1537. else
  1538. p->exit_signal = (clone_flags & CSIGNAL);
  1539. p->group_leader = p;
  1540. p->tgid = p->pid;
  1541. }
  1542. p->nr_dirtied = 0;
  1543. p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
  1544. p->dirty_paused_when = 0;
  1545. p->pdeath_signal = 0;
  1546. INIT_LIST_HEAD(&p->thread_group);
  1547. p->task_works = NULL;
  1548. cgroup_threadgroup_change_begin(current);
  1549. /*
  1550. * Ensure that the cgroup subsystem policies allow the new process to be
  1551. * forked. It should be noted the the new process's css_set can be changed
  1552. * between here and cgroup_post_fork() if an organisation operation is in
  1553. * progress.
  1554. */
  1555. retval = cgroup_can_fork(p);
  1556. if (retval)
  1557. goto bad_fork_free_pid;
  1558. /*
  1559. * Make it visible to the rest of the system, but dont wake it up yet.
  1560. * Need tasklist lock for parent etc handling!
  1561. */
  1562. write_lock_irq(&tasklist_lock);
  1563. /* CLONE_PARENT re-uses the old parent */
  1564. if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
  1565. p->real_parent = current->real_parent;
  1566. p->parent_exec_id = current->parent_exec_id;
  1567. } else {
  1568. p->real_parent = current;
  1569. p->parent_exec_id = current->self_exec_id;
  1570. }
  1571. klp_copy_process(p);
  1572. spin_lock(&current->sighand->siglock);
  1573. /*
  1574. * Copy seccomp details explicitly here, in case they were changed
  1575. * before holding sighand lock.
  1576. */
  1577. copy_seccomp(p);
  1578. /*
  1579. * Process group and session signals need to be delivered to just the
  1580. * parent before the fork or both the parent and the child after the
  1581. * fork. Restart if a signal comes in before we add the new process to
  1582. * it's process group.
  1583. * A fatal signal pending means that current will exit, so the new
  1584. * thread can't slip out of an OOM kill (or normal SIGKILL).
  1585. */
  1586. recalc_sigpending();
  1587. if (signal_pending(current)) {
  1588. retval = -ERESTARTNOINTR;
  1589. goto bad_fork_cancel_cgroup;
  1590. }
  1591. if (unlikely(!(ns_of_pid(pid)->nr_hashed & PIDNS_HASH_ADDING))) {
  1592. retval = -ENOMEM;
  1593. goto bad_fork_cancel_cgroup;
  1594. }
  1595. if (likely(p->pid)) {
  1596. ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
  1597. init_task_pid(p, PIDTYPE_PID, pid);
  1598. if (thread_group_leader(p)) {
  1599. init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
  1600. init_task_pid(p, PIDTYPE_SID, task_session(current));
  1601. if (is_child_reaper(pid)) {
  1602. ns_of_pid(pid)->child_reaper = p;
  1603. p->signal->flags |= SIGNAL_UNKILLABLE;
  1604. }
  1605. p->signal->leader_pid = pid;
  1606. p->signal->tty = tty_kref_get(current->signal->tty);
  1607. /*
  1608. * Inherit has_child_subreaper flag under the same
  1609. * tasklist_lock with adding child to the process tree
  1610. * for propagate_has_child_subreaper optimization.
