]>
Commit | Line | Data |
---|---|---|
1c79356b | 1 | /* |
39037602 | 2 | * Copyright (c) 2000-2016 Apple Inc. All rights reserved. |
1c79356b | 3 | * |
2d21ac55 | 4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
1c79356b | 5 | * |
2d21ac55 A |
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. | |
8f6c56a5 | 14 | * |
2d21ac55 A |
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 | |
8f6c56a5 A |
20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, | |
2d21ac55 A |
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. | |
8f6c56a5 | 25 | * |
2d21ac55 | 26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
1c79356b A |
27 | */ |
28 | /* | |
29 | * @OSF_FREE_COPYRIGHT@ | |
30 | */ | |
31 | /* | |
32 | * Mach Operating System | |
33 | * Copyright (c) 1991,1990,1989,1988 Carnegie Mellon University | |
34 | * All Rights Reserved. | |
35 | * | |
36 | * Permission to use, copy, modify and distribute this software and its | |
37 | * documentation is hereby granted, provided that both the copyright | |
38 | * notice and this permission notice appear in all copies of the | |
39 | * software, derivative works or modified versions, and any portions | |
40 | * thereof, and that both notices appear in supporting documentation. | |
41 | * | |
42 | * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" | |
43 | * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR | |
44 | * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. | |
45 | * | |
46 | * Carnegie Mellon requests users of this software to return to | |
47 | * | |
48 | * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU | |
49 | * School of Computer Science | |
50 | * Carnegie Mellon University | |
51 | * Pittsburgh PA 15213-3890 | |
52 | * | |
53 | * any improvements or extensions that they make and grant Carnegie Mellon | |
54 | * the rights to redistribute these changes. | |
55 | */ | |
56 | /* | |
57 | * File: kern/task.c | |
58 | * Author: Avadis Tevanian, Jr., Michael Wayne Young, David Golub, | |
59 | * David Black | |
60 | * | |
61 | * Task management primitives implementation. | |
62 | */ | |
63 | /* | |
64 | * Copyright (c) 1993 The University of Utah and | |
65 | * the Computer Systems Laboratory (CSL). All rights reserved. | |
66 | * | |
67 | * Permission to use, copy, modify and distribute this software and its | |
68 | * documentation is hereby granted, provided that both the copyright | |
69 | * notice and this permission notice appear in all copies of the | |
70 | * software, derivative works or modified versions, and any portions | |
71 | * thereof, and that both notices appear in supporting documentation. | |
72 | * | |
73 | * THE UNIVERSITY OF UTAH AND CSL ALLOW FREE USE OF THIS SOFTWARE IN ITS "AS | |
74 | * IS" CONDITION. THE UNIVERSITY OF UTAH AND CSL DISCLAIM ANY LIABILITY OF | |
75 | * ANY KIND FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. | |
76 | * | |
77 | * CSL requests users of this software to return to csl-dist@cs.utah.edu any | |
78 | * improvements that they make and grant CSL redistribution rights. | |
79 | * | |
80 | */ | |
2d21ac55 A |
81 | /* |
82 | * NOTICE: This file was modified by McAfee Research in 2004 to introduce | |
83 | * support for mandatory and extensible security protections. This notice | |
84 | * is included in support of clause 2.2 (b) of the Apple Public License, | |
85 | * Version 2.0. | |
86 | * Copyright (c) 2005 SPARTA, Inc. | |
87 | */ | |
1c79356b | 88 | |
91447636 | 89 | #include <mach/mach_types.h> |
1c79356b | 90 | #include <mach/boolean.h> |
91447636 | 91 | #include <mach/host_priv.h> |
1c79356b A |
92 | #include <mach/machine/vm_types.h> |
93 | #include <mach/vm_param.h> | |
3e170ce0 | 94 | #include <mach/mach_vm.h> |
1c79356b A |
95 | #include <mach/semaphore.h> |
96 | #include <mach/task_info.h> | |
97 | #include <mach/task_special_ports.h> | |
39037602 | 98 | #include <mach/sdt.h> |
91447636 | 99 | |
fe8ab488 | 100 | #include <ipc/ipc_importance.h> |
91447636 | 101 | #include <ipc/ipc_types.h> |
1c79356b A |
102 | #include <ipc/ipc_space.h> |
103 | #include <ipc/ipc_entry.h> | |
39236c6e | 104 | #include <ipc/ipc_hash.h> |
91447636 A |
105 | |
106 | #include <kern/kern_types.h> | |
1c79356b A |
107 | #include <kern/mach_param.h> |
108 | #include <kern/misc_protos.h> | |
109 | #include <kern/task.h> | |
110 | #include <kern/thread.h> | |
fe8ab488 | 111 | #include <kern/coalition.h> |
1c79356b A |
112 | #include <kern/zalloc.h> |
113 | #include <kern/kalloc.h> | |
3e170ce0 | 114 | #include <kern/kern_cdata.h> |
1c79356b A |
115 | #include <kern/processor.h> |
116 | #include <kern/sched_prim.h> /* for thread_wakeup */ | |
1c79356b | 117 | #include <kern/ipc_tt.h> |
1c79356b | 118 | #include <kern/host.h> |
91447636 A |
119 | #include <kern/clock.h> |
120 | #include <kern/timer.h> | |
1c79356b A |
121 | #include <kern/assert.h> |
122 | #include <kern/sync_lock.h> | |
2d21ac55 | 123 | #include <kern/affinity.h> |
39236c6e | 124 | #include <kern/exc_resource.h> |
3e170ce0 | 125 | #include <kern/machine.h> |
39037602 A |
126 | #include <kern/policy_internal.h> |
127 | ||
3e170ce0 | 128 | #include <corpses/task_corpse.h> |
39236c6e A |
129 | #if CONFIG_TELEMETRY |
130 | #include <kern/telemetry.h> | |
131 | #endif | |
91447636 A |
132 | |
133 | #include <vm/pmap.h> | |
134 | #include <vm/vm_map.h> | |
135 | #include <vm/vm_kern.h> /* for kernel_map, ipc_kernel_map */ | |
136 | #include <vm/vm_pageout.h> | |
2d21ac55 | 137 | #include <vm/vm_protos.h> |
39236c6e | 138 | #include <vm/vm_purgeable_internal.h> |
91447636 | 139 | |
39236c6e | 140 | #include <sys/resource.h> |
3e170ce0 A |
141 | #include <sys/signalvar.h> /* for coredump */ |
142 | ||
1c79356b A |
143 | /* |
144 | * Exported interfaces | |
145 | */ | |
146 | ||
147 | #include <mach/task_server.h> | |
148 | #include <mach/mach_host_server.h> | |
149 | #include <mach/host_security_server.h> | |
91447636 A |
150 | #include <mach/mach_port_server.h> |
151 | ||
2d21ac55 A |
152 | #include <vm/vm_shared_region.h> |
153 | ||
39236c6e A |
154 | #include <libkern/OSDebug.h> |
155 | #include <libkern/OSAtomic.h> | |
156 | ||
fe8ab488 A |
157 | #if CONFIG_ATM |
158 | #include <atm/atm_internal.h> | |
159 | #endif | |
160 | ||
39037602 A |
161 | #include <kern/sfi.h> /* picks up ledger.h */ |
162 | ||
163 | #if CONFIG_MACF | |
164 | #include <security/mac_mach_internal.h> | |
165 | #endif | |
fe8ab488 A |
166 | |
167 | #if KPERF | |
168 | extern int kpc_force_all_ctrs(task_t, int); | |
169 | #endif | |
170 | ||
b0d623f7 A |
171 | task_t kernel_task; |
172 | zone_t task_zone; | |
173 | lck_attr_t task_lck_attr; | |
174 | lck_grp_t task_lck_grp; | |
175 | lck_grp_attr_t task_lck_grp_attr; | |
176 | ||
39037602 A |
177 | extern int exc_via_corpse_forking; |
178 | extern int unify_corpse_blob_alloc; | |
179 | extern int corpse_for_fatal_memkill; | |
180 | ||
15129b1c A |
181 | /* Flag set by core audio when audio is playing. Used to stifle EXC_RESOURCE generation when active. */ |
182 | int audio_active = 0; | |
183 | ||
6d2010ae A |
184 | zinfo_usage_store_t tasks_tkm_private; |
185 | zinfo_usage_store_t tasks_tkm_shared; | |
186 | ||
4b17d6b6 | 187 | /* A container to accumulate statistics for expired tasks */ |
39236c6e A |
188 | expired_task_statistics_t dead_task_statistics; |
189 | lck_spin_t dead_task_statistics_lock; | |
4b17d6b6 | 190 | |
fe8ab488 A |
191 | ledger_template_t task_ledger_template = NULL; |
192 | ||
193 | struct _task_ledger_indices task_ledgers __attribute__((used)) = | |
39037602 | 194 | {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, |
fe8ab488 A |
195 | { 0 /* initialized at runtime */}, |
196 | #ifdef CONFIG_BANK | |
197 | -1, -1, | |
198 | #endif | |
39037602 | 199 | -1, -1, |
fe8ab488 A |
200 | }; |
201 | ||
4bd07ac2 A |
202 | /* System sleep state */ |
203 | boolean_t tasks_suspend_state; | |
204 | ||
205 | ||
316670eb | 206 | void init_task_ledgers(void); |
39236c6e A |
207 | void task_footprint_exceeded(int warning, __unused const void *param0, __unused const void *param1); |
208 | void task_wakeups_rate_exceeded(int warning, __unused const void *param0, __unused const void *param1); | |
39037602 A |
209 | void task_io_rate_exceeded(int warning, const void *param0, __unused const void *param1); |
210 | void __attribute__((noinline)) SENDING_NOTIFICATION__THIS_PROCESS_IS_CAUSING_TOO_MANY_WAKEUPS(void); | |
211 | void __attribute__((noinline)) PROC_CROSSED_HIGH_WATERMARK__SEND_EXC_RESOURCE_AND_SUSPEND(int max_footprint_mb, boolean_t is_fatal); | |
212 | void __attribute__((noinline)) SENDING_NOTIFICATION__THIS_PROCESS_IS_CAUSING_TOO_MUCH_IO(int flavor); | |
39236c6e A |
213 | |
214 | kern_return_t task_suspend_internal(task_t); | |
215 | kern_return_t task_resume_internal(task_t); | |
3e170ce0 | 216 | static kern_return_t task_start_halt_locked(task_t task, boolean_t should_mark_corpse); |
39037602 | 217 | int proc_list_uptrs(void *p, uint64_t *udata_buffer, int size); |
3e170ce0 | 218 | |
7e41aa88 | 219 | extern kern_return_t iokit_task_terminate(task_t task); |
39236c6e | 220 | |
3e170ce0 | 221 | extern kern_return_t exception_deliver(thread_t, exception_type_t, mach_exception_data_t, mach_msg_type_number_t, struct exception_action *, lck_mtx_t *); |
39037602 | 222 | extern void bsd_copythreadname(void *dst_uth, void *src_uth); |
39236c6e A |
223 | |
224 | // Warn tasks when they hit 80% of their memory limit. | |
225 | #define PHYS_FOOTPRINT_WARNING_LEVEL 80 | |
226 | ||
227 | #define TASK_WAKEUPS_MONITOR_DEFAULT_LIMIT 150 /* wakeups per second */ | |
228 | #define TASK_WAKEUPS_MONITOR_DEFAULT_INTERVAL 300 /* in seconds. */ | |
229 | ||
230 | /* | |
231 | * Level (in terms of percentage of the limit) at which the wakeups monitor triggers telemetry. | |
232 | * | |
233 | * (ie when the task's wakeups rate exceeds 70% of the limit, start taking user | |
234 | * stacktraces, aka micro-stackshots) | |
235 | */ | |
236 | #define TASK_WAKEUPS_MONITOR_DEFAULT_USTACKSHOTS_TRIGGER 70 | |
237 | ||
238 | int task_wakeups_monitor_interval; /* In seconds. Time period over which wakeups rate is observed */ | |
239 | int task_wakeups_monitor_rate; /* In hz. Maximum allowable wakeups per task before EXC_RESOURCE is sent */ | |
240 | ||
241 | int task_wakeups_monitor_ustackshots_trigger_pct; /* Percentage. Level at which we start gathering telemetry. */ | |
316670eb | 242 | |
39236c6e | 243 | int disable_exc_resource; /* Global override to supress EXC_RESOURCE for resource monitor violations. */ |
316670eb | 244 | |
3e170ce0 | 245 | ledger_amount_t max_task_footprint = 0; /* Per-task limit on physical memory consumption in bytes */ |
39037602 | 246 | int max_task_footprint_warning_level = 0; /* Per-task limit warning percentage */ |
3e170ce0 A |
247 | int max_task_footprint_mb = 0; /* Per-task limit on physical memory consumption in megabytes */ |
248 | ||
39037602 A |
249 | /* I/O Monitor Limits */ |
250 | #define IOMON_DEFAULT_LIMIT (20480ull) /* MB of logical/physical I/O */ | |
251 | #define IOMON_DEFAULT_INTERVAL (86400ull) /* in seconds */ | |
252 | ||
253 | uint64_t task_iomon_limit_mb; /* Per-task I/O monitor limit in MBs */ | |
254 | uint64_t task_iomon_interval_secs; /* Per-task I/O monitor interval in secs */ | |
255 | ||
256 | #define IO_TELEMETRY_DEFAULT_LIMIT (10ll * 1024ll * 1024ll) | |
257 | int64_t io_telemetry_limit; /* Threshold to take a microstackshot (0 indicated I/O telemetry is turned off) */ | |
258 | int64_t global_logical_writes_count = 0; /* Global count for logical writes */ | |
259 | static boolean_t global_update_logical_writes(int64_t); | |
260 | ||
fe8ab488 A |
261 | #if MACH_ASSERT |
262 | int pmap_ledgers_panic = 1; | |
263 | #endif /* MACH_ASSERT */ | |
264 | ||
b0d623f7 | 265 | int task_max = CONFIG_TASK_MAX; /* Max number of tasks */ |
1c79356b | 266 | |
39037602 | 267 | #if CONFIG_COREDUMP |
39236c6e | 268 | int hwm_user_cores = 0; /* high watermark violations generate user core files */ |
39037602 | 269 | #endif |
39236c6e A |
270 | |
271 | #ifdef MACH_BSD | |
272 | extern void proc_getexecutableuuid(void *, unsigned char *, unsigned long); | |
273 | extern int proc_pid(struct proc *p); | |
274 | extern int proc_selfpid(void); | |
275 | extern char *proc_name_address(struct proc *p); | |
3e170ce0 | 276 | extern uint64_t get_dispatchqueue_offset_from_proc(void *); |
39037602 A |
277 | |
278 | #if CONFIG_MEMORYSTATUS | |
3e170ce0 | 279 | extern void proc_memstat_terminated(struct proc* p, boolean_t set); |
39037602 A |
280 | extern boolean_t memorystatus_turnoff_exception_and_get_fatalness(boolean_t warning, const int max_footprint_mb); |
281 | extern void memorystatus_on_ledger_footprint_exceeded(int warning, boolean_t is_fatal); | |
282 | #endif /* CONFIG_MEMORYSTATUS */ | |
283 | ||
284 | #endif /* MACH_BSD */ | |
6d2010ae | 285 | |
1c79356b A |
286 | /* Forwards */ |
287 | ||
39037602 A |
288 | static void task_hold_locked(task_t task); |
289 | static void task_wait_locked(task_t task, boolean_t until_not_runnable); | |
290 | static void task_release_locked(task_t task); | |
291 | ||
292 | static void task_synchronizer_destroy_all(task_t task); | |
b0d623f7 | 293 | |
55e303ae A |
294 | void |
295 | task_backing_store_privileged( | |
296 | task_t task) | |
297 | { | |
298 | task_lock(task); | |
299 | task->priv_flags |= VM_BACKING_STORE_PRIV; | |
300 | task_unlock(task); | |
301 | return; | |
302 | } | |
303 | ||
91447636 A |
304 | |
305 | void | |
306 | task_set_64bit( | |
307 | task_t task, | |
308 | boolean_t is64bit) | |
309 | { | |
fe8ab488 | 310 | #if defined(__i386__) || defined(__x86_64__) || defined(__arm64__) |
2d21ac55 | 311 | thread_t thread; |
fe8ab488 | 312 | #endif /* defined(__i386__) || defined(__x86_64__) || defined(__arm64__) */ |
39236c6e A |
313 | |
314 | task_lock(task); | |
0c530ab8 A |
315 | |
316 | if (is64bit) { | |
2d21ac55 | 317 | if (task_has_64BitAddr(task)) |
39236c6e | 318 | goto out; |
91447636 | 319 | task_set_64BitAddr(task); |
91447636 | 320 | } else { |
2d21ac55 | 321 | if ( !task_has_64BitAddr(task)) |
39236c6e | 322 | goto out; |
91447636 | 323 | task_clear_64BitAddr(task); |
91447636 | 324 | } |
0c530ab8 A |
325 | /* FIXME: On x86, the thread save state flavor can diverge from the |
326 | * task's 64-bit feature flag due to the 32-bit/64-bit register save | |
327 | * state dichotomy. Since we can be pre-empted in this interval, | |
328 | * certain routines may observe the thread as being in an inconsistent | |
329 | * state with respect to its task's 64-bitness. | |
330 | */ | |
39236c6e | 331 | |
fe8ab488 | 332 | #if defined(__i386__) || defined(__x86_64__) || defined(__arm64__) |
0c530ab8 | 333 | queue_iterate(&task->threads, thread, thread_t, task_threads) { |
b0d623f7 | 334 | thread_mtx_lock(thread); |
2d21ac55 | 335 | machine_thread_switch_addrmode(thread); |
b0d623f7 | 336 | thread_mtx_unlock(thread); |
39037602 A |
337 | |
338 | if (thread == current_thread()) { | |
339 | uint64_t arg1, arg2; | |
340 | int urgency; | |
341 | spl_t spl = splsched(); | |
342 | /* | |
343 | * This call tell that the current thread changed it's 32bitness. | |
344 | * Other thread were no more on core when 32bitness was changed, | |
345 | * but current_thread() is on core and the previous call to | |
346 | * machine_thread_going_on_core() gave 32bitness which is now wrong. | |
347 | * | |
348 | * This is needed for bring-up, a different callback should be used | |
349 | * in the future. | |
350 | */ | |
351 | thread_lock(thread); | |
352 | urgency = thread_get_urgency(thread, &arg1, &arg2); | |
353 | machine_thread_going_on_core(thread, urgency, 0); | |
354 | thread_unlock(thread); | |
355 | splx(spl); | |
356 | } | |
0c530ab8 | 357 | } |
fe8ab488 | 358 | #endif /* defined(__i386__) || defined(__x86_64__) || defined(__arm64__) */ |
39236c6e A |
359 | |
360 | out: | |
b0d623f7 | 361 | task_unlock(task); |
91447636 A |
362 | } |
363 | ||
b0d623f7 A |
364 | |
365 | void | |
366 | task_set_dyld_info(task_t task, mach_vm_address_t addr, mach_vm_size_t size) | |
367 | { | |
368 | task_lock(task); | |
369 | task->all_image_info_addr = addr; | |
370 | task->all_image_info_size = size; | |
371 | task_unlock(task); | |
372 | } | |
373 | ||
fe8ab488 A |
374 | void |
375 | task_atm_reset(__unused task_t task) { | |
376 | ||
377 | #if CONFIG_ATM | |
378 | if (task->atm_context != NULL) { | |
379 | atm_task_descriptor_destroy(task->atm_context); | |
380 | task->atm_context = NULL; | |
381 | } | |
382 | #endif | |
383 | ||
384 | } | |
385 | ||
490019cf A |
386 | void |
387 | task_bank_reset(__unused task_t task) { | |
388 | ||
389 | #if CONFIG_BANK | |
390 | if (task->bank_context != NULL) { | |
391 | bank_task_destroy(task); | |
392 | } | |
393 | #endif | |
394 | ||
395 | } | |
396 | ||
397 | /* | |
398 | * NOTE: This should only be called when the P_LINTRANSIT | |
399 | * flag is set (the proc_trans lock is held) on the | |
400 | * proc associated with the task. | |
401 | */ | |
402 | void | |
403 | task_bank_init(__unused task_t task) { | |
404 | ||
405 | #if CONFIG_BANK | |
406 | if (task->bank_context != NULL) { | |
407 | panic("Task bank init called with non null bank context for task: %p and bank_context: %p", task, task->bank_context); | |
408 | } | |
409 | bank_task_initialize(task); | |
410 | #endif | |
411 | ||
412 | } | |
413 | ||
39236c6e A |
414 | #if TASK_REFERENCE_LEAK_DEBUG |
415 | #include <kern/btlog.h> | |
416 | ||
39236c6e A |
417 | static btlog_t *task_ref_btlog; |
418 | #define TASK_REF_OP_INCR 0x1 | |
419 | #define TASK_REF_OP_DECR 0x2 | |
420 | ||
39037602 | 421 | #define TASK_REF_NUM_RECORDS 100000 |
39236c6e A |
422 | #define TASK_REF_BTDEPTH 7 |
423 | ||
39236c6e A |
424 | void |
425 | task_reference_internal(task_t task) | |
426 | { | |
427 | void * bt[TASK_REF_BTDEPTH]; | |
428 | int numsaved = 0; | |
429 | ||
430 | numsaved = OSBacktrace(bt, TASK_REF_BTDEPTH); | |
431 | ||
432 | (void)hw_atomic_add(&(task)->ref_count, 1); | |
433 | btlog_add_entry(task_ref_btlog, task, TASK_REF_OP_INCR, | |
434 | bt, numsaved); | |
435 | } | |
436 | ||
437 | uint32_t | |
438 | task_deallocate_internal(task_t task) | |
439 | { | |
440 | void * bt[TASK_REF_BTDEPTH]; | |
441 | int numsaved = 0; | |
442 | ||
443 | numsaved = OSBacktrace(bt, TASK_REF_BTDEPTH); | |
444 | ||
445 | btlog_add_entry(task_ref_btlog, task, TASK_REF_OP_DECR, | |
446 | bt, numsaved); | |
447 | return hw_atomic_sub(&(task)->ref_count, 1); | |
448 | } | |
449 | ||
450 | #endif /* TASK_REFERENCE_LEAK_DEBUG */ | |
451 | ||
1c79356b A |
452 | void |
453 | task_init(void) | |
454 | { | |
b0d623f7 A |
455 | |
456 | lck_grp_attr_setdefault(&task_lck_grp_attr); | |
457 | lck_grp_init(&task_lck_grp, "task", &task_lck_grp_attr); | |
458 | lck_attr_setdefault(&task_lck_attr); | |
459 | lck_mtx_init(&tasks_threads_lock, &task_lck_grp, &task_lck_attr); | |
39037602 | 460 | lck_mtx_init(&tasks_corpse_lock, &task_lck_grp, &task_lck_attr); |
b0d623f7 | 461 | |
1c79356b A |
462 | task_zone = zinit( |
463 | sizeof(struct task), | |
b0d623f7 | 464 | task_max * sizeof(struct task), |
1c79356b A |
465 | TASK_CHUNK * sizeof(struct task), |
466 | "tasks"); | |
6d2010ae | 467 | |
0b4c1975 | 468 | zone_change(task_zone, Z_NOENCRYPT, TRUE); |
1c79356b | 469 | |
39037602 | 470 | |
39236c6e | 471 | /* |
fe8ab488 A |
472 | * Configure per-task memory limit. |
473 | * The boot-arg is interpreted as Megabytes, | |
474 | * and takes precedence over the device tree. | |
475 | * Setting the boot-arg to 0 disables task limits. | |
39236c6e | 476 | */ |
3e170ce0 A |
477 | if (!PE_parse_boot_argn("max_task_pmem", &max_task_footprint_mb, |
478 | sizeof (max_task_footprint_mb))) { | |
39236c6e A |
479 | /* |
480 | * No limit was found in boot-args, so go look in the device tree. | |
481 | */ | |
3e170ce0 A |
482 | if (!PE_get_default("kern.max_task_pmem", &max_task_footprint_mb, |
483 | sizeof(max_task_footprint_mb))) { | |
fe8ab488 A |
484 | /* |
485 | * No limit was found in device tree. | |
486 | */ | |
3e170ce0 | 487 | max_task_footprint_mb = 0; |
39236c6e A |
488 | } |
489 | } | |
490 | ||
3e170ce0 | 491 | if (max_task_footprint_mb != 0) { |
39037602 | 492 | #if CONFIG_MEMORYSTATUS |
3e170ce0 | 493 | if (max_task_footprint_mb < 50) { |
39236c6e | 494 | printf("Warning: max_task_pmem %d below minimum.\n", |
3e170ce0 A |
495 | max_task_footprint_mb); |
496 | max_task_footprint_mb = 50; | |
39236c6e A |
497 | } |
498 | printf("Limiting task physical memory footprint to %d MB\n", | |
3e170ce0 A |
499 | max_task_footprint_mb); |
500 | ||
501 | max_task_footprint = (ledger_amount_t)max_task_footprint_mb * 1024 * 1024; // Convert MB to bytes | |
39037602 A |
502 | |
503 | /* | |
504 | * Configure the per-task memory limit warning level. | |
505 | * This is computed as a percentage. | |
506 | */ | |
507 | max_task_footprint_warning_level = 0; | |
508 | ||
509 | if (max_mem < 0x40000000) { | |
510 | /* | |
511 | * On devices with < 1GB of memory: | |
512 | * -- set warnings to 50MB below the per-task limit. | |
513 | */ | |
514 | if (max_task_footprint_mb > 50) { | |
515 | max_task_footprint_warning_level = ((max_task_footprint_mb - 50) * 100) / max_task_footprint_mb; | |
516 | } | |
517 | } else { | |
518 | /* | |
519 | * On devices with >= 1GB of memory: | |
520 | * -- set warnings to 100MB below the per-task limit. | |
521 | */ | |
522 | if (max_task_footprint_mb > 100) { | |
523 | max_task_footprint_warning_level = ((max_task_footprint_mb - 100) * 100) / max_task_footprint_mb; | |
524 | } | |
525 | } | |
526 | ||
527 | /* | |
528 | * Never allow warning level to land below the default. | |
529 | */ | |
530 | if (max_task_footprint_warning_level < PHYS_FOOTPRINT_WARNING_LEVEL) { | |
531 | max_task_footprint_warning_level = PHYS_FOOTPRINT_WARNING_LEVEL; | |
532 | } | |
533 | ||
534 | printf("Limiting task physical memory warning to %d%%\n", max_task_footprint_warning_level); | |
535 | ||
39236c6e | 536 | #else |
39037602 A |
537 | printf("Warning: max_task_pmem specified, but jetsam not configured; ignoring.\n"); |
538 | #endif /* CONFIG_MEMORYSTATUS */ | |
39236c6e A |
539 | } |
540 | ||
fe8ab488 A |
541 | #if MACH_ASSERT |
542 | PE_parse_boot_argn("pmap_ledgers_panic", &pmap_ledgers_panic, | |
543 | sizeof (pmap_ledgers_panic)); | |
544 | #endif /* MACH_ASSERT */ | |
545 | ||
39037602 | 546 | #if CONFIG_COREDUMP |
39236c6e A |
547 | if (!PE_parse_boot_argn("hwm_user_cores", &hwm_user_cores, |
548 | sizeof (hwm_user_cores))) { | |
549 | hwm_user_cores = 0; | |
550 | } | |
39037602 | 551 | #endif |
a1c7dba1 | 552 | |
39236c6e A |
553 | proc_init_cpumon_params(); |
554 | ||
555 | if (!PE_parse_boot_argn("task_wakeups_monitor_rate", &task_wakeups_monitor_rate, sizeof (task_wakeups_monitor_rate))) { | |
556 | task_wakeups_monitor_rate = TASK_WAKEUPS_MONITOR_DEFAULT_LIMIT; | |
557 | } | |
558 | ||
559 | if (!PE_parse_boot_argn("task_wakeups_monitor_interval", &task_wakeups_monitor_interval, sizeof (task_wakeups_monitor_interval))) { | |
560 | task_wakeups_monitor_interval = TASK_WAKEUPS_MONITOR_DEFAULT_INTERVAL; | |
561 | } | |
562 | ||
563 | if (!PE_parse_boot_argn("task_wakeups_monitor_ustackshots_trigger_pct", &task_wakeups_monitor_ustackshots_trigger_pct, | |
564 | sizeof (task_wakeups_monitor_ustackshots_trigger_pct))) { | |
565 | task_wakeups_monitor_ustackshots_trigger_pct = TASK_WAKEUPS_MONITOR_DEFAULT_USTACKSHOTS_TRIGGER; | |
566 | } | |
567 | ||
568 | if (!PE_parse_boot_argn("disable_exc_resource", &disable_exc_resource, | |
569 | sizeof (disable_exc_resource))) { | |
570 | disable_exc_resource = 0; | |
571 | } | |
572 | ||
39037602 A |
573 | if (!PE_parse_boot_argn("task_iomon_limit_mb", &task_iomon_limit_mb, sizeof (task_iomon_limit_mb))) { |
574 | task_iomon_limit_mb = IOMON_DEFAULT_LIMIT; | |
575 | } | |
576 | ||
577 | if (!PE_parse_boot_argn("task_iomon_interval_secs", &task_iomon_interval_secs, sizeof (task_iomon_interval_secs))) { | |
578 | task_iomon_interval_secs = IOMON_DEFAULT_INTERVAL; | |
579 | } | |
580 | ||
581 | if (!PE_parse_boot_argn("io_telemetry_limit", &io_telemetry_limit, sizeof (io_telemetry_limit))) { | |
582 | io_telemetry_limit = IO_TELEMETRY_DEFAULT_LIMIT; | |
583 | } | |
584 | ||
fe8ab488 A |
585 | /* |
586 | * If we have coalitions, coalition_init() will call init_task_ledgers() as it | |
587 | * sets up the ledgers for the default coalition. If we don't have coalitions, | |
588 | * then we have to call it now. | |
589 | */ | |
590 | #if CONFIG_COALITIONS | |
591 | assert(task_ledger_template); | |
592 | #else /* CONFIG_COALITIONS */ | |
316670eb | 593 | init_task_ledgers(); |
fe8ab488 | 594 | #endif /* CONFIG_COALITIONS */ |
316670eb | 595 | |
39236c6e | 596 | #if TASK_REFERENCE_LEAK_DEBUG |
39037602 | 597 | task_ref_btlog = btlog_create(TASK_REF_NUM_RECORDS, TASK_REF_BTDEPTH, TRUE /* caller_will_remove_entries_for_element? */); |
39236c6e A |
598 | assert(task_ref_btlog); |
599 | #endif | |
600 | ||
1c79356b A |
601 | /* |
602 | * Create the kernel task as the first task. | |
1c79356b | 603 | */ |
b0d623f7 | 604 | #ifdef __LP64__ |
39037602 | 605 | if (task_create_internal(TASK_NULL, NULL, FALSE, TRUE, TF_NONE, &kernel_task) != KERN_SUCCESS) |
b0d623f7 | 606 | #else |
39037602 | 607 | if (task_create_internal(TASK_NULL, NULL, FALSE, FALSE, TF_NONE, &kernel_task) != KERN_SUCCESS) |
b0d623f7 | 608 | #endif |
1c79356b | 609 | panic("task_init\n"); |
55e303ae | 610 | |
1c79356b A |
611 | vm_map_deallocate(kernel_task->map); |
612 | kernel_task->map = kernel_map; | |
4b17d6b6 | 613 | lck_spin_init(&dead_task_statistics_lock, &task_lck_grp, &task_lck_attr); |
1c79356b A |
614 | } |
615 | ||
1c79356b A |
616 | /* |
617 | * Create a task running in the kernel address space. It may | |
618 | * have its own map of size mem_size and may have ipc privileges. | |
619 | */ | |
620 | kern_return_t | |
621 | kernel_task_create( | |
91447636 A |
622 | __unused task_t parent_task, |
623 | __unused vm_offset_t map_base, | |
624 | __unused vm_size_t map_size, | |
625 | __unused task_t *child_task) | |
1c79356b | 626 | { |
