exit.c 42 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662
  1. /*
  2. * linux/kernel/exit.c
  3. *
  4. * Copyright (C) 1991, 1992 Linus Torvalds
  5. */
  6. #include <linux/mm.h>
  7. #include <linux/slab.h>
  8. #include <linux/interrupt.h>
  9. #include <linux/smp_lock.h>
  10. #include <linux/module.h>
  11. #include <linux/capability.h>
  12. #include <linux/completion.h>
  13. #include <linux/personality.h>
  14. #include <linux/tty.h>
  15. #include <linux/mnt_namespace.h>
  16. #include <linux/key.h>
  17. #include <linux/security.h>
  18. #include <linux/cpu.h>
  19. #include <linux/acct.h>
  20. #include <linux/tsacct_kern.h>
  21. #include <linux/file.h>
  22. #include <linux/binfmts.h>
  23. #include <linux/nsproxy.h>
  24. #include <linux/pid_namespace.h>
  25. #include <linux/ptrace.h>
  26. #include <linux/profile.h>
  27. #include <linux/mount.h>
  28. #include <linux/proc_fs.h>
  29. #include <linux/mempolicy.h>
  30. #include <linux/taskstats_kern.h>
  31. #include <linux/delayacct.h>
  32. #include <linux/cpuset.h>
  33. #include <linux/syscalls.h>
  34. #include <linux/signal.h>
  35. #include <linux/posix-timers.h>
  36. #include <linux/cn_proc.h>
  37. #include <linux/mutex.h>
  38. #include <linux/futex.h>
  39. #include <linux/compat.h>
  40. #include <linux/pipe_fs_i.h>
  41. #include <linux/audit.h> /* for audit_free() */
  42. #include <linux/resource.h>
  43. #include <linux/blkdev.h>
  44. #include <asm/uaccess.h>
  45. #include <asm/unistd.h>
  46. #include <asm/pgtable.h>
  47. #include <asm/mmu_context.h>
  48. extern void sem_exit (void);
  49. static void exit_mm(struct task_struct * tsk);
  50. static void __unhash_process(struct task_struct *p)
  51. {
  52. nr_threads--;
  53. detach_pid(p, PIDTYPE_PID);
  54. if (thread_group_leader(p)) {
  55. detach_pid(p, PIDTYPE_PGID);
  56. detach_pid(p, PIDTYPE_SID);
  57. list_del_rcu(&p->tasks);
  58. __get_cpu_var(process_counts)--;
  59. }
  60. list_del_rcu(&p->thread_group);
  61. remove_parent(p);
  62. }
  63. /*
  64. * This function expects the tasklist_lock write-locked.
  65. */
  66. static void __exit_signal(struct task_struct *tsk)
  67. {
  68. struct signal_struct *sig = tsk->signal;
  69. struct sighand_struct *sighand;
  70. BUG_ON(!sig);
  71. BUG_ON(!atomic_read(&sig->count));
  72. rcu_read_lock();
  73. sighand = rcu_dereference(tsk->sighand);
  74. spin_lock(&sighand->siglock);
  75. posix_cpu_timers_exit(tsk);
  76. if (atomic_dec_and_test(&sig->count))
  77. posix_cpu_timers_exit_group(tsk);
  78. else {
  79. /*
  80. * If there is any task waiting for the group exit
  81. * then notify it:
  82. */
  83. if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
  84. wake_up_process(sig->group_exit_task);
  85. sig->group_exit_task = NULL;
  86. }
  87. if (tsk == sig->curr_target)
  88. sig->curr_target = next_thread(tsk);
  89. /*
  90. * Accumulate here the counters for all threads but the
  91. * group leader as they die, so they can be added into
  92. * the process-wide totals when those are taken.
  93. * The group leader stays around as a zombie as long
  94. * as there are other threads. When it gets reaped,
  95. * the exit.c code will add its counts into these totals.
  96. * We won't ever get here for the group leader, since it
  97. * will have been the last reference on the signal_struct.
  98. */
  99. sig->utime = cputime_add(sig->utime, tsk->utime);
  100. sig->stime = cputime_add(sig->stime, tsk->stime);
  101. sig->min_flt += tsk->min_flt;
  102. sig->maj_flt += tsk->maj_flt;
  103. sig->nvcsw += tsk->nvcsw;
  104. sig->nivcsw += tsk->nivcsw;
  105. sig->sched_time += tsk->sched_time;
  106. sig = NULL; /* Marker for below. */
  107. }
  108. __unhash_process(tsk);
  109. tsk->signal = NULL;
  110. tsk->sighand = NULL;
  111. spin_unlock(&sighand->siglock);
  112. rcu_read_unlock();
  113. __cleanup_sighand(sighand);
  114. clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
  115. flush_sigqueue(&tsk->pending);
  116. if (sig) {
  117. flush_sigqueue(&sig->shared_pending);
  118. taskstats_tgid_free(sig);
  119. __cleanup_signal(sig);
  120. }
  121. }
  122. static void delayed_put_task_struct(struct rcu_head *rhp)
  123. {
  124. put_task_struct(container_of(rhp, struct task_struct, rcu));
  125. }
  126. void release_task(struct task_struct * p)
  127. {
  128. struct task_struct *leader;
  129. int zap_leader;
  130. repeat:
  131. atomic_dec(&p->user->processes);
  132. write_lock_irq(&tasklist_lock);
  133. ptrace_unlink(p);
  134. BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
  135. __exit_signal(p);
  136. /*
  137. * If we are the last non-leader member of the thread
  138. * group, and the leader is zombie, then notify the
  139. * group leader's parent process. (if it wants notification.)
  140. */
  141. zap_leader = 0;
  142. leader = p->group_leader;
  143. if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
  144. BUG_ON(leader->exit_signal == -1);
  145. do_notify_parent(leader, leader->exit_signal);
  146. /*
  147. * If we were the last child thread and the leader has
  148. * exited already, and the leader's parent ignores SIGCHLD,
  149. * then we are the one who should release the leader.
  150. *
  151. * do_notify_parent() will have marked it self-reaping in
  152. * that case.
  153. */
  154. zap_leader = (leader->exit_signal == -1);
  155. }
  156. sched_exit(p);
  157. write_unlock_irq(&tasklist_lock);
  158. proc_flush_task(p);
  159. release_thread(p);
  160. call_rcu(&p->rcu, delayed_put_task_struct);
  161. p = leader;
  162. if (unlikely(zap_leader))
  163. goto repeat;
  164. }
  165. /*
  166. * This checks not only the pgrp, but falls back on the pid if no
  167. * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
  168. * without this...
