xref: /xnu-12377.61.12/osfmk/kern/host.c (revision 4d495c6e23c53686cf65f45067f79024cf5dcee8)
1 /*
2  * Copyright (c) 2000-2009 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 /*
29  * @OSF_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  [email protected]
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  */
58 
59 /*
60  *	host.c
61  *
62  *	Non-ipc host functions.
63  */
64 
65 #include <mach/mach_types.h>
66 #include <mach/boolean.h>
67 #include <mach/host_info.h>
68 #include <mach/host_special_ports.h>
69 #include <mach/kern_return.h>
70 #include <mach/machine.h>
71 #include <mach/port.h>
72 #include <ipc/ipc_policy.h>
73 #include <mach/processor_info.h>
74 #include <mach/vm_param.h>
75 #include <mach/processor.h>
76 #include <mach/mach_host_server.h>
77 #include <mach/host_priv_server.h>
78 #include <mach/vm_map.h>
79 #include <mach/task_info.h>
80 
81 #include <machine/commpage.h>
82 #include <machine/cpu_capabilities.h>
83 
84 #include <device/device_port.h>
85 
86 #include <kern/kern_types.h>
87 #include <kern/assert.h>
88 #include <kern/kalloc.h>
89 #include <kern/ecc.h>
90 #include <kern/host.h>
91 #include <kern/host_statistics.h>
92 #include <kern/ipc_host.h>
93 #include <kern/misc_protos.h>
94 #include <kern/sched.h>
95 #include <kern/processor.h>
96 
97 #include <vm/vm_map_xnu.h>
98 #include <vm/vm_purgeable_xnu.h>
99 #include <vm/vm_pageout.h>
100 #include <vm/vm_kern_xnu.h>
101 
102 #include <IOKit/IOBSD.h> // IOTaskHasEntitlement
103 #include <IOKit/IOKitKeys.h> // DriverKit entitlement strings
104 
105 #if CONFIG_ATM
106 #include <atm/atm_internal.h>
107 #endif
108 
109 #if CONFIG_MACF
110 #include <security/mac_mach_internal.h>
111 #endif
112 
113 #if CONFIG_CSR
114 #include <sys/csr.h>
115 #endif
116 
117 #include <pexpert/pexpert.h>
118 
119 SCALABLE_COUNTER_DEFINE(vm_statistics_zero_fill_count);        /* # of zero fill pages */
120 SCALABLE_COUNTER_DEFINE(vm_statistics_reactivations);          /* # of pages reactivated */
121 SCALABLE_COUNTER_DEFINE(vm_statistics_pageins);                /* # of pageins */
122 SCALABLE_COUNTER_DEFINE(vm_statistics_pageins_aborted);        /* # of pageins aborted */
123 SCALABLE_COUNTER_DEFINE(vm_statistics_pageins_requested);      /* # of pageins requested */
124 SCALABLE_COUNTER_DEFINE(vm_statistics_pageouts);               /* # of pageouts */
125 SCALABLE_COUNTER_DEFINE(vm_statistics_faults);                 /* # of faults */
126 SCALABLE_COUNTER_DEFINE(vm_statistics_cow_faults);             /* # of copy-on-writes */
127 SCALABLE_COUNTER_DEFINE(vm_statistics_lookups);                /* object cache lookups */
128 SCALABLE_COUNTER_DEFINE(vm_statistics_hits);                   /* object cache hits */
129 SCALABLE_COUNTER_DEFINE(vm_statistics_purges);                 /* # of pages purged */
130 SCALABLE_COUNTER_DEFINE(vm_statistics_decompressions);         /* # of pages decompressed */
131 SCALABLE_COUNTER_DEFINE(vm_statistics_compressions);           /* # of pages compressed */
132 SCALABLE_COUNTER_DEFINE(vm_statistics_swapins);                /* # of pages swapped in (via compression segments) */
133 SCALABLE_COUNTER_DEFINE(vm_statistics_swapouts);               /* # of pages swapped out (via compression segments) */
134 SCALABLE_COUNTER_DEFINE(vm_statistics_total_uncompressed_pages_in_compressor); /* # of pages (uncompressed) held within the compressor. */
135 SCALABLE_COUNTER_DEFINE(vm_page_grab_count);
136 SCALABLE_COUNTER_DEFINE(vm_page_grab_count_kern);
137 SCALABLE_COUNTER_DEFINE(vm_page_grab_count_iopl);
138 SCALABLE_COUNTER_DEFINE(vm_page_grab_count_upl);
139 
140 host_data_t realhost;
141 
142 static void
get_host_vm_stats(vm_statistics64_t out)143 get_host_vm_stats(vm_statistics64_t out)
144 {
145 	out->zero_fill_count = counter_load(&vm_statistics_zero_fill_count);
146 	out->reactivations = counter_load(&vm_statistics_reactivations);
147 	out->pageins = counter_load(&vm_statistics_pageins);
148 	out->pageouts = counter_load(&vm_statistics_pageouts);
149 	out->faults = counter_load(&vm_statistics_faults);
150 	out->cow_faults = counter_load(&vm_statistics_cow_faults);
151 	out->lookups = counter_load(&vm_statistics_lookups);
152 	out->hits = counter_load(&vm_statistics_hits);
153 	out->compressions = counter_load(&vm_statistics_compressions);
154 	out->decompressions = counter_load(&vm_statistics_decompressions);
155 	out->swapins = counter_load(&vm_statistics_swapins);
156 	out->swapouts = counter_load(&vm_statistics_swapouts);
157 }
158 vm_extmod_statistics_data_t host_extmod_statistics;
159 
160 kern_return_t
host_processors(host_priv_t host_priv,processor_array_t * out_array,mach_msg_type_number_t * countp)161 host_processors(host_priv_t host_priv, processor_array_t * out_array, mach_msg_type_number_t * countp)
162 {
163 	if (host_priv == HOST_PRIV_NULL) {
164 		return KERN_INVALID_ARGUMENT;
165 	}
166 
167 	unsigned int count = processor_count;
168 	assert(count != 0);
169 
170 	static_assert(sizeof(mach_port_t) == sizeof(processor_t));
171 
172 	mach_port_array_t ports = mach_port_array_alloc(count, Z_WAITOK);
173 	if (ports == NULL) {
174 		return KERN_RESOURCE_SHORTAGE;
175 	}
176 
177 	for (unsigned int i = 0; i < count; i++) {
178 		processor_t processor = processor_array[i];
179 		assert(processor != PROCESSOR_NULL);
180 
181 		/* do the conversion that Mig should handle */
182 		ports[i].port = convert_processor_to_port(processor);
183 	}
184 
185 	*countp = count;
186 	*out_array = ports;
187 
188 	return KERN_SUCCESS;
189 }
190 
191 extern int sched_allow_NO_SMT_threads;
192 
193 kern_return_t
host_info(host_t host,host_flavor_t flavor,host_info_t info,mach_msg_type_number_t * count)194 host_info(host_t host, host_flavor_t flavor, host_info_t info, mach_msg_type_number_t * count)
195 {
196 	if (host == HOST_NULL) {
197 		return KERN_INVALID_ARGUMENT;
198 	}
199 
200 	switch (flavor) {
201 	case HOST_BASIC_INFO: {
202 		host_basic_info_t basic_info;
203 		int master_id = master_processor->cpu_id;
204 
205 		/*
206 		 *	Basic information about this host.
