exit.c 41 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643
  1. /*
  2. * linux/kernel/exit.c
  3. *
  4. * Copyright (C) 1991, 1992 Linus Torvalds
  5. */
  6. #include <linux/config.h>
  7. #include <linux/mm.h>
  8. #include <linux/slab.h>
  9. #include <linux/interrupt.h>
  10. #include <linux/smp_lock.h>
  11. #include <linux/module.h>
  12. #include <linux/capability.h>
  13. #include <linux/completion.h>
  14. #include <linux/personality.h>
  15. #include <linux/tty.h>
  16. #include <linux/namespace.h>
  17. #include <linux/key.h>
  18. #include <linux/security.h>
  19. #include <linux/cpu.h>
  20. #include <linux/acct.h>
  21. #include <linux/file.h>
  22. #include <linux/binfmts.h>
  23. #include <linux/ptrace.h>
  24. #include <linux/profile.h>
  25. #include <linux/mount.h>
  26. #include <linux/proc_fs.h>
  27. #include <linux/mempolicy.h>
  28. #include <linux/cpuset.h>
  29. #include <linux/syscalls.h>
  30. #include <linux/signal.h>
  31. #include <linux/posix-timers.h>
  32. #include <linux/cn_proc.h>
  33. #include <linux/mutex.h>
  34. #include <linux/futex.h>
  35. #include <linux/compat.h>
  36. #include <linux/pipe_fs_i.h>
  37. #include <linux/audit.h> /* for audit_free() */
  38. #include <linux/resource.h>
  39. #include <asm/uaccess.h>
  40. #include <asm/unistd.h>
  41. #include <asm/pgtable.h>
  42. #include <asm/mmu_context.h>
  43. extern void sem_exit (void);
  44. extern struct task_struct *child_reaper;
  45. static void exit_mm(struct task_struct * tsk);
  46. static void __unhash_process(struct task_struct *p)
  47. {
  48. nr_threads--;
  49. detach_pid(p, PIDTYPE_PID);
  50. if (thread_group_leader(p)) {
  51. detach_pid(p, PIDTYPE_PGID);
  52. detach_pid(p, PIDTYPE_SID);
  53. list_del_rcu(&p->tasks);
  54. __get_cpu_var(process_counts)--;
  55. }
  56. list_del_rcu(&p->thread_group);
  57. remove_parent(p);
  58. }
  59. /*
  60. * This function expects the tasklist_lock write-locked.
  61. */
  62. static void __exit_signal(struct task_struct *tsk)
  63. {
  64. struct signal_struct *sig = tsk->signal;
  65. struct sighand_struct *sighand;
  66. BUG_ON(!sig);
  67. BUG_ON(!atomic_read(&sig->count));
  68. rcu_read_lock();
  69. sighand = rcu_dereference(tsk->sighand);
  70. spin_lock(&sighand->siglock);
  71. posix_cpu_timers_exit(tsk);
  72. if (atomic_dec_and_test(&sig->count))
  73. posix_cpu_timers_exit_group(tsk);
  74. else {
  75. /*
  76. * If there is any task waiting for the group exit
  77. * then notify it:
  78. */
  79. if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
  80. wake_up_process(sig->group_exit_task);
  81. sig->group_exit_task = NULL;
  82. }
  83. if (tsk == sig->curr_target)
  84. sig->curr_target = next_thread(tsk);
  85. /*
  86. * Accumulate here the counters for all threads but the
  87. * group leader as they die, so they can be added into
  88. * the process-wide totals when those are taken.
  89. * The group leader stays around as a zombie as long
  90. * as there are other threads. When it gets reaped,
  91. * the exit.c code will add its counts into these totals.
  92. * We won't ever get here for the group leader, since it
  93. * will have been the last reference on the signal_struct.
  94. */
  95. sig->utime = cputime_add(sig->utime, tsk->utime);
  96. sig->stime = cputime_add(sig->stime, tsk->stime);
  97. sig->min_flt += tsk->min_flt;
  98. sig->maj_flt += tsk->maj_flt;
  99. sig->nvcsw += tsk->nvcsw;
  100. sig->nivcsw += tsk->nivcsw;
  101. sig->sched_time += tsk->sched_time;
  102. sig = NULL; /* Marker for below. */
  103. }
  104. __unhash_process(tsk);
  105. tsk->signal = NULL;
  106. tsk->sighand = NULL;
  107. spin_unlock(&sighand->siglock);
  108. rcu_read_unlock();
  109. __cleanup_sighand(sighand);
  110. clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
  111. flush_sigqueue(&tsk->pending);
  112. if (sig) {
  113. flush_sigqueue(&sig->shared_pending);
  114. __cleanup_signal(sig);
  115. }
  116. }
  117. static void delayed_put_task_struct(struct rcu_head *rhp)
  118. {
  119. put_task_struct(container_of(rhp, struct task_struct, rcu));
  120. }
  121. void release_task(struct task_struct * p)
  122. {
  123. int zap_leader;
  124. task_t *leader;
  125. repeat:
  126. atomic_dec(&p->user->processes);
  127. write_lock_irq(&tasklist_lock);
  128. ptrace_unlink(p);
  129. BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
  130. __exit_signal(p);
  131. /*
  132. * If we are the last non-leader member of the thread
  133. * group, and the leader is zombie, then notify the
  134. * group leader's parent process. (if it wants notification.)
  135. */
  136. zap_leader = 0;
  137. leader = p->group_leader;
  138. if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
  139. BUG_ON(leader->exit_signal == -1);
  140. do_notify_parent(leader, leader->exit_signal);
  141. /*
  142. * If we were the last child thread and the leader has
  143. * exited already, and the leader's parent ignores SIGCHLD,
  144. * then we are the one who should release the leader.
  145. *
  146. * do_notify_parent() will have marked it self-reaping in
  147. * that case.
  148. */
  149. zap_leader = (leader->exit_signal == -1);
  150. }
  151. sched_exit(p);
  152. write_unlock_irq(&tasklist_lock);
  153. proc_flush_task(p);
  154. release_thread(p);
  155. call_rcu(&p->rcu, delayed_put_task_struct);
  156. p = leader;
  157. if (unlikely(zap_leader))
  158. goto repeat;
  159. }
  160. /*
  161. * This checks not only the pgrp, but falls back on the pid if no
  162. * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
  163. * without this...
  164. */
  165. int session_of_pgrp(int pgrp)
  166. {
  167. struct task_struct *p;
  168. int sid = -1;
  169. read_lock(&tasklist_lock);
  170. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  171. if (p->signal->session > 0) {
  172. sid = p->signal->session;
  173. goto out;
  174. }
  175. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  176. p = find_task_by_pid(pgrp);
  177. if (p)
  178. sid = p->signal->session;
  179. out:
  180. read_unlock(&tasklist_lock);
  181. return sid;
  182. }
  183. /*
  184. * Determine if a process group is "orphaned", according to the POSIX
  185. * definition in 2.2.2.52. Orphaned process groups are not to be affected
  186. * by terminal-generated stop signals. Newly orphaned process groups are
  187. * to receive a SIGHUP and a SIGCONT.