  169. */
  170. int session_of_pgrp(int pgrp)
  171. {
  172. struct task_struct *p;
  173. int sid = 0;
  174. read_lock(&tasklist_lock);
  175. p = find_task_by_pid_type(PIDTYPE_PGID, pgrp);
  176. if (p == NULL)
  177. p = find_task_by_pid(pgrp);
  178. if (p != NULL)
  179. sid = process_session(p);
  180. read_unlock(&tasklist_lock);
  181. return sid;
  182. }
  183. /*
  184. * Determine if a process group is "orphaned", according to the POSIX
  185. * definition in 2.2.2.52. Orphaned process groups are not to be affected
  186. * by terminal-generated stop signals. Newly orphaned process groups are
  187. * to receive a SIGHUP and a SIGCONT.
  188. *
  189. * "I ask you, have you ever known what it is to be an orphan?"
  190. */
  191. static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task)
  192. {
  193. struct task_struct *p;
  194. int ret = 1;
  195. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  196. if (p == ignored_task
  197. || p->exit_state
  198. || is_init(p->real_parent))
  199. continue;
  200. if (process_group(p->real_parent) != pgrp &&
  201. process_session(p->real_parent) == process_session(p)) {
  202. ret = 0;
  203. break;
  204. }
  205. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  206. return ret; /* (sighing) "Often!" */
  207. }
  208. int is_orphaned_pgrp(int pgrp)
  209. {
  210. int retval;
  211. read_lock(&tasklist_lock);
  212. retval = will_become_orphaned_pgrp(pgrp, NULL);
  213. read_unlock(&tasklist_lock);
  214. return retval;
  215. }
  216. static int has_stopped_jobs(int pgrp)
  217. {
  218. int retval = 0;
  219. struct task_struct *p;
  220. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  221. if (p->state != TASK_STOPPED)
  222. continue;
  223. retval = 1;
  224. break;
  225. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  226. return retval;
  227. }
  228. /**
  229. * reparent_to_init - Reparent the calling kernel thread to the init task
  230. * of the pid space that the thread belongs to.
  231. *
  232. * If a kernel thread is launched as a result of a system call, or if
  233. * it ever exits, it should generally reparent itself to init so that
  234. * it is correctly cleaned up on exit.
  235. *
  236. * The various task state such as scheduling policy and priority may have
  237. * been inherited from a user process, so we reset them to sane values here.
  238. *
  239. * NOTE that reparent_to_init() gives the caller full capabilities.
  240. */
  241. static void reparent_to_init(void)
  242. {
  243. write_lock_irq(&tasklist_lock);
  244. ptrace_unlink(current);
  245. /* Reparent to init */
  246. remove_parent(current);
  247. current->parent = child_reaper(current);
  248. current->real_parent = child_reaper(current);
  249. add_parent(current);
  250. /* Set the exit signal to SIGCHLD so we signal init on exit */
  251. current->exit_signal = SIGCHLD;
  252. if (!has_rt_policy(current) && (task_nice(current) < 0))
  253. set_user_nice(current, 0);
  254. /* cpus_allowed? */
  255. /* rt_priority? */
  256. /* signals? */
  257. security_task_reparent_to_init(current);
  258. memcpy(current->signal->rlim, init_task.signal->rlim,
  259. sizeof(current->signal->rlim));
  260. atomic_inc(&(INIT_USER->__count));
  261. write_unlock_irq(&tasklist_lock);
  262. switch_uid(INIT_USER);
  263. }
  264. void __set_special_pids(pid_t session, pid_t pgrp)
  265. {
  266. struct task_struct *curr = current->group_leader;
  267. if (process_session(curr) != session) {
  268. detach_pid(curr, PIDTYPE_SID);
  269. set_signal_session(curr->signal, session);
  270. attach_pid(curr, PIDTYPE_SID, session);
  271. }
  272. if (process_group(curr) != pgrp) {
  273. detach_pid(curr, PIDTYPE_PGID);
  274. curr->signal->pgrp = pgrp;
  275. attach_pid(curr, PIDTYPE_PGID, pgrp);
  276. }
  277. }
  278. static void set_special_pids(pid_t session, pid_t pgrp)
  279. {
  280. write_lock_irq(&tasklist_lock);
  281. __set_special_pids(session, pgrp);
  282. write_unlock_irq(&tasklist_lock);
  283. }
  284. /*
  285. * Let kernel threads use this to say that they
  286. * allow a certain signal (since daemonize() will
  287. * have disabled all of them by default).
  288. */
  289. int allow_signal(int sig)
  290. {
  291. if (!valid_signal(sig) || sig < 1)
  292. return -EINVAL;
  293. spin_lock_irq(&current->sighand->siglock);
  294. sigdelset(&current->blocked, sig);
  295. if (!current->mm) {
  296. /* Kernel threads handle their own signals.
  297. Let the signal code know it'll be handled, so
  298. that they don't get converted to SIGKILL or
  299. just silently dropped */
  300. current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
  301. }
  302. recalc_sigpending();
  303. spin_unlock_irq(&current->sighand->siglock);
  304. return 0;
  305. }
  306. EXPORT_SYMBOL(allow_signal);
  307. int disallow_signal(int sig)
  308. {
  309. if (!valid_signal(sig) || sig < 1)
  310. return -EINVAL;
  311. spin_lock_irq(&current->sighand->siglock);
  312. sigaddset(&current->blocked, sig);
  313. recalc_sigpending();
  314. spin_unlock_irq(&current->sighand->siglock);
  315. return 0;
  316. }
  317. EXPORT_SYMBOL(disallow_signal);
  318. /*
  319. * Put all the gunge required to become a kernel thread without
  320. * attached user resources in one place where it belongs.
  321. */
  322. void daemonize(const char *name, ...)
  323. {
  324. va_list args;
  325. struct fs_struct *fs;
  326. sigset_t blocked;
  327. va_start(args, name);
  328. vsnprintf(current->comm, sizeof(current->comm), name, args);
  329. va_end(args);
  330. /*
  331. * If we were started as result of loading a module, close all of the
  332. * user space pages. We don't need them, and if we didn't close them
  333. * they would be locked into memory.
  334. */
  335. exit_mm(current);
  336. set_special_pids(1, 1);
  337. proc_clear_tty(current);
  338. /* Block and flush all signals */
  339. sigfillset(&blocked);
  340. sigprocmask(SIG_BLOCK, &blocked, NULL);
  341. flush_signals(current);
  342. /* Become as one with the init task */
  343. exit_fs(current); /* current->fs->count--; */
  344. fs = init_task.fs;
  345. current->fs = fs;
  346. atomic_inc(&fs->count);
  347. exit_task_namespaces(current);
  348. current->nsproxy = init_task.nsproxy;
  349. get_task_namespaces(current);
  350. exit_files(current);
  351. current->files = init_task.files;
  352. atomic_inc(&current->files->count);
  353. reparent_to_init();
  354. }
  355. EXPORT_SYMBOL(daemonize);
  356. static void close_files(struct files_struct * files)
  357. {
  358. int i, j;
  359. struct fdtable *fdt;
  360. j = 0;
  361. /*
  362. * It is safe to dereference the fd table without RCU or
  363. * ->file_lock because this is the last reference to the
  364. * files structure.
