void
tcp_lro_init(void)
{
int i;
bzero(lro_flow_list, sizeof (struct lro_flow) * TCP_LRO_NUM_FLOWS);
for (i = 0; i < TCP_LRO_FLOW_MAP; i++) {
lro_flow_map[i] = TCP_LRO_FLOW_UNINIT;
}
/*
* allocate lock group attribute, group and attribute for tcp_lro_lock
*/
tcp_lro_mtx_grp_attr = lck_grp_attr_alloc_init();
tcp_lro_mtx_grp = lck_grp_alloc_init("tcplro", tcp_lro_mtx_grp_attr);
tcp_lro_mtx_attr = lck_attr_alloc_init();
lck_mtx_init(&tcp_lro_lock, tcp_lro_mtx_grp, tcp_lro_mtx_attr);
tcp_lro_timer = thread_call_allocate(tcp_lro_timer_proc, NULL);
if (tcp_lro_timer == NULL) {
panic_plain("%s: unable to allocate lro timer", __func__);
}
return;
}
void
flowadv_init(void)
{
STAILQ_INIT(&fadv_list);
/* Setup lock group and attribute for fadv_lock */
fadv_lock_grp_attr = lck_grp_attr_alloc_init();
fadv_lock_grp = lck_grp_alloc_init("fadv_lock", fadv_lock_grp_attr);
lck_mtx_init(&fadv_lock, fadv_lock_grp, NULL);
fadv_zone_size = P2ROUNDUP(sizeof (struct flowadv_fcentry),
sizeof (u_int64_t));
fadv_zone = zinit(fadv_zone_size,
FADV_ZONE_MAX * fadv_zone_size, 0, FADV_ZONE_NAME);
if (fadv_zone == NULL) {
panic("%s: failed allocating %s", __func__, FADV_ZONE_NAME);
/* NOTREACHED */
}
zone_change(fadv_zone, Z_EXPAND, TRUE);
zone_change(fadv_zone, Z_CALLERACCT, FALSE);
if (kernel_thread_start(flowadv_thread_func, NULL, &fadv_thread) !=
KERN_SUCCESS) {
panic("%s: couldn't create flow event advisory thread",
__func__);
/* NOTREACHED */
}
thread_deallocate(fadv_thread);
}
void
kpc_thread_init(void)
{
kpc_thread_lckgrp_attr = lck_grp_attr_alloc_init();
kpc_thread_lckgrp = lck_grp_alloc_init("kpc", kpc_thread_lckgrp_attr);
lck_mtx_init(&kpc_thread_lock, kpc_thread_lckgrp, LCK_ATTR_NULL);
}
void
mutex_init_EXT(
lck_mtx_t *mutex,
__unused unsigned short tag)
{
lck_mtx_init(mutex, &LockCompatGroup, LCK_ATTR_NULL);
}
/*
* "Decode" rv for use in the call to dtrace_probe()
*/
if (rval == ERESTART) {
munged_rv0 = -1LL; /* System call will be reissued in user mode. Make DTrace report a -1 return. */
munged_rv1 = -1LL;
} else if (rval != EJUSTRETURN) {
if (rval) {
munged_rv0 = -1LL; /* Mimic what libc will do. */
munged_rv1 = -1LL;
} else {
switch (sy->stsy_return_type) {
case _SYSCALL_RET_INT_T:
munged_rv0 = rv[0];
munged_rv1 = rv[1];
break;
case _SYSCALL_RET_UINT_T:
munged_rv0 = ((u_int)rv[0]);
munged_rv1 = ((u_int)rv[1]);
break;
case _SYSCALL_RET_OFF_T:
case _SYSCALL_RET_UINT64_T:
munged_rv0 = *(u_int64_t *)rv;
munged_rv1 = 0LL;
break;
case _SYSCALL_RET_ADDR_T:
case _SYSCALL_RET_SIZE_T:
case _SYSCALL_RET_SSIZE_T:
munged_rv0 = *(user_addr_t *)rv;
munged_rv1 = 0LL;
break;
case _SYSCALL_RET_NONE:
munged_rv0 = 0LL;
munged_rv1 = 0LL;
break;
default:
munged_rv0 = 0LL;
munged_rv1 = 0LL;
break;
}
}
} else {
munged_rv0 = 0LL;
munged_rv1 = 0LL;
}
(*systrace_probe)(id, munged_rv0, munged_rv0, munged_rv1, (uint64_t)rval, 0);
}
}
#endif /* __APPLE__ */
#define SYSTRACE_SHIFT 16
#define SYSTRACE_ISENTRY(x) ((int)(x) >> SYSTRACE_SHIFT)
#define SYSTRACE_SYSNUM(x) ((int)(x) & ((1 << SYSTRACE_SHIFT) - 1))
#define SYSTRACE_ENTRY(id) ((1 << SYSTRACE_SHIFT) | (id))
#define SYSTRACE_RETURN(id) (id)
#if ((1 << SYSTRACE_SHIFT) <= NSYSCALL)
#error 1 << SYSTRACE_SHIFT must exceed number of system calls
#endif
static dev_info_t *systrace_devi;
static dtrace_provider_id_t systrace_id;
#if !defined (__APPLE__)
static void
systrace_init(struct sysent *actual, systrace_sysent_t **interposed)
{
systrace_sysent_t *sysent = *interposed;
int i;
if (sysent == NULL) {
*interposed = sysent = kmem_zalloc(sizeof (systrace_sysent_t) *
NSYSCALL, KM_SLEEP);
}
for (i = 0; i < NSYSCALL; i++) {
struct sysent *a = &actual[i];
systrace_sysent_t *s = &sysent[i];
if (LOADABLE_SYSCALL(a) && !LOADED_SYSCALL(a))
continue;
if (a->sy_callc == dtrace_systrace_syscall)
continue;
#ifdef _SYSCALL32_IMPL
if (a->sy_callc == dtrace_systrace_syscall32)
continue;
#endif
s->stsy_underlying = a->sy_callc;
}
}
#else
#define systrace_init _systrace_init /* Avoid name clash with Darwin automagic conf symbol */
static void
systrace_init(struct sysent *actual, systrace_sysent_t **interposed)
{
systrace_sysent_t *ssysent = *interposed; /* Avoid sysent shadow warning
from bsd/sys/sysent.h */
int i;
if (ssysent == NULL) {
*interposed = ssysent = kmem_zalloc(sizeof (systrace_sysent_t) *
NSYSCALL, KM_SLEEP);
}
for (i = 0; i < NSYSCALL; i++) {
struct sysent *a = &actual[i];
systrace_sysent_t *s = &ssysent[i];
if (LOADABLE_SYSCALL(a) && !LOADED_SYSCALL(a))
continue;
if (a->sy_callc == dtrace_systrace_syscall)
continue;
#ifdef _SYSCALL32_IMPL
if (a->sy_callc == dtrace_systrace_syscall32)
continue;
#endif
s->stsy_underlying = a->sy_callc;
s->stsy_return_type = a->sy_return_type;
}
lck_mtx_init(&dtrace_systrace_lock, dtrace_lck_grp, dtrace_lck_attr);
}
void
kpc_common_init(void)
{
kpc_config_lckgrp_attr = lck_grp_attr_alloc_init();
kpc_config_lckgrp = lck_grp_alloc_init("kpc", kpc_config_lckgrp_attr);
lck_mtx_init(&kpc_config_lock, kpc_config_lckgrp, LCK_ATTR_NULL);
}
__private_extern__
void
hfs_chashinit_finish(struct hfsmount *hfsmp)
{
lck_mtx_init(&hfsmp->hfs_chash_mutex, chash_lck_grp, chash_lck_attr);
hfsmp->hfs_cnodehashtbl = hashinit(desiredvnodes / 4, M_HFSMNT, &hfsmp->hfs_cnodehash);
}
void
init_system_override()
{
sys_override_mtx_grp_attr = lck_grp_attr_alloc_init();
sys_override_mtx_grp = lck_grp_alloc_init("system_override", sys_override_mtx_grp_attr);
sys_override_mtx_attr = lck_attr_alloc_init();
lck_mtx_init(&sys_override_lock, sys_override_mtx_grp, sys_override_mtx_attr);
io_throttle_assert_cnt = cpu_throttle_assert_cnt = 0;
}
void
OSMalloc_init(
void)
{
queue_init(&OSMalloc_tag_list);
OSMalloc_tag_lck_grp = lck_grp_alloc_init("OSMalloc_tag", LCK_GRP_ATTR_NULL);
lck_mtx_init(&OSMalloc_tag_lock, OSMalloc_tag_lck_grp, LCK_ATTR_NULL);
}
/*
* thread_call_initialize:
*
* Initialize this module, called
* early during system initialization.
*/
void
thread_call_initialize(void)
{
thread_call_t call;
thread_call_group_t group = &thread_call_group0;
kern_return_t result;
thread_t thread;
int i;
spl_t s;
i = sizeof (thread_call_data_t);
thread_call_zone = zinit(i, 4096 * i, 16 * i, "thread_call");
zone_change(thread_call_zone, Z_CALLERACCT, FALSE);
zone_change(thread_call_zone, Z_NOENCRYPT, TRUE);
lck_attr_setdefault(&thread_call_lck_attr);
lck_grp_attr_setdefault(&thread_call_lck_grp_attr);
lck_grp_init(&thread_call_queues_lck_grp, "thread_call_queues", &thread_call_lck_grp_attr);
lck_grp_init(&thread_call_lck_grp, "thread_call", &thread_call_lck_grp_attr);
#if defined(__i386__) || defined(__x86_64__)
lck_mtx_init(&thread_call_lock_data, &thread_call_lck_grp, &thread_call_lck_attr);
#else
lck_spin_init(&thread_call_lock_data, &thread_call_lck_grp, &thread_call_lck_attr);
#endif
queue_init(&group->pending_queue);
queue_init(&group->delayed_queue);
s = splsched();
thread_call_lock_spin();
timer_call_setup(&group->delayed_timer, thread_call_delayed_timer, group);
wait_queue_init(&group->idle_wqueue, SYNC_POLICY_FIFO);
wait_queue_init(&group->daemon_wqueue, SYNC_POLICY_FIFO);
queue_init(&thread_call_internal_queue);
for (
call = internal_call_storage;
call < &internal_call_storage[internal_call_count];
call++) {
enqueue_tail(&thread_call_internal_queue, qe(call));
}
thread_call_daemon_awake = TRUE;
thread_call_unlock();
splx(s);
result = kernel_thread_start_priority((thread_continue_t)thread_call_daemon, group, BASEPRI_PREEMPT + 1, &thread);
if (result != KERN_SUCCESS)
panic("thread_call_initialize");
thread_deallocate(thread);
}
void telemetry_init(void)
{
kern_return_t ret;
uint32_t telemetry_notification_leeway;
lck_grp_init(&telemetry_lck_grp, "telemetry group", LCK_GRP_ATTR_NULL);
lck_mtx_init(&telemetry_mtx, &telemetry_lck_grp, LCK_ATTR_NULL);
if (!PE_parse_boot_argn("telemetry_buffer_size", &telemetry_buffer.size, sizeof(telemetry_buffer.size))) {
telemetry_buffer.size = TELEMETRY_DEFAULT_BUFFER_SIZE;
}
if (telemetry_buffer.size > TELEMETRY_MAX_BUFFER_SIZE)
telemetry_buffer.size = TELEMETRY_MAX_BUFFER_SIZE;
ret = kmem_alloc(kernel_map, &telemetry_buffer.buffer, telemetry_buffer.size, VM_KERN_MEMORY_DIAG);
if (ret != KERN_SUCCESS) {
kprintf("Telemetry: Allocation failed: %d\n", ret);
return;
}
bzero((void *) telemetry_buffer.buffer, telemetry_buffer.size);
if (!PE_parse_boot_argn("telemetry_notification_leeway", &telemetry_notification_leeway, sizeof(telemetry_notification_leeway))) {
/*
* By default, notify the user to collect the buffer when there is this much space left in the buffer.