  188. *
  189. * "I ask you, have you ever known what it is to be an orphan?"
  190. */
  191. static int will_become_orphaned_pgrp(int pgrp, task_t *ignored_task)
  192. {
  193. struct task_struct *p;
  194. int ret = 1;
  195. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  196. if (p == ignored_task
  197. || p->exit_state
  198. || p->real_parent->pid == 1)
  199. continue;
  200. if (process_group(p->real_parent) != pgrp
  201. && p->real_parent->signal->session == p->signal->session) {
  202. ret = 0;
  203. break;
  204. }
  205. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  206. return ret; /* (sighing) "Often!" */
  207. }
  208. int is_orphaned_pgrp(int pgrp)
  209. {
  210. int retval;
  211. read_lock(&tasklist_lock);
  212. retval = will_become_orphaned_pgrp(pgrp, NULL);
  213. read_unlock(&tasklist_lock);
  214. return retval;
  215. }
  216. static int has_stopped_jobs(int pgrp)
  217. {
  218. int retval = 0;
  219. struct task_struct *p;
  220. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  221. if (p->state != TASK_STOPPED)
  222. continue;
  223. /* If p is stopped by a debugger on a signal that won't
  224. stop it, then don't count p as stopped. This isn't
  225. perfect but it's a good approximation. */
  226. if (unlikely (p->ptrace)
  227. && p->exit_code != SIGSTOP
  228. && p->exit_code != SIGTSTP
  229. && p->exit_code != SIGTTOU
  230. && p->exit_code != SIGTTIN)
  231. continue;
  232. retval = 1;
  233. break;
  234. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  235. return retval;
  236. }
  237. /**
  238. * reparent_to_init - Reparent the calling kernel thread to the init task.
  239. *
  240. * If a kernel thread is launched as a result of a system call, or if
  241. * it ever exits, it should generally reparent itself to init so that
  242. * it is correctly cleaned up on exit.
  243. *
  244. * The various task state such as scheduling policy and priority may have
  245. * been inherited from a user process, so we reset them to sane values here.
  246. *
  247. * NOTE that reparent_to_init() gives the caller full capabilities.
  248. */
  249. static void reparent_to_init(void)
  250. {
  251. write_lock_irq(&tasklist_lock);
  252. ptrace_unlink(current);
  253. /* Reparent to init */
  254. remove_parent(current);
  255. current->parent = child_reaper;
  256. current->real_parent = child_reaper;
  257. add_parent(current);
  258. /* Set the exit signal to SIGCHLD so we signal init on exit */
  259. current->exit_signal = SIGCHLD;
  260. if ((current->policy == SCHED_NORMAL ||
  261. current->policy == SCHED_BATCH)
  262. && (task_nice(current) < 0))
  263. set_user_nice(current, 0);
  264. /* cpus_allowed? */
  265. /* rt_priority? */
  266. /* signals? */
  267. security_task_reparent_to_init(current);
  268. memcpy(current->signal->rlim, init_task.signal->rlim,
  269. sizeof(current->signal->rlim));
  270. atomic_inc(&(INIT_USER->__count));
  271. write_unlock_irq(&tasklist_lock);
  272. switch_uid(INIT_USER);
  273. }
  274. void __set_special_pids(pid_t session, pid_t pgrp)
  275. {
  276. struct task_struct *curr = current->group_leader;
  277. if (curr->signal->session != session) {
  278. detach_pid(curr, PIDTYPE_SID);
  279. curr->signal->session = session;
  280. attach_pid(curr, PIDTYPE_SID, session);
  281. }
  282. if (process_group(curr) != pgrp) {
  283. detach_pid(curr, PIDTYPE_PGID);
  284. curr->signal->pgrp = pgrp;
  285. attach_pid(curr, PIDTYPE_PGID, pgrp);
  286. }
  287. }
  288. void set_special_pids(pid_t session, pid_t pgrp)
  289. {
  290. write_lock_irq(&tasklist_lock);
  291. __set_special_pids(session, pgrp);
  292. write_unlock_irq(&tasklist_lock);
  293. }
  294. /*
  295. * Let kernel threads use this to say that they
  296. * allow a certain signal (since daemonize() will
  297. * have disabled all of them by default).
  298. */
  299. int allow_signal(int sig)
  300. {
  301. if (!valid_signal(sig) || sig < 1)
  302. return -EINVAL;
  303. spin_lock_irq(&current->sighand->siglock);
  304. sigdelset(&current->blocked, sig);
  305. if (!current->mm) {
  306. /* Kernel threads handle their own signals.
  307. Let the signal code know it'll be handled, so
  308. that they don't get converted to SIGKILL or
  309. just silently dropped */
  310. current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
  311. }
  312. recalc_sigpending();
  313. spin_unlock_irq(&current->sighand->siglock);
  314. return 0;
  315. }
  316. EXPORT_SYMBOL(allow_signal);
  317. int disallow_signal(int sig)
  318. {
  319. if (!valid_signal(sig) || sig < 1)
  320. return -EINVAL;
  321. spin_lock_irq(&current->sighand->siglock);
  322. sigaddset(&current->blocked, sig);
  323. recalc_sigpending();
  324. spin_unlock_irq(&current->sighand->siglock);
  325. return 0;
  326. }
  327. EXPORT_SYMBOL(disallow_signal);
  328. /*
  329. * Put all the gunge required to become a kernel thread without
  330. * attached user resources in one place where it belongs.
  331. */
  332. void daemonize(const char *name, ...)
  333. {
  334. va_list args;
  335. struct fs_struct *fs;
  336. sigset_t blocked;
  337. va_start(args, name);
  338. vsnprintf(current->comm, sizeof(current->comm), name, args);
  339. va_end(args);
  340. /*
  341. * If we were started as result of loading a module, close all of the
  342. * user space pages. We don't need them, and if we didn't close them
  343. * they would be locked into memory.
  344. */
  345. exit_mm(current);
  346. set_special_pids(1, 1);
  347. mutex_lock(&tty_mutex);
  348. current->signal->tty = NULL;
  349. mutex_unlock(&tty_mutex);
  350. /* Block and flush all signals */
  351. sigfillset(&blocked);
  352. sigprocmask(SIG_BLOCK, &blocked, NULL);
  353. flush_signals(current);
  354. /* Become as one with the init task */
  355. exit_fs(current); /* current->fs->count--; */
  356. fs = init_task.fs;
  357. current->fs = fs;
  358. atomic_inc(&fs->count);
  359. exit_namespace(current);
  360. current->namespace = init_task.namespace;
  361. get_namespace(current->namespace);
  362. exit_files(current);
  363. current->files = init_task.files;
  364. atomic_inc(&current->files->count);
  365. reparent_to_init();
  366. }
  367. EXPORT_SYMBOL(daemonize);
  368. static void close_files(struct files_struct * files)
  369. {
  370. int i, j;
  371. struct fdtable *fdt;
  372. j = 0;
  373. /*
  374. * It is safe to dereference the fd table without RCU or
  375. * ->file_lock because this is the last reference to the
  376. * files structure.
