]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (c) 2000-2016 Apple Inc. All rights reserved. | |
3 | * | |
4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ | |
5 | * | |
6 | * This file contains Original Code and/or Modifications of Original Code | |
7 | * as defined in and that are subject to the Apple Public Source License | |
8 | * Version 2.0 (the 'License'). You may not use this file except in | |
9 | * compliance with the License. The rights granted to you under the License | |
10 | * may not be used to create, or enable the creation or redistribution of, | |
11 | * unlawful or unlicensed copies of an Apple operating system, or to | |
12 | * circumvent, violate, or enable the circumvention or violation of, any | |
13 | * terms of an Apple operating system software license agreement. | |
14 | * | |
15 | * Please obtain a copy of the License at | |
16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. | |
17 | * | |
18 | * The Original Code and all software distributed under the License are | |
19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER | |
20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, | |
21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, | |
22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. | |
23 | * Please see the License for the specific language governing rights and | |
24 | * limitations under the License. | |
25 | * | |
26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ | |
27 | */ | |
28 | /* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */ | |
29 | /* | |
30 | * Copyright (c) 1982, 1986, 1989, 1991, 1993 | |
31 | * The Regents of the University of California. All rights reserved. | |
32 | * (c) UNIX System Laboratories, Inc. | |
33 | * All or some portions of this file are derived from material licensed | |
34 | * to the University of California by American Telephone and Telegraph | |
35 | * Co. or Unix System Laboratories, Inc. and are reproduced herein with | |
36 | * the permission of UNIX System Laboratories, Inc. | |
37 | * | |
38 | * Redistribution and use in source and binary forms, with or without | |
39 | * modification, are permitted provided that the following conditions | |
40 | * are met: | |
41 | * 1. Redistributions of source code must retain the above copyright | |
42 | * notice, this list of conditions and the following disclaimer. | |
43 | * 2. Redistributions in binary form must reproduce the above copyright | |
44 | * notice, this list of conditions and the following disclaimer in the | |
45 | * documentation and/or other materials provided with the distribution. | |
46 | * 3. All advertising materials mentioning features or use of this software | |
47 | * must display the following acknowledgement: | |
48 | * This product includes software developed by the University of | |
49 | * California, Berkeley and its contributors. | |
50 | * 4. Neither the name of the University nor the names of its contributors | |
51 | * may be used to endorse or promote products derived from this software | |
52 | * without specific prior written permission. | |
53 | * | |
54 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | |
55 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
56 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
57 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
58 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | |
59 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | |
60 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | |
61 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
62 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | |
63 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
64 | * SUCH DAMAGE. | |
65 | * | |
66 | * @(#)kern_exit.c 8.7 (Berkeley) 2/12/94 | |
67 | */ | |
68 | /* | |
69 | * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce | |
70 | * support for mandatory and extensible security protections. This notice | |
71 | * is included in support of clause 2.2 (b) of the Apple Public License, | |
72 | * Version 2.0. | |
73 | */ | |
74 | ||
75 | #include <machine/reg.h> | |
76 | #include <machine/psl.h> | |
77 | #include <stdatomic.h> | |
78 | ||
79 | #include "compat_43.h" | |
80 | ||
81 | #include <sys/param.h> | |
82 | #include <sys/systm.h> | |
83 | #include <sys/ioctl.h> | |
84 | #include <sys/proc_internal.h> | |
85 | #include <sys/proc.h> | |
86 | #include <sys/kauth.h> | |
87 | #include <sys/tty.h> | |
88 | #include <sys/time.h> | |
89 | #include <sys/resource.h> | |
90 | #include <sys/kernel.h> | |
91 | #include <sys/wait.h> | |
92 | #include <sys/file_internal.h> | |
93 | #include <sys/vnode_internal.h> | |
94 | #include <sys/syslog.h> | |
95 | #include <sys/malloc.h> | |
96 | #include <sys/resourcevar.h> | |
97 | #include <sys/ptrace.h> | |
98 | #include <sys/proc_info.h> | |
99 | #include <sys/reason.h> | |
100 | #include <sys/_types/_timeval64.h> | |
101 | #include <sys/user.h> | |
102 | #include <sys/aio_kern.h> | |
103 | #include <sys/sysproto.h> | |
104 | #include <sys/signalvar.h> | |
105 | #include <sys/kdebug.h> | |
106 | #include <sys/filedesc.h> /* fdfree */ | |
107 | #include <sys/acct.h> /* acct_process */ | |
108 | #include <sys/codesign.h> | |
109 | #include <sys/event.h> /* kevent_proc_copy_uptrs */ | |
110 | #include <sys/sdt.h> | |
111 | ||
112 | #include <security/audit/audit.h> | |
113 | #include <bsm/audit_kevents.h> | |
114 | ||
115 | #include <mach/mach_types.h> | |
116 | #include <mach/task.h> | |
117 | #include <mach/thread_act.h> | |
118 | ||
119 | #include <kern/exc_resource.h> | |
120 | #include <kern/kern_types.h> | |
121 | #include <kern/kalloc.h> | |
122 | #include <kern/task.h> | |
123 | #include <corpses/task_corpse.h> | |
124 | #include <kern/thread.h> | |
125 | #include <kern/thread_call.h> | |
126 | #include <kern/sched_prim.h> | |
127 | #include <kern/assert.h> | |
128 | #include <kern/policy_internal.h> | |
129 | #include <kern/exc_guard.h> | |
130 | ||
131 | #include <vm/vm_protos.h> | |
132 | #include <os/log.h> | |
133 | ||
134 | #include <pexpert/pexpert.h> | |
135 | ||
136 | #if SYSV_SHM | |
137 | #include <sys/shm_internal.h> /* shmexit */ | |
138 | #endif /* SYSV_SHM */ | |
139 | #if CONFIG_PERSONAS | |
140 | #include <sys/persona.h> | |
141 | #endif /* CONFIG_PERSONAS */ | |
142 | #if CONFIG_MEMORYSTATUS | |
143 | #include <sys/kern_memorystatus.h> | |
144 | #endif /* CONFIG_MEMORYSTATUS */ | |
145 | #if CONFIG_DTRACE | |
146 | /* Do not include dtrace.h, it redefines kmem_[alloc/free] */ | |
147 | void dtrace_proc_exit(proc_t p); | |
148 | #include <sys/dtrace_ptss.h> | |
149 | #endif /* CONFIG_DTRACE */ | |
150 | #if CONFIG_MACF | |
151 | #include <security/mac_framework.h> | |
152 | #include <security/mac_mach_internal.h> | |
153 | #include <sys/syscall.h> | |
154 | #endif /* CONFIG_MACF */ | |
155 | ||
156 | #if CONFIG_MEMORYSTATUS | |
157 | static void proc_memorystatus_remove(proc_t p); | |
158 | #endif /* CONFIG_MEMORYSTATUS */ | |
159 | void proc_prepareexit(proc_t p, int rv, boolean_t perf_notify); | |
160 | void gather_populate_corpse_crashinfo(proc_t p, task_t corpse_task, | |
161 | mach_exception_data_type_t code, mach_exception_data_type_t subcode, | |
162 | uint64_t *udata_buffer, int num_udata, void *reason); | |
163 | mach_exception_data_type_t proc_encode_exit_exception_code(proc_t p); | |
164 | void vfork_exit(proc_t p, int rv); | |
165 | __private_extern__ void munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p); | |
166 | __private_extern__ void munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p); | |
167 | static int reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock); | |
168 | static void populate_corpse_crashinfo(proc_t p, task_t corpse_task, | |
169 | struct rusage_superset *rup, mach_exception_data_type_t code, | |
170 | mach_exception_data_type_t subcode, uint64_t *udata_buffer, | |
171 | int num_udata, os_reason_t reason); | |
172 | static void proc_update_corpse_exception_codes(proc_t p, mach_exception_data_type_t *code, mach_exception_data_type_t *subcode); | |
173 | extern int proc_pidpathinfo_internal(proc_t p, uint64_t arg, char *buffer, uint32_t buffersize, int32_t *retval); | |
174 | static __attribute__((noinline)) void launchd_crashed_panic(proc_t p, int rv); | |
175 | extern void proc_piduniqidentifierinfo(proc_t p, struct proc_uniqidentifierinfo *p_uniqidinfo); | |
176 | extern void task_coalition_ids(task_t task, uint64_t ids[COALITION_NUM_TYPES]); | |
177 | extern uint64_t get_task_phys_footprint_limit(task_t); | |
178 | int proc_list_uptrs(void *p, uint64_t *udata_buffer, int size); | |
179 | extern uint64_t task_corpse_get_crashed_thread_id(task_t corpse_task); | |
180 | ||
181 | ||
182 | /* | |
183 | * Things which should have prototypes in headers, but don't | |
184 | */ | |
185 | void proc_exit(proc_t p); | |
186 | int wait1continue(int result); | |
187 | int waitidcontinue(int result); | |
188 | kern_return_t sys_perf_notify(thread_t thread, int pid); | |
189 | kern_return_t task_exception_notify(exception_type_t exception, | |
190 | mach_exception_data_type_t code, mach_exception_data_type_t subcode); | |
191 | kern_return_t task_violated_guard(mach_exception_code_t, mach_exception_subcode_t, void *); | |
192 | void delay(int); | |
193 | void gather_rusage_info(proc_t p, rusage_info_current *ru, int flavor); | |
194 | ||
195 | /* | |
196 | * NOTE: Source and target may *NOT* overlap! | |
197 | * XXX Should share code with bsd/dev/ppc/unix_signal.c | |
198 | */ | |
199 | void | |
200 | siginfo_user_to_user32(user_siginfo_t *in, user32_siginfo_t *out) | |
201 | { | |
202 | out->si_signo = in->si_signo; | |
203 | out->si_errno = in->si_errno; | |
204 | out->si_code = in->si_code; | |
205 | out->si_pid = in->si_pid; | |
206 | out->si_uid = in->si_uid; | |
207 | out->si_status = in->si_status; | |
208 | out->si_addr = CAST_DOWN_EXPLICIT(user32_addr_t, in->si_addr); | |
209 | /* following cast works for sival_int because of padding */ | |
210 | out->si_value.sival_ptr = CAST_DOWN_EXPLICIT(user32_addr_t, in->si_value.sival_ptr); | |
211 | out->si_band = in->si_band; /* range reduction */ | |
212 | } | |
213 | ||
214 | void | |
215 | siginfo_user_to_user64(user_siginfo_t *in, user64_siginfo_t *out) | |
216 | { | |
217 | out->si_signo = in->si_signo; | |
218 | out->si_errno = in->si_errno; | |
219 | out->si_code = in->si_code; | |
220 | out->si_pid = in->si_pid; | |
221 | out->si_uid = in->si_uid; | |
222 | out->si_status = in->si_status; | |
223 | out->si_addr = in->si_addr; | |
224 | /* following cast works for sival_int because of padding */ | |
225 | out->si_value.sival_ptr = in->si_value.sival_ptr; | |
226 | out->si_band = in->si_band; /* range reduction */ | |
227 | } | |
228 | ||
229 | static int | |
230 | copyoutsiginfo(user_siginfo_t *native, boolean_t is64, user_addr_t uaddr) | |
231 | { | |
232 | if (is64) { | |
233 | user64_siginfo_t sinfo64; | |
234 | ||
235 | bzero(&sinfo64, sizeof(sinfo64)); | |
236 | siginfo_user_to_user64(native, &sinfo64); | |
237 | return copyout(&sinfo64, uaddr, sizeof(sinfo64)); | |
238 | } else { | |
239 | user32_siginfo_t sinfo32; | |
240 | ||
241 | bzero(&sinfo32, sizeof(sinfo32)); | |
242 | siginfo_user_to_user32(native, &sinfo32); | |
243 | return copyout(&sinfo32, uaddr, sizeof(sinfo32)); | |
244 | } | |
245 | } | |
246 | ||
247 | void | |
248 | gather_populate_corpse_crashinfo(proc_t p, task_t corpse_task, | |
249 | mach_exception_data_type_t code, mach_exception_data_type_t subcode, | |
250 | uint64_t *udata_buffer, int num_udata, void *reason) | |
251 | { | |
252 | struct rusage_superset rup; | |
253 | ||
254 | gather_rusage_info(p, &rup.ri, RUSAGE_INFO_CURRENT); | |
255 | rup.ri.ri_phys_footprint = 0; | |
256 | populate_corpse_crashinfo(p, corpse_task, &rup, code, subcode, | |
257 | udata_buffer, num_udata, reason); | |
258 | } | |
259 | ||
260 | static void | |
261 | proc_update_corpse_exception_codes(proc_t p, mach_exception_data_type_t *code, mach_exception_data_type_t *subcode) | |
262 | { | |
263 | mach_exception_data_type_t code_update = *code; | |
264 | mach_exception_data_type_t subcode_update = *subcode; | |
265 | if (p->p_exit_reason == OS_REASON_NULL) { | |
266 | return; | |
267 | } | |
268 | ||
269 | switch (p->p_exit_reason->osr_namespace) { | |
270 | case OS_REASON_JETSAM: | |
271 | if (p->p_exit_reason->osr_code == JETSAM_REASON_MEMORY_PERPROCESSLIMIT) { | |
272 | /* Update the code with EXC_RESOURCE code for high memory watermark */ | |
273 | EXC_RESOURCE_ENCODE_TYPE(code_update, RESOURCE_TYPE_MEMORY); | |
274 | EXC_RESOURCE_ENCODE_FLAVOR(code_update, FLAVOR_HIGH_WATERMARK); | |
275 | EXC_RESOURCE_HWM_ENCODE_LIMIT(code_update, ((get_task_phys_footprint_limit(p->task)) >> 20)); | |
276 | subcode_update = 0; | |
277 | break; | |
278 | } | |
279 | ||
280 | break; | |
281 | default: | |
282 | break; | |
283 | } | |
284 | ||
285 | *code = code_update; | |
286 | *subcode = subcode_update; | |
287 | return; | |
288 | } | |
289 | ||
290 | mach_exception_data_type_t | |
291 | proc_encode_exit_exception_code(proc_t p) | |
292 | { | |
293 | uint64_t subcode = 0; | |
294 | ||
295 | if (p->p_exit_reason == OS_REASON_NULL) { | |
296 | return 0; | |
297 | } | |
298 | ||
299 | /* Embed first 32 bits of osr_namespace and osr_code in exception code */ | |
300 | ENCODE_OSR_NAMESPACE_TO_MACH_EXCEPTION_CODE(subcode, p->p_exit_reason->osr_namespace); | |
301 | ENCODE_OSR_CODE_TO_MACH_EXCEPTION_CODE(subcode, p->p_exit_reason->osr_code); | |
302 | return (mach_exception_data_type_t)subcode; | |
303 | } | |
304 | ||
305 | static void | |
306 | populate_corpse_crashinfo(proc_t p, task_t corpse_task, struct rusage_superset *rup, | |
307 | mach_exception_data_type_t code, mach_exception_data_type_t subcode, | |