*/
telemetry_notification_leeway = TELEMETRY_DEFAULT_NOTIFY_LEEWAY;
}
if (telemetry_notification_leeway >= telemetry_buffer.size) {
printf("telemetry: nonsensical telemetry_notification_leeway boot-arg %d changed to %d\n",
telemetry_notification_leeway, TELEMETRY_DEFAULT_NOTIFY_LEEWAY);
telemetry_notification_leeway = TELEMETRY_DEFAULT_NOTIFY_LEEWAY;
}
telemetry_buffer_notify_at = telemetry_buffer.size - telemetry_notification_leeway;
if (!PE_parse_boot_argn("telemetry_sample_rate", &telemetry_sample_rate, sizeof(telemetry_sample_rate))) {
telemetry_sample_rate = TELEMETRY_DEFAULT_SAMPLE_RATE;
}
/*
* To enable telemetry for all tasks, include "telemetry_sample_all_tasks=1" in boot-args.
*/
if (!PE_parse_boot_argn("telemetry_sample_all_tasks", &telemetry_sample_all_tasks, sizeof(telemetry_sample_all_tasks))) {
#if CONFIG_EMBEDDED && !(DEVELOPMENT || DEBUG)
telemetry_sample_all_tasks = FALSE;
#else
telemetry_sample_all_tasks = TRUE;
#endif /* CONFIG_EMBEDDED && !(DEVELOPMENT || DEBUG) */
}
kprintf("Telemetry: Sampling %stasks once per %u second%s\n",
(telemetry_sample_all_tasks) ? "all " : "",
telemetry_sample_rate, telemetry_sample_rate == 1 ? "" : "s");
}
/* Initialize the mutex governing access to the SysV msg subsystem */
__private_extern__ void
sysv_msg_lock_init( void )
{
sysv_msg_subsys_lck_grp_attr = lck_grp_attr_alloc_init();
sysv_msg_subsys_lck_grp = lck_grp_alloc_init("sysv_msg_subsys_lock", sysv_msg_subsys_lck_grp_attr);
sysv_msg_subsys_lck_attr = lck_attr_alloc_init();
lck_mtx_init(&sysv_msg_subsys_mutex, sysv_msg_subsys_lck_grp, sysv_msg_subsys_lck_attr);
}
lck_mtx_t *
lck_mtx_alloc_init(lck_grp_t *grp, lck_attr_t *attr)
{
lck_mtx_t *lck;
if ((lck = (lck_mtx_t *)kalloc(sizeof(lck_mtx_t))) != 0)
lck_mtx_init(lck, grp, attr);
return(lck);
}
int
nullfs_init_lck(lck_mtx_t * lck)
{
int error = 1;
if (lck && null_hashlck_grp && null_hashlck_attr) {
lck_mtx_init(lck, null_hashlck_grp, null_hashlck_attr);
error = 0;
}
return error;
}
void
kperf_bootstrap(void)
{
kperf_cfg_lckgrp_attr = lck_grp_attr_alloc_init();
kperf_cfg_lckgrp = lck_grp_alloc_init("kperf cfg",
kperf_cfg_lckgrp_attr);
lck_mtx_init(&kperf_cfg_lock, kperf_cfg_lckgrp, LCK_ATTR_NULL);
kperf_cfg_initted = TRUE;
}
void
eventhandler_lists_ctxt_init(struct eventhandler_lists_ctxt *evthdlr_lists_ctxt)
{
VERIFY(evthdlr_lists_ctxt != NULL);
TAILQ_INIT(&evthdlr_lists_ctxt->eventhandler_lists);
evthdlr_lists_ctxt->eventhandler_lists_initted = 1;
lck_mtx_init(&evthdlr_lists_ctxt->eventhandler_mutex,
eventhandler_mutex_grp, eventhandler_mutex_attr);
}
/*
* thread_call_initialize:
*
* Initialize this module, called
* early during system initialization.
*/
void
thread_call_initialize(void)
{
thread_call_t call;
kern_return_t result;
thread_t thread;
int i;
i = sizeof (thread_call_data_t);
thread_call_zone = zinit(i, 4096 * i, 16 * i, "thread_call");
zone_change(thread_call_zone, Z_CALLERACCT, FALSE);
zone_change(thread_call_zone, Z_NOENCRYPT, TRUE);
lck_attr_setdefault(&thread_call_lck_attr);
lck_grp_attr_setdefault(&thread_call_lck_grp_attr);
lck_grp_init(&thread_call_queues_lck_grp, "thread_call_queues", &thread_call_lck_grp_attr);
lck_grp_init(&thread_call_lck_grp, "thread_call", &thread_call_lck_grp_attr);
#if defined(__i386__) || defined(__x86_64__)
lck_mtx_init(&thread_call_lock_data, &thread_call_lck_grp, &thread_call_lck_attr);
#else
lck_spin_init(&thread_call_lock_data, &thread_call_lck_grp, &thread_call_lck_attr);
#endif
nanotime_to_absolutetime(0, THREAD_CALL_DEALLOC_INTERVAL_NS, &thread_call_dealloc_interval_abs);
wait_queue_init(&daemon_wqueue, SYNC_POLICY_FIFO);
thread_call_group_setup(&thread_call_groups[THREAD_CALL_PRIORITY_LOW], THREAD_CALL_PRIORITY_LOW, 0, TRUE);
thread_call_group_setup(&thread_call_groups[THREAD_CALL_PRIORITY_USER], THREAD_CALL_PRIORITY_USER, 0, TRUE);
thread_call_group_setup(&thread_call_groups[THREAD_CALL_PRIORITY_KERNEL], THREAD_CALL_PRIORITY_KERNEL, 1, TRUE);
thread_call_group_setup(&thread_call_groups[THREAD_CALL_PRIORITY_HIGH], THREAD_CALL_PRIORITY_HIGH, THREAD_CALL_THREAD_MIN, FALSE);
disable_ints_and_lock();
queue_init(&thread_call_internal_queue);
for (
call = internal_call_storage;
call < &internal_call_storage[INTERNAL_CALL_COUNT];
call++) {
enqueue_tail(&thread_call_internal_queue, qe(call));
}
thread_call_daemon_awake = TRUE;
enable_ints_and_unlock();
result = kernel_thread_start_priority((thread_continue_t)thread_call_daemon, NULL, BASEPRI_PREEMPT + 1, &thread);
if (result != KERN_SUCCESS)
panic("thread_call_initialize");
thread_deallocate(thread);
}
void ipc_host_init(void)
{
ipc_port_t port;
int i;
lck_mtx_init(&realhost.lock, &host_notify_lock_grp, &host_notify_lock_attr);
/*
* Allocate and set up the two host ports.
*/
port = ipc_port_alloc_kernel();
if (port == IP_NULL)
panic("ipc_host_init");
ipc_kobject_set(port, (ipc_kobject_t) &realhost, IKOT_HOST_SECURITY);
kernel_set_special_port(&realhost, HOST_SECURITY_PORT,
ipc_port_make_send(port));
port = ipc_port_alloc_kernel();
if (port == IP_NULL)
panic("ipc_host_init");
ipc_kobject_set(port, (ipc_kobject_t) &realhost, IKOT_HOST);
kernel_set_special_port(&realhost, HOST_PORT,
ipc_port_make_send(port));
port = ipc_port_alloc_kernel();
if (port == IP_NULL)
panic("ipc_host_init");
ipc_kobject_set(port, (ipc_kobject_t) &realhost, IKOT_HOST_PRIV);
kernel_set_special_port(&realhost, HOST_PRIV_PORT,
ipc_port_make_send(port));
/* the rest of the special ports will be set up later */
for (i = FIRST_EXCEPTION; i < EXC_TYPES_COUNT; i++) {
realhost.exc_actions[i].port = IP_NULL;
}/* for */
/*
* Set up ipc for default processor set.
*/
ipc_pset_init(&pset0);
ipc_pset_enable(&pset0);
/*
* And for master processor
*/
ipc_processor_init(master_processor);
ipc_processor_enable(master_processor);
}
/*
* Initialise cache headers
*/
int
nullfs_init(__unused struct vfsconf * vfsp)
{
NULLFSDEBUG("%s\n", __FUNCTION__);
/* assuming for now that this happens immediately and by default after fs
* installation */
null_hashlck_grp_attr = lck_grp_attr_alloc_init();
if (null_hashlck_grp_attr == NULL) {
goto error;
}
null_hashlck_grp = lck_grp_alloc_init("com.apple.filesystems.nullfs", null_hashlck_grp_attr);
if (null_hashlck_grp == NULL) {
goto error;
}
null_hashlck_attr = lck_attr_alloc_init();
if (null_hashlck_attr == NULL) {
goto error;
}
lck_mtx_init(&null_hashmtx, null_hashlck_grp, null_hashlck_attr);
null_node_hashtbl = hashinit(NULL_HASH_SIZE, M_TEMP, &null_hash_mask);
NULLFSDEBUG("%s finished\n", __FUNCTION__);
return (0);
error:
printf("NULLFS: failed to get lock element\n");
if (null_hashlck_grp_attr) {
lck_grp_attr_free(null_hashlck_grp_attr);
null_hashlck_grp_attr = NULL;
}
if (null_hashlck_grp) {
lck_grp_free(null_hashlck_grp);
null_hashlck_grp = NULL;
}
if (null_hashlck_attr) {
lck_attr_free(null_hashlck_attr);
null_hashlck_attr = NULL;
}
return KERN_FAILURE;
}
/*
* Creates a new reason and initializes it with the provided reason
* namespace and code. Also sets up the buffer and kcdata_descriptor
* associated with the reason. Returns a pointer to the newly created
* reason.
*
* Returns:
* REASON_NULL if unable to allocate a reason or initialize the nested buffer
* a pointer to the reason otherwise
*/
os_reason_t
os_reason_create(uint32_t osr_namespace, uint64_t osr_code)
{
os_reason_t new_reason = OS_REASON_NULL;
new_reason = (os_reason_t) zalloc(os_reason_zone);
if (new_reason == OS_REASON_NULL) {
#if OS_REASON_DEBUG
/*
* We rely on OS reasons to communicate important things such
* as process exit reason information, we should be aware
* when issues prevent us from allocating them.