  365. */
  366. fdt = files_fdtable(files);
  367. for (;;) {
  368. unsigned long set;
  369. i = j * __NFDBITS;
  370. if (i >= fdt->max_fds)
  371. break;
  372. set = fdt->open_fds->fds_bits[j++];
  373. while (set) {
  374. if (set & 1) {
  375. struct file * file = xchg(&fdt->fd[i], NULL);
  376. if (file)
  377. filp_close(file, files);
  378. }
  379. i++;
  380. set >>= 1;
  381. }
  382. }
  383. }
  384. struct files_struct *get_files_struct(struct task_struct *task)
  385. {
  386. struct files_struct *files;
  387. task_lock(task);
  388. files = task->files;
  389. if (files)
  390. atomic_inc(&files->count);
  391. task_unlock(task);
  392. return files;
  393. }
  394. void fastcall put_files_struct(struct files_struct *files)
  395. {
  396. struct fdtable *fdt;
  397. if (atomic_dec_and_test(&files->count)) {
  398. close_files(files);
  399. /*
  400. * Free the fd and fdset arrays if we expanded them.
  401. * If the fdtable was embedded, pass files for freeing
  402. * at the end of the RCU grace period. Otherwise,
  403. * you can free files immediately.
  404. */
  405. fdt = files_fdtable(files);
  406. if (fdt != &files->fdtab)
  407. kmem_cache_free(files_cachep, files);
  408. free_fdtable(fdt);
  409. }
  410. }
  411. EXPORT_SYMBOL(put_files_struct);
  412. void reset_files_struct(struct task_struct *tsk, struct files_struct *files)
  413. {
  414. struct files_struct *old;
  415. old = tsk->files;
  416. task_lock(tsk);
  417. tsk->files = files;
  418. task_unlock(tsk);
  419. put_files_struct(old);
  420. }
  421. EXPORT_SYMBOL(reset_files_struct);
  422. static inline void __exit_files(struct task_struct *tsk)
  423. {
  424. struct files_struct * files = tsk->files;
  425. if (files) {
  426. task_lock(tsk);
  427. tsk->files = NULL;
  428. task_unlock(tsk);
  429. put_files_struct(files);
  430. }
  431. }
  432. void exit_files(struct task_struct *tsk)
  433. {
  434. __exit_files(tsk);
  435. }
  436. static inline void __put_fs_struct(struct fs_struct *fs)
  437. {
  438. /* No need to hold fs->lock if we are killing it */
  439. if (atomic_dec_and_test(&fs->count)) {
  440. dput(fs->root);
  441. mntput(fs->rootmnt);
  442. dput(fs->pwd);
  443. mntput(fs->pwdmnt);
  444. if (fs->altroot) {
  445. dput(fs->altroot);
  446. mntput(fs->altrootmnt);
  447. }
  448. kmem_cache_free(fs_cachep, fs);
  449. }
  450. }
  451. void put_fs_struct(struct fs_struct *fs)
  452. {
  453. __put_fs_struct(fs);
  454. }
  455. static inline void __exit_fs(struct task_struct *tsk)
  456. {
  457. struct fs_struct * fs = tsk->fs;
  458. if (fs) {
  459. task_lock(tsk);
  460. tsk->fs = NULL;
  461. task_unlock(tsk);
  462. __put_fs_struct(fs);
  463. }
  464. }
  465. void exit_fs(struct task_struct *tsk)
  466. {
  467. __exit_fs(tsk);
  468. }
  469. EXPORT_SYMBOL_GPL(exit_fs);
  470. /*
  471. * Turn us into a lazy TLB process if we
  472. * aren't already..
  473. */
  474. static void exit_mm(struct task_struct * tsk)
  475. {
  476. struct mm_struct *mm = tsk->mm;
  477. mm_release(tsk, mm);
  478. if (!mm)
  479. return;
  480. /*
  481. * Serialize with any possible pending coredump.
  482. * We must hold mmap_sem around checking core_waiters
  483. * and clearing tsk->mm. The core-inducing thread
  484. * will increment core_waiters for each thread in the
  485. * group with ->mm != NULL.
  486. */
  487. down_read(&mm->mmap_sem);
  488. if (mm->core_waiters) {
  489. up_read(&mm->mmap_sem);
  490. down_write(&mm->mmap_sem);
  491. if (!--mm->core_waiters)
  492. complete(mm->core_startup_done);
  493. up_write(&mm->mmap_sem);
  494. wait_for_completion(&mm->core_done);
  495. down_read(&mm->mmap_sem);
  496. }
  497. atomic_inc(&mm->mm_count);
  498. BUG_ON(mm != tsk->active_mm);
  499. /* more a memory barrier than a real lock */
  500. task_lock(tsk);
  501. tsk->mm = NULL;
  502. up_read(&mm->mmap_sem);
  503. enter_lazy_tlb(mm, current);
  504. task_unlock(tsk);
  505. mmput(mm);
  506. }
  507. static inline void
  508. choose_new_parent(struct task_struct *p, struct task_struct *reaper)
  509. {
  510. /*
  511. * Make sure we're not reparenting to ourselves and that
  512. * the parent is not a zombie.
  513. */
  514. BUG_ON(p == reaper || reaper->exit_state);
  515. p->real_parent = reaper;
  516. }
  517. static void
  518. reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
  519. {
  520. if (p->pdeath_signal)
  521. /* We already hold the tasklist_lock here. */
  522. group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
  523. /* Move the child from its dying parent to the new one. */
  524. if (unlikely(traced)) {
  525. /* Preserve ptrace links if someone else is tracing this child. */
  526. list_del_init(&p->ptrace_list);
  527. if (p->parent != p->real_parent)
  528. list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
  529. } else {
  530. /* If this child is being traced, then we're the one tracing it
  531. * anyway, so let go of it.
  532. */
  533. p->ptrace = 0;
  534. remove_parent(p);
  535. p->parent = p->real_parent;
  536. add_parent(p);
  537. if (p->state == TASK_TRACED) {
  538. /*
  539. * If it was at a trace stop, turn it into
  540. * a normal stop since it's no longer being
  541. * traced.
  542. */
  543. ptrace_untrace(p);
  544. }
  545. }
  546. /* If this is a threaded reparent there is no need to
  547. * notify anyone anything has happened.
  548. */
  549. if (p->real_parent->group_leader == father->group_leader)
  550. return;
  551. /* We don't want people slaying init. */
  552. if (p->exit_signal != -1)
  553. p->exit_signal = SIGCHLD;
  554. /* If we'd notified the old parent about this child's death,
  555. * also notify the new parent.
