exit.c 42 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654
  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/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/ptrace.h>
  25. #include <linux/profile.h>
  26. #include <linux/mount.h>
  27. #include <linux/proc_fs.h>
  28. #include <linux/mempolicy.h>
  29. #include <linux/taskstats_kern.h>
  30. #include <linux/delayacct.h>
  31. #include <linux/cpuset.h>
  32. #include <linux/syscalls.h>
  33. #include <linux/signal.h>
  34. #include <linux/posix-timers.h>
  35. #include <linux/cn_proc.h>
  36. #include <linux/mutex.h>
  37. #include <linux/futex.h>
  38. #include <linux/compat.h>
  39. #include <linux/pipe_fs_i.h>
  40. #include <linux/audit.h> /* for audit_free() */
  41. #include <linux/resource.h>
  42. #include <linux/blkdev.h>
  43. #include <asm/uaccess.h>
  44. #include <asm/unistd.h>
  45. #include <asm/pgtable.h>
  46. #include <asm/mmu_context.h>
  47. extern void sem_exit (void);
  48. extern struct task_struct *child_reaper;
  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 = -1;
  174. read_lock(&tasklist_lock);
  175. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  176. if (p->signal->session > 0) {
  177. sid = p->signal->session;
  178. goto out;
  179. }
  180. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  181. p = find_task_by_pid(pgrp);
  182. if (p)
  183. sid = p->signal->session;
  184. out:
  185. read_unlock(&tasklist_lock);
  186. return sid;
  187. }
  188. /*
  189. * Determine if a process group is "orphaned", according to the POSIX
  190. * definition in 2.2.2.52. Orphaned process groups are not to be affected
  191. * by terminal-generated stop signals. Newly orphaned process groups are
  192. * to receive a SIGHUP and a SIGCONT.
  193. *
  194. * "I ask you, have you ever known what it is to be an orphan?"
  195. */
  196. static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task)
  197. {
  198. struct task_struct *p;
  199. int ret = 1;
  200. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  201. if (p == ignored_task
  202. || p->exit_state
  203. || is_init(p->real_parent))
  204. continue;
  205. if (process_group(p->real_parent) != pgrp
  206. && p->real_parent->signal->session == p->signal->session) {
  207. ret = 0;
  208. break;
  209. }
  210. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  211. return ret; /* (sighing) "Often!" */
  212. }
  213. int is_orphaned_pgrp(int pgrp)
  214. {
  215. int retval;
  216. read_lock(&tasklist_lock);
  217. retval = will_become_orphaned_pgrp(pgrp, NULL);
  218. read_unlock(&tasklist_lock);
  219. return retval;
  220. }
  221. static int has_stopped_jobs(int pgrp)
  222. {
  223. int retval = 0;
  224. struct task_struct *p;
  225. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  226. if (p->state != TASK_STOPPED)
  227. continue;
  228. retval = 1;
  229. break;
  230. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  231. return retval;
  232. }
  233. /**
  234. * reparent_to_init - Reparent the calling kernel thread to the init task.
  235. *
  236. * If a kernel thread is launched as a result of a system call, or if
  237. * it ever exits, it should generally reparent itself to init so that
  238. * it is correctly cleaned up on exit.
  239. *
  240. * The various task state such as scheduling policy and priority may have
  241. * been inherited from a user process, so we reset them to sane values here.
  242. *
  243. * NOTE that reparent_to_init() gives the caller full capabilities.
  244. */
  245. static void reparent_to_init(void)
  246. {
  247. write_lock_irq(&tasklist_lock);
  248. ptrace_unlink(current);
  249. /* Reparent to init */
  250. remove_parent(current);
  251. current->parent = child_reaper;
  252. current->real_parent = child_reaper;
  253. add_parent(current);
  254. /* Set the exit signal to SIGCHLD so we signal init on exit */
  255. current->exit_signal = SIGCHLD;
  256. if (!has_rt_policy(current) && (task_nice(current) < 0))
  257. set_user_nice(current, 0);
  258. /* cpus_allowed? */
  259. /* rt_priority? */
  260. /* signals? */
  261. security_task_reparent_to_init(current);
  262. memcpy(current->signal->rlim, init_task.signal->rlim,
  263. sizeof(current->signal->rlim));
  264. atomic_inc(&(INIT_USER->__count));
  265. write_unlock_irq(&tasklist_lock);
  266. switch_uid(INIT_USER);
  267. }
  268. void __set_special_pids(pid_t session, pid_t pgrp)
  269. {
  270. struct task_struct *curr = current->group_leader;
  271. if (curr->signal->session != session) {
  272. detach_pid(curr, PIDTYPE_SID);
  273. curr->signal->session = session;
  274. attach_pid(curr, PIDTYPE_SID, session);
  275. }
  276. if (process_group(curr) != pgrp) {
  277. detach_pid(curr, PIDTYPE_PGID);
  278. curr->signal->pgrp = pgrp;
  279. attach_pid(curr, PIDTYPE_PGID, pgrp);
  280. }
  281. }
  282. void set_special_pids(pid_t session, pid_t pgrp)
  283. {
  284. write_lock_irq(&tasklist_lock);
  285. __set_special_pids(session, pgrp);
  286. write_unlock_irq(&tasklist_lock);
  287. }
  288. /*
  289. * Let kernel threads use this to say that they
  290. * allow a certain signal (since daemonize() will
  291. * have disabled all of them by default).
  292. */
  293. int allow_signal(int sig)
  294. {
  295. if (!valid_signal(sig) || sig < 1)
  296. return -EINVAL;
  297. spin_lock_irq(&current->sighand->siglock);
  298. sigdelset(&current->blocked, sig);
  299. if (!current->mm) {
  300. /* Kernel threads handle their own signals.
