exit.c 41 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652
  1. /*
  2. * linux/kernel/exit.c
  3. *
  4. * Copyright (C) 1991, 1992 Linus Torvalds
  5. */
  6. #include <linux/mm.h>
  7. #include <linux/slab.h>
  8. #include <linux/interrupt.h>
  9. #include <linux/smp_lock.h>
  10. #include <linux/module.h>
  11. #include <linux/capability.h>
  12. #include <linux/completion.h>
  13. #include <linux/personality.h>
  14. #include <linux/tty.h>
  15. #include <linux/namespace.h>
  16. #include <linux/key.h>
  17. #include <linux/security.h>
  18. #include <linux/cpu.h>
  19. #include <linux/acct.h>
  20. #include <linux/tsacct_kern.h>
  21. #include <linux/file.h>
  22. #include <linux/binfmts.h>
  23. #include <linux/nsproxy.h>
  24. #include <linux/ptrace.h>
  25. #include <linux/profile.h>
  26. #include <linux/mount.h>
  27. #include <linux/proc_fs.h>
  28. #include <linux/mempolicy.h>
  29. #include <linux/taskstats_kern.h>
  30. #include <linux/delayacct.h>
  31. #include <linux/cpuset.h>
  32. #include <linux/syscalls.h>
  33. #include <linux/signal.h>
  34. #include <linux/posix-timers.h>
  35. #include <linux/cn_proc.h>
  36. #include <linux/mutex.h>
  37. #include <linux/futex.h>
  38. #include <linux/compat.h>
  39. #include <linux/pipe_fs_i.h>
  40. #include <linux/audit.h> /* for audit_free() */
  41. #include <linux/resource.h>
  42. #include <linux/blkdev.h>
  43. #include <asm/uaccess.h>
  44. #include <asm/unistd.h>
  45. #include <asm/pgtable.h>
  46. #include <asm/mmu_context.h>
  47. extern void sem_exit (void);
  48. extern struct task_struct *child_reaper;
  49. static void exit_mm(struct task_struct * tsk);
  50. static void __unhash_process(struct task_struct *p)
  51. {
  52. nr_threads--;
  53. detach_pid(p, PIDTYPE_PID);
  54. if (thread_group_leader(p)) {
  55. detach_pid(p, PIDTYPE_PGID);
  56. detach_pid(p, PIDTYPE_SID);
  57. list_del_rcu(&p->tasks);
  58. __get_cpu_var(process_counts)--;
  59. }
  60. list_del_rcu(&p->thread_group);
  61. remove_parent(p);
  62. }
  63. /*
  64. * This function expects the tasklist_lock write-locked.
  65. */
  66. static void __exit_signal(struct task_struct *tsk)
  67. {
  68. struct signal_struct *sig = tsk->signal;
  69. struct sighand_struct *sighand;
  70. BUG_ON(!sig);
  71. BUG_ON(!atomic_read(&sig->count));
  72. rcu_read_lock();
  73. sighand = rcu_dereference(tsk->sighand);
  74. spin_lock(&sighand->siglock);
  75. posix_cpu_timers_exit(tsk);
  76. if (atomic_dec_and_test(&sig->count))
  77. posix_cpu_timers_exit_group(tsk);
  78. else {
  79. /*
  80. * If there is any task waiting for the group exit
  81. * then notify it:
  82. */
  83. if (sig->group_exit_task && atomic_read(&sig->count) == sig->notify_count) {
  84. wake_up_process(sig->group_exit_task);
  85. sig->group_exit_task = NULL;
  86. }
  87. if (tsk == sig->curr_target)
  88. sig->curr_target = next_thread(tsk);
  89. /*
  90. * Accumulate here the counters for all threads but the
  91. * group leader as they die, so they can be added into
  92. * the process-wide totals when those are taken.
  93. * The group leader stays around as a zombie as long
  94. * as there are other threads. When it gets reaped,
  95. * the exit.c code will add its counts into these totals.
  96. * We won't ever get here for the group leader, since it
  97. * will have been the last reference on the signal_struct.
  98. */
  99. sig->utime = cputime_add(sig->utime, tsk->utime);
  100. sig->stime = cputime_add(sig->stime, tsk->stime);
  101. sig->min_flt += tsk->min_flt;
  102. sig->maj_flt += tsk->maj_flt;
  103. sig->nvcsw += tsk->nvcsw;
  104. sig->nivcsw += tsk->nivcsw;
  105. sig->sched_time += tsk->sched_time;
  106. sig = NULL; /* Marker for below. */
  107. }
  108. __unhash_process(tsk);
  109. tsk->signal = NULL;
  110. tsk->sighand = NULL;
  111. spin_unlock(&sighand->siglock);
  112. rcu_read_unlock();
  113. __cleanup_sighand(sighand);
  114. clear_tsk_thread_flag(tsk,TIF_SIGPENDING);
  115. flush_sigqueue(&tsk->pending);
  116. if (sig) {
  117. flush_sigqueue(&sig->shared_pending);
  118. taskstats_tgid_free(sig);
  119. __cleanup_signal(sig);
  120. }
  121. }
  122. static void delayed_put_task_struct(struct rcu_head *rhp)
  123. {
  124. put_task_struct(container_of(rhp, struct task_struct, rcu));
  125. }
  126. void release_task(struct task_struct * p)
  127. {
  128. struct task_struct *leader;
  129. int zap_leader;
  130. repeat:
  131. atomic_dec(&p->user->processes);
  132. write_lock_irq(&tasklist_lock);
  133. ptrace_unlink(p);
  134. BUG_ON(!list_empty(&p->ptrace_list) || !list_empty(&p->ptrace_children));
  135. __exit_signal(p);
  136. /*
  137. * If we are the last non-leader member of the thread
  138. * group, and the leader is zombie, then notify the
  139. * group leader's parent process. (if it wants notification.)
  140. */
  141. zap_leader = 0;
  142. leader = p->group_leader;
  143. if (leader != p && thread_group_empty(leader) && leader->exit_state == EXIT_ZOMBIE) {
  144. BUG_ON(leader->exit_signal == -1);
  145. do_notify_parent(leader, leader->exit_signal);
  146. /*
  147. * If we were the last child thread and the leader has
  148. * exited already, and the leader's parent ignores SIGCHLD,
  149. * then we are the one who should release the leader.
  150. *
  151. * do_notify_parent() will have marked it self-reaping in
  152. * that case.
  153. */
  154. zap_leader = (leader->exit_signal == -1);
  155. }
  156. sched_exit(p);
  157. write_unlock_irq(&tasklist_lock);
  158. proc_flush_task(p);
  159. release_thread(p);
  160. call_rcu(&p->rcu, delayed_put_task_struct);
  161. p = leader;
  162. if (unlikely(zap_leader))
  163. goto repeat;
  164. }
  165. /*
  166. * This checks not only the pgrp, but falls back on the pid if no
  167. * satisfactory pgrp is found. I dunno - gdb doesn't work correctly
  168. * without this...