  556. */
  557. if (!traced && p->exit_state == EXIT_ZOMBIE &&
  558. p->exit_signal != -1 && thread_group_empty(p))
  559. do_notify_parent(p, p->exit_signal);
  560. /*
  561. * process group orphan check
  562. * Case ii: Our child is in a different pgrp
  563. * than we are, and it was the only connection
  564. * outside, so the child pgrp is now orphaned.
  565. */
  566. if ((process_group(p) != process_group(father)) &&
  567. (process_session(p) == process_session(father))) {
  568. int pgrp = process_group(p);
  569. if (will_become_orphaned_pgrp(pgrp, NULL) &&
  570. has_stopped_jobs(pgrp)) {
  571. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
  572. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
  573. }
  574. }
  575. }
  576. /*
  577. * When we die, we re-parent all our children.
  578. * Try to give them to another thread in our thread
  579. * group, and if no such member exists, give it to
  580. * the child reaper process (ie "init") in our pid
  581. * space.
  582. */
  583. static void
  584. forget_original_parent(struct task_struct *father, struct list_head *to_release)
  585. {
  586. struct task_struct *p, *reaper = father;
  587. struct list_head *_p, *_n;
  588. do {
  589. reaper = next_thread(reaper);
  590. if (reaper == father) {
  591. reaper = child_reaper(father);
  592. break;
  593. }
  594. } while (reaper->exit_state);
  595. /*
  596. * There are only two places where our children can be:
  597. *
  598. * - in our child list
  599. * - in our ptraced child list
  600. *
  601. * Search them and reparent children.
  602. */
  603. list_for_each_safe(_p, _n, &father->children) {
  604. int ptrace;
  605. p = list_entry(_p, struct task_struct, sibling);
  606. ptrace = p->ptrace;
  607. /* if father isn't the real parent, then ptrace must be enabled */
  608. BUG_ON(father != p->real_parent && !ptrace);
  609. if (father == p->real_parent) {
  610. /* reparent with a reaper, real father it's us */
  611. choose_new_parent(p, reaper);
  612. reparent_thread(p, father, 0);
  613. } else {
  614. /* reparent ptraced task to its real parent */
  615. __ptrace_unlink (p);
  616. if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
  617. thread_group_empty(p))
  618. do_notify_parent(p, p->exit_signal);
  619. }
  620. /*
  621. * if the ptraced child is a zombie with exit_signal == -1
  622. * we must collect it before we exit, or it will remain
  623. * zombie forever since we prevented it from self-reap itself
  624. * while it was being traced by us, to be able to see it in wait4.
  625. */
  626. if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
  627. list_add(&p->ptrace_list, to_release);
  628. }
  629. list_for_each_safe(_p, _n, &father->ptrace_children) {
  630. p = list_entry(_p, struct task_struct, ptrace_list);
  631. choose_new_parent(p, reaper);
  632. reparent_thread(p, father, 1);
  633. }
  634. }
  635. /*
  636. * Send signals to all our closest relatives so that they know
  637. * to properly mourn us..
  638. */
  639. static void exit_notify(struct task_struct *tsk)
  640. {
  641. int state;
  642. struct task_struct *t;
  643. struct list_head ptrace_dead, *_p, *_n;
  644. if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
  645. && !thread_group_empty(tsk)) {
  646. /*
  647. * This occurs when there was a race between our exit
  648. * syscall and a group signal choosing us as the one to
  649. * wake up. It could be that we are the only thread
  650. * alerted to check for pending signals, but another thread
  651. * should be woken now to take the signal since we will not.
  652. * Now we'll wake all the threads in the group just to make
  653. * sure someone gets all the pending signals.
  654. */
  655. read_lock(&tasklist_lock);
  656. spin_lock_irq(&tsk->sighand->siglock);
  657. for (t = next_thread(tsk); t != tsk; t = next_thread(t))
  658. if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
  659. recalc_sigpending_tsk(t);
  660. if (signal_pending(t))
  661. signal_wake_up(t, 0);
  662. }
  663. spin_unlock_irq(&tsk->sighand->siglock);
  664. read_unlock(&tasklist_lock);
  665. }
  666. write_lock_irq(&tasklist_lock);
  667. /*
  668. * This does two things:
  669. *
  670. * A. Make init inherit all the child processes
  671. * B. Check to see if any process groups have become orphaned
  672. * as a result of our exiting, and if they have any stopped
  673. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  674. */
  675. INIT_LIST_HEAD(&ptrace_dead);
  676. forget_original_parent(tsk, &ptrace_dead);
  677. BUG_ON(!list_empty(&tsk->children));
  678. BUG_ON(!list_empty(&tsk->ptrace_children));
  679. /*
  680. * Check to see if any process groups have become orphaned
  681. * as a result of our exiting, and if they have any stopped
  682. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  683. *
  684. * Case i: Our father is in a different pgrp than we are
  685. * and we were the only connection outside, so our pgrp
  686. * is about to become orphaned.
  687. */
  688. t = tsk->real_parent;
  689. if ((process_group(t) != process_group(tsk)) &&
  690. (process_session(t) == process_session(tsk)) &&
  691. will_become_orphaned_pgrp(process_group(tsk), tsk) &&
  692. has_stopped_jobs(process_group(tsk))) {
  693. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
  694. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
  695. }
  696. /* Let father know we died
  697. *
  698. * Thread signals are configurable, but you aren't going to use
  699. * that to send signals to arbitary processes.
  700. * That stops right now.
  701. *
  702. * If the parent exec id doesn't match the exec id we saved
  703. * when we started then we know the parent has changed security
  704. * domain.
  705. *
  706. * If our self_exec id doesn't match our parent_exec_id then
  707. * we have changed execution domain as these two values started
  708. * the same after a fork.
  709. *
  710. */
  711. if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
  712. ( tsk->parent_exec_id != t->self_exec_id ||
  713. tsk->self_exec_id != tsk->parent_exec_id)
  714. && !capable(CAP_KILL))
  715. tsk->exit_signal = SIGCHLD;
  716. /* If something other than our normal parent is ptracing us, then
  717. * send it a SIGCHLD instead of honoring exit_signal. exit_signal
  718. * only has special meaning to our real parent.