  301. Let the signal code know it'll be handled, so
  302. that they don't get converted to SIGKILL or
  303. just silently dropped */
  304. current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
  305. }
  306. recalc_sigpending();
  307. spin_unlock_irq(&current->sighand->siglock);
  308. return 0;
  309. }
  310. EXPORT_SYMBOL(allow_signal);
  311. int disallow_signal(int sig)
  312. {
  313. if (!valid_signal(sig) || sig < 1)
  314. return -EINVAL;
  315. spin_lock_irq(&current->sighand->siglock);
  316. sigaddset(&current->blocked, sig);
  317. recalc_sigpending();
  318. spin_unlock_irq(&current->sighand->siglock);
  319. return 0;
  320. }
  321. EXPORT_SYMBOL(disallow_signal);
  322. /*
  323. * Put all the gunge required to become a kernel thread without
  324. * attached user resources in one place where it belongs.
  325. */
  326. void daemonize(const char *name, ...)
  327. {
  328. va_list args;
  329. struct fs_struct *fs;
  330. sigset_t blocked;
  331. va_start(args, name);
  332. vsnprintf(current->comm, sizeof(current->comm), name, args);
  333. va_end(args);
  334. /*
  335. * If we were started as result of loading a module, close all of the
  336. * user space pages. We don't need them, and if we didn't close them
  337. * they would be locked into memory.
  338. */
  339. exit_mm(current);
  340. set_special_pids(1, 1);
  341. mutex_lock(&tty_mutex);
  342. current->signal->tty = NULL;
  343. mutex_unlock(&tty_mutex);
  344. /* Block and flush all signals */
  345. sigfillset(&blocked);
  346. sigprocmask(SIG_BLOCK, &blocked, NULL);
  347. flush_signals(current);
  348. /* Become as one with the init task */
  349. exit_fs(current); /* current->fs->count--; */
  350. fs = init_task.fs;
  351. current->fs = fs;
  352. atomic_inc(&fs->count);
  353. exit_task_namespaces(current);
  354. current->nsproxy = init_task.nsproxy;
  355. get_task_namespaces(current);
  356. exit_files(current);
  357. current->files = init_task.files;
  358. atomic_inc(&current->files->count);
  359. reparent_to_init();
  360. }
  361. EXPORT_SYMBOL(daemonize);
  362. static void close_files(struct files_struct * files)
  363. {
  364. int i, j;
  365. struct fdtable *fdt;
  366. j = 0;
  367. /*
  368. * It is safe to dereference the fd table without RCU or
  369. * ->file_lock because this is the last reference to the
  370. * files structure.
  371. */
  372. fdt = files_fdtable(files);
  373. for (;;) {
  374. unsigned long set;
  375. i = j * __NFDBITS;
  376. if (i >= fdt->max_fdset || i >= fdt->max_fds)
  377. break;
  378. set = fdt->open_fds->fds_bits[j++];
  379. while (set) {
  380. if (set & 1) {
  381. struct file * file = xchg(&fdt->fd[i], NULL);
  382. if (file)
  383. filp_close(file, files);
  384. }
  385. i++;
  386. set >>= 1;
  387. }
  388. }
  389. }
  390. struct files_struct *get_files_struct(struct task_struct *task)
  391. {
  392. struct files_struct *files;
  393. task_lock(task);
  394. files = task->files;
  395. if (files)
  396. atomic_inc(&files->count);
  397. task_unlock(task);
  398. return files;
  399. }
  400. void fastcall put_files_struct(struct files_struct *files)
  401. {
  402. struct fdtable *fdt;
  403. if (atomic_dec_and_test(&files->count)) {
  404. close_files(files);
  405. /*
  406. * Free the fd and fdset arrays if we expanded them.
  407. * If the fdtable was embedded, pass files for freeing
  408. * at the end of the RCU grace period. Otherwise,
  409. * you can free files immediately.
  410. */
  411. fdt = files_fdtable(files);
  412. if (fdt == &files->fdtab)
  413. fdt->free_files = files;
  414. else
  415. kmem_cache_free(files_cachep, files);
  416. free_fdtable(fdt);
  417. }
  418. }
  419. EXPORT_SYMBOL(put_files_struct);
  420. void reset_files_struct(struct task_struct *tsk, struct files_struct *files)
  421. {
  422. struct files_struct *old;
  423. old = tsk->files;
  424. task_lock(tsk);
  425. tsk->files = files;
  426. task_unlock(tsk);
  427. put_files_struct(old);
  428. }
  429. EXPORT_SYMBOL(reset_files_struct);
  430. static inline void __exit_files(struct task_struct *tsk)
  431. {
  432. struct files_struct * files = tsk->files;
  433. if (files) {
  434. task_lock(tsk);
  435. tsk->files = NULL;
  436. task_unlock(tsk);
  437. put_files_struct(files);
  438. }
  439. }
  440. void exit_files(struct task_struct *tsk)
  441. {
  442. __exit_files(tsk);
  443. }
  444. static inline void __put_fs_struct(struct fs_struct *fs)
  445. {
  446. /* No need to hold fs->lock if we are killing it */
  447. if (atomic_dec_and_test(&fs->count)) {
  448. dput(fs->root);
  449. mntput(fs->rootmnt);
  450. dput(fs->pwd);
  451. mntput(fs->pwdmnt);
  452. if (fs->altroot) {
  453. dput(fs->altroot);
  454. mntput(fs->altrootmnt);
  455. }
  456. kmem_cache_free(fs_cachep, fs);
  457. }
  458. }
  459. void put_fs_struct(struct fs_struct *fs)
  460. {
  461. __put_fs_struct(fs);
  462. }
  463. static inline void __exit_fs(struct task_struct *tsk)
  464. {
  465. struct fs_struct * fs = tsk->fs;
  466. if (fs) {
  467. task_lock(tsk);
  468. tsk->fs = NULL;
  469. task_unlock(tsk);
  470. __put_fs_struct(fs);
  471. }
  472. }
  473. void exit_fs(struct task_struct *tsk)
  474. {
  475. __exit_fs(tsk);
  476. }
  477. EXPORT_SYMBOL_GPL(exit_fs);
  478. /*
  479. * Turn us into a lazy TLB process if we
  480. * aren't already..
  481. */
  482. static void exit_mm(struct task_struct * tsk)
  483. {
  484. struct mm_struct *mm = tsk->mm;
  485. mm_release(tsk, mm);
  486. if (!mm)
  487. return;
  488. /*
  489. * Serialize with any possible pending coredump.
