[apple/xnu.git] / bsd / kern / kern_memorystatus.c

/*
 * Copyright (c) 2006 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 * 
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. The rights granted to you under the License
 * may not be used to create, or enable the creation or redistribution of,
 * unlawful or unlicensed copies of an Apple operating system, or to
 * circumvent, violate, or enable the circumvention or violation of, any
 * terms of an Apple operating system software license agreement.
 * 
 * Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this file.
 * 
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 * 
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 *
 */

#include <sys/kern_event.h>
#include <sys/kern_memorystatus.h>

#include <kern/sched_prim.h>
#include <kern/lock.h>
#include <kern/task.h>
#include <kern/thread.h>
#include <libkern/libkern.h>
#include <mach/task.h>
#include <mach/task_info.h>
#include <sys/proc.h>
#include <sys/signal.h>
#include <sys/signalvar.h>
#include <sys/sysctl.h>
#include <sys/wait.h>

extern unsigned int    vm_page_free_count;
extern unsigned int    vm_page_active_count;
extern unsigned int    vm_page_inactive_count;
extern unsigned int    vm_page_purgeable_count;
extern unsigned int    vm_page_wire_count;

static void kern_memorystatus_thread(void);

int kern_memorystatus_wakeup = 0;
int kern_memorystatus_level = 0;
int kern_memorystatus_last_level = 0;
unsigned int kern_memorystatus_kev_failure_count = 0;
int kern_memorystatus_level_critical = 5;
#define kern_memorystatus_level_highwater (kern_memorystatus_level_critical + 5)

static struct {
	jetsam_kernel_stats_t stats;
	size_t entry_count;
	jetsam_snapshot_entry_t entries[kMaxSnapshotEntries];
} jetsam_snapshot;

static jetsam_priority_entry_t jetsam_priority_list[kMaxPriorityEntries];
#define jetsam_snapshot_list jetsam_snapshot.entries

static int jetsam_priority_list_index = 0;
static int jetsam_priority_list_count = 0;
static int jetsam_snapshot_list_count = 0;

static lck_mtx_t * jetsam_list_mlock;
static lck_attr_t * jetsam_lck_attr;
static lck_grp_t * jetsam_lck_grp;
static lck_grp_attr_t * jetsam_lck_grp_attr;

SYSCTL_INT(_kern, OID_AUTO, memorystatus_level, CTLFLAG_RD, &kern_memorystatus_level, 0, "");
SYSCTL_UINT(_kern, OID_AUTO, memorystatus_kev_failure_count, CTLFLAG_RD, &kern_memorystatus_kev_failure_count, 0, "");

__private_extern__ void
kern_memorystatus_init(void)
{
    jetsam_lck_attr = lck_attr_alloc_init();
    jetsam_lck_grp_attr= lck_grp_attr_alloc_init();
    jetsam_lck_grp = lck_grp_alloc_init("jetsam",  jetsam_lck_grp_attr);
    jetsam_list_mlock = lck_mtx_alloc_init(jetsam_lck_grp, jetsam_lck_attr);

	(void)kernel_thread(kernel_task, kern_memorystatus_thread);
}

static uint32_t
jetsam_task_page_count(task_t task)
{
	kern_return_t ret;
	static task_info_data_t data;
	static struct task_basic_info *info = (struct task_basic_info *)&data;
	static mach_msg_type_number_t count = TASK_BASIC_INFO_COUNT;

	ret = task_info(task, TASK_BASIC_INFO, (task_info_t)&data, &count);
	if (ret == KERN_SUCCESS) {
		return info->resident_size / PAGE_SIZE;
	}
	return 0;
}

static uint32_t
jetsam_flags_for_pid(pid_t pid)
{
	int i;

	for (i = 0; i < jetsam_priority_list_count; i++) {
		if (pid == jetsam_priority_list[i].pid) {
			return jetsam_priority_list[i].flags;
		}
	}
	return 0;
}

static void
jetsam_snapshot_procs(void)
{
	proc_t p;
	int i = 0;