  719. */
  720. if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
  721. int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
  722. do_notify_parent(tsk, signal);
  723. } else if (tsk->ptrace) {
  724. do_notify_parent(tsk, SIGCHLD);
  725. }
  726. state = EXIT_ZOMBIE;
  727. if (tsk->exit_signal == -1 &&
  728. (likely(tsk->ptrace == 0) ||
  729. unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
  730. state = EXIT_DEAD;
  731. tsk->exit_state = state;
  732. write_unlock_irq(&tasklist_lock);
  733. list_for_each_safe(_p, _n, &ptrace_dead) {
  734. list_del_init(_p);
  735. t = list_entry(_p, struct task_struct, ptrace_list);
  736. release_task(t);
  737. }
  738. /* If the process is dead, release it - nobody will wait for it */
  739. if (state == EXIT_DEAD)
  740. release_task(tsk);
  741. }
  742. fastcall NORET_TYPE void do_exit(long code)
  743. {
  744. struct task_struct *tsk = current;
  745. int group_dead;
  746. profile_task_exit(tsk);
  747. WARN_ON(atomic_read(&tsk->fs_excl));
  748. if (unlikely(in_interrupt()))
  749. panic("Aiee, killing interrupt handler!");
  750. if (unlikely(!tsk->pid))
  751. panic("Attempted to kill the idle task!");
  752. if (unlikely(tsk == child_reaper(tsk))) {
  753. if (tsk->nsproxy->pid_ns != &init_pid_ns)
  754. tsk->nsproxy->pid_ns->child_reaper = init_pid_ns.child_reaper;
  755. else
  756. panic("Attempted to kill init!");
  757. }
  758. if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
  759. current->ptrace_message = code;
  760. ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
  761. }
  762. /*
  763. * We're taking recursive faults here in do_exit. Safest is to just
  764. * leave this task alone and wait for reboot.
  765. */
  766. if (unlikely(tsk->flags & PF_EXITING)) {
  767. printk(KERN_ALERT
  768. "Fixing recursive fault but reboot is needed!\n");
  769. if (tsk->io_context)
  770. exit_io_context();
  771. set_current_state(TASK_UNINTERRUPTIBLE);
  772. schedule();
  773. }
  774. tsk->flags |= PF_EXITING;
  775. if (unlikely(in_atomic()))
  776. printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
  777. current->comm, current->pid,
  778. preempt_count());
  779. acct_update_integrals(tsk);
  780. if (tsk->mm) {
  781. update_hiwater_rss(tsk->mm);
  782. update_hiwater_vm(tsk->mm);
  783. }
  784. group_dead = atomic_dec_and_test(&tsk->signal->live);
  785. if (group_dead) {
  786. hrtimer_cancel(&tsk->signal->real_timer);
  787. exit_itimers(tsk->signal);
  788. }
  789. acct_collect(code, group_dead);
  790. if (unlikely(tsk->robust_list))
  791. exit_robust_list(tsk);
  792. #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
  793. if (unlikely(tsk->compat_robust_list))
  794. compat_exit_robust_list(tsk);
  795. #endif
  796. if (unlikely(tsk->audit_context))
  797. audit_free(tsk);
  798. taskstats_exit(tsk, group_dead);
  799. exit_mm(tsk);
  800. if (group_dead)
  801. acct_process();
  802. exit_sem(tsk);
  803. __exit_files(tsk);
  804. __exit_fs(tsk);
  805. exit_thread();
  806. cpuset_exit(tsk);
  807. exit_keys(tsk);
  808. if (group_dead && tsk->signal->leader)
  809. disassociate_ctty(1);
  810. module_put(task_thread_info(tsk)->exec_domain->module);
  811. if (tsk->binfmt)
  812. module_put(tsk->binfmt->module);
  813. tsk->exit_code = code;
  814. proc_exit_connector(tsk);
  815. exit_notify(tsk);
  816. exit_task_namespaces(tsk);
  817. #ifdef CONFIG_NUMA
  818. mpol_free(tsk->mempolicy);
  819. tsk->mempolicy = NULL;
  820. #endif
  821. /*
  822. * This must happen late, after the PID is not
  823. * hashed anymore:
  824. */
  825. if (unlikely(!list_empty(&tsk->pi_state_list)))
  826. exit_pi_state_list(tsk);
  827. if (unlikely(current->pi_state_cache))
  828. kfree(current->pi_state_cache);
  829. /*
  830. * Make sure we are holding no locks:
  831. */
  832. debug_check_no_locks_held(tsk);
  833. if (tsk->io_context)
  834. exit_io_context();
  835. if (tsk->splice_pipe)
  836. __free_pipe_info(tsk->splice_pipe);
  837. preempt_disable();
  838. /* causes final put_task_struct in finish_task_switch(). */
  839. tsk->state = TASK_DEAD;
  840. schedule();
  841. BUG();
  842. /* Avoid "noreturn function does return". */
  843. for (;;)
  844. cpu_relax(); /* For when BUG is null */
  845. }
  846. EXPORT_SYMBOL_GPL(do_exit);
  847. NORET_TYPE void complete_and_exit(struct completion *comp, long code)
  848. {
  849. if (comp)
  850. complete(comp);
  851. do_exit(code);
  852. }
  853. EXPORT_SYMBOL(complete_and_exit);
  854. asmlinkage long sys_exit(int error_code)
  855. {
  856. do_exit((error_code&0xff)<<8);
  857. }
  858. /*
  859. * Take down every thread in the group. This is called by fatal signals
  860. * as well as by sys_exit_group (below).
  861. */
  862. NORET_TYPE void
  863. do_group_exit(int exit_code)
  864. {
  865. BUG_ON(exit_code & 0x80); /* core dumps don't get here */
  866. if (current->signal->flags & SIGNAL_GROUP_EXIT)
  867. exit_code = current->signal->group_exit_code;
  868. else if (!thread_group_empty(current)) {
  869. struct signal_struct *const sig = current->signal;
  870. struct sighand_struct *const sighand = current->sighand;
  871. spin_lock_irq(&sighand->siglock);
  872. if (sig->flags & SIGNAL_GROUP_EXIT)
  873. /* Another thread got here before we took the lock. */
  874. exit_code = sig->group_exit_code;
  875. else {
  876. sig->group_exit_code = exit_code;
  877. zap_other_threads(current);
  878. }
  879. spin_unlock_irq(&sighand->siglock);
  880. }
  881. do_exit(exit_code);
  882. /* NOTREACHED */
  883. }
  884. /*
  885. * this kills every thread in the thread group. Note that any externally
  886. * wait4()-ing process will get the correct exit code - even if this
  887. * thread is not the thread group leader.