  1611. */
  1612. p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper ||
  1613. p->real_parent->signal->is_child_subreaper;
  1614. list_add_tail(&p->sibling, &p->real_parent->children);
  1615. list_add_tail_rcu(&p->tasks, &init_task.tasks);
  1616. attach_pid(p, PIDTYPE_PGID);
  1617. attach_pid(p, PIDTYPE_SID);
  1618. __this_cpu_inc(process_counts);
  1619. } else {
  1620. current->signal->nr_threads++;
  1621. atomic_inc(&current->signal->live);
  1622. atomic_inc(&current->signal->sigcnt);
  1623. list_add_tail_rcu(&p->thread_group,
  1624. &p->group_leader->thread_group);
  1625. list_add_tail_rcu(&p->thread_node,
  1626. &p->signal->thread_head);
  1627. }
  1628. attach_pid(p, PIDTYPE_PID);
  1629. nr_threads++;
  1630. }
  1631. total_forks++;
  1632. spin_unlock(&current->sighand->siglock);
  1633. syscall_tracepoint_update(p);
  1634. write_unlock_irq(&tasklist_lock);
  1635. proc_fork_connector(p);
  1636. cgroup_post_fork(p);
  1637. cgroup_threadgroup_change_end(current);
  1638. perf_event_fork(p);
  1639. trace_task_newtask(p, clone_flags);
  1640. uprobe_copy_process(p, clone_flags);
  1641. return p;
  1642. bad_fork_cancel_cgroup:
  1643. spin_unlock(&current->sighand->siglock);
  1644. write_unlock_irq(&tasklist_lock);
  1645. cgroup_cancel_fork(p);
  1646. bad_fork_free_pid:
  1647. cgroup_threadgroup_change_end(current);
  1648. if (pid != &init_struct_pid)
  1649. free_pid(pid);
  1650. bad_fork_cleanup_thread:
  1651. exit_thread(p);
  1652. bad_fork_cleanup_io:
  1653. if (p->io_context)
  1654. exit_io_context(p);
  1655. bad_fork_cleanup_namespaces:
  1656. exit_task_namespaces(p);
  1657. bad_fork_cleanup_mm:
  1658. if (p->mm)
  1659. mmput(p->mm);
  1660. bad_fork_cleanup_signal:
  1661. if (!(clone_flags & CLONE_THREAD))
  1662. free_signal_struct(p->signal);
  1663. bad_fork_cleanup_sighand:
  1664. __cleanup_sighand(p->sighand);
  1665. bad_fork_cleanup_fs:
  1666. exit_fs(p); /* blocking */
  1667. bad_fork_cleanup_files:
  1668. exit_files(p); /* blocking */
  1669. bad_fork_cleanup_semundo:
  1670. exit_sem(p);
  1671. bad_fork_cleanup_security:
  1672. security_task_free(p);
  1673. bad_fork_cleanup_audit:
  1674. audit_free(p);
  1675. bad_fork_cleanup_perf:
  1676. perf_event_free_task(p);
  1677. bad_fork_cleanup_policy:
  1678. #ifdef CONFIG_NUMA
  1679. mpol_put(p->mempolicy);
  1680. bad_fork_cleanup_threadgroup_lock:
  1681. #endif
  1682. delayacct_tsk_free(p);
  1683. bad_fork_cleanup_count:
  1684. atomic_dec(&p->cred->user->processes);
  1685. exit_creds(p);
  1686. bad_fork_free:
  1687. p->state = TASK_DEAD;
  1688. put_task_stack(p);
  1689. free_task(p);
  1690. fork_out:
  1691. return ERR_PTR(retval);
  1692. }
  1693. static inline void init_idle_pids(struct pid_link *links)
  1694. {
  1695. enum pid_type type;
  1696. for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
  1697. INIT_HLIST_NODE(&links[type].node); /* not really needed */
  1698. links[type].pid = &init_struct_pid;
  1699. }
  1700. }
  1701. struct task_struct *fork_idle(int cpu)
  1702. {
  1703. struct task_struct *task;
  1704. task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0,
  1705. cpu_to_node(cpu));
  1706. if (!IS_ERR(task)) {
  1707. init_idle_pids(task->pids);
  1708. init_idle(task, cpu);
  1709. }
  1710. return task;
  1711. }
  1712. /*
  1713. * Ok, this is the main fork-routine.
  1714. *
  1715. * It copies the process, and if successful kick-starts
  1716. * it and waits for it to finish using the VM if required.
  1717. */
  1718. long _do_fork(unsigned long clone_flags,
  1719. unsigned long stack_start,
  1720. unsigned long stack_size,
  1721. int __user *parent_tidptr,
  1722. int __user *child_tidptr,
  1723. unsigned long tls)
  1724. {
  1725. struct task_struct *p;
  1726. int trace = 0;
  1727. long nr;
  1728. /*
  1729. * Determine whether and which event to report to ptracer. When
  1730. * called from kernel_thread or CLONE_UNTRACED is explicitly
  1731. * requested, no event is reported; otherwise, report if the event
  1732. * for the type of forking is enabled.
  1733. */
  1734. if (!(clone_flags & CLONE_UNTRACED)) {
  1735. if (clone_flags & CLONE_VFORK)
  1736. trace = PTRACE_EVENT_VFORK;
  1737. else if ((clone_flags & CSIGNAL) != SIGCHLD)
  1738. trace = PTRACE_EVENT_CLONE;
  1739. else
  1740. trace = PTRACE_EVENT_FORK;
  1741. if (likely(!ptrace_event_enabled(current, trace)))
  1742. trace = 0;
  1743. }
  1744. p = copy_process(clone_flags, stack_start, stack_size,
  1745. child_tidptr, NULL, trace, tls, NUMA_NO_NODE);
  1746. add_latent_entropy();
  1747. /*
  1748. * Do this prior waking up the new thread - the thread pointer
  1749. * might get invalid after that point, if the thread exits quickly.