	jetsam_snapshot.stats.free_pages = vm_page_free_count;
	jetsam_snapshot.stats.active_pages = vm_page_active_count;
	jetsam_snapshot.stats.inactive_pages = vm_page_inactive_count;
	jetsam_snapshot.stats.purgeable_pages = vm_page_purgeable_count;
	jetsam_snapshot.stats.wired_pages = vm_page_wire_count;
	proc_list_lock();
	LIST_FOREACH(p, &allproc, p_list) {
		task_t task = p->task;
		jetsam_snapshot_list[i].pid = p->p_pid;
		jetsam_snapshot_list[i].pages = jetsam_task_page_count(task);
		jetsam_snapshot_list[i].flags = jetsam_flags_for_pid(p->p_pid);
		strlcpy(&jetsam_snapshot_list[i].name[0], p->p_comm, MAXCOMLEN+1);
#ifdef DEBUG
		printf("jetsam snapshot pid = %d, uuid = %02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x\n",
			p->p_pid, 
			p->p_uuid[0], p->p_uuid[1], p->p_uuid[2], p->p_uuid[3], p->p_uuid[4], p->p_uuid[5], p->p_uuid[6], p->p_uuid[7],
			p->p_uuid[8], p->p_uuid[9], p->p_uuid[10], p->p_uuid[11], p->p_uuid[12], p->p_uuid[13], p->p_uuid[14], p->p_uuid[15]);
#endif
		memcpy(&jetsam_snapshot_list[i].uuid[0], &p->p_uuid[0], sizeof(p->p_uuid));
		i++;
		if (i == kMaxSnapshotEntries) {
			break;
		} 	
	}
	proc_list_unlock();	
	jetsam_snapshot.entry_count = jetsam_snapshot_list_count = i - 1;
}

static void
jetsam_mark_pid_in_snapshot(pid_t pid, int flag)
{

	int i = 0;

	for (i = 0; i < jetsam_snapshot_list_count; i++) {
		if (jetsam_snapshot_list[i].pid == pid) {
			jetsam_snapshot_list[i].flags |= flag;
			return;
		}
	}
}

int
jetsam_kill_top_proc(void)
{
	proc_t p;

	if (jetsam_snapshot_list_count == 0) {
		jetsam_snapshot_procs();
	}
	lck_mtx_lock(jetsam_list_mlock);
	while (jetsam_priority_list_index < jetsam_priority_list_count) {
		pid_t aPid;
		aPid = jetsam_priority_list[jetsam_priority_list_index].pid;
		jetsam_priority_list_index++;
		/* skip empty slots in the list */
		if (aPid == 0) {
			continue; // with lock held
		}
		lck_mtx_unlock(jetsam_list_mlock);
		jetsam_mark_pid_in_snapshot(aPid, kJetsamFlagsKilled);
		p = proc_find(aPid);
		if (p != NULL) {
			printf("jetsam: killing pid %d [%s] - memory_status_level: %d - ", 
					aPid, (p->p_comm ? p->p_comm : "(unknown)"), kern_memorystatus_level);
			exit1(p, W_EXITCODE(0, SIGKILL), (int *)NULL);
			proc_rele(p);
#if DEBUG
			printf("jetsam: pid %d killed - memory_status_level: %d\n", aPid, kern_memorystatus_level);
#endif /* DEBUG */
			return 0;
		}
	    lck_mtx_lock(jetsam_list_mlock);
	}
	lck_mtx_unlock(jetsam_list_mlock);
	return -1;
}

static int
jetsam_kill_hiwat_proc(void)
{
	proc_t p;
	int i;
	if (jetsam_snapshot_list_count == 0) {
		jetsam_snapshot_procs();
	}
	lck_mtx_lock(jetsam_list_mlock);
	for (i = jetsam_priority_list_index; i < jetsam_priority_list_count; i++) {
		pid_t aPid;
		int32_t hiwat;
		aPid = jetsam_priority_list[i].pid;
		hiwat = jetsam_priority_list[i].hiwat_pages;	
		/* skip empty or non-hiwat slots in the list */
		if (aPid == 0 || (hiwat < 0)) {
			continue; // with lock held
		}
		lck_mtx_unlock(jetsam_list_mlock);
		p = proc_find(aPid);
		if (p != NULL) {
			int32_t pages = (int32_t)jetsam_task_page_count(p->task);
			if (pages > hiwat) {
#if DEBUG
				printf("jetsam: killing pid %d [%s] - %d pages > hiwat (%d)\n", aPid, p->p_comm, pages, hiwat);
#endif /* DEBUG */
				exit1(p, W_EXITCODE(0, SIGKILL), (int *)NULL);
				proc_rele(p);
#if DEBUG
				printf("jetsam: pid %d killed - memory_status_level: %d\n", aPid, kern_memorystatus_level);
#endif /* DEBUG */
				jetsam_mark_pid_in_snapshot(aPid, kJetsamFlagsKilledHiwat);
				jetsam_priority_list[i].pid = 0;
				return 0;
			} else {
				proc_rele(p);
			}