  377. */
  378. fdt = files_fdtable(files);
  379. for (;;) {
  380. unsigned long set;
  381. i = j * __NFDBITS;
  382. if (i >= fdt->max_fdset || i >= fdt->max_fds)
  383. break;
  384. set = fdt->open_fds->fds_bits[j++];
  385. while (set) {
  386. if (set & 1) {
  387. struct file * file = xchg(&fdt->fd[i], NULL);
  388. if (file)
  389. filp_close(file, files);
  390. }
  391. i++;
  392. set >>= 1;
  393. }
  394. }
  395. }
  396. struct files_struct *get_files_struct(struct task_struct *task)
  397. {
  398. struct files_struct *files;
  399. task_lock(task);
  400. files = task->files;
  401. if (files)
  402. atomic_inc(&files->count);
  403. task_unlock(task);
  404. return files;
  405. }
  406. void fastcall put_files_struct(struct files_struct *files)
  407. {
  408. struct fdtable *fdt;
  409. if (atomic_dec_and_test(&files->count)) {
  410. close_files(files);
  411. /*
  412. * Free the fd and fdset arrays if we expanded them.
  413. * If the fdtable was embedded, pass files for freeing
  414. * at the end of the RCU grace period. Otherwise,
  415. * you can free files immediately.
  416. */
  417. fdt = files_fdtable(files);
  418. if (fdt == &files->fdtab)
  419. fdt->free_files = files;
  420. else
  421. kmem_cache_free(files_cachep, files);
  422. free_fdtable(fdt);
  423. }
  424. }
  425. EXPORT_SYMBOL(put_files_struct);
  426. static inline void __exit_files(struct task_struct *tsk)
  427. {
  428. struct files_struct * files = tsk->files;
  429. if (files) {
  430. task_lock(tsk);
  431. tsk->files = NULL;
  432. task_unlock(tsk);
  433. put_files_struct(files);
  434. }
  435. }
  436. void exit_files(struct task_struct *tsk)
  437. {
  438. __exit_files(tsk);
  439. }
  440. static inline void __put_fs_struct(struct fs_struct *fs)
  441. {
  442. /* No need to hold fs->lock if we are killing it */
  443. if (atomic_dec_and_test(&fs->count)) {
  444. dput(fs->root);
  445. mntput(fs->rootmnt);
  446. dput(fs->pwd);
  447. mntput(fs->pwdmnt);
  448. if (fs->altroot) {
  449. dput(fs->altroot);
  450. mntput(fs->altrootmnt);
  451. }
  452. kmem_cache_free(fs_cachep, fs);
  453. }
  454. }
  455. void put_fs_struct(struct fs_struct *fs)
  456. {
  457. __put_fs_struct(fs);
  458. }
  459. static inline void __exit_fs(struct task_struct *tsk)
  460. {
  461. struct fs_struct * fs = tsk->fs;
  462. if (fs) {
  463. task_lock(tsk);
  464. tsk->fs = NULL;
  465. task_unlock(tsk);
  466. __put_fs_struct(fs);
  467. }
  468. }
  469. void exit_fs(struct task_struct *tsk)
  470. {
  471. __exit_fs(tsk);
  472. }
  473. EXPORT_SYMBOL_GPL(exit_fs);
  474. /*
  475. * Turn us into a lazy TLB process if we
  476. * aren't already..
  477. */
  478. static void exit_mm(struct task_struct * tsk)
  479. {
  480. struct mm_struct *mm = tsk->mm;
  481. mm_release(tsk, mm);
  482. if (!mm)
  483. return;
  484. /*
  485. * Serialize with any possible pending coredump.
  486. * We must hold mmap_sem around checking core_waiters
  487. * and clearing tsk->mm. The core-inducing thread
  488. * will increment core_waiters for each thread in the
  489. * group with ->mm != NULL.
  490. */
  491. down_read(&mm->mmap_sem);
  492. if (mm->core_waiters) {
  493. up_read(&mm->mmap_sem);
  494. down_write(&mm->mmap_sem);
  495. if (!--mm->core_waiters)
  496. complete(mm->core_startup_done);
  497. up_write(&mm->mmap_sem);
  498. wait_for_completion(&mm->core_done);
  499. down_read(&mm->mmap_sem);
  500. }
  501. atomic_inc(&mm->mm_count);
  502. BUG_ON(mm != tsk->active_mm);
  503. /* more a memory barrier than a real lock */
  504. task_lock(tsk);
  505. tsk->mm = NULL;
  506. up_read(&mm->mmap_sem);
  507. enter_lazy_tlb(mm, current);
  508. task_unlock(tsk);
  509. mmput(mm);
  510. }
  511. static inline void choose_new_parent(task_t *p, task_t *reaper)
  512. {
  513. /*
  514. * Make sure we're not reparenting to ourselves and that
  515. * the parent is not a zombie.
  516. */
  517. BUG_ON(p == reaper || reaper->exit_state);
  518. p->real_parent = reaper;
  519. }
  520. static void reparent_thread(task_t *p, task_t *father, int traced)
  521. {
  522. /* We don't want people slaying init. */
  523. if (p->exit_signal != -1)
  524. p->exit_signal = SIGCHLD;
  525. if (p->pdeath_signal)
  526. /* We already hold the tasklist_lock here. */
  527. group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
  528. /* Move the child from its dying parent to the new one. */
  529. if (unlikely(traced)) {
  530. /* Preserve ptrace links if someone else is tracing this child. */
  531. list_del_init(&p->ptrace_list);
  532. if (p->parent != p->real_parent)
  533. list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
  534. } else {
  535. /* If this child is being traced, then we're the one tracing it
  536. * anyway, so let go of it.
  537. */
  538. p->ptrace = 0;
  539. remove_parent(p);
  540. p->parent = p->real_parent;
  541. add_parent(p);
  542. /* If we'd notified the old parent about this child's death,
  543. * also notify the new parent.
  544. */
  545. if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
  546. thread_group_empty(p))
  547. do_notify_parent(p, p->exit_signal);
  548. else if (p->state == TASK_TRACED) {
  549. /*
  550. * If it was at a trace stop, turn it into
  551. * a normal stop since it's no longer being
  552. * traced.