  490. * We must hold mmap_sem around checking core_waiters
  491. * and clearing tsk->mm. The core-inducing thread
  492. * will increment core_waiters for each thread in the
  493. * group with ->mm != NULL.
  494. */
  495. down_read(&mm->mmap_sem);
  496. if (mm->core_waiters) {
  497. up_read(&mm->mmap_sem);
  498. down_write(&mm->mmap_sem);
  499. if (!--mm->core_waiters)
  500. complete(mm->core_startup_done);
  501. up_write(&mm->mmap_sem);
  502. wait_for_completion(&mm->core_done);
  503. down_read(&mm->mmap_sem);
  504. }
  505. atomic_inc(&mm->mm_count);
  506. BUG_ON(mm != tsk->active_mm);
  507. /* more a memory barrier than a real lock */
  508. task_lock(tsk);
  509. tsk->mm = NULL;
  510. up_read(&mm->mmap_sem);
  511. enter_lazy_tlb(mm, current);
  512. task_unlock(tsk);
  513. mmput(mm);
  514. }
  515. static inline void
  516. choose_new_parent(struct task_struct *p, struct task_struct *reaper)
  517. {
  518. /*
  519. * Make sure we're not reparenting to ourselves and that
  520. * the parent is not a zombie.
  521. */
  522. BUG_ON(p == reaper || reaper->exit_state);
  523. p->real_parent = reaper;
  524. }
  525. static void
  526. reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
  527. {
  528. /* We don't want people slaying init. */
  529. if (p->exit_signal != -1)
  530. p->exit_signal = SIGCHLD;
  531. if (p->pdeath_signal)
  532. /* We already hold the tasklist_lock here. */
  533. group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
  534. /* Move the child from its dying parent to the new one. */
  535. if (unlikely(traced)) {
  536. /* Preserve ptrace links if someone else is tracing this child. */
  537. list_del_init(&p->ptrace_list);
  538. if (p->parent != p->real_parent)
  539. list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
  540. } else {
  541. /* If this child is being traced, then we're the one tracing it
  542. * anyway, so let go of it.
  543. */
  544. p->ptrace = 0;
  545. remove_parent(p);
  546. p->parent = p->real_parent;
  547. add_parent(p);
  548. /* If we'd notified the old parent about this child's death,
  549. * also notify the new parent.
  550. */
  551. if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
  552. thread_group_empty(p))
  553. do_notify_parent(p, p->exit_signal);
  554. else if (p->state == TASK_TRACED) {
  555. /*
  556. * If it was at a trace stop, turn it into
  557. * a normal stop since it's no longer being
  558. * traced.
  559. */
  560. ptrace_untrace(p);
  561. }
  562. }
  563. /*
  564. * process group orphan check
  565. * Case ii: Our child is in a different pgrp
  566. * than we are, and it was the only connection
  567. * outside, so the child pgrp is now orphaned.
  568. */
  569. if ((process_group(p) != process_group(father)) &&
  570. (p->signal->session == father->signal->session)) {
  571. int pgrp = process_group(p);
  572. if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
  573. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
  574. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
  575. }
  576. }
  577. }
  578. /*
  579. * When we die, we re-parent all our children.
  580. * Try to give them to another thread in our thread
  581. * group, and if no such member exists, give it to
  582. * the global child reaper process (ie "init")
  583. */
  584. static void
  585. forget_original_parent(struct task_struct *father, struct list_head *to_release)
  586. {
  587. struct task_struct *p, *reaper = father;
  588. struct list_head *_p, *_n;
  589. do {
  590. reaper = next_thread(reaper);
  591. if (reaper == father) {
  592. reaper = child_reaper;
  593. break;
  594. }
  595. } while (reaper->exit_state);
  596. /*
  597. * There are only two places where our children can be:
  598. *
  599. * - in our child list
  600. * - in our ptraced child list
  601. *
  602. * Search them and reparent children.
  603. */
  604. list_for_each_safe(_p, _n, &father->children) {
  605. int ptrace;
  606. p = list_entry(_p, struct task_struct, sibling);
  607. ptrace = p->ptrace;
  608. /* if father isn't the real parent, then ptrace must be enabled */
  609. BUG_ON(father != p->real_parent && !ptrace);
  610. if (father == p->real_parent) {
  611. /* reparent with a reaper, real father it's us */
  612. choose_new_parent(p, reaper);
  613. reparent_thread(p, father, 0);
  614. } else {
  615. /* reparent ptraced task to its real parent */
  616. __ptrace_unlink (p);
  617. if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
  618. thread_group_empty(p))
  619. do_notify_parent(p, p->exit_signal);
  620. }
  621. /*
  622. * if the ptraced child is a zombie with exit_signal == -1
  623. * we must collect it before we exit, or it will remain
  624. * zombie forever since we prevented it from self-reap itself
  625. * while it was being traced by us, to be able to see it in wait4.
  626. */
  627. if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
  628. list_add(&p->ptrace_list, to_release);
  629. }
  630. list_for_each_safe(_p, _n, &father->ptrace_children) {
  631. p = list_entry(_p, struct task_struct, ptrace_list);
  632. choose_new_parent(p, reaper);
  633. reparent_thread(p, father, 1);
  634. }
  635. }
  636. /*
  637. * Send signals to all our closest relatives so that they know
  638. * to properly mourn us..
  639. */
  640. static void exit_notify(struct task_struct *tsk)
  641. {
  642. int state;
  643. struct task_struct *t;
  644. struct list_head ptrace_dead, *_p, *_n;
  645. if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
  646. && !thread_group_empty(tsk)) {
  647. /*
  648. * This occurs when there was a race between our exit
  649. * syscall and a group signal choosing us as the one to
  650. * wake up. It could be that we are the only thread
  651. * alerted to check for pending signals, but another thread
  652. * should be woken now to take the signal since we will not.
  653. * Now we'll wake all the threads in the group just to make
  654. * sure someone gets all the pending signals.