*/
if (os_reason_debug_disabled) {
kprintf("os_reason_create: failed to allocate reason with namespace: %u, code : %llu\n",
osr_namespace, osr_code);
} else {
panic("os_reason_create: failed to allocate reason with namespace: %u, code: %llu\n",
osr_namespace, osr_code);
}
#endif
return new_reason;
}
bzero(new_reason, sizeof(*new_reason));
new_reason->osr_namespace = osr_namespace;
new_reason->osr_code = osr_code;
new_reason->osr_flags = 0;
new_reason->osr_bufsize = 0;
new_reason->osr_kcd_buf = NULL;
lck_mtx_init(&new_reason->osr_lock, os_reason_lock_grp, os_reason_lock_attr);
new_reason->osr_refcount = 1;
return new_reason;
}
/*
* Initialise reassembly queue and fragment identifier.
*/
void
frag6_init(void)
{
/* ip6q_alloc() uses mbufs for IPv6 fragment queue structures */
_CASSERT(sizeof (struct ip6q) <= _MLEN);
/* ip6af_alloc() uses mbufs for IPv6 fragment queue structures */
_CASSERT(sizeof (struct ip6asfrag) <= _MLEN);
/* IPv6 fragment reassembly queue lock */
ip6qlock_grp_attr = lck_grp_attr_alloc_init();
ip6qlock_grp = lck_grp_alloc_init("ip6qlock", ip6qlock_grp_attr);
ip6qlock_attr = lck_attr_alloc_init();
lck_mtx_init(&ip6qlock, ip6qlock_grp, ip6qlock_attr);
lck_mtx_lock(&ip6qlock);
/* Initialize IPv6 reassembly queue. */
ip6q.ip6q_next = ip6q.ip6q_prev = &ip6q;
/* same limits as IPv4 */
ip6_maxfragpackets = nmbclusters / 32;
ip6_maxfrags = ip6_maxfragpackets * 2;
ip6q_updateparams();
lck_mtx_unlock(&ip6qlock);
}
/*
* This function is called very early on in the Mach startup, from the
* function start_kernel_threads() in osfmk/kern/startup.c. It's called
* in the context of the current (startup) task using a call to the
* function kernel_thread_create() to jump into start_kernel_threads().
* Internally, kernel_thread_create() calls thread_create_internal(),
* which calls uthread_alloc(). The function of uthread_alloc() is
* normally to allocate a uthread structure, and fill out the uu_sigmask,
* uu_context fields. It skips filling these out in the case of the "task"
* being "kernel_task", because the order of operation is inverted. To
* account for that, we need to manually fill in at least the contents
* of the uu_context.vc_ucred field so that the uthread structure can be
* used like any other.
*/
void
bsd_init(void)
{
struct uthread *ut;
unsigned int i;
#if __i386__ || __x86_64__
int error;
#endif
struct vfs_context context;
kern_return_t ret;
struct ucred temp_cred;
#define bsd_init_kprintf(x...) /* kprintf("bsd_init: " x) */
kernel_flock = funnel_alloc(KERNEL_FUNNEL);
if (kernel_flock == (funnel_t *)0 ) {
panic("bsd_init: Failed to allocate kernel funnel");
}
printf(copyright);
bsd_init_kprintf("calling kmeminit\n");
kmeminit();
bsd_init_kprintf("calling parse_bsd_args\n");
parse_bsd_args();
/* Initialize kauth subsystem before instancing the first credential */
bsd_init_kprintf("calling kauth_init\n");
kauth_init();
/* Initialize process and pgrp structures. */
bsd_init_kprintf("calling procinit\n");
procinit();
/* Initialize the ttys (MUST be before kminit()/bsd_autoconf()!)*/
tty_init();
kernproc = &proc0; /* implicitly bzero'ed */
/* kernel_task->proc = kernproc; */
set_bsdtask_info(kernel_task,(void *)kernproc);
/* give kernproc a name */
bsd_init_kprintf("calling process_name\n");
process_name("kernel_task", kernproc);
/* allocate proc lock group attribute and group */
bsd_init_kprintf("calling lck_grp_attr_alloc_init\n");
proc_lck_grp_attr= lck_grp_attr_alloc_init();
proc_lck_grp = lck_grp_alloc_init("proc", proc_lck_grp_attr);
#ifndef CONFIG_EMBEDDED
proc_slock_grp = lck_grp_alloc_init("proc-slock", proc_lck_grp_attr);
proc_fdmlock_grp = lck_grp_alloc_init("proc-fdmlock", proc_lck_grp_attr);
proc_mlock_grp = lck_grp_alloc_init("proc-mlock", proc_lck_grp_attr);
#endif
/* Allocate proc lock attribute */
proc_lck_attr = lck_attr_alloc_init();
#if 0
#if __PROC_INTERNAL_DEBUG
lck_attr_setdebug(proc_lck_attr);
#endif
#endif
#ifdef CONFIG_EMBEDDED
proc_list_mlock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr);
proc_klist_mlock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_mlock, proc_lck_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_fdmlock, proc_lck_grp, proc_lck_attr);
lck_spin_init(&kernproc->p_slock, proc_lck_grp, proc_lck_attr);
#else
proc_list_mlock = lck_mtx_alloc_init(proc_mlock_grp, proc_lck_attr);
proc_klist_mlock = lck_mtx_alloc_init(proc_mlock_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_mlock, proc_mlock_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_fdmlock, proc_fdmlock_grp, proc_lck_attr);
lck_spin_init(&kernproc->p_slock, proc_slock_grp, proc_lck_attr);
#endif
execargs_cache_lock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr);
execargs_cache_size = bsd_simul_execs;
execargs_free_count = bsd_simul_execs;
execargs_cache = (vm_offset_t *)kalloc(bsd_simul_execs * sizeof(vm_offset_t));
bzero(execargs_cache, bsd_simul_execs * sizeof(vm_offset_t));
if (current_task() != kernel_task)
printf("bsd_init: We have a problem, "
"current task is not kernel task\n");
bsd_init_kprintf("calling get_bsdthread_info\n");
ut = (uthread_t)get_bsdthread_info(current_thread());
#if CONFIG_MACF
/*
* Initialize the MAC Framework
*/
mac_policy_initbsd();
kernproc->p_mac_enforce = 0;
#endif /* MAC */
/*
* Create process 0.
*/
proc_list_lock();
LIST_INSERT_HEAD(&allproc, kernproc, p_list);
kernproc->p_pgrp = &pgrp0;
LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
LIST_INIT(&pgrp0.pg_members);
#ifdef CONFIG_EMBEDDED
lck_mtx_init(&pgrp0.pg_mlock, proc_lck_grp, proc_lck_attr);
#else
lck_mtx_init(&pgrp0.pg_mlock, proc_mlock_grp, proc_lck_attr);
#endif
/* There is no other bsd thread this point and is safe without pgrp lock */
LIST_INSERT_HEAD(&pgrp0.pg_members, kernproc, p_pglist);
kernproc->p_listflag |= P_LIST_INPGRP;
kernproc->p_pgrpid = 0;
pgrp0.pg_session = &session0;
pgrp0.pg_membercnt = 1;
session0.s_count = 1;
session0.s_leader = kernproc;
session0.s_listflags = 0;
#ifdef CONFIG_EMBEDDED
lck_mtx_init(&session0.s_mlock, proc_lck_grp, proc_lck_attr);
#else
lck_mtx_init(&session0.s_mlock, proc_mlock_grp, proc_lck_attr);
#endif
LIST_INSERT_HEAD(SESSHASH(0), &session0, s_hash);
proc_list_unlock();
#if CONFIG_LCTX
kernproc->p_lctx = NULL;
#endif
kernproc->task = kernel_task;
kernproc->p_stat = SRUN;
kernproc->p_flag = P_SYSTEM;
kernproc->p_nice = NZERO;
kernproc->p_pptr = kernproc;
TAILQ_INIT(&kernproc->p_uthlist);
TAILQ_INSERT_TAIL(&kernproc->p_uthlist, ut, uu_list);
kernproc->sigwait = FALSE;
kernproc->sigwait_thread = THREAD_NULL;
kernproc->exit_thread = THREAD_NULL;
kernproc->p_csflags = CS_VALID;
/*
* Create credential. This also Initializes the audit information.
*/
bsd_init_kprintf("calling bzero\n");
bzero(&temp_cred, sizeof(temp_cred));
temp_cred.cr_ngroups = 1;
temp_cred.cr_audit.as_aia_p = &audit_default_aia;
/* XXX the following will go away with cr_au */
temp_cred.cr_au.ai_auid = AU_DEFAUDITID;
bsd_init_kprintf("calling kauth_cred_create\n");
kernproc->p_ucred = kauth_cred_create(&temp_cred);
/* give the (already exisiting) initial thread a reference on it */
bsd_init_kprintf("calling kauth_cred_ref\n");
kauth_cred_ref(kernproc->p_ucred);
ut->uu_context.vc_ucred = kernproc->p_ucred;
ut->uu_context.vc_thread = current_thread();
TAILQ_INIT(&kernproc->p_aio_activeq);
TAILQ_INIT(&kernproc->p_aio_doneq);
kernproc->p_aio_total_count = 0;
kernproc->p_aio_active_count = 0;
bsd_init_kprintf("calling file_lock_init\n");
file_lock_init();
#if CONFIG_MACF
mac_cred_label_associate_kernel(kernproc->p_ucred);
mac_task_label_update_cred (kernproc->p_ucred, (struct task *) kernproc->task);
#endif
/* Create the file descriptor table. */
filedesc0.fd_refcnt = 1+1; /* +1 so shutdown will not _FREE_ZONE */
kernproc->p_fd = &filedesc0;
filedesc0.fd_cmask = cmask;
filedesc0.fd_knlistsize = -1;
filedesc0.fd_knlist = NULL;
filedesc0.fd_knhash = NULL;
filedesc0.fd_knhashmask = 0;
/* Create the limits structures. */
kernproc->p_limit = &limit0;
for (i = 0; i < sizeof(kernproc->p_rlimit)/sizeof(kernproc->p_rlimit[0]); i++)
limit0.pl_rlimit[i].rlim_cur =
limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = maxprocperuid;
limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
limit0.pl_rlimit[RLIMIT_STACK] = vm_initial_limit_stack;
limit0.pl_rlimit[RLIMIT_DATA] = vm_initial_limit_data;
limit0.pl_rlimit[RLIMIT_CORE] = vm_initial_limit_core;
limit0.pl_refcnt = 1;
kernproc->p_stats = &pstats0;
kernproc->p_sigacts = &sigacts0;
/*
* Charge root for two processes: init and mach_init.
*/
bsd_init_kprintf("calling chgproccnt\n");
(void)chgproccnt(0, 1);
/*
* Allocate a kernel submap for pageable memory
* for temporary copying (execve()).
*/
{
vm_offset_t minimum;
bsd_init_kprintf("calling kmem_suballoc\n");
ret = kmem_suballoc(kernel_map,
&minimum,
(vm_size_t)bsd_pageable_map_size,
TRUE,
VM_FLAGS_ANYWHERE,
&bsd_pageable_map);
if (ret != KERN_SUCCESS)
panic("bsd_init: Failed to allocate bsd pageable map");
}
/*
* Initialize buffers and hash links for buffers
*
* SIDE EFFECT: Starts a thread for bcleanbuf_thread(), so must
* happen after a credential has been associated with
* the kernel task.