207 		 */
208 		if (*count < HOST_BASIC_INFO_OLD_COUNT) {
209 			return KERN_FAILURE;
210 		}
211 
212 		basic_info = (host_basic_info_t)info;
213 
214 		basic_info->memory_size = machine_info.memory_size;
215 		basic_info->cpu_type = slot_type(master_id);
216 		basic_info->cpu_subtype = slot_subtype(master_id);
217 		basic_info->max_cpus = machine_info.max_cpus;
218 #if CONFIG_SCHED_SMT
219 		if (sched_allow_NO_SMT_threads && current_task()->t_flags & TF_NO_SMT) {
220 			basic_info->avail_cpus = primary_processor_avail_count_user;
221 		} else {
222 			basic_info->avail_cpus = processor_avail_count_user;
223 		}
224 #else
225 		basic_info->avail_cpus = processor_avail_count;
226 #endif
227 
228 
229 		if (*count >= HOST_BASIC_INFO_COUNT) {
230 			basic_info->cpu_threadtype = slot_threadtype(master_id);
231 			basic_info->physical_cpu = machine_info.physical_cpu;
232 			basic_info->physical_cpu_max = machine_info.physical_cpu_max;
233 #if defined(__x86_64__)
234 			basic_info->logical_cpu = basic_info->avail_cpus;
235 #else
236 			basic_info->logical_cpu = machine_info.logical_cpu;
237 #endif
238 			basic_info->logical_cpu_max = machine_info.logical_cpu_max;
239 			basic_info->max_mem = machine_info.max_mem;
240 
241 			*count = HOST_BASIC_INFO_COUNT;
242 		} else {
243 			*count = HOST_BASIC_INFO_OLD_COUNT;
244 		}
245 
246 		return KERN_SUCCESS;
247 	}
248 
249 	case HOST_SCHED_INFO: {
250 		host_sched_info_t sched_info;
251 		uint32_t quantum_time;
252 		uint64_t quantum_ns;
253 
254 		/*
255 		 *	Return scheduler information.
256 		 */
257 		if (*count < HOST_SCHED_INFO_COUNT) {
258 			return KERN_FAILURE;
259 		}
260 
261 		sched_info = (host_sched_info_t)info;
262 
263 		quantum_time = SCHED(initial_quantum_size)(THREAD_NULL);
264 		absolutetime_to_nanoseconds(quantum_time, &quantum_ns);
265 
266 		sched_info->min_timeout = sched_info->min_quantum = (uint32_t)(quantum_ns / 1000 / 1000);
267 
268 		*count = HOST_SCHED_INFO_COUNT;
269 
270 		return KERN_SUCCESS;
271 	}
272 
273 	case HOST_RESOURCE_SIZES: {
274 		/*
275 		 * Return sizes of kernel data structures
276 		 */
277 		if (*count < HOST_RESOURCE_SIZES_COUNT) {
278 			return KERN_FAILURE;
279 		}
280 
281 		/* XXX Fail until ledgers are implemented */
282 		return KERN_INVALID_ARGUMENT;
283 	}
284 
285 	case HOST_PRIORITY_INFO: {
286 		host_priority_info_t priority_info;
287 
288 		if (*count < HOST_PRIORITY_INFO_COUNT) {
289 			return KERN_FAILURE;
290 		}
291 
292 		priority_info = (host_priority_info_t)info;
293 
294 		priority_info->kernel_priority = MINPRI_KERNEL;
295 		priority_info->system_priority = MINPRI_KERNEL;
296 		priority_info->server_priority = MINPRI_RESERVED;
297 		priority_info->user_priority = BASEPRI_DEFAULT;
298 		priority_info->depress_priority = DEPRESSPRI;
299 		priority_info->idle_priority = IDLEPRI;
300 		priority_info->minimum_priority = MINPRI_USER;
301 		priority_info->maximum_priority = MAXPRI_RESERVED;
302 
303 		*count = HOST_PRIORITY_INFO_COUNT;
304 
305 		return KERN_SUCCESS;
306 	}
307 
308 	/*
309 	 * Gestalt for various trap facilities.
310 	 */
311 	case HOST_MACH_MSG_TRAP:
312 	case HOST_SEMAPHORE_TRAPS: {
313 		*count = 0;
314 		return KERN_SUCCESS;
315 	}
316 
317 	case HOST_CAN_HAS_DEBUGGER: {
318 		host_can_has_debugger_info_t can_has_debugger_info;
319 
320 		if (*count < HOST_CAN_HAS_DEBUGGER_COUNT) {
321 			return KERN_FAILURE;
322 		}
323 
324 		can_has_debugger_info = (host_can_has_debugger_info_t)info;
325 		can_has_debugger_info->can_has_debugger = PE_i_can_has_debugger(NULL);
326 		*count = HOST_CAN_HAS_DEBUGGER_COUNT;
327 
328 		return KERN_SUCCESS;
329 	}
330 
331 	case HOST_VM_PURGABLE: {
332 		if (*count < HOST_VM_PURGABLE_COUNT) {
333 			return KERN_FAILURE;
334 		}
335 
336 		vm_purgeable_stats((vm_purgeable_info_t)info, NULL);
337 
338 		*count = HOST_VM_PURGABLE_COUNT;
339 		return KERN_SUCCESS;
340 	}
341 
342 	case HOST_DEBUG_INFO_INTERNAL: {
343 #if DEVELOPMENT || DEBUG
344 		if (*count < HOST_DEBUG_INFO_INTERNAL_COUNT) {
345 			return KERN_FAILURE;
346 		}
347 
348 		host_debug_info_internal_t debug_info = (host_debug_info_internal_t)info;
349 		bzero(debug_info, sizeof(host_debug_info_internal_data_t));
350 		*count = HOST_DEBUG_INFO_INTERNAL_COUNT;
351 
352 #if CONFIG_COALITIONS
353 		debug_info->config_coalitions = 1;
354 #endif
355 		debug_info->config_bank = 1;
356 #if CONFIG_ATM
357 		debug_info->config_atm = 1;
358 #endif
359 #if CONFIG_CSR
360 		debug_info->config_csr = 1;
361 #endif
362 		return KERN_SUCCESS;
363 #else /* DEVELOPMENT || DEBUG */
364 		return KERN_NOT_SUPPORTED;
365 #endif
366 	}
367 
368 	case HOST_PREFERRED_USER_ARCH: {
369 		host_preferred_user_arch_t user_arch_info;
370 
371 		/*
372 		 *	Basic information about this host.