		}
		lck_mtx_lock(jetsam_list_mlock);
	}
	lck_mtx_unlock(jetsam_list_mlock);
	return -1;
}

static void
kern_memorystatus_thread(void)
{
	struct kev_msg ev_msg;
	jetsam_kernel_stats_t data;
	int ret;

	while(1) {

		while (kern_memorystatus_level <= kern_memorystatus_level_critical) {
			if (jetsam_kill_top_proc() < 0) {
				break;
			}
		}

		while (kern_memorystatus_level <= kern_memorystatus_level_highwater) {
			if (jetsam_kill_hiwat_proc() < 0) {
				break;
			}
		}
				
		kern_memorystatus_last_level = kern_memorystatus_level;

		ev_msg.vendor_code    = KEV_VENDOR_APPLE;
		ev_msg.kev_class      = KEV_SYSTEM_CLASS;
		ev_msg.kev_subclass   = KEV_MEMORYSTATUS_SUBCLASS;

		/* pass the memory status level (percent free) */
		ev_msg.event_code     = kMemoryStatusLevelNote;

		ev_msg.dv[0].data_length = sizeof kern_memorystatus_last_level;
		ev_msg.dv[0].data_ptr = &kern_memorystatus_last_level;
		ev_msg.dv[1].data_length = sizeof data;
		ev_msg.dv[1].data_ptr = &data;
		ev_msg.dv[2].data_length = 0;

		data.free_pages = vm_page_free_count;
		data.active_pages = vm_page_active_count;
		data.inactive_pages = vm_page_inactive_count;
		data.purgeable_pages = vm_page_purgeable_count;
		data.wired_pages = vm_page_wire_count;

		ret = kev_post_msg(&ev_msg);
		if (ret) {
			kern_memorystatus_kev_failure_count++;
			printf("%s: kev_post_msg() failed, err %d\n", __func__, ret);
		}

		if (jetsam_snapshot_list_count) {
			size_t snapshot_size =  sizeof(jetsam_kernel_stats_t) + sizeof(size_t) + sizeof(jetsam_snapshot_entry_t) * jetsam_snapshot_list_count;
			ev_msg.event_code = kMemoryStatusSnapshotNote;
			ev_msg.dv[0].data_length = sizeof snapshot_size;
			ev_msg.dv[0].data_ptr = &snapshot_size;
			ev_msg.dv[1].data_length = 0;

			ret = kev_post_msg(&ev_msg);
			if (ret) {
				kern_memorystatus_kev_failure_count++;
				printf("%s: kev_post_msg() failed, err %d\n", __func__, ret);
			}
		}

		if (kern_memorystatus_level >= kern_memorystatus_last_level + 5 ||
		    kern_memorystatus_level <= kern_memorystatus_last_level - 5)
			continue;

		assert_wait(&kern_memorystatus_wakeup, THREAD_UNINT);
		(void)thread_block((thread_continue_t)kern_memorystatus_thread);
	}
}

static int
sysctl_io_variable(struct sysctl_req *req, void *pValue, size_t currentsize, size_t maxsize, size_t *newsize)
{
    int error;

    /* Copy blob out */
    error = SYSCTL_OUT(req, pValue, currentsize);

    /* error or nothing to set */
    if (error || !req->newptr)
        return(error);

    if (req->newlen > maxsize) {
		return EINVAL;
	}
	error = SYSCTL_IN(req, pValue, req->newlen);

	if (!error) {
		*newsize = req->newlen;
	}

    return(error);
}

static int
sysctl_handle_kern_memorystatus_priority_list(__unused struct sysctl_oid *oid, __unused void *arg1, __unused int arg2, struct sysctl_req *req)
{
	int i, ret;
	jetsam_priority_entry_t temp_list[kMaxPriorityEntries];
	size_t newsize, currentsize;

	if (req->oldptr) {
		lck_mtx_lock(jetsam_list_mlock);
		for (i = 0; i < jetsam_priority_list_count; i++) {
			temp_list[i] = jetsam_priority_list[i];
		}
		lck_mtx_unlock(jetsam_list_mlock);
	}

	currentsize = sizeof(jetsam_priority_list[0]) * jetsam_priority_list_count;

	ret = sysctl_io_variable(req, &temp_list[0], currentsize, sizeof(temp_list), &newsize);