308 | uint64_t *udata_buffer, int num_udata, os_reason_t reason) | |
309 | { | |
310 | mach_vm_address_t uaddr = 0; | |
311 | mach_exception_data_type_t exc_codes[EXCEPTION_CODE_MAX]; | |
312 | exc_codes[0] = code; | |
313 | exc_codes[1] = subcode; | |
314 | cpu_type_t cputype; | |
315 | struct proc_uniqidentifierinfo p_uniqidinfo; | |
316 | struct proc_workqueueinfo pwqinfo; | |
317 | int retval = 0; | |
318 | uint64_t crashed_threadid = task_corpse_get_crashed_thread_id(corpse_task); | |
319 | unsigned int pflags = 0; | |
320 | uint64_t max_footprint_mb; | |
321 | uint64_t max_footprint; | |
322 | ||
323 | uint64_t ledger_internal; | |
324 | uint64_t ledger_internal_compressed; | |
325 | uint64_t ledger_iokit_mapped; | |
326 | uint64_t ledger_alternate_accounting; | |
327 | uint64_t ledger_alternate_accounting_compressed; | |
328 | uint64_t ledger_purgeable_nonvolatile; | |
329 | uint64_t ledger_purgeable_nonvolatile_compressed; | |
330 | uint64_t ledger_page_table; | |
331 | uint64_t ledger_phys_footprint; | |
332 | uint64_t ledger_phys_footprint_lifetime_max; | |
333 | uint64_t ledger_network_nonvolatile; | |
334 | uint64_t ledger_network_nonvolatile_compressed; | |
335 | uint64_t ledger_wired_mem; | |
336 | ||
337 | void *crash_info_ptr = task_get_corpseinfo(corpse_task); | |
338 | ||
339 | #if CONFIG_MEMORYSTATUS | |
340 | int memstat_dirty_flags = 0; | |
341 | #endif | |
342 | ||
343 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_EXCEPTION_CODES, sizeof(exc_codes), &uaddr)) { | |
344 | kcdata_memcpy(crash_info_ptr, uaddr, exc_codes, sizeof(exc_codes)); | |
345 | } | |
346 | ||
347 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PID, sizeof(p->p_pid), &uaddr)) { | |
348 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_pid, sizeof(p->p_pid)); | |
349 | } | |
350 | ||
351 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PPID, sizeof(p->p_ppid), &uaddr)) { | |
352 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_ppid, sizeof(p->p_ppid)); | |
353 | } | |
354 | ||
355 | /* Don't include the crashed thread ID if there's an exit reason that indicates it's irrelevant */ | |
356 | if ((p->p_exit_reason == OS_REASON_NULL) || !(p->p_exit_reason->osr_flags & OS_REASON_FLAG_NO_CRASHED_TID)) { | |
357 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_CRASHED_THREADID, sizeof(uint64_t), &uaddr)) { | |
358 | kcdata_memcpy(crash_info_ptr, uaddr, &crashed_threadid, sizeof(uint64_t)); | |
359 | } | |
360 | } | |
361 | ||
362 | static_assert(sizeof(struct proc_uniqidentifierinfo) == sizeof(struct crashinfo_proc_uniqidentifierinfo)); | |
363 | if (KERN_SUCCESS == | |
364 | kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_BSDINFOWITHUNIQID, sizeof(struct proc_uniqidentifierinfo), &uaddr)) { | |
365 | proc_piduniqidentifierinfo(p, &p_uniqidinfo); | |
366 | kcdata_memcpy(crash_info_ptr, uaddr, &p_uniqidinfo, sizeof(struct proc_uniqidentifierinfo)); | |
367 | } | |
368 | ||
369 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_RUSAGE_INFO, sizeof(rusage_info_current), &uaddr)) { | |
370 | kcdata_memcpy(crash_info_ptr, uaddr, &rup->ri, sizeof(rusage_info_current)); | |
371 | } | |
372 | ||
373 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_CSFLAGS, sizeof(p->p_csflags), &uaddr)) { | |
374 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_csflags, sizeof(p->p_csflags)); | |
375 | } | |
376 | ||
377 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_NAME, sizeof(p->p_comm), &uaddr)) { | |
378 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_comm, sizeof(p->p_comm)); | |
379 | } | |
380 | ||
381 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_STARTTIME, sizeof(p->p_start), &uaddr)) { | |
382 | struct timeval64 t64; | |
383 | t64.tv_sec = (int64_t)p->p_start.tv_sec; | |
384 | t64.tv_usec = (int64_t)p->p_start.tv_usec; | |
385 | kcdata_memcpy(crash_info_ptr, uaddr, &t64, sizeof(t64)); | |
386 | } | |
387 | ||
388 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_USERSTACK, sizeof(p->user_stack), &uaddr)) { | |
389 | kcdata_memcpy(crash_info_ptr, uaddr, &p->user_stack, sizeof(p->user_stack)); | |
390 | } | |
391 | ||
392 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_ARGSLEN, sizeof(p->p_argslen), &uaddr)) { | |
393 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_argslen, sizeof(p->p_argslen)); | |
394 | } | |
395 | ||
396 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_ARGC, sizeof(p->p_argc), &uaddr)) { | |
397 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_argc, sizeof(p->p_argc)); | |
398 | } | |
399 | ||
400 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_PATH, MAXPATHLEN, &uaddr)) { | |
401 | char *buf = (char *) kalloc(MAXPATHLEN); | |
402 | if (buf != NULL) { | |
403 | bzero(buf, MAXPATHLEN); | |
404 | proc_pidpathinfo_internal(p, 0, buf, MAXPATHLEN, &retval); | |
405 | kcdata_memcpy(crash_info_ptr, uaddr, buf, MAXPATHLEN); | |
406 | kfree(buf, MAXPATHLEN); | |
407 | } | |
408 | } | |
409 | ||
410 | pflags = p->p_flag & (P_LP64 | P_SUGID); | |
411 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_FLAGS, sizeof(pflags), &uaddr)) { | |
412 | kcdata_memcpy(crash_info_ptr, uaddr, &pflags, sizeof(pflags)); | |
413 | } | |
414 | ||
415 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_UID, sizeof(p->p_uid), &uaddr)) { | |
416 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_uid, sizeof(p->p_uid)); | |
417 | } | |
418 | ||
419 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_GID, sizeof(p->p_gid), &uaddr)) { | |
420 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_gid, sizeof(p->p_gid)); | |
421 | } | |
422 | ||
423 | cputype = cpu_type() & ~CPU_ARCH_MASK; | |
424 | if (IS_64BIT_PROCESS(p)) { | |
425 | cputype |= CPU_ARCH_ABI64; | |
426 | } | |
427 | ||
428 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_CPUTYPE, sizeof(cpu_type_t), &uaddr)) { | |
429 | kcdata_memcpy(crash_info_ptr, uaddr, &cputype, sizeof(cpu_type_t)); | |
430 | } | |
431 | ||
432 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_MEMORY_LIMIT, sizeof(max_footprint_mb), &uaddr)) { | |
433 | max_footprint = get_task_phys_footprint_limit(p->task); | |
434 | max_footprint_mb = max_footprint >> 20; | |
435 | kcdata_memcpy(crash_info_ptr, uaddr, &max_footprint_mb, sizeof(max_footprint_mb)); | |
436 | } | |
437 | ||
438 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_PHYS_FOOTPRINT_LIFETIME_MAX, sizeof(ledger_phys_footprint_lifetime_max), &uaddr)) { | |
439 | ledger_phys_footprint_lifetime_max = get_task_phys_footprint_lifetime_max(p->task); | |
440 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_phys_footprint_lifetime_max, sizeof(ledger_phys_footprint_lifetime_max)); | |
441 | } | |
442 | ||
443 | // In the forking case, the current ledger info is copied into the corpse while the original task is suspended for consistency | |
444 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_INTERNAL, sizeof(ledger_internal), &uaddr)) { | |
445 | ledger_internal = get_task_internal(corpse_task); | |
446 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_internal, sizeof(ledger_internal)); | |
447 | } | |
448 | ||
449 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_INTERNAL_COMPRESSED, sizeof(ledger_internal_compressed), &uaddr)) { | |
450 | ledger_internal_compressed = get_task_internal_compressed(corpse_task); | |
451 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_internal_compressed, sizeof(ledger_internal_compressed)); | |
452 | } | |
453 | ||
454 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_IOKIT_MAPPED, sizeof(ledger_iokit_mapped), &uaddr)) { | |
455 | ledger_iokit_mapped = get_task_iokit_mapped(corpse_task); | |
456 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_iokit_mapped, sizeof(ledger_iokit_mapped)); | |
457 | } | |
458 | ||
459 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_ALTERNATE_ACCOUNTING, sizeof(ledger_alternate_accounting), &uaddr)) { | |
460 | ledger_alternate_accounting = get_task_alternate_accounting(corpse_task); | |
461 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_alternate_accounting, sizeof(ledger_alternate_accounting)); | |
462 | } | |
463 | ||
464 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_ALTERNATE_ACCOUNTING_COMPRESSED, sizeof(ledger_alternate_accounting_compressed), &uaddr)) { | |
465 | ledger_alternate_accounting_compressed = get_task_alternate_accounting_compressed(corpse_task); | |
466 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_alternate_accounting_compressed, sizeof(ledger_alternate_accounting_compressed)); | |
467 | } | |
468 | ||
469 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_PURGEABLE_NONVOLATILE, sizeof(ledger_purgeable_nonvolatile), &uaddr)) { | |
470 | ledger_purgeable_nonvolatile = get_task_purgeable_nonvolatile(corpse_task); | |
471 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_purgeable_nonvolatile, sizeof(ledger_purgeable_nonvolatile)); | |
472 | } | |
473 | ||
474 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_PURGEABLE_NONVOLATILE_COMPRESSED, sizeof(ledger_purgeable_nonvolatile_compressed), &uaddr)) { | |
475 | ledger_purgeable_nonvolatile_compressed = get_task_purgeable_nonvolatile_compressed(corpse_task); | |
476 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_purgeable_nonvolatile_compressed, sizeof(ledger_purgeable_nonvolatile_compressed)); | |
477 | } | |
478 | ||
479 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_PAGE_TABLE, sizeof(ledger_page_table), &uaddr)) { | |
480 | ledger_page_table = get_task_page_table(corpse_task); | |
481 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_page_table, sizeof(ledger_page_table)); | |
482 | } | |
483 | ||
484 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_PHYS_FOOTPRINT, sizeof(ledger_phys_footprint), &uaddr)) { | |
485 | ledger_phys_footprint = get_task_phys_footprint(corpse_task); | |
486 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_phys_footprint, sizeof(ledger_phys_footprint)); | |
487 | } | |
488 | ||
489 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_NETWORK_NONVOLATILE, sizeof(ledger_network_nonvolatile), &uaddr)) { | |
490 | ledger_network_nonvolatile = get_task_network_nonvolatile(corpse_task); | |
491 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_network_nonvolatile, sizeof(ledger_network_nonvolatile)); | |
492 | } | |
493 | ||
494 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_NETWORK_NONVOLATILE_COMPRESSED, sizeof(ledger_network_nonvolatile_compressed), &uaddr)) { | |
495 | ledger_network_nonvolatile_compressed = get_task_network_nonvolatile_compressed(corpse_task); | |
496 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_network_nonvolatile_compressed, sizeof(ledger_network_nonvolatile_compressed)); | |
497 | } | |
498 | ||
499 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_LEDGER_WIRED_MEM, sizeof(ledger_wired_mem), &uaddr)) { | |
500 | ledger_wired_mem = get_task_wired_mem(corpse_task); | |
501 | kcdata_memcpy(crash_info_ptr, uaddr, &ledger_wired_mem, sizeof(ledger_wired_mem)); | |
502 | } | |
503 | ||
504 | bzero(&pwqinfo, sizeof(struct proc_workqueueinfo)); | |
505 | retval = fill_procworkqueue(p, &pwqinfo); | |
506 | if (retval == 0) { | |
507 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_WORKQUEUEINFO, sizeof(struct proc_workqueueinfo), &uaddr)) { | |
508 | kcdata_memcpy(crash_info_ptr, uaddr, &pwqinfo, sizeof(struct proc_workqueueinfo)); | |
509 | } | |
510 | } | |
511 | ||
512 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_RESPONSIBLE_PID, sizeof(p->p_responsible_pid), &uaddr)) { | |
513 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_responsible_pid, sizeof(p->p_responsible_pid)); | |
514 | } | |
515 | ||
516 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_PERSONA_ID, sizeof(uid_t), &uaddr)) { | |
517 | uid_t persona_id = proc_persona_id(p); | |
518 | kcdata_memcpy(crash_info_ptr, uaddr, &persona_id, sizeof(persona_id)); | |
519 | } | |
520 | ||
521 | #if CONFIG_COALITIONS | |
522 | if (KERN_SUCCESS == kcdata_get_memory_addr_for_array(crash_info_ptr, TASK_CRASHINFO_COALITION_ID, sizeof(uint64_t), COALITION_NUM_TYPES, &uaddr)) { | |
523 | uint64_t coalition_ids[COALITION_NUM_TYPES]; | |
524 | task_coalition_ids(p->task, coalition_ids); | |
525 | kcdata_memcpy(crash_info_ptr, uaddr, coalition_ids, sizeof(coalition_ids)); | |
526 | } | |
527 | #endif /* CONFIG_COALITIONS */ | |
528 | ||
529 | #if CONFIG_MEMORYSTATUS | |
530 | memstat_dirty_flags = memorystatus_dirty_get(p, FALSE); | |
531 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_DIRTY_FLAGS, sizeof(memstat_dirty_flags), &uaddr)) { | |
532 | kcdata_memcpy(crash_info_ptr, uaddr, &memstat_dirty_flags, sizeof(memstat_dirty_flags)); | |
533 | } | |
534 | #endif | |
535 | ||
536 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_MEMORY_LIMIT_INCREASE, sizeof(p->p_memlimit_increase), &uaddr)) { | |
537 | kcdata_memcpy(crash_info_ptr, uaddr, &p->p_memlimit_increase, sizeof(p->p_memlimit_increase)); | |
538 | } | |
539 | ||
540 | if (p->p_exit_reason != OS_REASON_NULL && reason == OS_REASON_NULL) { | |
541 | reason = p->p_exit_reason; | |
542 | } | |
543 | if (reason != OS_REASON_NULL) { | |
544 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, EXIT_REASON_SNAPSHOT, sizeof(struct exit_reason_snapshot), &uaddr)) { | |
545 | struct exit_reason_snapshot ers = { | |
546 | .ers_namespace = reason->osr_namespace, | |
547 | .ers_code = reason->osr_code, | |
548 | .ers_flags = reason->osr_flags | |
549 | }; | |
550 | ||
551 | kcdata_memcpy(crash_info_ptr, uaddr, &ers, sizeof(ers)); | |
552 | } | |
553 | ||
554 | if (reason->osr_kcd_buf != 0) { | |
555 | uint32_t reason_buf_size = kcdata_memory_get_used_bytes(&reason->osr_kcd_descriptor); | |
556 | assert(reason_buf_size != 0); | |
557 | ||
558 | if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, KCDATA_TYPE_NESTED_KCDATA, reason_buf_size, &uaddr)) { | |