  655. */
  656. read_lock(&tasklist_lock);
  657. spin_lock_irq(&tsk->sighand->siglock);
  658. for (t = next_thread(tsk); t != tsk; t = next_thread(t))
  659. if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
  660. recalc_sigpending_tsk(t);
  661. if (signal_pending(t))
  662. signal_wake_up(t, 0);
  663. }
  664. spin_unlock_irq(&tsk->sighand->siglock);
  665. read_unlock(&tasklist_lock);
  666. }
  667. write_lock_irq(&tasklist_lock);
  668. /*
  669. * This does two things:
  670. *
  671. * A. Make init inherit all the child processes
  672. * B. Check to see if any process groups have become orphaned
  673. * as a result of our exiting, and if they have any stopped
  674. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  675. */
  676. INIT_LIST_HEAD(&ptrace_dead);
  677. forget_original_parent(tsk, &ptrace_dead);
  678. BUG_ON(!list_empty(&tsk->children));
  679. BUG_ON(!list_empty(&tsk->ptrace_children));
  680. /*
  681. * Check to see if any process groups have become orphaned
  682. * as a result of our exiting, and if they have any stopped
  683. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  684. *
  685. * Case i: Our father is in a different pgrp than we are
  686. * and we were the only connection outside, so our pgrp
  687. * is about to become orphaned.
  688. */
  689. t = tsk->real_parent;
  690. if ((process_group(t) != process_group(tsk)) &&
  691. (t->signal->session == tsk->signal->session) &&
  692. will_become_orphaned_pgrp(process_group(tsk), tsk) &&
  693. has_stopped_jobs(process_group(tsk))) {
  694. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
  695. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
  696. }
  697. /* Let father know we died
  698. *
  699. * Thread signals are configurable, but you aren't going to use
  700. * that to send signals to arbitary processes.
  701. * That stops right now.
  702. *
  703. * If the parent exec id doesn't match the exec id we saved
  704. * when we started then we know the parent has changed security
  705. * domain.
  706. *
  707. * If our self_exec id doesn't match our parent_exec_id then
  708. * we have changed execution domain as these two values started
  709. * the same after a fork.
  710. *
  711. */
  712. if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
  713. ( tsk->parent_exec_id != t->self_exec_id ||
  714. tsk->self_exec_id != tsk->parent_exec_id)
  715. && !capable(CAP_KILL))
  716. tsk->exit_signal = SIGCHLD;
  717. /* If something other than our normal parent is ptracing us, then
  718. * send it a SIGCHLD instead of honoring exit_signal. exit_signal
  719. * only has special meaning to our real parent.
  720. */
  721. if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
  722. int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
  723. do_notify_parent(tsk, signal);
  724. } else if (tsk->ptrace) {
  725. do_notify_parent(tsk, SIGCHLD);
  726. }
  727. state = EXIT_ZOMBIE;
  728. if (tsk->exit_signal == -1 &&
  729. (likely(tsk->ptrace == 0) ||
  730. unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
  731. state = EXIT_DEAD;
  732. tsk->exit_state = state;
  733. write_unlock_irq(&tasklist_lock);
  734. list_for_each_safe(_p, _n, &ptrace_dead) {
  735. list_del_init(_p);
  736. t = list_entry(_p, struct task_struct, ptrace_list);
  737. release_task(t);
  738. }
  739. /* If the process is dead, release it - nobody will wait for it */
  740. if (state == EXIT_DEAD)
  741. release_task(tsk);
  742. }
  743. fastcall NORET_TYPE void do_exit(long code)
  744. {
  745. struct task_struct *tsk = current;
  746. int group_dead;
  747. profile_task_exit(tsk);
  748. WARN_ON(atomic_read(&tsk->fs_excl));
  749. if (unlikely(in_interrupt()))
  750. panic("Aiee, killing interrupt handler!");
  751. if (unlikely(!tsk->pid))
  752. panic("Attempted to kill the idle task!");
  753. if (unlikely(tsk == child_reaper))
  754. panic("Attempted to kill init!");
  755. if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
  756. current->ptrace_message = code;
  757. ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
  758. }
  759. /*
  760. * We're taking recursive faults here in do_exit. Safest is to just
  761. * leave this task alone and wait for reboot.
  762. */
  763. if (unlikely(tsk->flags & PF_EXITING)) {
  764. printk(KERN_ALERT
  765. "Fixing recursive fault but reboot is needed!\n");
  766. if (tsk->io_context)
  767. exit_io_context();
  768. set_current_state(TASK_UNINTERRUPTIBLE);
  769. schedule();
  770. }
  771. tsk->flags |= PF_EXITING;
  772. if (unlikely(in_atomic()))
  773. printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
  774. current->comm, current->pid,
  775. preempt_count());
  776. acct_update_integrals(tsk);
  777. if (tsk->mm) {
  778. update_hiwater_rss(tsk->mm);
  779. update_hiwater_vm(tsk->mm);
  780. }
  781. group_dead = atomic_dec_and_test(&tsk->signal->live);
  782. if (group_dead) {
  783. hrtimer_cancel(&tsk->signal->real_timer);
  784. exit_itimers(tsk->signal);
  785. }
  786. acct_collect(code, group_dead);
  787. if (unlikely(tsk->robust_list))
  788. exit_robust_list(tsk);
  789. #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
  790. if (unlikely(tsk->compat_robust_list))
  791. compat_exit_robust_list(tsk);
  792. #endif
  793. if (unlikely(tsk->audit_context))
  794. audit_free(tsk);
  795. taskstats_exit(tsk, group_dead);
  796. exit_mm(tsk);
  797. if (group_dead)
  798. acct_process();
  799. exit_sem(tsk);
  800. __exit_files(tsk);
  801. __exit_fs(tsk);
  802. exit_thread();
  803. cpuset_exit(tsk);
  804. exit_keys(tsk);
  805. if (group_dead && tsk->signal->leader)
  806. disassociate_ctty(1);
  807. module_put(task_thread_info(tsk)->exec_domain->module);
  808. if (tsk->binfmt)
  809. module_put(tsk->binfmt->module);
  810. tsk->exit_code = code;
  811. proc_exit_connector(tsk);
  812. exit_notify(tsk);
  813. exit_task_namespaces(tsk);
  814. #ifdef CONFIG_NUMA
  815. mpol_free(tsk->mempolicy);
  816. tsk->mempolicy = NULL;
  817. #endif
  818. /*
  819. * This must happen late, after the PID is not
  820. * hashed anymore:
  821. */
  822. if (unlikely(!list_empty(&tsk->pi_state_list)))
  823. exit_pi_state_list(tsk);
  824. if (unlikely(current->pi_state_cache))
  825. kfree(current->pi_state_cache);
  826. /*
  827. * Make sure we are holding no locks:
  828. */
  829. debug_check_no_locks_held(tsk);
  830. if (tsk->io_context)
  831. exit_io_context();
  832. if (tsk->splice_pipe)
  833. __free_pipe_info(tsk->splice_pipe);
  834. preempt_disable();
  835. /* causes final put_task_struct in finish_task_switch(). */
  836. tsk->state = TASK_DEAD;
  837. schedule();
  838. BUG();
  839. /* Avoid "noreturn function does return". */
  840. for (;;)
  841. cpu_relax(); /* For when BUG is null */
  842. }
  843. EXPORT_SYMBOL_GPL(do_exit);
  844. NORET_TYPE void complete_and_exit(struct completion *comp, long code)
  845. {
  846. if (comp)
  847. complete(comp);
  848. do_exit(code);
  849. }
  850. EXPORT_SYMBOL(complete_and_exit);
  851. asmlinkage long sys_exit(int error_code)
  852. {
  853. do_exit((error_code&0xff)<<8);
  854. }
  855. /*
  856. * Take down every thread in the group. This is called by fatal signals
  857. * as well as by sys_exit_group (below).