*/
bsd_init_kprintf("calling bsd_bufferinit\n");
bsd_bufferinit();
/* Initialize the execve() semaphore */
bsd_init_kprintf("calling semaphore_create\n");
if (ret != KERN_SUCCESS)
panic("bsd_init: Failed to create execve semaphore");
/*
* Initialize the calendar.
*/
bsd_init_kprintf("calling IOKitInitializeTime\n");
IOKitInitializeTime();
if (turn_on_log_leaks && !new_nkdbufs)
new_nkdbufs = 200000;
start_kern_tracing(new_nkdbufs);
if (turn_on_log_leaks)
log_leaks = 1;
bsd_init_kprintf("calling ubc_init\n");
ubc_init();
/* Initialize the file systems. */
bsd_init_kprintf("calling vfsinit\n");
vfsinit();
#if SOCKETS
/* Initialize per-CPU cache allocator */
mcache_init();
/* Initialize mbuf's. */
bsd_init_kprintf("calling mbinit\n");
mbinit();
net_str_id_init(); /* for mbuf tags */
#endif /* SOCKETS */
/*
* Initializes security event auditing.
* XXX: Should/could this occur later?
*/
#if CONFIG_AUDIT
bsd_init_kprintf("calling audit_init\n");
audit_init();
#endif
/* Initialize kqueues */
bsd_init_kprintf("calling knote_init\n");
knote_init();
/* Initialize for async IO */
bsd_init_kprintf("calling aio_init\n");
aio_init();
/* Initialize pipes */
bsd_init_kprintf("calling pipeinit\n");
pipeinit();
/* Initialize SysV shm subsystem locks; the subsystem proper is
* initialized through a sysctl.
*/
#if SYSV_SHM
bsd_init_kprintf("calling sysv_shm_lock_init\n");
sysv_shm_lock_init();
#endif
#if SYSV_SEM
bsd_init_kprintf("calling sysv_sem_lock_init\n");
sysv_sem_lock_init();
#endif
#if SYSV_MSG
bsd_init_kprintf("sysv_msg_lock_init\n");
sysv_msg_lock_init();
#endif
bsd_init_kprintf("calling pshm_lock_init\n");
pshm_lock_init();
bsd_init_kprintf("calling psem_lock_init\n");
psem_lock_init();
pthread_init();
/* POSIX Shm and Sem */
bsd_init_kprintf("calling pshm_cache_init\n");
pshm_cache_init();
bsd_init_kprintf("calling psem_cache_init\n");
psem_cache_init();
bsd_init_kprintf("calling time_zone_slock_init\n");
time_zone_slock_init();
/* Stack snapshot facility lock */
stackshot_lock_init();
/*
* Initialize protocols. Block reception of incoming packets
* until everything is ready.
*/
bsd_init_kprintf("calling sysctl_register_fixed\n");
sysctl_register_fixed();
bsd_init_kprintf("calling sysctl_mib_init\n");
sysctl_mib_init();
#if NETWORKING
bsd_init_kprintf("calling dlil_init\n");
dlil_init();
bsd_init_kprintf("calling proto_kpi_init\n");
proto_kpi_init();
#endif /* NETWORKING */
#if SOCKETS
bsd_init_kprintf("calling socketinit\n");
socketinit();
bsd_init_kprintf("calling domaininit\n");
domaininit();
#endif /* SOCKETS */
kernproc->p_fd->fd_cdir = NULL;
kernproc->p_fd->fd_rdir = NULL;
#if CONFIG_EMBEDDED
/* Initialize kernel memory status notifications */
bsd_init_kprintf("calling kern_memorystatus_init\n");
kern_memorystatus_init();
#endif
#ifdef GPROF
/* Initialize kernel profiling. */
kmstartup();
#endif
/* kick off timeout driven events by calling first time */
thread_wakeup(&lbolt);
timeout(lightning_bolt, 0, hz);
bsd_init_kprintf("calling bsd_autoconf\n");
bsd_autoconf();
#if CONFIG_DTRACE
dtrace_postinit();
#endif
/*
* We attach the loopback interface *way* down here to ensure
* it happens after autoconf(), otherwise it becomes the
* "primary" interface.
*/
#include <loop.h>
#if NLOOP > 0
bsd_init_kprintf("calling loopattach\n");
loopattach(); /* XXX */
#endif
#if PFLOG
/* Initialize packet filter log interface */
pfloginit();
#endif /* PFLOG */
#if NETHER > 0
/* Register the built-in dlil ethernet interface family */
bsd_init_kprintf("calling ether_family_init\n");
ether_family_init();
#endif /* ETHER */
#if NETWORKING
/* Call any kext code that wants to run just after network init */
bsd_init_kprintf("calling net_init_run\n");
net_init_run();
/* register user tunnel kernel control handler */
utun_register_control();
#endif /* NETWORKING */
bsd_init_kprintf("calling vnode_pager_bootstrap\n");
vnode_pager_bootstrap();
#if 0
/* XXX Hack for early debug stop */
printf("\nabout to sleep for 10 seconds\n");
IOSleep( 10 * 1000 );
/* Debugger("hello"); */
#endif
bsd_init_kprintf("calling inittodr\n");
inittodr(0);
#if CONFIG_EMBEDDED
{
/* print out early VM statistics */
kern_return_t kr1;
vm_statistics_data_t stat;
mach_msg_type_number_t count;
count = HOST_VM_INFO_COUNT;
kr1 = host_statistics(host_self(),
HOST_VM_INFO,
(host_info_t)&stat,
&count);
kprintf("Mach Virtual Memory Statistics (page size of 4096) bytes\n"
"Pages free:\t\t\t%u.\n"
"Pages active:\t\t\t%u.\n"
"Pages inactive:\t\t\t%u.\n"
"Pages wired down:\t\t%u.\n"
"\"Translation faults\":\t\t%u.\n"
"Pages copy-on-write:\t\t%u.\n"
"Pages zero filled:\t\t%u.\n"
"Pages reactivated:\t\t%u.\n"
"Pageins:\t\t\t%u.\n"
"Pageouts:\t\t\t%u.\n"
"Object cache: %u hits of %u lookups (%d%% hit rate)\n",
stat.free_count,
stat.active_count,
stat.inactive_count,
stat.wire_count,
stat.faults,
stat.cow_faults,
stat.zero_fill_count,
stat.reactivations,
stat.pageins,
stat.pageouts,
stat.hits,
stat.lookups,
(stat.hits == 0) ? 100 :
((stat.lookups * 100) / stat.hits));
}
#endif /* CONFIG_EMBEDDED */
/* Mount the root file system. */
while( TRUE) {
int err;
bsd_init_kprintf("calling setconf\n");
setconf();
bsd_init_kprintf("vfs_mountroot\n");
if (0 == (err = vfs_mountroot()))
break;
rootdevice[0] = '\0';
#if NFSCLIENT
if (mountroot == netboot_mountroot) {
PE_display_icon( 0, "noroot"); /* XXX a netboot-specific icon would be nicer */
vc_progress_set(FALSE, 0);
for (i=1; 1; i*=2) {
printf("bsd_init: failed to mount network root, error %d, %s\n",
err, PE_boot_args());
printf("We are hanging here...\n");
IOSleep(i*60*1000);
}
/*NOTREACHED*/
}
#endif
printf("cannot mount root, errno = %d\n", err);
boothowto |= RB_ASKNAME;
}
IOSecureBSDRoot(rootdevice);
context.vc_thread = current_thread();
context.vc_ucred = kernproc->p_ucred;
mountlist.tqh_first->mnt_flag |= MNT_ROOTFS;
bsd_init_kprintf("calling VFS_ROOT\n");
/* Get the vnode for '/'. Set fdp->fd_fd.fd_cdir to reference it. */
if (VFS_ROOT(mountlist.tqh_first, &rootvnode, &context))
panic("bsd_init: cannot find root vnode: %s", PE_boot_args());
rootvnode->v_flag |= VROOT;
(void)vnode_ref(rootvnode);
(void)vnode_put(rootvnode);
filedesc0.fd_cdir = rootvnode;
#if NFSCLIENT
if (mountroot == netboot_mountroot) {
int err;
/* post mount setup */
if ((err = netboot_setup()) != 0) {
PE_display_icon( 0, "noroot"); /* XXX a netboot-specific icon would be nicer */
vc_progress_set(FALSE, 0);
for (i=1; 1; i*=2) {
printf("bsd_init: NetBoot could not find root, error %d: %s\n",
err, PE_boot_args());
printf("We are hanging here...\n");
IOSleep(i*60*1000);
}
/*NOTREACHED*/
}
}
#endif
#if CONFIG_IMAGEBOOT
/*
* See if a system disk image is present. If so, mount it and
* switch the root vnode to point to it
*/
if(imageboot_needed()) {
int err;
/* An image was found */
if((err = imageboot_setup())) {
/*
* this is not fatal. Keep trying to root
* off the original media
*/
printf("%s: imageboot could not find root, %d\n",
__FUNCTION__, err);
}
}
#endif /* CONFIG_IMAGEBOOT */
/* set initial time; all other resource data is already zero'ed */
microtime(&kernproc->p_start);
kernproc->p_stats->p_start = kernproc->p_start; /* for compat */
#if DEVFS
{
char mounthere[] = "/dev"; /* !const because of internal casting */
bsd_init_kprintf("calling devfs_kernel_mount\n");
devfs_kernel_mount(mounthere);
}
#endif /* DEVFS */
/* Initialize signal state for process 0. */
bsd_init_kprintf("calling siginit\n");
siginit(kernproc);
bsd_init_kprintf("calling bsd_utaskbootstrap\n");
bsd_utaskbootstrap();
#if defined(__LP64__)
kernproc->p_flag |= P_LP64;
printf("Kernel is LP64\n");
#endif
#if __i386__ || __x86_64__
/* this should be done after the root filesystem is mounted */
error = set_archhandler(kernproc, CPU_TYPE_POWERPC);
// 10/30/08 - gab: <rdar://problem/6324501>
// if default 'translate' can't be found, see if the understudy is available
if (ENOENT == error) {
strlcpy(exec_archhandler_ppc.path, kRosettaStandIn_str, MAXPATHLEN);
error = set_archhandler(kernproc, CPU_TYPE_POWERPC);
}
if (error) /* XXX make more generic */
exec_archhandler_ppc.path[0] = 0;
#endif
bsd_init_kprintf("calling mountroot_post_hook\n");
/* invoke post-root-mount hook */
if (mountroot_post_hook != NULL)
mountroot_post_hook();
#if 0 /* not yet */
consider_zone_gc(FALSE);
#endif
bsd_init_kprintf("done\n");
}
/*
* Create a new thread.
* Doesn't start the thread running.