559 | kcdata_memcpy(crash_info_ptr, uaddr, reason->osr_kcd_buf, reason_buf_size); | |
560 | } | |
561 | } | |
562 | } | |
563 | ||
564 | if (num_udata > 0) { | |
565 | if (KERN_SUCCESS == kcdata_get_memory_addr_for_array(crash_info_ptr, TASK_CRASHINFO_UDATA_PTRS, | |
566 | sizeof(uint64_t), num_udata, &uaddr)) { | |
567 | kcdata_memcpy(crash_info_ptr, uaddr, udata_buffer, sizeof(uint64_t) * num_udata); | |
568 | } | |
569 | } | |
570 | } | |
571 | ||
572 | /* | |
573 | * We only parse exit reason kcdata blobs for launchd when it dies | |
574 | * and we're going to panic. | |
575 | * | |
576 | * Meant to be called immediately before panicking. | |
577 | */ | |
578 | char * | |
579 | launchd_exit_reason_get_string_desc(os_reason_t exit_reason) | |
580 | { | |
581 | kcdata_iter_t iter; | |
582 | ||
583 | if (exit_reason == OS_REASON_NULL || exit_reason->osr_kcd_buf == NULL || | |
584 | exit_reason->osr_bufsize == 0) { | |
585 | return NULL; | |
586 | } | |
587 | ||
588 | iter = kcdata_iter(exit_reason->osr_kcd_buf, exit_reason->osr_bufsize); | |
589 | if (!kcdata_iter_valid(iter)) { | |
590 | #if DEBUG || DEVELOPMENT | |
591 | printf("launchd exit reason has invalid exit reason buffer\n"); | |
592 | #endif | |
593 | return NULL; | |
594 | } | |
595 | ||
596 | if (kcdata_iter_type(iter) != KCDATA_BUFFER_BEGIN_OS_REASON) { | |
597 | #if DEBUG || DEVELOPMENT | |
598 | printf("launchd exit reason buffer type mismatch, expected %d got %d\n", | |
599 | KCDATA_BUFFER_BEGIN_OS_REASON, kcdata_iter_type(iter)); | |
600 | #endif | |
601 | return NULL; | |
602 | } | |
603 | ||
604 | iter = kcdata_iter_find_type(iter, EXIT_REASON_USER_DESC); | |
605 | if (!kcdata_iter_valid(iter)) { | |
606 | return NULL; | |
607 | } | |
608 | ||
609 | return (char *)kcdata_iter_payload(iter); | |
610 | } | |
611 | ||
612 | __abortlike | |
613 | static void | |
614 | launchd_crashed_panic(proc_t p, int rv) | |
615 | { | |
616 | char *launchd_exit_reason_desc = launchd_exit_reason_get_string_desc(p->p_exit_reason); | |
617 | ||
618 | if (p->p_exit_reason == OS_REASON_NULL) { | |
619 | printf("pid 1 exited -- no exit reason available -- (signal %d, exit %d)\n", | |
620 | WTERMSIG(rv), WEXITSTATUS(rv)); | |
621 | } else { | |
622 | printf("pid 1 exited -- exit reason namespace %d subcode 0x%llx, description %s\n", | |
623 | p->p_exit_reason->osr_namespace, p->p_exit_reason->osr_code, launchd_exit_reason_desc ? | |
624 | launchd_exit_reason_desc : "none"); | |
625 | } | |
626 | ||
627 | const char *launchd_crashed_prefix_str; | |
628 | ||
629 | if (strnstr(p->p_name, "preinit", sizeof(p->p_name))) { | |
630 | launchd_crashed_prefix_str = "LTE preinit process exited"; | |
631 | } else { | |
632 | launchd_crashed_prefix_str = "initproc exited"; | |
633 | } | |
634 | ||
635 | #if (DEVELOPMENT || DEBUG) && CONFIG_COREDUMP | |
636 | /* | |
637 | * For debugging purposes, generate a core file of initproc before | |
638 | * panicking. Leave at least 300 MB free on the root volume, and ignore | |
639 | * the process's corefile ulimit. fsync() the file to ensure it lands on disk | |
640 | * before the panic hits. | |
641 | */ | |
642 | ||
643 | int err; | |
644 | uint64_t coredump_start = mach_absolute_time(); | |
645 | uint64_t coredump_end; | |
646 | clock_sec_t tv_sec; | |
647 | clock_usec_t tv_usec; | |
648 | uint32_t tv_msec; | |
649 | ||
650 | ||
651 | err = coredump(p, 300, COREDUMP_IGNORE_ULIMIT | COREDUMP_FULLFSYNC); | |
652 | ||
653 | coredump_end = mach_absolute_time(); | |
654 | ||
655 | absolutetime_to_microtime(coredump_end - coredump_start, &tv_sec, &tv_usec); | |
656 | ||
657 | tv_msec = tv_usec / 1000; | |
658 | ||
659 | if (err != 0) { | |
660 | printf("Failed to generate initproc core file: error %d, took %d.%03d seconds\n", | |
661 | err, (uint32_t)tv_sec, tv_msec); | |
662 | } else { | |
663 | printf("Generated initproc core file in %d.%03d seconds\n", | |
664 | (uint32_t)tv_sec, tv_msec); | |
665 | } | |
666 | #endif /* (DEVELOPMENT || DEBUG) && CONFIG_COREDUMP */ | |
667 | ||
668 | sync(p, (void *)NULL, (int *)NULL); | |
669 | ||
670 | if (p->p_exit_reason == OS_REASON_NULL) { | |
671 | panic_with_options(0, NULL, DEBUGGER_OPTION_INITPROC_PANIC, "%s -- no exit reason available -- (signal %d, exit status %d %s)", | |
672 | launchd_crashed_prefix_str, WTERMSIG(rv), WEXITSTATUS(rv), ((p->p_csflags & CS_KILLED) ? "CS_KILLED" : "")); | |
673 | } else { | |
674 | panic_with_options(0, NULL, DEBUGGER_OPTION_INITPROC_PANIC, "%s %s -- exit reason namespace %d subcode 0x%llx description: %." LAUNCHD_PANIC_REASON_STRING_MAXLEN "s", | |
675 | ((p->p_csflags & CS_KILLED) ? "CS_KILLED" : ""), | |
676 | launchd_crashed_prefix_str, p->p_exit_reason->osr_namespace, p->p_exit_reason->osr_code, | |
677 | launchd_exit_reason_desc ? launchd_exit_reason_desc : "none"); | |
678 | } | |
679 | } | |
680 | ||
681 | #define OS_REASON_IFLAG_USER_FAULT 0x1 | |
682 | ||
683 | #define OS_REASON_TOTAL_USER_FAULTS_PER_PROC 5 | |
684 | ||
685 | static int | |
686 | abort_with_payload_internal(proc_t p, | |
687 | uint32_t reason_namespace, uint64_t reason_code, | |
688 | user_addr_t payload, uint32_t payload_size, | |
689 | user_addr_t reason_string, uint64_t reason_flags, | |
690 | uint32_t internal_flags) | |
691 | { | |
692 | os_reason_t exit_reason = OS_REASON_NULL; | |
693 | kern_return_t kr = KERN_SUCCESS; | |
694 | ||
695 | if (internal_flags & OS_REASON_IFLAG_USER_FAULT) { | |
696 | uint32_t old_value = atomic_load_explicit(&p->p_user_faults, | |
697 | memory_order_relaxed); | |
698 | for (;;) { | |
699 | if (old_value >= OS_REASON_TOTAL_USER_FAULTS_PER_PROC) { | |
700 | return EQFULL; | |
701 | } | |
702 | // this reloads the value in old_value | |
703 | if (atomic_compare_exchange_strong_explicit(&p->p_user_faults, | |
704 | &old_value, old_value + 1, memory_order_relaxed, | |
705 | memory_order_relaxed)) { | |
706 | break; | |
707 | } | |
708 | } | |
709 | } | |
710 | ||
711 | KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, | |
712 | p->p_pid, reason_namespace, | |
713 | reason_code, 0, 0); | |
714 | ||
715 | exit_reason = build_userspace_exit_reason(reason_namespace, reason_code, | |
716 | payload, payload_size, reason_string, reason_flags | OS_REASON_FLAG_ABORT); | |
717 | ||
718 | if (internal_flags & OS_REASON_IFLAG_USER_FAULT) { | |
719 | mach_exception_code_t code = 0; | |
720 | ||
721 | EXC_GUARD_ENCODE_TYPE(code, GUARD_TYPE_USER); /* simulated EXC_GUARD */ | |
722 | EXC_GUARD_ENCODE_FLAVOR(code, 0); | |
723 | EXC_GUARD_ENCODE_TARGET(code, reason_namespace); | |
724 | ||
725 | if (exit_reason == OS_REASON_NULL) { | |
726 | kr = KERN_RESOURCE_SHORTAGE; | |
727 | } else { | |
728 | kr = task_violated_guard(code, reason_code, exit_reason); | |
729 | } | |
730 | os_reason_free(exit_reason); | |
731 | } else { | |
732 | /* | |
733 | * We use SIGABRT (rather than calling exit directly from here) so that | |
734 | * the debugger can catch abort_with_{reason,payload} calls. | |
735 | */ | |
736 | psignal_try_thread_with_reason(p, current_thread(), SIGABRT, exit_reason); | |
737 | } | |
738 | ||
739 | switch (kr) { | |
740 | case KERN_SUCCESS: | |
741 | return 0; | |
742 | case KERN_NOT_SUPPORTED: | |
743 | return ENOTSUP; | |
744 | case KERN_INVALID_ARGUMENT: | |
745 | return EINVAL; | |
746 | case KERN_RESOURCE_SHORTAGE: | |
747 | default: | |
748 | return EBUSY; | |
749 | } | |
750 | } | |
751 | ||
752 | int | |
753 | abort_with_payload(struct proc *cur_proc, struct abort_with_payload_args *args, | |
754 | __unused void *retval) | |
755 | { | |
756 | abort_with_payload_internal(cur_proc, args->reason_namespace, | |
757 | args->reason_code, args->payload, args->payload_size, | |
758 | args->reason_string, args->reason_flags, 0); | |
759 | ||
760 | return 0; | |
761 | } | |
762 | ||
763 | int | |
764 | os_fault_with_payload(struct proc *cur_proc, | |
765 | struct os_fault_with_payload_args *args, __unused int *retval) | |
766 | { | |
767 | return abort_with_payload_internal(cur_proc, args->reason_namespace, | |
768 | args->reason_code, args->payload, args->payload_size, | |
769 | args->reason_string, args->reason_flags, OS_REASON_IFLAG_USER_FAULT); | |
770 | } | |
771 | ||
772 | ||
773 | /* | |
774 | * exit -- | |
775 | * Death of process. | |
776 | */ | |
777 | __attribute__((noreturn)) | |
778 | void | |
779 | exit(proc_t p, struct exit_args *uap, int *retval) | |
780 | { | |
781 | p->p_xhighbits = ((uint32_t)(uap->rval) & 0xFF000000) >> 24; | |
782 | exit1(p, W_EXITCODE(uap->rval, 0), retval); | |
783 | ||
784 | thread_exception_return(); | |
785 | /* NOTREACHED */ | |
786 | while (TRUE) { | |
787 | thread_block(THREAD_CONTINUE_NULL); | |
788 | } | |
789 | /* NOTREACHED */ | |
790 | } | |
791 | ||
792 | /* | |
793 | * Exit: deallocate address space and other resources, change proc state | |
794 | * to zombie, and unlink proc from allproc and parent's lists. Save exit | |
795 | * status and rusage for wait(). Check for child processes and orphan them. | |
796 | */ | |
797 | int | |
798 | exit1(proc_t p, int rv, int *retval) | |
799 | { | |
800 | return exit1_internal(p, rv, retval, TRUE, TRUE, 0); | |
801 | } | |
802 | ||
803 | int | |
804 | exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, boolean_t perf_notify, | |
805 | int jetsam_flags) | |
806 | { | |
807 | return exit_with_reason(p, rv, retval, thread_can_terminate, perf_notify, jetsam_flags, OS_REASON_NULL); | |
808 | } | |
809 | ||
810 | /* | |
811 | * NOTE: exit_with_reason drops a reference on the passed exit_reason | |
812 | */ | |
813 | int | |
814 | exit_with_reason(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, boolean_t perf_notify, | |
815 | int jetsam_flags, struct os_reason *exit_reason) | |
816 | { | |
817 | thread_t self = current_thread(); | |
818 | struct task *task = p->task; | |
819 | struct uthread *ut; | |
820 | int error = 0; | |
821 | ||
822 | /* | |
823 | * If a thread in this task has already | |
824 | * called exit(), then halt any others | |
825 | * right here. | |
826 | */ | |
827 | ||
828 | ut = get_bsdthread_info(self); | |
829 | if ((p == current_proc()) && | |
830 | (ut->uu_flag & UT_VFORK)) { | |
831 | os_reason_free(exit_reason); | |
832 | if (!thread_can_terminate) { | |
833 | return EINVAL; | |
834 | } | |
835 | ||
836 | vfork_exit(p, rv); | |
837 | vfork_return(p, retval, p->p_pid); | |
838 | unix_syscall_return(0); | |
839 | /* NOT REACHED */ | |
840 | } | |
841 | ||
842 | /* | |
843 | * The parameter list of audit_syscall_exit() was augmented to | |
844 | * take the Darwin syscall number as the first parameter, | |
845 | * which is currently required by mac_audit_postselect(). | |
846 | */ | |
847 | ||
848 | /* | |
849 | * The BSM token contains two components: an exit status as passed | |
850 | * to exit(), and a return value to indicate what sort of exit it | |
851 | * was. The exit status is WEXITSTATUS(rv), but it's not clear | |
852 | * what the return value is. | |
853 | */ | |
854 | AUDIT_ARG(exit, WEXITSTATUS(rv), 0); | |
855 | /* | |
856 | * TODO: what to audit here when jetsam calls exit and the uthread, | |
857 | * 'ut' does not belong to the proc, 'p'. | |
858 | */ | |
859 | AUDIT_SYSCALL_EXIT(SYS_exit, p, ut, 0); /* Exit is always successfull */ | |
860 | ||
861 | DTRACE_PROC1(exit, int, CLD_EXITED); | |
862 | ||
863 | /* mark process is going to exit and pull out of DBG/disk throttle */ | |
864 | /* TODO: This should be done after becoming exit thread */ | |
865 | proc_set_task_policy(p->task, TASK_POLICY_ATTRIBUTE, | |
866 | TASK_POLICY_TERMINATED, TASK_POLICY_ENABLE); | |
867 | ||
868 | proc_lock(p); | |
869 | error = proc_transstart(p, 1, (jetsam_flags ? 1 : 0)); | |
870 | if (error == EDEADLK) { | |
871 | /* | |
872 | * If proc_transstart() returns EDEADLK, then another thread | |
873 | * is either exec'ing or exiting. Return an error and allow | |
874 | * the other thread to continue. | |
875 | */ | |
876 | proc_unlock(p); | |
877 | os_reason_free(exit_reason); | |
878 | if (current_proc() == p) { | |
879 | if (p->exit_thread == self) { | |
880 | printf("exit_thread failed to exit, leaving process %s[%d] in unkillable limbo\n", | |
881 | p->p_comm, p->p_pid); | |
882 | } | |
883 | ||
884 | if (thread_can_terminate) { | |
885 | thread_exception_return(); | |
886 | } | |
887 | } | |
888 | ||
889 | return error; | |
890 | } | |
891 | ||
892 | while (p->exit_thread != self) { | |
893 | if (sig_try_locked(p) <= 0) { | |
894 | proc_transend(p, 1); | |
895 | os_reason_free(exit_reason); | |
896 | ||
897 | if (get_threadtask(self) != task) { | |
898 | proc_unlock(p); | |
899 | return 0; | |
900 | } | |
901 | proc_unlock(p); | |
902 | ||
903 | thread_terminate(self); | |
904 | if (!thread_can_terminate) { | |
905 | return 0; | |
906 | } | |
907 | ||
908 | thread_exception_return(); | |
909 | /* NOTREACHED */ | |
910 | } | |
911 | sig_lock_to_exit(p); | |
912 | } | |
913 | ||
914 | if (exit_reason != OS_REASON_NULL) { | |
915 | KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_COMMIT) | DBG_FUNC_NONE, | |
916 | p->p_pid, exit_reason->osr_namespace, | |
917 | exit_reason->osr_code, 0, 0); | |
918 | } | |
919 | ||
920 | assert(p->p_exit_reason == OS_REASON_NULL); | |
921 | p->p_exit_reason = exit_reason; | |
922 | ||
923 | p->p_lflag |= P_LEXIT; | |
924 | p->p_xstat = rv; | |
925 | p->p_lflag |= jetsam_flags; | |
926 | ||
927 | proc_transend(p, 1); | |
928 | proc_unlock(p); | |
929 | ||
930 | proc_prepareexit(p, rv, perf_notify); | |
931 | ||
932 | /* Last thread to terminate will call proc_exit() */ | |
933 | task_terminate_internal(task); | |
934 | ||
935 | return 0; | |
936 | } | |
937 | ||
938 | #if CONFIG_MEMORYSTATUS | |
939 | /* | |
940 | * Remove this process from jetsam bands for freezing or exiting. Note this will block, if the process | |