  1750. */
  1751. if (!IS_ERR(p)) {
  1752. struct completion vfork;
  1753. struct pid *pid;
  1754. trace_sched_process_fork(current, p);
  1755. pid = get_task_pid(p, PIDTYPE_PID);
  1756. nr = pid_vnr(pid);
  1757. if (clone_flags & CLONE_PARENT_SETTID)
  1758. put_user(nr, parent_tidptr);
  1759. if (clone_flags & CLONE_VFORK) {
  1760. p->vfork_done = &vfork;
  1761. init_completion(&vfork);
  1762. get_task_struct(p);
  1763. }
  1764. wake_up_new_task(p);
  1765. /* forking complete and child started to run, tell ptracer */
  1766. if (unlikely(trace))
  1767. ptrace_event_pid(trace, pid);
  1768. if (clone_flags & CLONE_VFORK) {
  1769. if (!wait_for_vfork_done(p, &vfork))
  1770. ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
  1771. }
  1772. put_pid(pid);
  1773. } else {
  1774. nr = PTR_ERR(p);
  1775. }
  1776. return nr;
  1777. }
  1778. #ifndef CONFIG_HAVE_COPY_THREAD_TLS
  1779. /* For compatibility with architectures that call do_fork directly rather than
  1780. * using the syscall entry points below. */
  1781. long do_fork(unsigned long clone_flags,
  1782. unsigned long stack_start,
  1783. unsigned long stack_size,
  1784. int __user *parent_tidptr,
  1785. int __user *child_tidptr)
  1786. {
  1787. return _do_fork(clone_flags, stack_start, stack_size,
  1788. parent_tidptr, child_tidptr, 0);
  1789. }
  1790. #endif
  1791. /*
  1792. * Create a kernel thread.
  1793. */
  1794. pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
  1795. {
  1796. return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn,
  1797. (unsigned long)arg, NULL, NULL, 0);
  1798. }
  1799. #ifdef __ARCH_WANT_SYS_FORK
  1800. SYSCALL_DEFINE0(fork)
  1801. {
  1802. #ifdef CONFIG_MMU
  1803. return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0);
  1804. #else
  1805. /* can not support in nommu mode */
  1806. return -EINVAL;
  1807. #endif
  1808. }
  1809. #endif
  1810. #ifdef __ARCH_WANT_SYS_VFORK
  1811. SYSCALL_DEFINE0(vfork)
  1812. {
  1813. return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0,
  1814. 0, NULL, NULL, 0);
  1815. }
  1816. #endif
  1817. #ifdef __ARCH_WANT_SYS_CLONE
  1818. #ifdef CONFIG_CLONE_BACKWARDS
  1819. SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
  1820. int __user *, parent_tidptr,
  1821. unsigned long, tls,
  1822. int __user *, child_tidptr)
  1823. #elif defined(CONFIG_CLONE_BACKWARDS2)
  1824. SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
  1825. int __user *, parent_tidptr,
  1826. int __user *, child_tidptr,
  1827. unsigned long, tls)
  1828. #elif defined(CONFIG_CLONE_BACKWARDS3)
  1829. SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
  1830. int, stack_size,
  1831. int __user *, parent_tidptr,
  1832. int __user *, child_tidptr,
  1833. unsigned long, tls)
  1834. #else
  1835. SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
  1836. int __user *, parent_tidptr,
  1837. int __user *, child_tidptr,
  1838. unsigned long, tls)
  1839. #endif
  1840. {
  1841. return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls);
  1842. }
  1843. #endif
  1844. void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data)
  1845. {
  1846. struct task_struct *leader, *parent, *child;
  1847. int res;
  1848. read_lock(&tasklist_lock);
  1849. leader = top = top->group_leader;
  1850. down:
  1851. for_each_thread(leader, parent) {
  1852. list_for_each_entry(child, &parent->children, sibling) {
  1853. res = visitor(child, data);
  1854. if (res) {
  1855. if (res < 0)
  1856. goto out;
  1857. leader = child;
  1858. goto down;
  1859. }
  1860. up:
  1861. ;
  1862. }
  1863. }
  1864. if (leader != top) {
  1865. child = leader;
  1866. parent = child->real_parent;
  1867. leader = parent->group_leader;
  1868. goto up;
  1869. }
  1870. out:
  1871. read_unlock(&tasklist_lock);
  1872. }
  1873. #ifndef ARCH_MIN_MMSTRUCT_ALIGN
  1874. #define ARCH_MIN_MMSTRUCT_ALIGN 0
  1875. #endif
  1876. static void sighand_ctor(void *data)
  1877. {
  1878. struct sighand_struct *sighand = data;
  1879. spin_lock_init(&sighand->siglock);
  1880. init_waitqueue_head(&sighand->signalfd_wqh);
  1881. }
  1882. void __init proc_caches_init(void)
  1883. {
  1884. sighand_cachep = kmem_cache_create("sighand_cache",
  1885. sizeof(struct sighand_struct), 0,
  1886. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU|
  1887. SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor);
  1888. signal_cachep = kmem_cache_create("signal_cache",
  1889. sizeof(struct signal_struct), 0,
  1890. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
  1891. NULL);
  1892. files_cachep = kmem_cache_create("files_cache",
  1893. sizeof(struct files_struct), 0,
  1894. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
  1895. NULL);
  1896. fs_cachep = kmem_cache_create("fs_cache",
  1897. sizeof(struct fs_struct), 0,
  1898. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
  1899. NULL);
  1900. /*
  1901. * FIXME! The "sizeof(struct mm_struct)" currently includes the
  1902. * whole struct cpumask for the OFFSTACK case. We could change
  1903. * this to *only* allocate as much of it as required by the
  1904. * maximum number of CPU's we can ever have. The cpumask_allocation
  1905. * is at the end of the structure, exactly for that reason.