373 		 */
374 		if (*count < HOST_PREFERRED_USER_ARCH_COUNT) {
375 			return KERN_FAILURE;
376 		}
377 
378 		user_arch_info = (host_preferred_user_arch_t)info;
379 
380 #if defined(PREFERRED_USER_CPU_TYPE) && defined(PREFERRED_USER_CPU_SUBTYPE)
381 		cpu_type_t preferred_cpu_type;
382 		cpu_subtype_t preferred_cpu_subtype;
383 		if (!PE_get_default("kern.preferred_cpu_type", &preferred_cpu_type, sizeof(cpu_type_t))) {
384 			preferred_cpu_type = PREFERRED_USER_CPU_TYPE;
385 		}
386 		if (!PE_get_default("kern.preferred_cpu_subtype", &preferred_cpu_subtype, sizeof(cpu_subtype_t))) {
387 			preferred_cpu_subtype = PREFERRED_USER_CPU_SUBTYPE;
388 		}
389 		user_arch_info->cpu_type    = preferred_cpu_type;
390 		user_arch_info->cpu_subtype = preferred_cpu_subtype;
391 #elif APPLEVIRTUALPLATFORM
392 		extern uint32_t force_arm64_32;
393 		if (force_arm64_32) {
394 			user_arch_info->cpu_type    = CPU_TYPE_ARM64_32;
395 			user_arch_info->cpu_subtype = CPU_SUBTYPE_ARM64_32_V8;
396 		} else {
397 			int master_id               = master_processor->cpu_id;
398 			user_arch_info->cpu_type    = slot_type(master_id);
399 			user_arch_info->cpu_subtype = slot_subtype(master_id);
400 		}
401 #else
402 		int master_id               = master_processor->cpu_id;
403 		user_arch_info->cpu_type    = slot_type(master_id);
404 		user_arch_info->cpu_subtype = slot_subtype(master_id);
405 #endif
406 
407 
408 		*count = HOST_PREFERRED_USER_ARCH_COUNT;
409 
410 		return KERN_SUCCESS;
411 	}
412 
413 	default: return KERN_INVALID_ARGUMENT;
414 	}
415 }
416 
417 kern_return_t host_statistics(host_t host, host_flavor_t flavor, host_info_t info, mach_msg_type_number_t * count);
418 
419 kern_return_t
host_statistics(host_t host,host_flavor_t flavor,host_info_t info,mach_msg_type_number_t * count)420 host_statistics(host_t host, host_flavor_t flavor, host_info_t info, mach_msg_type_number_t * count)
421 {
422 	if (host == HOST_NULL) {
423 		return KERN_INVALID_HOST;
424 	}
425 
426 	switch (flavor) {
427 	case HOST_LOAD_INFO: {
428 		host_load_info_t load_info;
429 
430 		if (*count < HOST_LOAD_INFO_COUNT) {
431 			return KERN_FAILURE;
432 		}
433 
434 		load_info = (host_load_info_t)info;
435 
436 		bcopy((char *)avenrun, (char *)load_info->avenrun, sizeof avenrun);
437 		bcopy((char *)mach_factor, (char *)load_info->mach_factor, sizeof mach_factor);
438 
439 		*count = HOST_LOAD_INFO_COUNT;
440 		return KERN_SUCCESS;
441 	}
442 
443 	case HOST_VM_INFO: {
444 		vm_statistics64_data_t host_vm_stat;
445 		vm_statistics_t stat32;
446 		mach_msg_type_number_t original_count;
447 		natural_t speculative_count = vm_page_speculative_count;
448 
449 		if (*count < HOST_VM_INFO_REV0_COUNT) {
450 			return KERN_FAILURE;
451 		}
452 
453 		get_host_vm_stats(&host_vm_stat);
454 
455 		stat32 = (vm_statistics_t)info;
456 
457 		stat32->free_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(vm_page_free_count + speculative_count);
458 		stat32->active_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(vm_page_active_count);
459 
460 		if (vm_page_local_q) {
461 			zpercpu_foreach(lq, vm_page_local_q) {
462 				stat32->active_count += VM_STATISTICS_TRUNCATE_TO_32_BIT(lq->vpl_count);
463 			}
464 		}
465 		stat32->inactive_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(vm_page_inactive_count);
466 #if !XNU_TARGET_OS_OSX
467 		stat32->wire_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(vm_page_wire_count);
468 #else /* !XNU_TARGET_OS_OSX */
469 		stat32->wire_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(vm_page_wire_count + vm_page_throttled_count + vm_lopage_free_count);
470 #endif /* !XNU_TARGET_OS_OSX */
471 		stat32->zero_fill_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.zero_fill_count);
472 		stat32->reactivations = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.reactivations);
473 		stat32->pageins = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.pageins);
474 		stat32->pageouts = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.pageouts);
475 		stat32->faults = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.faults);
476 		stat32->cow_faults = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.cow_faults);
477 		stat32->lookups = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.lookups);
478 		stat32->hits = VM_STATISTICS_TRUNCATE_TO_32_BIT(host_vm_stat.hits);
479 
480 		/*
481 		 * Fill in extra info added in later revisions of the
482 		 * vm_statistics data structure.  Fill in only what can fit
483 		 * in the data structure the caller gave us !
484 		 */
485 		original_count = *count;
486 		*count = HOST_VM_INFO_REV0_COUNT; /* rev0 already filled in */
487 		if (original_count >= HOST_VM_INFO_REV1_COUNT) {
488 			/* rev1 added "purgeable" info */
489 			stat32->purgeable_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(vm_page_purgeable_count);
490 			stat32->purges = VM_STATISTICS_TRUNCATE_TO_32_BIT(vm_page_purged_count);
491 			*count = HOST_VM_INFO_REV1_COUNT;
492 		}
493 
494 		if (original_count >= HOST_VM_INFO_REV2_COUNT) {
495 			/* rev2 added "speculative" info */
496 			stat32->speculative_count = VM_STATISTICS_TRUNCATE_TO_32_BIT(speculative_count);
497 			*count = HOST_VM_INFO_REV2_COUNT;
498 		}
499 
500 		/* rev3 changed some of the fields to be 64-bit*/
501 
502 		return KERN_SUCCESS;
503 	}
504 
505 	case HOST_CPU_LOAD_INFO: {
506 		host_cpu_load_info_t cpu_load_info;
507 
508 		if (*count < HOST_CPU_LOAD_INFO_COUNT) {
509 			return KERN_FAILURE;
510 		}
511 
512 #define GET_TICKS_VALUE(state, ticks)                                                      \
513 	MACRO_BEGIN cpu_load_info->cpu_ticks[(state)] += (uint32_t)(ticks / hz_tick_interval); \
514 	MACRO_END
515 #define GET_TICKS_VALUE_FROM_TIMER(processor, state, timer)                            \
516 	MACRO_BEGIN GET_TICKS_VALUE(state, timer_grab(&(processor)->timer)); \
517 	MACRO_END
518 
519 		cpu_load_info = (host_cpu_load_info_t)info;
520 		cpu_load_info->cpu_ticks[CPU_STATE_USER] = 0;
521 		cpu_load_info->cpu_ticks[CPU_STATE_SYSTEM] = 0;
522 		cpu_load_info->cpu_ticks[CPU_STATE_IDLE] = 0;
523 		cpu_load_info->cpu_ticks[CPU_STATE_NICE] = 0;
524 
525 		simple_lock(&processor_list_lock, LCK_GRP_NULL);
526 
527 		unsigned int pcount = processor_count;
528 
529 		for (unsigned int i = 0; i < pcount; i++) {
530 			processor_t processor = processor_array[i];
531 			assert(processor != PROCESSOR_NULL);
532 			processor_cpu_load_info(processor, cpu_load_info->cpu_ticks);
533 		}
534 		simple_unlock(&processor_list_lock);
535 
536 		*count = HOST_CPU_LOAD_INFO_COUNT;
537 
538 		return KERN_SUCCESS;
539 	}
540 
541 	case HOST_EXPIRED_TASK_INFO: {
542 		if (*count < TASK_POWER_INFO_COUNT) {
543 			return KERN_FAILURE;
544 		}
545 
546 		task_power_info_t tinfo1 = (task_power_info_t)info;
547 		task_power_info_v2_t tinfo2 = (task_power_info_v2_t)info;
548 
549 		tinfo1->task_interrupt_wakeups = dead_task_statistics.task_interrupt_wakeups;
550 		tinfo1->task_platform_idle_wakeups = dead_task_statistics.task_platform_idle_wakeups;
551 
552 		tinfo1->task_timer_wakeups_bin_1 = dead_task_statistics.task_timer_wakeups_bin_1;
553 
554 		tinfo1->task_timer_wakeups_bin_2 = dead_task_statistics.task_timer_wakeups_bin_2;
555 
556 		tinfo1->total_user = dead_task_statistics.total_user_time;
557 		tinfo1->total_system = dead_task_statistics.total_system_time;
558 		if (*count < TASK_POWER_INFO_V2_COUNT) {
559 			*count = TASK_POWER_INFO_COUNT;
560 		} else if (*count >= TASK_POWER_INFO_V2_COUNT) {
561 			tinfo2->gpu_energy.task_gpu_utilisation = dead_task_statistics.task_gpu_ns;
562 #if defined(__arm64__)
563 			tinfo2->task_energy = dead_task_statistics.task_energy;
564 			tinfo2->task_ptime = dead_task_statistics.total_ptime;
565 			tinfo2->task_pset_switches = dead_task_statistics.total_pset_switches;
566 #endif
567 			*count = TASK_POWER_INFO_V2_COUNT;
568 		}
569 
570 		return KERN_SUCCESS;
571 	}
572 	default: return KERN_INVALID_ARGUMENT;
573 	}
574 }
575 
576 extern uint32_t c_segment_pages_compressed;
577 
578 #define HOST_STATISTICS_TIME_WINDOW 1 /* seconds */
579 #define HOST_STATISTICS_MAX_REQUESTS 10 /* maximum number of requests per window */
580 #define HOST_STATISTICS_MIN_REQUESTS 2 /* minimum number of requests per window */
581 
582 uint64_t host_statistics_time_window;