  858. */
  859. NORET_TYPE void
  860. do_group_exit(int exit_code)
  861. {
  862. BUG_ON(exit_code & 0x80); /* core dumps don't get here */
  863. if (current->signal->flags & SIGNAL_GROUP_EXIT)
  864. exit_code = current->signal->group_exit_code;
  865. else if (!thread_group_empty(current)) {
  866. struct signal_struct *const sig = current->signal;
  867. struct sighand_struct *const sighand = current->sighand;
  868. spin_lock_irq(&sighand->siglock);
  869. if (sig->flags & SIGNAL_GROUP_EXIT)
  870. /* Another thread got here before we took the lock. */
  871. exit_code = sig->group_exit_code;
  872. else {
  873. sig->group_exit_code = exit_code;
  874. zap_other_threads(current);
  875. }
  876. spin_unlock_irq(&sighand->siglock);
  877. }
  878. do_exit(exit_code);
  879. /* NOTREACHED */
  880. }
  881. /*
  882. * this kills every thread in the thread group. Note that any externally
  883. * wait4()-ing process will get the correct exit code - even if this
  884. * thread is not the thread group leader.
  885. */
  886. asmlinkage void sys_exit_group(int error_code)
  887. {
  888. do_group_exit((error_code & 0xff) << 8);
  889. }
  890. static int eligible_child(pid_t pid, int options, struct task_struct *p)
  891. {
  892. if (pid > 0) {
  893. if (p->pid != pid)
  894. return 0;
  895. } else if (!pid) {
  896. if (process_group(p) != process_group(current))
  897. return 0;
  898. } else if (pid != -1) {
  899. if (process_group(p) != -pid)
  900. return 0;
  901. }
  902. /*
  903. * Do not consider detached threads that are
  904. * not ptraced:
  905. */
  906. if (p->exit_signal == -1 && !p->ptrace)
  907. return 0;
  908. /* Wait for all children (clone and not) if __WALL is set;
  909. * otherwise, wait for clone children *only* if __WCLONE is
  910. * set; otherwise, wait for non-clone children *only*. (Note:
  911. * A "clone" child here is one that reports to its parent
  912. * using a signal other than SIGCHLD.) */
  913. if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
  914. && !(options & __WALL))
  915. return 0;
  916. /*
  917. * Do not consider thread group leaders that are
  918. * in a non-empty thread group:
  919. */
  920. if (delay_group_leader(p))
  921. return 2;
  922. if (security_task_wait(p))
  923. return 0;
  924. return 1;
  925. }
  926. static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
  927. int why, int status,
  928. struct siginfo __user *infop,
  929. struct rusage __user *rusagep)
  930. {
  931. int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
  932. put_task_struct(p);
  933. if (!retval)
  934. retval = put_user(SIGCHLD, &infop->si_signo);
  935. if (!retval)
  936. retval = put_user(0, &infop->si_errno);
  937. if (!retval)
  938. retval = put_user((short)why, &infop->si_code);
  939. if (!retval)
  940. retval = put_user(pid, &infop->si_pid);
  941. if (!retval)
  942. retval = put_user(uid, &infop->si_uid);
  943. if (!retval)
  944. retval = put_user(status, &infop->si_status);
  945. if (!retval)
  946. retval = pid;
  947. return retval;
  948. }
  949. /*
  950. * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
  951. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  952. * the lock and this task is uninteresting. If we return nonzero, we have
  953. * released the lock and the system call should return.
  954. */
  955. static int wait_task_zombie(struct task_struct *p, int noreap,
  956. struct siginfo __user *infop,
  957. int __user *stat_addr, struct rusage __user *ru)
  958. {
  959. unsigned long state;
  960. int retval;
  961. int status;
  962. if (unlikely(noreap)) {
  963. pid_t pid = p->pid;
  964. uid_t uid = p->uid;
  965. int exit_code = p->exit_code;
  966. int why, status;
  967. if (unlikely(p->exit_state != EXIT_ZOMBIE))
  968. return 0;
  969. if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
  970. return 0;
  971. get_task_struct(p);
  972. read_unlock(&tasklist_lock);
  973. if ((exit_code & 0x7f) == 0) {
  974. why = CLD_EXITED;
  975. status = exit_code >> 8;
  976. } else {
  977. why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
  978. status = exit_code & 0x7f;
  979. }
  980. return wait_noreap_copyout(p, pid, uid, why,
  981. status, infop, ru);
  982. }
  983. /*
  984. * Try to move the task's state to DEAD
  985. * only one thread is allowed to do this:
  986. */
  987. state = xchg(&p->exit_state, EXIT_DEAD);
  988. if (state != EXIT_ZOMBIE) {
  989. BUG_ON(state != EXIT_DEAD);
  990. return 0;
  991. }
  992. if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
  993. /*
  994. * This can only happen in a race with a ptraced thread
  995. * dying on another processor.