  553. */
  554. ptrace_untrace(p);
  555. }
  556. }
  557. /*
  558. * process group orphan check
  559. * Case ii: Our child is in a different pgrp
  560. * than we are, and it was the only connection
  561. * outside, so the child pgrp is now orphaned.
  562. */
  563. if ((process_group(p) != process_group(father)) &&
  564. (p->signal->session == father->signal->session)) {
  565. int pgrp = process_group(p);
  566. if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
  567. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
  568. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
  569. }
  570. }
  571. }
  572. /*
  573. * When we die, we re-parent all our children.
  574. * Try to give them to another thread in our thread
  575. * group, and if no such member exists, give it to
  576. * the global child reaper process (ie "init")
  577. */
  578. static void forget_original_parent(struct task_struct * father,
  579. struct list_head *to_release)
  580. {
  581. struct task_struct *p, *reaper = father;
  582. struct list_head *_p, *_n;
  583. do {
  584. reaper = next_thread(reaper);
  585. if (reaper == father) {
  586. reaper = child_reaper;
  587. break;
  588. }
  589. } while (reaper->exit_state);
  590. /*
  591. * There are only two places where our children can be:
  592. *
  593. * - in our child list
  594. * - in our ptraced child list
  595. *
  596. * Search them and reparent children.
  597. */
  598. list_for_each_safe(_p, _n, &father->children) {
  599. int ptrace;
  600. p = list_entry(_p,struct task_struct,sibling);
  601. ptrace = p->ptrace;
  602. /* if father isn't the real parent, then ptrace must be enabled */
  603. BUG_ON(father != p->real_parent && !ptrace);
  604. if (father == p->real_parent) {
  605. /* reparent with a reaper, real father it's us */
  606. choose_new_parent(p, reaper);
  607. reparent_thread(p, father, 0);
  608. } else {
  609. /* reparent ptraced task to its real parent */
  610. __ptrace_unlink (p);
  611. if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
  612. thread_group_empty(p))
  613. do_notify_parent(p, p->exit_signal);
  614. }
  615. /*
  616. * if the ptraced child is a zombie with exit_signal == -1
  617. * we must collect it before we exit, or it will remain
  618. * zombie forever since we prevented it from self-reap itself
  619. * while it was being traced by us, to be able to see it in wait4.
  620. */
  621. if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
  622. list_add(&p->ptrace_list, to_release);
  623. }
  624. list_for_each_safe(_p, _n, &father->ptrace_children) {
  625. p = list_entry(_p,struct task_struct,ptrace_list);
  626. choose_new_parent(p, reaper);
  627. reparent_thread(p, father, 1);
  628. }
  629. }
  630. /*
  631. * Send signals to all our closest relatives so that they know
  632. * to properly mourn us..
  633. */
  634. static void exit_notify(struct task_struct *tsk)
  635. {
  636. int state;
  637. struct task_struct *t;
  638. struct list_head ptrace_dead, *_p, *_n;
  639. if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
  640. && !thread_group_empty(tsk)) {
  641. /*
  642. * This occurs when there was a race between our exit
  643. * syscall and a group signal choosing us as the one to
  644. * wake up. It could be that we are the only thread
  645. * alerted to check for pending signals, but another thread
  646. * should be woken now to take the signal since we will not.
  647. * Now we'll wake all the threads in the group just to make
  648. * sure someone gets all the pending signals.
  649. */
  650. read_lock(&tasklist_lock);
  651. spin_lock_irq(&tsk->sighand->siglock);
  652. for (t = next_thread(tsk); t != tsk; t = next_thread(t))
  653. if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
  654. recalc_sigpending_tsk(t);
  655. if (signal_pending(t))
  656. signal_wake_up(t, 0);
  657. }
  658. spin_unlock_irq(&tsk->sighand->siglock);
  659. read_unlock(&tasklist_lock);
  660. }
  661. write_lock_irq(&tasklist_lock);
  662. /*
  663. * This does two things:
  664. *
  665. * A. Make init inherit all the child processes
  666. * B. Check to see if any process groups have become orphaned
  667. * as a result of our exiting, and if they have any stopped
  668. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  669. */
  670. INIT_LIST_HEAD(&ptrace_dead);
  671. forget_original_parent(tsk, &ptrace_dead);
  672. BUG_ON(!list_empty(&tsk->children));
  673. BUG_ON(!list_empty(&tsk->ptrace_children));
  674. /*
  675. * Check to see if any process groups have become orphaned
  676. * as a result of our exiting, and if they have any stopped
  677. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  678. *
  679. * Case i: Our father is in a different pgrp than we are
  680. * and we were the only connection outside, so our pgrp
  681. * is about to become orphaned.
  682. */
  683. t = tsk->real_parent;
  684. if ((process_group(t) != process_group(tsk)) &&
  685. (t->signal->session == tsk->signal->session) &&
  686. will_become_orphaned_pgrp(process_group(tsk), tsk) &&
  687. has_stopped_jobs(process_group(tsk))) {
  688. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
  689. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
  690. }
  691. /* Let father know we died
  692. *
  693. * Thread signals are configurable, but you aren't going to use
  694. * that to send signals to arbitary processes.
  695. * That stops right now.
  696. *
  697. * If the parent exec id doesn't match the exec id we saved
  698. * when we started then we know the parent has changed security
  699. * domain.
  700. *
  701. * If our self_exec id doesn't match our parent_exec_id then
  702. * we have changed execution domain as these two values started
  703. * the same after a fork.
  704. *
  705. */
  706. if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
  707. ( tsk->parent_exec_id != t->self_exec_id ||
  708. tsk->self_exec_id != tsk->parent_exec_id)
  709. && !capable(CAP_KILL))
  710. tsk->exit_signal = SIGCHLD;
  711. /* If something other than our normal parent is ptracing us, then
  712. * send it a SIGCHLD instead of honoring exit_signal. exit_signal
  713. * only has special meaning to our real parent.
  714. */
  715. if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
  716. int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
  717. do_notify_parent(tsk, signal);
  718. } else if (tsk->ptrace) {
  719. do_notify_parent(tsk, SIGCHLD);
  720. }
  721. state = EXIT_ZOMBIE;
  722. if (tsk->exit_signal == -1 &&
  723. (likely(tsk->ptrace == 0) ||
  724. unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
  725. state = EXIT_DEAD;
  726. tsk->exit_state = state;
  727. write_unlock_irq(&tasklist_lock);
  728. list_for_each_safe(_p, _n, &ptrace_dead) {
  729. list_del_init(_p);
  730. t = list_entry(_p,struct task_struct,ptrace_list);
  731. release_task(t);
  732. }
  733. /* If the process is dead, release it - nobody will wait for it */
  734. if (state == EXIT_DEAD)
  735. release_task(tsk);
  736. }
  737. fastcall NORET_TYPE void do_exit(long code)
  738. {
  739. struct task_struct *tsk = current;
  740. int group_dead;
  741. profile_task_exit(tsk);
  742. WARN_ON(atomic_read(&tsk->fs_excl));
  743. if (unlikely(in_interrupt()))
  744. panic("Aiee, killing interrupt handler!");
  745. if (unlikely(!tsk->pid))
  746. panic("Attempted to kill the idle task!");
  747. if (unlikely(tsk == child_reaper))
  748. panic("Attempted to kill init!");
  749. if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
  750. current->ptrace_message = code;
  751. ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
  752. }
  753. /*
  754. * We're taking recursive faults here in do_exit. Safest is to just
  755. * leave this task alone and wait for reboot.