583 
584 static LCK_GRP_DECLARE(host_statistics_lck_grp, "host_statistics");
585 static LCK_MTX_DECLARE(host_statistics_lck, &host_statistics_lck_grp);
586 
587 #define HOST_VM_INFO64_REV0             0
588 #define HOST_VM_INFO64_REV1             1
589 #define HOST_EXTMOD_INFO64_REV0         2
590 #define HOST_LOAD_INFO_REV0             3
591 #define HOST_VM_INFO_REV0               4
592 #define HOST_VM_INFO_REV1               5
593 #define HOST_VM_INFO_REV2               6
594 #define HOST_CPU_LOAD_INFO_REV0         7
595 #define HOST_EXPIRED_TASK_INFO_REV0     8
596 #define HOST_EXPIRED_TASK_INFO_REV1     9
597 #define HOST_VM_COMPRESSOR_Q_LEN_REV0   10
598 #define HOST_VM_INFO64_REV2             11
599 #define NUM_HOST_INFO_DATA_TYPES        12
600 
601 static vm_statistics64_data_t host_vm_info64_rev0 = {};
602 static vm_statistics64_data_t host_vm_info64_rev1 = {};
603 static vm_statistics64_data_t host_vm_info64_rev2 = {};
604 static vm_extmod_statistics_data_t host_extmod_info64 = {};
605 static host_load_info_data_t host_load_info = {};
606 static vm_statistics_data_t host_vm_info_rev0 = {};
607 static vm_statistics_data_t host_vm_info_rev1 = {};
608 static vm_statistics_data_t host_vm_info_rev2 = {};
609 static host_cpu_load_info_data_t host_cpu_load_info = {};
610 static task_power_info_data_t host_expired_task_info = {};
611 static task_power_info_v2_data_t host_expired_task_info2 = {};
612 static vm_compressor_q_lens_data_t host_vm_compressor_q_lens = {};
613 
614 struct host_stats_cache {
615 	uint64_t last_access;
616 	uint64_t current_requests;
617 	uint64_t max_requests;
618 	uintptr_t data;
619 	mach_msg_type_number_t count; //NOTE count is in sizeof(integer_t)
620 };
621 
622 static struct host_stats_cache g_host_stats_cache[NUM_HOST_INFO_DATA_TYPES] = {
623 	[HOST_VM_INFO64_REV0] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_vm_info64_rev0, .count = HOST_VM_INFO64_REV0_COUNT },
624 	[HOST_VM_INFO64_REV1] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_vm_info64_rev1, .count = HOST_VM_INFO64_REV1_COUNT },
625 	[HOST_EXTMOD_INFO64_REV0] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_extmod_info64, .count = HOST_EXTMOD_INFO64_COUNT },
626 	[HOST_LOAD_INFO_REV0] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_load_info, .count = HOST_LOAD_INFO_COUNT },
627 	[HOST_VM_INFO_REV0] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_vm_info_rev0, .count = HOST_VM_INFO_REV0_COUNT },
628 	[HOST_VM_INFO_REV1] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_vm_info_rev1, .count = HOST_VM_INFO_REV1_COUNT },
629 	[HOST_VM_INFO_REV2] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_vm_info_rev2, .count = HOST_VM_INFO_REV2_COUNT },
630 	[HOST_CPU_LOAD_INFO_REV0] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_cpu_load_info, .count = HOST_CPU_LOAD_INFO_COUNT },
631 	[HOST_EXPIRED_TASK_INFO_REV0] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_expired_task_info, .count = TASK_POWER_INFO_COUNT },
632 	[HOST_EXPIRED_TASK_INFO_REV1] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_expired_task_info2, .count = TASK_POWER_INFO_V2_COUNT},
633 	[HOST_VM_COMPRESSOR_Q_LEN_REV0] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_vm_compressor_q_lens, .count = VM_COMPRESSOR_Q_LENS_COUNT},
634 	[HOST_VM_INFO64_REV2] = { .last_access = 0, .current_requests = 0, .max_requests = 0, .data = (uintptr_t)&host_vm_info64_rev2, .count = HOST_VM_INFO64_REV2_COUNT },
635 };
636 
637 
638 void
host_statistics_init(void)639 host_statistics_init(void)
640 {
641 	nanoseconds_to_absolutetime((HOST_STATISTICS_TIME_WINDOW * NSEC_PER_SEC), &host_statistics_time_window);
642 }
643 
644 static void
cache_host_statistics(int index,host_info64_t info)645 cache_host_statistics(int index, host_info64_t info)
646 {
647 	if (index < 0 || index >= NUM_HOST_INFO_DATA_TYPES) {
648 		return;
649 	}
650 
651 	if (task_get_platform_binary(current_task())) {
652 		return;
653 	}
654 
655 	memcpy((void *)g_host_stats_cache[index].data, info, g_host_stats_cache[index].count * sizeof(integer_t));
656 	return;
657 }
658 
659 static void
get_cached_info(int index,host_info64_t info,mach_msg_type_number_t * count)660 get_cached_info(int index, host_info64_t info, mach_msg_type_number_t* count)
661 {
662 	if (index < 0 || index >= NUM_HOST_INFO_DATA_TYPES) {
663 		*count = 0;
664 		return;
665 	}
666 
667 	*count = g_host_stats_cache[index].count;
668 	memcpy(info, (void *)g_host_stats_cache[index].data, g_host_stats_cache[index].count * sizeof(integer_t));
669 }
670 
671 static int
get_host_info_data_index(bool is_stat64,host_flavor_t flavor,mach_msg_type_number_t * count,kern_return_t * ret)672 get_host_info_data_index(bool is_stat64, host_flavor_t flavor, mach_msg_type_number_t* count, kern_return_t* ret)
673 {
674 	switch (flavor) {
675 	case HOST_VM_INFO64:
676 		if (!is_stat64) {
677 			*ret = KERN_INVALID_ARGUMENT;
678 			return -1;
679 		}
680 		if (*count < HOST_VM_INFO64_REV0_COUNT) {
681 			*ret = KERN_FAILURE;
682 			return -1;
683 		}
684 		if (*count >= HOST_VM_INFO64_REV2_COUNT) {
685 			return HOST_VM_INFO64_REV2;
686 		}
687 		if (*count >= HOST_VM_INFO64_REV1_COUNT) {
688 			return HOST_VM_INFO64_REV1;
689 		}
690 		return HOST_VM_INFO64_REV0;
691 
692 	case HOST_EXTMOD_INFO64:
693 		if (!is_stat64) {
694 			*ret = KERN_INVALID_ARGUMENT;
695 			return -1;
696 		}
697 		if (*count < HOST_EXTMOD_INFO64_COUNT) {
698 			*ret = KERN_FAILURE;
699 			return -1;
700 		}
701 		return HOST_EXTMOD_INFO64_REV0;
702 
703 	case HOST_LOAD_INFO:
704 		if (*count < HOST_LOAD_INFO_COUNT) {
705 			*ret = KERN_FAILURE;
706 			return -1;
707 		}
708 		return HOST_LOAD_INFO_REV0;
709 
710 	case HOST_VM_INFO:
711 		if (*count < HOST_VM_INFO_REV0_COUNT) {
712 			*ret = KERN_FAILURE;
713 			return -1;
714 		}
715 		if (*count >= HOST_VM_INFO_REV2_COUNT) {
716 			return HOST_VM_INFO_REV2;
717 		}
718 		if (*count >= HOST_VM_INFO_REV1_COUNT) {
719 			return HOST_VM_INFO_REV1;
720 		}
721 		return HOST_VM_INFO_REV0;
722 
723 	case HOST_CPU_LOAD_INFO:
724 		if (*count < HOST_CPU_LOAD_INFO_COUNT) {
725 			*ret = KERN_FAILURE;
726 			return -1;
727 		}
728 		return HOST_CPU_LOAD_INFO_REV0;
729 
730 	case HOST_EXPIRED_TASK_INFO:
731 		if (*count < TASK_POWER_INFO_COUNT) {
732 			*ret = KERN_FAILURE;
733 			return -1;
734 		}
735 		if (*count >= TASK_POWER_INFO_V2_COUNT) {
736 			return HOST_EXPIRED_TASK_INFO_REV1;
737 		}
738 		return HOST_EXPIRED_TASK_INFO_REV0;
739 
740 	case HOST_VM_COMPRESSOR_Q_LENS:
741 		if (*count < VM_COMPRESSOR_Q_LENS_COUNT) {
742 			*ret = KERN_FAILURE;
743 			return -1;
744 		}
745 		return HOST_VM_COMPRESSOR_Q_LEN_REV0;
746 
747 	default:
748 		*ret = KERN_INVALID_ARGUMENT;
749 		return -1;
750 	}
751 }
752 
753 static bool
rate_limit_host_statistics(bool is_stat64,host_flavor_t flavor,host_info64_t info,mach_msg_type_number_t * count,kern_return_t * ret,int * pindex)754 rate_limit_host_statistics(bool is_stat64, host_flavor_t flavor, host_info64_t info, mach_msg_type_number_t* count, kern_return_t* ret, int *pindex)
755 {
756 	task_t task = current_task();
757 
758 	assert(task != kernel_task);
759 
760 	*ret = KERN_SUCCESS;
761 	*pindex = -1;
762 
763 	/* Access control only for third party applications */
764 	if (task_get_platform_binary(task)) {
765 		return FALSE;
766 	}
767 
768 	/* Rate limit to HOST_STATISTICS_MAX_REQUESTS queries for each HOST_STATISTICS_TIME_WINDOW window of time */
769 	bool rate_limited = FALSE;
770 	bool set_last_access = TRUE;
771 
772 	/* there is a cache for every flavor */
773 	int index = get_host_info_data_index(is_stat64, flavor, count, ret);
774 	if (index == -1) {
775 		goto out;
776 	}
777 
778 	*pindex = index;
779 	lck_mtx_lock(&host_statistics_lck);
780 	if (g_host_stats_cache[index].last_access > mach_continuous_time() - host_statistics_time_window) {
781 		set_last_access = FALSE;
782 		if (g_host_stats_cache[index].current_requests++ >= g_host_stats_cache[index].max_requests) {
783 			rate_limited = TRUE;
784 			get_cached_info(index, info, count);
785 		}
786 	}
787 	if (set_last_access) {
788 		g_host_stats_cache[index].current_requests = 1;
789 		/*
790 		 * select a random number of requests (included between HOST_STATISTICS_MIN_REQUESTS and HOST_STATISTICS_MAX_REQUESTS)
791 		 * to let query host_statistics.