*/
static kern_return_t
thread_create_internal(
task_t parent_task,
integer_t priority,
thread_continue_t continuation,
int options,
#define TH_OPTION_NONE 0x00
#define TH_OPTION_NOCRED 0x01
#define TH_OPTION_NOSUSP 0x02
thread_t *out_thread)
{
thread_t new_thread;
static thread_t first_thread = THREAD_NULL;
/*
* Allocate a thread and initialize static fields
*/
if (first_thread == THREAD_NULL)
new_thread = first_thread = current_thread();
new_thread = (thread_t)zalloc(thread_zone);
if (new_thread == THREAD_NULL)
return (KERN_RESOURCE_SHORTAGE);
if (new_thread != first_thread)
*new_thread = thread_template;
#ifdef MACH_BSD
new_thread->uthread = uthread_alloc(parent_task, new_thread, (options & TH_OPTION_NOCRED) != 0);
if (new_thread->uthread == NULL) {
zfree(thread_zone, new_thread);
return (KERN_RESOURCE_SHORTAGE);
}
#endif /* MACH_BSD */
if (machine_thread_create(new_thread, parent_task) != KERN_SUCCESS) {
#ifdef MACH_BSD
void *ut = new_thread->uthread;
new_thread->uthread = NULL;
/* cred free may not be necessary */
uthread_cleanup(parent_task, ut, parent_task->bsd_info);
uthread_cred_free(ut);
uthread_zone_free(ut);
#endif /* MACH_BSD */
zfree(thread_zone, new_thread);
return (KERN_FAILURE);
}
new_thread->task = parent_task;
thread_lock_init(new_thread);
wake_lock_init(new_thread);
lck_mtx_init(&new_thread->mutex, &thread_lck_grp, &thread_lck_attr);
ipc_thread_init(new_thread);
queue_init(&new_thread->held_ulocks);
new_thread->continuation = continuation;
lck_mtx_lock(&tasks_threads_lock);
task_lock(parent_task);
if ( !parent_task->active || parent_task->halting ||
((options & TH_OPTION_NOSUSP) != 0 &&
parent_task->suspend_count > 0) ||
(parent_task->thread_count >= task_threadmax &&
parent_task != kernel_task) ) {
task_unlock(parent_task);
lck_mtx_unlock(&tasks_threads_lock);
#ifdef MACH_BSD
{
void *ut = new_thread->uthread;
new_thread->uthread = NULL;
uthread_cleanup(parent_task, ut, parent_task->bsd_info);
/* cred free may not be necessary */
uthread_cred_free(ut);
uthread_zone_free(ut);
}
#endif /* MACH_BSD */
ipc_thread_disable(new_thread);
ipc_thread_terminate(new_thread);
lck_mtx_destroy(&new_thread->mutex, &thread_lck_grp);
machine_thread_destroy(new_thread);
zfree(thread_zone, new_thread);
return (KERN_FAILURE);
}
/* New threads inherit any default state on the task */
machine_thread_inherit_taskwide(new_thread, parent_task);
task_reference_internal(parent_task);
if (new_thread->task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_PERTHR_LIMIT) {
/*
* This task has a per-thread CPU limit; make sure this new thread
* gets its limit set too, before it gets out of the kernel.
*/
set_astledger(new_thread);
}
new_thread->t_threadledger = LEDGER_NULL; /* per thread ledger is not inherited */
new_thread->t_ledger = new_thread->task->ledger;
if (new_thread->t_ledger)
ledger_reference(new_thread->t_ledger);
/* Cache the task's map */
new_thread->map = parent_task->map;
/* Chain the thread onto the task's list */
queue_enter(&parent_task->threads, new_thread, thread_t, task_threads);
parent_task->thread_count++;
/* So terminating threads don't need to take the task lock to decrement */
hw_atomic_add(&parent_task->active_thread_count, 1);
/* Protected by the tasks_threads_lock */
new_thread->thread_id = ++thread_unique_id;
queue_enter(&threads, new_thread, thread_t, threads);
threads_count++;
timer_call_setup(&new_thread->wait_timer, thread_timer_expire, new_thread);
timer_call_setup(&new_thread->depress_timer, thread_depress_expire, new_thread);
#if CONFIG_COUNTERS
/*
* If parent task has any reservations, they need to be propagated to this
* thread.
*/
new_thread->t_chud = (TASK_PMC_FLAG == (parent_task->t_chud & TASK_PMC_FLAG)) ?
THREAD_PMC_FLAG : 0U;
#endif
/* Set the thread's scheduling parameters */
new_thread->sched_mode = SCHED(initial_thread_sched_mode)(parent_task);
new_thread->sched_flags = 0;
new_thread->max_priority = parent_task->max_priority;
new_thread->task_priority = parent_task->priority;
new_thread->priority = (priority < 0)? parent_task->priority: priority;
if (new_thread->priority > new_thread->max_priority)
new_thread->priority = new_thread->max_priority;
#if CONFIG_EMBEDDED
if (new_thread->priority < MAXPRI_THROTTLE) {
new_thread->priority = MAXPRI_THROTTLE;
}
#endif /* CONFIG_EMBEDDED */
new_thread->importance =
new_thread->priority - new_thread->task_priority;
#if CONFIG_EMBEDDED
new_thread->saved_importance = new_thread->importance;
/* apple ios daemon starts all threads in darwin background */
if (parent_task->ext_appliedstate.apptype == PROC_POLICY_IOS_APPLE_DAEMON) {
/* Cannot use generic routines here so apply darwin bacground directly */
new_thread->policystate.hw_bg = TASK_POLICY_BACKGROUND_ATTRIBUTE_ALL;
/* set thread self backgrounding */
new_thread->appliedstate.hw_bg = new_thread->policystate.hw_bg;
/* priority will get recomputed suitably bit later */
new_thread->importance = INT_MIN;
/* to avoid changes to many pri compute routines, set the effect of those here */
new_thread->priority = MAXPRI_THROTTLE;
}
#endif /* CONFIG_EMBEDDED */
#if defined(CONFIG_SCHED_TRADITIONAL)
new_thread->sched_stamp = sched_tick;
new_thread->pri_shift = sched_pri_shift;
#endif
SCHED(compute_priority)(new_thread, FALSE);
new_thread->active = TRUE;
*out_thread = new_thread;
{
long dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4;
kdbg_trace_data(parent_task->bsd_info, &dbg_arg2);
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
TRACEDBG_CODE(DBG_TRACE_DATA, 1) | DBG_FUNC_NONE,
(vm_address_t)(uintptr_t)thread_tid(new_thread), dbg_arg2, 0, 0, 0);
kdbg_trace_string(parent_task->bsd_info,
&dbg_arg1, &dbg_arg2, &dbg_arg3, &dbg_arg4);
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
TRACEDBG_CODE(DBG_TRACE_STRING, 1) | DBG_FUNC_NONE,
dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4, 0);
}
DTRACE_PROC1(lwp__create, thread_t, *out_thread);
return (KERN_SUCCESS);
}
void
kalloc_init(
void)
{
kern_return_t retval;
vm_offset_t min;
vm_size_t size, kalloc_map_size;
register int i;
/*
* Scale the kalloc_map_size to physical memory size: stay below
* 1/8th the total zone map size, or 128 MB (for a 32-bit kernel).
*/
kalloc_map_size = (vm_size_t)(sane_size >> 5);
#if !__LP64__
if (kalloc_map_size > KALLOC_MAP_SIZE_MAX)
kalloc_map_size = KALLOC_MAP_SIZE_MAX;
#endif /* !__LP64__ */
if (kalloc_map_size < KALLOC_MAP_SIZE_MIN)
kalloc_map_size = KALLOC_MAP_SIZE_MIN;
retval = kmem_suballoc(kernel_map, &min, kalloc_map_size,
FALSE, VM_FLAGS_ANYWHERE | VM_FLAGS_PERMANENT,
&kalloc_map);
if (retval != KERN_SUCCESS)
panic("kalloc_init: kmem_suballoc failed");
kalloc_map_min = min;
kalloc_map_max = min + kalloc_map_size - 1;
/*
* Ensure that zones up to size 8192 bytes exist.
* This is desirable because messages are allocated
* with kalloc, and messages up through size 8192 are common.
*/
if (PAGE_SIZE < 16*1024)
kalloc_max = 16*1024;
else
kalloc_max = PAGE_SIZE;
kalloc_max_prerounded = kalloc_max / 2 + 1;
/* size it to be more than 16 times kalloc_max (256k) for allocations from kernel map */
kalloc_kernmap_size = (kalloc_max * 16) + 1;
kalloc_largest_allocated = kalloc_kernmap_size;
/*
* Allocate a zone for each size we are going to handle.
* We specify non-paged memory. Don't charge the caller
* for the allocation, as we aren't sure how the memory
* will be handled.
*/
for (i = 0; (size = k_zone_size[i]) < kalloc_max; i++) {
k_zone[i] = zinit(size, k_zone_max[i] * size, size,
k_zone_name[i]);
zone_change(k_zone[i], Z_CALLERACCT, FALSE);
}
/*
* Build the Direct LookUp Table for small allocations
*/
for (i = 0, size = 0; i <= N_K_ZDLUT; i++, size += KALLOC_MINALIGN) {
int zindex = 0;
while ((vm_size_t)k_zone_size[zindex] < size)
zindex++;
if (i == N_K_ZDLUT) {
k_zindex_start = zindex;
break;
}
k_zone_dlut[i] = (int8_t)zindex;
}
#ifdef KALLOC_DEBUG
printf("kalloc_init: k_zindex_start %d\n", k_zindex_start);
/*
* Do a quick synthesis to see how well/badly we can
* find-a-zone for a given size.
* Useful when debugging/tweaking the array of zone sizes.
* Cache misses probably more critical than compare-branches!
*/
for (i = 0; i < (int)N_K_ZONE; i++) {
vm_size_t testsize = (vm_size_t)k_zone_size[i] - 1;
int compare = 0;
int zindex;
if (testsize < MAX_SIZE_ZDLUT) {
compare += 1; /* 'if' (T) */
long dindex = INDEX_ZDLUT(testsize);
zindex = (int)k_zone_dlut[dindex];
} else if (testsize < kalloc_max_prerounded) {
compare += 2; /* 'if' (F), 'if' (T) */
zindex = k_zindex_start;
while ((vm_size_t)k_zone_size[zindex] < testsize) {
zindex++;
compare++; /* 'while' (T) */
}
compare++; /* 'while' (F) */
} else
break; /* not zone-backed */
zone_t z = k_zone[zindex];
printf("kalloc_init: req size %4lu: %11s took %d compare%s\n",
(unsigned long)testsize, z->zone_name, compare,
compare == 1 ? "" : "s");
}
#endif
kalloc_lck_grp = lck_grp_alloc_init("kalloc.large", LCK_GRP_ATTR_NULL);
lck_mtx_init(&kalloc_lock, kalloc_lck_grp, LCK_ATTR_NULL);
OSMalloc_init();
#ifdef MUTEX_ZONE
lck_mtx_zone = zinit(sizeof(struct _lck_mtx_), 1024*256, 4096, "lck_mtx");
#endif
}
/*
* forkproc
*
* Description: Create a new process structure, given a parent process
* structure.
*
* Parameters: parent_proc The parent process
*
* Returns: !NULL The new process structure
* NULL Error (insufficient free memory)
*
* Note: When successful, the newly created process structure is
* partially initialized; if a caller needs to deconstruct the
* returned structure, they must call forkproc_free() to do so.