  756. */
  757. if (unlikely(tsk->flags & PF_EXITING)) {
  758. printk(KERN_ALERT
  759. "Fixing recursive fault but reboot is needed!\n");
  760. if (tsk->io_context)
  761. exit_io_context();
  762. set_current_state(TASK_UNINTERRUPTIBLE);
  763. schedule();
  764. }
  765. tsk->flags |= PF_EXITING;
  766. if (unlikely(in_atomic()))
  767. printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
  768. current->comm, current->pid,
  769. preempt_count());
  770. acct_update_integrals(tsk);
  771. if (tsk->mm) {
  772. update_hiwater_rss(tsk->mm);
  773. update_hiwater_vm(tsk->mm);
  774. }
  775. group_dead = atomic_dec_and_test(&tsk->signal->live);
  776. if (group_dead) {
  777. hrtimer_cancel(&tsk->signal->real_timer);
  778. exit_itimers(tsk->signal);
  779. }
  780. acct_collect(code, group_dead);
  781. if (unlikely(tsk->robust_list))
  782. exit_robust_list(tsk);
  783. #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
  784. if (unlikely(tsk->compat_robust_list))
  785. compat_exit_robust_list(tsk);
  786. #endif
  787. if (unlikely(tsk->audit_context))
  788. audit_free(tsk);
  789. exit_mm(tsk);
  790. if (group_dead)
  791. acct_process();
  792. exit_sem(tsk);
  793. __exit_files(tsk);
  794. __exit_fs(tsk);
  795. exit_namespace(tsk);
  796. exit_thread();
  797. cpuset_exit(tsk);
  798. exit_keys(tsk);
  799. if (group_dead && tsk->signal->leader)
  800. disassociate_ctty(1);
  801. module_put(task_thread_info(tsk)->exec_domain->module);
  802. if (tsk->binfmt)
  803. module_put(tsk->binfmt->module);
  804. tsk->exit_code = code;
  805. proc_exit_connector(tsk);
  806. exit_notify(tsk);
  807. #ifdef CONFIG_NUMA
  808. mpol_free(tsk->mempolicy);
  809. tsk->mempolicy = NULL;
  810. #endif
  811. /*
  812. * This must happen late, after the PID is not
  813. * hashed anymore:
  814. */
  815. if (unlikely(!list_empty(&tsk->pi_state_list)))
  816. exit_pi_state_list(tsk);
  817. if (unlikely(current->pi_state_cache))
  818. kfree(current->pi_state_cache);
  819. /*
  820. * If DEBUG_MUTEXES is on, make sure we are holding no locks:
  821. */
  822. mutex_debug_check_no_locks_held(tsk);
  823. rt_mutex_debug_check_no_locks_held(tsk);
  824. if (tsk->io_context)
  825. exit_io_context();
  826. if (tsk->splice_pipe)
  827. __free_pipe_info(tsk->splice_pipe);
  828. /* PF_DEAD causes final put_task_struct after we schedule. */
  829. preempt_disable();
  830. BUG_ON(tsk->flags & PF_DEAD);
  831. tsk->flags |= PF_DEAD;
  832. schedule();
  833. BUG();
  834. /* Avoid "noreturn function does return". */
  835. for (;;) ;
  836. }
  837. EXPORT_SYMBOL_GPL(do_exit);
  838. NORET_TYPE void complete_and_exit(struct completion *comp, long code)
  839. {
  840. if (comp)
  841. complete(comp);
  842. do_exit(code);
  843. }
  844. EXPORT_SYMBOL(complete_and_exit);
  845. asmlinkage long sys_exit(int error_code)
  846. {
  847. do_exit((error_code&0xff)<<8);
  848. }
  849. /*
  850. * Take down every thread in the group. This is called by fatal signals
  851. * as well as by sys_exit_group (below).
  852. */
  853. NORET_TYPE void
  854. do_group_exit(int exit_code)
  855. {
  856. BUG_ON(exit_code & 0x80); /* core dumps don't get here */
  857. if (current->signal->flags & SIGNAL_GROUP_EXIT)
  858. exit_code = current->signal->group_exit_code;
  859. else if (!thread_group_empty(current)) {
  860. struct signal_struct *const sig = current->signal;
  861. struct sighand_struct *const sighand = current->sighand;
  862. spin_lock_irq(&sighand->siglock);
  863. if (sig->flags & SIGNAL_GROUP_EXIT)
  864. /* Another thread got here before we took the lock. */
  865. exit_code = sig->group_exit_code;
  866. else {
  867. sig->group_exit_code = exit_code;
  868. zap_other_threads(current);
  869. }
  870. spin_unlock_irq(&sighand->siglock);
  871. }
  872. do_exit(exit_code);
  873. /* NOTREACHED */
  874. }
  875. /*
  876. * this kills every thread in the thread group. Note that any externally
  877. * wait4()-ing process will get the correct exit code - even if this
  878. * thread is not the thread group leader.
  879. */
  880. asmlinkage void sys_exit_group(int error_code)
  881. {
  882. do_group_exit((error_code & 0xff) << 8);
  883. }
  884. static int eligible_child(pid_t pid, int options, task_t *p)
  885. {
  886. if (pid > 0) {
  887. if (p->pid != pid)
  888. return 0;
  889. } else if (!pid) {
  890. if (process_group(p) != process_group(current))
  891. return 0;
  892. } else if (pid != -1) {
  893. if (process_group(p) != -pid)
  894. return 0;
  895. }
  896. /*
  897. * Do not consider detached threads that are
  898. * not ptraced:
  899. */
  900. if (p->exit_signal == -1 && !p->ptrace)
  901. return 0;
  902. /* Wait for all children (clone and not) if __WALL is set;
  903. * otherwise, wait for clone children *only* if __WCLONE is
  904. * set; otherwise, wait for non-clone children *only*. (Note:
  905. * A "clone" child here is one that reports to its parent
  906. * using a signal other than SIGCHLD.) */
  907. if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
  908. && !(options & __WALL))
  909. return 0;
  910. /*
  911. * Do not consider thread group leaders that are
  912. * in a non-empty thread group:
  913. */
  914. if (current->tgid != p->tgid && delay_group_leader(p))
  915. return 2;
  916. if (security_task_wait(p))
  917. return 0;
  918. return 1;
  919. }
  920. static int wait_noreap_copyout(task_t *p, pid_t pid, uid_t uid,
  921. int why, int status,
  922. struct siginfo __user *infop,
  923. struct rusage __user *rusagep)
  924. {
  925. int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
  926. put_task_struct(p);
  927. if (!retval)
  928. retval = put_user(SIGCHLD, &infop->si_signo);
  929. if (!retval)
  930. retval = put_user(0, &infop->si_errno);
  931. if (!retval)
  932. retval = put_user((short)why, &infop->si_code);
  933. if (!retval)
  934. retval = put_user(pid, &infop->si_pid);
  935. if (!retval)
  936. retval = put_user(uid, &infop->si_uid);
  937. if (!retval)
  938. retval = put_user(status, &infop->si_status);
  939. if (!retval)
  940. retval = pid;
  941. return retval;
  942. }
  943. /*
  944. * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
  945. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  946. * the lock and this task is uninteresting. If we return nonzero, we have
  947. * released the lock and the system call should return.