	if (!ret && req->newptr) {
		jetsam_priority_list_count = newsize / sizeof(jetsam_priority_list[0]);
#if DEBUG 
		printf("set jetsam priority pids = { ");
		for (i = 0; i < jetsam_priority_list_count; i++) {
			printf("(%d, 0x%08x, %d) ", temp_list[i].pid, temp_list[i].flags, temp_list[i].hiwat_pages);
		}
		printf("}\n");
#endif /* DEBUG */
		lck_mtx_lock(jetsam_list_mlock);
		for (i = 0; i < jetsam_priority_list_count; i++) {
			jetsam_priority_list[i] = temp_list[i];
		}
		for (i = jetsam_priority_list_count; i < kMaxPriorityEntries; i++) {
			jetsam_priority_list[i].pid = 0;
			jetsam_priority_list[i].flags = 0;
			jetsam_priority_list[i].hiwat_pages = -1;
			jetsam_priority_list[i].hiwat_reserved1 = -1;
			jetsam_priority_list[i].hiwat_reserved2 = -1;
			jetsam_priority_list[i].hiwat_reserved3 = -1;
		}
		jetsam_priority_list_index = 0;
		lck_mtx_unlock(jetsam_list_mlock);
	}	
	return ret;
}

static int
sysctl_handle_kern_memorystatus_snapshot(__unused struct sysctl_oid *oid, __unused void *arg1, __unused int arg2, struct sysctl_req *req)
{
	int ret;
	size_t currentsize = 0;

	if (jetsam_snapshot_list_count > 0) {
		currentsize = sizeof(jetsam_kernel_stats_t) + sizeof(size_t) + sizeof(jetsam_snapshot_entry_t) * jetsam_snapshot_list_count;
	}
	if (!currentsize) {
		if (req->oldptr) {
#ifdef DEBUG
			printf("kern.memorystatus_snapshot returning EINVAL\n");
#endif
			return EINVAL;
		}
		else {
#ifdef DEBUG
			printf("kern.memorystatus_snapshot returning 0 for size\n");
#endif
		}
	} else {
#ifdef DEBUG
			printf("kern.memorystatus_snapshot returning %ld for size\n", (long)currentsize);
#endif
	}	
	ret = sysctl_io_variable(req, &jetsam_snapshot, currentsize, 0, NULL);
	if (!ret && req->oldptr) {
		jetsam_snapshot.entry_count = jetsam_snapshot_list_count = 0;
	}
	return ret;
}