*/
proc_t
forkproc(proc_t parent_proc)
{
proc_t child_proc; /* Our new process */
static int nextpid = 0, pidwrap = 0, nextpidversion = 0;
int error = 0;
struct session *sessp;
uthread_t parent_uthread = (uthread_t)get_bsdthread_info(current_thread());
MALLOC_ZONE(child_proc, proc_t , sizeof *child_proc, M_PROC, M_WAITOK);
if (child_proc == NULL) {
printf("forkproc: M_PROC zone exhausted\n");
goto bad;
}
/* zero it out as we need to insert in hash */
bzero(child_proc, sizeof *child_proc);
MALLOC_ZONE(child_proc->p_stats, struct pstats *,
sizeof *child_proc->p_stats, M_PSTATS, M_WAITOK);
if (child_proc->p_stats == NULL) {
printf("forkproc: M_SUBPROC zone exhausted (p_stats)\n");
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
MALLOC_ZONE(child_proc->p_sigacts, struct sigacts *,
sizeof *child_proc->p_sigacts, M_SIGACTS, M_WAITOK);
if (child_proc->p_sigacts == NULL) {
printf("forkproc: M_SUBPROC zone exhausted (p_sigacts)\n");
FREE_ZONE(child_proc->p_stats, sizeof *child_proc->p_stats, M_PSTATS);
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
/* allocate a callout for use by interval timers */
child_proc->p_rcall = thread_call_allocate((thread_call_func_t)realitexpire, child_proc);
if (child_proc->p_rcall == NULL) {
FREE_ZONE(child_proc->p_sigacts, sizeof *child_proc->p_sigacts, M_SIGACTS);
FREE_ZONE(child_proc->p_stats, sizeof *child_proc->p_stats, M_PSTATS);
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
/*
* Find an unused PID.
*/
proc_list_lock();
nextpid++;
retry:
/*
* If the process ID prototype has wrapped around,
* restart somewhat above 0, as the low-numbered procs
* tend to include daemons that don't exit.
*/
if (nextpid >= PID_MAX) {
nextpid = 100;
pidwrap = 1;
}
if (pidwrap != 0) {
/* if the pid stays in hash both for zombie and runniing state */
if (pfind_locked(nextpid) != PROC_NULL) {
nextpid++;
goto retry;
}
if (pgfind_internal(nextpid) != PGRP_NULL) {
nextpid++;
goto retry;
}
if (session_find_internal(nextpid) != SESSION_NULL) {
nextpid++;
goto retry;
}
}
nprocs++;
child_proc->p_pid = nextpid;
child_proc->p_idversion = nextpidversion++;
#if 1
if (child_proc->p_pid != 0) {
if (pfind_locked(child_proc->p_pid) != PROC_NULL)
panic("proc in the list already\n");
}
#endif
/* Insert in the hash */
child_proc->p_listflag |= (P_LIST_INHASH | P_LIST_INCREATE);
LIST_INSERT_HEAD(PIDHASH(child_proc->p_pid), child_proc, p_hash);
proc_list_unlock();
/*
* We've identified the PID we are going to use; initialize the new
* process structure.
*/
child_proc->p_stat = SIDL;
child_proc->p_pgrpid = PGRPID_DEAD;
/*
* The zero'ing of the proc was at the allocation time due to need
* for insertion to hash. Copy the section that is to be copied
* directly from the parent.
*/
bcopy(&parent_proc->p_startcopy, &child_proc->p_startcopy,
(unsigned) ((caddr_t)&child_proc->p_endcopy - (caddr_t)&child_proc->p_startcopy));
/*
* Some flags are inherited from the parent.
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
* The p_stats and p_sigacts substructs are set in vm_fork.
*/
child_proc->p_flag = (parent_proc->p_flag & (P_LP64 | P_TRANSLATED | P_AFFINITY));
if (parent_proc->p_flag & P_PROFIL)
startprofclock(child_proc);
/*
* Note that if the current thread has an assumed identity, this
* credential will be granted to the new process.
*/
child_proc->p_ucred = kauth_cred_get_with_ref();
#ifdef CONFIG_EMBEDDED
lck_mtx_init(&child_proc->p_mlock, proc_lck_grp, proc_lck_attr);
lck_mtx_init(&child_proc->p_fdmlock, proc_lck_grp, proc_lck_attr);
#if CONFIG_DTRACE
lck_mtx_init(&child_proc->p_dtrace_sprlock, proc_lck_grp, proc_lck_attr);
#endif
lck_spin_init(&child_proc->p_slock, proc_lck_grp, proc_lck_attr);
#else /* !CONFIG_EMBEDDED */
lck_mtx_init(&child_proc->p_mlock, proc_mlock_grp, proc_lck_attr);
lck_mtx_init(&child_proc->p_fdmlock, proc_fdmlock_grp, proc_lck_attr);
#if CONFIG_DTRACE
lck_mtx_init(&child_proc->p_dtrace_sprlock, proc_lck_grp, proc_lck_attr);
#endif
lck_spin_init(&child_proc->p_slock, proc_slock_grp, proc_lck_attr);
#endif /* !CONFIG_EMBEDDED */
klist_init(&child_proc->p_klist);
if (child_proc->p_textvp != NULLVP) {
/* bump references to the text vnode */
/* Need to hold iocount across the ref call */
if (vnode_getwithref(child_proc->p_textvp) == 0) {
error = vnode_ref(child_proc->p_textvp);
vnode_put(child_proc->p_textvp);
if (error != 0)
child_proc->p_textvp = NULLVP;
}
}
/*
* Copy the parents per process open file table to the child; if
* there is a per-thread current working directory, set the childs
* per-process current working directory to that instead of the
* parents.
*
* XXX may fail to copy descriptors to child
*/
child_proc->p_fd = fdcopy(parent_proc, parent_uthread->uu_cdir);
#if SYSV_SHM
if (parent_proc->vm_shm) {
/* XXX may fail to attach shm to child */
(void)shmfork(parent_proc, child_proc);
}
#endif
/*
* inherit the limit structure to child
*/
proc_limitfork(parent_proc, child_proc);
if (child_proc->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
uint64_t rlim_cur = child_proc->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur;
child_proc->p_rlim_cpu.tv_sec = (rlim_cur > __INT_MAX__) ? __INT_MAX__ : rlim_cur;
}
/* Intialize new process stats, including start time */
/* <rdar://6640543> non-zeroed portion contains garbage AFAICT */
bzero(&child_proc->p_stats->pstat_startzero,
(unsigned) ((caddr_t)&child_proc->p_stats->pstat_endzero -
(caddr_t)&child_proc->p_stats->pstat_startzero));
bzero(&child_proc->p_stats->user_p_prof, sizeof(struct user_uprof));
microtime(&child_proc->p_start);
child_proc->p_stats->p_start = child_proc->p_start; /* for compat */
if (parent_proc->p_sigacts != NULL)
(void)memcpy(child_proc->p_sigacts,
parent_proc->p_sigacts, sizeof *child_proc->p_sigacts);
else
(void)memset(child_proc->p_sigacts, 0, sizeof *child_proc->p_sigacts);
sessp = proc_session(parent_proc);
if (sessp->s_ttyvp != NULL && parent_proc->p_flag & P_CONTROLT)
OSBitOrAtomic(P_CONTROLT, &child_proc->p_flag);
session_rele(sessp);
/*
* block all signals to reach the process.
* no transition race should be occuring with the child yet,
* but indicate that the process is in (the creation) transition.
*/
proc_signalstart(child_proc, 0);
proc_transstart(child_proc, 0);
child_proc->p_pcaction = (parent_proc->p_pcaction) & P_PCMAX;
TAILQ_INIT(&child_proc->p_uthlist);
TAILQ_INIT(&child_proc->p_aio_activeq);
TAILQ_INIT(&child_proc->p_aio_doneq);
/* Inherit the parent flags for code sign */
child_proc->p_csflags = parent_proc->p_csflags;
/*
* All processes have work queue locks; cleaned up by
* reap_child_locked()
*/
workqueue_init_lock(child_proc);
/*
* Copy work queue information
*
* Note: This should probably only happen in the case where we are
* creating a child that is a copy of the parent; since this
* routine is called in the non-duplication case of vfork()
* or posix_spawn(), then this information should likely not
* be duplicated.
*
* <rdar://6640553> Work queue pointers that no longer point to code
*/
child_proc->p_wqthread = parent_proc->p_wqthread;
child_proc->p_threadstart = parent_proc->p_threadstart;
child_proc->p_pthsize = parent_proc->p_pthsize;
child_proc->p_targconc = parent_proc->p_targconc;
if ((parent_proc->p_lflag & P_LREGISTER) != 0) {
child_proc->p_lflag |= P_LREGISTER;
}
child_proc->p_dispatchqueue_offset = parent_proc->p_dispatchqueue_offset;
#if PSYNCH
pth_proc_hashinit(child_proc);
#endif /* PSYNCH */
#if CONFIG_LCTX
child_proc->p_lctx = NULL;
/* Add new process to login context (if any). */
if (parent_proc->p_lctx != NULL) {
/*
* <rdar://6640564> This should probably be delayed in the
* vfork() or posix_spawn() cases.