  169. */
  170. int session_of_pgrp(int pgrp)
  171. {
  172. struct task_struct *p;
  173. int sid = -1;
  174. read_lock(&tasklist_lock);
  175. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  176. if (p->signal->session > 0) {
  177. sid = p->signal->session;
  178. goto out;
  179. }
  180. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  181. p = find_task_by_pid(pgrp);
  182. if (p)
  183. sid = p->signal->session;
  184. out:
  185. read_unlock(&tasklist_lock);
  186. return sid;
  187. }
  188. /*
  189. * Determine if a process group is "orphaned", according to the POSIX
  190. * definition in 2.2.2.52. Orphaned process groups are not to be affected
  191. * by terminal-generated stop signals. Newly orphaned process groups are
  192. * to receive a SIGHUP and a SIGCONT.
  193. *
  194. * "I ask you, have you ever known what it is to be an orphan?"
  195. */
  196. static int will_become_orphaned_pgrp(int pgrp, struct task_struct *ignored_task)
  197. {
  198. struct task_struct *p;
  199. int ret = 1;
  200. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  201. if (p == ignored_task
  202. || p->exit_state
  203. || is_init(p->real_parent))
  204. continue;
  205. if (process_group(p->real_parent) != pgrp
  206. && p->real_parent->signal->session == p->signal->session) {
  207. ret = 0;
  208. break;
  209. }
  210. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  211. return ret; /* (sighing) "Often!" */
  212. }
  213. int is_orphaned_pgrp(int pgrp)
  214. {
  215. int retval;
  216. read_lock(&tasklist_lock);
  217. retval = will_become_orphaned_pgrp(pgrp, NULL);
  218. read_unlock(&tasklist_lock);
  219. return retval;
  220. }
  221. static int has_stopped_jobs(int pgrp)
  222. {
  223. int retval = 0;
  224. struct task_struct *p;
  225. do_each_task_pid(pgrp, PIDTYPE_PGID, p) {
  226. if (p->state != TASK_STOPPED)
  227. continue;
  228. retval = 1;
  229. break;
  230. } while_each_task_pid(pgrp, PIDTYPE_PGID, p);
  231. return retval;
  232. }
  233. /**
  234. * reparent_to_init - Reparent the calling kernel thread to the init task.
  235. *
  236. * If a kernel thread is launched as a result of a system call, or if
  237. * it ever exits, it should generally reparent itself to init so that
  238. * it is correctly cleaned up on exit.
  239. *
  240. * The various task state such as scheduling policy and priority may have
  241. * been inherited from a user process, so we reset them to sane values here.
  242. *
  243. * NOTE that reparent_to_init() gives the caller full capabilities.
  244. */
  245. static void reparent_to_init(void)
  246. {
  247. write_lock_irq(&tasklist_lock);
  248. ptrace_unlink(current);
  249. /* Reparent to init */
  250. remove_parent(current);
  251. current->parent = child_reaper;
  252. current->real_parent = child_reaper;
  253. add_parent(current);
  254. /* Set the exit signal to SIGCHLD so we signal init on exit */
  255. current->exit_signal = SIGCHLD;
  256. if (!has_rt_policy(current) && (task_nice(current) < 0))
  257. set_user_nice(current, 0);
  258. /* cpus_allowed? */
  259. /* rt_priority? */
  260. /* signals? */
  261. security_task_reparent_to_init(current);
  262. memcpy(current->signal->rlim, init_task.signal->rlim,
  263. sizeof(current->signal->rlim));
  264. atomic_inc(&(INIT_USER->__count));
  265. write_unlock_irq(&tasklist_lock);
  266. switch_uid(INIT_USER);
  267. }
  268. void __set_special_pids(pid_t session, pid_t pgrp)
  269. {
  270. struct task_struct *curr = current->group_leader;
  271. if (curr->signal->session != session) {
  272. detach_pid(curr, PIDTYPE_SID);
  273. curr->signal->session = session;
  274. attach_pid(curr, PIDTYPE_SID, session);
  275. }
  276. if (process_group(curr) != pgrp) {
  277. detach_pid(curr, PIDTYPE_PGID);
  278. curr->signal->pgrp = pgrp;
  279. attach_pid(curr, PIDTYPE_PGID, pgrp);
  280. }
  281. }
  282. static void set_special_pids(pid_t session, pid_t pgrp)
  283. {
  284. write_lock_irq(&tasklist_lock);
  285. __set_special_pids(session, pgrp);
  286. write_unlock_irq(&tasklist_lock);
  287. }
  288. /*
  289. * Let kernel threads use this to say that they
  290. * allow a certain signal (since daemonize() will
  291. * have disabled all of them by default).
  292. */
  293. int allow_signal(int sig)
  294. {
  295. if (!valid_signal(sig) || sig < 1)
  296. return -EINVAL;
  297. spin_lock_irq(&current->sighand->siglock);
  298. sigdelset(&current->blocked, sig);
  299. if (!current->mm) {
  300. /* Kernel threads handle their own signals.
  301. Let the signal code know it'll be handled, so
  302. that they don't get converted to SIGKILL or
  303. just silently dropped */
  304. current->sighand->action[(sig)-1].sa.sa_handler = (void __user *)2;
  305. }
  306. recalc_sigpending();
  307. spin_unlock_irq(&current->sighand->siglock);
  308. return 0;
  309. }
  310. EXPORT_SYMBOL(allow_signal);
  311. int disallow_signal(int sig)
  312. {
  313. if (!valid_signal(sig) || sig < 1)
  314. return -EINVAL;
  315. spin_lock_irq(&current->sighand->siglock);
  316. sigaddset(&current->blocked, sig);
  317. recalc_sigpending();
  318. spin_unlock_irq(&current->sighand->siglock);
  319. return 0;
  320. }
  321. EXPORT_SYMBOL(disallow_signal);
  322. /*
  323. * Put all the gunge required to become a kernel thread without
  324. * attached user resources in one place where it belongs.
  325. */
  326. void daemonize(const char *name, ...)
  327. {
  328. va_list args;
  329. struct fs_struct *fs;
  330. sigset_t blocked;
  331. va_start(args, name);
  332. vsnprintf(current->comm, sizeof(current->comm), name, args);
  333. va_end(args);
  334. /*
  335. * If we were started as result of loading a module, close all of the
  336. * user space pages. We don't need them, and if we didn't close them
  337. * they would be locked into memory.
  338. */
  339. exit_mm(current);
  340. set_special_pids(1, 1);
  341. proc_clear_tty(current);
  342. /* Block and flush all signals */
  343. sigfillset(&blocked);
  344. sigprocmask(SIG_BLOCK, &blocked, NULL);
  345. flush_signals(current);
  346. /* Become as one with the init task */
  347. exit_fs(current); /* current->fs->count--; */
  348. fs = init_task.fs;
  349. current->fs = fs;
  350. atomic_inc(&fs->count);
  351. exit_task_namespaces(current);
  352. current->nsproxy = init_task.nsproxy;
  353. get_task_namespaces(current);
  354. exit_files(current);
  355. current->files = init_task.files;
  356. atomic_inc(&current->files->count);
  357. reparent_to_init();
  358. }
  359. EXPORT_SYMBOL(daemonize);
  360. static void close_files(struct files_struct * files)
  361. {
  362. int i, j;
  363. struct fdtable *fdt;
  364. j = 0;
  365. /*
  366. * It is safe to dereference the fd table without RCU or
  367. * ->file_lock because this is the last reference to the
  368. * files structure.