941 | * is currently being frozen. | |
942 | * The proc_list_lock is held by the caller. | |
943 | * NB: If the process should be ineligible for future freezing or jetsaming the caller should first set | |
944 | * the p_listflag P_LIST_EXITED bit. | |
945 | */ | |
946 | static void | |
947 | proc_memorystatus_remove(proc_t p) | |
948 | { | |
949 | LCK_MTX_ASSERT(proc_list_mlock, LCK_MTX_ASSERT_OWNED); | |
950 | while (memorystatus_remove(p) == EAGAIN) { | |
951 | os_log(OS_LOG_DEFAULT, "memorystatus_remove: Process[%d] tried to exit while being frozen. Blocking exit until freeze completes.", p->p_pid); | |
952 | msleep(&p->p_memstat_state, proc_list_mlock, PWAIT, "proc_memorystatus_remove", NULL); | |
953 | } | |
954 | } | |
955 | #endif | |
956 | ||
957 | void | |
958 | proc_prepareexit(proc_t p, int rv, boolean_t perf_notify) | |
959 | { | |
960 | mach_exception_data_type_t code = 0, subcode = 0; | |
961 | ||
962 | struct uthread *ut; | |
963 | thread_t self = current_thread(); | |
964 | ut = get_bsdthread_info(self); | |
965 | struct rusage_superset *rup; | |
966 | int kr = 0; | |
967 | int create_corpse = FALSE; | |
968 | ||
969 | if (p == initproc) { | |
970 | launchd_crashed_panic(p, rv); | |
971 | /* NOTREACHED */ | |
972 | } | |
973 | ||
974 | /* | |
975 | * Generate a corefile/crashlog if: | |
976 | * The process doesn't have an exit reason that indicates no crash report should be created | |
977 | * AND any of the following are true: | |
978 | * - The process was terminated due to a fatal signal that generates a core | |
979 | * - The process was killed due to a code signing violation | |
980 | * - The process has an exit reason that indicates we should generate a crash report | |
981 | * | |
982 | * The first condition is necessary because abort_with_reason()/payload() use SIGABRT | |
983 | * (which normally triggers a core) but may indicate that no crash report should be created. | |
984 | */ | |
985 | if (!(PROC_HAS_EXITREASON(p) && (PROC_EXITREASON_FLAGS(p) & OS_REASON_FLAG_NO_CRASH_REPORT)) && | |
986 | (hassigprop(WTERMSIG(rv), SA_CORE) || ((p->p_csflags & CS_KILLED) != 0) || | |
987 | (PROC_HAS_EXITREASON(p) && (PROC_EXITREASON_FLAGS(p) & | |
988 | OS_REASON_FLAG_GENERATE_CRASH_REPORT)))) { | |
989 | /* | |
990 | * Workaround for processes checking up on PT_DENY_ATTACH: | |
991 | * should be backed out post-Leopard (details in 5431025). | |
992 | */ | |
993 | if ((SIGSEGV == WTERMSIG(rv)) && | |
994 | (p->p_pptr->p_lflag & P_LNOATTACH)) { | |
995 | goto skipcheck; | |
996 | } | |
997 | ||
998 | /* | |
999 | * Crash Reporter looks for the signal value, original exception | |
1000 | * type, and low 20 bits of the original code in code[0] | |
1001 | * (8, 4, and 20 bits respectively). code[1] is unmodified. | |
1002 | */ | |
1003 | code = ((WTERMSIG(rv) & 0xff) << 24) | | |
1004 | ((ut->uu_exception & 0x0f) << 20) | | |
1005 | ((int)ut->uu_code & 0xfffff); | |
1006 | subcode = ut->uu_subcode; | |
1007 | ||
1008 | kr = task_exception_notify(EXC_CRASH, code, subcode); | |
1009 | ||
1010 | /* Nobody handled EXC_CRASH?? remember to make corpse */ | |
1011 | if (kr != 0) { | |
1012 | create_corpse = TRUE; | |
1013 | } | |
1014 | } | |
1015 | ||
1016 | skipcheck: | |
1017 | /* Notify the perf server? */ | |
1018 | if (perf_notify) { | |
1019 | (void)sys_perf_notify(self, p->p_pid); | |
1020 | } | |
1021 | ||
1022 | ||
1023 | /* stash the usage into corpse data if making_corpse == true */ | |
1024 | if (create_corpse == TRUE) { | |
1025 | kr = task_mark_corpse(p->task); | |
1026 | if (kr != KERN_SUCCESS) { | |
1027 | if (kr == KERN_NO_SPACE) { | |
1028 | printf("Process[%d] has no vm space for corpse info.\n", p->p_pid); | |
1029 | } else if (kr == KERN_NOT_SUPPORTED) { | |
1030 | printf("Process[%d] was destined to be corpse. But corpse is disabled by config.\n", p->p_pid); | |
1031 | } else { | |
1032 | printf("Process[%d] crashed: %s. Too many corpses being created.\n", p->p_pid, p->p_comm); | |
1033 | } | |
1034 | create_corpse = FALSE; | |
1035 | } | |
1036 | } | |
1037 | ||
1038 | /* | |
1039 | * Before this process becomes a zombie, stash resource usage | |
1040 | * stats in the proc for external observers to query | |
1041 | * via proc_pid_rusage(). | |
1042 | * | |
1043 | * If the zombie allocation fails, just punt the stats. | |
1044 | */ | |
1045 | MALLOC_ZONE(rup, struct rusage_superset *, | |
1046 | sizeof(*rup), M_ZOMBIE, M_WAITOK); | |
1047 | if (rup != NULL) { | |
1048 | gather_rusage_info(p, &rup->ri, RUSAGE_INFO_CURRENT); | |
1049 | rup->ri.ri_phys_footprint = 0; | |
1050 | rup->ri.ri_proc_exit_abstime = mach_absolute_time(); | |
1051 | ||
1052 | /* | |
1053 | * Make the rusage_info visible to external observers | |
1054 | * only after it has been completely filled in. | |
1055 | */ | |
1056 | p->p_ru = rup; | |
1057 | } | |
1058 | if (create_corpse) { | |
1059 | int est_knotes = 0, num_knotes = 0; | |
1060 | uint64_t *buffer = NULL; | |
1061 | int buf_size = 0; | |
1062 | ||
1063 | /* Get all the udata pointers from kqueue */ | |
1064 | est_knotes = kevent_proc_copy_uptrs(p, NULL, 0); | |
1065 | if (est_knotes > 0) { | |
1066 | buf_size = (est_knotes + 32) * sizeof(uint64_t); | |
1067 | buffer = (uint64_t *) kalloc(buf_size); | |
1068 | num_knotes = kevent_proc_copy_uptrs(p, buffer, buf_size); | |
1069 | if (num_knotes > est_knotes + 32) { | |
1070 | num_knotes = est_knotes + 32; | |
1071 | } | |
1072 | } | |
1073 | ||
1074 | /* Update the code, subcode based on exit reason */ | |
1075 | proc_update_corpse_exception_codes(p, &code, &subcode); | |
1076 | populate_corpse_crashinfo(p, p->task, rup, | |
1077 | code, subcode, buffer, num_knotes, NULL); | |
1078 | if (buffer != NULL) { | |
1079 | kfree(buffer, buf_size); | |
1080 | } | |
1081 | } | |
1082 | /* | |
1083 | * Remove proc from allproc queue and from pidhash chain. | |
1084 | * Need to do this before we do anything that can block. | |
1085 | * Not doing causes things like mount() find this on allproc | |
1086 | * in partially cleaned state. | |
1087 | */ | |
1088 | ||
1089 | proc_list_lock(); | |
1090 | ||
1091 | #if CONFIG_MEMORYSTATUS | |
1092 | proc_memorystatus_remove(p); | |
1093 | #endif | |
1094 | ||
1095 | LIST_REMOVE(p, p_list); | |
1096 | LIST_INSERT_HEAD(&zombproc, p, p_list); /* Place onto zombproc. */ | |
1097 | /* will not be visible via proc_find */ | |
1098 | p->p_listflag |= P_LIST_EXITED; | |
1099 | ||
1100 | proc_list_unlock(); | |
1101 | ||
1102 | #ifdef PGINPROF | |
1103 | vmsizmon(); | |
1104 | #endif | |
1105 | /* | |
1106 | * If parent is waiting for us to exit or exec, | |
1107 | * P_LPPWAIT is set; we will wakeup the parent below. | |
1108 | */ | |
1109 | proc_lock(p); | |
1110 | p->p_lflag &= ~(P_LTRACED | P_LPPWAIT); | |
1111 | p->p_sigignore = ~(sigcantmask); | |
1112 | ut->uu_siglist = 0; | |
1113 | proc_unlock(p); | |
1114 | } | |
1115 | ||
1116 | void | |
1117 | proc_exit(proc_t p) | |
1118 | { | |
1119 | proc_t q; | |
1120 | proc_t pp; | |
1121 | struct task *task = p->task; | |
1122 | vnode_t tvp = NULLVP; | |
1123 | struct pgrp * pg; | |
1124 | struct session *sessp; | |
1125 | struct uthread * uth; | |
1126 | pid_t pid; | |
1127 | int exitval; | |
1128 | int knote_hint; | |
1129 | ||
1130 | uth = current_uthread(); | |
1131 | ||
1132 | proc_lock(p); | |
1133 | proc_transstart(p, 1, 0); | |
1134 | if (!(p->p_lflag & P_LEXIT)) { | |
1135 | /* | |
1136 | * This can happen if a thread_terminate() occurs | |
1137 | * in a single-threaded process. | |
1138 | */ | |
1139 | p->p_lflag |= P_LEXIT; | |
1140 | proc_transend(p, 1); | |
1141 | proc_unlock(p); | |
1142 | proc_prepareexit(p, 0, TRUE); | |
1143 | (void) task_terminate_internal(task); | |
1144 | proc_lock(p); | |
1145 | } else { | |
1146 | proc_transend(p, 1); | |
1147 | } | |
1148 | ||
1149 | p->p_lflag |= P_LPEXIT; | |
1150 | ||
1151 | /* | |
1152 | * Other kernel threads may be in the middle of signalling this process. | |
1153 | * Wait for those threads to wrap it up before making the process | |
1154 | * disappear on them. | |
1155 | */ | |
1156 | if ((p->p_lflag & P_LINSIGNAL) || (p->p_sigwaitcnt > 0)) { | |
1157 | p->p_sigwaitcnt++; | |
1158 | while ((p->p_lflag & P_LINSIGNAL) || (p->p_sigwaitcnt > 1)) { | |
1159 | msleep(&p->p_sigmask, &p->p_mlock, PWAIT, "proc_sigdrain", NULL); | |
1160 | } | |
1161 | p->p_sigwaitcnt--; | |
1162 | } | |
1163 | ||
1164 | proc_unlock(p); | |
1165 | pid = p->p_pid; | |
1166 | exitval = p->p_xstat; | |
1167 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON, | |
1168 | BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_START, | |
1169 | pid, exitval, 0, 0, 0); | |
1170 | ||
1171 | #if CONFIG_DTRACE | |
1172 | dtrace_proc_exit(p); | |
1173 | #endif | |
1174 | ||
1175 | /* | |
1176 | * need to cancel async IO requests that can be cancelled and wait for those | |
1177 | * already active. MAY BLOCK! | |
1178 | */ | |
1179 | ||
1180 | proc_refdrain(p); | |
1181 | ||
1182 | /* if any pending cpu limits action, clear it */ | |
1183 | task_clear_cpuusage(p->task, TRUE); | |
1184 | ||
1185 | workq_mark_exiting(p); | |
1186 | ||
1187 | _aio_exit( p ); | |
1188 | ||
1189 | /* | |
1190 | * Close open files and release open-file table. | |
1191 | * This may block! | |
1192 | */ | |
1193 | fdfree(p); | |
1194 | ||
1195 | /* | |
1196 | * Once all the knotes, kqueues & workloops are destroyed, get rid of the | |
1197 | * workqueue. | |
1198 | */ | |
1199 | workq_exit(p); | |
1200 | ||
1201 | if (uth->uu_lowpri_window) { | |
1202 | /* | |
1203 | * task is marked as a low priority I/O type | |
1204 | * and the I/O we issued while in flushing files on close | |
1205 | * collided with normal I/O operations... | |
1206 | * no need to throttle this thread since its going away | |
1207 | * but we do need to update our bookeeping w/r to throttled threads | |
1208 | */ | |
1209 | throttle_lowpri_io(0); | |
1210 | } | |
1211 | ||
1212 | if (p->p_lflag & P_LNSPACE_RESOLVER) { | |
1213 | /* | |
1214 | * The namespace resolver is exiting; there may be | |
1215 | * outstanding materialization requests to clean up. | |
1216 | */ | |
1217 | nspace_resolver_exited(p); | |
1218 | } | |
1219 | ||
1220 | #if SYSV_SHM | |
1221 | /* Close ref SYSV Shared memory*/ | |
1222 | if (p->vm_shm) { | |
1223 | shmexit(p); | |
1224 | } | |
1225 | #endif | |
1226 | #if SYSV_SEM | |
1227 | /* Release SYSV semaphores */ | |
1228 | semexit(p); | |
1229 | #endif | |
1230 | ||
1231 | #if PSYNCH | |
1232 | pth_proc_hashdelete(p); | |
1233 | #endif /* PSYNCH */ | |
1234 | ||
1235 | sessp = proc_session(p); | |
1236 | if (SESS_LEADER(p, sessp)) { | |
1237 | if (sessp->s_ttyvp != NULLVP) { | |
1238 | struct vnode *ttyvp; | |
1239 | int ttyvid; | |
1240 | int cttyflag = 0; | |
1241 | struct vfs_context context; | |
1242 | struct tty *tp; | |
1243 | ||
1244 | /* | |
1245 | * Controlling process. | |
1246 | * Signal foreground pgrp, | |
1247 | * drain controlling terminal | |
1248 | * and revoke access to controlling terminal. | |
1249 | */ | |
1250 | session_lock(sessp); | |
1251 | tp = SESSION_TP(sessp); | |
1252 | if ((tp != TTY_NULL) && (tp->t_session == sessp)) { | |
1253 | session_unlock(sessp); | |
1254 | ||
1255 | /* | |
1256 | * We're going to SIGHUP the foreground process | |
1257 | * group. It can't change from this point on | |
1258 | * until the revoke is complete. | |
1259 | * The process group changes under both the tty | |
1260 | * lock and proc_list_lock but we need only one | |
1261 | */ | |
1262 | tty_lock(tp); | |
1263 | ttysetpgrphup(tp); | |
1264 | tty_unlock(tp); | |
1265 | ||
1266 | tty_pgsignal(tp, SIGHUP, 1); | |
1267 | ||
1268 | session_lock(sessp); | |
1269 | tp = SESSION_TP(sessp); | |
1270 | } | |
1271 | cttyflag = sessp->s_flags & S_CTTYREF; | |
1272 | sessp->s_flags &= ~S_CTTYREF; | |
1273 | ttyvp = sessp->s_ttyvp; | |
1274 | ttyvid = sessp->s_ttyvid; | |
1275 | sessp->s_ttyvp = NULLVP; | |
1276 | sessp->s_ttyvid = 0; | |
1277 | sessp->s_ttyp = TTY_NULL; | |
1278 | sessp->s_ttypgrpid = NO_PID; | |
1279 | session_unlock(sessp); | |
1280 | ||
1281 | if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) { | |
1282 | if (tp != TTY_NULL) { | |
1283 | tty_lock(tp); | |
1284 | (void) ttywait(tp); | |
1285 | tty_unlock(tp); | |
1286 | } | |
1287 | context.vc_thread = proc_thread(p); /* XXX */ | |
1288 | context.vc_ucred = kauth_cred_proc_ref(p); | |
1289 | VNOP_REVOKE(ttyvp, REVOKEALL, &context); | |
1290 | if (cttyflag) { | |
1291 | /* | |
1292 | * Release the extra usecount taken in cttyopen. | |
1293 | * usecount should be released after VNOP_REVOKE is called. | |
1294 | * This usecount was taken to ensure that | |
1295 | * the VNOP_REVOKE results in a close to | |
1296 | * the tty since cttyclose is a no-op. | |
1297 | */ | |
1298 | vnode_rele(ttyvp); | |
1299 | } | |
1300 | vnode_put(ttyvp); | |
1301 | kauth_cred_unref(&context.vc_ucred); | |
1302 | ttyvp = NULLVP; | |
1303 | } | |
1304 | if (tp) { | |
1305 | /* | |
1306 | * This is cleared even if not set. This is also done in | |
1307 | * spec_close to ensure that the flag is cleared. | |
1308 | */ | |
1309 | tty_lock(tp); | |
1310 | ttyclrpgrphup(tp); | |
1311 | tty_unlock(tp); | |
1312 | ||
1313 | ttyfree(tp); | |
1314 | } | |
1315 | } | |
1316 | session_lock(sessp); | |
1317 | sessp->s_leader = NULL; | |
1318 | session_unlock(sessp); | |
1319 | } | |
1320 | session_rele(sessp); | |
1321 | ||
1322 | pg = proc_pgrp(p); | |
1323 | fixjobc(p, pg, 0); | |
1324 | pg_rele(pg); | |
1325 | ||
1326 | p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; | |
1327 | (void)acct_process(p); | |
1328 | ||
1329 | proc_list_lock(); | |
1330 | ||
1331 | if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT) { | |
1332 | p->p_listflag &= ~P_LIST_EXITCOUNT; | |