SYSCTL_PROC(_kern, OID_AUTO, memorystatus_priority_list, CTLTYPE_OPAQUE|CTLFLAG_RW, 0, 0, sysctl_handle_kern_memorystatus_priority_list, "S,jetsam_priorities", "");
SYSCTL_PROC(_kern, OID_AUTO, memorystatus_snapshot, CTLTYPE_OPAQUE|CTLFLAG_RD, 0, 0, sysctl_handle_kern_memorystatus_snapshot, "S,jetsam_snapshot", "");
Commit	Line	Data
2d21ac55 A	1	/*
	2	* Copyright (c) 2006 Apple Computer, 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	*/
2d21ac55 A	29
	30	#include <sys/kern_event.h>
	31	#include <sys/kern_memorystatus.h>
	32
	33	#include <kern/sched_prim.h>
b0d623f7	34	#include <kern/lock.h>
2d21ac55 A	35	#include <kern/task.h>
	36	#include <kern/thread.h>
	37	#include <libkern/libkern.h>
b0d623f7 A	38	#include <mach/task.h>
	39	#include <mach/task_info.h>
	40	#include <sys/proc.h>
	41	#include <sys/signal.h>
	42	#include <sys/signalvar.h>
2d21ac55	43	#include <sys/sysctl.h>
b0d623f7	44	#include <sys/wait.h>
2d21ac55	45
593a1d5f A	46	extern unsigned int vm_page_free_count;
	47	extern unsigned int vm_page_active_count;
	48	extern unsigned int vm_page_inactive_count;
	49	extern unsigned int vm_page_purgeable_count;
	50	extern unsigned int vm_page_wire_count;
	51
2d21ac55 A	52	static void kern_memorystatus_thread(void);
	53
	54	int kern_memorystatus_wakeup = 0;
2d21ac55 A	55	int kern_memorystatus_level = 0;
	56	int kern_memorystatus_last_level = 0;
	57	unsigned int kern_memorystatus_kev_failure_count = 0;
b0d623f7	58	int kern_memorystatus_level_critical = 5;
d1ecb069	59	#define kern_memorystatus_level_highwater (kern_memorystatus_level_critical + 5)
b0d623f7 A	60
	61	static struct {
	62	jetsam_kernel_stats_t stats;
	63	size_t entry_count;
	64	jetsam_snapshot_entry_t entries[kMaxSnapshotEntries];
	65	} jetsam_snapshot;
	66
	67	static jetsam_priority_entry_t jetsam_priority_list[kMaxPriorityEntries];
	68	#define jetsam_snapshot_list jetsam_snapshot.entries
	69
	70	static int jetsam_priority_list_index = 0;
	71	static int jetsam_priority_list_count = 0;
	72	static int jetsam_snapshot_list_count = 0;
	73
	74	static lck_mtx_t * jetsam_list_mlock;
	75	static lck_attr_t * jetsam_lck_attr;
	76	static lck_grp_t * jetsam_lck_grp;
	77	static lck_grp_attr_t * jetsam_lck_grp_attr;
2d21ac55 A	78
	79	SYSCTL_INT(_kern, OID_AUTO, memorystatus_level, CTLFLAG_RD, &kern_memorystatus_level, 0, "");
	80	SYSCTL_UINT(_kern, OID_AUTO, memorystatus_kev_failure_count, CTLFLAG_RD, &kern_memorystatus_kev_failure_count, 0, "");
	81
	82	__private_extern__ void
	83	kern_memorystatus_init(void)
	84	{
b0d623f7 A	85	jetsam_lck_attr = lck_attr_alloc_init();
	86	jetsam_lck_grp_attr= lck_grp_attr_alloc_init();
	87	jetsam_lck_grp = lck_grp_alloc_init("jetsam", jetsam_lck_grp_attr);
	88	jetsam_list_mlock = lck_mtx_alloc_init(jetsam_lck_grp, jetsam_lck_attr);
	89
2d21ac55 A	90	(void)kernel_thread(kernel_task, kern_memorystatus_thread);
	91	}
	92
b0d623f7 A	93	static uint32_t
	94	jetsam_task_page_count(task_t task)
	95	{
	96	kern_return_t ret;
	97	static task_info_data_t data;
	98	static struct task_basic_info info = (struct task_basic_info )&data;
	99	static mach_msg_type_number_t count = TASK_BASIC_INFO_COUNT;
	100
	101	ret = task_info(task, TASK_BASIC_INFO, (task_info_t)&data, &count);
	102	if (ret == KERN_SUCCESS) {
	103	return info->resident_size / PAGE_SIZE;
	104	}
	105	return 0;
	106	}
	107
	108	static uint32_t
	109	jetsam_flags_for_pid(pid_t pid)
	110	{
	111	int i;
	112
	113	for (i = 0; i < jetsam_priority_list_count; i++) {
	114	if (pid == jetsam_priority_list[i].pid) {