  369. */
  370. fdt = files_fdtable(files);
  371. for (;;) {
  372. unsigned long set;
  373. i = j * __NFDBITS;
  374. if (i >= fdt->max_fdset || i >= fdt->max_fds)
  375. break;
  376. set = fdt->open_fds->fds_bits[j++];
  377. while (set) {
  378. if (set & 1) {
  379. struct file * file = xchg(&fdt->fd[i], NULL);
  380. if (file)
  381. filp_close(file, files);
  382. }
  383. i++;
  384. set >>= 1;
  385. }
  386. }
  387. }
  388. struct files_struct *get_files_struct(struct task_struct *task)
  389. {
  390. struct files_struct *files;
  391. task_lock(task);
  392. files = task->files;
  393. if (files)
  394. atomic_inc(&files->count);
  395. task_unlock(task);
  396. return files;
  397. }
  398. void fastcall put_files_struct(struct files_struct *files)
  399. {
  400. struct fdtable *fdt;
  401. if (atomic_dec_and_test(&files->count)) {
  402. close_files(files);
  403. /*
  404. * Free the fd and fdset arrays if we expanded them.
  405. * If the fdtable was embedded, pass files for freeing
  406. * at the end of the RCU grace period. Otherwise,
  407. * you can free files immediately.
  408. */
  409. fdt = files_fdtable(files);
  410. if (fdt == &files->fdtab)
  411. fdt->free_files = files;
  412. else
  413. kmem_cache_free(files_cachep, files);
  414. free_fdtable(fdt);
  415. }
  416. }
  417. EXPORT_SYMBOL(put_files_struct);
  418. void reset_files_struct(struct task_struct *tsk, struct files_struct *files)
  419. {
  420. struct files_struct *old;
  421. old = tsk->files;
  422. task_lock(tsk);
  423. tsk->files = files;
  424. task_unlock(tsk);
  425. put_files_struct(old);
  426. }
  427. EXPORT_SYMBOL(reset_files_struct);
  428. static inline void __exit_files(struct task_struct *tsk)
  429. {
  430. struct files_struct * files = tsk->files;
  431. if (files) {
  432. task_lock(tsk);
  433. tsk->files = NULL;
  434. task_unlock(tsk);
  435. put_files_struct(files);
  436. }
  437. }
  438. void exit_files(struct task_struct *tsk)
  439. {
  440. __exit_files(tsk);
  441. }
  442. static inline void __put_fs_struct(struct fs_struct *fs)
  443. {
  444. /* No need to hold fs->lock if we are killing it */
  445. if (atomic_dec_and_test(&fs->count)) {
  446. dput(fs->root);
  447. mntput(fs->rootmnt);
  448. dput(fs->pwd);
  449. mntput(fs->pwdmnt);
  450. if (fs->altroot) {
  451. dput(fs->altroot);
  452. mntput(fs->altrootmnt);
  453. }
  454. kmem_cache_free(fs_cachep, fs);
  455. }
  456. }
  457. void put_fs_struct(struct fs_struct *fs)
  458. {
  459. __put_fs_struct(fs);
  460. }
  461. static inline void __exit_fs(struct task_struct *tsk)
  462. {
  463. struct fs_struct * fs = tsk->fs;
  464. if (fs) {
  465. task_lock(tsk);
  466. tsk->fs = NULL;
  467. task_unlock(tsk);
  468. __put_fs_struct(fs);
  469. }
  470. }
  471. void exit_fs(struct task_struct *tsk)
  472. {
  473. __exit_fs(tsk);
  474. }
  475. EXPORT_SYMBOL_GPL(exit_fs);
  476. /*
  477. * Turn us into a lazy TLB process if we
  478. * aren't already..
  479. */
  480. static void exit_mm(struct task_struct * tsk)
  481. {
  482. struct mm_struct *mm = tsk->mm;
  483. mm_release(tsk, mm);
  484. if (!mm)
  485. return;
  486. /*
  487. * Serialize with any possible pending coredump.
  488. * We must hold mmap_sem around checking core_waiters
  489. * and clearing tsk->mm. The core-inducing thread
  490. * will increment core_waiters for each thread in the
  491. * group with ->mm != NULL.
  492. */
  493. down_read(&mm->mmap_sem);
  494. if (mm->core_waiters) {
  495. up_read(&mm->mmap_sem);
  496. down_write(&mm->mmap_sem);
  497. if (!--mm->core_waiters)
  498. complete(mm->core_startup_done);
  499. up_write(&mm->mmap_sem);
  500. wait_for_completion(&mm->core_done);
  501. down_read(&mm->mmap_sem);
  502. }
  503. atomic_inc(&mm->mm_count);
  504. BUG_ON(mm != tsk->active_mm);
  505. /* more a memory barrier than a real lock */
  506. task_lock(tsk);
  507. tsk->mm = NULL;
  508. up_read(&mm->mmap_sem);
  509. enter_lazy_tlb(mm, current);
  510. task_unlock(tsk);
  511. mmput(mm);
  512. }
  513. static inline void
  514. choose_new_parent(struct task_struct *p, struct task_struct *reaper)
  515. {
  516. /*
  517. * Make sure we're not reparenting to ourselves and that
  518. * the parent is not a zombie.
  519. */
  520. BUG_ON(p == reaper || reaper->exit_state);
  521. p->real_parent = reaper;
  522. }
  523. static void
  524. reparent_thread(struct task_struct *p, struct task_struct *father, int traced)
  525. {
  526. /* We don't want people slaying init. */
  527. if (p->exit_signal != -1)
  528. p->exit_signal = SIGCHLD;
  529. if (p->pdeath_signal)
  530. /* We already hold the tasklist_lock here. */
  531. group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p);
  532. /* Move the child from its dying parent to the new one. */
  533. if (unlikely(traced)) {
  534. /* Preserve ptrace links if someone else is tracing this child. */
  535. list_del_init(&p->ptrace_list);
  536. if (p->parent != p->real_parent)
  537. list_add(&p->ptrace_list, &p->real_parent->ptrace_children);
  538. } else {
  539. /* If this child is being traced, then we're the one tracing it
  540. * anyway, so let go of it.
  541. */
  542. p->ptrace = 0;
  543. remove_parent(p);
  544. p->parent = p->real_parent;
  545. add_parent(p);
  546. /* If we'd notified the old parent about this child's death,
  547. * also notify the new parent.
  548. */
  549. if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
  550. thread_group_empty(p))
  551. do_notify_parent(p, p->exit_signal);
  552. else if (p->state == TASK_TRACED) {
  553. /*
  554. * If it was at a trace stop, turn it into
  555. * a normal stop since it's no longer being
  556. * traced.
  557. */
  558. ptrace_untrace(p);
  559. }
  560. }
  561. /*
  562. * process group orphan check
  563. * Case ii: Our child is in a different pgrp
  564. * than we are, and it was the only connection
  565. * outside, so the child pgrp is now orphaned.
  566. */
  567. if ((process_group(p) != process_group(father)) &&
  568. (p->signal->session == father->signal->session)) {
  569. int pgrp = process_group(p);
  570. if (will_become_orphaned_pgrp(pgrp, NULL) && has_stopped_jobs(pgrp)) {
  571. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, pgrp);
  572. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, pgrp);
  573. }
  574. }
  575. }
  576. /*
  577. * When we die, we re-parent all our children.