  948. */
  949. static int wait_task_zombie(task_t *p, int noreap,
  950. struct siginfo __user *infop,
  951. int __user *stat_addr, struct rusage __user *ru)
  952. {
  953. unsigned long state;
  954. int retval;
  955. int status;
  956. if (unlikely(noreap)) {
  957. pid_t pid = p->pid;
  958. uid_t uid = p->uid;
  959. int exit_code = p->exit_code;
  960. int why, status;
  961. if (unlikely(p->exit_state != EXIT_ZOMBIE))
  962. return 0;
  963. if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
  964. return 0;
  965. get_task_struct(p);
  966. read_unlock(&tasklist_lock);
  967. if ((exit_code & 0x7f) == 0) {
  968. why = CLD_EXITED;
  969. status = exit_code >> 8;
  970. } else {
  971. why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
  972. status = exit_code & 0x7f;
  973. }
  974. return wait_noreap_copyout(p, pid, uid, why,
  975. status, infop, ru);
  976. }
  977. /*
  978. * Try to move the task's state to DEAD
  979. * only one thread is allowed to do this:
  980. */
  981. state = xchg(&p->exit_state, EXIT_DEAD);
  982. if (state != EXIT_ZOMBIE) {
  983. BUG_ON(state != EXIT_DEAD);
  984. return 0;
  985. }
  986. if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
  987. /*
  988. * This can only happen in a race with a ptraced thread
  989. * dying on another processor.
  990. */
  991. return 0;
  992. }
  993. if (likely(p->real_parent == p->parent) && likely(p->signal)) {
  994. struct signal_struct *psig;
  995. struct signal_struct *sig;
  996. /*
  997. * The resource counters for the group leader are in its
  998. * own task_struct. Those for dead threads in the group
  999. * are in its signal_struct, as are those for the child
  1000. * processes it has previously reaped. All these
  1001. * accumulate in the parent's signal_struct c* fields.
  1002. *
  1003. * We don't bother to take a lock here to protect these
  1004. * p->signal fields, because they are only touched by
  1005. * __exit_signal, which runs with tasklist_lock
  1006. * write-locked anyway, and so is excluded here. We do
  1007. * need to protect the access to p->parent->signal fields,
  1008. * as other threads in the parent group can be right
  1009. * here reaping other children at the same time.
  1010. */
  1011. spin_lock_irq(&p->parent->sighand->siglock);
  1012. psig = p->parent->signal;
  1013. sig = p->signal;
  1014. psig->cutime =
  1015. cputime_add(psig->cutime,
  1016. cputime_add(p->utime,
  1017. cputime_add(sig->utime,
  1018. sig->cutime)));
  1019. psig->cstime =
  1020. cputime_add(psig->cstime,
  1021. cputime_add(p->stime,
  1022. cputime_add(sig->stime,
  1023. sig->cstime)));
  1024. psig->cmin_flt +=
  1025. p->min_flt + sig->min_flt + sig->cmin_flt;
  1026. psig->cmaj_flt +=
  1027. p->maj_flt + sig->maj_flt + sig->cmaj_flt;
  1028. psig->cnvcsw +=
  1029. p->nvcsw + sig->nvcsw + sig->cnvcsw;
  1030. psig->cnivcsw +=
  1031. p->nivcsw + sig->nivcsw + sig->cnivcsw;
  1032. spin_unlock_irq(&p->parent->sighand->siglock);
  1033. }
  1034. /*
  1035. * Now we are sure this task is interesting, and no other
  1036. * thread can reap it because we set its state to EXIT_DEAD.
  1037. */
  1038. read_unlock(&tasklist_lock);
  1039. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1040. status = (p->signal->flags & SIGNAL_GROUP_EXIT)
  1041. ? p->signal->group_exit_code : p->exit_code;
  1042. if (!retval && stat_addr)
  1043. retval = put_user(status, stat_addr);
  1044. if (!retval && infop)
  1045. retval = put_user(SIGCHLD, &infop->si_signo);
  1046. if (!retval && infop)
  1047. retval = put_user(0, &infop->si_errno);
  1048. if (!retval && infop) {
  1049. int why;
  1050. if ((status & 0x7f) == 0) {
  1051. why = CLD_EXITED;
  1052. status >>= 8;
  1053. } else {
  1054. why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
  1055. status &= 0x7f;
  1056. }
  1057. retval = put_user((short)why, &infop->si_code);
  1058. if (!retval)
  1059. retval = put_user(status, &infop->si_status);
  1060. }
  1061. if (!retval && infop)
  1062. retval = put_user(p->pid, &infop->si_pid);
  1063. if (!retval && infop)
  1064. retval = put_user(p->uid, &infop->si_uid);
  1065. if (retval) {
  1066. // TODO: is this safe?
  1067. p->exit_state = EXIT_ZOMBIE;
  1068. return retval;
  1069. }
  1070. retval = p->pid;
  1071. if (p->real_parent != p->parent) {
  1072. write_lock_irq(&tasklist_lock);
  1073. /* Double-check with lock held. */
  1074. if (p->real_parent != p->parent) {
  1075. __ptrace_unlink(p);
  1076. // TODO: is this safe?
  1077. p->exit_state = EXIT_ZOMBIE;
  1078. /*
  1079. * If this is not a detached task, notify the parent.
  1080. * If it's still not detached after that, don't release
  1081. * it now.
  1082. */
  1083. if (p->exit_signal != -1) {
  1084. do_notify_parent(p, p->exit_signal);
  1085. if (p->exit_signal != -1)
  1086. p = NULL;
  1087. }
  1088. }
  1089. write_unlock_irq(&tasklist_lock);
  1090. }
  1091. if (p != NULL)
  1092. release_task(p);
  1093. BUG_ON(!retval);
  1094. return retval;
  1095. }
  1096. /*
  1097. * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
  1098. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1099. * the lock and this task is uninteresting. If we return nonzero, we have
  1100. * released the lock and the system call should return.