55e303ae | 627 | return (KERN_INVALID_ARGUMENT); |
1c79356b A |
628 | } |
629 | ||
630 | kern_return_t | |
631 | task_create( | |
2d21ac55 | 632 | task_t parent_task, |
91447636 | 633 | __unused ledger_port_array_t ledger_ports, |
2d21ac55 A |
634 | __unused mach_msg_type_number_t num_ledger_ports, |
635 | __unused boolean_t inherit_memory, | |
636 | __unused task_t *child_task) /* OUT */ | |
1c79356b A |
637 | { |
638 | if (parent_task == TASK_NULL) | |
639 | return(KERN_INVALID_ARGUMENT); | |
640 | ||
2d21ac55 A |
641 | /* |
642 | * No longer supported: too many calls assume that a task has a valid | |
643 | * process attached. | |
644 | */ | |
645 | return(KERN_FAILURE); | |
1c79356b A |
646 | } |
647 | ||
648 | kern_return_t | |
649 | host_security_create_task_token( | |
91447636 | 650 | host_security_t host_security, |
2d21ac55 A |
651 | task_t parent_task, |
652 | __unused security_token_t sec_token, | |
653 | __unused audit_token_t audit_token, | |
654 | __unused host_priv_t host_priv, | |
91447636 A |
655 | __unused ledger_port_array_t ledger_ports, |
656 | __unused mach_msg_type_number_t num_ledger_ports, | |
2d21ac55 A |
657 | __unused boolean_t inherit_memory, |
658 | __unused task_t *child_task) /* OUT */ | |
1c79356b | 659 | { |
1c79356b A |
660 | if (parent_task == TASK_NULL) |
661 | return(KERN_INVALID_ARGUMENT); | |
662 | ||
663 | if (host_security == HOST_NULL) | |
664 | return(KERN_INVALID_SECURITY); | |
665 | ||
2d21ac55 A |
666 | /* |
667 | * No longer supported. | |
668 | */ | |
669 | return(KERN_FAILURE); | |
1c79356b A |
670 | } |
671 | ||
39236c6e A |
672 | /* |
673 | * Task ledgers | |
674 | * ------------ | |
675 | * | |
676 | * phys_footprint | |
677 | * Physical footprint: This is the sum of: | |
3e170ce0 A |
678 | * + (internal - alternate_accounting) |
679 | * + (internal_compressed - alternate_accounting_compressed) | |
fe8ab488 | 680 | * + iokit_mapped |
3e170ce0 A |
681 | * + purgeable_nonvolatile |
682 | * + purgeable_nonvolatile_compressed | |
39037602 | 683 | * + page_table |
39236c6e | 684 | * |
fe8ab488 A |
685 | * internal |
686 | * The task's anonymous memory, which on iOS is always resident. | |
687 | * | |
688 | * internal_compressed | |
689 | * Amount of this task's internal memory which is held by the compressor. | |
39236c6e A |
690 | * Such memory is no longer actually resident for the task [i.e., resident in its pmap], |
691 | * and could be either decompressed back into memory, or paged out to storage, depending | |
692 | * on our implementation. | |
fe8ab488 A |
693 | * |
694 | * iokit_mapped | |
695 | * IOKit mappings: The total size of all IOKit mappings in this task, regardless of | |
696 | clean/dirty or internal/external state]. | |
697 | * | |
698 | * alternate_accounting | |
699 | * The number of internal dirty pages which are part of IOKit mappings. By definition, these pages | |
700 | * are counted in both internal *and* iokit_mapped, so we must subtract them from the total to avoid | |
701 | * double counting. | |
39236c6e | 702 | */ |
316670eb A |
703 | void |
704 | init_task_ledgers(void) | |
705 | { | |
706 | ledger_template_t t; | |
707 | ||
708 | assert(task_ledger_template == NULL); | |
709 | assert(kernel_task == TASK_NULL); | |
710 | ||
39037602 A |
711 | #if MACH_ASSERT |
712 | PE_parse_boot_argn("pmap_ledgers_panic", &pmap_ledgers_panic, | |
713 | sizeof (pmap_ledgers_panic)); | |
714 | #endif /* MACH_ASSERT */ | |
715 | ||
316670eb A |
716 | if ((t = ledger_template_create("Per-task ledger")) == NULL) |
717 | panic("couldn't create task ledger template"); | |
718 | ||
719 | task_ledgers.cpu_time = ledger_entry_add(t, "cpu_time", "sched", "ns"); | |
720 | task_ledgers.tkm_private = ledger_entry_add(t, "tkm_private", | |
721 | "physmem", "bytes"); | |
722 | task_ledgers.tkm_shared = ledger_entry_add(t, "tkm_shared", "physmem", | |
723 | "bytes"); | |
724 | task_ledgers.phys_mem = ledger_entry_add(t, "phys_mem", "physmem", | |
725 | "bytes"); | |
726 | task_ledgers.wired_mem = ledger_entry_add(t, "wired_mem", "physmem", | |
727 | "bytes"); | |
fe8ab488 A |
728 | task_ledgers.internal = ledger_entry_add(t, "internal", "physmem", |
729 | "bytes"); | |
730 | task_ledgers.iokit_mapped = ledger_entry_add(t, "iokit_mapped", "mappings", | |
731 | "bytes"); | |
732 | task_ledgers.alternate_accounting = ledger_entry_add(t, "alternate_accounting", "physmem", | |
39236c6e | 733 | "bytes"); |
3e170ce0 A |
734 | task_ledgers.alternate_accounting_compressed = ledger_entry_add(t, "alternate_accounting_compressed", "physmem", |
735 | "bytes"); | |
39037602 A |
736 | task_ledgers.page_table = ledger_entry_add(t, "page_table", "physmem", |
737 | "bytes"); | |
39236c6e A |
738 | task_ledgers.phys_footprint = ledger_entry_add(t, "phys_footprint", "physmem", |
739 | "bytes"); | |
fe8ab488 | 740 | task_ledgers.internal_compressed = ledger_entry_add(t, "internal_compressed", "physmem", |
39236c6e | 741 | "bytes"); |
fe8ab488 A |
742 | task_ledgers.purgeable_volatile = ledger_entry_add(t, "purgeable_volatile", "physmem", "bytes"); |
743 | task_ledgers.purgeable_nonvolatile = ledger_entry_add(t, "purgeable_nonvolatile", "physmem", "bytes"); | |
744 | task_ledgers.purgeable_volatile_compressed = ledger_entry_add(t, "purgeable_volatile_compress", "physmem", "bytes"); | |
745 | task_ledgers.purgeable_nonvolatile_compressed = ledger_entry_add(t, "purgeable_nonvolatile_compress", "physmem", "bytes"); | |
4b17d6b6 | 746 | task_ledgers.platform_idle_wakeups = ledger_entry_add(t, "platform_idle_wakeups", "power", |
39236c6e | 747 | "count"); |
4b17d6b6 | 748 | task_ledgers.interrupt_wakeups = ledger_entry_add(t, "interrupt_wakeups", "power", |
39236c6e | 749 | "count"); |
fe8ab488 | 750 | |
3e170ce0 | 751 | #if CONFIG_SCHED_SFI |
fe8ab488 A |
752 | sfi_class_id_t class_id, ledger_alias; |
753 | for (class_id = SFI_CLASS_UNSPECIFIED; class_id < MAX_SFI_CLASS_ID; class_id++) { | |
754 | task_ledgers.sfi_wait_times[class_id] = -1; | |
755 | } | |
756 | ||
757 | /* don't account for UNSPECIFIED */ | |
758 | for (class_id = SFI_CLASS_UNSPECIFIED + 1; class_id < MAX_SFI_CLASS_ID; class_id++) { | |
759 | ledger_alias = sfi_get_ledger_alias_for_class(class_id); | |
760 | if (ledger_alias != SFI_CLASS_UNSPECIFIED) { | |
761 | /* Check to see if alias has been registered yet */ | |
762 | if (task_ledgers.sfi_wait_times[ledger_alias] != -1) { | |
763 | task_ledgers.sfi_wait_times[class_id] = task_ledgers.sfi_wait_times[ledger_alias]; | |
764 | } else { | |
765 | /* Otherwise, initialize it first */ | |
766 | task_ledgers.sfi_wait_times[class_id] = task_ledgers.sfi_wait_times[ledger_alias] = sfi_ledger_entry_add(t, ledger_alias); | |
767 | } | |
768 | } else { | |
769 | task_ledgers.sfi_wait_times[class_id] = sfi_ledger_entry_add(t, class_id); | |
770 | } | |
771 | ||
772 | if (task_ledgers.sfi_wait_times[class_id] < 0) { | |
773 | panic("couldn't create entries for task ledger template for SFI class 0x%x", class_id); | |
774 | } | |
775 | } | |
316670eb | 776 | |
3e170ce0 A |
777 | assert(task_ledgers.sfi_wait_times[MAX_SFI_CLASS_ID -1] != -1); |
778 | #endif /* CONFIG_SCHED_SFI */ | |
779 | ||
fe8ab488 A |
780 | #ifdef CONFIG_BANK |
781 | task_ledgers.cpu_time_billed_to_me = ledger_entry_add(t, "cpu_time_billed_to_me", "sched", "ns"); | |
782 | task_ledgers.cpu_time_billed_to_others = ledger_entry_add(t, "cpu_time_billed_to_others", "sched", "ns"); | |
783 | #endif | |
39037602 A |
784 | task_ledgers.physical_writes = ledger_entry_add(t, "physical_writes", "res", "bytes"); |
785 | task_ledgers.logical_writes = ledger_entry_add(t, "logical_writes", "res", "bytes"); | |
786 | ||
fe8ab488 A |
787 | if ((task_ledgers.cpu_time < 0) || |
788 | (task_ledgers.tkm_private < 0) || | |
789 | (task_ledgers.tkm_shared < 0) || | |
790 | (task_ledgers.phys_mem < 0) || | |
791 | (task_ledgers.wired_mem < 0) || | |
792 | (task_ledgers.internal < 0) || | |
793 | (task_ledgers.iokit_mapped < 0) || | |
794 | (task_ledgers.alternate_accounting < 0) || | |
3e170ce0 | 795 | (task_ledgers.alternate_accounting_compressed < 0) || |
39037602 | 796 | (task_ledgers.page_table < 0) || |
fe8ab488 A |
797 | (task_ledgers.phys_footprint < 0) || |
798 | (task_ledgers.internal_compressed < 0) || | |
799 | (task_ledgers.purgeable_volatile < 0) || | |
800 | (task_ledgers.purgeable_nonvolatile < 0) || | |
801 | (task_ledgers.purgeable_volatile_compressed < 0) || | |
802 | (task_ledgers.purgeable_nonvolatile_compressed < 0) || | |
803 | (task_ledgers.platform_idle_wakeups < 0) || | |
39037602 | 804 | (task_ledgers.interrupt_wakeups < 0) || |
fe8ab488 | 805 | #ifdef CONFIG_BANK |
39037602 | 806 | (task_ledgers.cpu_time_billed_to_me < 0) || (task_ledgers.cpu_time_billed_to_others < 0) || |
fe8ab488 | 807 | #endif |
39037602 A |
808 | (task_ledgers.physical_writes < 0) || |
809 | (task_ledgers.logical_writes < 0) | |
fe8ab488 | 810 | ) { |
316670eb A |
811 | panic("couldn't create entries for task ledger template"); |
812 | } | |
813 | ||
39037602 A |
814 | ledger_track_credit_only(t, task_ledgers.phys_footprint); |
815 | ledger_track_credit_only(t, task_ledgers.internal); | |
816 | ledger_track_credit_only(t, task_ledgers.internal_compressed); | |
817 | ledger_track_credit_only(t, task_ledgers.iokit_mapped); | |
818 | ledger_track_credit_only(t, task_ledgers.alternate_accounting); | |
819 | ledger_track_credit_only(t, task_ledgers.alternate_accounting_compressed); | |
820 | ledger_track_credit_only(t, task_ledgers.purgeable_volatile); | |
821 | ledger_track_credit_only(t, task_ledgers.purgeable_nonvolatile); | |
822 | ledger_track_credit_only(t, task_ledgers.purgeable_volatile_compressed); | |
823 | ledger_track_credit_only(t, task_ledgers.purgeable_nonvolatile_compressed); | |
824 | ||
39236c6e | 825 | ledger_track_maximum(t, task_ledgers.phys_footprint, 60); |
fe8ab488 A |
826 | #if MACH_ASSERT |
827 | if (pmap_ledgers_panic) { | |
828 | ledger_panic_on_negative(t, task_ledgers.phys_footprint); | |
39037602 | 829 | ledger_panic_on_negative(t, task_ledgers.page_table); |
fe8ab488 A |
830 | ledger_panic_on_negative(t, task_ledgers.internal); |
831 | ledger_panic_on_negative(t, task_ledgers.internal_compressed); | |
832 | ledger_panic_on_negative(t, task_ledgers.iokit_mapped); | |
833 | ledger_panic_on_negative(t, task_ledgers.alternate_accounting); | |
3e170ce0 | 834 | ledger_panic_on_negative(t, task_ledgers.alternate_accounting_compressed); |
fe8ab488 A |
835 | ledger_panic_on_negative(t, task_ledgers.purgeable_volatile); |
836 | ledger_panic_on_negative(t, task_ledgers.purgeable_nonvolatile); | |
837 | ledger_panic_on_negative(t, task_ledgers.purgeable_volatile_compressed); | |
838 | ledger_panic_on_negative(t, task_ledgers.purgeable_nonvolatile_compressed); | |
839 | } | |
840 | #endif /* MACH_ASSERT */ | |
39236c6e | 841 | |
39037602 | 842 | #if CONFIG_MEMORYSTATUS |
39236c6e | 843 | ledger_set_callback(t, task_ledgers.phys_footprint, task_footprint_exceeded, NULL, NULL); |
39037602 | 844 | #endif /* CONFIG_MEMORYSTATUS */ |
39236c6e A |
845 | |
846 | ledger_set_callback(t, task_ledgers.interrupt_wakeups, | |
847 | task_wakeups_rate_exceeded, NULL, NULL); | |
39037602 A |
848 | ledger_set_callback(t, task_ledgers.physical_writes, task_io_rate_exceeded, (void *)FLAVOR_IO_PHYSICAL_WRITES, NULL); |
849 | ledger_set_callback(t, task_ledgers.logical_writes, task_io_rate_exceeded, (void *)FLAVOR_IO_LOGICAL_WRITES, NULL); | |
316670eb A |
850 | task_ledger_template = t; |
851 | } | |
852 | ||
1c79356b | 853 | kern_return_t |
55e303ae | 854 | task_create_internal( |
1c79356b | 855 | task_t parent_task, |
3e170ce0 | 856 | coalition_t *parent_coalitions __unused, |
1c79356b | 857 | boolean_t inherit_memory, |
0c530ab8 | 858 | boolean_t is_64bit, |
39037602 | 859 | uint32_t t_flags, |
1c79356b A |
860 | task_t *child_task) /* OUT */ |
861 | { | |
2d21ac55 A |
862 | task_t new_task; |
863 | vm_shared_region_t shared_region; | |
316670eb | 864 | ledger_t ledger = NULL; |
1c79356b A |
865 | |
866 | new_task = (task_t) zalloc(task_zone); | |
867 | ||
868 | if (new_task == TASK_NULL) | |
869 | return(KERN_RESOURCE_SHORTAGE); | |
870 | ||
871 | /* one ref for just being alive; one for our caller */ | |
872 | new_task->ref_count = 2; | |
873 | ||
316670eb A |
874 | /* allocate with active entries */ |
875 | assert(task_ledger_template != NULL); | |
876 | if ((ledger = ledger_instantiate(task_ledger_template, | |
877 | LEDGER_CREATE_ACTIVE_ENTRIES)) == NULL) { | |
878 | zfree(task_zone, new_task); | |
879 | return(KERN_RESOURCE_SHORTAGE); | |
880 | } | |
39236c6e | 881 | |
316670eb A |
882 | new_task->ledger = ledger; |
883 | ||
fe8ab488 A |
884 | #if defined(CONFIG_SCHED_MULTIQ) |
885 | new_task->sched_group = sched_group_create(); | |
886 | #endif | |
887 | ||
b0d623f7 | 888 | /* if inherit_memory is true, parent_task MUST not be NULL */ |
39037602 A |
889 | if (!(t_flags & TF_CORPSE_FORK) && inherit_memory) |
890 | new_task->map = vm_map_fork(ledger, parent_task->map, 0); | |
1c79356b | 891 | else |
316670eb A |
892 | new_task->map = vm_map_create(pmap_create(ledger, 0, is_64bit), |
893 | (vm_map_offset_t)(VM_MIN_ADDRESS), | |
894 | (vm_map_offset_t)(VM_MAX_ADDRESS), TRUE); | |
1c79356b | 895 | |
2d21ac55 A |
896 | /* Inherit memlock limit from parent */ |
897 | if (parent_task) | |
b0d623f7 | 898 | vm_map_set_user_wire_limit(new_task->map, (vm_size_t)parent_task->map->user_wire_limit); |
2d21ac55 | 899 | |
b0d623f7 | 900 | lck_mtx_init(&new_task->lock, &task_lck_grp, &task_lck_attr); |
55e303ae | 901 | queue_init(&new_task->threads); |
1c79356b | 902 | new_task->suspend_count = 0; |
55e303ae | 903 | new_task->thread_count = 0; |
55e303ae | 904 | new_task->active_thread_count = 0; |
1c79356b | 905 | new_task->user_stop_count = 0; |
39236c6e | 906 | new_task->legacy_stop_count = 0; |
1c79356b | 907 | new_task->active = TRUE; |
b0d623f7 | 908 | new_task->halting = FALSE; |
2d21ac55 | 909 | new_task->user_data = NULL; |
55e303ae | 910 | new_task->priv_flags = 0; |
39037602 | 911 | new_task->t_flags = t_flags; |
39236c6e | 912 | new_task->importance = 0; |
39037602 A |
913 | new_task->corpse_info_kernel = NULL; |
914 | new_task->exec_token = 0; | |
1c79356b | 915 | |
fe8ab488 A |
916 | #if CONFIG_ATM |
917 | new_task->atm_context = NULL; | |
918 | #endif | |
919 | #if CONFIG_BANK | |
920 | new_task->bank_context = NULL; | |
921 | #endif | |
922 | ||
1c79356b | 923 | #ifdef MACH_BSD |
2d21ac55 | 924 | new_task->bsd_info = NULL; |
3e170ce0 | 925 | new_task->corpse_info = NULL; |
1c79356b A |
926 | #endif /* MACH_BSD */ |
927 | ||
39037602 A |
928 | #if CONFIG_MACF |
929 | new_task->crash_label = NULL; | |
930 | #endif | |
931 | ||
932 | #if CONFIG_MEMORYSTATUS | |
39236c6e A |
933 | if (max_task_footprint != 0) { |
934 | ledger_set_limit(ledger, task_ledgers.phys_footprint, max_task_footprint, PHYS_FOOTPRINT_WARNING_LEVEL); | |
935 | } | |
39037602 | 936 | #endif /* CONFIG_MEMORYSTATUS */ |
39236c6e A |
937 | |
938 | if (task_wakeups_monitor_rate != 0) { | |
939 | uint32_t flags = WAKEMON_ENABLE | WAKEMON_SET_DEFAULTS; | |
940 | int32_t rate; // Ignored because of WAKEMON_SET_DEFAULTS | |
941 | task_wakeups_monitor_ctl(new_task, &flags, &rate); | |
942 | } | |
943 | ||
39037602 A |
944 | #if CONFIG_IO_ACCOUNTING |
945 | uint32_t flags = IOMON_ENABLE; | |
946 | task_io_monitor_ctl(new_task, &flags); | |
947 | #endif /* CONFIG_IO_ACCOUNTING */ | |
948 | ||
b0d623f7 | 949 | #if defined(__i386__) || defined(__x86_64__) |
0c530ab8 A |
950 | new_task->i386_ldt = 0; |
951 | #endif | |
952 | ||
39236c6e | 953 | new_task->task_debug = NULL; |
55e303ae | 954 | |
39037602 A |
955 | #if DEVELOPMENT || DEBUG |
956 | new_task->task_unnested = FALSE; | |
957 | new_task->task_disconnected_count = 0; | |
958 | #endif | |
1c79356b | 959 | queue_init(&new_task->semaphore_list); |
1c79356b | 960 | new_task->semaphores_owned = 0; |
1c79356b | 961 | |
1c79356b A |
962 | ipc_task_init(new_task, parent_task); |
963 | ||
2d21ac55 | 964 | new_task->vtimers = 0; |
1c79356b | 965 | |
2d21ac55 A |
966 | new_task->shared_region = NULL; |
967 | ||
968 | new_task->affinity_space = NULL; | |
1c79356b | 969 | |
316670eb A |
970 | new_task->pidsuspended = FALSE; |
971 | new_task->frozen = FALSE; | |
39236c6e | 972 | new_task->changing_freeze_state = FALSE; |
316670eb A |
973 | new_task->rusage_cpu_flags = 0; |
974 | new_task->rusage_cpu_percentage = 0; | |
975 | new_task->rusage_cpu_interval = 0; | |
976 | new_task->rusage_cpu_deadline = 0; | |
977 | new_task->rusage_cpu_callt = NULL; | |
39236c6e A |
978 | #if MACH_ASSERT |
979 | new_task->suspends_outstanding = 0; | |
980 | #endif | |
981 | ||
fe8ab488 A |
982 | #if HYPERVISOR |
983 | new_task->hv_task_target = NULL; | |
984 | #endif /* HYPERVISOR */ | |
985 | ||
316670eb | 986 | |
39236c6e A |
987 | new_task->mem_notify_reserved = 0; |
988 | #if IMPORTANCE_INHERITANCE | |
fe8ab488 | 989 | new_task->task_imp_base = NULL; |
39236c6e A |
990 | #endif /* IMPORTANCE_INHERITANCE */ |
991 | ||
992 | #if defined(__x86_64__) | |
db609669 | 993 | new_task->uexc_range_start = new_task->uexc_range_size = new_task->uexc_handler = 0; |
39236c6e A |
994 | #endif |
995 | ||
996 | new_task->requested_policy = default_task_requested_policy; | |
997 | new_task->effective_policy = default_task_effective_policy; | |
db609669 | 998 | |
2d21ac55 | 999 | if (parent_task != TASK_NULL) { |
1c79356b | 1000 | new_task->sec_token = parent_task->sec_token; |
55e303ae | 1001 | new_task->audit_token = parent_task->audit_token; |
1c79356b | 1002 | |
2d21ac55 A |
1003 | /* inherit the parent's shared region */ |
1004 | shared_region = vm_shared_region_get(parent_task); | |
1005 | vm_shared_region_set(new_task, shared_region); | |
1c79356b | 1006 | |
91447636 A |
1007 | if(task_has_64BitAddr(parent_task)) |
1008 | task_set_64BitAddr(new_task); | |
b0d623f7 A |
1009 | new_task->all_image_info_addr = parent_task->all_image_info_addr; |
1010 | new_task->all_image_info_size = parent_task->all_image_info_size; | |
0c530ab8 | 1011 | |
b0d623f7 | 1012 | #if defined(__i386__) || defined(__x86_64__) |
0c530ab8 A |
1013 | if (inherit_memory && parent_task->i386_ldt) |
1014 | new_task->i386_ldt = user_ldt_copy(parent_task->i386_ldt); | |
1015 | #endif | |
2d21ac55 A |
1016 | if (inherit_memory && parent_task->affinity_space) |
1017 | task_affinity_create(parent_task, new_task); | |
b0d623f7 A |
1018 | |
1019 | new_task->pset_hint = parent_task->pset_hint = task_choose_pset(parent_task); | |
39236c6e A |
1020 | |
1021 | #if IMPORTANCE_INHERITANCE | |
fe8ab488 A |
1022 | ipc_importance_task_t new_task_imp = IIT_NULL; |
1023 | ||
1024 | if (task_is_marked_importance_donor(parent_task)) { | |
1025 | new_task_imp = ipc_importance_for_task(new_task, FALSE); | |
1026 | assert(IIT_NULL != new_task_imp); | |
1027 | ipc_importance_task_mark_donor(new_task_imp, TRUE); | |
1028 | } | |
39236c6e | 1029 | /* Embedded doesn't want this to inherit */ |
fe8ab488 A |
1030 | if (task_is_marked_importance_receiver(parent_task)) { |
1031 | if (IIT_NULL == new_task_imp) | |
1032 | new_task_imp = ipc_importance_for_task(new_task, FALSE); | |
1033 | assert(IIT_NULL != new_task_imp); | |
1034 | ipc_importance_task_mark_receiver(new_task_imp, TRUE); | |
1035 | } | |
1036 | if (task_is_marked_importance_denap_receiver(parent_task)) { | |
1037 | if (IIT_NULL == new_task_imp) | |
1038 | new_task_imp = ipc_importance_for_task(new_task, FALSE); | |
1039 | assert(IIT_NULL != new_task_imp); | |
1040 | ipc_importance_task_mark_denap_receiver(new_task_imp, TRUE); | |
1041 | } | |
1042 | ||
1043 | if (IIT_NULL != new_task_imp) { | |
1044 | assert(new_task->task_imp_base == new_task_imp); | |
1045 | ipc_importance_task_release(new_task_imp); | |
1046 | } | |
39236c6e A |
1047 | #endif /* IMPORTANCE_INHERITANCE */ |
1048 | ||
fe8ab488 A |
1049 | new_task->priority = BASEPRI_DEFAULT; |
1050 | new_task->max_priority = MAXPRI_USER; | |
1051 | ||
39037602 | 1052 | task_policy_create(new_task, parent_task); |
39236c6e | 1053 | } else { |
1c79356b | 1054 | new_task->sec_token = KERNEL_SECURITY_TOKEN; |
55e303ae | 1055 | new_task->audit_token = KERNEL_AUDIT_TOKEN; |
b0d623f7 A |
1056 | #ifdef __LP64__ |
1057 | if(is_64bit) | |
1058 | task_set_64BitAddr(new_task); | |
1059 | #endif | |
6d2010ae A |
1060 | new_task->all_image_info_addr = (mach_vm_address_t)0; |
1061 | new_task->all_image_info_size = (mach_vm_size_t)0; | |
b0d623f7 A |
1062 | |
1063 | new_task->pset_hint = PROCESSOR_SET_NULL; | |
fe8ab488 A |
1064 | |
1065 | if (kernel_task == TASK_NULL) { | |
1066 | new_task->priority = BASEPRI_KERNEL; | |
1067 | new_task->max_priority = MAXPRI_KERNEL; | |
1068 | } else { | |
1069 | new_task->priority = BASEPRI_DEFAULT; | |
1070 | new_task->max_priority = MAXPRI_USER; | |
1071 | } | |
1c79356b A |
1072 | } |
1073 | ||
3e170ce0 A |
1074 | bzero(new_task->coalition, sizeof(new_task->coalition)); |
1075 | for (int i = 0; i < COALITION_NUM_TYPES; i++) | |
1076 | queue_chain_init(new_task->task_coalition[i]); | |
fe8ab488 A |
1077 | |
1078 | /* Allocate I/O Statistics */ | |
1079 | new_task->task_io_stats = (io_stat_info_t)kalloc(sizeof(struct io_stat_info)); | |
1080 | assert(new_task->task_io_stats != NULL); | |
1081 | bzero(new_task->task_io_stats, sizeof(struct io_stat_info)); | |
1082 | ||
1083 | bzero(&(new_task->cpu_time_qos_stats), sizeof(struct _cpu_time_qos_stats)); | |
1084 | ||
6d2010ae | 1085 | bzero(&new_task->extmod_statistics, sizeof(new_task->extmod_statistics)); |
3e170ce0 | 1086 | |
39037602 A |
1087 | /* Copy resource acc. info from Parent for Corpe Forked task. */ |
1088 | if (parent_task != NULL && (t_flags & TF_CORPSE_FORK)) { | |
1089 | new_task->total_user_time = parent_task->total_user_time; | |
1090 | new_task->total_system_time = parent_task->total_system_time; | |
1091 | ledger_rollup(new_task->ledger, parent_task->ledger); | |
1092 | new_task->faults = parent_task->faults; | |
1093 | new_task->pageins = parent_task->pageins; | |
1094 | new_task->cow_faults = parent_task->cow_faults; | |
1095 | new_task->messages_sent = parent_task->messages_sent; | |
1096 | new_task->messages_received = parent_task->messages_received; | |
1097 | new_task->syscalls_mach = parent_task->syscalls_mach; | |
1098 | new_task->syscalls_unix = parent_task->syscalls_unix; | |
1099 | new_task->c_switch = parent_task->c_switch; | |
1100 | new_task->p_switch = parent_task->p_switch; | |
1101 | new_task->ps_switch = parent_task->ps_switch; | |
1102 | new_task->extmod_statistics = parent_task->extmod_statistics; | |
1103 | new_task->low_mem_notified_warn = parent_task->low_mem_notified_warn; | |
1104 | new_task->low_mem_notified_critical = parent_task->low_mem_notified_critical; | |
1105 | new_task->purged_memory_warn = parent_task->purged_memory_warn; | |
1106 | new_task->purged_memory_critical = parent_task->purged_memory_critical; | |
1107 | new_task->low_mem_privileged_listener = parent_task->low_mem_privileged_listener; | |
1108 | *new_task->task_io_stats = *parent_task->task_io_stats; | |
1109 | new_task->cpu_time_qos_stats = parent_task->cpu_time_qos_stats; | |
1110 | new_task->task_timer_wakeups_bin_1 = parent_task->task_timer_wakeups_bin_1; | |
1111 | new_task->task_timer_wakeups_bin_2 = parent_task->task_timer_wakeups_bin_2; | |
1112 | new_task->task_gpu_ns = parent_task->task_gpu_ns; | |
1113 | new_task->task_immediate_writes = parent_task->task_immediate_writes; | |
1114 | new_task->task_deferred_writes = parent_task->task_deferred_writes; | |
1115 | new_task->task_invalidated_writes = parent_task->task_invalidated_writes; | |
1116 | new_task->task_metadata_writes = parent_task->task_metadata_writes; | |
1117 | new_task->task_energy = parent_task->task_energy; | |
1118 | } else { | |
1119 | /* Initialize to zero for standard fork/spawn case */ | |
1120 | new_task->total_user_time = 0; | |
1121 | new_task->total_system_time = 0; | |
1122 | new_task->faults = 0; | |
1123 | new_task->pageins = 0; | |
1124 | new_task->cow_faults = 0; | |
1125 | new_task->messages_sent = 0; | |
1126 | new_task->messages_received = 0; | |
1127 | new_task->syscalls_mach = 0; | |
1128 | new_task->syscalls_unix = 0; | |
1129 | new_task->c_switch = 0; | |
1130 | new_task->p_switch = 0; | |
1131 | new_task->ps_switch = 0; | |
1132 | new_task->low_mem_notified_warn = 0; | |
1133 | new_task->low_mem_notified_critical = 0; | |
1134 | new_task->purged_memory_warn = 0; | |
1135 | new_task->purged_memory_critical = 0; | |
1136 | new_task->low_mem_privileged_listener = 0; | |
1137 | new_task->task_timer_wakeups_bin_1 = 0; | |
1138 | new_task->task_timer_wakeups_bin_2 = 0; | |
1139 | new_task->task_gpu_ns = 0; | |
1140 | new_task->task_immediate_writes = 0; | |
1141 | new_task->task_deferred_writes = 0; | |
1142 | new_task->task_invalidated_writes = 0; | |
1143 | new_task->task_metadata_writes = 0; | |
1144 | new_task->task_energy = 0; | |
1145 | } | |
3e170ce0 | 1146 | |
39037602 A |
1147 | |
1148 | #if CONFIG_COALITIONS | |
1149 | if (!(t_flags & TF_CORPSE_FORK)) { | |
1150 | /* TODO: there is no graceful failure path here... */ | |
1151 | if (parent_coalitions && parent_coalitions[COALITION_TYPE_RESOURCE]) { | |
1152 | coalitions_adopt_task(parent_coalitions, new_task); | |
1153 | } else if (parent_task && parent_task->coalition[COALITION_TYPE_RESOURCE]) { | |
1154 | /* | |
1155 | * all tasks at least have a resource coalition, so | |
1156 | * if the parent has one then inherit all coalitions | |
1157 | * the parent is a part of | |
1158 | */ | |
1159 | coalitions_adopt_task(parent_task->coalition, new_task); | |
1160 | } else { | |
1161 | /* TODO: assert that new_task will be PID 1 (launchd) */ | |
1162 | coalitions_adopt_init_task(new_task); | |
1163 | } | |
3e170ce0 | 1164 | } else { |
39037602 | 1165 | coalitions_adopt_corpse_task(new_task); |
3e170ce0 A |
1166 | } |
1167 | ||
1168 | if (new_task->coalition[COALITION_TYPE_RESOURCE] == COALITION_NULL) { | |
1169 | panic("created task is not a member of a resource coalition"); | |
1170 | } | |
1171 | #endif /* CONFIG_COALITIONS */ | |