  578. * Try to give them to another thread in our thread
  579. * group, and if no such member exists, give it to
  580. * the global child reaper process (ie "init")
  581. */
  582. static void
  583. forget_original_parent(struct task_struct *father, struct list_head *to_release)
  584. {
  585. struct task_struct *p, *reaper = father;
  586. struct list_head *_p, *_n;
  587. do {
  588. reaper = next_thread(reaper);
  589. if (reaper == father) {
  590. reaper = child_reaper;
  591. break;
  592. }
  593. } while (reaper->exit_state);
  594. /*
  595. * There are only two places where our children can be:
  596. *
  597. * - in our child list
  598. * - in our ptraced child list
  599. *
  600. * Search them and reparent children.
  601. */
  602. list_for_each_safe(_p, _n, &father->children) {
  603. int ptrace;
  604. p = list_entry(_p, struct task_struct, sibling);
  605. ptrace = p->ptrace;
  606. /* if father isn't the real parent, then ptrace must be enabled */
  607. BUG_ON(father != p->real_parent && !ptrace);
  608. if (father == p->real_parent) {
  609. /* reparent with a reaper, real father it's us */
  610. choose_new_parent(p, reaper);
  611. reparent_thread(p, father, 0);
  612. } else {
  613. /* reparent ptraced task to its real parent */
  614. __ptrace_unlink (p);
  615. if (p->exit_state == EXIT_ZOMBIE && p->exit_signal != -1 &&
  616. thread_group_empty(p))
  617. do_notify_parent(p, p->exit_signal);
  618. }
  619. /*
  620. * if the ptraced child is a zombie with exit_signal == -1
  621. * we must collect it before we exit, or it will remain
  622. * zombie forever since we prevented it from self-reap itself
  623. * while it was being traced by us, to be able to see it in wait4.
  624. */
  625. if (unlikely(ptrace && p->exit_state == EXIT_ZOMBIE && p->exit_signal == -1))
  626. list_add(&p->ptrace_list, to_release);
  627. }
  628. list_for_each_safe(_p, _n, &father->ptrace_children) {
  629. p = list_entry(_p, struct task_struct, ptrace_list);
  630. choose_new_parent(p, reaper);
  631. reparent_thread(p, father, 1);
  632. }
  633. }
  634. /*
  635. * Send signals to all our closest relatives so that they know
  636. * to properly mourn us..
  637. */
  638. static void exit_notify(struct task_struct *tsk)
  639. {
  640. int state;
  641. struct task_struct *t;
  642. struct list_head ptrace_dead, *_p, *_n;
  643. if (signal_pending(tsk) && !(tsk->signal->flags & SIGNAL_GROUP_EXIT)
  644. && !thread_group_empty(tsk)) {
  645. /*
  646. * This occurs when there was a race between our exit
  647. * syscall and a group signal choosing us as the one to
  648. * wake up. It could be that we are the only thread
  649. * alerted to check for pending signals, but another thread
  650. * should be woken now to take the signal since we will not.
  651. * Now we'll wake all the threads in the group just to make
  652. * sure someone gets all the pending signals.
  653. */
  654. read_lock(&tasklist_lock);
  655. spin_lock_irq(&tsk->sighand->siglock);
  656. for (t = next_thread(tsk); t != tsk; t = next_thread(t))
  657. if (!signal_pending(t) && !(t->flags & PF_EXITING)) {
  658. recalc_sigpending_tsk(t);
  659. if (signal_pending(t))
  660. signal_wake_up(t, 0);
  661. }
  662. spin_unlock_irq(&tsk->sighand->siglock);
  663. read_unlock(&tasklist_lock);
  664. }
  665. write_lock_irq(&tasklist_lock);
  666. /*
  667. * This does two things:
  668. *
  669. * A. Make init inherit all the child processes
  670. * B. Check to see if any process groups have become orphaned
  671. * as a result of our exiting, and if they have any stopped
  672. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  673. */
  674. INIT_LIST_HEAD(&ptrace_dead);
  675. forget_original_parent(tsk, &ptrace_dead);
  676. BUG_ON(!list_empty(&tsk->children));
  677. BUG_ON(!list_empty(&tsk->ptrace_children));
  678. /*
  679. * Check to see if any process groups have become orphaned
  680. * as a result of our exiting, and if they have any stopped
  681. * jobs, send them a SIGHUP and then a SIGCONT. (POSIX 3.2.2.2)
  682. *
  683. * Case i: Our father is in a different pgrp than we are
  684. * and we were the only connection outside, so our pgrp
  685. * is about to become orphaned.
  686. */
  687. t = tsk->real_parent;
  688. if ((process_group(t) != process_group(tsk)) &&
  689. (t->signal->session == tsk->signal->session) &&
  690. will_become_orphaned_pgrp(process_group(tsk), tsk) &&
  691. has_stopped_jobs(process_group(tsk))) {
  692. __kill_pg_info(SIGHUP, SEND_SIG_PRIV, process_group(tsk));
  693. __kill_pg_info(SIGCONT, SEND_SIG_PRIV, process_group(tsk));
  694. }
  695. /* Let father know we died
  696. *
  697. * Thread signals are configurable, but you aren't going to use
  698. * that to send signals to arbitary processes.
  699. * That stops right now.
  700. *
  701. * If the parent exec id doesn't match the exec id we saved
  702. * when we started then we know the parent has changed security
  703. * domain.
  704. *
  705. * If our self_exec id doesn't match our parent_exec_id then
  706. * we have changed execution domain as these two values started
  707. * the same after a fork.
  708. *
  709. */
  710. if (tsk->exit_signal != SIGCHLD && tsk->exit_signal != -1 &&
  711. ( tsk->parent_exec_id != t->self_exec_id ||
  712. tsk->self_exec_id != tsk->parent_exec_id)
  713. && !capable(CAP_KILL))
  714. tsk->exit_signal = SIGCHLD;
  715. /* If something other than our normal parent is ptracing us, then
  716. * send it a SIGCHLD instead of honoring exit_signal. exit_signal
  717. * only has special meaning to our real parent.
  718. */
  719. if (tsk->exit_signal != -1 && thread_group_empty(tsk)) {
  720. int signal = tsk->parent == tsk->real_parent ? tsk->exit_signal : SIGCHLD;
  721. do_notify_parent(tsk, signal);
  722. } else if (tsk->ptrace) {
  723. do_notify_parent(tsk, SIGCHLD);
  724. }
  725. state = EXIT_ZOMBIE;
  726. if (tsk->exit_signal == -1 &&
  727. (likely(tsk->ptrace == 0) ||
  728. unlikely(tsk->parent->signal->flags & SIGNAL_GROUP_EXIT)))
  729. state = EXIT_DEAD;
  730. tsk->exit_state = state;
  731. write_unlock_irq(&tasklist_lock);
  732. list_for_each_safe(_p, _n, &ptrace_dead) {
  733. list_del_init(_p);
  734. t = list_entry(_p, struct task_struct, ptrace_list);
  735. release_task(t);
  736. }
  737. /* If the process is dead, release it - nobody will wait for it */
  738. if (state == EXIT_DEAD)
  739. release_task(tsk);
  740. }
  741. fastcall NORET_TYPE void do_exit(long code)
  742. {
  743. struct task_struct *tsk = current;
  744. int group_dead;
  745. profile_task_exit(tsk);
  746. WARN_ON(atomic_read(&tsk->fs_excl));
  747. if (unlikely(in_interrupt()))
  748. panic("Aiee, killing interrupt handler!");
  749. if (unlikely(!tsk->pid))
  750. panic("Attempted to kill the idle task!");
  751. if (unlikely(tsk == child_reaper))
  752. panic("Attempted to kill init!");
  753. if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
  754. current->ptrace_message = code;
  755. ptrace_notify((PTRACE_EVENT_EXIT << 8) | SIGTRAP);
  756. }
  757. /*
  758. * We're taking recursive faults here in do_exit. Safest is to just
  759. * leave this task alone and wait for reboot.