	115	return jetsam_priority_list[i].flags;
	116	}
	117	}
	118	return 0;
	119	}
	120
	121	static void
	122	jetsam_snapshot_procs(void)
	123	{
	124	proc_t p;
	125	int i = 0;
	126
	127	jetsam_snapshot.stats.free_pages = vm_page_free_count;
	128	jetsam_snapshot.stats.active_pages = vm_page_active_count;
	129	jetsam_snapshot.stats.inactive_pages = vm_page_inactive_count;
	130	jetsam_snapshot.stats.purgeable_pages = vm_page_purgeable_count;
	131	jetsam_snapshot.stats.wired_pages = vm_page_wire_count;
	132	proc_list_lock();
	133	LIST_FOREACH(p, &allproc, p_list) {
	134	task_t task = p->task;
	135	jetsam_snapshot_list[i].pid = p->p_pid;
	136	jetsam_snapshot_list[i].pages = jetsam_task_page_count(task);
	137	jetsam_snapshot_list[i].flags = jetsam_flags_for_pid(p->p_pid);
	138	strlcpy(&jetsam_snapshot_list[i].name[0], p->p_comm, MAXCOMLEN+1);
	139	#ifdef DEBUG
	140	printf("jetsam snapshot pid = %d, uuid = %02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x\n",
	141	p->p_pid,
	142	p->p_uuid[0], p->p_uuid[1], p->p_uuid[2], p->p_uuid[3], p->p_uuid[4], p->p_uuid[5], p->p_uuid[6], p->p_uuid[7],
	143	p->p_uuid[8], p->p_uuid[9], p->p_uuid[10], p->p_uuid[11], p->p_uuid[12], p->p_uuid[13], p->p_uuid[14], p->p_uuid[15]);
	144	#endif
	145	memcpy(&jetsam_snapshot_list[i].uuid[0], &p->p_uuid[0], sizeof(p->p_uuid));
	146	i++;
	147	if (i == kMaxSnapshotEntries) {
	148	break;
	149	}
	150	}
	151	proc_list_unlock();
	152	jetsam_snapshot.entry_count = jetsam_snapshot_list_count = i - 1;
	153	}
	154
	155	static void
d1ecb069	156	jetsam_mark_pid_in_snapshot(pid_t pid, int flag)
b0d623f7 A	157	{
	158
	159	int i = 0;
	160
	161	for (i = 0; i < jetsam_snapshot_list_count; i++) {
	162	if (jetsam_snapshot_list[i].pid == pid) {
d1ecb069	163	jetsam_snapshot_list[i].flags \|= flag;
b0d623f7 A	164	return;
	165	}
	166	}
	167	}
	168
d1ecb069	169	int
b0d623f7 A	170	jetsam_kill_top_proc(void)
	171	{
	172	proc_t p;
	173
	174	if (jetsam_snapshot_list_count == 0) {
	175	jetsam_snapshot_procs();
	176	}
	177	lck_mtx_lock(jetsam_list_mlock);
	178	while (jetsam_priority_list_index < jetsam_priority_list_count) {
	179	pid_t aPid;
	180	aPid = jetsam_priority_list[jetsam_priority_list_index].pid;
	181	jetsam_priority_list_index++;
	182	/* skip empty slots in the list */
	183	if (aPid == 0) {
	184	continue; // with lock held
	185	}
	186	lck_mtx_unlock(jetsam_list_mlock);
d1ecb069	187	jetsam_mark_pid_in_snapshot(aPid, kJetsamFlagsKilled);
b0d623f7 A	188	p = proc_find(aPid);
b0d623f7 A	189	if (p != NULL) {
d1ecb069 A	190	printf("jetsam: killing pid %d [%s] - memory_status_level: %d - ",
d1ecb069 A	191	aPid, (p->p_comm ? p->p_comm : "(unknown)"), kern_memorystatus_level);
b0d623f7 A	192	exit1(p, W_EXITCODE(0, SIGKILL), (int *)NULL);
	193	proc_rele(p);
	194	#if DEBUG
	195	printf("jetsam: pid %d killed - memory_status_level: %d\n", aPid, kern_memorystatus_level);
	196	#endif /* DEBUG */
	197	return 0;
	198	}
	199	lck_mtx_lock(jetsam_list_mlock);
	200	}
	201	lck_mtx_unlock(jetsam_list_mlock);
	202	return -1;
	203	}
	204
d1ecb069 A	205	static int
	206	jetsam_kill_hiwat_proc(void)
	207	{
	208	proc_t p;
	209	int i;
	210	if (jetsam_snapshot_list_count == 0) {
	211	jetsam_snapshot_procs();
	212	}
	213	lck_mtx_lock(jetsam_list_mlock);
	214	for (i = jetsam_priority_list_index; i < jetsam_priority_list_count; i++) {
	215	pid_t aPid;
	216	int32_t hiwat;
	217	aPid = jetsam_priority_list[i].pid;
	218	hiwat = jetsam_priority_list[i].hiwat_pages;