1172 | ||
1173 | new_task->dispatchqueue_offset = 0; | |
1174 | if (parent_task != NULL) { | |
1175 | new_task->dispatchqueue_offset = parent_task->dispatchqueue_offset; | |
1176 | } | |
1c79356b | 1177 | |
55e303ae A |
1178 | if (vm_backing_store_low && parent_task != NULL) |
1179 | new_task->priv_flags |= (parent_task->priv_flags&VM_BACKING_STORE_PRIV); | |
1c79356b | 1180 | |
39236c6e | 1181 | new_task->task_volatile_objects = 0; |
fe8ab488 A |
1182 | new_task->task_nonvolatile_objects = 0; |
1183 | new_task->task_purgeable_disowning = FALSE; | |
1184 | new_task->task_purgeable_disowned = FALSE; | |
39236c6e | 1185 | |
39037602 A |
1186 | #if CONFIG_SECLUDED_MEMORY |
1187 | new_task->task_can_use_secluded_mem = FALSE; | |
1188 | new_task->task_could_use_secluded_mem = FALSE; | |
1189 | new_task->task_could_also_use_secluded_mem = FALSE; | |
1190 | #endif /* CONFIG_SECLUDED_MEMORY */ | |
1191 | ||
7e41aa88 A |
1192 | queue_init(&new_task->io_user_clients); |
1193 | ||
1c79356b A |
1194 | ipc_task_enable(new_task); |
1195 | ||
3e170ce0 A |
1196 | lck_mtx_lock(&tasks_threads_lock); |
1197 | queue_enter(&tasks, new_task, task_t, tasks); | |
1198 | tasks_count++; | |
4bd07ac2 A |
1199 | if (tasks_suspend_state) { |
1200 | task_suspend_internal(new_task); | |
1201 | } | |
3e170ce0 A |
1202 | lck_mtx_unlock(&tasks_threads_lock); |
1203 | ||
1c79356b A |
1204 | *child_task = new_task; |
1205 | return(KERN_SUCCESS); | |
1206 | } | |
1207 | ||
fe8ab488 A |
1208 | int task_dropped_imp_count = 0; |
1209 | ||
1c79356b | 1210 | /* |
91447636 | 1211 | * task_deallocate: |
1c79356b | 1212 | * |
91447636 | 1213 | * Drop a reference on a task. |
1c79356b A |
1214 | */ |
1215 | void | |
9bccf70c | 1216 | task_deallocate( |
1c79356b A |
1217 | task_t task) |
1218 | { | |
4b17d6b6 | 1219 | ledger_amount_t credit, debit, interrupt_wakeups, platform_idle_wakeups; |
fe8ab488 | 1220 | uint32_t refs; |
316670eb | 1221 | |
9bccf70c A |
1222 | if (task == TASK_NULL) |
1223 | return; | |
1224 | ||
fe8ab488 A |
1225 | refs = task_deallocate_internal(task); |
1226 | ||
1227 | #if IMPORTANCE_INHERITANCE | |
1228 | if (refs > 1) | |
9bccf70c | 1229 | return; |
fe8ab488 A |
1230 | |
1231 | if (refs == 1) { | |
1232 | /* | |
1233 | * If last ref potentially comes from the task's importance, | |
1234 | * disconnect it. But more task refs may be added before | |
1235 | * that completes, so wait for the reference to go to zero | |
1236 | * naturually (it may happen on a recursive task_deallocate() | |
1237 | * from the ipc_importance_disconnect_task() call). | |
1238 | */ | |
1239 | if (IIT_NULL != task->task_imp_base) | |
1240 | ipc_importance_disconnect_task(task); | |
1241 | return; | |
1242 | } | |
1243 | #else | |
1244 | if (refs > 0) | |
1245 | return; | |
1246 | #endif /* IMPORTANCE_INHERITANCE */ | |
1c79356b | 1247 | |
6d2010ae A |
1248 | lck_mtx_lock(&tasks_threads_lock); |
1249 | queue_remove(&terminated_tasks, task, task_t, tasks); | |
39236c6e | 1250 | terminated_tasks_count--; |
6d2010ae A |
1251 | lck_mtx_unlock(&tasks_threads_lock); |
1252 | ||
fe8ab488 A |
1253 | /* |
1254 | * remove the reference on atm descriptor | |
1255 | */ | |
490019cf | 1256 | task_atm_reset(task); |
fe8ab488 | 1257 | |
fe8ab488 A |
1258 | /* |
1259 | * remove the reference on bank context | |
1260 | */ | |
490019cf | 1261 | task_bank_reset(task); |
fe8ab488 A |
1262 | |
1263 | if (task->task_io_stats) | |
1264 | kfree(task->task_io_stats, sizeof(struct io_stat_info)); | |
1265 | ||
316670eb A |
1266 | /* |
1267 | * Give the machine dependent code a chance | |
1268 | * to perform cleanup before ripping apart | |
1269 | * the task. | |
1270 | */ | |
1271 | machine_task_terminate(task); | |
1272 | ||
9bccf70c A |
1273 | ipc_task_terminate(task); |
1274 | ||
7e41aa88 A |
1275 | /* let iokit know */ |
1276 | iokit_task_terminate(task); | |
1277 | ||
2d21ac55 A |
1278 | if (task->affinity_space) |
1279 | task_affinity_deallocate(task); | |
1280 | ||
fe8ab488 A |
1281 | #if MACH_ASSERT |
1282 | if (task->ledger != NULL && | |
1283 | task->map != NULL && | |
1284 | task->map->pmap != NULL && | |
1285 | task->map->pmap->ledger != NULL) { | |
1286 | assert(task->ledger == task->map->pmap->ledger); | |
1287 | } | |
1288 | #endif /* MACH_ASSERT */ | |
1289 | ||
1290 | vm_purgeable_disown(task); | |
1291 | assert(task->task_purgeable_disowned); | |
1292 | if (task->task_volatile_objects != 0 || | |
1293 | task->task_nonvolatile_objects != 0) { | |
1294 | panic("task_deallocate(%p): " | |
1295 | "volatile_objects=%d nonvolatile_objects=%d\n", | |
1296 | task, | |
1297 | task->task_volatile_objects, | |
1298 | task->task_nonvolatile_objects); | |
1299 | } | |
1300 | ||
1c79356b A |
1301 | vm_map_deallocate(task->map); |
1302 | is_release(task->itk_space); | |
1c79356b | 1303 | |
4b17d6b6 A |
1304 | ledger_get_entries(task->ledger, task_ledgers.interrupt_wakeups, |
1305 | &interrupt_wakeups, &debit); | |
1306 | ledger_get_entries(task->ledger, task_ledgers.platform_idle_wakeups, | |
1307 | &platform_idle_wakeups, &debit); | |
1308 | ||
fe8ab488 A |
1309 | #if defined(CONFIG_SCHED_MULTIQ) |
1310 | sched_group_destroy(task->sched_group); | |
1311 | #endif | |
1312 | ||
4b17d6b6 A |
1313 | /* Accumulate statistics for dead tasks */ |
1314 | lck_spin_lock(&dead_task_statistics_lock); | |
1315 | dead_task_statistics.total_user_time += task->total_user_time; | |
1316 | dead_task_statistics.total_system_time += task->total_system_time; | |
1317 | ||
1318 | dead_task_statistics.task_interrupt_wakeups += interrupt_wakeups; | |
1319 | dead_task_statistics.task_platform_idle_wakeups += platform_idle_wakeups; | |
1320 | ||
1321 | dead_task_statistics.task_timer_wakeups_bin_1 += task->task_timer_wakeups_bin_1; | |
1322 | dead_task_statistics.task_timer_wakeups_bin_2 += task->task_timer_wakeups_bin_2; | |
1323 | ||
1324 | lck_spin_unlock(&dead_task_statistics_lock); | |
b0d623f7 A |
1325 | lck_mtx_destroy(&task->lock, &task_lck_grp); |
1326 | ||
316670eb A |
1327 | if (!ledger_get_entries(task->ledger, task_ledgers.tkm_private, &credit, |
1328 | &debit)) { | |
1329 | OSAddAtomic64(credit, (int64_t *)&tasks_tkm_private.alloc); | |
1330 | OSAddAtomic64(debit, (int64_t *)&tasks_tkm_private.free); | |
1331 | } | |
1332 | if (!ledger_get_entries(task->ledger, task_ledgers.tkm_shared, &credit, | |
1333 | &debit)) { | |
1334 | OSAddAtomic64(credit, (int64_t *)&tasks_tkm_shared.alloc); | |
1335 | OSAddAtomic64(debit, (int64_t *)&tasks_tkm_shared.free); | |
1336 | } | |
1337 | ledger_dereference(task->ledger); | |
39236c6e A |
1338 | |
1339 | #if TASK_REFERENCE_LEAK_DEBUG | |
1340 | btlog_remove_entries_for_element(task_ref_btlog, task); | |
1341 | #endif | |
1342 | ||
fe8ab488 | 1343 | #if CONFIG_COALITIONS |
3e170ce0 | 1344 | task_release_coalitions(task); |
fe8ab488 A |
1345 | #endif /* CONFIG_COALITIONS */ |
1346 | ||
3e170ce0 A |
1347 | bzero(task->coalition, sizeof(task->coalition)); |
1348 | ||
1349 | #if MACH_BSD | |
1350 | /* clean up collected information since last reference to task is gone */ | |
1351 | if (task->corpse_info) { | |
39037602 | 1352 | task_crashinfo_destroy(task->corpse_info, RELEASE_CORPSE_REF); |
3e170ce0 A |
1353 | task->corpse_info = NULL; |
1354 | } | |
1355 | #endif | |
39037602 A |
1356 | if (task->corpse_info_kernel) { |
1357 | kfree(task->corpse_info_kernel, CORPSEINFO_ALLOCATION_SIZE); | |
1358 | } | |
1359 | ||
1360 | #if CONFIG_MACF | |
1361 | if (task->crash_label) { | |
1362 | mac_exc_action_label_task_destroy(task); | |
1363 | } | |
1364 | #endif | |
39236c6e | 1365 | |
91447636 | 1366 | zfree(task_zone, task); |
1c79356b A |
1367 | } |
1368 | ||
0c530ab8 A |
1369 | /* |
1370 | * task_name_deallocate: | |
1371 | * | |
1372 | * Drop a reference on a task name. | |
1373 | */ | |
1374 | void | |
1375 | task_name_deallocate( | |
1376 | task_name_t task_name) | |
1377 | { | |
1378 | return(task_deallocate((task_t)task_name)); | |
1379 | } | |
1380 | ||
39236c6e A |
1381 | /* |
1382 | * task_suspension_token_deallocate: | |
1383 | * | |
1384 | * Drop a reference on a task suspension token. | |
1385 | */ | |
1386 | void | |
1387 | task_suspension_token_deallocate( | |
1388 | task_suspension_token_t token) | |
1389 | { | |
1390 | return(task_deallocate((task_t)token)); | |
1391 | } | |
0c530ab8 | 1392 | |
3e170ce0 A |
1393 | |
1394 | /* | |
1395 | * task_collect_crash_info: | |
1396 | * | |
1397 | * collect crash info from bsd and mach based data | |
1398 | */ | |
1399 | kern_return_t | |
39037602 | 1400 | task_collect_crash_info(task_t task, struct proc *proc, int is_corpse_fork) |
3e170ce0 A |
1401 | { |
1402 | kern_return_t kr = KERN_SUCCESS; | |
1403 | ||
1404 | kcdata_descriptor_t crash_data = NULL; | |
1405 | kcdata_descriptor_t crash_data_release = NULL; | |
1406 | mach_msg_type_number_t size = CORPSEINFO_ALLOCATION_SIZE; | |
39037602 A |
1407 | mach_vm_offset_t crash_data_ptr = 0; |
1408 | void *crash_data_kernel = NULL; | |
1409 | void *crash_data_kernel_release = NULL; | |
1410 | int corpse_blob_kernel_alloc = (is_corpse_fork || unify_corpse_blob_alloc); | |
3e170ce0 A |
1411 | |
1412 | if (!corpses_enabled()) { | |
1413 | return KERN_NOT_SUPPORTED; | |
1414 | } | |
1415 | ||
1416 | task_lock(task); | |
39037602 A |
1417 | |
1418 | assert(is_corpse_fork || task->bsd_info != NULL); | |
1419 | if (task->corpse_info == NULL && (is_corpse_fork || task->bsd_info != NULL)) { | |
1420 | #if CONFIG_MACF | |
1421 | /* Update the corpse label, used by the exception delivery mac hook */ | |
1422 | mac_exc_action_label_task_update(task, proc); | |
1423 | #endif | |
3e170ce0 | 1424 | task_unlock(task); |
3e170ce0 | 1425 | |
39037602 A |
1426 | if (!corpse_blob_kernel_alloc) { |
1427 | /* map crash data memory in task's vm map */ | |
1428 | kr = mach_vm_allocate(task->map, &crash_data_ptr, size, (VM_MAKE_TAG(VM_MEMORY_CORPSEINFO) | VM_FLAGS_ANYWHERE)); | |
1429 | } else { | |
1430 | crash_data_kernel = (void *) kalloc(CORPSEINFO_ALLOCATION_SIZE); | |
1431 | if (crash_data_kernel == 0) | |
1432 | kr = KERN_RESOURCE_SHORTAGE; | |
1433 | bzero(crash_data_kernel, CORPSEINFO_ALLOCATION_SIZE); | |
1434 | crash_data_ptr = (mach_vm_offset_t) crash_data_kernel; | |
1435 | } | |
3e170ce0 A |
1436 | if (kr != KERN_SUCCESS) |
1437 | goto out_no_lock; | |
1438 | ||
39037602 A |
1439 | /* Do not get a corpse ref for corpse fork */ |
1440 | crash_data = task_crashinfo_alloc_init((mach_vm_address_t)crash_data_ptr, size, is_corpse_fork ? !GET_CORPSE_REF : GET_CORPSE_REF, corpse_blob_kernel_alloc ? KCFLAG_USE_MEMCOPY: KCFLAG_USE_COPYOUT); | |
3e170ce0 A |
1441 | if (crash_data) { |
1442 | task_lock(task); | |
1443 | crash_data_release = task->corpse_info; | |
39037602 | 1444 | crash_data_kernel_release = task->corpse_info_kernel; |
3e170ce0 | 1445 | task->corpse_info = crash_data; |
39037602 A |
1446 | task->corpse_info_kernel = crash_data_kernel; |
1447 | ||
3e170ce0 A |
1448 | task_unlock(task); |
1449 | kr = KERN_SUCCESS; | |
1450 | } else { | |
1451 | /* if failed to create corpse info, free the mapping */ | |
39037602 A |
1452 | if (!corpse_blob_kernel_alloc) { |
1453 | if (KERN_SUCCESS != mach_vm_deallocate(task->map, crash_data_ptr, size)) { | |
1454 | printf("mach_vm_deallocate failed to clear corpse_data for pid %d.\n", task_pid(task)); | |
1455 | } | |
1456 | } else { | |
1457 | kfree(crash_data_kernel, CORPSEINFO_ALLOCATION_SIZE); | |
3e170ce0 A |
1458 | } |
1459 | kr = KERN_FAILURE; | |
1460 | } | |
1461 | ||
1462 | if (crash_data_release != NULL) { | |
39037602 A |
1463 | task_crashinfo_destroy(crash_data_release, is_corpse_fork ? !RELEASE_CORPSE_REF : RELEASE_CORPSE_REF); |
1464 | } | |
1465 | if (crash_data_kernel_release != NULL) { | |
1466 | kfree(crash_data_kernel_release, CORPSEINFO_ALLOCATION_SIZE); | |
3e170ce0 A |
1467 | } |
1468 | } else { | |
1469 | task_unlock(task); | |
1470 | } | |
1471 | ||
1472 | out_no_lock: | |
1473 | return kr; | |
1474 | } | |
1475 | ||
1476 | /* | |
1477 | * task_deliver_crash_notification: | |
1478 | * | |
1479 | * Makes outcall to registered host port for a corpse. | |
1480 | */ | |
1481 | kern_return_t | |
39037602 | 1482 | task_deliver_crash_notification(task_t task, thread_t thread, mach_exception_data_type_t subcode) |
3e170ce0 A |
1483 | { |
1484 | kcdata_descriptor_t crash_info = task->corpse_info; | |
1485 | thread_t th_iter = NULL; | |
1486 | kern_return_t kr = KERN_SUCCESS; | |
1487 | wait_interrupt_t wsave; | |
1488 | mach_exception_data_type_t code[EXCEPTION_CODE_MAX]; | |
39037602 | 1489 | ipc_port_t task_port, old_notify; |
3e170ce0 A |
1490 | |
1491 | if (crash_info == NULL) | |
1492 | return KERN_FAILURE; | |
1493 | ||
3e170ce0 | 1494 | task_lock(task); |
39037602 A |
1495 | if (task_is_a_corpse_fork(task)) { |
1496 | /* Populate code with EXC_RESOURCE for corpse fork */ | |
1497 | code[0] = EXC_RESOURCE; | |
1498 | code[1] = subcode; | |
1499 | } else if (unify_corpse_blob_alloc) { | |
1500 | /* Populate code with EXC_CRASH for corpses */ | |
1501 | code[0] = EXC_CRASH; | |
1502 | code[1] = 0; | |
1503 | /* Update the code[1] if the boot-arg corpse_for_fatal_memkill is set */ | |
1504 | if (corpse_for_fatal_memkill) { | |
1505 | code[1] = subcode; | |
1506 | } | |
1507 | } else { | |
1508 | /* Populate code with address and length for EXC_CRASH */ | |
1509 | code[0] = crash_info->kcd_addr_begin; | |
1510 | code[1] = crash_info->kcd_length; | |
1511 | } | |
3e170ce0 A |
1512 | queue_iterate(&task->threads, th_iter, thread_t, task_threads) |
1513 | { | |
39037602 A |
1514 | if (th_iter->corpse_dup == FALSE) { |
1515 | ipc_thread_reset(th_iter); | |
1516 | } | |
3e170ce0 A |
1517 | } |
1518 | task_unlock(task); | |
1519 | ||
39037602 A |
1520 | /* Arm the no-sender notification for taskport */ |
1521 | task_reference(task); | |
1522 | task_port = convert_task_to_port(task); | |
1523 | ip_lock(task_port); | |
1524 | assert(ip_active(task_port)); | |
1525 | ipc_port_nsrequest(task_port, task_port->ip_mscount, ipc_port_make_sonce_locked(task_port), &old_notify); | |
1526 | /* port unlocked */ | |
1527 | assert(IP_NULL == old_notify); | |
1528 | ||
3e170ce0 | 1529 | wsave = thread_interrupt_level(THREAD_UNINT); |
39037602 | 1530 | kr = exception_triage_thread(EXC_CORPSE_NOTIFY, code, EXCEPTION_CODE_MAX, thread); |
3e170ce0 A |
1531 | if (kr != KERN_SUCCESS) { |
1532 | printf("Failed to send exception EXC_CORPSE_NOTIFY. error code: %d for pid %d\n", kr, task_pid(task)); | |
1533 | } | |
1534 | ||
3e170ce0 | 1535 | (void)thread_interrupt_level(wsave); |
3e170ce0 | 1536 | |
39037602 A |
1537 | /* |
1538 | * Drop the send right on task port, will fire the | |
1539 | * no-sender notification if exception deliver failed. | |
1540 | */ | |
1541 | ipc_port_release_send(task_port); | |
3e170ce0 A |
1542 | return kr; |
1543 | } | |
1544 | ||
1c79356b A |
1545 | /* |
1546 | * task_terminate: | |
1547 | * | |
1548 | * Terminate the specified task. See comments on thread_terminate | |
1549 | * (kern/thread.c) about problems with terminating the "current task." | |
1550 | */ | |
1551 | ||
1552 | kern_return_t | |
1553 | task_terminate( | |
1554 | task_t task) | |
1555 | { | |
1556 | if (task == TASK_NULL) | |
91447636 A |
1557 | return (KERN_INVALID_ARGUMENT); |
1558 | ||
1c79356b | 1559 | if (task->bsd_info) |
91447636 A |
1560 | return (KERN_FAILURE); |
1561 | ||
1c79356b A |
1562 | return (task_terminate_internal(task)); |
1563 | } | |
1564 | ||
fe8ab488 A |
1565 | #if MACH_ASSERT |
1566 | extern int proc_pid(struct proc *); | |
1567 | extern void proc_name_kdp(task_t t, char *buf, int size); | |
1568 | #endif /* MACH_ASSERT */ | |
1569 | ||
1570 | #define VM_MAP_PARTIAL_REAP 0x54 /* 0x150 */ | |
1571 | static void | |
1572 | __unused task_partial_reap(task_t task, __unused int pid) | |
1573 | { | |
1574 | unsigned int reclaimed_resident = 0; | |
1575 | unsigned int reclaimed_compressed = 0; | |
1576 | uint64_t task_page_count; | |
1577 | ||
1578 | task_page_count = (get_task_phys_footprint(task) / PAGE_SIZE_64); | |
1579 | ||
1580 | KERNEL_DEBUG_CONSTANT((MACHDBG_CODE(DBG_MACH_VM, VM_MAP_PARTIAL_REAP) | DBG_FUNC_START), | |
1581 | pid, task_page_count, 0, 0, 0); | |
1582 | ||
1583 | vm_map_partial_reap(task->map, &reclaimed_resident, &reclaimed_compressed); | |
1584 | ||
1585 | KERNEL_DEBUG_CONSTANT((MACHDBG_CODE(DBG_MACH_VM, VM_MAP_PARTIAL_REAP) | DBG_FUNC_END), | |
1586 | pid, reclaimed_resident, reclaimed_compressed, 0, 0); | |
1587 | } | |
1588 | ||
3e170ce0 A |
1589 | kern_return_t |
1590 | task_mark_corpse(task_t task) | |
1591 | { | |
1592 | kern_return_t kr = KERN_SUCCESS; | |
1593 | thread_t self_thread; | |
1594 | (void) self_thread; | |
1595 | wait_interrupt_t wsave; | |
1596 | ||
1597 | assert(task != kernel_task); | |
1598 | assert(task == current_task()); | |
1599 | assert(!task_is_a_corpse(task)); | |
1600 | ||
39037602 | 1601 | kr = task_collect_crash_info(task, (struct proc*)task->bsd_info, FALSE); |
3e170ce0 A |
1602 | if (kr != KERN_SUCCESS) { |
1603 | return kr; | |
1604 | } | |
1605 | ||
1606 | self_thread = current_thread(); | |
1607 | ||
1608 | wsave = thread_interrupt_level(THREAD_UNINT); | |
1609 | task_lock(task); | |
1610 | ||
1611 | task_set_corpse_pending_report(task); | |
1612 | task_set_corpse(task); | |
1613 | ||
1614 | kr = task_start_halt_locked(task, TRUE); | |
1615 | assert(kr == KERN_SUCCESS); | |
39037602 | 1616 | |
3e170ce0 | 1617 | ipc_task_reset(task); |
39037602 A |
1618 | /* Remove the naked send right for task port, needed to arm no sender notification */ |
1619 | task_set_special_port(task, TASK_KERNEL_PORT, IPC_PORT_NULL); | |
3e170ce0 A |
1620 | ipc_task_enable(task); |
1621 | ||
1622 | task_unlock(task); | |
1623 | /* terminate the ipc space */ | |
1624 | ipc_space_terminate(task->itk_space); | |
39037602 A |
1625 | |
1626 | /* Add it to global corpse task list */ | |
1627 | task_add_to_corpse_task_list(task); | |
3e170ce0 A |
1628 | |
1629 | task_start_halt(task); | |
1630 | thread_terminate_internal(self_thread); | |
39037602 | 1631 | |
3e170ce0 A |
1632 | (void) thread_interrupt_level(wsave); |
1633 | assert(task->halting == TRUE); | |
1634 | return kr; | |
1635 | } | |
1636 | ||
39037602 A |
1637 | /* |
1638 | * task_clear_corpse | |
1639 | * | |
1640 | * Clears the corpse pending bit on task. | |
1641 | * Removes inspection bit on the threads. | |
1642 | */ | |
1643 | void | |
1644 | task_clear_corpse(task_t task) | |
1645 | { | |
1646 | thread_t th_iter = NULL; | |
1647 | ||
1648 | task_lock(task); | |
1649 | queue_iterate(&task->threads, th_iter, thread_t, task_threads) | |
1650 | { | |
1651 | thread_mtx_lock(th_iter); | |
1652 | th_iter->inspection = FALSE; | |
1653 | thread_mtx_unlock(th_iter); | |
1654 | } | |
1655 | ||
1656 | thread_terminate_crashed_threads(); | |
1657 | /* remove the pending corpse report flag */ | |
1658 | task_clear_corpse_pending_report(task); | |
1659 | ||
1660 | task_unlock(task); | |
1661 | } | |
1662 | ||
1663 | /* | |
1664 | * task_port_notify | |
1665 | * | |
1666 | * Called whenever the Mach port system detects no-senders on | |
1667 | * the task port of a corpse. | |
1668 | * Each notification that comes in should terminate the task (corpse). | |
1669 | */ | |
1670 | void | |
1671 | task_port_notify(mach_msg_header_t *msg) | |
1672 | { | |
1673 | mach_no_senders_notification_t *notification = (void *)msg; | |
1674 | ipc_port_t port = notification->not_header.msgh_remote_port; | |
1675 | task_t task; | |
1676 | ||
1677 | assert(ip_active(port)); | |
1678 | assert(IKOT_TASK == ip_kotype(port)); | |
1679 | task = (task_t) port->ip_kobject; | |
1680 | ||
1681 | assert(task_is_a_corpse(task)); | |
1682 | ||
1683 | /* Remove the task from global corpse task list */ | |
1684 | task_remove_from_corpse_task_list(task); | |
1685 | ||
1686 | task_clear_corpse(task); | |
1687 | task_terminate_internal(task); | |
1688 | } | |
1689 | ||
1690 | /* | |
1691 | * task_wait_till_threads_terminate_locked | |
1692 | * | |
1693 | * Wait till all the threads in the task are terminated. | |
1694 | * Might release the task lock and re-acquire it. | |
1695 | */ | |
1696 | void | |
1697 | task_wait_till_threads_terminate_locked(task_t task) | |
1698 | { | |
1699 | /* wait for all the threads in the task to terminate */ | |
1700 | while (task->active_thread_count != 0) { | |
1701 | assert_wait((event_t)&task->active_thread_count, THREAD_UNINT); | |
1702 | task_unlock(task); | |
1703 | thread_block(THREAD_CONTINUE_NULL); | |
1704 | ||
1705 | task_lock(task); | |
1706 | } | |
1707 | } | |
1708 | ||
1709 | /* | |
1710 | * task_duplicate_map_and_threads | |
1711 | * | |
1712 | * Copy vmmap of source task. | |
1713 | * Copy active threads from source task to destination task. | |
1714 | * Source task would be suspended during the copy. | |
1715 | */ | |
1716 | kern_return_t | |
1717 | task_duplicate_map_and_threads( | |
1718 | task_t task, | |
1719 | void *p, | |
1720 | task_t new_task, | |
1721 | thread_t *thread_ret, | |
1722 | int is64bit, | |
1723 | uint64_t **udata_buffer, | |
1724 | int *size, | |
1725 | int *num_udata) | |
1726 | { | |
1727 | kern_return_t kr = KERN_SUCCESS; | |
1728 | int active; | |
1729 | thread_t thread, self, thread_return = THREAD_NULL; | |
1730 | thread_t new_thread = THREAD_NULL; | |
1731 | thread_t *thread_array; | |
1732 | uint32_t active_thread_count = 0, array_count = 0, i; | |
1733 | vm_map_t oldmap; | |
1734 | uint64_t *buffer = NULL; | |
1735 | int buf_size = 0; | |
1736 | int est_knotes = 0, num_knotes = 0; | |
1737 | ||
1738 | self = current_thread(); | |
1739 | ||
1740 | /* | |
1741 | * Suspend the task to copy thread state, use the internal | |
1742 | * variant so that no user-space process can resume | |
1743 | * the task from under us | |
1744 | */ | |
1745 | kr = task_suspend_internal(task); | |
1746 | if (kr != KERN_SUCCESS) { | |
1747 | return kr; | |
1748 | } | |
1749 | ||
1750 | if (task->map->disable_vmentry_reuse == TRUE) { | |
1751 | /* | |
1752 | * Quite likely GuardMalloc (or some debugging tool) | |
1753 | * is being used on this task. And it has gone through | |
1754 | * its limit. Making a corpse will likely encounter | |
1755 | * a lot of VM entries that will need COW. | |
1756 | * | |
1757 | * Skip it. | |
1758 | */ | |
1759 | task_resume_internal(task); | |
1760 | return KERN_FAILURE; | |
1761 | } | |
1762 | ||
1763 | /* Setup new task's vmmap, switch from parent task's map to it COW map */ | |
1764 | oldmap = new_task->map; | |
1765 | new_task->map = vm_map_fork(new_task->ledger, | |
1766 | task->map, | |
1767 | (VM_MAP_FORK_SHARE_IF_INHERIT_NONE | | |
1768 | VM_MAP_FORK_PRESERVE_PURGEABLE)); | |
1769 | vm_map_deallocate(oldmap); | |
1770 | ||
1771 | if (is64bit) { | |
1772 | vm_map_set_64bit(get_task_map(new_task)); | |
1773 | } else { | |
1774 | vm_map_set_32bit(get_task_map(new_task)); | |
1775 | } | |
1776 | ||
1777 | /* Get all the udata pointers from kqueue */ | |
1778 | est_knotes = proc_list_uptrs(p, NULL, 0); | |
1779 | if (est_knotes > 0) { | |
1780 | buf_size = (est_knotes + 32) * sizeof(uint64_t); | |
1781 | buffer = (uint64_t *) kalloc(buf_size); | |
1782 | num_knotes = proc_list_uptrs(p, buffer, buf_size); | |
1783 | if (num_knotes > est_knotes + 32) { | |
1784 | num_knotes = est_knotes + 32; | |
1785 | } | |
1786 | } | |
1787 | ||
1788 | active_thread_count = task->active_thread_count; | |
1789 | if (active_thread_count == 0) { | |
1790 | if (buffer != NULL) { | |
1791 | kfree(buffer, buf_size); | |
1792 | } | |
1793 | task_resume_internal(task); | |
1794 | return KERN_FAILURE; | |
1795 | } | |
1796 | ||
1797 | thread_array = (thread_t *) kalloc(sizeof(thread_t) * active_thread_count); | |
1798 | ||
1799 | /* Iterate all the threads and drop the task lock before calling thread_create_with_continuation */ | |
1800 | task_lock(task); | |
1801 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
1802 | /* Skip inactive threads */ | |
1803 | active = thread->active; | |
1804 | if (!active) { | |
1805 | continue; | |
1806 | } | |
1807 | ||
1808 | if (array_count >= active_thread_count) { | |
1809 | break; | |
1810 | } | |
1811 | ||
1812 | thread_array[array_count++] = thread; | |
1813 | thread_reference(thread); | |
1814 | } | |
1815 | task_unlock(task); | |
1816 | ||
1817 | for (i = 0; i < array_count; i++) { | |
1818 | ||
1819 | kr = thread_create_with_continuation(new_task, &new_thread, (thread_continue_t)thread_corpse_continue); | |
1820 | if (kr != KERN_SUCCESS) { | |
1821 | break; | |
1822 | } | |
1823 | ||
1824 | /* Equivalent of current thread in corpse */ | |
1825 | if (thread_array[i] == self) { | |
1826 | thread_return = new_thread; | |
1827 | } else { | |
1828 | /* drop the extra ref returned by thread_create_with_continuation */ | |
1829 | thread_deallocate(new_thread); | |
1830 | } | |
1831 | ||
1832 | kr = thread_dup2(thread_array[i], new_thread); | |
1833 | if (kr != KERN_SUCCESS) { | |
1834 | thread_mtx_lock(new_thread); | |
1835 | new_thread->corpse_dup = TRUE; | |
1836 | thread_mtx_unlock(new_thread); | |
1837 | continue; | |
1838 | } | |
1839 | ||
1840 | /* Copy thread name */ | |
1841 | bsd_copythreadname(new_thread->uthread, thread_array[i]->uthread); | |
1842 | thread_copy_resource_info(new_thread, thread_array[i]); | |
1843 | } | |
1844 | ||
1845 | task_resume_internal(task); | |
1846 | ||
1847 | for (i = 0; i < array_count; i++) { | |
1848 | thread_deallocate(thread_array[i]); | |
1849 | } | |
1850 | kfree(thread_array, sizeof(thread_t) * active_thread_count); | |
1851 | ||
1852 | if (kr == KERN_SUCCESS) { | |
1853 | *thread_ret = thread_return; | |
1854 | *udata_buffer = buffer; | |
1855 | *size = buf_size; | |
1856 | *num_udata = num_knotes; | |
1857 | } else { | |
1858 | if (thread_return != THREAD_NULL) { | |
1859 | thread_deallocate(thread_return); | |
1860 | } | |
1861 | if (buffer != NULL) { | |
1862 | kfree(buffer, buf_size); | |
1863 | } | |
1864 | } | |
1865 | ||
1866 | return kr; | |
1867 | } | |
1868 | ||
1869 | #if CONFIG_SECLUDED_MEMORY | |
1870 | extern void task_set_can_use_secluded_mem_locked( | |