  760. */
  761. if (unlikely(tsk->flags & PF_EXITING)) {
  762. printk(KERN_ALERT
  763. "Fixing recursive fault but reboot is needed!\n");
  764. if (tsk->io_context)
  765. exit_io_context();
  766. set_current_state(TASK_UNINTERRUPTIBLE);
  767. schedule();
  768. }
  769. tsk->flags |= PF_EXITING;
  770. if (unlikely(in_atomic()))
  771. printk(KERN_INFO "note: %s[%d] exited with preempt_count %d\n",
  772. current->comm, current->pid,
  773. preempt_count());
  774. acct_update_integrals(tsk);
  775. if (tsk->mm) {
  776. update_hiwater_rss(tsk->mm);
  777. update_hiwater_vm(tsk->mm);
  778. }
  779. group_dead = atomic_dec_and_test(&tsk->signal->live);
  780. if (group_dead) {
  781. hrtimer_cancel(&tsk->signal->real_timer);
  782. exit_itimers(tsk->signal);
  783. }
  784. acct_collect(code, group_dead);
  785. if (unlikely(tsk->robust_list))
  786. exit_robust_list(tsk);
  787. #if defined(CONFIG_FUTEX) && defined(CONFIG_COMPAT)
  788. if (unlikely(tsk->compat_robust_list))
  789. compat_exit_robust_list(tsk);
  790. #endif
  791. if (unlikely(tsk->audit_context))
  792. audit_free(tsk);
  793. taskstats_exit(tsk, group_dead);
  794. exit_mm(tsk);
  795. if (group_dead)
  796. acct_process();
  797. exit_sem(tsk);
  798. __exit_files(tsk);
  799. __exit_fs(tsk);
  800. exit_thread();
  801. cpuset_exit(tsk);
  802. exit_keys(tsk);
  803. if (group_dead && tsk->signal->leader)
  804. disassociate_ctty(1);
  805. module_put(task_thread_info(tsk)->exec_domain->module);
  806. if (tsk->binfmt)
  807. module_put(tsk->binfmt->module);
  808. tsk->exit_code = code;
  809. proc_exit_connector(tsk);
  810. exit_notify(tsk);
  811. exit_task_namespaces(tsk);
  812. #ifdef CONFIG_NUMA
  813. mpol_free(tsk->mempolicy);
  814. tsk->mempolicy = NULL;
  815. #endif
  816. /*
  817. * This must happen late, after the PID is not
  818. * hashed anymore:
  819. */
  820. if (unlikely(!list_empty(&tsk->pi_state_list)))
  821. exit_pi_state_list(tsk);
  822. if (unlikely(current->pi_state_cache))
  823. kfree(current->pi_state_cache);
  824. /*
  825. * Make sure we are holding no locks:
  826. */
  827. debug_check_no_locks_held(tsk);
  828. if (tsk->io_context)
  829. exit_io_context();
  830. if (tsk->splice_pipe)
  831. __free_pipe_info(tsk->splice_pipe);
  832. preempt_disable();
  833. /* causes final put_task_struct in finish_task_switch(). */
  834. tsk->state = TASK_DEAD;
  835. schedule();
  836. BUG();
  837. /* Avoid "noreturn function does return". */
  838. for (;;)
  839. cpu_relax(); /* For when BUG is null */
  840. }
  841. EXPORT_SYMBOL_GPL(do_exit);
  842. NORET_TYPE void complete_and_exit(struct completion *comp, long code)
  843. {
  844. if (comp)
  845. complete(comp);
  846. do_exit(code);
  847. }
  848. EXPORT_SYMBOL(complete_and_exit);
  849. asmlinkage long sys_exit(int error_code)
  850. {
  851. do_exit((error_code&0xff)<<8);
  852. }
  853. /*
  854. * Take down every thread in the group. This is called by fatal signals
  855. * as well as by sys_exit_group (below).
  856. */
  857. NORET_TYPE void
  858. do_group_exit(int exit_code)
  859. {
  860. BUG_ON(exit_code & 0x80); /* core dumps don't get here */
  861. if (current->signal->flags & SIGNAL_GROUP_EXIT)
  862. exit_code = current->signal->group_exit_code;
  863. else if (!thread_group_empty(current)) {
  864. struct signal_struct *const sig = current->signal;
  865. struct sighand_struct *const sighand = current->sighand;
  866. spin_lock_irq(&sighand->siglock);
  867. if (sig->flags & SIGNAL_GROUP_EXIT)
  868. /* Another thread got here before we took the lock. */
  869. exit_code = sig->group_exit_code;
  870. else {
  871. sig->group_exit_code = exit_code;
  872. zap_other_threads(current);
  873. }
  874. spin_unlock_irq(&sighand->siglock);
  875. }
  876. do_exit(exit_code);
  877. /* NOTREACHED */
  878. }
  879. /*
  880. * this kills every thread in the thread group. Note that any externally
  881. * wait4()-ing process will get the correct exit code - even if this
  882. * thread is not the thread group leader.
  883. */
  884. asmlinkage void sys_exit_group(int error_code)
  885. {
  886. do_group_exit((error_code & 0xff) << 8);
  887. }
  888. static int eligible_child(pid_t pid, int options, struct task_struct *p)
  889. {
  890. if (pid > 0) {
  891. if (p->pid != pid)
  892. return 0;
  893. } else if (!pid) {
  894. if (process_group(p) != process_group(current))
  895. return 0;
  896. } else if (pid != -1) {
  897. if (process_group(p) != -pid)
  898. return 0;
  899. }
  900. /*
  901. * Do not consider detached threads that are
  902. * not ptraced:
  903. */
  904. if (p->exit_signal == -1 && !p->ptrace)
  905. return 0;
  906. /* Wait for all children (clone and not) if __WALL is set;
  907. * otherwise, wait for clone children *only* if __WCLONE is
  908. * set; otherwise, wait for non-clone children *only*. (Note:
  909. * A "clone" child here is one that reports to its parent
  910. * using a signal other than SIGCHLD.) */
  911. if (((p->exit_signal != SIGCHLD) ^ ((options & __WCLONE) != 0))
  912. && !(options & __WALL))
  913. return 0;
  914. /*
  915. * Do not consider thread group leaders that are
  916. * in a non-empty thread group:
  917. */
  918. if (delay_group_leader(p))
  919. return 2;
  920. if (security_task_wait(p))
  921. return 0;
  922. return 1;
  923. }
  924. static int wait_noreap_copyout(struct task_struct *p, pid_t pid, uid_t uid,
  925. int why, int status,
  926. struct siginfo __user *infop,
  927. struct rusage __user *rusagep)
  928. {
  929. int retval = rusagep ? getrusage(p, RUSAGE_BOTH, rusagep) : 0;
  930. put_task_struct(p);
  931. if (!retval)
  932. retval = put_user(SIGCHLD, &infop->si_signo);
  933. if (!retval)
  934. retval = put_user(0, &infop->si_errno);
  935. if (!retval)
  936. retval = put_user((short)why, &infop->si_code);
  937. if (!retval)
  938. retval = put_user(pid, &infop->si_pid);
  939. if (!retval)
  940. retval = put_user(uid, &infop->si_uid);
  941. if (!retval)
  942. retval = put_user(status, &infop->si_status);
  943. if (!retval)
  944. retval = pid;
  945. return retval;
  946. }
  947. /*
  948. * Handle sys_wait4 work for one task in state EXIT_ZOMBIE. We hold
  949. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  950. * the lock and this task is uninteresting. If we return nonzero, we have
  951. * released the lock and the system call should return.