792 		 * In this way it is not possible to infer looking at when the a cached copy changes if host_statistics was called on
793 		 * the provious window.
794 		 */
795 		g_host_stats_cache[index].max_requests = (mach_absolute_time() % (HOST_STATISTICS_MAX_REQUESTS - HOST_STATISTICS_MIN_REQUESTS + 1)) + HOST_STATISTICS_MIN_REQUESTS;
796 		g_host_stats_cache[index].last_access = mach_continuous_time();
797 	}
798 	lck_mtx_unlock(&host_statistics_lck);
799 out:
800 	return rate_limited;
801 }
802 
803 kern_return_t
vm_stats(void * info,unsigned int * count)804 vm_stats(void *info, unsigned int *count)
805 {
806 	vm_statistics64_data_t host_vm_stat;
807 	mach_msg_type_number_t original_count;
808 	unsigned int local_q_internal_count;
809 	unsigned int local_q_external_count;
810 	natural_t speculative_count = vm_page_speculative_count;
811 	natural_t throttled_count = vm_page_throttled_count;
812 
813 	if (*count < HOST_VM_INFO64_REV0_COUNT) {
814 		return KERN_FAILURE;
815 	}
816 	get_host_vm_stats(&host_vm_stat);
817 
818 	vm_statistics64_t stat = (vm_statistics64_t)info;
819 
820 	stat->free_count = vm_page_free_count + speculative_count;
821 	stat->active_count = vm_page_active_count;
822 
823 	local_q_internal_count = 0;
824 	local_q_external_count = 0;
825 	if (vm_page_local_q) {
826 		zpercpu_foreach(lq, vm_page_local_q) {
827 			stat->active_count += lq->vpl_count;
828 			local_q_internal_count += lq->vpl_internal_count;
829 			local_q_external_count += lq->vpl_external_count;
830 		}
831 	}
832 	stat->inactive_count = vm_page_inactive_count;
833 #if !XNU_TARGET_OS_OSX
834 	stat->wire_count = vm_page_wire_count;
835 #else /* !XNU_TARGET_OS_OSX */
836 	stat->wire_count = vm_page_wire_count + throttled_count + vm_lopage_free_count;
837 #endif /* !XNU_TARGET_OS_OSX */
838 	stat->zero_fill_count = host_vm_stat.zero_fill_count;
839 	stat->reactivations = host_vm_stat.reactivations;
840 	stat->pageins = host_vm_stat.pageins;
841 	stat->pageouts = host_vm_stat.pageouts;
842 	stat->faults = host_vm_stat.faults;
843 	stat->cow_faults = host_vm_stat.cow_faults;
844 	stat->lookups = host_vm_stat.lookups;
845 	stat->hits = host_vm_stat.hits;
846 
847 	stat->purgeable_count = vm_page_purgeable_count;
848 	stat->purges = vm_page_purged_count;
849 
850 	stat->speculative_count = speculative_count;
851 
852 	/*
853 	 * Fill in extra info added in later revisions of the
854 	 * vm_statistics data structure.  Fill in only what can fit
855 	 * in the data structure the caller gave us !
856 	 */
857 	original_count = *count;
858 	*count = HOST_VM_INFO64_REV0_COUNT; /* rev0 already filled in */
859 	if (original_count >= HOST_VM_INFO64_REV1_COUNT) {
860 		/* rev1 added "throttled count" */
861 		stat->throttled_count = throttled_count;
862 		/* rev1 added "compression" info */
863 		stat->compressor_page_count = VM_PAGE_COMPRESSOR_COUNT;
864 		stat->compressions = host_vm_stat.compressions;
865 		stat->decompressions = host_vm_stat.decompressions;
866 		stat->swapins = host_vm_stat.swapins;
867 		stat->swapouts = host_vm_stat.swapouts;
868 		/* rev1 added:
869 		 * "external page count"
870 		 * "anonymous page count"
871 		 * "total # of pages (uncompressed) held in the compressor"
872 		 */
873 		stat->external_page_count = (vm_page_pageable_external_count + local_q_external_count);
874 		stat->internal_page_count = (vm_page_pageable_internal_count + local_q_internal_count);
875 		stat->total_uncompressed_pages_in_compressor = c_segment_pages_compressed;
876 		*count = HOST_VM_INFO64_REV1_COUNT;
877 	}
878 	if (original_count >= HOST_VM_INFO64_REV2_COUNT) {
879 		stat->swapped_count = os_atomic_load(&vm_page_swapped_count, relaxed);
880 		*count = HOST_VM_INFO64_REV2_COUNT;
881 	}
882 
883 	return KERN_SUCCESS;
884 }
885 
886 #if DEVELOPMENT || DEBUG
887 extern uint32_t        c_segment_count;
888 extern uint32_t        c_age_count;
889 extern uint32_t        c_early_swappedin_count, c_regular_swappedin_count, c_late_swappedin_count;
890 extern uint32_t        c_early_swapout_count, c_regular_swapout_count, c_late_swapout_count;
891 extern uint32_t        c_swapio_count;
892 extern uint32_t        c_swappedout_count;
893 extern uint32_t        c_swappedout_sparse_count;
894 extern uint32_t        c_major_count;
895 extern uint32_t        c_filling_count;
896 extern uint32_t        c_empty_count;
897 extern uint32_t        c_bad_count;
898 extern uint32_t        c_minor_count;
899 extern uint32_t        c_segments_available;
900 
901 static kern_return_t
vm_compressor_queue_lens(void * info,unsigned int * count)902 vm_compressor_queue_lens(void *info, unsigned int *count)
903 {
904 	if (*count < VM_COMPRESSOR_Q_LENS_COUNT) {
905 		return KERN_NO_SPACE;
906 	}
907 
908 	struct vm_compressor_q_lens *qc = (struct vm_compressor_q_lens *)info;
909 	qc->qcc_segments_available = c_segments_available;
910 	qc->qcc_segment_count = c_segment_count;
911 	qc->qcc_age_count = c_age_count;
912 	qc->qcc_early_swappedin_count = c_early_swappedin_count;
913 	qc->qcc_regular_swappedin_count = c_regular_swappedin_count;
914 	qc->qcc_late_swappedin_count = c_late_swappedin_count;
915 	qc->qcc_early_swapout_count = c_early_swapout_count;
916 	qc->qcc_regular_swapout_count = c_regular_swapout_count;
917 	qc->qcc_late_swapout_count = c_late_swapout_count;
918 	qc->qcc_swapio_count = c_swapio_count;
919 	qc->qcc_swappedout_count = c_swappedout_count;
920 	qc->qcc_swappedout_sparse_count = c_swappedout_sparse_count;
921 	qc->qcc_major_count = c_major_count;
922 	qc->qcc_filling_count = c_filling_count;
923 	qc->qcc_empty_count = c_empty_count;
924 	qc->qcc_bad_count = c_bad_count;
925 	qc->qcc_minor_count = c_minor_count;
926 
927 	*count = VM_COMPRESSOR_Q_LENS_COUNT;
928 