1871 | task_t task, | |
1872 | boolean_t can_use_secluded_mem); | |
1873 | #endif /* CONFIG_SECLUDED_MEMORY */ | |
1874 | ||
1c79356b A |
1875 | kern_return_t |
1876 | task_terminate_internal( | |
91447636 | 1877 | task_t task) |
1c79356b | 1878 | { |
91447636 A |
1879 | thread_t thread, self; |
1880 | task_t self_task; | |
1881 | boolean_t interrupt_save; | |
fe8ab488 | 1882 | int pid = 0; |
1c79356b A |
1883 | |
1884 | assert(task != kernel_task); | |
1885 | ||
91447636 A |
1886 | self = current_thread(); |
1887 | self_task = self->task; | |
1c79356b A |
1888 | |
1889 | /* | |
1890 | * Get the task locked and make sure that we are not racing | |
1891 | * with someone else trying to terminate us. | |
1892 | */ | |
91447636 | 1893 | if (task == self_task) |
1c79356b | 1894 | task_lock(task); |
91447636 A |
1895 | else |
1896 | if (task < self_task) { | |
1c79356b | 1897 | task_lock(task); |
91447636 A |
1898 | task_lock(self_task); |
1899 | } | |
1900 | else { | |
1901 | task_lock(self_task); | |
1c79356b A |
1902 | task_lock(task); |
1903 | } | |
1904 | ||
39037602 A |
1905 | #if CONFIG_SECLUDED_MEMORY |
1906 | if (task->task_can_use_secluded_mem) { | |
1907 | task_set_can_use_secluded_mem_locked(task, FALSE); | |
1908 | } | |
1909 | task->task_could_use_secluded_mem = FALSE; | |
1910 | task->task_could_also_use_secluded_mem = FALSE; | |
1911 | #endif /* CONFIG_SECLUDED_MEMORY */ | |
1912 | ||
6d2010ae | 1913 | if (!task->active) { |
1c79356b | 1914 | /* |
6d2010ae | 1915 | * Task is already being terminated. |
1c79356b A |
1916 | * Just return an error. If we are dying, this will |
1917 | * just get us to our AST special handler and that | |
1918 | * will get us to finalize the termination of ourselves. | |
1919 | */ | |
1920 | task_unlock(task); | |
91447636 A |
1921 | if (self_task != task) |
1922 | task_unlock(self_task); | |
1923 | ||
1924 | return (KERN_FAILURE); | |
1c79356b | 1925 | } |
91447636 | 1926 | |
3e170ce0 A |
1927 | if (task_corpse_pending_report(task)) { |
1928 | /* | |
1929 | * Task is marked for reporting as corpse. | |
1930 | * Just return an error. This will | |
1931 | * just get us to our AST special handler and that | |
1932 | * will get us to finish the path to death | |
1933 | */ | |
1934 | task_unlock(task); | |
1935 | if (self_task != task) | |
1936 | task_unlock(self_task); | |
1937 | ||
1938 | return (KERN_FAILURE); | |
1939 | } | |
1940 | ||
91447636 A |
1941 | if (self_task != task) |
1942 | task_unlock(self_task); | |
1c79356b | 1943 | |
e7c99d92 A |
1944 | /* |
1945 | * Make sure the current thread does not get aborted out of | |
1946 | * the waits inside these operations. | |
1947 | */ | |
9bccf70c | 1948 | interrupt_save = thread_interrupt_level(THREAD_UNINT); |
e7c99d92 | 1949 | |
1c79356b A |
1950 | /* |
1951 | * Indicate that we want all the threads to stop executing | |
1952 | * at user space by holding the task (we would have held | |
1953 | * each thread independently in thread_terminate_internal - | |
1954 | * but this way we may be more likely to already find it | |
1955 | * held there). Mark the task inactive, and prevent | |
1956 | * further task operations via the task port. | |
1957 | */ | |
1958 | task_hold_locked(task); | |
1959 | task->active = FALSE; | |
1960 | ipc_task_disable(task); | |
1961 | ||
39236c6e A |
1962 | #if CONFIG_TELEMETRY |
1963 | /* | |
1964 | * Notify telemetry that this task is going away. | |
1965 | */ | |
1966 | telemetry_task_ctl_locked(task, TF_TELEMETRY, 0); | |
1967 | #endif | |
1968 | ||
1c79356b | 1969 | /* |
91447636 A |
1970 | * Terminate each thread in the task. |
1971 | */ | |
1972 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
1973 | thread_terminate_internal(thread); | |
1c79356b | 1974 | } |
e7c99d92 | 1975 | |
fe8ab488 A |
1976 | #ifdef MACH_BSD |
1977 | if (task->bsd_info != NULL) { | |
1978 | pid = proc_pid(task->bsd_info); | |
1979 | } | |
1980 | #endif /* MACH_BSD */ | |
1981 | ||
316670eb A |
1982 | task_unlock(task); |
1983 | ||
39037602 | 1984 | proc_set_task_policy(task, TASK_POLICY_ATTRIBUTE, |
3e170ce0 | 1985 | TASK_POLICY_TERMINATED, TASK_POLICY_ENABLE); |
a1c7dba1 | 1986 | |
fe8ab488 A |
1987 | /* Early object reap phase */ |
1988 | ||
1989 | // PR-17045188: Revisit implementation | |
1990 | // task_partial_reap(task, pid); | |
1991 | ||
1c79356b A |
1992 | |
1993 | /* | |
1994 | * Destroy all synchronizers owned by the task. | |
1995 | */ | |
1996 | task_synchronizer_destroy_all(task); | |
1997 | ||
1c79356b A |
1998 | /* |
1999 | * Destroy the IPC space, leaving just a reference for it. | |
2000 | */ | |
316670eb | 2001 | ipc_space_terminate(task->itk_space); |
1c79356b | 2002 | |
fe8ab488 A |
2003 | #if 00 |
2004 | /* if some ledgers go negative on tear-down again... */ | |
2005 | ledger_disable_panic_on_negative(task->map->pmap->ledger, | |
2006 | task_ledgers.phys_footprint); | |
2007 | ledger_disable_panic_on_negative(task->map->pmap->ledger, | |
2008 | task_ledgers.internal); | |
2009 | ledger_disable_panic_on_negative(task->map->pmap->ledger, | |
2010 | task_ledgers.internal_compressed); | |
2011 | ledger_disable_panic_on_negative(task->map->pmap->ledger, | |
2012 | task_ledgers.iokit_mapped); | |
2013 | ledger_disable_panic_on_negative(task->map->pmap->ledger, | |
2014 | task_ledgers.alternate_accounting); | |
3e170ce0 A |
2015 | ledger_disable_panic_on_negative(task->map->pmap->ledger, |
2016 | task_ledgers.alternate_accounting_compressed); | |
fe8ab488 | 2017 | #endif |
91447636 | 2018 | |
1c79356b A |
2019 | /* |
2020 | * If the current thread is a member of the task | |
2021 | * being terminated, then the last reference to | |
2022 | * the task will not be dropped until the thread | |
2023 | * is finally reaped. To avoid incurring the | |
2024 | * expense of removing the address space regions | |
2025 | * at reap time, we do it explictly here. | |
2026 | */ | |
3e170ce0 A |
2027 | |
2028 | vm_map_lock(task->map); | |
2029 | vm_map_disable_hole_optimization(task->map); | |
2030 | vm_map_unlock(task->map); | |
2031 | ||
2d21ac55 A |
2032 | vm_map_remove(task->map, |
2033 | task->map->min_offset, | |
2034 | task->map->max_offset, | |
3e170ce0 A |
2035 | /* no unnesting on final cleanup: */ |
2036 | VM_MAP_REMOVE_NO_UNNESTING); | |
1c79356b | 2037 | |
2d21ac55 A |
2038 | /* release our shared region */ |
2039 | vm_shared_region_set(task, NULL); | |
9bccf70c | 2040 | |
3e170ce0 | 2041 | |
fe8ab488 A |
2042 | #if MACH_ASSERT |
2043 | /* | |
2044 | * Identify the pmap's process, in case the pmap ledgers drift | |
2045 | * and we have to report it. | |
2046 | */ | |
2047 | char procname[17]; | |
2048 | if (task->bsd_info) { | |
2049 | pid = proc_pid(task->bsd_info); | |
2050 | proc_name_kdp(task, procname, sizeof (procname)); | |
2051 | } else { | |
2052 | pid = 0; | |
2053 | strlcpy(procname, "<unknown>", sizeof (procname)); | |
2054 | } | |
2055 | pmap_set_process(task->map->pmap, pid, procname); | |
2056 | #endif /* MACH_ASSERT */ | |
2057 | ||
b0d623f7 | 2058 | lck_mtx_lock(&tasks_threads_lock); |
2d21ac55 | 2059 | queue_remove(&tasks, task, task_t, tasks); |
6d2010ae | 2060 | queue_enter(&terminated_tasks, task, task_t, tasks); |
2d21ac55 | 2061 | tasks_count--; |
39236c6e | 2062 | terminated_tasks_count++; |
b0d623f7 | 2063 | lck_mtx_unlock(&tasks_threads_lock); |
9bccf70c | 2064 | |
1c79356b | 2065 | /* |
e7c99d92 A |
2066 | * We no longer need to guard against being aborted, so restore |
2067 | * the previous interruptible state. | |
2068 | */ | |
9bccf70c | 2069 | thread_interrupt_level(interrupt_save); |
e7c99d92 | 2070 | |
fe8ab488 A |
2071 | #if KPERF |
2072 | /* force the task to release all ctrs */ | |
2073 | if (task->t_chud & TASK_KPC_FORCED_ALL_CTRS) | |
2074 | kpc_force_all_ctrs(task, 0); | |
2075 | #endif | |
2076 | ||
2077 | #if CONFIG_COALITIONS | |
2078 | /* | |
3e170ce0 | 2079 | * Leave our coalitions. (drop activation but not reference) |
fe8ab488 | 2080 | */ |
3e170ce0 | 2081 | coalitions_remove_task(task); |
fe8ab488 A |
2082 | #endif |
2083 | ||
e7c99d92 A |
2084 | /* |
2085 | * Get rid of the task active reference on itself. | |
1c79356b | 2086 | */ |
1c79356b A |
2087 | task_deallocate(task); |
2088 | ||
91447636 | 2089 | return (KERN_SUCCESS); |
1c79356b A |
2090 | } |
2091 | ||
4bd07ac2 A |
2092 | void |
2093 | tasks_system_suspend(boolean_t suspend) | |
2094 | { | |
2095 | task_t task; | |
2096 | ||
2097 | lck_mtx_lock(&tasks_threads_lock); | |
2098 | assert(tasks_suspend_state != suspend); | |
2099 | tasks_suspend_state = suspend; | |
2100 | queue_iterate(&tasks, task, task_t, tasks) { | |
2101 | if (task == kernel_task) { | |
2102 | continue; | |
2103 | } | |
2104 | suspend ? task_suspend_internal(task) : task_resume_internal(task); | |
2105 | } | |
2106 | lck_mtx_unlock(&tasks_threads_lock); | |
2107 | } | |
2108 | ||
1c79356b | 2109 | /* |
b0d623f7 | 2110 | * task_start_halt: |
91447636 A |
2111 | * |
2112 | * Shut the current task down (except for the current thread) in | |
2113 | * preparation for dramatic changes to the task (probably exec). | |
b0d623f7 A |
2114 | * We hold the task and mark all other threads in the task for |
2115 | * termination. | |
1c79356b A |
2116 | */ |
2117 | kern_return_t | |
3e170ce0 A |
2118 | task_start_halt(task_t task) |
2119 | { | |
2120 | kern_return_t kr = KERN_SUCCESS; | |
2121 | task_lock(task); | |
2122 | kr = task_start_halt_locked(task, FALSE); | |
2123 | task_unlock(task); | |
2124 | return kr; | |
2125 | } | |
2126 | ||
2127 | static kern_return_t | |
2128 | task_start_halt_locked(task_t task, boolean_t should_mark_corpse) | |
1c79356b | 2129 | { |
3e170ce0 A |
2130 | thread_t thread, self; |
2131 | uint64_t dispatchqueue_offset; | |
1c79356b A |
2132 | |
2133 | assert(task != kernel_task); | |
2134 | ||
91447636 | 2135 | self = current_thread(); |
1c79356b | 2136 | |
39037602 | 2137 | if (task != self->task && !task_is_a_corpse_fork(task)) |
91447636 | 2138 | return (KERN_INVALID_ARGUMENT); |
1c79356b | 2139 | |
b0d623f7 | 2140 | if (task->halting || !task->active || !self->active) { |
1c79356b | 2141 | /* |
3e170ce0 A |
2142 | * Task or current thread is already being terminated. |
2143 | * Hurry up and return out of the current kernel context | |
2144 | * so that we run our AST special handler to terminate | |
2145 | * ourselves. | |
1c79356b | 2146 | */ |
91447636 | 2147 | return (KERN_FAILURE); |
1c79356b A |
2148 | } |
2149 | ||
b0d623f7 A |
2150 | task->halting = TRUE; |
2151 | ||
3e170ce0 A |
2152 | /* |
2153 | * Mark all the threads to keep them from starting any more | |
2154 | * user-level execution. The thread_terminate_internal code | |
2155 | * would do this on a thread by thread basis anyway, but this | |
2156 | * gives us a better chance of not having to wait there. | |
2157 | */ | |
2158 | task_hold_locked(task); | |
2159 | dispatchqueue_offset = get_dispatchqueue_offset_from_proc(task->bsd_info); | |
1c79356b | 2160 | |
3e170ce0 A |
2161 | /* |
2162 | * Terminate all the other threads in the task. | |
2163 | */ | |
2164 | queue_iterate(&task->threads, thread, thread_t, task_threads) | |
2165 | { | |
2166 | if (should_mark_corpse) { | |
2167 | thread_mtx_lock(thread); | |
2168 | thread->inspection = TRUE; | |
2169 | thread_mtx_unlock(thread); | |
1c79356b | 2170 | } |
3e170ce0 A |
2171 | if (thread != self) |
2172 | thread_terminate_internal(thread); | |
1c79356b | 2173 | } |
3e170ce0 A |
2174 | task->dispatchqueue_offset = dispatchqueue_offset; |
2175 | ||
2176 | task_release_locked(task); | |
2177 | ||
b0d623f7 A |
2178 | return KERN_SUCCESS; |
2179 | } | |
2180 | ||
2181 | ||
2182 | /* | |
2183 | * task_complete_halt: | |
2184 | * | |
2185 | * Complete task halt by waiting for threads to terminate, then clean | |
2186 | * up task resources (VM, port namespace, etc...) and then let the | |
2187 | * current thread go in the (practically empty) task context. | |
2188 | */ | |
2189 | void | |
2190 | task_complete_halt(task_t task) | |
2191 | { | |
2192 | task_lock(task); | |
2193 | assert(task->halting); | |
2194 | assert(task == current_task()); | |
e7c99d92 | 2195 | |
b0d623f7 A |
2196 | /* |
2197 | * Wait for the other threads to get shut down. | |
2198 | * When the last other thread is reaped, we'll be | |
316670eb | 2199 | * woken up. |
b0d623f7 A |
2200 | */ |
2201 | if (task->thread_count > 1) { | |
2202 | assert_wait((event_t)&task->halting, THREAD_UNINT); | |
2203 | task_unlock(task); | |
2204 | thread_block(THREAD_CONTINUE_NULL); | |
2205 | } else { | |
2206 | task_unlock(task); | |
2207 | } | |
1c79356b | 2208 | |
316670eb A |
2209 | /* |
2210 | * Give the machine dependent code a chance | |
2211 | * to perform cleanup of task-level resources | |
2212 | * associated with the current thread before | |
2213 | * ripping apart the task. | |
2214 | */ | |
2215 | machine_task_terminate(task); | |
2216 | ||
1c79356b A |
2217 | /* |
2218 | * Destroy all synchronizers owned by the task. | |
2219 | */ | |
2220 | task_synchronizer_destroy_all(task); | |
2221 | ||
2222 | /* | |
9bccf70c A |
2223 | * Destroy the contents of the IPC space, leaving just |
2224 | * a reference for it. | |
e7c99d92 | 2225 | */ |
55e303ae | 2226 | ipc_space_clean(task->itk_space); |
1c79356b A |
2227 | |
2228 | /* | |
2229 | * Clean out the address space, as we are going to be | |
2230 | * getting a new one. | |
2231 | */ | |
91447636 | 2232 | vm_map_remove(task->map, task->map->min_offset, |
3e170ce0 A |
2233 | task->map->max_offset, |
2234 | /* no unnesting on final cleanup: */ | |
2235 | VM_MAP_REMOVE_NO_UNNESTING); | |
1c79356b | 2236 | |
39037602 A |
2237 | /* |
2238 | * Kick out any IOKitUser handles to the task. At best they're stale, | |
2239 | * at worst someone is racing a SUID exec. | |
2240 | */ | |
2241 | iokit_task_terminate(task); | |
2242 | ||
b0d623f7 | 2243 | task->halting = FALSE; |
1c79356b A |
2244 | } |
2245 | ||
2246 | /* | |
2247 | * task_hold_locked: | |
2248 | * | |
2249 | * Suspend execution of the specified task. | |
2250 | * This is a recursive-style suspension of the task, a count of | |
2251 | * suspends is maintained. | |
2252 | * | |
2253 | * CONDITIONS: the task is locked and active. | |
2254 | */ | |
2255 | void | |
2256 | task_hold_locked( | |
39037602 | 2257 | task_t task) |
1c79356b | 2258 | { |
39037602 | 2259 | thread_t thread; |
1c79356b A |
2260 | |
2261 | assert(task->active); | |
2262 | ||
9bccf70c A |
2263 | if (task->suspend_count++ > 0) |
2264 | return; | |
1c79356b A |
2265 | |
2266 | /* | |
91447636 | 2267 | * Iterate through all the threads and hold them. |
1c79356b | 2268 | */ |
91447636 A |
2269 | queue_iterate(&task->threads, thread, thread_t, task_threads) { |
2270 | thread_mtx_lock(thread); | |
2271 | thread_hold(thread); | |
2272 | thread_mtx_unlock(thread); | |
1c79356b A |
2273 | } |
2274 | } | |
2275 | ||
2276 | /* | |
2277 | * task_hold: | |
2278 | * | |
2279 | * Same as the internal routine above, except that is must lock | |
2280 | * and verify that the task is active. This differs from task_suspend | |
2281 | * in that it places a kernel hold on the task rather than just a | |
2282 | * user-level hold. This keeps users from over resuming and setting | |
2283 | * it running out from under the kernel. | |
2284 | * | |
2285 | * CONDITIONS: the caller holds a reference on the task | |
2286 | */ | |
2287 | kern_return_t | |
91447636 | 2288 | task_hold( |
39037602 | 2289 | task_t task) |
1c79356b | 2290 | { |
1c79356b A |
2291 | if (task == TASK_NULL) |
2292 | return (KERN_INVALID_ARGUMENT); | |
91447636 | 2293 | |
1c79356b | 2294 | task_lock(task); |
91447636 | 2295 | |
1c79356b A |
2296 | if (!task->active) { |
2297 | task_unlock(task); | |
91447636 | 2298 | |
1c79356b A |
2299 | return (KERN_FAILURE); |
2300 | } | |
1c79356b | 2301 | |
91447636 A |
2302 | task_hold_locked(task); |
2303 | task_unlock(task); | |
2304 | ||
2305 | return (KERN_SUCCESS); | |
1c79356b A |
2306 | } |
2307 | ||
316670eb A |
2308 | kern_return_t |
2309 | task_wait( | |
2310 | task_t task, | |
2311 | boolean_t until_not_runnable) | |
2312 | { | |
2313 | if (task == TASK_NULL) | |
2314 | return (KERN_INVALID_ARGUMENT); | |
2315 | ||
2316 | task_lock(task); | |
2317 | ||
2318 | if (!task->active) { | |
2319 | task_unlock(task); | |
2320 | ||
2321 | return (KERN_FAILURE); | |
2322 | } | |
2323 | ||
2324 | task_wait_locked(task, until_not_runnable); | |
2325 | task_unlock(task); | |
2326 | ||
2327 | return (KERN_SUCCESS); | |
2328 | } | |
2329 | ||
1c79356b | 2330 | /* |
91447636 A |
2331 | * task_wait_locked: |
2332 | * | |
1c79356b A |
2333 | * Wait for all threads in task to stop. |
2334 | * | |
2335 | * Conditions: | |
2336 | * Called with task locked, active, and held. | |
2337 | */ | |
2338 | void | |
2339 | task_wait_locked( | |
39037602 | 2340 | task_t task, |
316670eb | 2341 | boolean_t until_not_runnable) |
1c79356b | 2342 | { |
39037602 | 2343 | thread_t thread, self; |
1c79356b A |
2344 | |
2345 | assert(task->active); | |
2346 | assert(task->suspend_count > 0); | |
2347 | ||
91447636 A |
2348 | self = current_thread(); |
2349 | ||
1c79356b | 2350 | /* |
91447636 | 2351 | * Iterate through all the threads and wait for them to |
1c79356b A |
2352 | * stop. Do not wait for the current thread if it is within |
2353 | * the task. | |
2354 | */ | |
91447636 A |
2355 | queue_iterate(&task->threads, thread, thread_t, task_threads) { |
2356 | if (thread != self) | |
316670eb | 2357 | thread_wait(thread, until_not_runnable); |
1c79356b A |
2358 | } |
2359 | } | |
2360 | ||
2361 | /* | |
2362 | * task_release_locked: | |
2363 | * | |
2364 | * Release a kernel hold on a task. | |
2365 | * | |
2366 | * CONDITIONS: the task is locked and active | |
2367 | */ | |
2368 | void | |
2369 | task_release_locked( | |
39037602 | 2370 | task_t task) |
1c79356b | 2371 | { |
39037602 | 2372 | thread_t thread; |
1c79356b A |
2373 | |
2374 | assert(task->active); | |
9bccf70c | 2375 | assert(task->suspend_count > 0); |
1c79356b | 2376 | |
9bccf70c A |
2377 | if (--task->suspend_count > 0) |
2378 | return; | |
1c79356b | 2379 | |
91447636 A |
2380 | queue_iterate(&task->threads, thread, thread_t, task_threads) { |
2381 | thread_mtx_lock(thread); | |
2382 | thread_release(thread); | |
2383 | thread_mtx_unlock(thread); | |
1c79356b A |
2384 | } |
2385 | } | |
2386 | ||
2387 | /* | |
2388 | * task_release: | |
2389 | * | |
2390 | * Same as the internal routine above, except that it must lock | |
2391 | * and verify that the task is active. | |
2392 | * | |
2393 | * CONDITIONS: The caller holds a reference to the task | |
2394 | */ | |
2395 | kern_return_t | |
91447636 A |
2396 | task_release( |
2397 | task_t task) | |
1c79356b | 2398 | { |
1c79356b A |
2399 | if (task == TASK_NULL) |
2400 | return (KERN_INVALID_ARGUMENT); | |
91447636 | 2401 | |
1c79356b | 2402 | task_lock(task); |
91447636 | 2403 | |
1c79356b A |
2404 | if (!task->active) { |
2405 | task_unlock(task); | |
91447636 | 2406 | |
1c79356b A |
2407 | return (KERN_FAILURE); |
2408 | } | |
1c79356b | 2409 | |
91447636 A |
2410 | task_release_locked(task); |
2411 | task_unlock(task); | |
2412 | ||
2413 | return (KERN_SUCCESS); | |
1c79356b A |
2414 | } |
2415 | ||
2416 | kern_return_t | |
2417 | task_threads( | |
91447636 A |
2418 | task_t task, |
2419 | thread_act_array_t *threads_out, | |
1c79356b A |
2420 | mach_msg_type_number_t *count) |
2421 | { | |
91447636 | 2422 | mach_msg_type_number_t actual; |
2d21ac55 | 2423 | thread_t *thread_list; |
91447636 A |
2424 | thread_t thread; |
2425 | vm_size_t size, size_needed; | |
2426 | void *addr; | |
2427 | unsigned int i, j; | |
1c79356b A |
2428 | |
2429 | if (task == TASK_NULL) | |
91447636 | 2430 | return (KERN_INVALID_ARGUMENT); |
1c79356b | 2431 | |
2d21ac55 | 2432 | size = 0; addr = NULL; |
1c79356b A |
2433 | |
2434 | for (;;) { | |
2435 | task_lock(task); | |
2436 | if (!task->active) { | |
2437 | task_unlock(task); | |
91447636 | 2438 | |
1c79356b A |
2439 | if (size != 0) |
2440 | kfree(addr, size); | |
91447636 A |
2441 | |
2442 | return (KERN_FAILURE); | |
1c79356b A |
2443 | } |
2444 | ||
55e303ae | 2445 | actual = task->thread_count; |
1c79356b A |
2446 | |
2447 | /* do we have the memory we need? */ | |
91447636 | 2448 | size_needed = actual * sizeof (mach_port_t); |
1c79356b A |
2449 | if (size_needed <= size) |
2450 | break; | |
2451 | ||
2452 | /* unlock the task and allocate more memory */ | |
2453 | task_unlock(task); | |
2454 | ||
2455 | if (size != 0) | |
2456 | kfree(addr, size); | |
2457 | ||
2458 | assert(size_needed > 0); | |
2459 | size = size_needed; | |
2460 | ||
2461 | addr = kalloc(size); | |
2462 | if (addr == 0) | |
91447636 | 2463 | return (KERN_RESOURCE_SHORTAGE); |
1c79356b A |
2464 | } |
2465 | ||
2466 | /* OK, have memory and the task is locked & active */ | |
2d21ac55 | 2467 | thread_list = (thread_t *)addr; |
91447636 A |
2468 | |
2469 | i = j = 0; | |
2470 | ||
2471 | for (thread = (thread_t)queue_first(&task->threads); i < actual; | |
2472 | ++i, thread = (thread_t)queue_next(&thread->task_threads)) { | |
2473 | thread_reference_internal(thread); | |
2d21ac55 | 2474 | thread_list[j++] = thread; |
1c79356b | 2475 | } |
91447636 A |
2476 | |
2477 | assert(queue_end(&task->threads, (queue_entry_t)thread)); | |
1c79356b A |
2478 | |
2479 | actual = j; | |
91447636 | 2480 | size_needed = actual * sizeof (mach_port_t); |
1c79356b | 2481 | |
91447636 | 2482 | /* can unlock task now that we've got the thread refs */ |
1c79356b A |
2483 | task_unlock(task); |
2484 | ||
2485 | if (actual == 0) { | |
91447636 | 2486 | /* no threads, so return null pointer and deallocate memory */ |
1c79356b | 2487 | |
2d21ac55 | 2488 | *threads_out = NULL; |
1c79356b A |
2489 | *count = 0; |
2490 | ||
2491 | if (size != 0) | |
2492 | kfree(addr, size); | |
91447636 A |
2493 | } |
2494 | else { | |
1c79356b A |
2495 | /* if we allocated too much, must copy */ |
2496 | ||
2497 | if (size_needed < size) { | |
91447636 | 2498 | void *newaddr; |
1c79356b A |
2499 | |
2500 | newaddr = kalloc(size_needed); | |
2501 | if (newaddr == 0) { | |
91447636 | 2502 | for (i = 0; i < actual; ++i) |
2d21ac55 | 2503 | thread_deallocate(thread_list[i]); |
1c79356b | 2504 | kfree(addr, size); |
91447636 | 2505 | return (KERN_RESOURCE_SHORTAGE); |
1c79356b A |
2506 | } |
2507 | ||
91447636 | 2508 | bcopy(addr, newaddr, size_needed); |
1c79356b | 2509 | kfree(addr, size); |
2d21ac55 | 2510 | thread_list = (thread_t *)newaddr; |
1c79356b A |
2511 | } |
2512 | ||
2d21ac55 | 2513 | *threads_out = thread_list; |
1c79356b A |
2514 | *count = actual; |
2515 | ||
2516 | /* do the conversion that Mig should handle */ | |
2517 | ||
91447636 | 2518 | for (i = 0; i < actual; ++i) |
2d21ac55 | 2519 | ((ipc_port_t *) thread_list)[i] = convert_thread_to_port(thread_list[i]); |
1c79356b A |
2520 | } |
2521 | ||
91447636 | 2522 | return (KERN_SUCCESS); |
1c79356b A |
2523 | } |
2524 | ||
39236c6e A |
2525 | #define TASK_HOLD_NORMAL 0 |
2526 | #define TASK_HOLD_PIDSUSPEND 1 | |
2527 | #define TASK_HOLD_LEGACY 2 | |
2528 | #define TASK_HOLD_LEGACY_ALL 3 | |
2529 | ||
316670eb A |
2530 | static kern_return_t |
2531 | place_task_hold ( | |
39037602 | 2532 | task_t task, |
39236c6e | 2533 | int mode) |
316670eb | 2534 | { |
39037602 | 2535 | if (!task->active && !task_is_a_corpse(task)) { |
1c79356b A |
2536 | return (KERN_FAILURE); |
2537 | } | |
91447636 | 2538 | |
39037602 A |
2539 | /* Return success for corpse task */ |
2540 | if (task_is_a_corpse(task)) { | |
2541 | return KERN_SUCCESS; | |
2542 | } | |
2543 | ||
39236c6e A |
2544 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, |
2545 | MACHDBG_CODE(DBG_MACH_IPC,MACH_TASK_SUSPEND) | DBG_FUNC_NONE, | |
3e170ce0 | 2546 | task_pid(task), ((thread_t)queue_first(&task->threads))->thread_id, |
39236c6e A |
2547 | task->user_stop_count, task->user_stop_count + 1, 0); |
2548 | ||
2549 | #if MACH_ASSERT | |
2550 | current_task()->suspends_outstanding++; | |
2551 | #endif | |
2552 | ||
2553 | if (mode == TASK_HOLD_LEGACY) | |
2554 | task->legacy_stop_count++; | |
2555 | ||
91447636 | 2556 | if (task->user_stop_count++ > 0) { |
1c79356b A |
2557 | /* |
2558 | * If the stop count was positive, the task is | |
2559 | * already stopped and we can exit. | |
2560 | */ | |
1c79356b A |
2561 | return (KERN_SUCCESS); |
2562 | } | |
2563 | ||
2564 | /* | |
2565 | * Put a kernel-level hold on the threads in the task (all | |
2566 | * user-level task suspensions added together represent a | |
2567 | * single kernel-level hold). We then wait for the threads | |
2568 | * to stop executing user code. | |
2569 | */ | |
2570 | task_hold_locked(task); | |
39236c6e | 2571 | task_wait_locked(task, FALSE); |
316670eb A |
2572 | |
2573 | return (KERN_SUCCESS); | |
2574 | } | |
2575 | ||
2576 | static kern_return_t | |
2577 | release_task_hold ( | |
39037602 | 2578 | task_t task, |
39236c6e | 2579 | int mode) |
316670eb | 2580 | { |
39037602 | 2581 | boolean_t release = FALSE; |
316670eb | 2582 | |
39037602 | 2583 | if (!task->active && !task_is_a_corpse(task)) { |
316670eb A |
2584 | return (KERN_FAILURE); |
2585 | } | |
39037602 A |
2586 | |
2587 | /* Return success for corpse task */ | |
2588 | if (task_is_a_corpse(task)) { | |
2589 | return KERN_SUCCESS; | |
2590 | } | |
316670eb | 2591 | |
39236c6e | 2592 | if (mode == TASK_HOLD_PIDSUSPEND) { |
316670eb | 2593 | if (task->pidsuspended == FALSE) { |
39236c6e | 2594 | return (KERN_FAILURE); |
316670eb A |
2595 | } |
2596 | task->pidsuspended = FALSE; | |
2597 | } | |
2598 | ||
39236c6e A |
2599 | if (task->user_stop_count > (task->pidsuspended ? 1 : 0)) { |
2600 | ||
2601 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, | |
2602 | MACHDBG_CODE(DBG_MACH_IPC,MACH_TASK_RESUME) | DBG_FUNC_NONE, | |
3e170ce0 | 2603 | task_pid(task), ((thread_t)queue_first(&task->threads))->thread_id, |
39236c6e A |
2604 | task->user_stop_count, mode, task->legacy_stop_count); |
2605 | ||
2606 | #if MACH_ASSERT | |
2607 | /* | |
2608 | * This is obviously not robust; if we suspend one task and then resume a different one, | |