  1101. */
  1102. static int wait_task_stopped(task_t *p, int delayed_group_leader, int noreap,
  1103. struct siginfo __user *infop,
  1104. int __user *stat_addr, struct rusage __user *ru)
  1105. {
  1106. int retval, exit_code;
  1107. if (!p->exit_code)
  1108. return 0;
  1109. if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
  1110. p->signal && p->signal->group_stop_count > 0)
  1111. /*
  1112. * A group stop is in progress and this is the group leader.
  1113. * We won't report until all threads have stopped.
  1114. */
  1115. return 0;
  1116. /*
  1117. * Now we are pretty sure this task is interesting.
  1118. * Make sure it doesn't get reaped out from under us while we
  1119. * give up the lock and then examine it below. We don't want to
  1120. * keep holding onto the tasklist_lock while we call getrusage and
  1121. * possibly take page faults for user memory.
  1122. */
  1123. get_task_struct(p);
  1124. read_unlock(&tasklist_lock);
  1125. if (unlikely(noreap)) {
  1126. pid_t pid = p->pid;
  1127. uid_t uid = p->uid;
  1128. int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
  1129. exit_code = p->exit_code;
  1130. if (unlikely(!exit_code) ||
  1131. unlikely(p->state & TASK_TRACED))
  1132. goto bail_ref;
  1133. return wait_noreap_copyout(p, pid, uid,
  1134. why, (exit_code << 8) | 0x7f,
  1135. infop, ru);
  1136. }
  1137. write_lock_irq(&tasklist_lock);
  1138. /*
  1139. * This uses xchg to be atomic with the thread resuming and setting
  1140. * it. It must also be done with the write lock held to prevent a
  1141. * race with the EXIT_ZOMBIE case.
  1142. */
  1143. exit_code = xchg(&p->exit_code, 0);
  1144. if (unlikely(p->exit_state)) {
  1145. /*
  1146. * The task resumed and then died. Let the next iteration
  1147. * catch it in EXIT_ZOMBIE. Note that exit_code might
  1148. * already be zero here if it resumed and did _exit(0).
  1149. * The task itself is dead and won't touch exit_code again;
  1150. * other processors in this function are locked out.
  1151. */
  1152. p->exit_code = exit_code;
  1153. exit_code = 0;
  1154. }
  1155. if (unlikely(exit_code == 0)) {
  1156. /*
  1157. * Another thread in this function got to it first, or it
  1158. * resumed, or it resumed and then died.
  1159. */
  1160. write_unlock_irq(&tasklist_lock);
  1161. bail_ref:
  1162. put_task_struct(p);
  1163. /*
  1164. * We are returning to the wait loop without having successfully
  1165. * removed the process and having released the lock. We cannot
  1166. * continue, since the "p" task pointer is potentially stale.
  1167. *
  1168. * Return -EAGAIN, and do_wait() will restart the loop from the
  1169. * beginning. Do _not_ re-acquire the lock.
  1170. */
  1171. return -EAGAIN;
  1172. }
  1173. /* move to end of parent's list to avoid starvation */
  1174. remove_parent(p);
  1175. add_parent(p);
  1176. write_unlock_irq(&tasklist_lock);
  1177. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1178. if (!retval && stat_addr)
  1179. retval = put_user((exit_code << 8) | 0x7f, stat_addr);
  1180. if (!retval && infop)
  1181. retval = put_user(SIGCHLD, &infop->si_signo);
  1182. if (!retval && infop)
  1183. retval = put_user(0, &infop->si_errno);
  1184. if (!retval && infop)
  1185. retval = put_user((short)((p->ptrace & PT_PTRACED)
  1186. ? CLD_TRAPPED : CLD_STOPPED),
  1187. &infop->si_code);
  1188. if (!retval && infop)
  1189. retval = put_user(exit_code, &infop->si_status);
  1190. if (!retval && infop)
  1191. retval = put_user(p->pid, &infop->si_pid);
  1192. if (!retval && infop)
  1193. retval = put_user(p->uid, &infop->si_uid);
  1194. if (!retval)
  1195. retval = p->pid;
  1196. put_task_struct(p);
  1197. BUG_ON(!retval);
  1198. return retval;
  1199. }
  1200. /*
  1201. * Handle do_wait work for one task in a live, non-stopped state.
  1202. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1203. * the lock and this task is uninteresting. If we return nonzero, we have
  1204. * released the lock and the system call should return.
  1205. */
  1206. static int wait_task_continued(task_t *p, int noreap,
  1207. struct siginfo __user *infop,
  1208. int __user *stat_addr, struct rusage __user *ru)
  1209. {
  1210. int retval;
  1211. pid_t pid;
  1212. uid_t uid;
  1213. if (unlikely(!p->signal))
  1214. return 0;
  1215. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
  1216. return 0;
  1217. spin_lock_irq(&p->sighand->siglock);
  1218. /* Re-check with the lock held. */
  1219. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
  1220. spin_unlock_irq(&p->sighand->siglock);
  1221. return 0;
  1222. }
  1223. if (!noreap)
  1224. p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
  1225. spin_unlock_irq(&p->sighand->siglock);
  1226. pid = p->pid;
  1227. uid = p->uid;
  1228. get_task_struct(p);
  1229. read_unlock(&tasklist_lock);
  1230. if (!infop) {
  1231. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1232. put_task_struct(p);
  1233. if (!retval && stat_addr)
  1234. retval = put_user(0xffff, stat_addr);
  1235. if (!retval)
  1236. retval = p->pid;
  1237. } else {
  1238. retval = wait_noreap_copyout(p, pid, uid,
  1239. CLD_CONTINUED, SIGCONT,
  1240. infop, ru);
  1241. BUG_ON(retval == 0);
  1242. }
  1243. return retval;
  1244. }
  1245. static inline int my_ptrace_child(struct task_struct *p)
  1246. {
  1247. if (!(p->ptrace & PT_PTRACED))
  1248. return 0;
  1249. if (!(p->ptrace & PT_ATTACHED))
  1250. return 1;
  1251. /*
  1252. * This child was PTRACE_ATTACH'd. We should be seeing it only if
  1253. * we are the attacher. If we are the real parent, this is a race
  1254. * inside ptrace_attach. It is waiting for the tasklist_lock,
  1255. * which we have to switch the parent links, but has already set
  1256. * the flags in p->ptrace.