929 	return KERN_SUCCESS;
930 }
931 
932 #endif /* DEVELOPMENT || DEBUG */
933 
934 kern_return_t host_statistics64(host_t host, host_flavor_t flavor, host_info_t info, mach_msg_type_number_t * count);
935 
936 kern_return_t
host_statistics64(host_t host,host_flavor_t flavor,host_info64_t info,mach_msg_type_number_t * count)937 host_statistics64(host_t host, host_flavor_t flavor, host_info64_t info, mach_msg_type_number_t * count)
938 {
939 	if (host == HOST_NULL) {
940 		return KERN_INVALID_HOST;
941 	}
942 
943 	switch (flavor) {
944 	case HOST_VM_INFO64: /* We were asked to get vm_statistics64 */
945 		return vm_stats(info, count);
946 
947 	case HOST_EXTMOD_INFO64: /* We were asked to get vm_statistics64 */
948 	{
949 		vm_extmod_statistics_t out_extmod_statistics;
950 
951 		if (*count < HOST_EXTMOD_INFO64_COUNT) {
952 			return KERN_FAILURE;
953 		}
954 
955 		out_extmod_statistics = (vm_extmod_statistics_t)info;
956 		*out_extmod_statistics = host_extmod_statistics;
957 
958 		*count = HOST_EXTMOD_INFO64_COUNT;
959 
960 		return KERN_SUCCESS;
961 	}
962 
963 	case HOST_VM_COMPRESSOR_Q_LENS:
964 #if DEVELOPMENT || DEBUG
965 		return vm_compressor_queue_lens(info, count);
966 #else
967 		return KERN_NOT_SUPPORTED;
968 #endif
969 
970 	default: /* If we didn't recognize the flavor, send to host_statistics */
971 		return host_statistics(host, flavor, (host_info_t)info, count);
972 	}
973 }
974 
975 kern_return_t
host_statistics64_from_user(host_t host,host_flavor_t flavor,host_info64_t info,mach_msg_type_number_t * count)976 host_statistics64_from_user(host_t host, host_flavor_t flavor, host_info64_t info, mach_msg_type_number_t * count)
977 {
978 	kern_return_t ret = KERN_SUCCESS;
979 	int index;
980 
981 	if (host == HOST_NULL) {
982 		return KERN_INVALID_HOST;
983 	}
984 
985 	if (rate_limit_host_statistics(TRUE, flavor, info, count, &ret, &index)) {
986 		return ret;
987 	}
988 
989 	if (ret != KERN_SUCCESS) {
990 		return ret;
991 	}
992 
993 	ret = host_statistics64(host, flavor, info, count);
994 
995 	if (ret == KERN_SUCCESS) {
996 		cache_host_statistics(index, info);
997 	}
998 
999 	return ret;
1000 }
1001 
1002 kern_return_t
host_statistics_from_user(host_t host,host_flavor_t flavor,host_info64_t info,mach_msg_type_number_t * count)1003 host_statistics_from_user(host_t host, host_flavor_t flavor, host_info64_t info, mach_msg_type_number_t * count)
1004 {
1005 	kern_return_t ret = KERN_SUCCESS;
1006 	int index;
1007 
1008 	if (host == HOST_NULL) {
1009 		return KERN_INVALID_HOST;
1010 	}
1011 
1012 	if (rate_limit_host_statistics(FALSE, flavor, info, count, &ret, &index)) {
1013 		return ret;
1014 	}
1015 
1016 	if (ret != KERN_SUCCESS) {
1017 		return ret;
1018 	}
1019 
1020 	ret = host_statistics(host, flavor, info, count);
1021 
1022 	if (ret == KERN_SUCCESS) {
1023 		cache_host_statistics(index, info);
1024 	}
1025 
1026 	return ret;
1027 }
1028 
1029 /*
1030  * Get host statistics that require privilege.
1031  * None for now, just call the un-privileged version.
1032  */
1033 kern_return_t
host_priv_statistics(host_priv_t host_priv,host_flavor_t flavor,host_info_t info,mach_msg_type_number_t * count)1034 host_priv_statistics(host_priv_t host_priv, host_flavor_t flavor, host_info_t info, mach_msg_type_number_t * count)
1035 {
1036 	return host_statistics((host_t)host_priv, flavor, info, count);
1037 }
1038 
1039 kern_return_t
set_sched_stats_active(boolean_t active)1040 set_sched_stats_active(boolean_t active)
1041 {
1042 	sched_stats_active = active;
1043 	return KERN_SUCCESS;
1044 }
1045 
1046 kern_return_t
get_sched_statistics(struct _processor_statistics_np * out,uint32_t * count)1047 get_sched_statistics(struct _processor_statistics_np * out, uint32_t * count)
1048 {
1049 	uint32_t pos = 0;
1050 
1051 	if (!sched_stats_active) {
1052 		return KERN_FAILURE;
1053 	}
1054 
1055 	percpu_foreach_base(pcpu_base) {
1056 		struct sched_statistics stats;
1057 		processor_t processor;
1058 
1059 		pos += sizeof(struct _processor_statistics_np);
1060 		if (pos > *count) {
1061 			return KERN_FAILURE;
1062 		}
1063 
1064 		stats = *PERCPU_GET_WITH_BASE(pcpu_base, sched_stats);
1065 		processor = PERCPU_GET_WITH_BASE(pcpu_base, processor);
1066 
1067 		out->ps_cpuid = processor->cpu_id;
1068 		out->ps_csw_count = stats.csw_count;
1069 		out->ps_preempt_count = stats.preempt_count;
1070 		out->ps_preempted_rt_count = stats.preempted_rt_count;
1071 		out->ps_preempted_by_rt_count = stats.preempted_by_rt_count;
1072 		out->ps_rt_sched_count = stats.rt_sched_count;
1073 		out->ps_interrupt_count = stats.interrupt_count;
1074 		out->ps_ipi_count = stats.ipi_count;
1075 		out->ps_timer_pop_count = stats.timer_pop_count;
1076 		out->ps_runq_count_sum = SCHED(processor_runq_stats_count_sum)(processor);
1077 		out->ps_idle_transitions = stats.idle_transitions;
1078 		out->ps_quantum_timer_expirations = stats.quantum_timer_expirations;
1079 
1080 		out++;
1081 	}
1082 
1083 	/* And include RT Queue information */
1084 	pos += sizeof(struct _processor_statistics_np);
1085 	if (pos > *count) {
1086 		return KERN_FAILURE;
1087 	}
1088 
1089 	bzero(out, sizeof(*out));
1090 	out->ps_cpuid = (-1);
1091 	out->ps_runq_count_sum = SCHED(rt_runq_count_sum)();
1092 	out++;
1093 
1094 	*count = pos;
1095 
1096 	return KERN_SUCCESS;
1097 }
1098 
1099 kern_return_t
host_page_size(host_t host,vm_size_t * out_page_size)1100 host_page_size(host_t host, vm_size_t * out_page_size)
1101 {
1102 	if (host == HOST_NULL) {
1103 		return KERN_INVALID_ARGUMENT;
1104 	}
1105 
1106 	*out_page_size = PAGE_SIZE;
1107 
1108 	return KERN_SUCCESS;
1109 }
1110 
1111 /*
1112  *	Return kernel version string (more than you ever
1113  *	wanted to know about what version of the kernel this is).