  996. */
  997. return 0;
  998. }
  999. if (likely(p->real_parent == p->parent) && likely(p->signal)) {
  1000. struct signal_struct *psig;
  1001. struct signal_struct *sig;
  1002. /*
  1003. * The resource counters for the group leader are in its
  1004. * own task_struct. Those for dead threads in the group
  1005. * are in its signal_struct, as are those for the child
  1006. * processes it has previously reaped. All these
  1007. * accumulate in the parent's signal_struct c* fields.
  1008. *
  1009. * We don't bother to take a lock here to protect these
  1010. * p->signal fields, because they are only touched by
  1011. * __exit_signal, which runs with tasklist_lock
  1012. * write-locked anyway, and so is excluded here. We do
  1013. * need to protect the access to p->parent->signal fields,
  1014. * as other threads in the parent group can be right
  1015. * here reaping other children at the same time.
  1016. */
  1017. spin_lock_irq(&p->parent->sighand->siglock);
  1018. psig = p->parent->signal;
  1019. sig = p->signal;
  1020. psig->cutime =
  1021. cputime_add(psig->cutime,
  1022. cputime_add(p->utime,
  1023. cputime_add(sig->utime,
  1024. sig->cutime)));
  1025. psig->cstime =
  1026. cputime_add(psig->cstime,
  1027. cputime_add(p->stime,
  1028. cputime_add(sig->stime,
  1029. sig->cstime)));
  1030. psig->cmin_flt +=
  1031. p->min_flt + sig->min_flt + sig->cmin_flt;
  1032. psig->cmaj_flt +=
  1033. p->maj_flt + sig->maj_flt + sig->cmaj_flt;
  1034. psig->cnvcsw +=
  1035. p->nvcsw + sig->nvcsw + sig->cnvcsw;
  1036. psig->cnivcsw +=
  1037. p->nivcsw + sig->nivcsw + sig->cnivcsw;
  1038. spin_unlock_irq(&p->parent->sighand->siglock);
  1039. }
  1040. /*
  1041. * Now we are sure this task is interesting, and no other
  1042. * thread can reap it because we set its state to EXIT_DEAD.
  1043. */
  1044. read_unlock(&tasklist_lock);
  1045. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1046. status = (p->signal->flags & SIGNAL_GROUP_EXIT)
  1047. ? p->signal->group_exit_code : p->exit_code;
  1048. if (!retval && stat_addr)
  1049. retval = put_user(status, stat_addr);
  1050. if (!retval && infop)
  1051. retval = put_user(SIGCHLD, &infop->si_signo);
  1052. if (!retval && infop)
  1053. retval = put_user(0, &infop->si_errno);
  1054. if (!retval && infop) {
  1055. int why;
  1056. if ((status & 0x7f) == 0) {
  1057. why = CLD_EXITED;
  1058. status >>= 8;
  1059. } else {
  1060. why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
  1061. status &= 0x7f;
  1062. }
  1063. retval = put_user((short)why, &infop->si_code);
  1064. if (!retval)
  1065. retval = put_user(status, &infop->si_status);
  1066. }
  1067. if (!retval && infop)
  1068. retval = put_user(p->pid, &infop->si_pid);
  1069. if (!retval && infop)
  1070. retval = put_user(p->uid, &infop->si_uid);
  1071. if (retval) {
  1072. // TODO: is this safe?
  1073. p->exit_state = EXIT_ZOMBIE;
  1074. return retval;
  1075. }
  1076. retval = p->pid;
  1077. if (p->real_parent != p->parent) {
  1078. write_lock_irq(&tasklist_lock);
  1079. /* Double-check with lock held. */
  1080. if (p->real_parent != p->parent) {
  1081. __ptrace_unlink(p);
  1082. // TODO: is this safe?
  1083. p->exit_state = EXIT_ZOMBIE;
  1084. /*
  1085. * If this is not a detached task, notify the parent.
  1086. * If it's still not detached after that, don't release
  1087. * it now.
  1088. */
  1089. if (p->exit_signal != -1) {
  1090. do_notify_parent(p, p->exit_signal);
  1091. if (p->exit_signal != -1)
  1092. p = NULL;
  1093. }
  1094. }
  1095. write_unlock_irq(&tasklist_lock);
  1096. }
  1097. if (p != NULL)
  1098. release_task(p);
  1099. BUG_ON(!retval);
  1100. return retval;
  1101. }
  1102. /*
  1103. * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
  1104. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1105. * the lock and this task is uninteresting. If we return nonzero, we have
  1106. * released the lock and the system call should return.
  1107. */
  1108. static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
  1109. int noreap, struct siginfo __user *infop,
  1110. int __user *stat_addr, struct rusage __user *ru)
  1111. {
  1112. int retval, exit_code;
  1113. if (!p->exit_code)
  1114. return 0;
  1115. if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
  1116. p->signal && p->signal->group_stop_count > 0)
  1117. /*
  1118. * A group stop is in progress and this is the group leader.
  1119. * We won't report until all threads have stopped.
  1120. */
  1121. return 0;
  1122. /*
  1123. * Now we are pretty sure this task is interesting.
  1124. * Make sure it doesn't get reaped out from under us while we
  1125. * give up the lock and then examine it below. We don't want to
  1126. * keep holding onto the tasklist_lock while we call getrusage and
  1127. * possibly take page faults for user memory.