1333 | proc_shutdown_exitcount--; | |
1334 | if (proc_shutdown_exitcount == 0) { | |
1335 | wakeup(&proc_shutdown_exitcount); | |
1336 | } | |
1337 | } | |
1338 | ||
1339 | /* wait till parentrefs are dropped and grant no more */ | |
1340 | proc_childdrainstart(p); | |
1341 | while ((q = p->p_children.lh_first) != NULL) { | |
1342 | int reparentedtoinit = (q->p_listflag & P_LIST_DEADPARENT) ? 1 : 0; | |
1343 | if (q->p_stat == SZOMB) { | |
1344 | if (p != q->p_pptr) { | |
1345 | panic("parent child linkage broken"); | |
1346 | } | |
1347 | /* check for sysctl zomb lookup */ | |
1348 | while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) { | |
1349 | msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0); | |
1350 | } | |
1351 | q->p_listflag |= P_LIST_WAITING; | |
1352 | /* | |
1353 | * This is a named reference and it is not granted | |
1354 | * if the reap is already in progress. So we get | |
1355 | * the reference here exclusively and their can be | |
1356 | * no waiters. So there is no need for a wakeup | |
1357 | * after we are done. Also the reap frees the structure | |
1358 | * and the proc struct cannot be used for wakeups as well. | |
1359 | * It is safe to use q here as this is system reap | |
1360 | */ | |
1361 | (void)reap_child_locked(p, q, 1, reparentedtoinit, 1, 0); | |
1362 | } else { | |
1363 | /* | |
1364 | * Traced processes are killed | |
1365 | * since their existence means someone is messing up. | |
1366 | */ | |
1367 | if (q->p_lflag & P_LTRACED) { | |
1368 | struct proc *opp; | |
1369 | ||
1370 | /* | |
1371 | * Take a reference on the child process to | |
1372 | * ensure it doesn't exit and disappear between | |
1373 | * the time we drop the list_lock and attempt | |
1374 | * to acquire its proc_lock. | |
1375 | */ | |
1376 | if (proc_ref_locked(q) != q) { | |
1377 | continue; | |
1378 | } | |
1379 | ||
1380 | proc_list_unlock(); | |
1381 | ||
1382 | opp = proc_find(q->p_oppid); | |
1383 | if (opp != PROC_NULL) { | |
1384 | proc_list_lock(); | |
1385 | q->p_oppid = 0; | |
1386 | proc_list_unlock(); | |
1387 | proc_reparentlocked(q, opp, 0, 0); | |
1388 | proc_rele(opp); | |
1389 | } else { | |
1390 | /* original parent exited while traced */ | |
1391 | proc_list_lock(); | |
1392 | q->p_listflag |= P_LIST_DEADPARENT; | |
1393 | q->p_oppid = 0; | |
1394 | proc_list_unlock(); | |
1395 | proc_reparentlocked(q, initproc, 0, 0); | |
1396 | } | |
1397 | ||
1398 | proc_lock(q); | |
1399 | q->p_lflag &= ~P_LTRACED; | |
1400 | ||
1401 | if (q->sigwait_thread) { | |
1402 | thread_t thread = q->sigwait_thread; | |
1403 | ||
1404 | proc_unlock(q); | |
1405 | /* | |
1406 | * The sigwait_thread could be stopped at a | |
1407 | * breakpoint. Wake it up to kill. | |
1408 | * Need to do this as it could be a thread which is not | |
1409 | * the first thread in the task. So any attempts to kill | |
1410 | * the process would result into a deadlock on q->sigwait. | |
1411 | */ | |
1412 | thread_resume(thread); | |
1413 | clear_wait(thread, THREAD_INTERRUPTED); | |
1414 | threadsignal(thread, SIGKILL, 0, TRUE); | |
1415 | } else { | |
1416 | proc_unlock(q); | |
1417 | } | |
1418 | ||
1419 | psignal(q, SIGKILL); | |
1420 | proc_list_lock(); | |
1421 | proc_rele_locked(q); | |
1422 | } else { | |
1423 | q->p_listflag |= P_LIST_DEADPARENT; | |
1424 | proc_reparentlocked(q, initproc, 0, 1); | |
1425 | } | |
1426 | } | |
1427 | } | |
1428 | ||
1429 | proc_childdrainend(p); | |
1430 | proc_list_unlock(); | |
1431 | ||
1432 | #if CONFIG_MACF | |
1433 | /* | |
1434 | * Notify MAC policies that proc is dead. | |
1435 | * This should be replaced with proper label management | |
1436 | * (rdar://problem/32126399). | |
1437 | */ | |
1438 | mac_proc_notify_exit(p); | |
1439 | #endif | |
1440 | ||
1441 | /* | |
1442 | * Release reference to text vnode | |
1443 | */ | |
1444 | tvp = p->p_textvp; | |
1445 | p->p_textvp = NULL; | |
1446 | if (tvp != NULLVP) { | |
1447 | vnode_rele(tvp); | |
1448 | } | |
1449 | ||
1450 | /* | |
1451 | * Save exit status and final rusage info, adding in child rusage | |
1452 | * info and self times. If we were unable to allocate a zombie | |
1453 | * structure, this information is lost. | |
1454 | */ | |
1455 | if (p->p_ru != NULL) { | |
1456 | calcru(p, &p->p_stats->p_ru.ru_utime, &p->p_stats->p_ru.ru_stime, NULL); | |
1457 | p->p_ru->ru = p->p_stats->p_ru; | |
1458 | ||
1459 | ruadd(&(p->p_ru->ru), &p->p_stats->p_cru); | |
1460 | } | |
1461 | ||
1462 | /* | |
1463 | * Free up profiling buffers. | |
1464 | */ | |
1465 | { | |
1466 | struct uprof *p0 = &p->p_stats->p_prof, *p1, *pn; | |
1467 | ||
1468 | p1 = p0->pr_next; | |
1469 | p0->pr_next = NULL; | |
1470 | p0->pr_scale = 0; | |
1471 | ||
1472 | for (; p1 != NULL; p1 = pn) { | |
1473 | pn = p1->pr_next; | |
1474 | kfree(p1, sizeof *p1); | |
1475 | } | |
1476 | } | |
1477 | ||
1478 | proc_free_realitimer(p); | |
1479 | ||
1480 | /* | |
1481 | * Other substructures are freed from wait(). | |
1482 | */ | |
1483 | FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS); | |
1484 | p->p_stats = NULL; | |
1485 | ||
1486 | FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS); | |
1487 | p->p_sigacts = NULL; | |
1488 | ||
1489 | proc_limitdrop(p, 1); | |
1490 | p->p_limit = NULL; | |
1491 | ||
1492 | /* | |
1493 | * Finish up by terminating the task | |
1494 | * and halt this thread (only if a | |
1495 | * member of the task exiting). | |
1496 | */ | |
1497 | p->task = TASK_NULL; | |
1498 | set_bsdtask_info(task, NULL); | |
1499 | ||
1500 | knote_hint = NOTE_EXIT | (p->p_xstat & 0xffff); | |
1501 | proc_knote(p, knote_hint); | |
1502 | ||
1503 | /* mark the thread as the one that is doing proc_exit | |
1504 | * no need to hold proc lock in uthread_free | |
1505 | */ | |
1506 | uth->uu_flag |= UT_PROCEXIT; | |
1507 | /* | |
1508 | * Notify parent that we're gone. | |
1509 | */ | |
1510 | pp = proc_parent(p); | |
1511 | if (pp->p_flag & P_NOCLDWAIT) { | |
1512 | if (p->p_ru != NULL) { | |
1513 | proc_lock(pp); | |
1514 | #if 3839178 | |
1515 | /* | |
1516 | * If the parent is ignoring SIGCHLD, then POSIX requires | |
1517 | * us to not add the resource usage to the parent process - | |
1518 | * we are only going to hand it off to init to get reaped. | |
1519 | * We should contest the standard in this case on the basis | |
1520 | * of RLIMIT_CPU. | |
1521 | */ | |
1522 | #else /* !3839178 */ | |
1523 | /* | |
1524 | * Add child resource usage to parent before giving | |
1525 | * zombie to init. If we were unable to allocate a | |
1526 | * zombie structure, this information is lost. | |
1527 | */ | |
1528 | ruadd(&pp->p_stats->p_cru, &p->p_ru->ru); | |
1529 | #endif /* !3839178 */ | |
1530 | update_rusage_info_child(&pp->p_stats->ri_child, &p->p_ru->ri); | |
1531 | proc_unlock(pp); | |
1532 | } | |
1533 | ||
1534 | /* kernel can reap this one, no need to move it to launchd */ | |
1535 | proc_list_lock(); | |
1536 | p->p_listflag |= P_LIST_DEADPARENT; | |
1537 | proc_list_unlock(); | |
1538 | } | |
1539 | if ((p->p_listflag & P_LIST_DEADPARENT) == 0 || p->p_oppid) { | |
1540 | if (pp != initproc) { | |
1541 | proc_lock(pp); | |
1542 | pp->si_pid = p->p_pid; | |
1543 | pp->p_xhighbits = p->p_xhighbits; | |
1544 | p->p_xhighbits = 0; | |
1545 | pp->si_status = p->p_xstat; | |
1546 | pp->si_code = CLD_EXITED; | |
1547 | /* | |
1548 | * p_ucred usage is safe as it is an exiting process | |
1549 | * and reference is dropped in reap | |
1550 | */ | |
1551 | pp->si_uid = kauth_cred_getruid(p->p_ucred); | |
1552 | proc_unlock(pp); | |
1553 | } | |
1554 | /* mark as a zombie */ | |
1555 | /* No need to take proc lock as all refs are drained and | |
1556 | * no one except parent (reaping ) can look at this. | |
1557 | * The write is to an int and is coherent. Also parent is | |
1558 | * keyed off of list lock for reaping | |
1559 | */ | |
1560 | DTRACE_PROC2(exited, proc_t, p, int, exitval); | |
1561 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON, | |
1562 | BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END, | |
1563 | pid, exitval, 0, 0, 0); | |
1564 | p->p_stat = SZOMB; | |
1565 | /* | |
1566 | * The current process can be reaped so, no one | |
1567 | * can depend on this | |
1568 | */ | |
1569 | ||
1570 | psignal(pp, SIGCHLD); | |
1571 | ||
1572 | /* and now wakeup the parent */ | |
1573 | proc_list_lock(); | |
1574 | wakeup((caddr_t)pp); | |
1575 | proc_list_unlock(); | |
1576 | } else { | |
1577 | /* should be fine as parent proc would be initproc */ | |
1578 | /* mark as a zombie */ | |
1579 | /* No need to take proc lock as all refs are drained and | |
1580 | * no one except parent (reaping ) can look at this. | |
1581 | * The write is to an int and is coherent. Also parent is | |
1582 | * keyed off of list lock for reaping | |
1583 | */ | |
1584 | DTRACE_PROC2(exited, proc_t, p, int, exitval); | |
1585 | proc_list_lock(); | |
1586 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON, | |
1587 | BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END, | |
1588 | pid, exitval, 0, 0, 0); | |
1589 | /* check for sysctl zomb lookup */ | |
1590 | while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) { | |
1591 | msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0); | |
1592 | } | |
1593 | /* safe to use p as this is a system reap */ | |
1594 | p->p_stat = SZOMB; | |
1595 | p->p_listflag |= P_LIST_WAITING; | |
1596 | ||
1597 | /* | |
1598 | * This is a named reference and it is not granted | |
1599 | * if the reap is already in progress. So we get | |
1600 | * the reference here exclusively and their can be | |
1601 | * no waiters. So there is no need for a wakeup | |
1602 | * after we are done. AlsO the reap frees the structure | |
1603 | * and the proc struct cannot be used for wakeups as well. | |
1604 | * It is safe to use p here as this is system reap | |
1605 | */ | |
1606 | (void)reap_child_locked(pp, p, 1, 0, 1, 1); | |
1607 | /* list lock dropped by reap_child_locked */ | |
1608 | } | |
1609 | if (uth->uu_lowpri_window) { | |
1610 | /* | |
1611 | * task is marked as a low priority I/O type and we've | |
1612 | * somehow picked up another throttle during exit processing... | |
1613 | * no need to throttle this thread since its going away | |
1614 | * but we do need to update our bookeeping w/r to throttled threads | |
1615 | */ | |
1616 | throttle_lowpri_io(0); | |
1617 | } | |
1618 | ||
1619 | proc_rele(pp); | |
1620 | } | |
1621 | ||
1622 | ||
1623 | /* | |
1624 | * reap_child_locked | |
1625 | * | |
1626 | * Description: Given a process from which all status information needed | |
1627 | * has already been extracted, if the process is a ptrace | |
1628 | * attach process, detach it and give it back to its real | |
1629 | * parent, else recover all resources remaining associated | |
1630 | * with it. | |
1631 | * | |
1632 | * Parameters: proc_t parent Parent of process being reaped | |
1633 | * proc_t child Process to reap | |
1634 | * | |
1635 | * Returns: 0 Process was not reaped because it | |
1636 | * came from an attach | |
1637 | * 1 Process was reaped | |
1638 | */ | |
1639 | static int | |
1640 | reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock) | |
1641 | { | |
1642 | proc_t trace_parent = PROC_NULL; /* Traced parent process, if tracing */ | |
1643 | ||
1644 | if (locked == 1) { | |
1645 | proc_list_unlock(); | |
1646 | } | |
1647 | ||
1648 | /* | |
1649 | * If we got the child via a ptrace 'attach', | |
1650 | * we need to give it back to the old parent. | |
1651 | * | |
1652 | * Exception: someone who has been reparented to launchd before being | |
1653 | * ptraced can simply be reaped, refer to radar 5677288 | |
1654 | * p_oppid -> ptraced | |
1655 | * trace_parent == initproc -> away from launchd | |
1656 | * reparentedtoinit -> came to launchd by reparenting | |
1657 | */ | |
1658 | if (child->p_oppid) { | |
1659 | int knote_hint; | |
1660 | pid_t oppid; | |
1661 | ||
1662 | proc_lock(child); | |
1663 | oppid = child->p_oppid; | |
1664 | child->p_oppid = 0; | |
1665 | knote_hint = NOTE_EXIT | (child->p_xstat & 0xffff); | |
1666 | proc_unlock(child); | |
1667 | ||
1668 | if ((trace_parent = proc_find(oppid)) | |
1669 | && !((trace_parent == initproc) && reparentedtoinit)) { | |
1670 | if (trace_parent != initproc) { | |
1671 | /* | |
1672 | * proc internal fileds and p_ucred usage safe | |
1673 | * here as child is dead and is not reaped or | |
1674 | * reparented yet | |
1675 | */ | |
1676 | proc_lock(trace_parent); | |
1677 | trace_parent->si_pid = child->p_pid; | |
1678 | trace_parent->si_status = child->p_xstat; | |
1679 | trace_parent->si_code = CLD_CONTINUED; | |
1680 | trace_parent->si_uid = kauth_cred_getruid(child->p_ucred); | |
1681 | proc_unlock(trace_parent); | |
1682 | } | |
1683 | proc_reparentlocked(child, trace_parent, 1, 0); | |
1684 | ||
1685 | /* resend knote to original parent (and others) after reparenting */ | |
1686 | proc_knote(child, knote_hint); | |
1687 | ||
1688 | psignal(trace_parent, SIGCHLD); | |
1689 | proc_list_lock(); | |
1690 | wakeup((caddr_t)trace_parent); | |
1691 | child->p_listflag &= ~P_LIST_WAITING; | |
1692 | wakeup(&child->p_stat); | |
1693 | proc_list_unlock(); | |
1694 | proc_rele(trace_parent); | |
1695 | if ((locked == 1) && (droplock == 0)) { | |
1696 | proc_list_lock(); | |
1697 | } | |
1698 | return 0; | |
1699 | } | |
1700 | ||
1701 | /* | |
1702 | * If we can't reparent (e.g. the original parent exited while child was being debugged, or | |
1703 | * original parent is the same as the debugger currently exiting), we still need to satisfy | |
1704 | * the knote lifecycle for other observers on the system. While the debugger was attached, | |
1705 | * the NOTE_EXIT would not have been broadcast during initial child termination. | |
1706 | */ | |
1707 | proc_knote(child, knote_hint); | |