1114  */
1115 extern char version[];
1116 
1117 kern_return_t
host_kernel_version(host_t host,kernel_version_t out_version)1118 host_kernel_version(host_t host, kernel_version_t out_version)
1119 {
1120 	if (host == HOST_NULL) {
1121 		return KERN_INVALID_ARGUMENT;
1122 	}
1123 
1124 	(void)strncpy(out_version, version, sizeof(kernel_version_t));
1125 
1126 	return KERN_SUCCESS;
1127 }
1128 
1129 /*
1130  *	host_processor_sets:
1131  *
1132  *	List all processor sets on the host.
1133  */
1134 kern_return_t
host_processor_sets(host_priv_t host_priv,processor_set_name_array_t * pset_list,mach_msg_type_number_t * count)1135 host_processor_sets(host_priv_t host_priv, processor_set_name_array_t * pset_list, mach_msg_type_number_t * count)
1136 {
1137 	mach_port_array_t ports;
1138 
1139 	if (host_priv == HOST_PRIV_NULL) {
1140 		return KERN_INVALID_ARGUMENT;
1141 	}
1142 
1143 	/*
1144 	 *	Allocate memory.  Can be pageable because it won't be
1145 	 *	touched while holding a lock.
1146 	 */
1147 
1148 	ports = mach_port_array_alloc(1, Z_WAITOK | Z_NOFAIL);
1149 
1150 	/* do the conversion that Mig should handle */
1151 	ports[0].port = convert_pset_name_to_port(&pset0);
1152 
1153 	*pset_list = ports;
1154 	*count = 1;
1155 
1156 	return KERN_SUCCESS;
1157 }
1158 
1159 /*
1160  *	host_processor_set_priv:
1161  *
1162  *	Return control port for given processor set.
1163  */
1164 kern_return_t
host_processor_set_priv(host_priv_t host_priv,processor_set_t pset_name,processor_set_t * pset)1165 host_processor_set_priv(host_priv_t host_priv, processor_set_t pset_name, processor_set_t * pset)
1166 {
1167 	if (host_priv == HOST_PRIV_NULL || pset_name == PROCESSOR_SET_NULL) {
1168 		*pset = PROCESSOR_SET_NULL;
1169 
1170 		return KERN_INVALID_ARGUMENT;
1171 	}
1172 
1173 	*pset = pset_name;
1174 
1175 	return KERN_SUCCESS;
1176 }
1177 
1178 /*
1179  *	host_processor_info
1180  *
1181  *	Return info about the processors on this host.  It will return
1182  *	the number of processors, and the specific type of info requested
1183  *	in an OOL array.
1184  */
1185 kern_return_t
host_processor_info(host_t host,processor_flavor_t flavor,natural_t * out_pcount,processor_info_array_t * out_array,mach_msg_type_number_t * out_array_count)1186 host_processor_info(host_t host,
1187     processor_flavor_t flavor,
1188     natural_t * out_pcount,
1189     processor_info_array_t * out_array,
1190     mach_msg_type_number_t * out_array_count)
1191 {
1192 	kern_return_t result;
1193 	host_t thost;
1194 	processor_info_t info;
1195 	unsigned int icount;
1196 	unsigned int pcount;
1197 	vm_offset_t addr;
1198 	vm_size_t size, needed;
1199 	vm_map_copy_t copy;
1200 
1201 	if (host == HOST_NULL) {
1202 		return KERN_INVALID_ARGUMENT;
1203 	}
1204 
1205 	result = processor_info_count(flavor, &icount);
1206 	if (result != KERN_SUCCESS) {
1207 		return result;
1208 	}
1209 
1210 	pcount = processor_count;
1211 	assert(pcount != 0);
1212 
1213 	needed = pcount * icount * sizeof(natural_t);
1214 	size = vm_map_round_page(needed, VM_MAP_PAGE_MASK(ipc_kernel_map));
1215 	result = kmem_alloc(ipc_kernel_map, &addr, size, KMA_DATA, VM_KERN_MEMORY_IPC);
1216 	if (result != KERN_SUCCESS) {
1217 		return KERN_RESOURCE_SHORTAGE;
1218 	}
1219 
1220 	info = (processor_info_t)addr;
1221 
1222 	for (unsigned int i = 0; i < pcount; i++) {
1223 		processor_t processor = processor_array[i];
1224 		assert(processor != PROCESSOR_NULL);
1225 
1226 		unsigned int tcount = icount;
1227 
1228 		result = processor_info(processor, flavor, &thost, info, &tcount);
1229 		if (result != KERN_SUCCESS) {
1230 			kmem_free(ipc_kernel_map, addr, size);
1231 			return result;
1232 		}
1233 		info += icount;
1234 	}
1235 
1236 	if (size != needed) {
1237 		bzero((char *)addr + needed, size - needed);
1238 	}
1239 
1240 	result = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(addr, VM_MAP_PAGE_MASK(ipc_kernel_map)),
1241 	    vm_map_round_page(addr + size, VM_MAP_PAGE_MASK(ipc_kernel_map)), FALSE);
1242 	assert(result == KERN_SUCCESS);
1243 	result = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)addr, (vm_map_size_t)needed, TRUE, &copy);
1244 	assert(result == KERN_SUCCESS);
1245 
1246 	*out_pcount = pcount;
1247 	*out_array = (processor_info_array_t)copy;
1248 	*out_array_count = pcount * icount;
1249 
1250 	return KERN_SUCCESS;
1251 }
1252 
1253 static bool
is_valid_host_special_port(int id)1254 is_valid_host_special_port(int id)
1255 {
1256 	return (id <= HOST_MAX_SPECIAL_PORT) &&
1257 	       (id >= HOST_MIN_SPECIAL_PORT) &&
1258 	       ((id <= HOST_LAST_SPECIAL_KERNEL_PORT) || (id > HOST_MAX_SPECIAL_KERNEL_PORT));
1259 }
1260 
1261 /*
1262  *      Kernel interface for setting a special port.
1263  */
1264 kern_return_t
kernel_set_special_port(host_priv_t host_priv,int id,ipc_port_t port)1265 kernel_set_special_port(host_priv_t host_priv, int id, ipc_port_t port)
1266 {
1267 	ipc_port_t old_port;
1268 
1269 	if (!is_valid_host_special_port(id)) {
1270 		panic("attempted to set invalid special port %d", id);
1271 	}
1272 
1273 	if (id == HOST_NODE_PORT) {
1274 		return KERN_NOT_SUPPORTED;
1275 	}
1276 
1277 	host_lock(host_priv);
1278 	old_port = host_priv->special[id];
1279 	host_priv->special[id] = port;
1280 	host_unlock(host_priv);
1281 
1282 	if (IP_VALID(old_port)) {
1283 		ipc_port_release_send(old_port);
1284 	}
1285 
1286 
1287 	return KERN_SUCCESS;
1288 }
1289 
1290 /*
1291  *      Kernel interface for retrieving a special port.
1292  */
1293 kern_return_t
kernel_get_special_port(host_priv_t host_priv,int id,ipc_port_t * portp)1294 kernel_get_special_port(host_priv_t host_priv, int id, ipc_port_t * portp)
1295 {
1296 	if (!is_valid_host_special_port(id)) {
1297 		panic("attempted to get invalid special port %d", id);
1298 	}
1299 
1300 	host_lock(host_priv);
1301 	*portp = host_priv->special[id];
1302 	host_unlock(host_priv);
1303 	return KERN_SUCCESS;
1304 }
1305 
1306 /*
1307  *      User interface for setting a special port.