2609 | * we'll fly under the radar. This is only meant to catch the common case of a crashed | |
2610 | * or buggy suspender. | |
2611 | */ | |
2612 | current_task()->suspends_outstanding--; | |
2613 | #endif | |
2614 | ||
2615 | if (mode == TASK_HOLD_LEGACY_ALL) { | |
2616 | if (task->legacy_stop_count >= task->user_stop_count) { | |
2617 | task->user_stop_count = 0; | |
2618 | release = TRUE; | |
2619 | } else { | |
2620 | task->user_stop_count -= task->legacy_stop_count; | |
2621 | } | |
2622 | task->legacy_stop_count = 0; | |
2623 | } else { | |
2624 | if (mode == TASK_HOLD_LEGACY && task->legacy_stop_count > 0) | |
2625 | task->legacy_stop_count--; | |
2626 | if (--task->user_stop_count == 0) | |
2627 | release = TRUE; | |
316670eb A |
2628 | } |
2629 | } | |
2630 | else { | |
2631 | return (KERN_FAILURE); | |
2632 | } | |
2633 | ||
2634 | /* | |
2635 | * Release the task if necessary. | |
2636 | */ | |
2637 | if (release) | |
2638 | task_release_locked(task); | |
2639 | ||
2640 | return (KERN_SUCCESS); | |
2641 | } | |
2642 | ||
39236c6e | 2643 | |
316670eb A |
2644 | /* |
2645 | * task_suspend: | |
2646 | * | |
39236c6e A |
2647 | * Implement an (old-fashioned) user-level suspension on a task. |
2648 | * | |
2649 | * Because the user isn't expecting to have to manage a suspension | |
2650 | * token, we'll track it for him in the kernel in the form of a naked | |
2651 | * send right to the task's resume port. All such send rights | |
2652 | * account for a single suspension against the task (unlike task_suspend2() | |
2653 | * where each caller gets a unique suspension count represented by a | |
2654 | * unique send-once right). | |
316670eb A |
2655 | * |
2656 | * Conditions: | |
2657 | * The caller holds a reference to the task | |
2658 | */ | |
2659 | kern_return_t | |
2660 | task_suspend( | |
39037602 | 2661 | task_t task) |
316670eb | 2662 | { |
39236c6e A |
2663 | kern_return_t kr; |
2664 | mach_port_t port, send, old_notify; | |
2665 | mach_port_name_t name; | |
2666 | ||
316670eb A |
2667 | if (task == TASK_NULL || task == kernel_task) |
2668 | return (KERN_INVALID_ARGUMENT); | |
2669 | ||
2670 | task_lock(task); | |
2671 | ||
39236c6e A |
2672 | /* |
2673 | * Claim a send right on the task resume port, and request a no-senders | |
2674 | * notification on that port (if none outstanding). | |
2675 | */ | |
2676 | if (task->itk_resume == IP_NULL) { | |
2677 | task->itk_resume = ipc_port_alloc_kernel(); | |
2678 | if (!IP_VALID(task->itk_resume)) | |
2679 | panic("failed to create resume port"); | |
2680 | ipc_kobject_set(task->itk_resume, (ipc_kobject_t)task, IKOT_TASK_RESUME); | |
2681 | } | |
2682 | ||
2683 | port = task->itk_resume; | |
2684 | ip_lock(port); | |
2685 | assert(ip_active(port)); | |
2686 | ||
2687 | send = ipc_port_make_send_locked(port); | |
2688 | assert(IP_VALID(send)); | |
2689 | ||
2690 | if (port->ip_nsrequest == IP_NULL) { | |
2691 | ipc_port_nsrequest(port, port->ip_mscount, ipc_port_make_sonce_locked(port), &old_notify); | |
2692 | assert(old_notify == IP_NULL); | |
2693 | /* port unlocked */ | |
2694 | } else { | |
2695 | ip_unlock(port); | |
2696 | } | |
2697 | ||
2698 | /* | |
2699 | * place a legacy hold on the task. | |
2700 | */ | |
2701 | kr = place_task_hold(task, TASK_HOLD_LEGACY); | |
2702 | if (kr != KERN_SUCCESS) { | |
2703 | task_unlock(task); | |
2704 | ipc_port_release_send(send); | |
2705 | return kr; | |
2706 | } | |
91447636 | 2707 | |
1c79356b | 2708 | task_unlock(task); |
91447636 | 2709 | |
39236c6e A |
2710 | /* |
2711 | * Copyout the send right into the calling task's IPC space. It won't know it is there, | |
2712 | * but we'll look it up when calling a traditional resume. Any IPC operations that | |
2713 | * deallocate the send right will auto-release the suspension. | |
2714 | */ | |
2715 | if ((kr = ipc_kmsg_copyout_object(current_task()->itk_space, (ipc_object_t)send, | |
2716 | MACH_MSG_TYPE_MOVE_SEND, &name)) != KERN_SUCCESS) { | |
3e170ce0 A |
2717 | printf("warning: %s(%d) failed to copyout suspension token for pid %d with error: %d\n", |
2718 | proc_name_address(current_task()->bsd_info), proc_pid(current_task()->bsd_info), | |
2719 | task_pid(task), kr); | |
39236c6e A |
2720 | return (kr); |
2721 | } | |
2722 | ||
316670eb | 2723 | return (kr); |
1c79356b A |
2724 | } |
2725 | ||
2726 | /* | |
91447636 | 2727 | * task_resume: |
39236c6e | 2728 | * Release a user hold on a task. |
1c79356b A |
2729 | * |
2730 | * Conditions: | |
2731 | * The caller holds a reference to the task | |
2732 | */ | |
2733 | kern_return_t | |
91447636 | 2734 | task_resume( |
39037602 | 2735 | task_t task) |
1c79356b | 2736 | { |
316670eb | 2737 | kern_return_t kr; |
39236c6e A |
2738 | mach_port_name_t resume_port_name; |
2739 | ipc_entry_t resume_port_entry; | |
2740 | ipc_space_t space = current_task()->itk_space; | |
2741 | ||
2742 | if (task == TASK_NULL || task == kernel_task ) | |
2743 | return (KERN_INVALID_ARGUMENT); | |
2744 | ||
2745 | /* release a legacy task hold */ | |
2746 | task_lock(task); | |
2747 | kr = release_task_hold(task, TASK_HOLD_LEGACY); | |
2748 | task_unlock(task); | |
2749 | ||
2750 | is_write_lock(space); | |
2751 | if (is_active(space) && IP_VALID(task->itk_resume) && | |
2752 | ipc_hash_lookup(space, (ipc_object_t)task->itk_resume, &resume_port_name, &resume_port_entry) == TRUE) { | |
2753 | /* | |
2754 | * We found a suspension token in the caller's IPC space. Release a send right to indicate that | |
2755 | * we are holding one less legacy hold on the task from this caller. If the release failed, | |
2756 | * go ahead and drop all the rights, as someone either already released our holds or the task | |
2757 | * is gone. | |
2758 | */ | |
2759 | if (kr == KERN_SUCCESS) | |
2760 | ipc_right_dealloc(space, resume_port_name, resume_port_entry); | |
2761 | else | |
2762 | ipc_right_destroy(space, resume_port_name, resume_port_entry, FALSE, 0); | |
2763 | /* space unlocked */ | |
2764 | } else { | |
2765 | is_write_unlock(space); | |
2766 | if (kr == KERN_SUCCESS) | |
3e170ce0 | 2767 | printf("warning: %s(%d) performed out-of-band resume on pid %d\n", |
39236c6e | 2768 | proc_name_address(current_task()->bsd_info), proc_pid(current_task()->bsd_info), |
3e170ce0 | 2769 | task_pid(task)); |
39236c6e A |
2770 | } |
2771 | ||
2772 | return kr; | |
2773 | } | |
1c79356b | 2774 | |
39236c6e A |
2775 | /* |
2776 | * Suspend the target task. | |
2777 | * Making/holding a token/reference/port is the callers responsibility. | |
2778 | */ | |
2779 | kern_return_t | |
2780 | task_suspend_internal(task_t task) | |
2781 | { | |
2782 | kern_return_t kr; | |
2783 | ||
91447636 A |
2784 | if (task == TASK_NULL || task == kernel_task) |
2785 | return (KERN_INVALID_ARGUMENT); | |
1c79356b | 2786 | |
1c79356b | 2787 | task_lock(task); |
39236c6e A |
2788 | kr = place_task_hold(task, TASK_HOLD_NORMAL); |
2789 | task_unlock(task); | |
2790 | return (kr); | |
2791 | } | |
2792 | ||
2793 | /* | |
2794 | * Suspend the target task, and return a suspension token. The token | |
2795 | * represents a reference on the suspended task. | |
2796 | */ | |
2797 | kern_return_t | |
2798 | task_suspend2( | |
39037602 | 2799 | task_t task, |
39236c6e A |
2800 | task_suspension_token_t *suspend_token) |
2801 | { | |
2802 | kern_return_t kr; | |
2803 | ||
2804 | kr = task_suspend_internal(task); | |
2805 | if (kr != KERN_SUCCESS) { | |
2806 | *suspend_token = TASK_NULL; | |
2807 | return (kr); | |
2808 | } | |
2809 | ||
2810 | /* | |
2811 | * Take a reference on the target task and return that to the caller | |
2812 | * as a "suspension token," which can be converted into an SO right to | |
2813 | * the now-suspended task's resume port. | |
2814 | */ | |
2815 | task_reference_internal(task); | |
2816 | *suspend_token = task; | |
2817 | ||
2818 | return (KERN_SUCCESS); | |
2819 | } | |
2820 | ||
2821 | /* | |
2822 | * Resume the task | |
2823 | * (reference/token/port management is caller's responsibility). | |
2824 | */ | |
2825 | kern_return_t | |
2826 | task_resume_internal( | |
39037602 | 2827 | task_suspension_token_t task) |
39236c6e A |
2828 | { |
2829 | kern_return_t kr; | |
91447636 | 2830 | |
39236c6e A |
2831 | if (task == TASK_NULL || task == kernel_task) |
2832 | return (KERN_INVALID_ARGUMENT); | |
91447636 | 2833 | |
39236c6e A |
2834 | task_lock(task); |
2835 | kr = release_task_hold(task, TASK_HOLD_NORMAL); | |
316670eb | 2836 | task_unlock(task); |
39236c6e A |
2837 | return (kr); |
2838 | } | |
2839 | ||
2840 | /* | |
2841 | * Resume the task using a suspension token. Consumes the token's ref. | |
2842 | */ | |
2843 | kern_return_t | |
2844 | task_resume2( | |
39037602 | 2845 | task_suspension_token_t task) |
39236c6e A |
2846 | { |
2847 | kern_return_t kr; | |
2848 | ||
2849 | kr = task_resume_internal(task); | |
2850 | task_suspension_token_deallocate(task); | |
91447636 | 2851 | |
316670eb A |
2852 | return (kr); |
2853 | } | |
2854 | ||
39236c6e A |
2855 | boolean_t |
2856 | task_suspension_notify(mach_msg_header_t *request_header) | |
2857 | { | |
2858 | ipc_port_t port = (ipc_port_t) request_header->msgh_remote_port; | |
2859 | task_t task = convert_port_to_task_suspension_token(port); | |
2860 | mach_msg_type_number_t not_count; | |
2861 | ||
2862 | if (task == TASK_NULL || task == kernel_task) | |
2863 | return TRUE; /* nothing to do */ | |
2864 | ||
2865 | switch (request_header->msgh_id) { | |
2866 | ||
2867 | case MACH_NOTIFY_SEND_ONCE: | |
2868 | /* release the hold held by this specific send-once right */ | |
2869 | task_lock(task); | |
2870 | release_task_hold(task, TASK_HOLD_NORMAL); | |
2871 | task_unlock(task); | |
2872 | break; | |
2873 | ||
2874 | case MACH_NOTIFY_NO_SENDERS: | |
2875 | not_count = ((mach_no_senders_notification_t *)request_header)->not_count; | |
2876 | ||
2877 | task_lock(task); | |
2878 | ip_lock(port); | |
2879 | if (port->ip_mscount == not_count) { | |
2880 | ||
2881 | /* release all the [remaining] outstanding legacy holds */ | |
2882 | assert(port->ip_nsrequest == IP_NULL); | |
2883 | ip_unlock(port); | |
2884 | release_task_hold(task, TASK_HOLD_LEGACY_ALL); | |
2885 | task_unlock(task); | |
2886 | ||
2887 | } else if (port->ip_nsrequest == IP_NULL) { | |
2888 | ipc_port_t old_notify; | |
2889 | ||
2890 | task_unlock(task); | |
2891 | /* new send rights, re-arm notification at current make-send count */ | |
2892 | ipc_port_nsrequest(port, port->ip_mscount, ipc_port_make_sonce_locked(port), &old_notify); | |
2893 | assert(old_notify == IP_NULL); | |
2894 | /* port unlocked */ | |
2895 | } else { | |
2896 | ip_unlock(port); | |
2897 | task_unlock(task); | |
2898 | } | |
2899 | break; | |
2900 | ||
2901 | default: | |
2902 | break; | |
2903 | } | |
2904 | ||
2905 | task_suspension_token_deallocate(task); /* drop token reference */ | |
2906 | return TRUE; | |
2907 | } | |
2908 | ||
316670eb A |
2909 | kern_return_t |
2910 | task_pidsuspend_locked(task_t task) | |
2911 | { | |
2912 | kern_return_t kr; | |
2913 | ||
2914 | if (task->pidsuspended) { | |
2915 | kr = KERN_FAILURE; | |
2916 | goto out; | |
1c79356b | 2917 | } |
91447636 | 2918 | |
316670eb A |
2919 | task->pidsuspended = TRUE; |
2920 | ||
39236c6e | 2921 | kr = place_task_hold(task, TASK_HOLD_PIDSUSPEND); |
316670eb A |
2922 | if (kr != KERN_SUCCESS) { |
2923 | task->pidsuspended = FALSE; | |
1c79356b | 2924 | } |
316670eb A |
2925 | out: |
2926 | return(kr); | |
2927 | } | |
1c79356b | 2928 | |
316670eb A |
2929 | |
2930 | /* | |
2931 | * task_pidsuspend: | |
2932 | * | |
2933 | * Suspends a task by placing a hold on its threads. | |
2934 | * | |
2935 | * Conditions: | |
2936 | * The caller holds a reference to the task | |
2937 | */ | |
2938 | kern_return_t | |
2939 | task_pidsuspend( | |
39037602 | 2940 | task_t task) |
316670eb A |
2941 | { |
2942 | kern_return_t kr; | |
2943 | ||
2944 | if (task == TASK_NULL || task == kernel_task) | |
2945 | return (KERN_INVALID_ARGUMENT); | |
2946 | ||
2947 | task_lock(task); | |
2948 | ||
2949 | kr = task_pidsuspend_locked(task); | |
1c79356b A |
2950 | |
2951 | task_unlock(task); | |
91447636 | 2952 | |
316670eb A |
2953 | return (kr); |
2954 | } | |
2955 | ||
316670eb A |
2956 | /* |
2957 | * task_pidresume: | |
2958 | * Resumes a previously suspended task. | |
2959 | * | |
2960 | * Conditions: | |
2961 | * The caller holds a reference to the task | |
2962 | */ | |
2963 | kern_return_t | |
2964 | task_pidresume( | |
39037602 | 2965 | task_t task) |
316670eb A |
2966 | { |
2967 | kern_return_t kr; | |
316670eb A |
2968 | |
2969 | if (task == TASK_NULL || task == kernel_task) | |
2970 | return (KERN_INVALID_ARGUMENT); | |
2971 | ||
2972 | task_lock(task); | |
2973 | ||
39037602 | 2974 | #if CONFIG_FREEZE |
316670eb | 2975 | |
39236c6e | 2976 | while (task->changing_freeze_state) { |
316670eb | 2977 | |
39236c6e A |
2978 | assert_wait((event_t)&task->changing_freeze_state, THREAD_UNINT); |
2979 | task_unlock(task); | |
2980 | thread_block(THREAD_CONTINUE_NULL); | |
316670eb | 2981 | |
39236c6e | 2982 | task_lock(task); |
316670eb | 2983 | } |
39236c6e A |
2984 | task->changing_freeze_state = TRUE; |
2985 | #endif | |
2986 | ||
2987 | kr = release_task_hold(task, TASK_HOLD_PIDSUSPEND); | |
2988 | ||
2989 | task_unlock(task); | |
2990 | ||
39037602 | 2991 | #if CONFIG_FREEZE |
39236c6e | 2992 | |
39236c6e A |
2993 | task_lock(task); |
2994 | ||
2995 | if (kr == KERN_SUCCESS) | |
2996 | task->frozen = FALSE; | |
2997 | task->changing_freeze_state = FALSE; | |
2998 | thread_wakeup(&task->changing_freeze_state); | |
2999 | ||
3000 | task_unlock(task); | |
316670eb A |
3001 | #endif |
3002 | ||
3003 | return (kr); | |
1c79356b A |
3004 | } |
3005 | ||
39037602 A |
3006 | |
3007 | #if DEVELOPMENT || DEBUG | |
3008 | ||
3009 | extern void IOSleep(int); | |
3010 | ||
3011 | kern_return_t | |
3012 | task_disconnect_page_mappings(task_t task) | |
3013 | { | |
3014 | int n; | |
3015 | ||
3016 | if (task == TASK_NULL || task == kernel_task) | |
3017 | return (KERN_INVALID_ARGUMENT); | |
3018 | ||
3019 | /* | |
3020 | * this function is used to strip all of the mappings from | |
3021 | * the pmap for the specified task to force the task to | |
3022 | * re-fault all of the pages it is actively using... this | |
3023 | * allows us to approximate the true working set of the | |
3024 | * specified task. We only engage if at least 1 of the | |
3025 | * threads in the task is runnable, but we want to continuously | |
3026 | * sweep (at least for a while - I've arbitrarily set the limit at | |
3027 | * 100 sweeps to be re-looked at as we gain experience) to get a better | |
3028 | * view into what areas within a page are being visited (as opposed to only | |
3029 | * seeing the first fault of a page after the task becomes | |
3030 | * runnable)... in the future I may | |
3031 | * try to block until awakened by a thread in this task | |
3032 | * being made runnable, but for now we'll periodically poll from the | |
3033 | * user level debug tool driving the sysctl | |
3034 | */ | |
3035 | for (n = 0; n < 100; n++) { | |
3036 | thread_t thread; | |
3037 | boolean_t runnable; | |
3038 | boolean_t do_unnest; | |
3039 | int page_count; | |
3040 | ||
3041 | runnable = FALSE; | |
3042 | do_unnest = FALSE; | |
3043 | ||
3044 | task_lock(task); | |
3045 | ||
3046 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
3047 | ||
3048 | if (thread->state & TH_RUN) { | |
3049 | runnable = TRUE; | |
3050 | break; | |
3051 | } | |
3052 | } | |
3053 | if (n == 0) | |
3054 | task->task_disconnected_count++; | |
3055 | ||
3056 | if (task->task_unnested == FALSE) { | |
3057 | if (runnable == TRUE) { | |
3058 | task->task_unnested = TRUE; | |
3059 | do_unnest = TRUE; | |
3060 | } | |
3061 | } | |
3062 | task_unlock(task); | |
3063 | ||
3064 | if (runnable == FALSE) | |
3065 | break; | |
3066 | ||
3067 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, (MACHDBG_CODE(DBG_MACH_WORKINGSET, VM_DISCONNECT_TASK_PAGE_MAPPINGS)) | DBG_FUNC_START, | |
3068 | task, do_unnest, task->task_disconnected_count, 0, 0); | |
3069 | ||
3070 | page_count = vm_map_disconnect_page_mappings(task->map, do_unnest); | |
3071 | ||
3072 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, (MACHDBG_CODE(DBG_MACH_WORKINGSET, VM_DISCONNECT_TASK_PAGE_MAPPINGS)) | DBG_FUNC_END, | |
3073 | task, page_count, 0, 0, 0); | |
3074 | ||
3075 | if ((n % 5) == 4) | |
3076 | IOSleep(1); | |
3077 | } | |
3078 | return (KERN_SUCCESS); | |
3079 | } | |
3080 | ||
3081 | #endif | |
3082 | ||
3083 | ||
6d2010ae A |
3084 | #if CONFIG_FREEZE |
3085 | ||
3086 | /* | |
3087 | * task_freeze: | |
3088 | * | |
316670eb | 3089 | * Freeze a task. |
6d2010ae A |
3090 | * |
3091 | * Conditions: | |
3092 | * The caller holds a reference to the task | |
3093 | */ | |
3e170ce0 A |
3094 | extern void vm_wake_compactor_swapper(); |
3095 | extern queue_head_t c_swapout_list_head; | |
3096 | ||
6d2010ae A |
3097 | kern_return_t |
3098 | task_freeze( | |
39037602 | 3099 | task_t task, |
6d2010ae A |
3100 | uint32_t *purgeable_count, |
3101 | uint32_t *wired_count, | |
3102 | uint32_t *clean_count, | |
3103 | uint32_t *dirty_count, | |
316670eb | 3104 | uint32_t dirty_budget, |
6d2010ae A |
3105 | boolean_t *shared, |
3106 | boolean_t walk_only) | |
3107 | { | |
39037602 | 3108 | kern_return_t kr = KERN_SUCCESS; |
316670eb | 3109 | |
6d2010ae A |
3110 | if (task == TASK_NULL || task == kernel_task) |
3111 | return (KERN_INVALID_ARGUMENT); | |
3112 | ||
316670eb A |
3113 | task_lock(task); |
3114 | ||
39236c6e A |
3115 | while (task->changing_freeze_state) { |
3116 | ||
3117 | assert_wait((event_t)&task->changing_freeze_state, THREAD_UNINT); | |
3118 | task_unlock(task); | |
3119 | thread_block(THREAD_CONTINUE_NULL); | |
3120 | ||
3121 | task_lock(task); | |
3122 | } | |
316670eb | 3123 | if (task->frozen) { |
39236c6e A |
3124 | task_unlock(task); |
3125 | return (KERN_FAILURE); | |
316670eb | 3126 | } |
39236c6e | 3127 | task->changing_freeze_state = TRUE; |
316670eb A |
3128 | |
3129 | task_unlock(task); | |
3130 | ||
6d2010ae | 3131 | if (walk_only) { |
39037602 | 3132 | panic("task_freeze - walk_only == TRUE"); |
6d2010ae | 3133 | } else { |
316670eb | 3134 | kr = vm_map_freeze(task->map, purgeable_count, wired_count, clean_count, dirty_count, dirty_budget, shared); |
6d2010ae A |
3135 | } |
3136 | ||
39236c6e A |
3137 | task_lock(task); |
3138 | ||
3139 | if (walk_only == FALSE && kr == KERN_SUCCESS) | |
3140 | task->frozen = TRUE; | |
3141 | task->changing_freeze_state = FALSE; | |
3142 | thread_wakeup(&task->changing_freeze_state); | |
3143 | ||
3144 | task_unlock(task); | |
3145 | ||
39037602 | 3146 | if (VM_CONFIG_COMPRESSOR_IS_PRESENT) { |
3e170ce0 A |
3147 | vm_wake_compactor_swapper(); |
3148 | /* | |
3149 | * We do an explicit wakeup of the swapout thread here | |
3150 | * because the compact_and_swap routines don't have | |
3151 | * knowledge about these kind of "per-task packed c_segs" | |
3152 | * and so will not be evaluating whether we need to do | |
3153 | * a wakeup there. | |
3154 | */ | |
3155 | thread_wakeup((event_t)&c_swapout_list_head); | |
3156 | } | |
3157 | ||
316670eb | 3158 | return (kr); |
6d2010ae A |
3159 | } |
3160 | ||
3161 | /* | |
3162 | * task_thaw: | |
3163 | * | |
3164 | * Thaw a currently frozen task. | |
3165 | * | |
3166 | * Conditions: | |
3167 | * The caller holds a reference to the task | |
3168 | */ | |
3169 | kern_return_t | |
3170 | task_thaw( | |
39037602 | 3171 | task_t task) |
6d2010ae A |
3172 | { |
3173 | if (task == TASK_NULL || task == kernel_task) | |
3174 | return (KERN_INVALID_ARGUMENT); | |
3175 | ||
316670eb A |
3176 | task_lock(task); |
3177 | ||
39236c6e A |
3178 | while (task->changing_freeze_state) { |
3179 | ||
3180 | assert_wait((event_t)&task->changing_freeze_state, THREAD_UNINT); | |
3181 | task_unlock(task); | |
3182 | thread_block(THREAD_CONTINUE_NULL); | |
3183 | ||
3184 | task_lock(task); | |
3185 | } | |
316670eb | 3186 | if (!task->frozen) { |
39236c6e A |
3187 | task_unlock(task); |
3188 | return (KERN_FAILURE); | |
316670eb | 3189 | } |
39037602 | 3190 | task->frozen = FALSE; |
39236c6e | 3191 | |
316670eb A |
3192 | task_unlock(task); |
3193 | ||
39037602 | 3194 | return (KERN_SUCCESS); |
6d2010ae A |
3195 | } |
3196 | ||
3197 | #endif /* CONFIG_FREEZE */ | |
3198 | ||
1c79356b A |
3199 | kern_return_t |
3200 | host_security_set_task_token( | |
3201 | host_security_t host_security, | |
3202 | task_t task, | |
3203 | security_token_t sec_token, | |
55e303ae | 3204 | audit_token_t audit_token, |
1c79356b A |
3205 | host_priv_t host_priv) |
3206 | { | |
55e303ae | 3207 | ipc_port_t host_port; |
1c79356b A |
3208 | kern_return_t kr; |
3209 | ||
3210 | if (task == TASK_NULL) | |
3211 | return(KERN_INVALID_ARGUMENT); | |
3212 | ||
3213 | if (host_security == HOST_NULL) | |
3214 | return(KERN_INVALID_SECURITY); | |
3215 | ||
3216 | task_lock(task); | |
3217 | task->sec_token = sec_token; | |
55e303ae | 3218 | task->audit_token = audit_token; |
39236c6e A |
3219 | |
3220 | task_unlock(task); | |
1c79356b A |
3221 | |
3222 | if (host_priv != HOST_PRIV_NULL) { | |
55e303ae | 3223 | kr = host_get_host_priv_port(host_priv, &host_port); |
1c79356b | 3224 | } else { |
55e303ae | 3225 | kr = host_get_host_port(host_priv_self(), &host_port); |
1c79356b | 3226 | } |
55e303ae A |
3227 | assert(kr == KERN_SUCCESS); |
3228 | kr = task_set_special_port(task, TASK_HOST_PORT, host_port); | |
1c79356b A |
3229 | return(kr); |
3230 | } | |
3231 | ||
fe8ab488 A |
3232 | kern_return_t |
3233 | task_send_trace_memory( | |
3234 | task_t target_task, | |
3235 | __unused uint32_t pid, | |
3236 | __unused uint64_t uniqueid) | |
3237 | { | |
3238 | kern_return_t kr = KERN_INVALID_ARGUMENT; | |
3239 | if (target_task == TASK_NULL) | |
3240 | return (KERN_INVALID_ARGUMENT); | |
3241 | ||
3242 | #if CONFIG_ATM | |
3243 | kr = atm_send_proc_inspect_notification(target_task, | |
3244 | pid, | |
3245 | uniqueid); | |
3246 | ||
3247 | #endif | |
3248 | return (kr); | |
3249 | } | |
1c79356b A |
3250 | /* |
3251 | * This routine was added, pretty much exclusively, for registering the | |
3252 | * RPC glue vector for in-kernel short circuited tasks. Rather than | |
3253 | * removing it completely, I have only disabled that feature (which was | |
3254 | * the only feature at the time). It just appears that we are going to | |
3255 | * want to add some user data to tasks in the future (i.e. bsd info, | |
3256 | * task names, etc...), so I left it in the formal task interface. | |
3257 | */ | |
3258 | kern_return_t | |
3259 | task_set_info( | |
3260 | task_t task, | |
3261 | task_flavor_t flavor, | |
91447636 A |
3262 | __unused task_info_t task_info_in, /* pointer to IN array */ |
3263 | __unused mach_msg_type_number_t task_info_count) | |
1c79356b | 3264 | { |
1c79356b A |
3265 | if (task == TASK_NULL) |
3266 | return(KERN_INVALID_ARGUMENT); | |
3267 | ||
3268 | switch (flavor) { | |
fe8ab488 A |
3269 | |
3270 | #if CONFIG_ATM | |
3271 | case TASK_TRACE_MEMORY_INFO: | |
3272 | { | |
3273 | if (task_info_count != TASK_TRACE_MEMORY_INFO_COUNT) | |
3274 | return (KERN_INVALID_ARGUMENT); | |
3275 | ||
3276 | assert(task_info_in != NULL); | |
3277 | task_trace_memory_info_t mem_info; | |
3278 | mem_info = (task_trace_memory_info_t) task_info_in; | |
3279 | kern_return_t kr = atm_register_trace_memory(task, | |
3280 | mem_info->user_memory_address, | |
3e170ce0 | 3281 | mem_info->buffer_size); |
fe8ab488 | 3282 | return kr; |
fe8ab488 A |
3283 | } |
3284 | ||
3285 | #endif | |
1c79356b A |
3286 | default: |
3287 | return (KERN_INVALID_ARGUMENT); | |
3288 | } | |
3289 | return (KERN_SUCCESS); | |
3290 | } | |
3291 | ||
3e170ce0 | 3292 | int radar_20146450 = 1; |
1c79356b A |
3293 | kern_return_t |
3294 | task_info( | |
39236c6e A |
3295 | task_t task, |
3296 | task_flavor_t flavor, | |
3297 | task_info_t task_info_out, | |
1c79356b A |
3298 | mach_msg_type_number_t *task_info_count) |
3299 | { | |
b0d623f7 | 3300 | kern_return_t error = KERN_SUCCESS; |
39037602 | 3301 | mach_msg_type_number_t original_task_info_count; |
b0d623f7 | 3302 | |
1c79356b | 3303 | if (task == TASK_NULL) |
91447636 | 3304 | return (KERN_INVALID_ARGUMENT); |
1c79356b | 3305 | |
39037602 | 3306 | original_task_info_count = *task_info_count; |
b0d623f7 A |
3307 | task_lock(task); |
3308 | ||
3309 | if ((task != current_task()) && (!task->active)) { | |
3310 | task_unlock(task); | |
3311 | return (KERN_INVALID_ARGUMENT); | |
3312 | } | |
3313 | ||
1c79356b A |
3314 | switch (flavor) { |
3315 | ||
91447636 | 3316 | case TASK_BASIC_INFO_32: |
2d21ac55 | 3317 | case TASK_BASIC2_INFO_32: |
91447636 A |
3318 | { |
3319 | task_basic_info_32_t basic_info; | |
b0d623f7 A |
3320 | vm_map_t map; |
3321 | clock_sec_t secs; | |
3322 | clock_usec_t usecs; | |
1c79356b | 3323 | |
b0d623f7 A |
3324 | if (*task_info_count < TASK_BASIC_INFO_32_COUNT) { |
3325 | error = KERN_INVALID_ARGUMENT; | |
3326 | break; | |
3327 | } | |
1c79356b | 3328 | |
91447636 | 3329 | basic_info = (task_basic_info_32_t)task_info_out; |
1c79356b | 3330 | |
91447636 | 3331 | map = (task == kernel_task)? kernel_map: task->map; |
b0d623f7 | 3332 | basic_info->virtual_size = (typeof(basic_info->virtual_size))map->size; |
2d21ac55 A |
3333 | if (flavor == TASK_BASIC2_INFO_32) { |
3334 | /* | |
3335 | * The "BASIC2" flavor gets the maximum resident | |
3336 | * size instead of the current resident size... | |
3337 | */ | |
3338 | basic_info->resident_size = pmap_resident_max(map->pmap); | |
3339 | } else { | |
3340 | basic_info->resident_size = pmap_resident_count(map->pmap); | |
3341 | } | |
3342 | basic_info->resident_size *= PAGE_SIZE; | |
1c79356b | 3343 | |
0b4e3aa0 A |
3344 | basic_info->policy = ((task != kernel_task)? |
3345 | POLICY_TIMESHARE: POLICY_RR); | |
1c79356b | 3346 | basic_info->suspend_count = task->user_stop_count; |
91447636 | 3347 | |
b0d623f7 A |
3348 | absolutetime_to_microtime(task->total_user_time, &secs, &usecs); |
3349 | basic_info->user_time.seconds = | |