  952. */
  953. static int wait_task_zombie(struct task_struct *p, int noreap,
  954. struct siginfo __user *infop,
  955. int __user *stat_addr, struct rusage __user *ru)
  956. {
  957. unsigned long state;
  958. int retval;
  959. int status;
  960. if (unlikely(noreap)) {
  961. pid_t pid = p->pid;
  962. uid_t uid = p->uid;
  963. int exit_code = p->exit_code;
  964. int why, status;
  965. if (unlikely(p->exit_state != EXIT_ZOMBIE))
  966. return 0;
  967. if (unlikely(p->exit_signal == -1 && p->ptrace == 0))
  968. return 0;
  969. get_task_struct(p);
  970. read_unlock(&tasklist_lock);
  971. if ((exit_code & 0x7f) == 0) {
  972. why = CLD_EXITED;
  973. status = exit_code >> 8;
  974. } else {
  975. why = (exit_code & 0x80) ? CLD_DUMPED : CLD_KILLED;
  976. status = exit_code & 0x7f;
  977. }
  978. return wait_noreap_copyout(p, pid, uid, why,
  979. status, infop, ru);
  980. }
  981. /*
  982. * Try to move the task's state to DEAD
  983. * only one thread is allowed to do this:
  984. */
  985. state = xchg(&p->exit_state, EXIT_DEAD);
  986. if (state != EXIT_ZOMBIE) {
  987. BUG_ON(state != EXIT_DEAD);
  988. return 0;
  989. }
  990. if (unlikely(p->exit_signal == -1 && p->ptrace == 0)) {
  991. /*
  992. * This can only happen in a race with a ptraced thread
  993. * dying on another processor.
  994. */
  995. return 0;
  996. }
  997. if (likely(p->real_parent == p->parent) && likely(p->signal)) {
  998. struct signal_struct *psig;
  999. struct signal_struct *sig;
  1000. /*
  1001. * The resource counters for the group leader are in its
  1002. * own task_struct. Those for dead threads in the group
  1003. * are in its signal_struct, as are those for the child
  1004. * processes it has previously reaped. All these
  1005. * accumulate in the parent's signal_struct c* fields.
  1006. *
  1007. * We don't bother to take a lock here to protect these
  1008. * p->signal fields, because they are only touched by
  1009. * __exit_signal, which runs with tasklist_lock
  1010. * write-locked anyway, and so is excluded here. We do
  1011. * need to protect the access to p->parent->signal fields,
  1012. * as other threads in the parent group can be right
  1013. * here reaping other children at the same time.
  1014. */
  1015. spin_lock_irq(&p->parent->sighand->siglock);
  1016. psig = p->parent->signal;
  1017. sig = p->signal;
  1018. psig->cutime =
  1019. cputime_add(psig->cutime,
  1020. cputime_add(p->utime,
  1021. cputime_add(sig->utime,
  1022. sig->cutime)));
  1023. psig->cstime =
  1024. cputime_add(psig->cstime,
  1025. cputime_add(p->stime,
  1026. cputime_add(sig->stime,
  1027. sig->cstime)));
  1028. psig->cmin_flt +=
  1029. p->min_flt + sig->min_flt + sig->cmin_flt;
  1030. psig->cmaj_flt +=
  1031. p->maj_flt + sig->maj_flt + sig->cmaj_flt;
  1032. psig->cnvcsw +=
  1033. p->nvcsw + sig->nvcsw + sig->cnvcsw;
  1034. psig->cnivcsw +=
  1035. p->nivcsw + sig->nivcsw + sig->cnivcsw;
  1036. spin_unlock_irq(&p->parent->sighand->siglock);
  1037. }
  1038. /*
  1039. * Now we are sure this task is interesting, and no other
  1040. * thread can reap it because we set its state to EXIT_DEAD.
  1041. */
  1042. read_unlock(&tasklist_lock);
  1043. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1044. status = (p->signal->flags & SIGNAL_GROUP_EXIT)
  1045. ? p->signal->group_exit_code : p->exit_code;
  1046. if (!retval && stat_addr)
  1047. retval = put_user(status, stat_addr);
  1048. if (!retval && infop)
  1049. retval = put_user(SIGCHLD, &infop->si_signo);
  1050. if (!retval && infop)
  1051. retval = put_user(0, &infop->si_errno);
  1052. if (!retval && infop) {
  1053. int why;
  1054. if ((status & 0x7f) == 0) {
  1055. why = CLD_EXITED;
  1056. status >>= 8;
  1057. } else {
  1058. why = (status & 0x80) ? CLD_DUMPED : CLD_KILLED;
  1059. status &= 0x7f;
  1060. }
  1061. retval = put_user((short)why, &infop->si_code);
  1062. if (!retval)
  1063. retval = put_user(status, &infop->si_status);
  1064. }
  1065. if (!retval && infop)
  1066. retval = put_user(p->pid, &infop->si_pid);
  1067. if (!retval && infop)
  1068. retval = put_user(p->uid, &infop->si_uid);
  1069. if (retval) {
  1070. // TODO: is this safe?
  1071. p->exit_state = EXIT_ZOMBIE;
  1072. return retval;
  1073. }
  1074. retval = p->pid;
  1075. if (p->real_parent != p->parent) {
  1076. write_lock_irq(&tasklist_lock);
  1077. /* Double-check with lock held. */
  1078. if (p->real_parent != p->parent) {
  1079. __ptrace_unlink(p);
  1080. // TODO: is this safe?
  1081. p->exit_state = EXIT_ZOMBIE;
  1082. /*
  1083. * If this is not a detached task, notify the parent.
  1084. * If it's still not detached after that, don't release
  1085. * it now.
  1086. */
  1087. if (p->exit_signal != -1) {
  1088. do_notify_parent(p, p->exit_signal);
  1089. if (p->exit_signal != -1)
  1090. p = NULL;
  1091. }
  1092. }
  1093. write_unlock_irq(&tasklist_lock);
  1094. }
  1095. if (p != NULL)
  1096. release_task(p);
  1097. BUG_ON(!retval);
  1098. return retval;
  1099. }
  1100. /*
  1101. * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold
  1102. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1103. * the lock and this task is uninteresting. If we return nonzero, we have
  1104. * released the lock and the system call should return.