  1906. */
  1907. mm_cachep = kmem_cache_create("mm_struct",
  1908. sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
  1909. SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
  1910. NULL);
  1911. vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
  1912. mmap_init();
  1913. nsproxy_cache_init();
  1914. }
  1915. /*
  1916. * Check constraints on flags passed to the unshare system call.
  1917. */
  1918. static int check_unshare_flags(unsigned long unshare_flags)
  1919. {
  1920. if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
  1921. CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
  1922. CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
  1923. CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
  1924. return -EINVAL;
  1925. /*
  1926. * Not implemented, but pretend it works if there is nothing
  1927. * to unshare. Note that unsharing the address space or the
  1928. * signal handlers also need to unshare the signal queues (aka
  1929. * CLONE_THREAD).
  1930. */
  1931. if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
  1932. if (!thread_group_empty(current))
  1933. return -EINVAL;
  1934. }
  1935. if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
  1936. if (atomic_read(&current->sighand->count) > 1)
  1937. return -EINVAL;
  1938. }
  1939. if (unshare_flags & CLONE_VM) {
  1940. if (!current_is_single_threaded())
  1941. return -EINVAL;
  1942. }
  1943. return 0;
  1944. }
  1945. /*
  1946. * Unshare the filesystem structure if it is being shared
  1947. */
  1948. static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
  1949. {
  1950. struct fs_struct *fs = current->fs;
  1951. if (!(unshare_flags & CLONE_FS) || !fs)
  1952. return 0;
  1953. /* don't need lock here; in the worst case we'll do useless copy */
  1954. if (fs->users == 1)
  1955. return 0;
  1956. *new_fsp = copy_fs_struct(fs);
  1957. if (!*new_fsp)
  1958. return -ENOMEM;
  1959. return 0;
  1960. }
  1961. /*
  1962. * Unshare file descriptor table if it is being shared
  1963. */
  1964. static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
  1965. {
  1966. struct files_struct *fd = current->files;
  1967. int error = 0;
  1968. if ((unshare_flags & CLONE_FILES) &&
  1969. (fd && atomic_read(&fd->count) > 1)) {
  1970. *new_fdp = dup_fd(fd, &error);
  1971. if (!*new_fdp)
  1972. return error;
  1973. }
  1974. return 0;
  1975. }
  1976. /*
  1977. * unshare allows a process to 'unshare' part of the process
  1978. * context which was originally shared using clone. copy_*
  1979. * functions used by do_fork() cannot be used here directly
  1980. * because they modify an inactive task_struct that is being
  1981. * constructed. Here we are modifying the current, active,
  1982. * task_struct.
  1983. */
  1984. SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
  1985. {
  1986. struct fs_struct *fs, *new_fs = NULL;
  1987. struct files_struct *fd, *new_fd = NULL;
  1988. struct cred *new_cred = NULL;
  1989. struct nsproxy *new_nsproxy = NULL;
  1990. int do_sysvsem = 0;
  1991. int err;
  1992. /*
  1993. * If unsharing a user namespace must also unshare the thread group
  1994. * and unshare the filesystem root and working directories.
  1995. */
  1996. if (unshare_flags & CLONE_NEWUSER)
  1997. unshare_flags |= CLONE_THREAD | CLONE_FS;
  1998. /*
  1999. * If unsharing vm, must also unshare signal handlers.