3350 | (typeof(basic_info->user_time.seconds))secs; | |
3351 | basic_info->user_time.microseconds = usecs; | |
3352 | ||
3353 | absolutetime_to_microtime(task->total_system_time, &secs, &usecs); | |
3354 | basic_info->system_time.seconds = | |
3355 | (typeof(basic_info->system_time.seconds))secs; | |
3356 | basic_info->system_time.microseconds = usecs; | |
1c79356b | 3357 | |
91447636 | 3358 | *task_info_count = TASK_BASIC_INFO_32_COUNT; |
1c79356b | 3359 | break; |
91447636 | 3360 | } |
1c79356b | 3361 | |
91447636 A |
3362 | case TASK_BASIC_INFO_64: |
3363 | { | |
3364 | task_basic_info_64_t basic_info; | |
b0d623f7 A |
3365 | vm_map_t map; |
3366 | clock_sec_t secs; | |
3367 | clock_usec_t usecs; | |
1c79356b | 3368 | |
b0d623f7 A |
3369 | if (*task_info_count < TASK_BASIC_INFO_64_COUNT) { |
3370 | error = KERN_INVALID_ARGUMENT; | |
3371 | break; | |
3372 | } | |
91447636 A |
3373 | |
3374 | basic_info = (task_basic_info_64_t)task_info_out; | |
3375 | ||
3376 | map = (task == kernel_task)? kernel_map: task->map; | |
3377 | basic_info->virtual_size = map->size; | |
2d21ac55 A |
3378 | basic_info->resident_size = |
3379 | (mach_vm_size_t)(pmap_resident_count(map->pmap)) | |
3380 | * PAGE_SIZE_64; | |
91447636 | 3381 | |
91447636 A |
3382 | basic_info->policy = ((task != kernel_task)? |
3383 | POLICY_TIMESHARE: POLICY_RR); | |
3384 | basic_info->suspend_count = task->user_stop_count; | |
3385 | ||
b0d623f7 A |
3386 | absolutetime_to_microtime(task->total_user_time, &secs, &usecs); |
3387 | basic_info->user_time.seconds = | |
3388 | (typeof(basic_info->user_time.seconds))secs; | |
3389 | basic_info->user_time.microseconds = usecs; | |
3390 | ||
3391 | absolutetime_to_microtime(task->total_system_time, &secs, &usecs); | |
3392 | basic_info->system_time.seconds = | |
3393 | (typeof(basic_info->system_time.seconds))secs; | |
3394 | basic_info->system_time.microseconds = usecs; | |
91447636 A |
3395 | |
3396 | *task_info_count = TASK_BASIC_INFO_64_COUNT; | |
3397 | break; | |
3398 | } | |
3399 | ||
316670eb A |
3400 | case MACH_TASK_BASIC_INFO: |
3401 | { | |
3402 | mach_task_basic_info_t basic_info; | |
3403 | vm_map_t map; | |
3404 | clock_sec_t secs; | |
3405 | clock_usec_t usecs; | |
3406 | ||
3407 | if (*task_info_count < MACH_TASK_BASIC_INFO_COUNT) { | |
3408 | error = KERN_INVALID_ARGUMENT; | |
3409 | break; | |
3410 | } | |
3411 | ||
3412 | basic_info = (mach_task_basic_info_t)task_info_out; | |
3413 | ||
3414 | map = (task == kernel_task) ? kernel_map : task->map; | |
3415 | ||
3416 | basic_info->virtual_size = map->size; | |
3417 | ||
3418 | basic_info->resident_size = | |
3419 | (mach_vm_size_t)(pmap_resident_count(map->pmap)); | |
3420 | basic_info->resident_size *= PAGE_SIZE_64; | |
3421 | ||
3422 | basic_info->resident_size_max = | |
3423 | (mach_vm_size_t)(pmap_resident_max(map->pmap)); | |
3424 | basic_info->resident_size_max *= PAGE_SIZE_64; | |
3425 | ||
3426 | basic_info->policy = ((task != kernel_task) ? | |
3427 | POLICY_TIMESHARE : POLICY_RR); | |
3428 | ||
3429 | basic_info->suspend_count = task->user_stop_count; | |
3430 | ||
3431 | absolutetime_to_microtime(task->total_user_time, &secs, &usecs); | |
3432 | basic_info->user_time.seconds = | |
3433 | (typeof(basic_info->user_time.seconds))secs; | |
3434 | basic_info->user_time.microseconds = usecs; | |
3435 | ||
3436 | absolutetime_to_microtime(task->total_system_time, &secs, &usecs); | |
3437 | basic_info->system_time.seconds = | |
3438 | (typeof(basic_info->system_time.seconds))secs; | |
3439 | basic_info->system_time.microseconds = usecs; | |
3440 | ||
3441 | *task_info_count = MACH_TASK_BASIC_INFO_COUNT; | |
3442 | break; | |
3443 | } | |
3444 | ||
91447636 A |
3445 | case TASK_THREAD_TIMES_INFO: |
3446 | { | |
39037602 A |
3447 | task_thread_times_info_t times_info; |
3448 | thread_t thread; | |
91447636 | 3449 | |
b0d623f7 A |
3450 | if (*task_info_count < TASK_THREAD_TIMES_INFO_COUNT) { |
3451 | error = KERN_INVALID_ARGUMENT; | |
3452 | break; | |
3453 | } | |
1c79356b A |
3454 | |
3455 | times_info = (task_thread_times_info_t) task_info_out; | |
3456 | times_info->user_time.seconds = 0; | |
3457 | times_info->user_time.microseconds = 0; | |
3458 | times_info->system_time.seconds = 0; | |
3459 | times_info->system_time.microseconds = 0; | |
3460 | ||
1c79356b | 3461 | |
91447636 | 3462 | queue_iterate(&task->threads, thread, thread_t, task_threads) { |
39236c6e | 3463 | time_value_t user_time, system_time; |
1c79356b | 3464 | |
39236c6e A |
3465 | if (thread->options & TH_OPT_IDLE_THREAD) |
3466 | continue; | |
1c79356b | 3467 | |
39236c6e | 3468 | thread_read_times(thread, &user_time, &system_time); |
91447636 | 3469 | |
39236c6e A |
3470 | time_value_add(×_info->user_time, &user_time); |
3471 | time_value_add(×_info->system_time, &system_time); | |
3472 | } | |
1c79356b A |
3473 | |
3474 | *task_info_count = TASK_THREAD_TIMES_INFO_COUNT; | |
3475 | break; | |
91447636 A |
3476 | } |
3477 | ||
3478 | case TASK_ABSOLUTETIME_INFO: | |
3479 | { | |
3480 | task_absolutetime_info_t info; | |
39037602 | 3481 | thread_t thread; |
91447636 | 3482 | |
b0d623f7 A |
3483 | if (*task_info_count < TASK_ABSOLUTETIME_INFO_COUNT) { |
3484 | error = KERN_INVALID_ARGUMENT; | |
3485 | break; | |
3486 | } | |
91447636 A |
3487 | |
3488 | info = (task_absolutetime_info_t)task_info_out; | |
3489 | info->threads_user = info->threads_system = 0; | |
3490 | ||
91447636 A |
3491 | |
3492 | info->total_user = task->total_user_time; | |
3493 | info->total_system = task->total_system_time; | |
3494 | ||
3495 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
3496 | uint64_t tval; | |
316670eb A |
3497 | spl_t x; |
3498 | ||
39236c6e A |
3499 | if (thread->options & TH_OPT_IDLE_THREAD) |
3500 | continue; | |
3501 | ||
316670eb A |
3502 | x = splsched(); |
3503 | thread_lock(thread); | |
91447636 A |
3504 | |
3505 | tval = timer_grab(&thread->user_timer); | |
3506 | info->threads_user += tval; | |
3507 | info->total_user += tval; | |
3508 | ||
3509 | tval = timer_grab(&thread->system_timer); | |
316670eb A |
3510 | if (thread->precise_user_kernel_time) { |
3511 | info->threads_system += tval; | |
3512 | info->total_system += tval; | |
3513 | } else { | |
3514 | /* system_timer may represent either sys or user */ | |
3515 | info->threads_user += tval; | |
3516 | info->total_user += tval; | |
3517 | } | |
3518 | ||
3519 | thread_unlock(thread); | |
3520 | splx(x); | |
91447636 A |
3521 | } |
3522 | ||
91447636 A |
3523 | |
3524 | *task_info_count = TASK_ABSOLUTETIME_INFO_COUNT; | |
3525 | break; | |
3526 | } | |
1c79356b | 3527 | |
b0d623f7 A |
3528 | case TASK_DYLD_INFO: |
3529 | { | |
3530 | task_dyld_info_t info; | |
3531 | ||
6d2010ae A |
3532 | /* |
3533 | * We added the format field to TASK_DYLD_INFO output. For | |
3534 | * temporary backward compatibility, accept the fact that | |
3535 | * clients may ask for the old version - distinquished by the | |
3536 | * size of the expected result structure. | |
3537 | */ | |
3538 | #define TASK_LEGACY_DYLD_INFO_COUNT \ | |
3539 | offsetof(struct task_dyld_info, all_image_info_format)/sizeof(natural_t) | |
3540 | ||
3541 | if (*task_info_count < TASK_LEGACY_DYLD_INFO_COUNT) { | |
b0d623f7 A |
3542 | error = KERN_INVALID_ARGUMENT; |
3543 | break; | |
3544 | } | |
6d2010ae | 3545 | |
b0d623f7 A |
3546 | info = (task_dyld_info_t)task_info_out; |
3547 | info->all_image_info_addr = task->all_image_info_addr; | |
3548 | info->all_image_info_size = task->all_image_info_size; | |
6d2010ae A |
3549 | |
3550 | /* only set format on output for those expecting it */ | |
3551 | if (*task_info_count >= TASK_DYLD_INFO_COUNT) { | |
3552 | info->all_image_info_format = task_has_64BitAddr(task) ? | |
3553 | TASK_DYLD_ALL_IMAGE_INFO_64 : | |
3554 | TASK_DYLD_ALL_IMAGE_INFO_32 ; | |
3555 | *task_info_count = TASK_DYLD_INFO_COUNT; | |
3556 | } else { | |
3557 | *task_info_count = TASK_LEGACY_DYLD_INFO_COUNT; | |
3558 | } | |
b0d623f7 A |
3559 | break; |
3560 | } | |
3561 | ||
6d2010ae A |
3562 | case TASK_EXTMOD_INFO: |
3563 | { | |
3564 | task_extmod_info_t info; | |
3565 | void *p; | |
3566 | ||
3567 | if (*task_info_count < TASK_EXTMOD_INFO_COUNT) { | |
3568 | error = KERN_INVALID_ARGUMENT; | |
3569 | break; | |
3570 | } | |
3571 | ||
3572 | info = (task_extmod_info_t)task_info_out; | |
3573 | ||
3574 | p = get_bsdtask_info(task); | |
3575 | if (p) { | |
3576 | proc_getexecutableuuid(p, info->task_uuid, sizeof(info->task_uuid)); | |
3577 | } else { | |
3578 | bzero(info->task_uuid, sizeof(info->task_uuid)); | |
3579 | } | |
3580 | info->extmod_statistics = task->extmod_statistics; | |
3581 | *task_info_count = TASK_EXTMOD_INFO_COUNT; | |
3582 | ||
3583 | break; | |
3584 | } | |
3585 | ||
3586 | case TASK_KERNELMEMORY_INFO: | |
3587 | { | |
3588 | task_kernelmemory_info_t tkm_info; | |
316670eb | 3589 | ledger_amount_t credit, debit; |
6d2010ae A |
3590 | |
3591 | if (*task_info_count < TASK_KERNELMEMORY_INFO_COUNT) { | |
3592 | error = KERN_INVALID_ARGUMENT; | |
3593 | break; | |
3594 | } | |
3595 | ||
3596 | tkm_info = (task_kernelmemory_info_t) task_info_out; | |
316670eb A |
3597 | tkm_info->total_palloc = 0; |
3598 | tkm_info->total_pfree = 0; | |
3599 | tkm_info->total_salloc = 0; | |
3600 | tkm_info->total_sfree = 0; | |
6d2010ae A |
3601 | |
3602 | if (task == kernel_task) { | |
3603 | /* | |
3604 | * All shared allocs/frees from other tasks count against | |
3605 | * the kernel private memory usage. If we are looking up | |
3606 | * info for the kernel task, gather from everywhere. | |
3607 | */ | |
3608 | task_unlock(task); | |
3609 | ||
3610 | /* start by accounting for all the terminated tasks against the kernel */ | |
3611 | tkm_info->total_palloc = tasks_tkm_private.alloc + tasks_tkm_shared.alloc; | |
3612 | tkm_info->total_pfree = tasks_tkm_private.free + tasks_tkm_shared.free; | |
6d2010ae A |
3613 | |
3614 | /* count all other task/thread shared alloc/free against the kernel */ | |
3615 | lck_mtx_lock(&tasks_threads_lock); | |
316670eb A |
3616 | |
3617 | /* XXX this really shouldn't be using the function parameter 'task' as a local var! */ | |
6d2010ae A |
3618 | queue_iterate(&tasks, task, task_t, tasks) { |
3619 | if (task == kernel_task) { | |
316670eb A |
3620 | if (ledger_get_entries(task->ledger, |
3621 | task_ledgers.tkm_private, &credit, | |
3622 | &debit) == KERN_SUCCESS) { | |
3623 | tkm_info->total_palloc += credit; | |
3624 | tkm_info->total_pfree += debit; | |
3625 | } | |
6d2010ae | 3626 | } |
316670eb A |
3627 | if (!ledger_get_entries(task->ledger, |
3628 | task_ledgers.tkm_shared, &credit, &debit)) { | |
3629 | tkm_info->total_palloc += credit; | |
3630 | tkm_info->total_pfree += debit; | |
6d2010ae | 3631 | } |
6d2010ae A |
3632 | } |
3633 | lck_mtx_unlock(&tasks_threads_lock); | |
3634 | } else { | |
316670eb A |
3635 | if (!ledger_get_entries(task->ledger, |
3636 | task_ledgers.tkm_private, &credit, &debit)) { | |
3637 | tkm_info->total_palloc = credit; | |
3638 | tkm_info->total_pfree = debit; | |
3639 | } | |
3640 | if (!ledger_get_entries(task->ledger, | |
3641 | task_ledgers.tkm_shared, &credit, &debit)) { | |
3642 | tkm_info->total_salloc = credit; | |
3643 | tkm_info->total_sfree = debit; | |
6d2010ae A |
3644 | } |
3645 | task_unlock(task); | |
3646 | } | |
3647 | ||
3648 | *task_info_count = TASK_KERNELMEMORY_INFO_COUNT; | |
3649 | return KERN_SUCCESS; | |
3650 | } | |
3651 | ||
91447636 A |
3652 | /* OBSOLETE */ |
3653 | case TASK_SCHED_FIFO_INFO: | |
3654 | { | |
1c79356b | 3655 | |
b0d623f7 A |
3656 | if (*task_info_count < POLICY_FIFO_BASE_COUNT) { |
3657 | error = KERN_INVALID_ARGUMENT; | |
3658 | break; | |
3659 | } | |
1c79356b | 3660 | |
b0d623f7 | 3661 | error = KERN_INVALID_POLICY; |
6d2010ae | 3662 | break; |
91447636 | 3663 | } |
1c79356b | 3664 | |
91447636 A |
3665 | /* OBSOLETE */ |
3666 | case TASK_SCHED_RR_INFO: | |
3667 | { | |
39037602 | 3668 | policy_rr_base_t rr_base; |
6d2010ae A |
3669 | uint32_t quantum_time; |
3670 | uint64_t quantum_ns; | |
1c79356b | 3671 | |
b0d623f7 A |
3672 | if (*task_info_count < POLICY_RR_BASE_COUNT) { |
3673 | error = KERN_INVALID_ARGUMENT; | |
3674 | break; | |
3675 | } | |
1c79356b A |
3676 | |
3677 | rr_base = (policy_rr_base_t) task_info_out; | |
3678 | ||
0b4e3aa0 | 3679 | if (task != kernel_task) { |
b0d623f7 A |
3680 | error = KERN_INVALID_POLICY; |
3681 | break; | |
1c79356b A |
3682 | } |
3683 | ||
3684 | rr_base->base_priority = task->priority; | |
1c79356b | 3685 | |
6d2010ae A |
3686 | quantum_time = SCHED(initial_quantum_size)(THREAD_NULL); |
3687 | absolutetime_to_nanoseconds(quantum_time, &quantum_ns); | |
3688 | ||
3689 | rr_base->quantum = (uint32_t)(quantum_ns / 1000 / 1000); | |
1c79356b A |
3690 | |
3691 | *task_info_count = POLICY_RR_BASE_COUNT; | |
3692 | break; | |
91447636 | 3693 | } |
1c79356b | 3694 | |
91447636 A |
3695 | /* OBSOLETE */ |
3696 | case TASK_SCHED_TIMESHARE_INFO: | |
3697 | { | |
39037602 | 3698 | policy_timeshare_base_t ts_base; |
1c79356b | 3699 | |
b0d623f7 A |
3700 | if (*task_info_count < POLICY_TIMESHARE_BASE_COUNT) { |
3701 | error = KERN_INVALID_ARGUMENT; | |
3702 | break; | |
3703 | } | |
1c79356b A |
3704 | |
3705 | ts_base = (policy_timeshare_base_t) task_info_out; | |
3706 | ||
0b4e3aa0 | 3707 | if (task == kernel_task) { |
b0d623f7 A |
3708 | error = KERN_INVALID_POLICY; |
3709 | break; | |
1c79356b A |
3710 | } |
3711 | ||
3712 | ts_base->base_priority = task->priority; | |
1c79356b A |
3713 | |
3714 | *task_info_count = POLICY_TIMESHARE_BASE_COUNT; | |
3715 | break; | |
91447636 | 3716 | } |
1c79356b | 3717 | |
91447636 A |
3718 | case TASK_SECURITY_TOKEN: |
3719 | { | |
39037602 | 3720 | security_token_t *sec_token_p; |
1c79356b | 3721 | |
b0d623f7 A |
3722 | if (*task_info_count < TASK_SECURITY_TOKEN_COUNT) { |
3723 | error = KERN_INVALID_ARGUMENT; | |
3724 | break; | |
3725 | } | |
1c79356b A |
3726 | |
3727 | sec_token_p = (security_token_t *) task_info_out; | |
3728 | ||
1c79356b | 3729 | *sec_token_p = task->sec_token; |
1c79356b A |
3730 | |
3731 | *task_info_count = TASK_SECURITY_TOKEN_COUNT; | |
91447636 A |
3732 | break; |
3733 | } | |
1c79356b | 3734 | |
91447636 A |
3735 | case TASK_AUDIT_TOKEN: |
3736 | { | |
39037602 | 3737 | audit_token_t *audit_token_p; |
55e303ae | 3738 | |
b0d623f7 A |
3739 | if (*task_info_count < TASK_AUDIT_TOKEN_COUNT) { |
3740 | error = KERN_INVALID_ARGUMENT; | |
3741 | break; | |
3742 | } | |
55e303ae A |
3743 | |
3744 | audit_token_p = (audit_token_t *) task_info_out; | |
3745 | ||
55e303ae | 3746 | *audit_token_p = task->audit_token; |
55e303ae A |
3747 | |
3748 | *task_info_count = TASK_AUDIT_TOKEN_COUNT; | |
91447636 A |
3749 | break; |
3750 | } | |
55e303ae | 3751 | |
91447636 | 3752 | case TASK_SCHED_INFO: |
b0d623f7 | 3753 | error = KERN_INVALID_ARGUMENT; |
6d2010ae | 3754 | break; |
1c79356b | 3755 | |
91447636 A |
3756 | case TASK_EVENTS_INFO: |
3757 | { | |
39037602 A |
3758 | task_events_info_t events_info; |
3759 | thread_t thread; | |
1c79356b | 3760 | |
b0d623f7 A |
3761 | if (*task_info_count < TASK_EVENTS_INFO_COUNT) { |
3762 | error = KERN_INVALID_ARGUMENT; | |
3763 | break; | |
3764 | } | |
1c79356b A |
3765 | |
3766 | events_info = (task_events_info_t) task_info_out; | |
3767 | ||
2d21ac55 | 3768 | |
1c79356b A |
3769 | events_info->faults = task->faults; |
3770 | events_info->pageins = task->pageins; | |
3771 | events_info->cow_faults = task->cow_faults; | |
3772 | events_info->messages_sent = task->messages_sent; | |
3773 | events_info->messages_received = task->messages_received; | |
3774 | events_info->syscalls_mach = task->syscalls_mach; | |
3775 | events_info->syscalls_unix = task->syscalls_unix; | |
2d21ac55 A |
3776 | |
3777 | events_info->csw = task->c_switch; | |
3778 | ||
3779 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
6d2010ae A |
3780 | events_info->csw += thread->c_switch; |
3781 | events_info->syscalls_mach += thread->syscalls_mach; | |
3782 | events_info->syscalls_unix += thread->syscalls_unix; | |
2d21ac55 A |
3783 | } |
3784 | ||
1c79356b A |
3785 | |
3786 | *task_info_count = TASK_EVENTS_INFO_COUNT; | |
3787 | break; | |
91447636 | 3788 | } |
2d21ac55 A |
3789 | case TASK_AFFINITY_TAG_INFO: |
3790 | { | |
b0d623f7 A |
3791 | if (*task_info_count < TASK_AFFINITY_TAG_INFO_COUNT) { |
3792 | error = KERN_INVALID_ARGUMENT; | |
3793 | break; | |
3794 | } | |
2d21ac55 | 3795 | |
b0d623f7 | 3796 | error = task_affinity_info(task, task_info_out, task_info_count); |
6d2010ae | 3797 | break; |
2d21ac55 | 3798 | } |
4b17d6b6 A |
3799 | case TASK_POWER_INFO: |
3800 | { | |
4b17d6b6 A |
3801 | if (*task_info_count < TASK_POWER_INFO_COUNT) { |
3802 | error = KERN_INVALID_ARGUMENT; | |
3803 | break; | |
3804 | } | |
3805 | ||
39037602 | 3806 | task_power_info_locked(task, (task_power_info_t)task_info_out, NULL, NULL); |
fe8ab488 A |
3807 | break; |
3808 | } | |
3809 | ||
3810 | case TASK_POWER_INFO_V2: | |
3811 | { | |
3812 | if (*task_info_count < TASK_POWER_INFO_V2_COUNT) { | |
3813 | error = KERN_INVALID_ARGUMENT; | |
3814 | break; | |
3815 | } | |
3816 | task_power_info_v2_t tpiv2 = (task_power_info_v2_t) task_info_out; | |
39037602 A |
3817 | |
3818 | uint64_t *task_energy = NULL; | |
3819 | task_power_info_locked(task, &tpiv2->cpu_energy, &tpiv2->gpu_energy, task_energy); | |
39236c6e A |
3820 | break; |
3821 | } | |
4b17d6b6 | 3822 | |
39236c6e A |
3823 | case TASK_VM_INFO: |
3824 | case TASK_VM_INFO_PURGEABLE: | |
3825 | { | |
3826 | task_vm_info_t vm_info; | |
3827 | vm_map_t map; | |
4b17d6b6 | 3828 | |
3e170ce0 | 3829 | if (*task_info_count < TASK_VM_INFO_REV0_COUNT) { |
39236c6e A |
3830 | error = KERN_INVALID_ARGUMENT; |
3831 | break; | |
3832 | } | |
4b17d6b6 | 3833 | |
39236c6e | 3834 | vm_info = (task_vm_info_t)task_info_out; |
4b17d6b6 | 3835 | |
39236c6e A |
3836 | if (task == kernel_task) { |
3837 | map = kernel_map; | |
3838 | /* no lock */ | |
3839 | } else { | |
3840 | map = task->map; | |
3841 | vm_map_lock_read(map); | |
3842 | } | |
4b17d6b6 | 3843 | |
39236c6e A |
3844 | vm_info->virtual_size = (typeof(vm_info->virtual_size))map->size; |
3845 | vm_info->region_count = map->hdr.nentries; | |
3846 | vm_info->page_size = vm_map_page_size(map); | |
3847 | ||
3848 | vm_info->resident_size = pmap_resident_count(map->pmap); | |
3849 | vm_info->resident_size *= PAGE_SIZE; | |
3850 | vm_info->resident_size_peak = pmap_resident_max(map->pmap); | |
3851 | vm_info->resident_size_peak *= PAGE_SIZE; | |
3852 | ||
3853 | #define _VM_INFO(_name) \ | |
3854 | vm_info->_name = ((mach_vm_size_t) map->pmap->stats._name) * PAGE_SIZE | |
3855 | ||
3856 | _VM_INFO(device); | |
3857 | _VM_INFO(device_peak); | |
3858 | _VM_INFO(external); | |
3859 | _VM_INFO(external_peak); | |
3860 | _VM_INFO(internal); | |
3861 | _VM_INFO(internal_peak); | |
3862 | _VM_INFO(reusable); | |
3863 | _VM_INFO(reusable_peak); | |
3864 | _VM_INFO(compressed); | |
3865 | _VM_INFO(compressed_peak); | |
3866 | _VM_INFO(compressed_lifetime); | |
3867 | ||
3868 | vm_info->purgeable_volatile_pmap = 0; | |
3869 | vm_info->purgeable_volatile_resident = 0; | |
3870 | vm_info->purgeable_volatile_virtual = 0; | |
3871 | if (task == kernel_task) { | |
3872 | /* | |
3873 | * We do not maintain the detailed stats for the | |
3874 | * kernel_pmap, so just count everything as | |
3875 | * "internal"... | |
3876 | */ | |
3877 | vm_info->internal = vm_info->resident_size; | |
3878 | /* | |
3879 | * ... but since the memory held by the VM compressor | |
3880 | * in the kernel address space ought to be attributed | |
3881 | * to user-space tasks, we subtract it from "internal" | |
3882 | * to give memory reporting tools a more accurate idea | |
3883 | * of what the kernel itself is actually using, instead | |
3884 | * of making it look like the kernel is leaking memory | |
3885 | * when the system is under memory pressure. | |
3886 | */ | |
3887 | vm_info->internal -= (VM_PAGE_COMPRESSOR_COUNT * | |
3888 | PAGE_SIZE); | |
3889 | } else { | |
3890 | mach_vm_size_t volatile_virtual_size; | |
3891 | mach_vm_size_t volatile_resident_size; | |
3e170ce0 | 3892 | mach_vm_size_t volatile_compressed_size; |
39236c6e | 3893 | mach_vm_size_t volatile_pmap_size; |
3e170ce0 | 3894 | mach_vm_size_t volatile_compressed_pmap_size; |
39236c6e A |
3895 | kern_return_t kr; |
3896 | ||
3897 | if (flavor == TASK_VM_INFO_PURGEABLE) { | |
3898 | kr = vm_map_query_volatile( | |
3899 | map, | |
3900 | &volatile_virtual_size, | |
3901 | &volatile_resident_size, | |
3e170ce0 A |
3902 | &volatile_compressed_size, |
3903 | &volatile_pmap_size, | |
3904 | &volatile_compressed_pmap_size); | |
39236c6e A |
3905 | if (kr == KERN_SUCCESS) { |
3906 | vm_info->purgeable_volatile_pmap = | |
3907 | volatile_pmap_size; | |
3e170ce0 A |
3908 | if (radar_20146450) { |
3909 | vm_info->compressed -= | |
3910 | volatile_compressed_pmap_size; | |
3911 | } | |
39236c6e A |
3912 | vm_info->purgeable_volatile_resident = |
3913 | volatile_resident_size; | |
3914 | vm_info->purgeable_volatile_virtual = | |
3915 | volatile_virtual_size; | |
3916 | } | |
4b17d6b6 | 3917 | } |
4b17d6b6 | 3918 | } |
39037602 | 3919 | *task_info_count = TASK_VM_INFO_REV0_COUNT; |
39236c6e | 3920 | |
39037602 A |
3921 | if (original_task_info_count >= TASK_VM_INFO_REV1_COUNT) { |
3922 | vm_info->phys_footprint = | |
3923 | (mach_vm_size_t) get_task_phys_footprint(task); | |
3924 | *task_info_count = TASK_VM_INFO_REV1_COUNT; | |
3925 | } | |
3926 | if (original_task_info_count >= TASK_VM_INFO_REV2_COUNT) { | |
3927 | vm_info->min_address = map->min_offset; | |
3928 | vm_info->max_address = map->max_offset; | |
3929 | *task_info_count = TASK_VM_INFO_REV2_COUNT; | |
3930 | } | |
3931 | ||
3932 | if (task != kernel_task) { | |
3933 | vm_map_unlock_read(map); | |
3e170ce0 A |
3934 | } |
3935 | ||
4b17d6b6 A |
3936 | break; |
3937 | } | |
3938 | ||
fe8ab488 A |
3939 | case TASK_WAIT_STATE_INFO: |
3940 | { | |
3941 | /* | |
3942 | * Deprecated flavor. Currently allowing some results until all users | |
3943 | * stop calling it. The results may not be accurate. | |
3944 | */ | |
3945 | task_wait_state_info_t wait_state_info; | |
3946 | uint64_t total_sfi_ledger_val = 0; | |
3947 | ||
3948 | if (*task_info_count < TASK_WAIT_STATE_INFO_COUNT) { | |
3949 | error = KERN_INVALID_ARGUMENT; | |
3950 | break; | |
3951 | } | |
3952 | ||
3953 | wait_state_info = (task_wait_state_info_t) task_info_out; | |
3954 | ||
3955 | wait_state_info->total_wait_state_time = 0; | |
3956 | bzero(wait_state_info->_reserved, sizeof(wait_state_info->_reserved)); | |
3957 | ||
3e170ce0 | 3958 | #if CONFIG_SCHED_SFI |
fe8ab488 A |
3959 | int i, prev_lentry = -1; |
3960 | int64_t val_credit, val_debit; | |
3961 | ||
3962 | for (i = 0; i < MAX_SFI_CLASS_ID; i++){ | |
3963 | val_credit =0; | |
3964 | /* | |
3965 | * checking with prev_lentry != entry ensures adjacent classes | |
3966 | * which share the same ledger do not add wait times twice. | |
3967 | * Note: Use ledger() call to get data for each individual sfi class. | |
3968 | */ | |
3969 | if (prev_lentry != task_ledgers.sfi_wait_times[i] && | |
3970 | KERN_SUCCESS == ledger_get_entries(task->ledger, | |
3971 | task_ledgers.sfi_wait_times[i], &val_credit, &val_debit)) { | |
3972 | total_sfi_ledger_val += val_credit; | |
3973 | } | |
3974 | prev_lentry = task_ledgers.sfi_wait_times[i]; | |
3975 | } | |
3976 | ||
3e170ce0 | 3977 | #endif /* CONFIG_SCHED_SFI */ |
fe8ab488 A |
3978 | wait_state_info->total_wait_sfi_state_time = total_sfi_ledger_val; |
3979 | *task_info_count = TASK_WAIT_STATE_INFO_COUNT; | |
3980 | ||
3981 | break; | |
3982 | } | |
3e170ce0 A |
3983 | case TASK_VM_INFO_PURGEABLE_ACCOUNT: |
3984 | { | |
3985 | #if DEVELOPMENT || DEBUG | |
3986 | pvm_account_info_t acnt_info; | |
3987 | ||
3988 | if (*task_info_count < PVM_ACCOUNT_INFO_COUNT) { | |
3989 | error = KERN_INVALID_ARGUMENT; | |
3990 | break; | |
3991 | } | |
fe8ab488 | 3992 | |
3e170ce0 A |
3993 | if (task_info_out == NULL) { |
3994 | error = KERN_INVALID_ARGUMENT; | |
3995 | break; | |
3996 | } | |
3997 | ||
3998 | acnt_info = (pvm_account_info_t) task_info_out; | |
3999 | ||
4000 | error = vm_purgeable_account(task, acnt_info); | |
4001 | ||
4002 | *task_info_count = PVM_ACCOUNT_INFO_COUNT; | |
4003 | ||
4004 | break; | |
4005 | #else /* DEVELOPMENT || DEBUG */ | |
4006 | error = KERN_NOT_SUPPORTED; | |
4007 | break; | |
4008 | #endif /* DEVELOPMENT || DEBUG */ | |
4009 | } | |
4010 | case TASK_FLAGS_INFO: | |
4011 | { | |
4012 | task_flags_info_t flags_info; | |
4013 | ||
4014 | if (*task_info_count < TASK_FLAGS_INFO_COUNT) { | |
4015 | error = KERN_INVALID_ARGUMENT; | |
4016 | break; | |
4017 | } | |
4018 | ||
4019 | flags_info = (task_flags_info_t)task_info_out; | |
4020 | ||
4021 | /* only publish the 64-bit flag of the task */ | |
4022 | flags_info->flags = task->t_flags & TF_64B_ADDR; | |
4023 | ||
4024 | *task_info_count = TASK_FLAGS_INFO_COUNT; | |
4025 | break; | |
4026 | } | |
4027 | ||
4028 | case TASK_DEBUG_INFO_INTERNAL: | |
4029 | { | |
4030 | #if DEVELOPMENT || DEBUG | |
4031 | task_debug_info_internal_t dbg_info; | |
4032 | if (*task_info_count < TASK_DEBUG_INFO_INTERNAL_COUNT) { | |
4033 | error = KERN_NOT_SUPPORTED; | |
4034 | break; | |
4035 | } | |
4036 | ||
4037 | if (task_info_out == NULL) { | |
4038 | error = KERN_INVALID_ARGUMENT; | |
4039 | break; | |
4040 | } | |
4041 | dbg_info = (task_debug_info_internal_t) task_info_out; | |
4042 | dbg_info->ipc_space_size = 0; | |
4043 | if (task->itk_space){ | |
4044 | dbg_info->ipc_space_size = task->itk_space->is_table_size; | |
4045 | } | |
4046 | ||
4047 | error = KERN_SUCCESS; | |
4048 | *task_info_count = TASK_DEBUG_INFO_INTERNAL_COUNT; | |
4049 | break; | |
4050 | #else /* DEVELOPMENT || DEBUG */ | |
4051 | error = KERN_NOT_SUPPORTED; | |
4052 | break; | |
4053 | #endif /* DEVELOPMENT || DEBUG */ | |
4054 | } | |
91447636 | 4055 | default: |
b0d623f7 | 4056 | error = KERN_INVALID_ARGUMENT; |
1c79356b A |
4057 | } |
4058 | ||
b0d623f7 A |
4059 | task_unlock(task); |
4060 | return (error); | |