1708 | ||
1709 | if (trace_parent != PROC_NULL) { | |
1710 | proc_rele(trace_parent); | |
1711 | } | |
1712 | } | |
1713 | ||
1714 | #pragma clang diagnostic push | |
1715 | #pragma clang diagnostic ignored "-Wdeprecated-declarations" | |
1716 | proc_knote(child, NOTE_REAP); | |
1717 | #pragma clang diagnostic pop | |
1718 | ||
1719 | proc_knote_drain(child); | |
1720 | ||
1721 | child->p_xstat = 0; | |
1722 | if (child->p_ru) { | |
1723 | proc_lock(parent); | |
1724 | #if 3839178 | |
1725 | /* | |
1726 | * If the parent is ignoring SIGCHLD, then POSIX requires | |
1727 | * us to not add the resource usage to the parent process - | |
1728 | * we are only going to hand it off to init to get reaped. | |
1729 | * We should contest the standard in this case on the basis | |
1730 | * of RLIMIT_CPU. | |
1731 | */ | |
1732 | if (!(parent->p_flag & P_NOCLDWAIT)) | |
1733 | #endif /* 3839178 */ | |
1734 | ruadd(&parent->p_stats->p_cru, &child->p_ru->ru); | |
1735 | update_rusage_info_child(&parent->p_stats->ri_child, &child->p_ru->ri); | |
1736 | proc_unlock(parent); | |
1737 | FREE_ZONE(child->p_ru, sizeof *child->p_ru, M_ZOMBIE); | |
1738 | child->p_ru = NULL; | |
1739 | } else { | |
1740 | printf("Warning : lost p_ru for %s\n", child->p_comm); | |
1741 | } | |
1742 | ||
1743 | AUDIT_SESSION_PROCEXIT(child); | |
1744 | ||
1745 | /* | |
1746 | * Decrement the count of procs running with this uid. | |
1747 | * p_ucred usage is safe here as it is an exited process. | |
1748 | * and refernce is dropped after these calls down below | |
1749 | * (locking protection is provided by list lock held in chgproccnt) | |
1750 | */ | |
1751 | #if CONFIG_PERSONAS | |
1752 | /* | |
1753 | * persona_proc_drop calls chgproccnt(-1) on the persona uid, | |
1754 | * and (+1) on the child->p_ucred uid | |
1755 | */ | |
1756 | persona_proc_drop(child); | |
1757 | #endif | |
1758 | (void)chgproccnt(kauth_cred_getruid(child->p_ucred), -1); | |
1759 | ||
1760 | os_reason_free(child->p_exit_reason); | |
1761 | ||
1762 | /* | |
1763 | * Free up credentials. | |
1764 | */ | |
1765 | if (IS_VALID_CRED(child->p_ucred)) { | |
1766 | kauth_cred_unref(&child->p_ucred); | |
1767 | } | |
1768 | ||
1769 | /* XXXX Note NOT SAFE TO USE p_ucred from this point onwards */ | |
1770 | ||
1771 | /* | |
1772 | * Finally finished with old proc entry. | |
1773 | * Unlink it from its process group and free it. | |
1774 | */ | |
1775 | leavepgrp(child); | |
1776 | ||
1777 | proc_list_lock(); | |
1778 | LIST_REMOVE(child, p_list); /* off zombproc */ | |
1779 | parent->p_childrencnt--; | |
1780 | LIST_REMOVE(child, p_sibling); | |
1781 | /* If there are no more children wakeup parent */ | |
1782 | if ((deadparent != 0) && (LIST_EMPTY(&parent->p_children))) { | |
1783 | wakeup((caddr_t)parent); /* with list lock held */ | |
1784 | } | |
1785 | child->p_listflag &= ~P_LIST_WAITING; | |
1786 | wakeup(&child->p_stat); | |
1787 | ||
1788 | /* Take it out of process hash */ | |
1789 | LIST_REMOVE(child, p_hash); | |
1790 | child->p_listflag &= ~P_LIST_INHASH; | |
1791 | proc_checkdeadrefs(child); | |
1792 | nprocs--; | |
1793 | ||
1794 | if (deadparent) { | |
1795 | /* | |
1796 | * If a child zombie is being reaped because its parent | |
1797 | * is exiting, make sure we update the list flag | |
1798 | */ | |
1799 | child->p_listflag |= P_LIST_DEADPARENT; | |
1800 | } | |
1801 | ||
1802 | proc_list_unlock(); | |
1803 | ||
1804 | ||
1805 | lck_mtx_destroy(&child->p_mlock, proc_mlock_grp); | |
1806 | lck_mtx_destroy(&child->p_ucred_mlock, proc_ucred_mlock_grp); | |
1807 | lck_mtx_destroy(&child->p_fdmlock, proc_fdmlock_grp); | |
1808 | #if CONFIG_DTRACE | |
1809 | lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp); | |
1810 | #endif | |
1811 | lck_spin_destroy(&child->p_slock, proc_slock_grp); | |
1812 | ||
1813 | FREE_ZONE(child, sizeof *child, M_PROC); | |
1814 | if ((locked == 1) && (droplock == 0)) { | |
1815 | proc_list_lock(); | |
1816 | } | |
1817 | ||
1818 | return 1; | |
1819 | } | |
1820 | ||
1821 | ||
1822 | int | |
1823 | wait1continue(int result) | |
1824 | { | |
1825 | proc_t p; | |
1826 | thread_t thread; | |
1827 | uthread_t uth; | |
1828 | struct _wait4_data *wait4_data; | |
1829 | struct wait4_nocancel_args *uap; | |
1830 | int *retval; | |
1831 | ||
1832 | if (result) { | |
1833 | return result; | |
1834 | } | |
1835 | ||
1836 | p = current_proc(); | |
1837 | thread = current_thread(); | |
1838 | uth = (struct uthread *)get_bsdthread_info(thread); | |
1839 | ||
1840 | wait4_data = &uth->uu_save.uus_wait4_data; | |
1841 | uap = wait4_data->args; | |
1842 | retval = wait4_data->retval; | |
1843 | return wait4_nocancel(p, uap, retval); | |
1844 | } | |
1845 | ||
1846 | int | |
1847 | wait4(proc_t q, struct wait4_args *uap, int32_t *retval) | |
1848 | { | |
1849 | __pthread_testcancel(1); | |
1850 | return wait4_nocancel(q, (struct wait4_nocancel_args *)uap, retval); | |
1851 | } | |
1852 | ||
1853 | int | |
1854 | wait4_nocancel(proc_t q, struct wait4_nocancel_args *uap, int32_t *retval) | |
1855 | { | |
1856 | int nfound; | |
1857 | int sibling_count; | |
1858 | proc_t p; | |
1859 | int status, error; | |
1860 | uthread_t uth; | |
1861 | struct _wait4_data *wait4_data; | |
1862 | ||
1863 | AUDIT_ARG(pid, uap->pid); | |
1864 | ||
1865 | if (uap->pid == 0) { | |
1866 | uap->pid = -q->p_pgrpid; | |
1867 | } | |
1868 | ||
1869 | loop: | |
1870 | proc_list_lock(); | |
1871 | loop1: | |
1872 | nfound = 0; | |
1873 | sibling_count = 0; | |
1874 | ||
1875 | PCHILDREN_FOREACH(q, p) { | |
1876 | if (p->p_sibling.le_next != 0) { | |
1877 | sibling_count++; | |
1878 | } | |
1879 | if (uap->pid != WAIT_ANY && | |
1880 | p->p_pid != uap->pid && | |
1881 | p->p_pgrpid != -(uap->pid)) { | |
1882 | continue; | |
1883 | } | |
1884 | ||
1885 | nfound++; | |
1886 | ||
1887 | /* XXX This is racy because we don't get the lock!!!! */ | |
1888 | ||
1889 | if (p->p_listflag & P_LIST_WAITING) { | |
1890 | /* we're not using a continuation here but we still need to stash | |
1891 | * the args for stackshot. */ | |
1892 | uth = current_uthread(); | |
1893 | wait4_data = &uth->uu_save.uus_wait4_data; | |
1894 | wait4_data->args = uap; | |
1895 | thread_set_pending_block_hint(current_thread(), kThreadWaitOnProcess); | |
1896 | ||
1897 | (void)msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0); | |
1898 | goto loop1; | |
1899 | } | |
1900 | p->p_listflag |= P_LIST_WAITING; /* only allow single thread to wait() */ | |
1901 | ||
1902 | ||
1903 | if (p->p_stat == SZOMB) { | |
1904 | int reparentedtoinit = (p->p_listflag & P_LIST_DEADPARENT) ? 1 : 0; | |
1905 | ||
1906 | proc_list_unlock(); | |
1907 | #if CONFIG_MACF | |
1908 | if ((error = mac_proc_check_wait(q, p)) != 0) { | |
1909 | goto out; | |
1910 | } | |
1911 | #endif | |
1912 | retval[0] = p->p_pid; | |
1913 | if (uap->status) { | |
1914 | /* Legacy apps expect only 8 bits of status */ | |
1915 | status = 0xffff & p->p_xstat; /* convert to int */ | |
1916 | error = copyout((caddr_t)&status, | |
1917 | uap->status, | |
1918 | sizeof(status)); | |
1919 | if (error) { | |
1920 | goto out; | |
1921 | } | |
1922 | } | |
1923 | if (uap->rusage) { | |
1924 | if (p->p_ru == NULL) { | |
1925 | error = ENOMEM; | |
1926 | } else { | |
1927 | if (IS_64BIT_PROCESS(q)) { | |
1928 | struct user64_rusage my_rusage = {}; | |
1929 | munge_user64_rusage(&p->p_ru->ru, &my_rusage); | |
1930 | error = copyout((caddr_t)&my_rusage, | |
1931 | uap->rusage, | |
1932 | sizeof(my_rusage)); | |
1933 | } else { | |
1934 | struct user32_rusage my_rusage = {}; | |
1935 | munge_user32_rusage(&p->p_ru->ru, &my_rusage); | |
1936 | error = copyout((caddr_t)&my_rusage, | |
1937 | uap->rusage, | |
1938 | sizeof(my_rusage)); | |
1939 | } | |
1940 | } | |
1941 | /* information unavailable? */ | |
1942 | if (error) { | |
1943 | goto out; | |
1944 | } | |
1945 | } | |
1946 | ||
1947 | /* Conformance change for 6577252. | |
1948 | * When SIGCHLD is blocked and wait() returns because the status | |
1949 | * of a child process is available and there are no other | |
1950 | * children processes, then any pending SIGCHLD signal is cleared. | |
1951 | */ | |
1952 | if (sibling_count == 0) { | |
1953 | int mask = sigmask(SIGCHLD); | |
1954 | uth = current_uthread(); | |
1955 | ||
1956 | if ((uth->uu_sigmask & mask) != 0) { | |
1957 | /* we are blocking SIGCHLD signals. clear any pending SIGCHLD. | |
1958 | * This locking looks funny but it is protecting access to the | |
1959 | * thread via p_uthlist. | |
1960 | */ | |
1961 | proc_lock(q); | |
1962 | uth->uu_siglist &= ~mask; /* clear pending signal */ | |
1963 | proc_unlock(q); | |
1964 | } | |
1965 | } | |
1966 | ||
1967 | /* Clean up */ | |
1968 | (void)reap_child_locked(q, p, 0, reparentedtoinit, 0, 0); | |
1969 | ||
1970 | return 0; | |
1971 | } | |
1972 | if (p->p_stat == SSTOP && (p->p_lflag & P_LWAITED) == 0 && | |
1973 | (p->p_lflag & P_LTRACED || uap->options & WUNTRACED)) { | |
1974 | proc_list_unlock(); | |
1975 | #if CONFIG_MACF | |
1976 | if ((error = mac_proc_check_wait(q, p)) != 0) { | |
1977 | goto out; | |
1978 | } | |
1979 | #endif | |
1980 | proc_lock(p); | |
1981 | p->p_lflag |= P_LWAITED; | |
1982 | proc_unlock(p); | |
1983 | retval[0] = p->p_pid; | |
1984 | if (uap->status) { | |
1985 | status = W_STOPCODE(p->p_xstat); | |
1986 | error = copyout((caddr_t)&status, | |
1987 | uap->status, | |
1988 | sizeof(status)); | |
1989 | } else { | |
1990 | error = 0; | |
1991 | } | |
1992 | goto out; | |
1993 | } | |
1994 | /* | |
1995 | * If we are waiting for continued processses, and this | |
1996 | * process was continued | |
1997 | */ | |
1998 | if ((uap->options & WCONTINUED) && | |
1999 | (p->p_flag & P_CONTINUED)) { | |
2000 | proc_list_unlock(); | |
2001 | #if CONFIG_MACF | |
2002 | if ((error = mac_proc_check_wait(q, p)) != 0) { | |
2003 | goto out; | |
2004 | } | |
2005 | #endif | |
2006 | ||
2007 | /* Prevent other process for waiting for this event */ | |
2008 | OSBitAndAtomic(~((uint32_t)P_CONTINUED), &p->p_flag); | |
2009 | retval[0] = p->p_pid; | |
2010 | if (uap->status) { | |
2011 | status = W_STOPCODE(SIGCONT); | |
2012 | error = copyout((caddr_t)&status, | |
2013 | uap->status, | |
2014 | sizeof(status)); | |
2015 | } else { | |
2016 | error = 0; | |
2017 | } | |
2018 | goto out; | |
2019 | } | |
2020 | p->p_listflag &= ~P_LIST_WAITING; | |
2021 | wakeup(&p->p_stat); | |
2022 | } | |
2023 | /* list lock is held when we get here any which way */ | |
2024 | if (nfound == 0) { | |
2025 | proc_list_unlock(); | |
2026 | return ECHILD; | |
2027 | } | |
2028 | ||
2029 | if (uap->options & WNOHANG) { | |
2030 | retval[0] = 0; | |
2031 | proc_list_unlock(); | |
2032 | return 0; | |
2033 | } | |
2034 | ||
2035 | /* Save arguments for continuation. Backing storage is in uthread->uu_arg, and will not be deallocated */ | |
2036 | uth = current_uthread(); | |
2037 | wait4_data = &uth->uu_save.uus_wait4_data; | |
2038 | wait4_data->args = uap; | |
2039 | wait4_data->retval = retval; | |
2040 | ||
2041 | thread_set_pending_block_hint(current_thread(), kThreadWaitOnProcess); | |
2042 | if ((error = msleep0((caddr_t)q, proc_list_mlock, PWAIT | PCATCH | PDROP, "wait", 0, wait1continue))) { | |
2043 | return error; | |
2044 | } | |
2045 | ||
2046 | goto loop; | |
2047 | out: | |
2048 | proc_list_lock(); | |
2049 | p->p_listflag &= ~P_LIST_WAITING; | |
2050 | wakeup(&p->p_stat); | |
2051 | proc_list_unlock(); | |
2052 | return error; | |
2053 | } | |
2054 | ||
2055 | #if DEBUG | |
2056 | #define ASSERT_LCK_MTX_OWNED(lock) \ | |
2057 | lck_mtx_assert(lock, LCK_MTX_ASSERT_OWNED) | |
2058 | #else | |
2059 | #define ASSERT_LCK_MTX_OWNED(lock) /* nothing */ | |
2060 | #endif | |
2061 | ||
2062 | int | |
2063 | waitidcontinue(int result) | |
2064 | { | |
2065 | proc_t p; | |
2066 | thread_t thread; | |
2067 | uthread_t uth; | |
2068 | struct _waitid_data *waitid_data; | |
2069 | struct waitid_nocancel_args *uap; | |
2070 | int *retval; | |
2071 | ||
2072 | if (result) { | |
2073 | return result; | |
2074 | } | |
2075 | ||
2076 | p = current_proc(); | |
2077 | thread = current_thread(); | |
2078 | uth = (struct uthread *)get_bsdthread_info(thread); | |
2079 | ||
2080 | waitid_data = &uth->uu_save.uus_waitid_data; | |
2081 | uap = waitid_data->args; | |
2082 | retval = waitid_data->retval; | |
2083 | return waitid_nocancel(p, uap, retval); | |
2084 | } | |
2085 | ||
2086 | /* | |
2087 | * Description: Suspend the calling thread until one child of the process | |
2088 | * containing the calling thread changes state. | |
2089 | * | |
2090 | * Parameters: uap->idtype one of P_PID, P_PGID, P_ALL | |
2091 | * uap->id pid_t or gid_t or ignored | |
2092 | * uap->infop Address of siginfo_t struct in | |
2093 | * user space into which to return status | |
2094 | * uap->options flag values | |
2095 | * | |
2096 | * Returns: 0 Success | |
2097 | * !0 Error returning status to user space | |
2098 | */ | |
2099 | int | |
2100 | waitid(proc_t q, struct waitid_args *uap, int32_t *retval) | |
2101 | { | |
2102 | __pthread_testcancel(1); | |
2103 | return waitid_nocancel(q, (struct waitid_nocancel_args *)uap, retval); | |
2104 | } | |
2105 | ||
2106 | int | |
2107 | waitid_nocancel(proc_t q, struct waitid_nocancel_args *uap, | |
2108 | __unused int32_t *retval) | |
2109 | { | |
2110 | user_siginfo_t siginfo; /* siginfo data to return to caller */ | |
2111 | boolean_t caller64 = IS_64BIT_PROCESS(q); | |
2112 | int nfound; | |
2113 | proc_t p; | |
2114 | int error; | |
2115 | uthread_t uth; | |
2116 | struct _waitid_data *waitid_data; | |
2117 | ||
2118 | if (uap->options == 0 || | |
2119 | (uap->options & ~(WNOHANG | WNOWAIT | WCONTINUED | WSTOPPED | WEXITED))) { | |
2120 | return EINVAL; /* bits set that aren't recognized */ | |
2121 | } | |
2122 | switch (uap->idtype) { | |