  888. */
  889. asmlinkage void sys_exit_group(int error_code)
  890. {
  891. do_group_exit((error_code & 0xff) << 8);
  892. }
  893. static int eligible_child(pid_t pid, int options, struct task_struct *p)
  894. {
  895. if (pid > 0) {
  896. if (p->pid != pid)
  897. return 0;
  898. } else if (!pid) {
  899. if (process_group(p) != process_group(current))
  900. return 0;
  901. } else if (pid != -1) {
  902. if (process_group(p) != -pid)
  903. return 0;
  904. }
  905. /*
  906. * Do not consider detached threads that are
  907. * not ptraced:
  908. */
  909. if (p->exit_signal == -1 && !p->ptrace)
  910. return 0;
  911. /* Wait for all children (clone and not) if __WALL is set;
  912. * otherwise, wait for clone children *only* if __WCLONE is
  913. * set; otherwise, wait for non-clone children *only*. (Note:
  914. * A "clone" child here is one that reports to its parent
  915. * using a signal other than SIGCHLD.) */
  916. if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
  917. && !(options & __WALL))
  918. return 0;
  919. /*
  920. * Do not consider thread group leaders that are
  921. * in a non-empty thread group:
  922. */
  923. if (delay_group_leader(p))
  924. return 2;
  925. if (security_task_wait(p))
  926. return 0;
  927. return 1;
  928. }
  929. static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
  930. int why, int status,
  931. struct siginfo __user *infop,
  932. struct rusage __user *rusagep)
  933. {
  934. int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
  935. put_task_struct(p);
  936. if (!retval)
  937. retval = put_user(SIGCHLD, &infop->si_signo);
  938. if (!retval)
  939. retval = put_user(0, &infop->si_errno);
  940. if (!retval)
  941. retval = put_user((short)why, &infop->si_code);
  942. if (!retval)
  943. retval = put_user(pid, &infop->si_pid);
  944. if (!retval)
  945. retval = put_user(uid, &infop->si_uid);
  946. if (!retval)
  947. retval = put_user(status, &infop->si_status);
  948. if (!retval)
  949. retval = pid;
  950. return retval;
  951. }
  952. /*
  953. * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
  954. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  955. * the lock and this task is uninteresting. If we return nonzero, we have
  956. * released the lock and the system call should return.
  957. */
  958. static int wait_task_zombie(struct task_struct *p, int noreap,
  959. struct siginfo __user *infop,
  960. int __user *stat_addr, struct rusage __user *ru)
  961. {
  962. unsigned long state;
  963. int retval;
  964. int status;
  965. if (unlikely(noreap)) {
  966. pid_t pid = p->pid;
  967. uid_t uid = p->uid;
  968. int exit_code = p->exit_code;
  969. int why, status;
  970. if (unlikely(p->exit_state != EXIT_ZOMBIE))
  971. return 0;
  972. if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
  973. return 0;
  974. get_task_struct(p);
  975. read_unlock(&tasklist_lock);
  976. if ((exit_code & 0x7f) == 0) {
  977. why = CLD_EXITED;
  978. status = exit_code >> 8;
  979. } else {
  980. why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
  981. status = exit_code & 0x7f;
  982. }
  983. return wait_noreap_copyout(p, pid, uid, why,
  984. status, infop, ru);
  985. }
  986. /*
  987. * Try to move the task's state to DEAD
  988. * only one thread is allowed to do this:
  989. */
  990. state = xchg(&p->exit_state, EXIT_DEAD);
  991. if (state != EXIT_ZOMBIE) {
  992. BUG_ON(state != EXIT_DEAD);
  993. return 0;
  994. }
  995. if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
  996. /*
  997. * This can only happen in a race with a ptraced thread
  998. * dying on another processor.
  999. */
  1000. return 0;
  1001. }
  1002. if (likely(p->real_parent == p->parent) && likely(p->signal)) {
  1003. struct signal_struct *psig;
  1004. struct signal_struct *sig;
  1005. /*
  1006. * The resource counters for the group leader are in its
  1007. * own task_struct. Those for dead threads in the group
  1008. * are in its signal_struct, as are those for the child
  1009. * processes it has previously reaped. All these
  1010. * accumulate in the parent's signal_struct c* fields.
  1011. *
  1012. * We don't bother to take a lock here to protect these
  1013. * p->signal fields, because they are only touched by
  1014. * __exit_signal, which runs with tasklist_lock
  1015. * write-locked anyway, and so is excluded here. We do
  1016. * need to protect the access to p->parent->signal fields,
  1017. * as other threads in the parent group can be right
  1018. * here reaping other children at the same time.
  1019. */
  1020. spin_lock_irq(&p->parent->sighand->siglock);
  1021. psig = p->parent->signal;
  1022. sig = p->signal;
  1023. psig->cutime =
  1024. cputime_add(psig->cutime,
  1025. cputime_add(p->utime,
  1026. cputime_add(sig->utime,
  1027. sig->cutime)));
  1028. psig->cstime =
  1029. cputime_add(psig->cstime,
  1030. cputime_add(p->stime,
  1031. cputime_add(sig->stime,
  1032. sig->cstime)));
  1033. psig->cmin_flt +=
  1034. p->min_flt + sig->min_flt + sig->cmin_flt;
  1035. psig->cmaj_flt +=
  1036. p->maj_flt + sig->maj_flt + sig->cmaj_flt;
  1037. psig->cnvcsw +=
  1038. p->nvcsw + sig->nvcsw + sig->cnvcsw;
  1039. psig->cnivcsw +=
  1040. p->nivcsw + sig->nivcsw + sig->cnivcsw;
  1041. spin_unlock_irq(&p->parent->sighand->siglock);
  1042. }
  1043. /*
  1044. * Now we are sure this task is interesting, and no other
  1045. * thread can reap it because we set its state to EXIT_DEAD.
  1046. */
  1047. read_unlock(&tasklist_lock);
  1048. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1049. status = (p->signal->flags & SIGNAL_GROUP_EXIT)
  1050. ? p->signal->group_exit_code : p->exit_code;
  1051. if (!retval && stat_addr)
  1052. retval = put_user(status, stat_addr);
  1053. if (!retval && infop)
  1054. retval = put_user(SIGCHLD, &infop->si_signo);
  1055. if (!retval && infop)
  1056. retval = put_user(0, &infop->si_errno);
  1057. if (!retval && infop) {
  1058. int why;
  1059. if ((status & 0x7f) == 0) {
  1060. why = CLD_EXITED;
  1061. status >>= 8;
  1062. } else {
  1063. why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
  1064. status &= 0x7f;
  1065. }
  1066. retval = put_user((short)why, &infop->si_code);
  1067. if (!retval)
  1068. retval = put_user(status, &infop->si_status);
  1069. }
  1070. if (!retval && infop)
  1071. retval = put_user(p->pid, &infop->si_pid);
  1072. if (!retval && infop)
  1073. retval = put_user(p->uid, &infop->si_uid);
  1074. if (retval) {
  1075. // TODO: is this safe?
  1076. p->exit_state = EXIT_ZOMBIE;
  1077. return retval;
  1078. }
  1079. retval = p->pid;
  1080. if (p->real_parent != p->parent) {
  1081. write_lock_irq(&tasklist_lock);
  1082. /* Double-check with lock held. */
  1083. if (p->real_parent != p->parent) {
  1084. __ptrace_unlink(p);
  1085. // TODO: is this safe?
  1086. p->exit_state = EXIT_ZOMBIE;
  1087. /*
  1088. * If this is not a detached task, notify the parent.
  1089. * If it's still not detached after that, don't release
  1090. * it now.