1c79356b A |
4061 | } |
4062 | ||
39236c6e A |
4063 | /* |
4064 | * task_power_info | |
4065 | * | |
4066 | * Returns power stats for the task. | |
4067 | * Note: Called with task locked. | |
4068 | */ | |
4069 | void | |
4070 | task_power_info_locked( | |
4071 | task_t task, | |
fe8ab488 | 4072 | task_power_info_t info, |
39037602 A |
4073 | gpu_energy_data_t ginfo, |
4074 | uint64_t *task_energy) | |
39236c6e A |
4075 | { |
4076 | thread_t thread; | |
4077 | ledger_amount_t tmp; | |
4078 | ||
4079 | task_lock_assert_owned(task); | |
4080 | ||
4081 | ledger_get_entries(task->ledger, task_ledgers.interrupt_wakeups, | |
4082 | (ledger_amount_t *)&info->task_interrupt_wakeups, &tmp); | |
4083 | ledger_get_entries(task->ledger, task_ledgers.platform_idle_wakeups, | |
4084 | (ledger_amount_t *)&info->task_platform_idle_wakeups, &tmp); | |
4085 | ||
4086 | info->task_timer_wakeups_bin_1 = task->task_timer_wakeups_bin_1; | |
4087 | info->task_timer_wakeups_bin_2 = task->task_timer_wakeups_bin_2; | |
4088 | ||
4089 | info->total_user = task->total_user_time; | |
4090 | info->total_system = task->total_system_time; | |
4091 | ||
39037602 A |
4092 | if (task_energy) { |
4093 | *task_energy = task->task_energy; | |
4094 | } | |
4095 | ||
fe8ab488 A |
4096 | if (ginfo) { |
4097 | ginfo->task_gpu_utilisation = task->task_gpu_ns; | |
4098 | } | |
4099 | ||
39236c6e A |
4100 | queue_iterate(&task->threads, thread, thread_t, task_threads) { |
4101 | uint64_t tval; | |
4102 | spl_t x; | |
4103 | ||
4104 | if (thread->options & TH_OPT_IDLE_THREAD) | |
4105 | continue; | |
4106 | ||
4107 | x = splsched(); | |
4108 | thread_lock(thread); | |
4109 | ||
4110 | info->task_timer_wakeups_bin_1 += thread->thread_timer_wakeups_bin_1; | |
4111 | info->task_timer_wakeups_bin_2 += thread->thread_timer_wakeups_bin_2; | |
4112 | ||
39037602 A |
4113 | if (task_energy) { |
4114 | *task_energy += ml_energy_stat(thread); | |
4115 | } | |
4116 | ||
39236c6e A |
4117 | tval = timer_grab(&thread->user_timer); |
4118 | info->total_user += tval; | |
4119 | ||
4120 | tval = timer_grab(&thread->system_timer); | |
4121 | if (thread->precise_user_kernel_time) { | |
4122 | info->total_system += tval; | |
4123 | } else { | |
4124 | /* system_timer may represent either sys or user */ | |
4125 | info->total_user += tval; | |
4126 | } | |
4127 | ||
fe8ab488 A |
4128 | if (ginfo) { |
4129 | ginfo->task_gpu_utilisation += ml_gpu_stat(thread); | |
4130 | } | |
4131 | thread_unlock(thread); | |
4132 | splx(x); | |
4133 | } | |
4134 | } | |
4135 | ||
4136 | /* | |
4137 | * task_gpu_utilisation | |
4138 | * | |
4139 | * Returns the total gpu time used by the all the threads of the task | |
4140 | * (both dead and alive) | |
4141 | */ | |
4142 | uint64_t | |
4143 | task_gpu_utilisation( | |
4144 | task_t task) | |
4145 | { | |
4146 | uint64_t gpu_time = 0; | |
4147 | thread_t thread; | |
4148 | ||
4149 | task_lock(task); | |
4150 | gpu_time += task->task_gpu_ns; | |
4151 | ||
4152 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
4153 | spl_t x; | |
4154 | x = splsched(); | |
4155 | thread_lock(thread); | |
4156 | gpu_time += ml_gpu_stat(thread); | |
39236c6e A |
4157 | thread_unlock(thread); |
4158 | splx(x); | |
4159 | } | |
fe8ab488 A |
4160 | |
4161 | task_unlock(task); | |
4162 | return gpu_time; | |
39236c6e A |
4163 | } |
4164 | ||
39037602 A |
4165 | /* |
4166 | * task_energy | |
4167 | * | |
4168 | * Returns the total energy used by the all the threads of the task | |
4169 | * (both dead and alive) | |
4170 | */ | |
4171 | uint64_t | |
4172 | task_energy( | |
4173 | task_t task) | |
4174 | { | |
4175 | uint64_t energy = 0; | |
4176 | thread_t thread; | |
4177 | ||
4178 | task_lock(task); | |
4179 | energy += task->task_energy; | |
4180 | ||
4181 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
4182 | spl_t x; | |
4183 | x = splsched(); | |
4184 | thread_lock(thread); | |
4185 | energy += ml_energy_stat(thread); | |
4186 | thread_unlock(thread); | |
4187 | splx(x); | |
4188 | } | |
4189 | ||
4190 | task_unlock(task); | |
4191 | return energy; | |
4192 | } | |
4193 | ||
39236c6e A |
4194 | kern_return_t |
4195 | task_purgable_info( | |
4196 | task_t task, | |
4197 | task_purgable_info_t *stats) | |
4198 | { | |
4199 | if (task == TASK_NULL || stats == NULL) | |
4200 | return KERN_INVALID_ARGUMENT; | |
4201 | /* Take task reference */ | |
4202 | task_reference(task); | |
4203 | vm_purgeable_stats((vm_purgeable_info_t)stats, task); | |
4204 | /* Drop task reference */ | |
4205 | task_deallocate(task); | |
4206 | return KERN_SUCCESS; | |
4207 | } | |
4208 | ||
2d21ac55 A |
4209 | void |
4210 | task_vtimer_set( | |
4211 | task_t task, | |
4212 | integer_t which) | |
4213 | { | |
4214 | thread_t thread; | |
316670eb | 4215 | spl_t x; |
2d21ac55 | 4216 | |
2d21ac55 A |
4217 | task_lock(task); |
4218 | ||
4219 | task->vtimers |= which; | |
4220 | ||
4221 | switch (which) { | |
4222 | ||
4223 | case TASK_VTIMER_USER: | |
4224 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
316670eb A |
4225 | x = splsched(); |
4226 | thread_lock(thread); | |
4227 | if (thread->precise_user_kernel_time) | |
4228 | thread->vtimer_user_save = timer_grab(&thread->user_timer); | |
4229 | else | |
4230 | thread->vtimer_user_save = timer_grab(&thread->system_timer); | |
4231 | thread_unlock(thread); | |
4232 | splx(x); | |
2d21ac55 A |
4233 | } |
4234 | break; | |
4235 | ||
4236 | case TASK_VTIMER_PROF: | |
4237 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
316670eb A |
4238 | x = splsched(); |
4239 | thread_lock(thread); | |
2d21ac55 A |
4240 | thread->vtimer_prof_save = timer_grab(&thread->user_timer); |
4241 | thread->vtimer_prof_save += timer_grab(&thread->system_timer); | |
316670eb A |
4242 | thread_unlock(thread); |
4243 | splx(x); | |
2d21ac55 A |
4244 | } |
4245 | break; | |
4246 | ||
4247 | case TASK_VTIMER_RLIM: | |
4248 | queue_iterate(&task->threads, thread, thread_t, task_threads) { | |
316670eb A |
4249 | x = splsched(); |
4250 | thread_lock(thread); | |
2d21ac55 A |
4251 | thread->vtimer_rlim_save = timer_grab(&thread->user_timer); |
4252 | thread->vtimer_rlim_save += timer_grab(&thread->system_timer); | |
316670eb A |
4253 | thread_unlock(thread); |
4254 | splx(x); | |
2d21ac55 A |
4255 | } |
4256 | break; | |
4257 | } | |
4258 | ||
4259 | task_unlock(task); | |
4260 | } | |
4261 | ||
4262 | void | |
4263 | task_vtimer_clear( | |
4264 | task_t task, | |
4265 | integer_t which) | |
4266 | { | |
4267 | assert(task == current_task()); | |
4268 | ||
4269 | task_lock(task); | |
4270 | ||
4271 | task->vtimers &= ~which; | |
4272 | ||
4273 | task_unlock(task); | |
4274 | } | |
4275 | ||
4276 | void | |
4277 | task_vtimer_update( | |
4278 | __unused | |
4279 | task_t task, | |
4280 | integer_t which, | |
4281 | uint32_t *microsecs) | |
4282 | { | |
4283 | thread_t thread = current_thread(); | |
39037602 A |
4284 | uint32_t tdelt = 0; |
4285 | clock_sec_t secs = 0; | |
2d21ac55 A |
4286 | uint64_t tsum; |
4287 | ||
4288 | assert(task == current_task()); | |
4289 | ||
39037602 A |
4290 | spl_t s = splsched(); |
4291 | thread_lock(thread); | |
2d21ac55 | 4292 | |
39037602 A |
4293 | if ((task->vtimers & which) != (uint32_t)which) { |
4294 | thread_unlock(thread); | |
4295 | splx(s); | |
4296 | return; | |
4297 | } | |
2d21ac55 A |
4298 | |
4299 | switch (which) { | |
4300 | ||
4301 | case TASK_VTIMER_USER: | |
316670eb A |
4302 | if (thread->precise_user_kernel_time) { |
4303 | tdelt = (uint32_t)timer_delta(&thread->user_timer, | |
4304 | &thread->vtimer_user_save); | |
4305 | } else { | |
4306 | tdelt = (uint32_t)timer_delta(&thread->system_timer, | |
2d21ac55 | 4307 | &thread->vtimer_user_save); |
316670eb | 4308 | } |
b0d623f7 | 4309 | absolutetime_to_microtime(tdelt, &secs, microsecs); |
2d21ac55 A |
4310 | break; |
4311 | ||
4312 | case TASK_VTIMER_PROF: | |
4313 | tsum = timer_grab(&thread->user_timer); | |
4314 | tsum += timer_grab(&thread->system_timer); | |
b0d623f7 A |
4315 | tdelt = (uint32_t)(tsum - thread->vtimer_prof_save); |
4316 | absolutetime_to_microtime(tdelt, &secs, microsecs); | |
4317 | /* if the time delta is smaller than a usec, ignore */ | |
4318 | if (*microsecs != 0) | |
4319 | thread->vtimer_prof_save = tsum; | |
2d21ac55 A |
4320 | break; |
4321 | ||
4322 | case TASK_VTIMER_RLIM: | |
4323 | tsum = timer_grab(&thread->user_timer); | |
4324 | tsum += timer_grab(&thread->system_timer); | |
b0d623f7 | 4325 | tdelt = (uint32_t)(tsum - thread->vtimer_rlim_save); |
2d21ac55 | 4326 | thread->vtimer_rlim_save = tsum; |
b0d623f7 | 4327 | absolutetime_to_microtime(tdelt, &secs, microsecs); |
2d21ac55 A |
4328 | break; |
4329 | } | |
4330 | ||
39037602 A |
4331 | thread_unlock(thread); |
4332 | splx(s); | |
2d21ac55 A |
4333 | } |
4334 | ||
1c79356b A |
4335 | /* |
4336 | * task_assign: | |
4337 | * | |
4338 | * Change the assigned processor set for the task | |
4339 | */ | |
4340 | kern_return_t | |
4341 | task_assign( | |
91447636 A |
4342 | __unused task_t task, |
4343 | __unused processor_set_t new_pset, | |
4344 | __unused boolean_t assign_threads) | |
1c79356b | 4345 | { |
1c79356b A |
4346 | return(KERN_FAILURE); |
4347 | } | |
4348 | ||
4349 | /* | |
4350 | * task_assign_default: | |
4351 | * | |
4352 | * Version of task_assign to assign to default processor set. | |
4353 | */ | |
4354 | kern_return_t | |
4355 | task_assign_default( | |
4356 | task_t task, | |
4357 | boolean_t assign_threads) | |
4358 | { | |
2d21ac55 | 4359 | return (task_assign(task, &pset0, assign_threads)); |
1c79356b A |
4360 | } |
4361 | ||
4362 | /* | |
4363 | * task_get_assignment | |
4364 | * | |
4365 | * Return name of processor set that task is assigned to. | |
4366 | */ | |
4367 | kern_return_t | |
4368 | task_get_assignment( | |
4369 | task_t task, | |
4370 | processor_set_t *pset) | |
4371 | { | |
7e41aa88 A |
4372 | if (!task || !task->active) |
4373 | return KERN_FAILURE; | |
1c79356b | 4374 | |
2d21ac55 A |
4375 | *pset = &pset0; |
4376 | ||
7e41aa88 | 4377 | return KERN_SUCCESS; |
1c79356b A |
4378 | } |
4379 | ||
3e170ce0 A |
4380 | uint64_t |
4381 | get_task_dispatchqueue_offset( | |
4382 | task_t task) | |
4383 | { | |
4384 | return task->dispatchqueue_offset; | |
4385 | } | |
1c79356b A |
4386 | |
4387 | /* | |
4388 | * task_policy | |
4389 | * | |
4390 | * Set scheduling policy and parameters, both base and limit, for | |
4391 | * the given task. Policy must be a policy which is enabled for the | |
4392 | * processor set. Change contained threads if requested. | |
4393 | */ | |
4394 | kern_return_t | |
4395 | task_policy( | |
91447636 A |
4396 | __unused task_t task, |
4397 | __unused policy_t policy_id, | |
4398 | __unused policy_base_t base, | |
4399 | __unused mach_msg_type_number_t count, | |
4400 | __unused boolean_t set_limit, | |
4401 | __unused boolean_t change) | |
1c79356b A |
4402 | { |
4403 | return(KERN_FAILURE); | |
4404 | } | |
4405 | ||
4406 | /* | |
4407 | * task_set_policy | |
4408 | * | |
4409 | * Set scheduling policy and parameters, both base and limit, for | |
4410 | * the given task. Policy can be any policy implemented by the | |
4411 | * processor set, whether enabled or not. Change contained threads | |
4412 | * if requested. | |
4413 | */ | |
4414 | kern_return_t | |
4415 | task_set_policy( | |
91447636 A |
4416 | __unused task_t task, |
4417 | __unused processor_set_t pset, | |
4418 | __unused policy_t policy_id, | |
4419 | __unused policy_base_t base, | |
4420 | __unused mach_msg_type_number_t base_count, | |
4421 | __unused policy_limit_t limit, | |
4422 | __unused mach_msg_type_number_t limit_count, | |
4423 | __unused boolean_t change) | |
1c79356b A |
4424 | { |
4425 | return(KERN_FAILURE); | |
4426 | } | |
4427 | ||
91447636 A |
4428 | kern_return_t |
4429 | task_set_ras_pc( | |
4430 | __unused task_t task, | |
4431 | __unused vm_offset_t pc, | |
4432 | __unused vm_offset_t endpc) | |
4433 | { | |
1c79356b | 4434 | return KERN_FAILURE; |
1c79356b A |
4435 | } |
4436 | ||
4437 | void | |
4438 | task_synchronizer_destroy_all(task_t task) | |
4439 | { | |
1c79356b A |
4440 | /* |
4441 | * Destroy owned semaphores | |
4442 | */ | |
4bd07ac2 | 4443 | semaphore_destroy_all(task); |
1c79356b A |
4444 | } |
4445 | ||
b0d623f7 A |
4446 | /* |
4447 | * Install default (machine-dependent) initial thread state | |
4448 | * on the task. Subsequent thread creation will have this initial | |
4449 | * state set on the thread by machine_thread_inherit_taskwide(). | |
4450 | * Flavors and structures are exactly the same as those to thread_set_state() | |
4451 | */ | |
4452 | kern_return_t | |
4453 | task_set_state( | |
4454 | task_t task, | |
4455 | int flavor, | |
4456 | thread_state_t state, | |
4457 | mach_msg_type_number_t state_count) | |
4458 | { | |
4459 | kern_return_t ret; | |
4460 | ||
4461 | if (task == TASK_NULL) { | |
4462 | return (KERN_INVALID_ARGUMENT); | |
4463 | } | |
4464 | ||
4465 | task_lock(task); | |
4466 | ||
4467 | if (!task->active) { | |
4468 | task_unlock(task); | |
4469 | return (KERN_FAILURE); | |
4470 | } | |
4471 | ||
4472 | ret = machine_task_set_state(task, flavor, state, state_count); | |
4473 | ||
4474 | task_unlock(task); | |
4475 | return ret; | |
4476 | } | |
4477 | ||
4478 | /* | |
4479 | * Examine the default (machine-dependent) initial thread state | |
4480 | * on the task, as set by task_set_state(). Flavors and structures | |
4481 | * are exactly the same as those passed to thread_get_state(). | |
4482 | */ | |
4483 | kern_return_t | |
4484 | task_get_state( | |
4485 | task_t task, | |
4486 | int flavor, | |
4487 | thread_state_t state, | |
4488 | mach_msg_type_number_t *state_count) | |
4489 | { | |
4490 | kern_return_t ret; | |
4491 | ||
4492 | if (task == TASK_NULL) { | |
4493 | return (KERN_INVALID_ARGUMENT); | |
4494 | } | |
4495 | ||
4496 | task_lock(task); | |
4497 | ||
4498 | if (!task->active) { | |
4499 | task_unlock(task); | |
4500 | return (KERN_FAILURE); | |
4501 | } | |
4502 | ||
4503 | ret = machine_task_get_state(task, flavor, state, state_count); | |
4504 | ||
4505 | task_unlock(task); | |
4506 | return ret; | |
4507 | } | |
4508 | ||
39037602 | 4509 | #if CONFIG_MEMORYSTATUS |
39236c6e A |
4510 | #define HWM_USERCORE_MINSPACE 250 // free space (in MB) required *after* core file creation |
4511 | ||
4512 | void __attribute__((noinline)) | |
39037602 | 4513 | PROC_CROSSED_HIGH_WATERMARK__SEND_EXC_RESOURCE_AND_SUSPEND(int max_footprint_mb, boolean_t is_fatal) |
39236c6e A |
4514 | { |
4515 | task_t task = current_task(); | |
4516 | int pid = 0; | |
3e170ce0 | 4517 | const char *procname = "unknown"; |
39236c6e A |
4518 | mach_exception_data_type_t code[EXCEPTION_CODE_MAX]; |
4519 | ||
4520 | #ifdef MACH_BSD | |
4521 | pid = proc_selfpid(); | |
fe8ab488 A |
4522 | |
4523 | if (pid == 1) { | |
4524 | /* | |
4525 | * Cannot have ReportCrash analyzing | |
4526 | * a suspended initproc. | |
4527 | */ | |
4528 | return; | |
4529 | } | |
4530 | ||
39236c6e A |
4531 | if (task->bsd_info != NULL) |
4532 | procname = proc_name_address(current_task()->bsd_info); | |
4533 | #endif | |
39037602 | 4534 | #if CONFIG_COREDUMP |
39236c6e A |
4535 | if (hwm_user_cores) { |
4536 | int error; | |
4537 | uint64_t starttime, end; | |
4538 | clock_sec_t secs = 0; | |
4539 | uint32_t microsecs = 0; | |
4540 | ||
4541 | starttime = mach_absolute_time(); | |
4542 | /* | |
4543 | * Trigger a coredump of this process. Don't proceed unless we know we won't | |
4544 | * be filling up the disk; and ignore the core size resource limit for this | |
4545 | * core file. | |
4546 | */ | |
3e170ce0 | 4547 | if ((error = coredump(current_task()->bsd_info, HWM_USERCORE_MINSPACE, COREDUMP_IGNORE_ULIMIT)) != 0) { |
39236c6e A |
4548 | printf("couldn't take coredump of %s[%d]: %d\n", procname, pid, error); |
4549 | } | |
4550 | /* | |
4551 | * coredump() leaves the task suspended. | |
4552 | */ | |
4553 | task_resume_internal(current_task()); | |
4554 | ||
4555 | end = mach_absolute_time(); | |
4556 | absolutetime_to_microtime(end - starttime, &secs, µsecs); | |
4557 | printf("coredump of %s[%d] taken in %d secs %d microsecs\n", | |
4558 | proc_name_address(current_task()->bsd_info), pid, (int)secs, microsecs); | |
4559 | } | |
39037602 | 4560 | #endif /* CONFIG_COREDUMP */ |
39236c6e A |
4561 | |
4562 | if (disable_exc_resource) { | |
4563 | printf("process %s[%d] crossed memory high watermark (%d MB); EXC_RESOURCE " | |
4564 | "supressed by a boot-arg.\n", procname, pid, max_footprint_mb); | |
4565 | return; | |
4566 | } | |
4567 | ||
3e170ce0 A |
4568 | /* |
4569 | * A task that has triggered an EXC_RESOURCE, should not be | |
4570 | * jetsammed when the device is under memory pressure. Here | |
4571 | * we set the P_MEMSTAT_TERMINATED flag so that the process | |
4572 | * will be skipped if the memorystatus_thread wakes up. | |
4573 | */ | |
4574 | proc_memstat_terminated(current_task()->bsd_info, TRUE); | |
4575 | ||
39236c6e A |
4576 | printf("process %s[%d] crossed memory high watermark (%d MB); sending " |
4577 | "EXC_RESOURCE.\n", procname, pid, max_footprint_mb); | |
4578 | ||
4579 | code[0] = code[1] = 0; | |
4580 | EXC_RESOURCE_ENCODE_TYPE(code[0], RESOURCE_TYPE_MEMORY); | |
4581 | EXC_RESOURCE_ENCODE_FLAVOR(code[0], FLAVOR_HIGH_WATERMARK); | |
4582 | EXC_RESOURCE_HWM_ENCODE_LIMIT(code[0], max_footprint_mb); | |
3e170ce0 | 4583 | |
39037602 A |
4584 | /* Do not generate a corpse fork if the violation is a fatal one */ |
4585 | if (is_fatal || exc_via_corpse_forking == 0) { | |
4586 | /* Do not send a EXC_RESOURCE is corpse_for_fatal_memkill is set */ | |
4587 | if (corpse_for_fatal_memkill == 0) { | |
4588 | /* | |
4589 | * Use the _internal_ variant so that no user-space | |
4590 | * process can resume our task from under us. | |
4591 | */ | |
4592 | task_suspend_internal(task); | |
4593 | exception_triage(EXC_RESOURCE, code, EXCEPTION_CODE_MAX); | |
4594 | task_resume_internal(task); | |
4595 | } | |
4596 | } else { | |
4597 | task_enqueue_exception_with_corpse(task, code, EXCEPTION_CODE_MAX); | |
4598 | } | |
3e170ce0 A |
4599 | |
4600 | /* | |
4601 | * After the EXC_RESOURCE has been handled, we must clear the | |
4602 | * P_MEMSTAT_TERMINATED flag so that the process can again be | |
4603 | * considered for jetsam if the memorystatus_thread wakes up. | |
4604 | */ | |
4605 | proc_memstat_terminated(current_task()->bsd_info, FALSE); /* clear the flag */ | |
39236c6e A |
4606 | } |
4607 | ||
4608 | /* | |
4609 | * Callback invoked when a task exceeds its physical footprint limit. | |
4610 | */ | |
4611 | void | |
4612 | task_footprint_exceeded(int warning, __unused const void *param0, __unused const void *param1) | |
4613 | { | |
fe8ab488 | 4614 | ledger_amount_t max_footprint, max_footprint_mb; |
fe8ab488 | 4615 | task_t task; |
39037602 A |
4616 | boolean_t is_fatal; |
4617 | boolean_t trigger_exception; | |
39236c6e A |
4618 | |
4619 | if (warning == LEDGER_WARNING_DIPPED_BELOW) { | |
4620 | /* | |
4621 | * Task memory limits only provide a warning on the way up. | |
4622 | */ | |
4623 | return; | |
4624 | } | |
4625 | ||
fe8ab488 A |
4626 | task = current_task(); |
4627 | ||
4628 | ledger_get_limit(task->ledger, task_ledgers.phys_footprint, &max_footprint); | |
4629 | max_footprint_mb = max_footprint >> 20; | |
4630 | ||
4631 | /* | |
39037602 | 4632 | * Capture the trigger exception flag before turning off the exception. |
fe8ab488 | 4633 | */ |
39037602 A |
4634 | trigger_exception = task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_PHYS_FOOTPRINT_EXCEPTION ? TRUE : FALSE; |
4635 | ||
4636 | is_fatal = memorystatus_turnoff_exception_and_get_fatalness((warning == LEDGER_WARNING_ROSE_ABOVE) ? TRUE : FALSE, (int)max_footprint_mb); | |
39236c6e A |
4637 | |
4638 | /* | |
4639 | * If this an actual violation (not a warning), | |
4640 | * generate a non-fatal high watermark EXC_RESOURCE. | |
4641 | */ | |
39037602 A |
4642 | if ((warning == 0) && trigger_exception) { |
4643 | PROC_CROSSED_HIGH_WATERMARK__SEND_EXC_RESOURCE_AND_SUSPEND((int)max_footprint_mb, is_fatal); | |
39236c6e A |
4644 | } |
4645 | ||
4646 | memorystatus_on_ledger_footprint_exceeded((warning == LEDGER_WARNING_ROSE_ABOVE) ? TRUE : FALSE, | |
39037602 | 4647 | is_fatal); |
39236c6e A |
4648 | } |
4649 | ||
4650 | extern int proc_check_footprint_priv(void); | |
4651 | ||
4652 | kern_return_t | |
4653 | task_set_phys_footprint_limit( | |
4654 | task_t task, | |
4655 | int new_limit_mb, | |
4656 | int *old_limit_mb) | |
4657 | { | |
4658 | kern_return_t error; | |
4659 | ||
4660 | if ((error = proc_check_footprint_priv())) { | |
4661 | return (KERN_NO_ACCESS); | |
4662 | } | |
4663 | ||
4664 | return task_set_phys_footprint_limit_internal(task, new_limit_mb, old_limit_mb, FALSE); | |
4665 | } | |
4666 | ||
3e170ce0 A |
4667 | kern_return_t |
4668 | task_convert_phys_footprint_limit( | |
4669 | int limit_mb, | |
4670 | int *converted_limit_mb) | |
4671 | { | |
4672 | if (limit_mb == -1) { | |
4673 | /* | |
4674 | * No limit | |
4675 | */ | |
4676 | if (max_task_footprint != 0) { | |
4677 | *converted_limit_mb = (int)(max_task_footprint / 1024 / 1024); /* bytes to MB */ | |
4678 | } else { | |
4679 | *converted_limit_mb = (int)(LEDGER_LIMIT_INFINITY >> 20); | |
4680 | } | |
4681 | } else { | |
4682 | /* nothing to convert */ | |
4683 | *converted_limit_mb = limit_mb; | |
4684 | } | |
4685 | return (KERN_SUCCESS); | |
4686 | } | |
4687 | ||
4688 | ||
39236c6e A |
4689 | kern_return_t |
4690 | task_set_phys_footprint_limit_internal( | |
4691 | task_t task, | |
4692 | int new_limit_mb, | |
4693 | int *old_limit_mb, | |
4694 | boolean_t trigger_exception) | |
4695 | { | |
4696 | ledger_amount_t old; | |
4697 | ||
4698 | ledger_get_limit(task->ledger, task_ledgers.phys_footprint, &old); | |
4699 | ||
4700 | if (old_limit_mb) { | |
3e170ce0 A |
4701 | /* |
4702 | * Check that limit >> 20 will not give an "unexpected" 32-bit | |
4703 | * result. There are, however, implicit assumptions that -1 mb limit | |
4704 | * equates to LEDGER_LIMIT_INFINITY. | |
4705 | */ | |
4706 | assert(((old & 0xFFF0000000000000LL) == 0) || (old == LEDGER_LIMIT_INFINITY)); | |
4707 | *old_limit_mb = (int)(old >> 20); | |
39236c6e A |
4708 | } |
4709 | ||
4710 | if (new_limit_mb == -1) { | |
4711 | /* | |
4712 | * Caller wishes to remove the limit. | |
4713 | */ | |
4714 | ledger_set_limit(task->ledger, task_ledgers.phys_footprint, | |
4715 | max_task_footprint ? max_task_footprint : LEDGER_LIMIT_INFINITY, | |
39037602 | 4716 | max_task_footprint ? max_task_footprint_warning_level : 0); |
39236c6e A |
4717 | return (KERN_SUCCESS); |
4718 | } | |
4719 | ||
4720 | #ifdef CONFIG_NOMONITORS | |
4721 | return (KERN_SUCCESS); | |
4722 | #endif /* CONFIG_NOMONITORS */ | |
4723 | ||
4724 | task_lock(task); | |
4725 | ||
4726 | if (trigger_exception) { | |
4727 | task->rusage_cpu_flags |= TASK_RUSECPU_FLAGS_PHYS_FOOTPRINT_EXCEPTION; | |
4728 | } else { | |
4729 | task->rusage_cpu_flags &= ~TASK_RUSECPU_FLAGS_PHYS_FOOTPRINT_EXCEPTION; | |
4730 | } | |
4731 | ||
4732 | ledger_set_limit(task->ledger, task_ledgers.phys_footprint, | |
4733 | (ledger_amount_t)new_limit_mb << 20, PHYS_FOOTPRINT_WARNING_LEVEL); | |
4734 | ||
3e170ce0 A |
4735 | if (task == current_task()) { |
4736 | ledger_check_new_balance(task->ledger, task_ledgers.phys_footprint); | |
4737 | } | |
4738 | ||
39236c6e A |
4739 | task_unlock(task); |
4740 | ||
4741 | return (KERN_SUCCESS); | |
4742 | } | |
4743 | ||
4744 | kern_return_t | |
4745 | task_get_phys_footprint_limit( | |
4746 | task_t task, | |
4747 | int *limit_mb) | |
4748 | { | |
4749 | ledger_amount_t limit; | |
4750 | ||
4751 | ledger_get_limit(task->ledger, task_ledgers.phys_footprint, &limit); | |
3e170ce0 A |
4752 | /* |
4753 | * Check that limit >> 20 will not give an "unexpected" signed, 32-bit | |
4754 | * result. There are, however, implicit assumptions that -1 mb limit | |
4755 | * equates to LEDGER_LIMIT_INFINITY. | |
4756 | */ | |
4757 | assert(((limit & 0xFFF0000000000000LL) == 0) || (limit == LEDGER_LIMIT_INFINITY)); | |
4758 | *limit_mb = (int)(limit >> 20); | |
39236c6e A |
4759 | |
4760 | return (KERN_SUCCESS); | |
4761 | } | |
39037602 | 4762 | #else /* CONFIG_MEMORYSTATUS */ |
39236c6e A |
4763 | kern_return_t |
4764 | task_set_phys_footprint_limit( | |
4765 | __unused task_t task, | |
4766 | __unused int new_limit_mb, | |
4767 | __unused int *old_limit_mb) | |
4768 | { | |
4769 | return (KERN_FAILURE); | |
4770 | } | |
4771 | ||
4772 | kern_return_t | |
4773 | task_get_phys_footprint_limit( | |
4774 | __unused task_t task, | |
4775 | __unused int *limit_mb) | |
4776 | { | |
4777 | return (KERN_FAILURE); | |
4778 | } | |
39037602 | 4779 | #endif /* CONFIG_MEMORYSTATUS */ |
b0d623f7 | 4780 | |
1c79356b A |
4781 | /* |
4782 | * We need to export some functions to other components that | |
4783 | * are currently implemented in macros within the osfmk | |
4784 | * component. Just export them as functions of the same name. | |
4785 | */ | |
4786 | boolean_t is_kerneltask(task_t t) | |
4787 | { | |
4788 | if (t == kernel_task) | |
55e303ae A |
4789 | return (TRUE); |
4790 | ||
4791 | return (FALSE); | |
1c79356b A |
4792 | } |
4793 | ||
39037602 | 4794 | boolean_t is_corpsetask(task_t t) |
b0d623f7 | 4795 | { |
39037602 | 4796 | return (task_is_a_corpse(t)); |
b0d623f7 A |
4797 | } |
4798 | ||
1c79356b | 4799 | #undef current_task |
91447636 A |
4800 | task_t current_task(void); |
4801 | task_t current_task(void) | |
1c79356b A |
4802 | { |
4803 | return (current_task_fast()); | |
4804 | } | |
91447636 A |
4805 | |
4806 | #undef task_reference | |
4807 | void task_reference(task_t task); | |
4808 | void | |
4809 | task_reference( | |
4810 | task_t task) | |
4811 | { | |
4812 | if (task != TASK_NULL) | |
4813 | task_reference_internal(task); | |
4814 | } | |
2d21ac55 | 4815 | |
3e170ce0 A |
4816 | /* defined in bsd/kern/kern_prot.c */ |
4817 | extern int get_audit_token_pid(audit_token_t *audit_token); | |
4818 | ||
4819 | int task_pid(task_t task) | |
4820 | { | |