  1128. */
  1129. get_task_struct(p);
  1130. read_unlock(&tasklist_lock);
  1131. if (unlikely(noreap)) {
  1132. pid_t pid = p->pid;
  1133. uid_t uid = p->uid;
  1134. int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
  1135. exit_code = p->exit_code;
  1136. if (unlikely(!exit_code) ||
  1137. unlikely(p->state & TASK_TRACED))
  1138. goto bail_ref;
  1139. return wait_noreap_copyout(p, pid, uid,
  1140. why, (exit_code << 8) | 0x7f,
  1141. infop, ru);
  1142. }
  1143. write_lock_irq(&tasklist_lock);
  1144. /*
  1145. * This uses xchg to be atomic with the thread resuming and setting
  1146. * it. It must also be done with the write lock held to prevent a
  1147. * race with the EXIT_ZOMBIE case.
  1148. */
  1149. exit_code = xchg(&p->exit_code, 0);
  1150. if (unlikely(p->exit_state)) {
  1151. /*
  1152. * The task resumed and then died. Let the next iteration
  1153. * catch it in EXIT_ZOMBIE. Note that exit_code might
  1154. * already be zero here if it resumed and did _exit(0).
  1155. * The task itself is dead and won't touch exit_code again;
  1156. * other processors in this function are locked out.
  1157. */
  1158. p->exit_code = exit_code;
  1159. exit_code = 0;
  1160. }
  1161. if (unlikely(exit_code == 0)) {
  1162. /*
  1163. * Another thread in this function got to it first, or it
  1164. * resumed, or it resumed and then died.
  1165. */
  1166. write_unlock_irq(&tasklist_lock);
  1167. bail_ref:
  1168. put_task_struct(p);
  1169. /*
  1170. * We are returning to the wait loop without having successfully
  1171. * removed the process and having released the lock. We cannot
  1172. * continue, since the "p" task pointer is potentially stale.
  1173. *
  1174. * Return -EAGAIN, and do_wait() will restart the loop from the
  1175. * beginning. Do _not_ re-acquire the lock.
  1176. */
  1177. return -EAGAIN;
  1178. }
  1179. /* move to end of parent's list to avoid starvation */
  1180. remove_parent(p);
  1181. add_parent(p);
  1182. write_unlock_irq(&tasklist_lock);
  1183. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1184. if (!retval && stat_addr)
  1185. retval = put_user((exit_code << 8) | 0x7f, stat_addr);
  1186. if (!retval && infop)
  1187. retval = put_user(SIGCHLD, &infop->si_signo);
  1188. if (!retval && infop)
  1189. retval = put_user(0, &infop->si_errno);
  1190. if (!retval && infop)
  1191. retval = put_user((short)((p->ptrace & PT_PTRACED)
  1192. ? CLD_TRAPPED : CLD_STOPPED),
  1193. &infop->si_code);
  1194. if (!retval && infop)
  1195. retval = put_user(exit_code, &infop->si_status);
  1196. if (!retval && infop)
  1197. retval = put_user(p->pid, &infop->si_pid);
  1198. if (!retval && infop)
  1199. retval = put_user(p->uid, &infop->si_uid);
  1200. if (!retval)
  1201. retval = p->pid;
  1202. put_task_struct(p);
  1203. BUG_ON(!retval);
  1204. return retval;
  1205. }
  1206. /*
  1207. * Handle do_wait work for one task in a live, non-stopped state.
  1208. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1209. * the lock and this task is uninteresting. If we return nonzero, we have
  1210. * released the lock and the system call should return.
  1211. */
  1212. static int wait_task_continued(struct task_struct *p, int noreap,
  1213. struct siginfo __user *infop,
  1214. int __user *stat_addr, struct rusage __user *ru)
  1215. {
  1216. int retval;
  1217. pid_t pid;
  1218. uid_t uid;
  1219. if (unlikely(!p->signal))
  1220. return 0;
  1221. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
  1222. return 0;
  1223. spin_lock_irq(&p->sighand->siglock);
  1224. /* Re-check with the lock held. */
  1225. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
  1226. spin_unlock_irq(&p->sighand->siglock);
  1227. return 0;
  1228. }
  1229. if (!noreap)
  1230. p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
  1231. spin_unlock_irq(&p->sighand->siglock);
  1232. pid = p->pid;
  1233. uid = p->uid;
  1234. get_task_struct(p);
  1235. read_unlock(&tasklist_lock);
  1236. if (!infop) {
  1237. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1238. put_task_struct(p);
  1239. if (!retval && stat_addr)
  1240. retval = put_user(0xffff, stat_addr);
  1241. if (!retval)
  1242. retval = p->pid;
  1243. } else {
  1244. retval = wait_noreap_copyout(p, pid, uid,
  1245. CLD_CONTINUED, SIGCONT,
  1246. infop, ru);
  1247. BUG_ON(retval == 0);
  1248. }
  1249. return retval;
  1250. }
  1251. static inline int my_ptrace_child(struct task_struct *p)
  1252. {
  1253. if (!(p->ptrace & PT_PTRACED))
  1254. return 0;
  1255. if (!(p->ptrace & PT_ATTACHED))
  1256. return 1;
  1257. /*
  1258. * This child was PTRACE_ATTACH'd. We should be seeing it only if
  1259. * we are the attacher. If we are the real parent, this is a race
  1260. * inside ptrace_attach. It is waiting for the tasklist_lock,
  1261. * which we have to switch the parent links, but has already set
  1262. * the flags in p->ptrace.
  1263. */
  1264. return (p->parent != p->real_parent);
  1265. }
  1266. static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
  1267. int __user *stat_addr, struct rusage __user *ru)
  1268. {
  1269. DECLARE_WAITQUEUE(wait, current);
  1270. struct task_struct *tsk;
  1271. int flag, retval;
  1272. add_wait_queue(&current->signal->wait_chldexit,&wait);
  1273. repeat:
  1274. /*
  1275. * We will set this flag if we see any child that might later
  1276. * match our criteria, even if we are not able to reap it yet.