  2000. */
  2001. if (unshare_flags & CLONE_VM)
  2002. unshare_flags |= CLONE_SIGHAND;
  2003. /*
  2004. * If unsharing a signal handlers, must also unshare the signal queues.
  2005. */
  2006. if (unshare_flags & CLONE_SIGHAND)
  2007. unshare_flags |= CLONE_THREAD;
  2008. /*
  2009. * If unsharing namespace, must also unshare filesystem information.
  2010. */
  2011. if (unshare_flags & CLONE_NEWNS)
  2012. unshare_flags |= CLONE_FS;
  2013. err = check_unshare_flags(unshare_flags);
  2014. if (err)
  2015. goto bad_unshare_out;
  2016. /*
  2017. * CLONE_NEWIPC must also detach from the undolist: after switching
  2018. * to a new ipc namespace, the semaphore arrays from the old
  2019. * namespace are unreachable.
  2020. */
  2021. if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
  2022. do_sysvsem = 1;
  2023. err = unshare_fs(unshare_flags, &new_fs);
  2024. if (err)
  2025. goto bad_unshare_out;
  2026. err = unshare_fd(unshare_flags, &new_fd);
  2027. if (err)
  2028. goto bad_unshare_cleanup_fs;
  2029. err = unshare_userns(unshare_flags, &new_cred);
  2030. if (err)
  2031. goto bad_unshare_cleanup_fd;
  2032. err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
  2033. new_cred, new_fs);
  2034. if (err)
  2035. goto bad_unshare_cleanup_cred;
  2036. if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
  2037. if (do_sysvsem) {
  2038. /*
  2039. * CLONE_SYSVSEM is equivalent to sys_exit().
  2040. */
  2041. exit_sem(current);
  2042. }
  2043. if (unshare_flags & CLONE_NEWIPC) {
  2044. /* Orphan segments in old ns (see sem above). */
  2045. exit_shm(current);
  2046. shm_init_task(current);
  2047. }
  2048. if (new_nsproxy)
  2049. switch_task_namespaces(current, new_nsproxy);
  2050. task_lock(current);
  2051. if (new_fs) {
  2052. fs = current->fs;
  2053. spin_lock(&fs->lock);
  2054. current->fs = new_fs;
  2055. if (--fs->users)
  2056. new_fs = NULL;
  2057. else
  2058. new_fs = fs;
  2059. spin_unlock(&fs->lock);
  2060. }
  2061. if (new_fd) {
  2062. fd = current->files;
  2063. current->files = new_fd;
  2064. new_fd = fd;
  2065. }
  2066. task_unlock(current);
  2067. if (new_cred) {
  2068. /* Install the new user namespace */
  2069. commit_creds(new_cred);
  2070. new_cred = NULL;
  2071. }
  2072. }
  2073. perf_event_namespaces(current);
  2074. bad_unshare_cleanup_cred:
  2075. if (new_cred)
  2076. put_cred(new_cred);
  2077. bad_unshare_cleanup_fd:
  2078. if (new_fd)
  2079. put_files_struct(new_fd);
  2080. bad_unshare_cleanup_fs:
  2081. if (new_fs)
  2082. free_fs_struct(new_fs);
  2083. bad_unshare_out:
  2084. return err;
  2085. }
  2086. /*
  2087. * Helper to unshare the files of the current task.
  2088. * We don't want to expose copy_files internals to
  2089. * the exec layer of the kernel.
  2090. */
  2091. int unshare_files(struct files_struct **displaced)
  2092. {
  2093. struct task_struct *task = current;
  2094. struct files_struct *copy = NULL;
  2095. int error;
  2096. error = unshare_fd(CLONE_FILES, &copy);
  2097. if (error || !copy) {
  2098. *displaced = NULL;
  2099. return error;
  2100. }
  2101. *displaced = task->files;
  2102. task_lock(task);
  2103. task->files = copy;
  2104. task_unlock(task);
  2105. return 0;
  2106. }
  2107. int sysctl_max_threads(struct ctl_table *table, int write,
  2108. void __user *buffer, size_t *lenp, loff_t *ppos)
  2109. {
  2110. struct ctl_table t;
  2111. int ret;
  2112. int threads = max_threads;
  2113. int min = MIN_THREADS;
  2114. int max = MAX_THREADS;
  2115. t = *table;
  2116. t.data = &threads;
  2117. t.extra1 = &min;
  2118. t.extra2 = &max;
  2119. ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
  2120. if (ret || !write)
  2121. return ret;
  2122. set_max_threads(threads);
  2123. return 0;
  2124. }