4821 | if (task) | |
4822 | return get_audit_token_pid(&task->audit_token); | |
4823 | return -1; | |
4824 | } | |
4825 | ||
4826 | ||
39037602 A |
4827 | /* |
4828 | * This routine finds a thread in a task by its unique id | |
4829 | * Returns a referenced thread or THREAD_NULL if the thread was not found | |
4830 | * | |
4831 | * TODO: This is super inefficient - it's an O(threads in task) list walk! | |
4832 | * We should make a tid hash, or transition all tid clients to thread ports | |
4833 | * | |
4834 | * Precondition: No locks held (will take task lock) | |
6d2010ae A |
4835 | */ |
4836 | thread_t | |
4837 | task_findtid(task_t task, uint64_t tid) | |
4838 | { | |
39037602 A |
4839 | thread_t self = current_thread(); |
4840 | thread_t found_thread = THREAD_NULL; | |
4841 | thread_t iter_thread = THREAD_NULL; | |
6d2010ae | 4842 | |
39037602 A |
4843 | /* Short-circuit the lookup if we're looking up ourselves */ |
4844 | if (tid == self->thread_id || tid == TID_NULL) { | |
4845 | assert(self->task == task); | |
4846 | ||
4847 | thread_reference(self); | |
4848 | ||
4849 | return self; | |
6d2010ae | 4850 | } |
39037602 A |
4851 | |
4852 | task_lock(task); | |
4853 | ||
4854 | queue_iterate(&task->threads, iter_thread, thread_t, task_threads) { | |
4855 | if (iter_thread->thread_id == tid) { | |
4856 | found_thread = iter_thread; | |
4857 | thread_reference(found_thread); | |
4858 | break; | |
4859 | } | |
4860 | } | |
4861 | ||
4862 | task_unlock(task); | |
4863 | ||
4864 | return (found_thread); | |
6d2010ae A |
4865 | } |
4866 | ||
39037602 | 4867 | |
39236c6e A |
4868 | /* |
4869 | * Control the CPU usage monitor for a task. | |
4870 | */ | |
4871 | kern_return_t | |
4872 | task_cpu_usage_monitor_ctl(task_t task, uint32_t *flags) | |
4873 | { | |
4874 | int error = KERN_SUCCESS; | |
4875 | ||
4876 | if (*flags & CPUMON_MAKE_FATAL) { | |
4877 | task->rusage_cpu_flags |= TASK_RUSECPU_FLAGS_FATAL_CPUMON; | |
4878 | } else { | |
4879 | error = KERN_INVALID_ARGUMENT; | |
4880 | } | |
4881 | ||
4882 | return error; | |
4883 | } | |
4884 | ||
4885 | /* | |
4886 | * Control the wakeups monitor for a task. | |
4887 | */ | |
4888 | kern_return_t | |
4889 | task_wakeups_monitor_ctl(task_t task, uint32_t *flags, int32_t *rate_hz) | |
4890 | { | |
4891 | ledger_t ledger = task->ledger; | |
4892 | ||
4893 | task_lock(task); | |
4894 | if (*flags & WAKEMON_GET_PARAMS) { | |
4895 | ledger_amount_t limit; | |
4896 | uint64_t period; | |
4897 | ||
4898 | ledger_get_limit(ledger, task_ledgers.interrupt_wakeups, &limit); | |
4899 | ledger_get_period(ledger, task_ledgers.interrupt_wakeups, &period); | |
4900 | ||
4901 | if (limit != LEDGER_LIMIT_INFINITY) { | |
4902 | /* | |
4903 | * An active limit means the wakeups monitor is enabled. | |
4904 | */ | |
4905 | *rate_hz = (int32_t)(limit / (int64_t)(period / NSEC_PER_SEC)); | |
4906 | *flags = WAKEMON_ENABLE; | |
4907 | if (task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_FATAL_WAKEUPSMON) { | |
4908 | *flags |= WAKEMON_MAKE_FATAL; | |
4909 | } | |
4910 | } else { | |
4911 | *flags = WAKEMON_DISABLE; | |
4912 | *rate_hz = -1; | |
4913 | } | |
4914 | ||
4915 | /* | |
4916 | * If WAKEMON_GET_PARAMS is present in flags, all other flags are ignored. | |
4917 | */ | |
4918 | task_unlock(task); | |
4919 | return KERN_SUCCESS; | |
4920 | } | |
4921 | ||
4922 | if (*flags & WAKEMON_ENABLE) { | |
4923 | if (*flags & WAKEMON_SET_DEFAULTS) { | |
4924 | *rate_hz = task_wakeups_monitor_rate; | |
4925 | } | |
4926 | ||
4927 | #ifndef CONFIG_NOMONITORS | |
4928 | if (*flags & WAKEMON_MAKE_FATAL) { | |
4929 | task->rusage_cpu_flags |= TASK_RUSECPU_FLAGS_FATAL_WAKEUPSMON; | |
4930 | } | |
4931 | #endif /* CONFIG_NOMONITORS */ | |
4932 | ||
39037602 | 4933 | if (*rate_hz <= 0) { |
39236c6e A |
4934 | task_unlock(task); |
4935 | return KERN_INVALID_ARGUMENT; | |
4936 | } | |
4937 | ||
4938 | #ifndef CONFIG_NOMONITORS | |
4939 | ledger_set_limit(ledger, task_ledgers.interrupt_wakeups, *rate_hz * task_wakeups_monitor_interval, | |
4940 | task_wakeups_monitor_ustackshots_trigger_pct); | |
4941 | ledger_set_period(ledger, task_ledgers.interrupt_wakeups, task_wakeups_monitor_interval * NSEC_PER_SEC); | |
4942 | ledger_enable_callback(ledger, task_ledgers.interrupt_wakeups); | |
4943 | #endif /* CONFIG_NOMONITORS */ | |
4944 | } else if (*flags & WAKEMON_DISABLE) { | |
4945 | /* | |
4946 | * Caller wishes to disable wakeups monitor on the task. | |
4947 | * | |
4948 | * Disable telemetry if it was triggered by the wakeups monitor, and | |
4949 | * remove the limit & callback on the wakeups ledger entry. | |
4950 | */ | |
4951 | #if CONFIG_TELEMETRY | |
490019cf | 4952 | telemetry_task_ctl_locked(task, TF_WAKEMON_WARNING, 0); |
39236c6e A |
4953 | #endif |
4954 | ledger_disable_refill(ledger, task_ledgers.interrupt_wakeups); | |
4955 | ledger_disable_callback(ledger, task_ledgers.interrupt_wakeups); | |
4956 | } | |
4957 | ||
4958 | task_unlock(task); | |
4959 | return KERN_SUCCESS; | |
4960 | } | |
4961 | ||
4962 | void | |
4963 | task_wakeups_rate_exceeded(int warning, __unused const void *param0, __unused const void *param1) | |
4964 | { | |
4965 | if (warning == LEDGER_WARNING_ROSE_ABOVE) { | |
4966 | #if CONFIG_TELEMETRY | |
4967 | /* | |
4968 | * This task is in danger of violating the wakeups monitor. Enable telemetry on this task | |
4969 | * so there are micro-stackshots available if and when EXC_RESOURCE is triggered. | |
4970 | */ | |
4971 | telemetry_task_ctl(current_task(), TF_WAKEMON_WARNING, 1); | |
4972 | #endif | |
4973 | return; | |
4974 | } | |
4975 | ||
4976 | #if CONFIG_TELEMETRY | |
4977 | /* | |
4978 | * If the balance has dipped below the warning level (LEDGER_WARNING_DIPPED_BELOW) or | |
4979 | * exceeded the limit, turn telemetry off for the task. | |
4980 | */ | |
4981 | telemetry_task_ctl(current_task(), TF_WAKEMON_WARNING, 0); | |
4982 | #endif | |
4983 | ||
4984 | if (warning == 0) { | |
39037602 | 4985 | SENDING_NOTIFICATION__THIS_PROCESS_IS_CAUSING_TOO_MANY_WAKEUPS(); |
39236c6e A |
4986 | } |
4987 | } | |
4988 | ||
4989 | void __attribute__((noinline)) | |
39037602 | 4990 | SENDING_NOTIFICATION__THIS_PROCESS_IS_CAUSING_TOO_MANY_WAKEUPS(void) |
39236c6e | 4991 | { |
39037602 A |
4992 | task_t task = current_task(); |
4993 | int pid = 0; | |
4994 | const char *procname = "unknown"; | |
4995 | boolean_t fatal; | |
4996 | kern_return_t kr; | |
4997 | #ifdef EXC_RESOURCE_MONITORS | |
4998 | mach_exception_data_type_t code[EXCEPTION_CODE_MAX]; | |
4999 | #endif /* EXC_RESOURCE_MONITORS */ | |
5000 | struct ledger_entry_info lei; | |
39236c6e A |
5001 | |
5002 | #ifdef MACH_BSD | |
5003 | pid = proc_selfpid(); | |
5004 | if (task->bsd_info != NULL) | |
5005 | procname = proc_name_address(current_task()->bsd_info); | |
5006 | #endif | |
5007 | ||
5008 | ledger_get_entry_info(task->ledger, task_ledgers.interrupt_wakeups, &lei); | |
5009 | ||
5010 | /* | |
5011 | * Disable the exception notification so we don't overwhelm | |
5012 | * the listener with an endless stream of redundant exceptions. | |
39037602 | 5013 | * TODO: detect whether another thread is already reporting the violation. |
39236c6e A |
5014 | */ |
5015 | uint32_t flags = WAKEMON_DISABLE; | |
5016 | task_wakeups_monitor_ctl(task, &flags, NULL); | |
5017 | ||
39037602 A |
5018 | fatal = task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_FATAL_WAKEUPSMON; |
5019 | trace_resource_violation(RMON_CPUWAKES_VIOLATED, &lei); | |
5020 | printf("process %s[%d] caught waking the CPU %llu times " | |
5021 | "over ~%llu seconds, averaging %llu wakes / second and " | |
5022 | "violating a %slimit of %llu wakes over %llu seconds.\n", | |
5023 | procname, pid, | |
5024 | lei.lei_balance, lei.lei_last_refill / NSEC_PER_SEC, | |
5025 | lei.lei_last_refill == 0 ? 0 : | |
5026 | (NSEC_PER_SEC * lei.lei_balance / lei.lei_last_refill), | |
5027 | fatal ? "FATAL " : "", | |
5028 | lei.lei_limit, lei.lei_refill_period / NSEC_PER_SEC); | |
5029 | ||
5030 | kr = send_resource_violation(send_cpu_wakes_violation, task, &lei, | |
5031 | fatal ? kRNFatalLimitFlag : 0); | |
5032 | if (kr) { | |
5033 | printf("send_resource_violation(CPU wakes, ...): error %#x\n", kr); | |
5034 | } | |
5035 | ||
5036 | #ifdef EXC_RESOURCE_MONITORS | |
39236c6e A |
5037 | if (disable_exc_resource) { |
5038 | printf("process %s[%d] caught causing excessive wakeups. EXC_RESOURCE " | |
5039 | "supressed by a boot-arg\n", procname, pid); | |
5040 | return; | |
5041 | } | |
15129b1c A |
5042 | if (audio_active) { |
5043 | printf("process %s[%d] caught causing excessive wakeups. EXC_RESOURCE " | |
5044 | "supressed due to audio playback\n", procname, pid); | |
5045 | return; | |
5046 | } | |
39037602 A |
5047 | if (lei.lei_last_refill == 0) { |
5048 | printf("process %s[%d] caught causing excessive wakeups. EXC_RESOURCE " | |
5049 | "supressed due to lei.lei_last_refill = 0 \n", procname, pid); | |
5050 | } | |
39236c6e A |
5051 | |
5052 | code[0] = code[1] = 0; | |
5053 | EXC_RESOURCE_ENCODE_TYPE(code[0], RESOURCE_TYPE_WAKEUPS); | |
5054 | EXC_RESOURCE_ENCODE_FLAVOR(code[0], FLAVOR_WAKEUPS_MONITOR); | |
39037602 A |
5055 | EXC_RESOURCE_CPUMONITOR_ENCODE_WAKEUPS_PERMITTED(code[0], |
5056 | NSEC_PER_SEC * lei.lei_limit / lei.lei_refill_period); | |
5057 | EXC_RESOURCE_CPUMONITOR_ENCODE_OBSERVATION_INTERVAL(code[0], | |
5058 | lei.lei_last_refill); | |
5059 | EXC_RESOURCE_CPUMONITOR_ENCODE_WAKEUPS_OBSERVED(code[1], | |
5060 | NSEC_PER_SEC * lei.lei_balance / lei.lei_last_refill); | |
39236c6e | 5061 | exception_triage(EXC_RESOURCE, code, EXCEPTION_CODE_MAX); |
39037602 | 5062 | #endif /* EXC_RESOURCE_MONITORS */ |
39236c6e | 5063 | |
39037602 | 5064 | if (fatal) { |
39236c6e A |
5065 | task_terminate_internal(task); |
5066 | } | |
5067 | } | |
fe8ab488 | 5068 | |
39037602 A |
5069 | static boolean_t |
5070 | global_update_logical_writes(int64_t io_delta) | |
fe8ab488 | 5071 | { |
39037602 A |
5072 | int64_t old_count, new_count; |
5073 | boolean_t needs_telemetry; | |
5074 | ||
5075 | do { | |
5076 | new_count = old_count = global_logical_writes_count; | |
5077 | new_count += io_delta; | |
5078 | if (new_count >= io_telemetry_limit) { | |
5079 | new_count = 0; | |
5080 | needs_telemetry = TRUE; | |
5081 | } else { | |
5082 | needs_telemetry = FALSE; | |
5083 | } | |
5084 | } while(!OSCompareAndSwap64(old_count, new_count, &global_logical_writes_count)); | |
5085 | return needs_telemetry; | |
5086 | } | |
fe8ab488 | 5087 | |
39037602 A |
5088 | void task_update_logical_writes(task_t task, uint32_t io_size, int flags, void *vp) |
5089 | { | |
5090 | int64_t io_delta = 0; | |
5091 | boolean_t needs_telemetry = FALSE; | |
fe8ab488 | 5092 | |
39037602 A |
5093 | if ((!task) || (!io_size) || (!vp)) |
5094 | return; | |
5095 | ||
5096 | KERNEL_DEBUG_CONSTANT((MACHDBG_CODE(DBG_MACH_VM, VM_DATA_WRITE)) | DBG_FUNC_NONE, | |
5097 | task_pid(task), io_size, flags, (uintptr_t)VM_KERNEL_ADDRPERM(vp), 0); | |
5098 | DTRACE_IO4(logical_writes, struct task *, task, uint32_t, io_size, int, flags, vnode *, vp); | |
5099 | switch(flags) { | |
5100 | case TASK_WRITE_IMMEDIATE: | |
5101 | OSAddAtomic64(io_size, (SInt64 *)&(task->task_immediate_writes)); | |
5102 | ledger_credit(task->ledger, task_ledgers.logical_writes, io_size); | |
5103 | break; | |
5104 | case TASK_WRITE_DEFERRED: | |
5105 | OSAddAtomic64(io_size, (SInt64 *)&(task->task_deferred_writes)); | |
5106 | ledger_credit(task->ledger, task_ledgers.logical_writes, io_size); | |
5107 | break; | |
5108 | case TASK_WRITE_INVALIDATED: | |
5109 | OSAddAtomic64(io_size, (SInt64 *)&(task->task_invalidated_writes)); | |
5110 | ledger_debit(task->ledger, task_ledgers.logical_writes, io_size); | |
5111 | break; | |
5112 | case TASK_WRITE_METADATA: | |
5113 | OSAddAtomic64(io_size, (SInt64 *)&(task->task_metadata_writes)); | |
5114 | ledger_credit(task->ledger, task_ledgers.logical_writes, io_size); | |
5115 | break; | |
fe8ab488 | 5116 | } |
39037602 A |
5117 | |
5118 | io_delta = (flags == TASK_WRITE_INVALIDATED) ? ((int64_t)io_size * -1ll) : ((int64_t)io_size); | |
5119 | if (io_telemetry_limit != 0) { | |
5120 | /* If io_telemetry_limit is 0, disable global updates and I/O telemetry */ | |
5121 | needs_telemetry = global_update_logical_writes(io_delta); | |
5122 | if (needs_telemetry) { | |
5123 | act_set_io_telemetry_ast(current_thread()); | |
5124 | } | |
fe8ab488 | 5125 | } |
39037602 | 5126 | } |
fe8ab488 | 5127 | |
39037602 A |
5128 | /* |
5129 | * Control the I/O monitor for a task. | |
5130 | */ | |
5131 | kern_return_t | |
5132 | task_io_monitor_ctl(task_t task, uint32_t *flags) | |
5133 | { | |
5134 | ledger_t ledger = task->ledger; | |
fe8ab488 | 5135 | |
39037602 A |
5136 | task_lock(task); |
5137 | if (*flags & IOMON_ENABLE) { | |
5138 | /* Configure the physical I/O ledger */ | |
5139 | ledger_set_limit(ledger, task_ledgers.physical_writes, (task_iomon_limit_mb * 1024 * 1024), 0); | |
5140 | ledger_set_period(ledger, task_ledgers.physical_writes, (task_iomon_interval_secs * NSEC_PER_SEC)); | |
5141 | ||
5142 | /* Configure the logical I/O ledger */ | |
5143 | ledger_set_limit(ledger, task_ledgers.logical_writes, (task_iomon_limit_mb * 1024 * 1024), 0); | |
5144 | ledger_set_period(ledger, task_ledgers.logical_writes, (task_iomon_interval_secs * NSEC_PER_SEC)); | |
5145 | ||
5146 | } else if (*flags & IOMON_DISABLE) { | |
5147 | /* | |
5148 | * Caller wishes to disable I/O monitor on the task. | |
5149 | */ | |
5150 | ledger_disable_refill(ledger, task_ledgers.physical_writes); | |
5151 | ledger_disable_callback(ledger, task_ledgers.physical_writes); | |
5152 | ledger_disable_refill(ledger, task_ledgers.logical_writes); | |
5153 | ledger_disable_callback(ledger, task_ledgers.logical_writes); | |
5154 | } | |
fe8ab488 | 5155 | |
39037602 | 5156 | task_unlock(task); |
fe8ab488 A |
5157 | return KERN_SUCCESS; |
5158 | } | |
5159 | ||
39037602 A |
5160 | void |
5161 | task_io_rate_exceeded(int warning, const void *param0, __unused const void *param1) | |
5162 | { | |
5163 | if (warning == 0) { | |
5164 | SENDING_NOTIFICATION__THIS_PROCESS_IS_CAUSING_TOO_MUCH_IO((int)param0); | |
5165 | } | |
5166 | } | |
5167 | ||
5168 | void __attribute__((noinline)) SENDING_NOTIFICATION__THIS_PROCESS_IS_CAUSING_TOO_MUCH_IO(int flavor) | |
5169 | { | |
5170 | int pid = 0; | |
5171 | task_t task = current_task(); | |
5172 | #ifdef EXC_RESOURCE_MONITORS | |
5173 | mach_exception_data_type_t code[EXCEPTION_CODE_MAX]; | |
5174 | #endif /* EXC_RESOURCE_MONITORS */ | |
5175 | struct ledger_entry_info lei; | |
5176 | kern_return_t kr; | |
5177 | ||
5178 | #ifdef MACH_BSD | |
5179 | pid = proc_selfpid(); | |
5180 | #endif | |
5181 | /* | |
5182 | * Get the ledger entry info. We need to do this before disabling the exception | |
5183 | * to get correct values for all fields. | |
5184 | */ | |
5185 | switch(flavor) { | |
5186 | case FLAVOR_IO_PHYSICAL_WRITES: | |
5187 | ledger_get_entry_info(task->ledger, task_ledgers.physical_writes, &lei); | |
5188 | break; | |
5189 | case FLAVOR_IO_LOGICAL_WRITES: | |
5190 | ledger_get_entry_info(task->ledger, task_ledgers.logical_writes, &lei); | |
5191 | break; | |
5192 | } | |
5193 | ||
5194 | ||
5195 | /* | |
5196 | * Disable the exception notification so we don't overwhelm | |
5197 | * the listener with an endless stream of redundant exceptions. | |
5198 | * TODO: detect whether another thread is already reporting the violation. | |
5199 | */ | |
5200 | uint32_t flags = IOMON_DISABLE; | |
5201 | task_io_monitor_ctl(task, &flags); | |
5202 | ||
5203 | if (flavor == FLAVOR_IO_LOGICAL_WRITES) { | |
5204 | trace_resource_violation(RMON_LOGWRITES_VIOLATED, &lei); | |
5205 | } | |
5206 | printf("process [%d] caught causing excessive I/O (flavor: %d). Task I/O: %lld MB. [Limit : %lld MB per %lld secs]\n", | |
5207 | pid, flavor, (lei.lei_balance / (1024 * 1024)), (lei.lei_limit / (1024 * 1024)), (lei.lei_refill_period / NSEC_PER_SEC)); | |
5208 | ||
5209 | kr = send_resource_violation(send_disk_writes_violation, task, &lei, kRNFlagsNone); | |
5210 | if (kr) { | |
5211 | printf("send_resource_violation(disk_writes, ...): error %#x\n", kr); | |
5212 | } | |
5213 | ||
5214 | #ifdef EXC_RESOURCE_MONITORS | |
5215 | code[0] = code[1] = 0; | |
5216 | EXC_RESOURCE_ENCODE_TYPE(code[0], RESOURCE_TYPE_IO); | |
5217 | EXC_RESOURCE_ENCODE_FLAVOR(code[0], flavor); | |
5218 | EXC_RESOURCE_IO_ENCODE_INTERVAL(code[0], (lei.lei_refill_period / NSEC_PER_SEC)); | |
5219 | EXC_RESOURCE_IO_ENCODE_LIMIT(code[0], (lei.lei_limit / (1024 * 1024))); | |
5220 | EXC_RESOURCE_IO_ENCODE_OBSERVED(code[1], (lei.lei_balance / (1024 * 1024))); | |
5221 | exception_triage(EXC_RESOURCE, code, EXCEPTION_CODE_MAX); | |
5222 | #endif /* EXC_RESOURCE_MONITORS */ | |
5223 | } | |
5224 | ||
fe8ab488 A |
5225 | /* Placeholders for the task set/get voucher interfaces */ |
5226 | kern_return_t | |
5227 | task_get_mach_voucher( | |
5228 | task_t task, | |
5229 | mach_voucher_selector_t __unused which, | |
5230 | ipc_voucher_t *voucher) | |
5231 | { | |
5232 | if (TASK_NULL == task) | |
5233 | return KERN_INVALID_TASK; | |
5234 | ||
5235 | *voucher = NULL; | |
5236 | return KERN_SUCCESS; | |
5237 | } | |
5238 | ||
5239 | kern_return_t | |
5240 | task_set_mach_voucher( | |
5241 | task_t task, | |
5242 | ipc_voucher_t __unused voucher) | |
5243 | { | |
5244 | if (TASK_NULL == task) | |
5245 | return KERN_INVALID_TASK; | |
5246 | ||
5247 | return KERN_SUCCESS; | |
5248 | } | |
5249 | ||
5250 | kern_return_t | |
5251 | task_swap_mach_voucher( | |
5252 | task_t task, | |
5253 | ipc_voucher_t new_voucher, | |
5254 | ipc_voucher_t *in_out_old_voucher) | |
5255 | { | |
5256 | if (TASK_NULL == task) | |
5257 | return KERN_INVALID_TASK; | |
5258 | ||
5259 | *in_out_old_voucher = new_voucher; | |
5260 | return KERN_SUCCESS; | |
5261 | } | |
5262 | ||
5263 | void task_set_gpu_denied(task_t task, boolean_t denied) | |
5264 | { | |
5265 | task_lock(task); | |
5266 | ||
5267 | if (denied) { | |
5268 | task->t_flags |= TF_GPU_DENIED; | |
5269 | } else { | |
5270 | task->t_flags &= ~TF_GPU_DENIED; | |
5271 | } | |
5272 | ||
5273 | task_unlock(task); | |
5274 | } | |
5275 | ||
5276 | boolean_t task_is_gpu_denied(task_t task) | |
5277 | { | |
5278 | /* We don't need the lock to read this flag */ | |
5279 | return (task->t_flags & TF_GPU_DENIED) ? TRUE : FALSE; | |
5280 | } | |
4bd07ac2 | 5281 | |
39037602 A |
5282 | |
5283 | uint64_t get_task_memory_region_count(task_t task) | |
4bd07ac2 | 5284 | { |
39037602 A |
5285 | vm_map_t map; |
5286 | map = (task == kernel_task) ? kernel_map: task->map; | |
5287 | return((uint64_t)get_map_nentries(map)); | |
5288 | } | |
5289 | ||
5290 | static void | |
5291 | kdebug_trace_dyld_internal(uint32_t base_code, | |
5292 | struct dyld_kernel_image_info *info) | |
5293 | { | |
5294 | static_assert(sizeof(info->uuid) >= 16); | |
5295 | ||
5296 | #if defined(__LP64__) | |
5297 | uint64_t *uuid = (uint64_t *)&(info->uuid); | |
5298 | ||
5299 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, | |
5300 | KDBG_EVENTID(DBG_DYLD, DBG_DYLD_UUID, base_code), uuid[0], | |
5301 | uuid[1], info->load_addr, | |
5302 | (uint64_t)info->fsid.val[0] | ((uint64_t)info->fsid.val[1] << 32), | |
5303 | 0); | |
5304 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, | |
5305 | KDBG_EVENTID(DBG_DYLD, DBG_DYLD_UUID, base_code + 1), | |
5306 | (uint64_t)info->fsobjid.fid_objno | | |
5307 | ((uint64_t)info->fsobjid.fid_generation << 32), | |
5308 | 0, 0, 0, 0); | |
5309 | #else /* defined(__LP64__) */ | |
5310 | uint32_t *uuid = (uint32_t *)&(info->uuid); | |
5311 | ||
5312 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, | |
5313 | KDBG_EVENTID(DBG_DYLD, DBG_DYLD_UUID, base_code + 2), uuid[0], | |
5314 | uuid[1], uuid[2], uuid[3], 0); | |
5315 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, | |
5316 | KDBG_EVENTID(DBG_DYLD, DBG_DYLD_UUID, base_code + 3), | |
5317 | (uint32_t)info->load_addr, info->fsid.val[0], info->fsid.val[1], | |
5318 | info->fsobjid.fid_objno, 0); | |
5319 | KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, | |
5320 | KDBG_EVENTID(DBG_DYLD, DBG_DYLD_UUID, base_code + 4), | |
5321 | info->fsobjid.fid_generation, 0, 0, 0, 0); | |
5322 | #endif /* !defined(__LP64__) */ | |
5323 | } | |
5324 | ||
5325 | static kern_return_t | |
5326 | kdebug_trace_dyld(task_t task, uint32_t base_code, | |
5327 | vm_map_copy_t infos_copy, mach_msg_type_number_t infos_len) | |
5328 | { | |
5329 | kern_return_t kr; | |
5330 | dyld_kernel_image_info_array_t infos; | |
5331 | vm_map_offset_t map_data; | |
5332 | vm_offset_t data; | |
5333 | ||
5334 | assert(infos_copy != NULL); | |
5335 | ||
5336 | if (task == NULL || task != current_task()) { | |
5337 | return KERN_INVALID_TASK; | |
4bd07ac2 | 5338 | } |
39037602 A |
5339 | |
5340 | kr = vm_map_copyout(ipc_kernel_map, &map_data, (vm_map_copy_t)infos_copy); | |
5341 | if (kr != KERN_SUCCESS) { | |
5342 | return kr; | |
5343 | } | |
5344 | ||
5345 | infos = CAST_DOWN(dyld_kernel_image_info_array_t, map_data); | |
5346 | ||
5347 | for (mach_msg_type_number_t i = 0; i < infos_len; i++) { | |
5348 | kdebug_trace_dyld_internal(base_code, &(infos[i])); | |
5349 | } | |
5350 | ||
5351 | data = CAST_DOWN(vm_offset_t, map_data); | |
5352 | mach_vm_deallocate(ipc_kernel_map, data, infos_len * sizeof(infos[0])); | |
5353 | return KERN_SUCCESS; | |
5354 | } | |
5355 | ||
5356 | kern_return_t | |
5357 | task_register_dyld_image_infos(task_t task, | |
5358 | dyld_kernel_image_info_array_t infos_copy, | |
5359 | mach_msg_type_number_t infos_len) | |
5360 | { | |
5361 | return kdebug_trace_dyld(task, DBG_DYLD_UUID_MAP_A, | |
5362 | (vm_map_copy_t)infos_copy, infos_len); | |
5363 | } | |
5364 | ||
5365 | kern_return_t | |
5366 | task_unregister_dyld_image_infos(task_t task, | |
5367 | dyld_kernel_image_info_array_t infos_copy, | |
5368 | mach_msg_type_number_t infos_len) | |
5369 | { | |
5370 | return kdebug_trace_dyld(task, DBG_DYLD_UUID_UNMAP_A, | |
5371 | (vm_map_copy_t)infos_copy, infos_len); | |
5372 | } | |
5373 | ||
5374 | kern_return_t | |
5375 | task_get_dyld_image_infos(__unused task_t task, | |
5376 | __unused dyld_kernel_image_info_array_t * dyld_images, | |
5377 | __unused mach_msg_type_number_t * dyld_imagesCnt) | |
5378 | { | |
5379 | return KERN_NOT_SUPPORTED; | |
5380 | } | |
5381 | ||
5382 | kern_return_t | |
5383 | task_register_dyld_shared_cache_image_info(task_t task, | |
5384 | dyld_kernel_image_info_t cache_img, | |
5385 | __unused boolean_t no_cache, | |
5386 | __unused boolean_t private_cache) | |
5387 | { | |
5388 | if (task == NULL || task != current_task()) { | |
5389 | return KERN_INVALID_TASK; | |
5390 | } | |
5391 | ||
5392 | kdebug_trace_dyld_internal(DBG_DYLD_UUID_SHARED_CACHE_A, &cache_img); | |
5393 | return KERN_SUCCESS; | |
5394 | } | |
5395 | ||
5396 | kern_return_t | |
5397 | task_register_dyld_set_dyld_state(__unused task_t task, | |
5398 | __unused uint8_t dyld_state) | |
5399 | { | |
5400 | return KERN_NOT_SUPPORTED; | |
5401 | } | |
5402 | ||
5403 | kern_return_t | |
5404 | task_register_dyld_get_process_state(__unused task_t task, | |
5405 | __unused dyld_kernel_process_info_t * dyld_process_state) | |
5406 | { | |
5407 | return KERN_NOT_SUPPORTED; | |
5408 | } | |
5409 | ||
5410 | #if CONFIG_SECLUDED_MEMORY | |
5411 | int num_tasks_can_use_secluded_mem = 0; | |
5412 | ||
5413 | void | |
5414 | task_set_can_use_secluded_mem( | |
5415 | task_t task, | |
5416 | boolean_t can_use_secluded_mem) | |
5417 | { | |
5418 | if (!task->task_could_use_secluded_mem) { | |
5419 | return; | |
5420 | } | |
5421 | task_lock(task); | |
5422 | task_set_can_use_secluded_mem_locked(task, can_use_secluded_mem); | |
5423 | task_unlock(task); | |
5424 | } | |
5425 | ||
5426 | void | |
5427 | task_set_can_use_secluded_mem_locked( | |
5428 | task_t task, | |
5429 | boolean_t can_use_secluded_mem) | |
5430 | { | |
5431 | assert(task->task_could_use_secluded_mem); | |
5432 | if (can_use_secluded_mem && | |
5433 | secluded_for_apps && /* global boot-arg */ | |
5434 | !task->task_can_use_secluded_mem) { | |
5435 | assert(num_tasks_can_use_secluded_mem >= 0); | |
5436 | OSAddAtomic(+1, | |
5437 | (volatile SInt32 *)&num_tasks_can_use_secluded_mem); | |
5438 | task->task_can_use_secluded_mem = TRUE; | |
5439 | } else if (!can_use_secluded_mem && | |
5440 | task->task_can_use_secluded_mem) { | |
5441 | assert(num_tasks_can_use_secluded_mem > 0); | |
5442 | OSAddAtomic(-1, | |
5443 | (volatile SInt32 *)&num_tasks_can_use_secluded_mem); | |
5444 | task->task_can_use_secluded_mem = FALSE; | |
5445 | } | |
5446 | } | |
5447 | ||
5448 | void | |
5449 | task_set_could_use_secluded_mem( | |
5450 | task_t task, | |
5451 | boolean_t could_use_secluded_mem) | |
5452 | { | |
5453 | task->task_could_use_secluded_mem = could_use_secluded_mem; | |
5454 | } | |
5455 | ||
5456 | void | |
5457 | task_set_could_also_use_secluded_mem( | |
5458 | task_t task, | |
5459 | boolean_t could_also_use_secluded_mem) | |
5460 | { | |
5461 | task->task_could_also_use_secluded_mem = could_also_use_secluded_mem; | |
5462 | } | |
5463 | ||
5464 | boolean_t | |
5465 | task_can_use_secluded_mem( | |
5466 | task_t task) | |
5467 | { | |
5468 | if (task->task_can_use_secluded_mem) { | |
5469 | assert(task->task_could_use_secluded_mem); | |
5470 | assert(num_tasks_can_use_secluded_mem > 0); | |
5471 | return TRUE; | |
5472 | } | |
5473 | if (task->task_could_also_use_secluded_mem && | |
5474 | num_tasks_can_use_secluded_mem > 0) { | |
5475 | assert(num_tasks_can_use_secluded_mem > 0); | |
5476 | return TRUE; | |
5477 | } | |
5478 | return FALSE; | |
5479 | } | |
5480 | ||
5481 | boolean_t | |
5482 | task_could_use_secluded_mem( | |
5483 | task_t task) | |
5484 | { | |
5485 | return task->task_could_use_secluded_mem; | |
4bd07ac2 | 5486 | } |
39037602 | 5487 | #endif /* CONFIG_SECLUDED_MEMORY */ |
7e41aa88 A |
5488 | |
5489 | queue_head_t * | |
5490 | task_io_user_clients(task_t task) | |
5491 | { | |
39037602 | 5492 | return (&task->io_user_clients); |
7e41aa88 | 5493 | } |