2123 | case P_PID: /* child with process ID equal to... */ | |
2124 | case P_PGID: /* child with process group ID equal to... */ | |
2125 | if (((int)uap->id) < 0) { | |
2126 | return EINVAL; | |
2127 | } | |
2128 | break; | |
2129 | case P_ALL: /* any child */ | |
2130 | break; | |
2131 | } | |
2132 | ||
2133 | loop: | |
2134 | proc_list_lock(); | |
2135 | loop1: | |
2136 | nfound = 0; | |
2137 | ||
2138 | PCHILDREN_FOREACH(q, p) { | |
2139 | switch (uap->idtype) { | |
2140 | case P_PID: /* child with process ID equal to... */ | |
2141 | if (p->p_pid != (pid_t)uap->id) { | |
2142 | continue; | |
2143 | } | |
2144 | break; | |
2145 | case P_PGID: /* child with process group ID equal to... */ | |
2146 | if (p->p_pgrpid != (pid_t)uap->id) { | |
2147 | continue; | |
2148 | } | |
2149 | break; | |
2150 | case P_ALL: /* any child */ | |
2151 | break; | |
2152 | } | |
2153 | ||
2154 | /* XXX This is racy because we don't get the lock!!!! */ | |
2155 | ||
2156 | /* | |
2157 | * Wait collision; go to sleep and restart; used to maintain | |
2158 | * the single return for waited process guarantee. | |
2159 | */ | |
2160 | if (p->p_listflag & P_LIST_WAITING) { | |
2161 | (void) msleep(&p->p_stat, proc_list_mlock, | |
2162 | PWAIT, "waitidcoll", 0); | |
2163 | goto loop1; | |
2164 | } | |
2165 | p->p_listflag |= P_LIST_WAITING; /* mark busy */ | |
2166 | ||
2167 | nfound++; | |
2168 | ||
2169 | bzero(&siginfo, sizeof(siginfo)); | |
2170 | ||
2171 | switch (p->p_stat) { | |
2172 | case SZOMB: /* Exited */ | |
2173 | if (!(uap->options & WEXITED)) { | |
2174 | break; | |
2175 | } | |
2176 | proc_list_unlock(); | |
2177 | #if CONFIG_MACF | |
2178 | if ((error = mac_proc_check_wait(q, p)) != 0) { | |
2179 | goto out; | |
2180 | } | |
2181 | #endif | |
2182 | siginfo.si_signo = SIGCHLD; | |
2183 | siginfo.si_pid = p->p_pid; | |
2184 | siginfo.si_status = (WEXITSTATUS(p->p_xstat) & 0x00FFFFFF) | (((uint32_t)(p->p_xhighbits) << 24) & 0xFF000000); | |
2185 | p->p_xhighbits = 0; | |
2186 | if (WIFSIGNALED(p->p_xstat)) { | |
2187 | siginfo.si_code = WCOREDUMP(p->p_xstat) ? | |
2188 | CLD_DUMPED : CLD_KILLED; | |
2189 | } else { | |
2190 | siginfo.si_code = CLD_EXITED; | |
2191 | } | |
2192 | ||
2193 | if ((error = copyoutsiginfo(&siginfo, | |
2194 | caller64, uap->infop)) != 0) { | |
2195 | goto out; | |
2196 | } | |
2197 | ||
2198 | /* Prevent other process for waiting for this event? */ | |
2199 | if (!(uap->options & WNOWAIT)) { | |
2200 | (void) reap_child_locked(q, p, 0, 0, 0, 0); | |
2201 | return 0; | |
2202 | } | |
2203 | goto out; | |
2204 | ||
2205 | case SSTOP: /* Stopped */ | |
2206 | /* | |
2207 | * If we are not interested in stopped processes, then | |
2208 | * ignore this one. | |
2209 | */ | |
2210 | if (!(uap->options & WSTOPPED)) { | |
2211 | break; | |
2212 | } | |
2213 | ||
2214 | /* | |
2215 | * If someone has already waited it, we lost a race | |
2216 | * to be the one to return status. | |
2217 | */ | |
2218 | if ((p->p_lflag & P_LWAITED) != 0) { | |
2219 | break; | |
2220 | } | |
2221 | proc_list_unlock(); | |
2222 | #if CONFIG_MACF | |
2223 | if ((error = mac_proc_check_wait(q, p)) != 0) { | |
2224 | goto out; | |
2225 | } | |
2226 | #endif | |
2227 | siginfo.si_signo = SIGCHLD; | |
2228 | siginfo.si_pid = p->p_pid; | |
2229 | siginfo.si_status = p->p_xstat; /* signal number */ | |
2230 | siginfo.si_code = CLD_STOPPED; | |
2231 | ||
2232 | if ((error = copyoutsiginfo(&siginfo, | |
2233 | caller64, uap->infop)) != 0) { | |
2234 | goto out; | |
2235 | } | |
2236 | ||
2237 | /* Prevent other process for waiting for this event? */ | |
2238 | if (!(uap->options & WNOWAIT)) { | |
2239 | proc_lock(p); | |
2240 | p->p_lflag |= P_LWAITED; | |
2241 | proc_unlock(p); | |
2242 | } | |
2243 | goto out; | |
2244 | ||
2245 | default: /* All other states => Continued */ | |
2246 | if (!(uap->options & WCONTINUED)) { | |
2247 | break; | |
2248 | } | |
2249 | ||
2250 | /* | |
2251 | * If the flag isn't set, then this process has not | |
2252 | * been stopped and continued, or the status has | |
2253 | * already been reaped by another caller of waitid(). | |
2254 | */ | |
2255 | if ((p->p_flag & P_CONTINUED) == 0) { | |
2256 | break; | |
2257 | } | |
2258 | proc_list_unlock(); | |
2259 | #if CONFIG_MACF | |
2260 | if ((error = mac_proc_check_wait(q, p)) != 0) { | |
2261 | goto out; | |
2262 | } | |
2263 | #endif | |
2264 | siginfo.si_signo = SIGCHLD; | |
2265 | siginfo.si_code = CLD_CONTINUED; | |
2266 | proc_lock(p); | |
2267 | siginfo.si_pid = p->p_contproc; | |
2268 | siginfo.si_status = p->p_xstat; | |
2269 | proc_unlock(p); | |
2270 | ||
2271 | if ((error = copyoutsiginfo(&siginfo, | |
2272 | caller64, uap->infop)) != 0) { | |
2273 | goto out; | |
2274 | } | |
2275 | ||
2276 | /* Prevent other process for waiting for this event? */ | |
2277 | if (!(uap->options & WNOWAIT)) { | |
2278 | OSBitAndAtomic(~((uint32_t)P_CONTINUED), | |
2279 | &p->p_flag); | |
2280 | } | |
2281 | goto out; | |
2282 | } | |
2283 | ASSERT_LCK_MTX_OWNED(proc_list_mlock); | |
2284 | ||
2285 | /* Not a process we are interested in; go on to next child */ | |
2286 | ||
2287 | p->p_listflag &= ~P_LIST_WAITING; | |
2288 | wakeup(&p->p_stat); | |
2289 | } | |
2290 | ASSERT_LCK_MTX_OWNED(proc_list_mlock); | |
2291 | ||
2292 | /* No child processes that could possibly satisfy the request? */ | |
2293 | ||
2294 | if (nfound == 0) { | |
2295 | proc_list_unlock(); | |
2296 | return ECHILD; | |
2297 | } | |
2298 | ||
2299 | if (uap->options & WNOHANG) { | |
2300 | proc_list_unlock(); | |
2301 | #if CONFIG_MACF | |
2302 | if ((error = mac_proc_check_wait(q, p)) != 0) { | |
2303 | return error; | |
2304 | } | |
2305 | #endif | |
2306 | /* | |
2307 | * The state of the siginfo structure in this case | |
2308 | * is undefined. Some implementations bzero it, some | |
2309 | * (like here) leave it untouched for efficiency. | |
2310 | * | |
2311 | * Thus the most portable check for "no matching pid with | |
2312 | * WNOHANG" is to store a zero into si_pid before | |
2313 | * invocation, then check for a non-zero value afterwards. | |
2314 | */ | |
2315 | return 0; | |
2316 | } | |
2317 | ||
2318 | /* Save arguments for continuation. Backing storage is in uthread->uu_arg, and will not be deallocated */ | |
2319 | uth = current_uthread(); | |
2320 | waitid_data = &uth->uu_save.uus_waitid_data; | |
2321 | waitid_data->args = uap; | |
2322 | waitid_data->retval = retval; | |
2323 | ||
2324 | if ((error = msleep0(q, proc_list_mlock, | |
2325 | PWAIT | PCATCH | PDROP, "waitid", 0, waitidcontinue)) != 0) { | |
2326 | return error; | |
2327 | } | |
2328 | ||
2329 | goto loop; | |
2330 | out: | |
2331 | proc_list_lock(); | |
2332 | p->p_listflag &= ~P_LIST_WAITING; | |
2333 | wakeup(&p->p_stat); | |
2334 | proc_list_unlock(); | |
2335 | return error; | |
2336 | } | |
2337 | ||
2338 | /* | |
2339 | * make process 'parent' the new parent of process 'child'. | |
2340 | */ | |
2341 | void | |
2342 | proc_reparentlocked(proc_t child, proc_t parent, int signallable, int locked) | |
2343 | { | |
2344 | proc_t oldparent = PROC_NULL; | |
2345 | ||
2346 | if (child->p_pptr == parent) { | |
2347 | return; | |
2348 | } | |
2349 | ||
2350 | if (locked == 0) { | |
2351 | proc_list_lock(); | |
2352 | } | |
2353 | ||
2354 | oldparent = child->p_pptr; | |
2355 | #if __PROC_INTERNAL_DEBUG | |
2356 | if (oldparent == PROC_NULL) { | |
2357 | panic("proc_reparent: process %p does not have a parent\n", child); | |
2358 | } | |
2359 | #endif | |
2360 | ||
2361 | LIST_REMOVE(child, p_sibling); | |
2362 | #if __PROC_INTERNAL_DEBUG | |
2363 | if (oldparent->p_childrencnt == 0) { | |
2364 | panic("process children count already 0\n"); | |
2365 | } | |
2366 | #endif | |
2367 | oldparent->p_childrencnt--; | |
2368 | #if __PROC_INTERNAL_DEBUG | |
2369 | if (oldparent->p_childrencnt < 0) { | |
2370 | panic("process children count -ve\n"); | |
2371 | } | |
2372 | #endif | |
2373 | LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); | |
2374 | parent->p_childrencnt++; | |
2375 | child->p_pptr = parent; | |
2376 | child->p_ppid = parent->p_pid; | |
2377 | ||
2378 | proc_list_unlock(); | |
2379 | ||
2380 | if ((signallable != 0) && (initproc == parent) && (child->p_stat == SZOMB)) { | |
2381 | psignal(initproc, SIGCHLD); | |
2382 | } | |
2383 | if (locked == 1) { | |
2384 | proc_list_lock(); | |
2385 | } | |
2386 | } | |
2387 | ||
2388 | /* | |
2389 | * Exit: deallocate address space and other resources, change proc state | |
2390 | * to zombie, and unlink proc from allproc and parent's lists. Save exit | |
2391 | * status and rusage for wait(). Check for child processes and orphan them. | |
2392 | */ | |
2393 | ||
2394 | void | |
2395 | vfork_exit(proc_t p, int rv) | |
2396 | { | |
2397 | vfork_exit_internal(p, rv, 0); | |
2398 | } | |
2399 | ||
2400 | void | |
2401 | vfork_exit_internal(proc_t p, int rv, int forceexit) | |
2402 | { | |
2403 | thread_t self = current_thread(); | |
2404 | #ifdef FIXME | |
2405 | struct task *task = p->task; | |
2406 | #endif | |
2407 | struct uthread *ut; | |
2408 | ||
2409 | /* | |
2410 | * If a thread in this task has already | |
2411 | * called exit(), then halt any others | |
2412 | * right here. | |
2413 | */ | |
2414 | ||
2415 | ut = get_bsdthread_info(self); | |
2416 | ||
2417 | ||
2418 | proc_lock(p); | |
2419 | if ((p->p_lflag & P_LPEXIT) == P_LPEXIT) { | |
2420 | /* | |
2421 | * This happens when a parent exits/killed and vfork is in progress | |
2422 | * other threads. But shutdown code for ex has already called exit1() | |
2423 | */ | |
2424 | proc_unlock(p); | |
2425 | return; | |
2426 | } | |
2427 | p->p_lflag |= (P_LEXIT | P_LPEXIT); | |
2428 | proc_unlock(p); | |
2429 | ||
2430 | if (forceexit == 0) { | |
2431 | /* | |
2432 | * parent of a vfork child has already called exit() and the | |
2433 | * thread that has vfork in proress terminates. So there is no | |
2434 | * separate address space here and it has already been marked for | |
2435 | * termination. This was never covered before and could cause problems | |
2436 | * if we block here for outside code. | |
2437 | */ | |
2438 | /* Notify the perf server */ | |
2439 | (void)sys_perf_notify(self, p->p_pid); | |
2440 | } | |
2441 | ||
2442 | /* | |
2443 | * Remove proc from allproc queue and from pidhash chain. | |
2444 | * Need to do this before we do anything that can block. | |
2445 | * Not doing causes things like mount() find this on allproc | |
2446 | * in partially cleaned state. | |
2447 | */ | |
2448 | ||
2449 | proc_list_lock(); | |
2450 | ||
2451 | #if CONFIG_MEMORYSTATUS | |
2452 | proc_memorystatus_remove(p); | |
2453 | #endif | |
2454 | ||
2455 | LIST_REMOVE(p, p_list); | |
2456 | LIST_INSERT_HEAD(&zombproc, p, p_list); /* Place onto zombproc. */ | |
2457 | /* will not be visible via proc_find */ | |
2458 | p->p_listflag |= P_LIST_EXITED; | |
2459 | ||
2460 | proc_list_unlock(); | |
2461 | ||
2462 | proc_lock(p); | |
2463 | p->p_xstat = rv; | |
2464 | p->p_lflag &= ~(P_LTRACED | P_LPPWAIT); | |
2465 | p->p_sigignore = ~0; | |
2466 | proc_unlock(p); | |
2467 | ||
2468 | ut->uu_siglist = 0; | |
2469 | ||
2470 | /* begin vproc_exit */ | |
2471 | ||
2472 | proc_t q; | |
2473 | proc_t pp; | |
2474 | ||
2475 | vnode_t tvp; | |
2476 | ||
2477 | struct pgrp * pg; | |
2478 | struct session *sessp; | |
2479 | struct rusage_superset *rup; | |
2480 | ||
2481 | /* XXX Zombie allocation may fail, in which case stats get lost */ | |
2482 | MALLOC_ZONE(rup, struct rusage_superset *, | |
2483 | sizeof(*rup), M_ZOMBIE, M_WAITOK); | |
2484 | ||
2485 | proc_refdrain(p); | |
2486 | ||
2487 | /* | |
2488 | * Close open files and release open-file table. | |
2489 | * This may block! | |
2490 | */ | |
2491 | fdfree(p); | |
2492 | ||
2493 | sessp = proc_session(p); | |
2494 | if (SESS_LEADER(p, sessp)) { | |
2495 | if (sessp->s_ttyvp != NULLVP) { | |
2496 | struct vnode *ttyvp; | |
2497 | int ttyvid; | |
2498 | int cttyflag = 0; | |
2499 | struct vfs_context context; | |
2500 | struct tty *tp; | |
2501 | ||
2502 | /* | |
2503 | * Controlling process. | |
2504 | * Signal foreground pgrp, | |
2505 | * drain controlling terminal | |
2506 | * and revoke access to controlling terminal. | |
2507 | */ | |
2508 | session_lock(sessp); | |
2509 | tp = SESSION_TP(sessp); | |
2510 | if ((tp != TTY_NULL) && (tp->t_session == sessp)) { | |
2511 | session_unlock(sessp); | |
2512 | ||
2513 | /* | |
2514 | * We're going to SIGHUP the foreground process | |
2515 | * group. It can't change from this point on | |
2516 | * until the revoke is complete. | |
2517 | * The process group changes under both the tty | |
2518 | * lock and proc_list_lock but we need only one | |
2519 | */ | |
2520 | tty_lock(tp); | |
2521 | ttysetpgrphup(tp); | |
2522 | tty_unlock(tp); | |
2523 | ||
2524 | tty_pgsignal(tp, SIGHUP, 1); | |
2525 | ||
2526 | session_lock(sessp); | |
2527 | tp = SESSION_TP(sessp); | |
2528 | } | |
2529 | cttyflag = sessp->s_flags & S_CTTYREF; | |
2530 | sessp->s_flags &= ~S_CTTYREF; | |
2531 | ttyvp = sessp->s_ttyvp; | |
2532 | ttyvid = sessp->s_ttyvid; | |
2533 | sessp->s_ttyvp = NULL; | |
2534 | sessp->s_ttyvid = 0; | |
2535 | sessp->s_ttyp = TTY_NULL; | |
2536 | sessp->s_ttypgrpid = NO_PID; | |
2537 | session_unlock(sessp); | |
2538 | ||
2539 | if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) { | |
2540 | if (tp != TTY_NULL) { | |
2541 | tty_lock(tp); | |
2542 | (void) ttywait(tp); | |
2543 | tty_unlock(tp); | |
2544 | } | |
2545 | context.vc_thread = proc_thread(p); /* XXX */ | |
2546 | context.vc_ucred = kauth_cred_proc_ref(p); | |
2547 | VNOP_REVOKE(ttyvp, REVOKEALL, &context); | |
2548 | if (cttyflag) { | |
2549 | /* | |
2550 | * Release the extra usecount taken in cttyopen. | |
2551 | * usecount should be released after VNOP_REVOKE is called. | |