1308  *
1309  *      Only permits the user to set a user-owned special port
1310  *      ID, rejecting a kernel-owned special port ID.
1311  *
1312  *      A special kernel port cannot be set up using this
1313  *      routine; use kernel_set_special_port() instead.
1314  */
1315 kern_return_t
host_set_special_port_from_user(host_priv_t host_priv,int id,ipc_port_t port)1316 host_set_special_port_from_user(host_priv_t host_priv, int id, ipc_port_t port)
1317 {
1318 	if (host_priv == HOST_PRIV_NULL || id <= HOST_MAX_SPECIAL_KERNEL_PORT || id > HOST_MAX_SPECIAL_PORT) {
1319 		return KERN_INVALID_ARGUMENT;
1320 	}
1321 
1322 	if (task_is_driver(current_task())) {
1323 		return KERN_NO_ACCESS;
1324 	}
1325 
1326 	/*
1327 	 * rdar://70585367
1328 	 * disallow immovable send so other process can't retrieve it through host_get_special_port()
1329 	 */
1330 	if (!ipc_can_stash_naked_send(port)) {
1331 		return KERN_DENIED;
1332 	}
1333 
1334 	return host_set_special_port(host_priv, id, port);
1335 }
1336 
1337 kern_return_t
host_set_special_port(host_priv_t host_priv,int id,ipc_port_t port)1338 host_set_special_port(host_priv_t host_priv, int id, ipc_port_t port)
1339 {
1340 	if (host_priv == HOST_PRIV_NULL || id <= HOST_MAX_SPECIAL_KERNEL_PORT || id > HOST_MAX_SPECIAL_PORT) {
1341 		return KERN_INVALID_ARGUMENT;
1342 	}
1343 
1344 	if (current_task() != kernel_task && !task_is_initproc(current_task())) {
1345 		bool allowed = (id == HOST_TELEMETRY_PORT &&
1346 		    IOTaskHasEntitlement(current_task(), "com.apple.private.xpc.launchd.event-monitor"));
1347 #if CONFIG_CSR
1348 		if (!allowed) {
1349 			allowed = (csr_check(CSR_ALLOW_TASK_FOR_PID) == 0);
1350 		}
1351 #endif
1352 		if (!allowed) {
1353 			return KERN_NO_ACCESS;
1354 		}
1355 	}
1356 
1357 #if CONFIG_MACF
1358 	if (mac_task_check_set_host_special_port(current_task(), id, port) != 0) {
1359 		return KERN_NO_ACCESS;
1360 	}
1361 #endif
1362 
1363 	return kernel_set_special_port(host_priv, id, port);
1364 }
1365 
1366 /*
1367  *      User interface for retrieving a special port.
1368  *
1369  *      Note that there is nothing to prevent a user special
1370  *      port from disappearing after it has been discovered by
1371  *      the caller; thus, using a special port can always result
1372  *      in a "port not valid" error.
1373  */
1374 
1375 kern_return_t
host_get_special_port_from_user(host_priv_t host_priv,__unused int node,int id,ipc_port_t * portp)1376 host_get_special_port_from_user(host_priv_t host_priv, __unused int node, int id, ipc_port_t * portp)
1377 {
1378 	if (host_priv == HOST_PRIV_NULL || id == HOST_SECURITY_PORT || id > HOST_MAX_SPECIAL_PORT || id < HOST_MIN_SPECIAL_PORT) {
1379 		return KERN_INVALID_ARGUMENT;
1380 	}
1381 
1382 	task_t task = current_task();
1383 	if (task && task_is_driver(task) && id > HOST_MAX_SPECIAL_KERNEL_PORT) {
1384 		/* allow HID drivers to get the sysdiagnose port for keychord handling */
1385 		if (id == HOST_SYSDIAGNOSE_PORT &&
1386 		    IOCurrentTaskHasEntitlement(kIODriverKitHIDFamilyEventServiceEntitlementKey)) {
1387 			goto get_special_port;
1388 		}
1389 		return KERN_NO_ACCESS;
1390 	}
1391 get_special_port:
1392 	return host_get_special_port(host_priv, node, id, portp);
1393 }
1394 
1395 kern_return_t
host_get_special_port(host_priv_t host_priv,__unused int node,int id,ipc_port_t * portp)1396 host_get_special_port(host_priv_t host_priv, __unused int node, int id, ipc_port_t * portp)
1397 {
1398 	ipc_port_t port;
1399 
1400 	if (host_priv == HOST_PRIV_NULL || id == HOST_SECURITY_PORT || id > HOST_MAX_SPECIAL_PORT || id < HOST_MIN_SPECIAL_PORT) {
1401 		return KERN_INVALID_ARGUMENT;
1402 	}
1403 
1404 	host_lock(host_priv);
1405 	port = realhost.special[id];
1406 	switch (id) {
1407 	case HOST_PORT:
1408 		*portp = ipc_kobject_copy_send(port, &realhost, IKOT_HOST);
1409 		break;
1410 	case HOST_PRIV_PORT:
1411 		*portp = ipc_kobject_copy_send(port, &realhost, IKOT_HOST_PRIV);
1412 		break;
1413 	case HOST_IO_MAIN_PORT:
1414 		*portp = ipc_port_copy_send_any(main_device_port);
1415 		break;
1416 	default:
1417 		*portp = ipc_port_copy_send_mqueue(port);
1418 		break;
1419 	}
1420 	host_unlock(host_priv);
1421 
1422 	return KERN_SUCCESS;
1423 }
1424 
1425 /*
1426  *	host_get_io_main
1427  *
1428  *	Return the IO main access port for this host.
1429  */
1430 kern_return_t
host_get_io_main(host_t host,io_main_t * io_mainp)1431 host_get_io_main(host_t host, io_main_t * io_mainp)
1432 {
1433 	if (host == HOST_NULL) {
1434 		return KERN_INVALID_ARGUMENT;
1435 	}
1436 
1437 	return host_get_io_main_port(host_priv_self(), io_mainp);
1438 }
1439 
1440 host_t
host_self(void)1441 host_self(void)
1442 {
1443 	return &realhost;
1444 }
1445 
1446 host_priv_t
host_priv_self(void)1447 host_priv_self(void)
1448 {
1449 	return &realhost;
1450 }
1451 
1452 kern_return_t
host_set_atm_diagnostic_flag(host_t host,uint32_t diagnostic_flag)1453 host_set_atm_diagnostic_flag(host_t host, uint32_t diagnostic_flag)
1454 {
1455 	if (host == HOST_NULL) {
1456 		return KERN_INVALID_ARGUMENT;
1457 	}
1458 
1459 	if (!IOCurrentTaskHasEntitlement("com.apple.private.set-atm-diagnostic-flag")) {
1460 		return KERN_NO_ACCESS;
1461 	}
1462 
1463 #if CONFIG_ATM
1464 	return atm_set_diagnostic_config(diagnostic_flag);
1465 #else
1466 	(void)diagnostic_flag;
1467 	return KERN_NOT_SUPPORTED;
1468 #endif
1469 }
1470 
1471 kern_return_t
host_set_multiuser_config_flags(host_priv_t host_priv,uint32_t multiuser_config)1472 host_set_multiuser_config_flags(host_priv_t host_priv, uint32_t multiuser_config)
1473 {
1474 #if !defined(XNU_TARGET_OS_OSX)
1475 	if (host_priv == HOST_PRIV_NULL) {
1476 		return KERN_INVALID_ARGUMENT;
1477 	}
1478 
1479 	/*
1480 	 * multiuser bit is extensively used for sharedIpad mode.
1481 	 * Caller sets the sharedIPad or other mutiuser modes.
1482 	 * Any override during commpage setting is not suitable anymore.
1483 	 */
1484 	commpage_update_multiuser_config(multiuser_config);
1485 	return KERN_SUCCESS;
1486 #else
1487 	(void)host_priv;
1488 	(void)multiuser_config;
1489 	return KERN_NOT_SUPPORTED;
1490 #endif
1491 }
1492