  1091. */
  1092. if (p->exit_signal != -1) {
  1093. do_notify_parent(p, p->exit_signal);
  1094. if (p->exit_signal != -1)
  1095. p = NULL;
  1096. }
  1097. }
  1098. write_unlock_irq(&tasklist_lock);
  1099. }
  1100. if (p != NULL)
  1101. release_task(p);
  1102. BUG_ON(!retval);
  1103. return retval;
  1104. }
  1105. /*
  1106. * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
  1107. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1108. * the lock and this task is uninteresting. If we return nonzero, we have
  1109. * released the lock and the system call should return.
  1110. */
  1111. static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
  1112. int noreap, struct siginfo __user *infop,
  1113. int __user *stat_addr, struct rusage __user *ru)
  1114. {
  1115. int retval, exit_code;
  1116. if (!p->exit_code)
  1117. return 0;
  1118. if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
  1119. p->signal && p->signal->group_stop_count > 0)
  1120. /*
  1121. * A group stop is in progress and this is the group leader.
  1122. * We won't report until all threads have stopped.
  1123. */
  1124. return 0;
  1125. /*
  1126. * Now we are pretty sure this task is interesting.
  1127. * Make sure it doesn't get reaped out from under us while we
  1128. * give up the lock and then examine it below. We don't want to
  1129. * keep holding onto the tasklist_lock while we call getrusage and
  1130. * possibly take page faults for user memory.
  1131. */
  1132. get_task_struct(p);
  1133. read_unlock(&tasklist_lock);
  1134. if (unlikely(noreap)) {
  1135. pid_t pid = p->pid;
  1136. uid_t uid = p->uid;
  1137. int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
  1138. exit_code = p->exit_code;
  1139. if (unlikely(!exit_code) ||
  1140. unlikely(p->state & TASK_TRACED))
  1141. goto bail_ref;
  1142. return wait_noreap_copyout(p, pid, uid,
  1143. why, (exit_code << 8) | 0x7f,
  1144. infop, ru);
  1145. }
  1146. write_lock_irq(&tasklist_lock);
  1147. /*
  1148. * This uses xchg to be atomic with the thread resuming and setting
  1149. * it. It must also be done with the write lock held to prevent a
  1150. * race with the EXIT_ZOMBIE case.
  1151. */
  1152. exit_code = xchg(&p->exit_code, 0);
  1153. if (unlikely(p->exit_state)) {
  1154. /*
  1155. * The task resumed and then died. Let the next iteration
  1156. * catch it in EXIT_ZOMBIE. Note that exit_code might
  1157. * already be zero here if it resumed and did _exit(0).
  1158. * The task itself is dead and won't touch exit_code again;
  1159. * other processors in this function are locked out.
  1160. */
  1161. p->exit_code = exit_code;
  1162. exit_code = 0;
  1163. }
  1164. if (unlikely(exit_code == 0)) {
  1165. /*
  1166. * Another thread in this function got to it first, or it
  1167. * resumed, or it resumed and then died.
  1168. */
  1169. write_unlock_irq(&tasklist_lock);
  1170. bail_ref:
  1171. put_task_struct(p);
  1172. /*
  1173. * We are returning to the wait loop without having successfully
  1174. * removed the process and having released the lock. We cannot
  1175. * continue, since the "p" task pointer is potentially stale.
  1176. *
  1177. * Return -EAGAIN, and do_wait() will restart the loop from the
  1178. * beginning. Do _not_ re-acquire the lock.
  1179. */
  1180. return -EAGAIN;
  1181. }
  1182. /* move to end of parent's list to avoid starvation */
  1183. remove_parent(p);
  1184. add_parent(p);
  1185. write_unlock_irq(&tasklist_lock);
  1186. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1187. if (!retval && stat_addr)
  1188. retval = put_user((exit_code << 8) | 0x7f, stat_addr);
  1189. if (!retval && infop)
  1190. retval = put_user(SIGCHLD, &infop->si_signo);
  1191. if (!retval && infop)
  1192. retval = put_user(0, &infop->si_errno);
  1193. if (!retval && infop)
  1194. retval = put_user((short)((p->ptrace & PT_PTRACED)
  1195. ? CLD_TRAPPED : CLD_STOPPED),
  1196. &infop->si_code);
  1197. if (!retval && infop)
  1198. retval = put_user(exit_code, &infop->si_status);
  1199. if (!retval && infop)
  1200. retval = put_user(p->pid, &infop->si_pid);
  1201. if (!retval && infop)
  1202. retval = put_user(p->uid, &infop->si_uid);
  1203. if (!retval)
  1204. retval = p->pid;
  1205. put_task_struct(p);
  1206. BUG_ON(!retval);
  1207. return retval;
  1208. }
  1209. /*
  1210. * Handle do_wait work for one task in a live, non-stopped state.
  1211. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1212. * the lock and this task is uninteresting. If we return nonzero, we have
  1213. * released the lock and the system call should return.
  1214. */
  1215. static int wait_task_continued(struct task_struct *p, int noreap,
  1216. struct siginfo __user *infop,
  1217. int __user *stat_addr, struct rusage __user *ru)
  1218. {
  1219. int retval;
  1220. pid_t pid;
  1221. uid_t uid;
  1222. if (unlikely(!p->signal))
  1223. return 0;
  1224. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
  1225. return 0;
  1226. spin_lock_irq(&p->sighand->siglock);
  1227. /* Re-check with the lock held. */
  1228. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
  1229. spin_unlock_irq(&p->sighand->siglock);
  1230. return 0;
  1231. }
  1232. if (!noreap)
  1233. p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
  1234. spin_unlock_irq(&p->sighand->siglock);
  1235. pid = p->pid;
  1236. uid = p->uid;
  1237. get_task_struct(p);
  1238. read_unlock(&tasklist_lock);
  1239. if (!infop) {
  1240. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1241. put_task_struct(p);
  1242. if (!retval && stat_addr)
  1243. retval = put_user(0xffff, stat_addr);
  1244. if (!retval)
  1245. retval = p->pid;
  1246. } else {
  1247. retval = wait_noreap_copyout(p, pid, uid,
  1248. CLD_CONTINUED, SIGCONT,
  1249. infop, ru);
  1250. BUG_ON(retval == 0);
  1251. }
  1252. return retval;
  1253. }
  1254. static inline int my_ptrace_child(struct task_struct *p)
  1255. {
  1256. if (!(p->ptrace & PT_PTRACED))
  1257. return 0;
  1258. if (!(p->ptrace & PT_ATTACHED))
  1259. return 1;
  1260. /*
  1261. * This child was PTRACE_ATTACH'd. We should be seeing it only if
  1262. * we are the attacher. If we are the real parent, this is a race
  1263. * inside ptrace_attach. It is waiting for the tasklist_lock,
  1264. * which we have to switch the parent links, but has already set
  1265. * the flags in p->ptrace.