	219	/* skip empty or non-hiwat slots in the list */
	220	if (aPid == 0 \|\| (hiwat < 0)) {
	221	continue; // with lock held
	222	}
	223	lck_mtx_unlock(jetsam_list_mlock);
	224	p = proc_find(aPid);
	225	if (p != NULL) {
	226	int32_t pages = (int32_t)jetsam_task_page_count(p->task);
	227	if (pages > hiwat) {
	228	#if DEBUG
	229	printf("jetsam: killing pid %d [%s] - %d pages > hiwat (%d)\n", aPid, p->p_comm, pages, hiwat);
	230	#endif /* DEBUG */
	231	exit1(p, W_EXITCODE(0, SIGKILL), (int *)NULL);
	232	proc_rele(p);
	233	#if DEBUG
	234	printf("jetsam: pid %d killed - memory_status_level: %d\n", aPid, kern_memorystatus_level);
	235	#endif /* DEBUG */
	236	jetsam_mark_pid_in_snapshot(aPid, kJetsamFlagsKilledHiwat);
	237	jetsam_priority_list[i].pid = 0;
	238	return 0;
	239	} else {
	240	proc_rele(p);
	241	}
	242
	243	}
	244	lck_mtx_lock(jetsam_list_mlock);
	245	}
	246	lck_mtx_unlock(jetsam_list_mlock);
	247	return -1;
	248	}
	249
2d21ac55 A	250	static void
	251	kern_memorystatus_thread(void)
	252	{
	253	struct kev_msg ev_msg;
b0d623f7	254	jetsam_kernel_stats_t data;
2d21ac55 A	255	int ret;
	256
	257	while(1) {
b0d623f7 A	258
	259	while (kern_memorystatus_level <= kern_memorystatus_level_critical) {
	260	if (jetsam_kill_top_proc() < 0) {
	261	break;
	262	}
	263	}
	264
d1ecb069 A	265	while (kern_memorystatus_level <= kern_memorystatus_level_highwater) {
	266	if (jetsam_kill_hiwat_proc() < 0) {
	267	break;
	268	}
	269	}
	270
2d21ac55 A	271	kern_memorystatus_last_level = kern_memorystatus_level;
	272
	273	ev_msg.vendor_code = KEV_VENDOR_APPLE;
	274	ev_msg.kev_class = KEV_SYSTEM_CLASS;
	275	ev_msg.kev_subclass = KEV_MEMORYSTATUS_SUBCLASS;
	276
b0d623f7 A	277	/* pass the memory status level (percent free) */
b0d623f7 A	278	ev_msg.event_code = kMemoryStatusLevelNote;
2d21ac55	279
b0d623f7 A	280	ev_msg.dv[0].data_length = sizeof kern_memorystatus_last_level;
	281	ev_msg.dv[0].data_ptr = &kern_memorystatus_last_level;
	282	ev_msg.dv[1].data_length = sizeof data;
	283	ev_msg.dv[1].data_ptr = &data;
	284	ev_msg.dv[2].data_length = 0;
593a1d5f A	285
	286	data.free_pages = vm_page_free_count;
	287	data.active_pages = vm_page_active_count;
	288	data.inactive_pages = vm_page_inactive_count;
	289	data.purgeable_pages = vm_page_purgeable_count;
	290	data.wired_pages = vm_page_wire_count;
2d21ac55 A	291
	292	ret = kev_post_msg(&ev_msg);
	293	if (ret) {
	294	kern_memorystatus_kev_failure_count++;
	295	printf("%s: kev_post_msg() failed, err %d\n", __func__, ret);
	296	}
	297
b0d623f7 A	298	if (jetsam_snapshot_list_count) {
	299	size_t snapshot_size = sizeof(jetsam_kernel_stats_t) + sizeof(size_t) + sizeof(jetsam_snapshot_entry_t) * jetsam_snapshot_list_count;
	300	ev_msg.event_code = kMemoryStatusSnapshotNote;
	301	ev_msg.dv[0].data_length = sizeof snapshot_size;
	302	ev_msg.dv[0].data_ptr = &snapshot_size;
	303	ev_msg.dv[1].data_length = 0;
	304
	305	ret = kev_post_msg(&ev_msg);
	306	if (ret) {
	307	kern_memorystatus_kev_failure_count++;
	308	printf("%s: kev_post_msg() failed, err %d\n", __func__, ret);
	309	}
	310	}
	311
2d21ac55 A	312	if (kern_memorystatus_level >= kern_memorystatus_last_level + 5 \|\|
	313	kern_memorystatus_level <= kern_memorystatus_last_level - 5)
	314	continue;
	315
	316	assert_wait(&kern_memorystatus_wakeup, THREAD_UNINT);
	317	(void)thread_block((thread_continue_t)kern_memorystatus_thread);
	318	}
	319	}
b0d623f7 A	320
	321	static int
	322	sysctl_io_variable(struct sysctl_req req, void pValue, size_t currentsize, size_t maxsize, size_t *newsize)
