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