2552 | * This usecount was taken to ensure that | |
2553 | * the VNOP_REVOKE results in a close to | |
2554 | * the tty since cttyclose is a no-op. | |
2555 | */ | |
2556 | vnode_rele(ttyvp); | |
2557 | } | |
2558 | vnode_put(ttyvp); | |
2559 | kauth_cred_unref(&context.vc_ucred); | |
2560 | ttyvp = NULLVP; | |
2561 | } | |
2562 | if (tp) { | |
2563 | /* | |
2564 | * This is cleared even if not set. This is also done in | |
2565 | * spec_close to ensure that the flag is cleared. | |
2566 | */ | |
2567 | tty_lock(tp); | |
2568 | ttyclrpgrphup(tp); | |
2569 | tty_unlock(tp); | |
2570 | ||
2571 | ttyfree(tp); | |
2572 | } | |
2573 | } | |
2574 | session_lock(sessp); | |
2575 | sessp->s_leader = NULL; | |
2576 | session_unlock(sessp); | |
2577 | } | |
2578 | session_rele(sessp); | |
2579 | ||
2580 | pg = proc_pgrp(p); | |
2581 | fixjobc(p, pg, 0); | |
2582 | pg_rele(pg); | |
2583 | ||
2584 | p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; | |
2585 | ||
2586 | proc_list_lock(); | |
2587 | proc_childdrainstart(p); | |
2588 | while ((q = p->p_children.lh_first) != NULL) { | |
2589 | if (q->p_stat == SZOMB) { | |
2590 | if (p != q->p_pptr) { | |
2591 | panic("parent child linkage broken"); | |
2592 | } | |
2593 | /* check for lookups by zomb sysctl */ | |
2594 | while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) { | |
2595 | msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0); | |
2596 | } | |
2597 | q->p_listflag |= P_LIST_WAITING; | |
2598 | /* | |
2599 | * This is a named reference and it is not granted | |
2600 | * if the reap is already in progress. So we get | |
2601 | * the reference here exclusively and their can be | |
2602 | * no waiters. So there is no need for a wakeup | |
2603 | * after we are done. AlsO the reap frees the structure | |
2604 | * and the proc struct cannot be used for wakeups as well. | |
2605 | * It is safe to use q here as this is system reap | |
2606 | */ | |
2607 | (void)reap_child_locked(p, q, 1, 0, 1, 0); | |
2608 | } else { | |
2609 | /* | |
2610 | * Traced processes are killed | |
2611 | * since their existence means someone is messing up. | |
2612 | */ | |
2613 | if (q->p_lflag & P_LTRACED) { | |
2614 | struct proc *opp; | |
2615 | ||
2616 | proc_list_unlock(); | |
2617 | ||
2618 | opp = proc_find(q->p_oppid); | |
2619 | if (opp != PROC_NULL) { | |
2620 | proc_list_lock(); | |
2621 | q->p_oppid = 0; | |
2622 | proc_list_unlock(); | |
2623 | proc_reparentlocked(q, opp, 0, 0); | |
2624 | proc_rele(opp); | |
2625 | } else { | |
2626 | /* original parent exited while traced */ | |
2627 | proc_list_lock(); | |
2628 | q->p_listflag |= P_LIST_DEADPARENT; | |
2629 | q->p_oppid = 0; | |
2630 | proc_list_unlock(); | |
2631 | proc_reparentlocked(q, initproc, 0, 0); | |
2632 | } | |
2633 | ||
2634 | proc_lock(q); | |
2635 | q->p_lflag &= ~P_LTRACED; | |
2636 | ||
2637 | if (q->sigwait_thread) { | |
2638 | thread_t thread = q->sigwait_thread; | |
2639 | ||
2640 | proc_unlock(q); | |
2641 | /* | |
2642 | * The sigwait_thread could be stopped at a | |
2643 | * breakpoint. Wake it up to kill. | |
2644 | * Need to do this as it could be a thread which is not | |
2645 | * the first thread in the task. So any attempts to kill | |
2646 | * the process would result into a deadlock on q->sigwait. | |
2647 | */ | |
2648 | thread_resume(thread); | |
2649 | clear_wait(thread, THREAD_INTERRUPTED); | |
2650 | threadsignal(thread, SIGKILL, 0, TRUE); | |
2651 | } else { | |
2652 | proc_unlock(q); | |
2653 | } | |
2654 | ||
2655 | psignal(q, SIGKILL); | |
2656 | proc_list_lock(); | |
2657 | } else { | |
2658 | q->p_listflag |= P_LIST_DEADPARENT; | |
2659 | proc_reparentlocked(q, initproc, 0, 1); | |
2660 | } | |
2661 | } | |
2662 | } | |
2663 | ||
2664 | proc_childdrainend(p); | |
2665 | proc_list_unlock(); | |
2666 | ||
2667 | /* | |
2668 | * Release reference to text vnode | |
2669 | */ | |
2670 | tvp = p->p_textvp; | |
2671 | p->p_textvp = NULL; | |
2672 | if (tvp != NULLVP) { | |
2673 | vnode_rele(tvp); | |
2674 | } | |
2675 | ||
2676 | /* | |
2677 | * Save exit status and final rusage info, adding in child rusage | |
2678 | * info and self times. If we were unable to allocate a zombie | |
2679 | * structure, this information is lost. | |
2680 | */ | |
2681 | if (rup != NULL) { | |
2682 | rup->ru = p->p_stats->p_ru; | |
2683 | timerclear(&rup->ru.ru_utime); | |
2684 | timerclear(&rup->ru.ru_stime); | |
2685 | ||
2686 | #ifdef FIXME | |
2687 | if (task) { | |
2688 | mach_task_basic_info_data_t tinfo; | |
2689 | task_thread_times_info_data_t ttimesinfo; | |
2690 | int task_info_stuff, task_ttimes_stuff; | |
2691 | struct timeval ut, st; | |
2692 | ||
2693 | task_info_stuff = MACH_TASK_BASIC_INFO_COUNT; | |
2694 | task_info(task, MACH_TASK_BASIC_INFO, | |
2695 | &tinfo, &task_info_stuff); | |
2696 | p->p_ru->ru.ru_utime.tv_sec = tinfo.user_time.seconds; | |
2697 | p->p_ru->ru.ru_utime.tv_usec = tinfo.user_time.microseconds; | |
2698 | p->p_ru->ru.ru_stime.tv_sec = tinfo.system_time.seconds; | |
2699 | p->p_ru->ru.ru_stime.tv_usec = tinfo.system_time.microseconds; | |
2700 | ||
2701 | task_ttimes_stuff = TASK_THREAD_TIMES_INFO_COUNT; | |
2702 | task_info(task, TASK_THREAD_TIMES_INFO, | |
2703 | &ttimesinfo, &task_ttimes_stuff); | |
2704 | ||
2705 | ut.tv_sec = ttimesinfo.user_time.seconds; | |
2706 | ut.tv_usec = ttimesinfo.user_time.microseconds; | |
2707 | st.tv_sec = ttimesinfo.system_time.seconds; | |
2708 | st.tv_usec = ttimesinfo.system_time.microseconds; | |
2709 | timeradd(&ut, &p->p_ru->ru.ru_utime, &p->p_ru->ru.ru_utime); | |
2710 | timeradd(&st, &p->p_ru->ru.ru_stime, &p->p_ru->ru.ru_stime); | |
2711 | } | |
2712 | #endif /* FIXME */ | |
2713 | ||
2714 | ruadd(&rup->ru, &p->p_stats->p_cru); | |
2715 | ||
2716 | gather_rusage_info(p, &rup->ri, RUSAGE_INFO_CURRENT); | |
2717 | rup->ri.ri_phys_footprint = 0; | |
2718 | rup->ri.ri_proc_exit_abstime = mach_absolute_time(); | |
2719 | ||
2720 | /* | |
2721 | * Now that we have filled in the rusage info, make it | |
2722 | * visible to an external observer via proc_pid_rusage(). | |
2723 | */ | |
2724 | p->p_ru = rup; | |
2725 | } | |
2726 | ||
2727 | /* | |
2728 | * Free up profiling buffers. | |
2729 | */ | |
2730 | { | |
2731 | struct uprof *p0 = &p->p_stats->p_prof, *p1, *pn; | |
2732 | ||
2733 | p1 = p0->pr_next; | |
2734 | p0->pr_next = NULL; | |
2735 | p0->pr_scale = 0; | |
2736 | ||
2737 | for (; p1 != NULL; p1 = pn) { | |
2738 | pn = p1->pr_next; | |
2739 | kfree(p1, sizeof *p1); | |
2740 | } | |
2741 | } | |
2742 | ||
2743 | #if PSYNCH | |
2744 | pth_proc_hashdelete(p); | |
2745 | #endif /* PSYNCH */ | |
2746 | ||
2747 | proc_free_realitimer(p); | |
2748 | ||
2749 | /* | |
2750 | * Other substructures are freed from wait(). | |
2751 | */ | |
2752 | FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS); | |
2753 | p->p_stats = NULL; | |
2754 | ||
2755 | FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS); | |
2756 | p->p_sigacts = NULL; | |
2757 | ||
2758 | proc_limitdrop(p, 1); | |
2759 | p->p_limit = NULL; | |
2760 | ||
2761 | /* | |
2762 | * Finish up by terminating the task | |
2763 | * and halt this thread (only if a | |
2764 | * member of the task exiting). | |
2765 | */ | |
2766 | p->task = TASK_NULL; | |
2767 | ||
2768 | /* | |
2769 | * Notify parent that we're gone. | |
2770 | */ | |
2771 | pp = proc_parent(p); | |
2772 | if ((p->p_listflag & P_LIST_DEADPARENT) == 0) { | |
2773 | if (pp != initproc) { | |
2774 | proc_lock(pp); | |
2775 | pp->si_pid = p->p_pid; | |
2776 | pp->p_xhighbits = p->p_xhighbits; | |
2777 | p->p_xhighbits = 0; | |
2778 | pp->si_status = p->p_xstat; | |
2779 | pp->si_code = CLD_EXITED; | |
2780 | /* | |
2781 | * p_ucred usage is safe as it is an exiting process | |
2782 | * and reference is dropped in reap | |
2783 | */ | |
2784 | pp->si_uid = kauth_cred_getruid(p->p_ucred); | |
2785 | proc_unlock(pp); | |
2786 | } | |
2787 | /* mark as a zombie */ | |
2788 | /* mark as a zombie */ | |
2789 | /* No need to take proc lock as all refs are drained and | |
2790 | * no one except parent (reaping ) can look at this. | |
2791 | * The write is to an int and is coherent. Also parent is | |
2792 | * keyed off of list lock for reaping | |
2793 | */ | |
2794 | p->p_stat = SZOMB; | |
2795 | ||
2796 | psignal(pp, SIGCHLD); | |
2797 | ||
2798 | /* and now wakeup the parent */ | |
2799 | proc_list_lock(); | |
2800 | wakeup((caddr_t)pp); | |
2801 | proc_list_unlock(); | |
2802 | } else { | |
2803 | proc_list_lock(); | |
2804 | /* check for lookups by zomb sysctl */ | |
2805 | while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) { | |
2806 | msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0); | |
2807 | } | |
2808 | p->p_stat = SZOMB; | |
2809 | p->p_listflag |= P_LIST_WAITING; | |
2810 | ||
2811 | /* | |
2812 | * This is a named reference and it is not granted | |
2813 | * if the reap is already in progress. So we get | |
2814 | * the reference here exclusively and their can be | |
2815 | * no waiters. So there is no need for a wakeup | |
2816 | * after we are done. AlsO the reap frees the structure | |
2817 | * and the proc struct cannot be used for wakeups as well. | |
2818 | * It is safe to use p here as this is system reap | |
2819 | */ | |
2820 | (void)reap_child_locked(pp, p, 0, 0, 1, 1); | |
2821 | /* list lock dropped by reap_child_locked */ | |
2822 | } | |
2823 | proc_rele(pp); | |
2824 | } | |
2825 | ||
2826 | ||
2827 | /* | |
2828 | * munge_rusage | |
2829 | * LP64 support - long is 64 bits if we are dealing with a 64 bit user | |
2830 | * process. We munge the kernel version of rusage into the | |
2831 | * 64 bit version. | |
2832 | */ | |
2833 | __private_extern__ void | |
2834 | munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p) | |
2835 | { | |
2836 | /* Zero-out struct so that padding is cleared */ | |
2837 | bzero(a_user_rusage_p, sizeof(struct user64_rusage)); | |
2838 | ||
2839 | /* timeval changes size, so utime and stime need special handling */ | |
2840 | a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec; | |
2841 | a_user_rusage_p->ru_utime.tv_usec = a_rusage_p->ru_utime.tv_usec; | |
2842 | a_user_rusage_p->ru_stime.tv_sec = a_rusage_p->ru_stime.tv_sec; | |
2843 | a_user_rusage_p->ru_stime.tv_usec = a_rusage_p->ru_stime.tv_usec; | |
2844 | /* | |
2845 | * everything else can be a direct assign, since there is no loss | |
2846 | * of precision implied boing 32->64. | |
2847 | */ | |
2848 | a_user_rusage_p->ru_maxrss = a_rusage_p->ru_maxrss; | |
2849 | a_user_rusage_p->ru_ixrss = a_rusage_p->ru_ixrss; | |
2850 | a_user_rusage_p->ru_idrss = a_rusage_p->ru_idrss; | |
2851 | a_user_rusage_p->ru_isrss = a_rusage_p->ru_isrss; | |
2852 | a_user_rusage_p->ru_minflt = a_rusage_p->ru_minflt; | |
2853 | a_user_rusage_p->ru_majflt = a_rusage_p->ru_majflt; | |
2854 | a_user_rusage_p->ru_nswap = a_rusage_p->ru_nswap; | |
2855 | a_user_rusage_p->ru_inblock = a_rusage_p->ru_inblock; | |
2856 | a_user_rusage_p->ru_oublock = a_rusage_p->ru_oublock; | |
2857 | a_user_rusage_p->ru_msgsnd = a_rusage_p->ru_msgsnd; | |
2858 | a_user_rusage_p->ru_msgrcv = a_rusage_p->ru_msgrcv; | |
2859 | a_user_rusage_p->ru_nsignals = a_rusage_p->ru_nsignals; | |
2860 | a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw; | |
2861 | a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw; | |
2862 | } | |
2863 | ||
2864 | /* For a 64-bit kernel and 32-bit userspace, munging may be needed */ | |
2865 | __private_extern__ void | |
2866 | munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p) | |
2867 | { | |
2868 | bzero(a_user_rusage_p, sizeof(struct user32_rusage)); | |
2869 | ||
2870 | /* timeval changes size, so utime and stime need special handling */ | |
2871 | a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec; | |
2872 | a_user_rusage_p->ru_utime.tv_usec = a_rusage_p->ru_utime.tv_usec; | |
2873 | a_user_rusage_p->ru_stime.tv_sec = a_rusage_p->ru_stime.tv_sec; | |
2874 | a_user_rusage_p->ru_stime.tv_usec = a_rusage_p->ru_stime.tv_usec; | |
2875 | /* | |
2876 | * everything else can be a direct assign. We currently ignore | |
2877 | * the loss of precision | |
2878 | */ | |
2879 | a_user_rusage_p->ru_maxrss = a_rusage_p->ru_maxrss; | |
2880 | a_user_rusage_p->ru_ixrss = a_rusage_p->ru_ixrss; | |
2881 | a_user_rusage_p->ru_idrss = a_rusage_p->ru_idrss; | |
2882 | a_user_rusage_p->ru_isrss = a_rusage_p->ru_isrss; | |
2883 | a_user_rusage_p->ru_minflt = a_rusage_p->ru_minflt; | |
2884 | a_user_rusage_p->ru_majflt = a_rusage_p->ru_majflt; | |
2885 | a_user_rusage_p->ru_nswap = a_rusage_p->ru_nswap; | |
2886 | a_user_rusage_p->ru_inblock = a_rusage_p->ru_inblock; | |
2887 | a_user_rusage_p->ru_oublock = a_rusage_p->ru_oublock; | |
2888 | a_user_rusage_p->ru_msgsnd = a_rusage_p->ru_msgsnd; | |
2889 | a_user_rusage_p->ru_msgrcv = a_rusage_p->ru_msgrcv; | |
2890 | a_user_rusage_p->ru_nsignals = a_rusage_p->ru_nsignals; | |
2891 | a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw; | |
2892 | a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw; | |
2893 | } | |
2894 | ||
2895 | void | |
2896 | kdp_wait4_find_process(thread_t thread, __unused event64_t wait_event, thread_waitinfo_t *waitinfo) | |
2897 | { | |
2898 | assert(thread != NULL); | |
2899 | assert(waitinfo != NULL); | |
2900 | ||
2901 | struct uthread *ut = get_bsdthread_info(thread); | |
2902 | waitinfo->context = 0; | |
2903 | // ensure wmesg is consistent with a thread waiting in wait4 | |
2904 | assert(!strcmp(ut->uu_wmesg, "waitcoll") || !strcmp(ut->uu_wmesg, "wait")); | |
2905 | struct wait4_nocancel_args *args = ut->uu_save.uus_wait4_data.args; | |
2906 | // May not actually contain a pid; this is just the argument to wait4. | |
2907 | // See man wait4 for other valid wait4 arguments. | |
2908 | waitinfo->owner = args->pid; | |
2909 | } |