	323	{
	324	int error;
	325
	326	/* Copy blob out */
	327	error = SYSCTL_OUT(req, pValue, currentsize);
	328
	329	/* error or nothing to set */
	330	if (error \|\| !req->newptr)
	331	return(error);
	332
	333	if (req->newlen > maxsize) {
	334	return EINVAL;
	335	}
	336	error = SYSCTL_IN(req, pValue, req->newlen);
	337
	338	if (!error) {
	339	*newsize = req->newlen;
	340	}
	341
	342	return(error);
	343	}
	344
	345	static int
	346	sysctl_handle_kern_memorystatus_priority_list(__unused struct sysctl_oid oid, __unused void arg1, __unused int arg2, struct sysctl_req *req)
	347	{
	348	int i, ret;
	349	jetsam_priority_entry_t temp_list[kMaxPriorityEntries];
	350	size_t newsize, currentsize;
	351
	352	if (req->oldptr) {
	353	lck_mtx_lock(jetsam_list_mlock);
	354	for (i = 0; i < jetsam_priority_list_count; i++) {
	355	temp_list[i] = jetsam_priority_list[i];
	356	}
	357	lck_mtx_unlock(jetsam_list_mlock);
	358	}
	359
	360	currentsize = sizeof(jetsam_priority_list[0]) * jetsam_priority_list_count;
	361
	362	ret = sysctl_io_variable(req, &temp_list[0], currentsize, sizeof(temp_list), &newsize);
	363
	364	if (!ret && req->newptr) {
	365	jetsam_priority_list_count = newsize / sizeof(jetsam_priority_list[0]);
	366	#if DEBUG
	367	printf("set jetsam priority pids = { ");
	368	for (i = 0; i < jetsam_priority_list_count; i++) {
b7266188	369	printf("(%d, 0x%08x, %d) ", temp_list[i].pid, temp_list[i].flags, temp_list[i].hiwat_pages);
b0d623f7 A	370	}
	371	printf("}\n");
	372	#endif /* DEBUG */
	373	lck_mtx_lock(jetsam_list_mlock);
	374	for (i = 0; i < jetsam_priority_list_count; i++) {
	375	jetsam_priority_list[i] = temp_list[i];
	376	}
	377	for (i = jetsam_priority_list_count; i < kMaxPriorityEntries; i++) {
	378	jetsam_priority_list[i].pid = 0;
	379	jetsam_priority_list[i].flags = 0;
b7266188 A	380	jetsam_priority_list[i].hiwat_pages = -1;
	381	jetsam_priority_list[i].hiwat_reserved1 = -1;
	382	jetsam_priority_list[i].hiwat_reserved2 = -1;
	383	jetsam_priority_list[i].hiwat_reserved3 = -1;
b0d623f7 A	384	}
	385	jetsam_priority_list_index = 0;
	386	lck_mtx_unlock(jetsam_list_mlock);
	387	}
	388	return ret;
	389	}
	390
	391	static int
	392	sysctl_handle_kern_memorystatus_snapshot(__unused struct sysctl_oid oid, __unused void arg1, __unused int arg2, struct sysctl_req *req)
	393	{
	394	int ret;
	395	size_t currentsize = 0;
	396
	397	if (jetsam_snapshot_list_count > 0) {
	398	currentsize = sizeof(jetsam_kernel_stats_t) + sizeof(size_t) + sizeof(jetsam_snapshot_entry_t) * jetsam_snapshot_list_count;
	399	}
	400	if (!currentsize) {
	401	if (req->oldptr) {
	402	#ifdef DEBUG
	403	printf("kern.memorystatus_snapshot returning EINVAL\n");
	404	#endif
	405	return EINVAL;
	406	}
	407	else {
	408	#ifdef DEBUG
	409	printf("kern.memorystatus_snapshot returning 0 for size\n");
	410	#endif
	411	}
	412	} else {
	413	#ifdef DEBUG
	414	printf("kern.memorystatus_snapshot returning %ld for size\n", (long)currentsize);
	415	#endif
	416	}
	417	ret = sysctl_io_variable(req, &jetsam_snapshot, currentsize, 0, NULL);
	418	if (!ret && req->oldptr) {
	419	jetsam_snapshot.entry_count = jetsam_snapshot_list_count = 0;
	420	}
	421	return ret;
	422	}
	423
	424	SYSCTL_PROC(_kern, OID_AUTO, memorystatus_priority_list, CTLTYPE_OPAQUE\|CTLFLAG_RW, 0, 0, sysctl_handle_kern_memorystatus_priority_list, "S,jetsam_priorities", "");
	425	SYSCTL_PROC(_kern, OID_AUTO, memorystatus_snapshot, CTLTYPE_OPAQUE\|CTLFLAG_RD, 0, 0, sysctl_handle_kern_memorystatus_snapshot, "S,jetsam_snapshot", "");