  1257. */
  1258. return (p->parent != p->real_parent);
  1259. }
  1260. static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
  1261. int __user *stat_addr, struct rusage __user *ru)
  1262. {
  1263. DECLARE_WAITQUEUE(wait, current);
  1264. struct task_struct *tsk;
  1265. int flag, retval;
  1266. add_wait_queue(&current->signal->wait_chldexit,&wait);
  1267. repeat:
  1268. /*
  1269. * We will set this flag if we see any child that might later
  1270. * match our criteria, even if we are not able to reap it yet.
  1271. */
  1272. flag = 0;
  1273. current->state = TASK_INTERRUPTIBLE;
  1274. read_lock(&tasklist_lock);
  1275. tsk = current;
  1276. do {
  1277. struct task_struct *p;
  1278. struct list_head *_p;
  1279. int ret;
  1280. list_for_each(_p,&tsk->children) {
  1281. p = list_entry(_p,struct task_struct,sibling);
  1282. ret = eligible_child(pid, options, p);
  1283. if (!ret)
  1284. continue;
  1285. switch (p->state) {
  1286. case TASK_TRACED:
  1287. /*
  1288. * When we hit the race with PTRACE_ATTACH,
  1289. * we will not report this child. But the
  1290. * race means it has not yet been moved to
  1291. * our ptrace_children list, so we need to
  1292. * set the flag here to avoid a spurious ECHILD
  1293. * when the race happens with the only child.
  1294. */
  1295. flag = 1;
  1296. if (!my_ptrace_child(p))
  1297. continue;
  1298. /*FALLTHROUGH*/
  1299. case TASK_STOPPED:
  1300. /*
  1301. * It's stopped now, so it might later
  1302. * continue, exit, or stop again.
  1303. */
  1304. flag = 1;
  1305. if (!(options & WUNTRACED) &&
  1306. !my_ptrace_child(p))
  1307. continue;
  1308. retval = wait_task_stopped(p, ret == 2,
  1309. (options & WNOWAIT),
  1310. infop,
  1311. stat_addr, ru);
  1312. if (retval == -EAGAIN)
  1313. goto repeat;
  1314. if (retval != 0) /* He released the lock. */
  1315. goto end;
  1316. break;
  1317. default:
  1318. // case EXIT_DEAD:
  1319. if (p->exit_state == EXIT_DEAD)
  1320. continue;
  1321. // case EXIT_ZOMBIE:
  1322. if (p->exit_state == EXIT_ZOMBIE) {
  1323. /*
  1324. * Eligible but we cannot release
  1325. * it yet:
  1326. */
  1327. if (ret == 2)
  1328. goto check_continued;
  1329. if (!likely(options & WEXITED))
  1330. continue;
  1331. retval = wait_task_zombie(
  1332. p, (options & WNOWAIT),
  1333. infop, stat_addr, ru);
  1334. /* He released the lock. */
  1335. if (retval != 0)
  1336. goto end;
  1337. break;
  1338. }
  1339. check_continued:
  1340. /*
  1341. * It's running now, so it might later
  1342. * exit, stop, or stop and then continue.
  1343. */
  1344. flag = 1;
  1345. if (!unlikely(options & WCONTINUED))
  1346. continue;
  1347. retval = wait_task_continued(
  1348. p, (options & WNOWAIT),
  1349. infop, stat_addr, ru);
  1350. if (retval != 0) /* He released the lock. */
  1351. goto end;
  1352. break;
  1353. }
  1354. }
  1355. if (!flag) {
  1356. list_for_each(_p, &tsk->ptrace_children) {
  1357. p = list_entry(_p, struct task_struct,
  1358. ptrace_list);
  1359. if (!eligible_child(pid, options, p))
  1360. continue;
  1361. flag = 1;
  1362. break;
  1363. }
  1364. }
  1365. if (options & __WNOTHREAD)
  1366. break;
  1367. tsk = next_thread(tsk);
  1368. BUG_ON(tsk->signal != current->signal);
  1369. } while (tsk != current);
  1370. read_unlock(&tasklist_lock);
  1371. if (flag) {
  1372. retval = 0;
  1373. if (options & WNOHANG)
  1374. goto end;
  1375. retval = -ERESTARTSYS;
  1376. if (signal_pending(current))
  1377. goto end;
  1378. schedule();
  1379. goto repeat;
  1380. }
  1381. retval = -ECHILD;
  1382. end:
  1383. current->state = TASK_RUNNING;
  1384. remove_wait_queue(&current->signal->wait_chldexit,&wait);
  1385. if (infop) {
  1386. if (retval > 0)
  1387. retval = 0;
  1388. else {
  1389. /*
  1390. * For a WNOHANG return, clear out all the fields
  1391. * we would set so the user can easily tell the
  1392. * difference.
  1393. */
  1394. if (!retval)
  1395. retval = put_user(0, &infop->si_signo);
  1396. if (!retval)
  1397. retval = put_user(0, &infop->si_errno);
  1398. if (!retval)
  1399. retval = put_user(0, &infop->si_code);
  1400. if (!retval)
  1401. retval = put_user(0, &infop->si_pid);
  1402. if (!retval)
  1403. retval = put_user(0, &infop->si_uid);
  1404. if (!retval)
  1405. retval = put_user(0, &infop->si_status);
  1406. }
  1407. }
  1408. return retval;
  1409. }
  1410. asmlinkage long sys_waitid(int which, pid_t pid,
  1411. struct siginfo __user *infop, int options,
  1412. struct rusage __user *ru)
  1413. {
  1414. long ret;
  1415. if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
  1416. return -EINVAL;
  1417. if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
  1418. return -EINVAL;
  1419. switch (which) {
  1420. case P_ALL:
  1421. pid = -1;
  1422. break;
  1423. case P_PID:
  1424. if (pid <= 0)
  1425. return -EINVAL;
  1426. break;
  1427. case P_PGID:
  1428. if (pid <= 0)
  1429. return -EINVAL;
  1430. pid = -pid;
  1431. break;
  1432. default:
  1433. return -EINVAL;
  1434. }
  1435. ret = do_wait(pid, options, infop, NULL, ru);
  1436. /* avoid REGPARM breakage on x86: */
  1437. prevent_tail_call(ret);
  1438. return ret;
  1439. }
  1440. asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
  1441. int options, struct rusage __user *ru)
  1442. {
  1443. long ret;
  1444. if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
  1445. __WNOTHREAD|__WCLONE|__WALL))
  1446. return -EINVAL;
  1447. ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
  1448. /* avoid REGPARM breakage on x86: */
  1449. prevent_tail_call(ret);
  1450. return ret;
  1451. }
  1452. #ifdef __ARCH_WANT_SYS_WAITPID
  1453. /*
  1454. * sys_waitpid() remains for compatibility. waitpid() should be
  1455. * implemented by calling sys_wait4() from libc.a.
  1456. */
  1457. asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
  1458. {
  1459. return sys_wait4(pid, stat_addr, options, NULL);
  1460. }
  1461. #endif