  1105. */
  1106. static int wait_task_stopped(struct task_struct *p, int delayed_group_leader,
  1107. int noreap, struct siginfo __user *infop,
  1108. int __user *stat_addr, struct rusage __user *ru)
  1109. {
  1110. int retval, exit_code;
  1111. if (!p->exit_code)
  1112. return 0;
  1113. if (delayed_group_leader && !(p->ptrace & PT_PTRACED) &&
  1114. p->signal && p->signal->group_stop_count > 0)
  1115. /*
  1116. * A group stop is in progress and this is the group leader.
  1117. * We won't report until all threads have stopped.
  1118. */
  1119. return 0;
  1120. /*
  1121. * Now we are pretty sure this task is interesting.
  1122. * Make sure it doesn't get reaped out from under us while we
  1123. * give up the lock and then examine it below. We don't want to
  1124. * keep holding onto the tasklist_lock while we call getrusage and
  1125. * possibly take page faults for user memory.
  1126. */
  1127. get_task_struct(p);
  1128. read_unlock(&tasklist_lock);
  1129. if (unlikely(noreap)) {
  1130. pid_t pid = p->pid;
  1131. uid_t uid = p->uid;
  1132. int why = (p->ptrace & PT_PTRACED) ? CLD_TRAPPED : CLD_STOPPED;
  1133. exit_code = p->exit_code;
  1134. if (unlikely(!exit_code) ||
  1135. unlikely(p->state & TASK_TRACED))
  1136. goto bail_ref;
  1137. return wait_noreap_copyout(p, pid, uid,
  1138. why, (exit_code << 8) | 0x7f,
  1139. infop, ru);
  1140. }
  1141. write_lock_irq(&tasklist_lock);
  1142. /*
  1143. * This uses xchg to be atomic with the thread resuming and setting
  1144. * it. It must also be done with the write lock held to prevent a
  1145. * race with the EXIT_ZOMBIE case.
  1146. */
  1147. exit_code = xchg(&p->exit_code, 0);
  1148. if (unlikely(p->exit_state)) {
  1149. /*
  1150. * The task resumed and then died. Let the next iteration
  1151. * catch it in EXIT_ZOMBIE. Note that exit_code might
  1152. * already be zero here if it resumed and did _exit(0).
  1153. * The task itself is dead and won't touch exit_code again;
  1154. * other processors in this function are locked out.
  1155. */
  1156. p->exit_code = exit_code;
  1157. exit_code = 0;
  1158. }
  1159. if (unlikely(exit_code == 0)) {
  1160. /*
  1161. * Another thread in this function got to it first, or it
  1162. * resumed, or it resumed and then died.
  1163. */
  1164. write_unlock_irq(&tasklist_lock);
  1165. bail_ref:
  1166. put_task_struct(p);
  1167. /*
  1168. * We are returning to the wait loop without having successfully
  1169. * removed the process and having released the lock. We cannot
  1170. * continue, since the "p" task pointer is potentially stale.
  1171. *
  1172. * Return -EAGAIN, and do_wait() will restart the loop from the
  1173. * beginning. Do _not_ re-acquire the lock.
  1174. */
  1175. return -EAGAIN;
  1176. }
  1177. /* move to end of parent's list to avoid starvation */
  1178. remove_parent(p);
  1179. add_parent(p);
  1180. write_unlock_irq(&tasklist_lock);
  1181. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1182. if (!retval && stat_addr)
  1183. retval = put_user((exit_code << 8) | 0x7f, stat_addr);
  1184. if (!retval && infop)
  1185. retval = put_user(SIGCHLD, &infop->si_signo);
  1186. if (!retval && infop)
  1187. retval = put_user(0, &infop->si_errno);
  1188. if (!retval && infop)
  1189. retval = put_user((short)((p->ptrace & PT_PTRACED)
  1190. ? CLD_TRAPPED : CLD_STOPPED),
  1191. &infop->si_code);
  1192. if (!retval && infop)
  1193. retval = put_user(exit_code, &infop->si_status);
  1194. if (!retval && infop)
  1195. retval = put_user(p->pid, &infop->si_pid);
  1196. if (!retval && infop)
  1197. retval = put_user(p->uid, &infop->si_uid);
  1198. if (!retval)
  1199. retval = p->pid;
  1200. put_task_struct(p);
  1201. BUG_ON(!retval);
  1202. return retval;
  1203. }
  1204. /*
  1205. * Handle do_wait work for one task in a live, non-stopped state.
  1206. * read_lock(&tasklist_lock) on entry. If we return zero, we still hold
  1207. * the lock and this task is uninteresting. If we return nonzero, we have
  1208. * released the lock and the system call should return.
  1209. */
  1210. static int wait_task_continued(struct task_struct *p, int noreap,
  1211. struct siginfo __user *infop,
  1212. int __user *stat_addr, struct rusage __user *ru)
  1213. {
  1214. int retval;
  1215. pid_t pid;
  1216. uid_t uid;
  1217. if (unlikely(!p->signal))
  1218. return 0;
  1219. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED))
  1220. return 0;
  1221. spin_lock_irq(&p->sighand->siglock);
  1222. /* Re-check with the lock held. */
  1223. if (!(p->signal->flags & SIGNAL_STOP_CONTINUED)) {
  1224. spin_unlock_irq(&p->sighand->siglock);
  1225. return 0;
  1226. }
  1227. if (!noreap)
  1228. p->signal->flags &= ~SIGNAL_STOP_CONTINUED;
  1229. spin_unlock_irq(&p->sighand->siglock);
  1230. pid = p->pid;
  1231. uid = p->uid;
  1232. get_task_struct(p);
  1233. read_unlock(&tasklist_lock);
  1234. if (!infop) {
  1235. retval = ru ? getrusage(p, RUSAGE_BOTH, ru) : 0;
  1236. put_task_struct(p);
  1237. if (!retval && stat_addr)
  1238. retval = put_user(0xffff, stat_addr);
  1239. if (!retval)
  1240. retval = p->pid;
  1241. } else {
  1242. retval = wait_noreap_copyout(p, pid, uid,
  1243. CLD_CONTINUED, SIGCONT,
  1244. infop, ru);
  1245. BUG_ON(retval == 0);
  1246. }
  1247. return retval;
  1248. }
  1249. static inline int my_ptrace_child(struct task_struct *p)
  1250. {
  1251. if (!(p->ptrace & PT_PTRACED))
  1252. return 0;
  1253. if (!(p->ptrace & PT_ATTACHED))
  1254. return 1;
  1255. /*
  1256. * This child was PTRACE_ATTACH'd. We should be seeing it only if
  1257. * we are the attacher. If we are the real parent, this is a race
  1258. * inside ptrace_attach. It is waiting for the tasklist_lock,
  1259. * which we have to switch the parent links, but has already set
  1260. * the flags in p->ptrace.