*/
LCTX_LOCK(parent_proc->p_lctx);
enterlctx(child_proc, parent_proc->p_lctx, 0);
}
#endif
bad:
return(child_proc);
}
void
bsd_init(void)
{
struct uthread *ut;
unsigned int i;
struct vfs_context context;
kern_return_t ret;
struct ucred temp_cred;
struct posix_cred temp_pcred;
#if NFSCLIENT || CONFIG_IMAGEBOOT
boolean_t netboot = FALSE;
#endif
#define bsd_init_kprintf(x...) /* kprintf("bsd_init: " x) */
throttle_init();
printf(copyright);
bsd_init_kprintf("calling kmeminit\n");
kmeminit();
bsd_init_kprintf("calling parse_bsd_args\n");
parse_bsd_args();
#if CONFIG_DEV_KMEM
bsd_init_kprintf("calling dev_kmem_init\n");
dev_kmem_init();
#endif
/* Initialize kauth subsystem before instancing the first credential */
bsd_init_kprintf("calling kauth_init\n");
kauth_init();
/* Initialize process and pgrp structures. */
bsd_init_kprintf("calling procinit\n");
procinit();
/* Initialize the ttys (MUST be before kminit()/bsd_autoconf()!)*/
tty_init();
kernproc = &proc0; /* implicitly bzero'ed */
/* kernel_task->proc = kernproc; */
set_bsdtask_info(kernel_task,(void *)kernproc);
/* give kernproc a name */
bsd_init_kprintf("calling process_name\n");
process_name("kernel_task", kernproc);
/* allocate proc lock group attribute and group */
bsd_init_kprintf("calling lck_grp_attr_alloc_init\n");
proc_lck_grp_attr= lck_grp_attr_alloc_init();
proc_lck_grp = lck_grp_alloc_init("proc", proc_lck_grp_attr);
#if CONFIG_FINE_LOCK_GROUPS
proc_slock_grp = lck_grp_alloc_init("proc-slock", proc_lck_grp_attr);
proc_fdmlock_grp = lck_grp_alloc_init("proc-fdmlock", proc_lck_grp_attr);
proc_ucred_mlock_grp = lck_grp_alloc_init("proc-ucred-mlock", proc_lck_grp_attr);
proc_mlock_grp = lck_grp_alloc_init("proc-mlock", proc_lck_grp_attr);
#endif
/* Allocate proc lock attribute */
proc_lck_attr = lck_attr_alloc_init();
#if 0
#if __PROC_INTERNAL_DEBUG
lck_attr_setdebug(proc_lck_attr);
#endif
#endif
#if CONFIG_FINE_LOCK_GROUPS
proc_list_mlock = lck_mtx_alloc_init(proc_mlock_grp, proc_lck_attr);
proc_klist_mlock = lck_mtx_alloc_init(proc_mlock_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_mlock, proc_mlock_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_fdmlock, proc_fdmlock_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_ucred_mlock, proc_ucred_mlock_grp, proc_lck_attr);
lck_spin_init(&kernproc->p_slock, proc_slock_grp, proc_lck_attr);
#else
proc_list_mlock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr);
proc_klist_mlock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_mlock, proc_lck_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_fdmlock, proc_lck_grp, proc_lck_attr);
lck_mtx_init(&kernproc->p_ucred_mlock, proc_lck_grp, proc_lck_attr);
lck_spin_init(&kernproc->p_slock, proc_lck_grp, proc_lck_attr);
#endif
assert(bsd_simul_execs != 0);
execargs_cache_lock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr);
execargs_cache_size = bsd_simul_execs;
execargs_free_count = bsd_simul_execs;
execargs_cache = (vm_offset_t *)kalloc(bsd_simul_execs * sizeof(vm_offset_t));
bzero(execargs_cache, bsd_simul_execs * sizeof(vm_offset_t));
if (current_task() != kernel_task)
printf("bsd_init: We have a problem, "
"current task is not kernel task\n");
bsd_init_kprintf("calling get_bsdthread_info\n");
ut = (uthread_t)get_bsdthread_info(current_thread());
#if CONFIG_MACF
/*
* Initialize the MAC Framework
*/
mac_policy_initbsd();
kernproc->p_mac_enforce = 0;
#if defined (__i386__) || defined (__x86_64__)
/*
* We currently only support this on i386/x86_64, as that is the
* only lock code we have instrumented so far.
*/
check_policy_init(policy_check_flags);
#endif
#endif /* MAC */
/* Initialize System Override call */
init_system_override();
/*
* Create process 0.
*/
proc_list_lock();
LIST_INSERT_HEAD(&allproc, kernproc, p_list);
kernproc->p_pgrp = &pgrp0;
LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash);
LIST_INIT(&pgrp0.pg_members);
#ifdef CONFIG_FINE_LOCK_GROUPS
lck_mtx_init(&pgrp0.pg_mlock, proc_mlock_grp, proc_lck_attr);
#else
lck_mtx_init(&pgrp0.pg_mlock, proc_lck_grp, proc_lck_attr);
#endif
/* There is no other bsd thread this point and is safe without pgrp lock */
LIST_INSERT_HEAD(&pgrp0.pg_members, kernproc, p_pglist);
kernproc->p_listflag |= P_LIST_INPGRP;
kernproc->p_pgrpid = 0;
kernproc->p_uniqueid = 0;
pgrp0.pg_session = &session0;
pgrp0.pg_membercnt = 1;
session0.s_count = 1;
session0.s_leader = kernproc;
session0.s_listflags = 0;
#ifdef CONFIG_FINE_LOCK_GROUPS
lck_mtx_init(&session0.s_mlock, proc_mlock_grp, proc_lck_attr);
#else
lck_mtx_init(&session0.s_mlock, proc_lck_grp, proc_lck_attr);
#endif
LIST_INSERT_HEAD(SESSHASH(0), &session0, s_hash);
proc_list_unlock();
kernproc->task = kernel_task;
kernproc->p_stat = SRUN;
kernproc->p_flag = P_SYSTEM;
kernproc->p_lflag = 0;
kernproc->p_ladvflag = 0;
#if DEVELOPMENT || DEBUG
if (bootarg_disable_aslr)
kernproc->p_flag |= P_DISABLE_ASLR;
#endif
kernproc->p_nice = NZERO;
kernproc->p_pptr = kernproc;
TAILQ_INIT(&kernproc->p_uthlist);
TAILQ_INSERT_TAIL(&kernproc->p_uthlist, ut, uu_list);
kernproc->sigwait = FALSE;
kernproc->sigwait_thread = THREAD_NULL;
kernproc->exit_thread = THREAD_NULL;
kernproc->p_csflags = CS_VALID;
/*
* Create credential. This also Initializes the audit information.
*/
bsd_init_kprintf("calling bzero\n");
bzero(&temp_cred, sizeof(temp_cred));
bzero(&temp_pcred, sizeof(temp_pcred));
temp_pcred.cr_ngroups = 1;
/* kern_proc, shouldn't call up to DS for group membership */
temp_pcred.cr_flags = CRF_NOMEMBERD;
temp_cred.cr_audit.as_aia_p = audit_default_aia_p;
bsd_init_kprintf("calling kauth_cred_create\n");
/*
* We have to label the temp cred before we create from it to
* properly set cr_ngroups, or the create will fail.
*/
posix_cred_label(&temp_cred, &temp_pcred);
kernproc->p_ucred = kauth_cred_create(&temp_cred);
/* update cred on proc */
PROC_UPDATE_CREDS_ONPROC(kernproc);
/* give the (already exisiting) initial thread a reference on it */
bsd_init_kprintf("calling kauth_cred_ref\n");
kauth_cred_ref(kernproc->p_ucred);
ut->uu_context.vc_ucred = kernproc->p_ucred;
ut->uu_context.vc_thread = current_thread();
TAILQ_INIT(&kernproc->p_aio_activeq);
TAILQ_INIT(&kernproc->p_aio_doneq);
kernproc->p_aio_total_count = 0;
kernproc->p_aio_active_count = 0;
bsd_init_kprintf("calling file_lock_init\n");
file_lock_init();
#if CONFIG_MACF
mac_cred_label_associate_kernel(kernproc->p_ucred);
#endif
/* Create the file descriptor table. */
kernproc->p_fd = &filedesc0;
filedesc0.fd_cmask = cmask;
filedesc0.fd_knlistsize = -1;
filedesc0.fd_knlist = NULL;
filedesc0.fd_knhash = NULL;
filedesc0.fd_knhashmask = 0;
/* Create the limits structures. */
kernproc->p_limit = &limit0;
for (i = 0; i < sizeof(kernproc->p_rlimit)/sizeof(kernproc->p_rlimit[0]); i++)
limit0.pl_rlimit[i].rlim_cur =
limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = maxprocperuid;
limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
limit0.pl_rlimit[RLIMIT_STACK] = vm_initial_limit_stack;
limit0.pl_rlimit[RLIMIT_DATA] = vm_initial_limit_data;
limit0.pl_rlimit[RLIMIT_CORE] = vm_initial_limit_core;
limit0.pl_refcnt = 1;
kernproc->p_stats = &pstats0;
kernproc->p_sigacts = &sigacts0;
/*
* Charge root for one process: launchd.
*/
bsd_init_kprintf("calling chgproccnt\n");
(void)chgproccnt(0, 1);
/*
* Allocate a kernel submap for pageable memory
* for temporary copying (execve()).
*/
{
vm_offset_t minimum;
bsd_init_kprintf("calling kmem_suballoc\n");
assert(bsd_pageable_map_size != 0);
ret = kmem_suballoc(kernel_map,
&minimum,
(vm_size_t)bsd_pageable_map_size,
TRUE,
VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_KERN_MEMORY_BSD),
&bsd_pageable_map);
if (ret != KERN_SUCCESS)
panic("bsd_init: Failed to allocate bsd pageable map");
}
/*
* Initialize buffers and hash links for buffers
*
* SIDE EFFECT: Starts a thread for bcleanbuf_thread(), so must
* happen after a credential has been associated with
* the kernel task.
*/
bsd_init_kprintf("calling bsd_bufferinit\n");
bsd_bufferinit();
/* Initialize the execve() semaphore */
bsd_init_kprintf("calling semaphore_create\n");
if (ret != KERN_SUCCESS)
panic("bsd_init: Failed to create execve semaphore");
/*
* Initialize the calendar.
*/
bsd_init_kprintf("calling IOKitInitializeTime\n");
IOKitInitializeTime();
bsd_init_kprintf("calling ubc_init\n");
ubc_init();
/*
* Initialize device-switches.
*/
bsd_init_kprintf("calling devsw_init() \n");
devsw_init();
/* Initialize the file systems. */
bsd_init_kprintf("calling vfsinit\n");
vfsinit();
#if CONFIG_PROC_UUID_POLICY
/* Initial proc_uuid_policy subsystem */
bsd_init_kprintf("calling proc_uuid_policy_init()\n");
proc_uuid_policy_init();
#endif
#if SOCKETS
/* Initialize per-CPU cache allocator */
mcache_init();
/* Initialize mbuf's. */
bsd_init_kprintf("calling mbinit\n");
mbinit();
net_str_id_init(); /* for mbuf tags */
#endif /* SOCKETS */
/*
* Initializes security event auditing.
* XXX: Should/could this occur later?
*/
#if CONFIG_AUDIT
bsd_init_kprintf("calling audit_init\n");
audit_init();
#endif
/* Initialize kqueues */
bsd_init_kprintf("calling knote_init\n");
knote_init();
/* Initialize for async IO */
bsd_init_kprintf("calling aio_init\n");
aio_init();
/* Initialize pipes */
bsd_init_kprintf("calling pipeinit\n");
pipeinit();
/* Initialize SysV shm subsystem locks; the subsystem proper is
* initialized through a sysctl.
*/
#if SYSV_SHM
bsd_init_kprintf("calling sysv_shm_lock_init\n");
sysv_shm_lock_init();
#endif
#if SYSV_SEM
bsd_init_kprintf("calling sysv_sem_lock_init\n");
sysv_sem_lock_init();
#endif
#if SYSV_MSG
bsd_init_kprintf("sysv_msg_lock_init\n");
sysv_msg_lock_init();
#endif
bsd_init_kprintf("calling pshm_lock_init\n");
pshm_lock_init();
bsd_init_kprintf("calling psem_lock_init\n");
psem_lock_init();
pthread_init();
/* POSIX Shm and Sem */
bsd_init_kprintf("calling pshm_cache_init\n");
pshm_cache_init();
bsd_init_kprintf("calling psem_cache_init\n");
psem_cache_init();
bsd_init_kprintf("calling time_zone_slock_init\n");
time_zone_slock_init();
bsd_init_kprintf("calling select_waitq_init\n");
select_waitq_init();
/*
* Initialize protocols. Block reception of incoming packets
* until everything is ready.