  1266. */
  1267. return (p->parent != p->real_parent);
  1268. }
  1269. static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
  1270. int __user *stat_addr, struct rusage __user *ru)
  1271. {
  1272. DECLARE_WAITQUEUE(wait, current);
  1273. struct task_struct *tsk;
  1274. int flag, retval;
  1275. add_wait_queue(&current->signal->wait_chldexit,&wait);
  1276. repeat:
  1277. /*
  1278. * We will set this flag if we see any child that might later
  1279. * match our criteria, even if we are not able to reap it yet.
  1280. */
  1281. flag = 0;
  1282. current->state = TASK_INTERRUPTIBLE;
  1283. read_lock(&tasklist_lock);
  1284. tsk = current;
  1285. do {
  1286. struct task_struct *p;
  1287. struct list_head *_p;
  1288. int ret;
  1289. list_for_each(_p,&tsk->children) {
  1290. p = list_entry(_p, struct task_struct, sibling);
  1291. ret = eligible_child(pid, options, p);
  1292. if (!ret)
  1293. continue;
  1294. switch (p->state) {
  1295. case TASK_TRACED:
  1296. /*
  1297. * When we hit the race with PTRACE_ATTACH,
  1298. * we will not report this child. But the
  1299. * race means it has not yet been moved to
  1300. * our ptrace_children list, so we need to
  1301. * set the flag here to avoid a spurious ECHILD
  1302. * when the race happens with the only child.
  1303. */
  1304. flag = 1;
  1305. if (!my_ptrace_child(p))
  1306. continue;
  1307. /*FALLTHROUGH*/
  1308. case TASK_STOPPED:
  1309. /*
  1310. * It's stopped now, so it might later
  1311. * continue, exit, or stop again.
  1312. */
  1313. flag = 1;
  1314. if (!(options & WUNTRACED) &&
  1315. !my_ptrace_child(p))
  1316. continue;
  1317. retval = wait_task_stopped(p, ret == 2,
  1318. (options & WNOWAIT),
  1319. infop,
  1320. stat_addr, ru);
  1321. if (retval == -EAGAIN)
  1322. goto repeat;
  1323. if (retval != 0) /* He released the lock. */
  1324. goto end;
  1325. break;
  1326. default:
  1327. // case EXIT_DEAD:
  1328. if (p->exit_state == EXIT_DEAD)
  1329. continue;
  1330. // case EXIT_ZOMBIE:
  1331. if (p->exit_state == EXIT_ZOMBIE) {
  1332. /*
  1333. * Eligible but we cannot release
  1334. * it yet:
  1335. */
  1336. if (ret == 2)
  1337. goto check_continued;
  1338. if (!likely(options & WEXITED))
  1339. continue;
  1340. retval = wait_task_zombie(
  1341. p, (options & WNOWAIT),
  1342. infop, stat_addr, ru);
  1343. /* He released the lock. */
  1344. if (retval != 0)
  1345. goto end;
  1346. break;
  1347. }
  1348. check_continued:
  1349. /*
  1350. * It's running now, so it might later
  1351. * exit, stop, or stop and then continue.
  1352. */
  1353. flag = 1;
  1354. if (!unlikely(options & WCONTINUED))
  1355. continue;
  1356. retval = wait_task_continued(
  1357. p, (options & WNOWAIT),
  1358. infop, stat_addr, ru);
  1359. if (retval != 0) /* He released the lock. */
  1360. goto end;
  1361. break;
  1362. }
  1363. }
  1364. if (!flag) {
  1365. list_for_each(_p, &tsk->ptrace_children) {
  1366. p = list_entry(_p, struct task_struct,
  1367. ptrace_list);
  1368. if (!eligible_child(pid, options, p))
  1369. continue;
  1370. flag = 1;
  1371. break;
  1372. }
  1373. }
  1374. if (options & __WNOTHREAD)
  1375. break;
  1376. tsk = next_thread(tsk);
  1377. BUG_ON(tsk->signal != current->signal);
  1378. } while (tsk != current);
  1379. read_unlock(&tasklist_lock);
  1380. if (flag) {
  1381. retval = 0;
  1382. if (options & WNOHANG)
  1383. goto end;
  1384. retval = -ERESTARTSYS;
  1385. if (signal_pending(current))
  1386. goto end;
  1387. schedule();
  1388. goto repeat;
  1389. }
  1390. retval = -ECHILD;
  1391. end:
  1392. current->state = TASK_RUNNING;
  1393. remove_wait_queue(&current->signal->wait_chldexit,&wait);
  1394. if (infop) {
  1395. if (retval > 0)
  1396. retval = 0;
  1397. else {
  1398. /*
  1399. * For a WNOHANG return, clear out all the fields
  1400. * we would set so the user can easily tell the
  1401. * difference.
  1402. */
  1403. if (!retval)
  1404. retval = put_user(0, &infop->si_signo);
  1405. if (!retval)
  1406. retval = put_user(0, &infop->si_errno);
  1407. if (!retval)
  1408. retval = put_user(0, &infop->si_code);
  1409. if (!retval)
  1410. retval = put_user(0, &infop->si_pid);
  1411. if (!retval)
  1412. retval = put_user(0, &infop->si_uid);
  1413. if (!retval)
  1414. retval = put_user(0, &infop->si_status);
  1415. }
  1416. }
  1417. return retval;
  1418. }
  1419. asmlinkage long sys_waitid(int which, pid_t pid,
  1420. struct siginfo __user *infop, int options,
  1421. struct rusage __user *ru)
  1422. {
  1423. long ret;
  1424. if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
  1425. return -EINVAL;
  1426. if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
  1427. return -EINVAL;
  1428. switch (which) {
  1429. case P_ALL:
  1430. pid = -1;
  1431. break;
  1432. case P_PID:
  1433. if (pid <= 0)
  1434. return -EINVAL;
  1435. break;
  1436. case P_PGID:
  1437. if (pid <= 0)
  1438. return -EINVAL;
  1439. pid = -pid;
  1440. break;
  1441. default:
  1442. return -EINVAL;
  1443. }
  1444. ret = do_wait(pid, options, infop, NULL, ru);
  1445. /* avoid REGPARM breakage on x86: */
  1446. prevent_tail_call(ret);
  1447. return ret;
  1448. }
  1449. asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
  1450. int options, struct rusage __user *ru)
  1451. {
  1452. long ret;
  1453. if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
  1454. __WNOTHREAD|__WCLONE|__WALL))
  1455. return -EINVAL;
  1456. ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
  1457. /* avoid REGPARM breakage on x86: */
  1458. prevent_tail_call(ret);
  1459. return ret;
  1460. }
  1461. #ifdef __ARCH_WANT_SYS_WAITPID
  1462. /*
  1463. * sys_waitpid() remains for compatibility. waitpid() should be
  1464. * implemented by calling sys_wait4() from libc.a.
  1465. */
  1466. asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
  1467. {
  1468. return sys_wait4(pid, stat_addr, options, NULL);
  1469. }
  1470. #endif