  1261. */
  1262. return (p->parent != p->real_parent);
  1263. }
  1264. static long do_wait(pid_t pid, int options, struct siginfo __user *infop,
  1265. int __user *stat_addr, struct rusage __user *ru)
  1266. {
  1267. DECLARE_WAITQUEUE(wait, current);
  1268. struct task_struct *tsk;
  1269. int flag, retval;
  1270. add_wait_queue(&current->signal->wait_chldexit,&wait);
  1271. repeat:
  1272. /*
  1273. * We will set this flag if we see any child that might later
  1274. * match our criteria, even if we are not able to reap it yet.
  1275. */
  1276. flag = 0;
  1277. current->state = TASK_INTERRUPTIBLE;
  1278. read_lock(&tasklist_lock);
  1279. tsk = current;
  1280. do {
  1281. struct task_struct *p;
  1282. struct list_head *_p;
  1283. int ret;
  1284. list_for_each(_p,&tsk->children) {
  1285. p = list_entry(_p, struct task_struct, sibling);
  1286. ret = eligible_child(pid, options, p);
  1287. if (!ret)
  1288. continue;
  1289. switch (p->state) {
  1290. case TASK_TRACED:
  1291. /*
  1292. * When we hit the race with PTRACE_ATTACH,
  1293. * we will not report this child. But the
  1294. * race means it has not yet been moved to
  1295. * our ptrace_children list, so we need to
  1296. * set the flag here to avoid a spurious ECHILD
  1297. * when the race happens with the only child.
  1298. */
  1299. flag = 1;
  1300. if (!my_ptrace_child(p))
  1301. continue;
  1302. /*FALLTHROUGH*/
  1303. case TASK_STOPPED:
  1304. /*
  1305. * It's stopped now, so it might later
  1306. * continue, exit, or stop again.
  1307. */
  1308. flag = 1;
  1309. if (!(options & WUNTRACED) &&
  1310. !my_ptrace_child(p))
  1311. continue;
  1312. retval = wait_task_stopped(p, ret == 2,
  1313. (options & WNOWAIT),
  1314. infop,
  1315. stat_addr, ru);
  1316. if (retval == -EAGAIN)
  1317. goto repeat;
  1318. if (retval != 0) /* He released the lock. */
  1319. goto end;
  1320. break;
  1321. default:
  1322. // case EXIT_DEAD:
  1323. if (p->exit_state == EXIT_DEAD)
  1324. continue;
  1325. // case EXIT_ZOMBIE:
  1326. if (p->exit_state == EXIT_ZOMBIE) {
  1327. /*
  1328. * Eligible but we cannot release
  1329. * it yet:
  1330. */
  1331. if (ret == 2)
  1332. goto check_continued;
  1333. if (!likely(options & WEXITED))
  1334. continue;
  1335. retval = wait_task_zombie(
  1336. p, (options & WNOWAIT),
  1337. infop, stat_addr, ru);
  1338. /* He released the lock. */
  1339. if (retval != 0)
  1340. goto end;
  1341. break;
  1342. }
  1343. check_continued:
  1344. /*
  1345. * It's running now, so it might later
  1346. * exit, stop, or stop and then continue.
  1347. */
  1348. flag = 1;
  1349. if (!unlikely(options & WCONTINUED))
  1350. continue;
  1351. retval = wait_task_continued(
  1352. p, (options & WNOWAIT),
  1353. infop, stat_addr, ru);
  1354. if (retval != 0) /* He released the lock. */
  1355. goto end;
  1356. break;
  1357. }
  1358. }
  1359. if (!flag) {
  1360. list_for_each(_p, &tsk->ptrace_children) {
  1361. p = list_entry(_p, struct task_struct,
  1362. ptrace_list);
  1363. if (!eligible_child(pid, options, p))
  1364. continue;
  1365. flag = 1;
  1366. break;
  1367. }
  1368. }
  1369. if (options & __WNOTHREAD)
  1370. break;
  1371. tsk = next_thread(tsk);
  1372. BUG_ON(tsk->signal != current->signal);
  1373. } while (tsk != current);
  1374. read_unlock(&tasklist_lock);
  1375. if (flag) {
  1376. retval = 0;
  1377. if (options & WNOHANG)
  1378. goto end;
  1379. retval = -ERESTARTSYS;
  1380. if (signal_pending(current))
  1381. goto end;
  1382. schedule();
  1383. goto repeat;
  1384. }
  1385. retval = -ECHILD;
  1386. end:
  1387. current->state = TASK_RUNNING;
  1388. remove_wait_queue(&current->signal->wait_chldexit,&wait);
  1389. if (infop) {
  1390. if (retval > 0)
  1391. retval = 0;
  1392. else {
  1393. /*
  1394. * For a WNOHANG return, clear out all the fields
  1395. * we would set so the user can easily tell the
  1396. * difference.
  1397. */
  1398. if (!retval)
  1399. retval = put_user(0, &infop->si_signo);
  1400. if (!retval)
  1401. retval = put_user(0, &infop->si_errno);
  1402. if (!retval)
  1403. retval = put_user(0, &infop->si_code);
  1404. if (!retval)
  1405. retval = put_user(0, &infop->si_pid);
  1406. if (!retval)
  1407. retval = put_user(0, &infop->si_uid);
  1408. if (!retval)
  1409. retval = put_user(0, &infop->si_status);
  1410. }
  1411. }
  1412. return retval;
  1413. }
  1414. asmlinkage long sys_waitid(int which, pid_t pid,
  1415. struct siginfo __user *infop, int options,
  1416. struct rusage __user *ru)
  1417. {
  1418. long ret;
  1419. if (options & ~(WNOHANG|WNOWAIT|WEXITED|WSTOPPED|WCONTINUED))
  1420. return -EINVAL;
  1421. if (!(options & (WEXITED|WSTOPPED|WCONTINUED)))
  1422. return -EINVAL;
  1423. switch (which) {
  1424. case P_ALL:
  1425. pid = -1;
  1426. break;
  1427. case P_PID:
  1428. if (pid <= 0)
  1429. return -EINVAL;
  1430. break;
  1431. case P_PGID:
  1432. if (pid <= 0)
  1433. return -EINVAL;
  1434. pid = -pid;
  1435. break;
  1436. default:
  1437. return -EINVAL;
  1438. }
  1439. ret = do_wait(pid, options, infop, NULL, ru);
  1440. /* avoid REGPARM breakage on x86: */
  1441. prevent_tail_call(ret);
  1442. return ret;
  1443. }
  1444. asmlinkage long sys_wait4(pid_t pid, int __user *stat_addr,
  1445. int options, struct rusage __user *ru)
  1446. {
  1447. long ret;
  1448. if (options & ~(WNOHANG|WUNTRACED|WCONTINUED|
  1449. __WNOTHREAD|__WCLONE|__WALL))
  1450. return -EINVAL;
  1451. ret = do_wait(pid, options | WEXITED, NULL, stat_addr, ru);
  1452. /* avoid REGPARM breakage on x86: */
  1453. prevent_tail_call(ret);
  1454. return ret;
  1455. }
  1456. #ifdef __ARCH_WANT_SYS_WAITPID
  1457. /*
  1458. * sys_waitpid() remains for compatibility. waitpid() should be
  1459. * implemented by calling sys_wait4() from libc.a.
  1460. */
  1461. asmlinkage long sys_waitpid(pid_t pid, int __user *stat_addr, int options)
  1462. {
  1463. return sys_wait4(pid, stat_addr, options, NULL);
  1464. }
  1465. #endif