  1277. */
  1278. flag = 0;
  1279. current->state = TASK_INTERRUPTIBLE;
  1280. read_lock(&tasklist_lock);
  1281. tsk = current;
  1282. do {
  1283. struct task_struct *p;
  1284. struct list_head *_p;
  1285. int ret;
  1286. list_for_each(_p,&tsk->children) {
  1287. p = list_entry(_p, struct task_struct, sibling);
  1288. ret = eligible_child(pid, options, p);
  1289. if (!ret)
  1290. continue;
  1291. switch (p->state) {
  1292. case TASK_TRACED:
  1293. /*
  1294. * When we hit the race with PTRACE_ATTACH,
  1295. * we will not report this child. But the
  1296. * race means it has not yet been moved to
  1297. * our ptrace_children list, so we need to
  1298. * set the flag here to avoid a spurious ECHILD
  1299. * when the race happens with the only child.
  1300. */
  1301. flag = 1;
  1302. if (!my_ptrace_child(p))
  1303. continue;
  1304. /*FALLTHROUGH*/
  1305. case TASK_STOPPED:
  1306. /*
  1307. * It's stopped now, so it might later
  1308. * continue, exit, or stop again.
  1309. */
  1310. flag = 1;
  1311. if (!(options & WUNTRACED) &&
  1312. !my_ptrace_child(p))
  1313. continue;
  1314. retval = wait_task_stopped(p, ret == 2,
  1315. (options & WNOWAIT),
  1316. infop,
  1317. stat_addr, ru);
  1318. if (retval == -EAGAIN)
  1319. goto repeat;
  1320. if (retval != 0) /* He released the lock. */
  1321. goto end;
  1322. break;
  1323. default:
  1324. // case EXIT_DEAD:
  1325. if (p->exit_state == EXIT_DEAD)
  1326. continue;
  1327. // case EXIT_ZOMBIE:
  1328. if (p->exit_state == EXIT_ZOMBIE) {
  1329. /*
  1330. * Eligible but we cannot release
  1331. * it yet:
  1332. */
  1333. if (ret == 2)
  1334. goto check_continued;
  1335. if (!likely(options & WEXITED))
  1336. continue;
  1337. retval = wait_task_zombie(
  1338. p, (options & WNOWAIT),
  1339. infop, stat_addr, ru);
  1340. /* He released the lock. */
  1341. if (retval != 0)
  1342. goto end;
  1343. break;
  1344. }
  1345. check_continued:
  1346. /*
  1347. * It's running now, so it might later
  1348. * exit, stop, or stop and then continue.
  1349. */
  1350. flag = 1;
  1351. if (!unlikely(options & WCONTINUED))
  1352. continue;
  1353. retval = wait_task_continued(
  1354. p, (options & WNOWAIT),
  1355. infop, stat_addr, ru);
  1356. if (retval != 0) /* He released the lock. */
  1357. goto end;
  1358. break;
  1359. }
  1360. }
  1361. if (!flag) {
  1362. list_for_each(_p, &tsk->ptrace_children) {
  1363. p = list_entry(_p, struct task_struct,
  1364. ptrace_list);
  1365. if (!eligible_child(pid, options, p))
  1366. continue;
  1367. flag = 1;
  1368. break;
  1369. }
  1370. }
  1371. if (options & __WNOTHREAD)
  1372. break;
  1373. tsk = next_thread(tsk);
  1374. BUG_ON(tsk->signal != current->signal);
  1375. } while (tsk != current);
  1376. read_unlock(&tasklist_lock);
  1377. if (flag) {
  1378. retval = 0;
  1379. if (options & WNOHANG)
  1380. goto end;
  1381. retval = -ERESTARTSYS;
  1382. if (signal_pending(current))
  1383. goto end;
  1384. schedule();
  1385. goto repeat;
  1386. }
  1387. retval = -ECHILD;
  1388. end:
  1389. current->state = TASK_RUNNING;
  1390. remove_wait_queue(&current->signal->wait_chldexit,&wait);
  1391. if (infop) {
  1392. if (retval > 0)
  1393. retval = 0;
  1394. else {
  1395. /*
  1396. * For a WNOHANG return, clear out all the fields
  1397. * we would set so the user can easily tell the
  1398. * difference.
  1399. */
  1400. if (!retval)
  1401. retval = put_user(0, &infop->si_signo);
  1402. if (!retval)
  1403. retval = put_user(0, &infop->si_errno);
  1404. if (!retval)
  1405. retval = put_user(0, &infop->si_code);
  1406. if (!retval)
  1407. retval = put_user(0, &infop->si_pid);
  1408. if (!retval)
  1409. retval = put_user(0, &infop->si_uid);
  1410. if (!retval)
  1411. retval = put_user(0, &infop->si_status);
  1412. }
  1413. }
  1414. return retval;
  1415. }
  1416. asmlinkage long sys_waitid(int which, pid_t pid,
  1417. struct siginfo __user *infop, int options,
  1418. struct rusage __user *ru)
  1419. {
  1420. long ret;
  1421. if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
  1422. return -EINVAL;
  1423. if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
  1424. return -EINVAL;
  1425. switch (which) {
  1426. case P_ALL:
  1427. pid = -1;
  1428. break;
  1429. case P_PID:
  1430. if (pid <= 0)
  1431. return -EINVAL;
  1432. break;
  1433. case P_PGID:
  1434. if (pid <= 0)
  1435. return -EINVAL;
  1436. pid = -pid;
  1437. break;
  1438. default:
  1439. return -EINVAL;
  1440. }
  1441. ret = do_wait(pid, options, infop, NULL, ru);
  1442. /* avoid REGPARM breakage on x86: */
  1443. prevent_tail_call(ret);
  1444. return ret;
  1445. }
  1446. asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
  1447. int options, struct rusage __user *ru)
  1448. {
  1449. long ret;
  1450. if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
  1451. __WNOTHREAD|__WCLONE|__WALL))
  1452. return -EINVAL;
  1453. ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
  1454. /* avoid REGPARM breakage on x86: */
  1455. prevent_tail_call(ret);
  1456. return ret;
  1457. }
  1458. #ifdef __ARCH_WANT_SYS_WAITPID
  1459. /*
  1460. * sys_waitpid() remains for compatibility. waitpid() should be
  1461. * implemented by calling sys_wait4() from libc.a.
  1462. */
  1463. asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
  1464. {
  1465. return sys_wait4(pid, stat_addr, options, NULL);
  1466. }
  1467. #endif