*/
bsd_init_kprintf("calling sysctl_register_fixed\n");
sysctl_register_fixed();
bsd_init_kprintf("calling sysctl_mib_init\n");
sysctl_mib_init();
#if NETWORKING
bsd_init_kprintf("calling dlil_init\n");
dlil_init();
bsd_init_kprintf("calling proto_kpi_init\n");
proto_kpi_init();
#endif /* NETWORKING */
#if SOCKETS
bsd_init_kprintf("calling socketinit\n");
socketinit();
bsd_init_kprintf("calling domaininit\n");
domaininit();
iptap_init();
#if FLOW_DIVERT
flow_divert_init();
#endif /* FLOW_DIVERT */
#endif /* SOCKETS */
kernproc->p_fd->fd_cdir = NULL;
kernproc->p_fd->fd_rdir = NULL;
#if CONFIG_FREEZE
#ifndef CONFIG_MEMORYSTATUS
#error "CONFIG_FREEZE defined without matching CONFIG_MEMORYSTATUS"
#endif
/* Initialise background freezing */
bsd_init_kprintf("calling memorystatus_freeze_init\n");
memorystatus_freeze_init();
#endif
#if CONFIG_MEMORYSTATUS
/* Initialize kernel memory status notifications */
bsd_init_kprintf("calling memorystatus_init\n");
memorystatus_init();
#endif /* CONFIG_MEMORYSTATUS */
bsd_init_kprintf("calling macx_init\n");
macx_init();
bsd_init_kprintf("calling acct_init\n");
acct_init();
#ifdef GPROF
/* Initialize kernel profiling. */
kmstartup();
#endif
bsd_init_kprintf("calling bsd_autoconf\n");
bsd_autoconf();
#if CONFIG_DTRACE
dtrace_postinit();
#endif
/*
* We attach the loopback interface *way* down here to ensure
* it happens after autoconf(), otherwise it becomes the
* "primary" interface.
*/
#include <loop.h>
#if NLOOP > 0
bsd_init_kprintf("calling loopattach\n");
loopattach(); /* XXX */
#endif
#if NGIF
/* Initialize gif interface (after lo0) */
gif_init();
#endif
#if PFLOG
/* Initialize packet filter log interface */
pfloginit();
#endif /* PFLOG */
#if NETHER > 0
/* Register the built-in dlil ethernet interface family */
bsd_init_kprintf("calling ether_family_init\n");
ether_family_init();
#endif /* ETHER */
#if NETWORKING
/* Call any kext code that wants to run just after network init */
bsd_init_kprintf("calling net_init_run\n");
net_init_run();
#if CONTENT_FILTER
cfil_init();
#endif
#if PACKET_MANGLER
pkt_mnglr_init();
#endif
#if NECP
/* Initialize Network Extension Control Policies */
necp_init();
#endif
netagent_init();
/* register user tunnel kernel control handler */
utun_register_control();
#if IPSEC
ipsec_register_control();
#endif /* IPSEC */
netsrc_init();
nstat_init();
tcp_cc_init();
#if MPTCP
mptcp_control_register();
#endif /* MPTCP */
#endif /* NETWORKING */
bsd_init_kprintf("calling vnode_pager_bootstrap\n");
vnode_pager_bootstrap();
bsd_init_kprintf("calling inittodr\n");
inittodr(0);
/* Mount the root file system. */
while( TRUE) {
int err;
bsd_init_kprintf("calling setconf\n");
setconf();
#if NFSCLIENT
netboot = (mountroot == netboot_mountroot);
#endif
bsd_init_kprintf("vfs_mountroot\n");
if (0 == (err = vfs_mountroot()))
break;
rootdevice[0] = '\0';
#if NFSCLIENT
if (netboot) {
PE_display_icon( 0, "noroot"); /* XXX a netboot-specific icon would be nicer */
vc_progress_set(FALSE, 0);
for (i=1; 1; i*=2) {
printf("bsd_init: failed to mount network root, error %d, %s\n",
err, PE_boot_args());
printf("We are hanging here...\n");
IOSleep(i*60*1000);
}
/*NOTREACHED*/
}
#endif
printf("cannot mount root, errno = %d\n", err);
boothowto |= RB_ASKNAME;
}
IOSecureBSDRoot(rootdevice);
context.vc_thread = current_thread();
context.vc_ucred = kernproc->p_ucred;
mountlist.tqh_first->mnt_flag |= MNT_ROOTFS;
bsd_init_kprintf("calling VFS_ROOT\n");
/* Get the vnode for '/'. Set fdp->fd_fd.fd_cdir to reference it. */
if (VFS_ROOT(mountlist.tqh_first, &rootvnode, &context))
panic("bsd_init: cannot find root vnode: %s", PE_boot_args());
rootvnode->v_flag |= VROOT;
(void)vnode_ref(rootvnode);
(void)vnode_put(rootvnode);
filedesc0.fd_cdir = rootvnode;
#if NFSCLIENT
if (netboot) {
int err;
netboot = TRUE;
/* post mount setup */
if ((err = netboot_setup()) != 0) {
PE_display_icon( 0, "noroot"); /* XXX a netboot-specific icon would be nicer */
vc_progress_set(FALSE, 0);
for (i=1; 1; i*=2) {
printf("bsd_init: NetBoot could not find root, error %d: %s\n",
err, PE_boot_args());
printf("We are hanging here...\n");
IOSleep(i*60*1000);
}
/*NOTREACHED*/
}
}
#endif
#if CONFIG_IMAGEBOOT
/*
* See if a system disk image is present. If so, mount it and
* switch the root vnode to point to it
*/
if (netboot == FALSE && imageboot_needed()) {
/*
* An image was found. No turning back: we're booted
* with a kernel from the disk image.
*/
imageboot_setup();
}
#endif /* CONFIG_IMAGEBOOT */
/* set initial time; all other resource data is already zero'ed */
microtime_with_abstime(&kernproc->p_start, &kernproc->p_stats->ps_start);
#if DEVFS
{
char mounthere[] = "/dev"; /* !const because of internal casting */
bsd_init_kprintf("calling devfs_kernel_mount\n");
devfs_kernel_mount(mounthere);
}
#endif /* DEVFS */
/* Initialize signal state for process 0. */
bsd_init_kprintf("calling siginit\n");
siginit(kernproc);
bsd_init_kprintf("calling bsd_utaskbootstrap\n");
bsd_utaskbootstrap();
#if defined(__LP64__)
kernproc->p_flag |= P_LP64;
#endif
pal_kernel_announce();
bsd_init_kprintf("calling mountroot_post_hook\n");
/* invoke post-root-mount hook */
if (mountroot_post_hook != NULL)
mountroot_post_hook();
#if 0 /* not yet */
consider_zone_gc(FALSE);
#endif
bsd_init_kprintf("done\n");
}
void
scope6_init(lck_grp_t *grp, lck_attr_t *attr)
{
bzero(&sid_default, sizeof(sid_default));
lck_mtx_init(&scope6_lock, grp, attr);
}
/*
* Create a new thread.
* Doesn't start the thread running.
*/
static kern_return_t
thread_create_internal(
task_t parent_task,
integer_t priority,
thread_continue_t continuation,
int options,
#define TH_OPTION_NONE 0x00
#define TH_OPTION_NOCRED 0x01
#define TH_OPTION_NOSUSP 0x02
thread_t *out_thread)
{
thread_t new_thread;
static thread_t first_thread;
/*
* Allocate a thread and initialize static fields
*/
if (first_thread == THREAD_NULL)
new_thread = first_thread = current_thread();
else
new_thread = (thread_t)zalloc(thread_zone);
if (new_thread == THREAD_NULL)
return (KERN_RESOURCE_SHORTAGE);
if (new_thread != first_thread)
*new_thread = thread_template;
#ifdef MACH_BSD
new_thread->uthread = uthread_alloc(parent_task, new_thread, (options & TH_OPTION_NOCRED) != 0);
if (new_thread->uthread == NULL) {
zfree(thread_zone, new_thread);
return (KERN_RESOURCE_SHORTAGE);
}
#endif /* MACH_BSD */
if (machine_thread_create(new_thread, parent_task) != KERN_SUCCESS) {
#ifdef MACH_BSD
void *ut = new_thread->uthread;
new_thread->uthread = NULL;
/* cred free may not be necessary */
uthread_cleanup(parent_task, ut, parent_task->bsd_info);
uthread_cred_free(ut);
uthread_zone_free(ut);
#endif /* MACH_BSD */
zfree(thread_zone, new_thread);
return (KERN_FAILURE);
}
new_thread->task = parent_task;
thread_lock_init(new_thread);
wake_lock_init(new_thread);
lck_mtx_init(&new_thread->mutex, &thread_lck_grp, &thread_lck_attr);
ipc_thread_init(new_thread);
queue_init(&new_thread->held_ulocks);
new_thread->continuation = continuation;
lck_mtx_lock(&tasks_threads_lock);
task_lock(parent_task);
if ( !parent_task->active || parent_task->halting ||
((options & TH_OPTION_NOSUSP) != 0 &&
parent_task->suspend_count > 0) ||
(parent_task->thread_count >= task_threadmax &&
parent_task != kernel_task) ) {
task_unlock(parent_task);
lck_mtx_unlock(&tasks_threads_lock);
#ifdef MACH_BSD
{
void *ut = new_thread->uthread;
new_thread->uthread = NULL;
uthread_cleanup(parent_task, ut, parent_task->bsd_info);
/* cred free may not be necessary */
uthread_cred_free(ut);
uthread_zone_free(ut);
}
#endif /* MACH_BSD */
ipc_thread_disable(new_thread);
ipc_thread_terminate(new_thread);
lck_mtx_destroy(&new_thread->mutex, &thread_lck_grp);
machine_thread_destroy(new_thread);
zfree(thread_zone, new_thread);
return (KERN_FAILURE);
}
/* New threads inherit any default state on the task */
machine_thread_inherit_taskwide(new_thread, parent_task);
task_reference_internal(parent_task);
/* Cache the task's map */
new_thread->map = parent_task->map;
/* Chain the thread onto the task's list */
queue_enter(&parent_task->threads, new_thread, thread_t, task_threads);
parent_task->thread_count++;
/* So terminating threads don't need to take the task lock to decrement */
hw_atomic_add(&parent_task->active_thread_count, 1);
/* Protected by the tasks_threads_lock */
new_thread->thread_id = ++thread_unique_id;
queue_enter(&threads, new_thread, thread_t, threads);
threads_count++;
timer_call_setup(&new_thread->wait_timer, thread_timer_expire, new_thread);
timer_call_setup(&new_thread->depress_timer, thread_depress_expire, new_thread);
#if CONFIG_COUNTERS
/*
* If parent task has any reservations, they need to be propagated to this
* thread.
*/
new_thread->t_chud = (TASK_PMC_FLAG == (parent_task->t_chud & TASK_PMC_FLAG)) ?
THREAD_PMC_FLAG : 0U;
#endif
/* Set the thread's scheduling parameters */
if (parent_task != kernel_task)
new_thread->sched_mode |= TH_MODE_TIMESHARE;
new_thread->max_priority = parent_task->max_priority;
new_thread->task_priority = parent_task->priority;
new_thread->priority = (priority < 0)? parent_task->priority: priority;
if (new_thread->priority > new_thread->max_priority)
new_thread->priority = new_thread->max_priority;
new_thread->importance =
new_thread->priority - new_thread->task_priority;
new_thread->sched_stamp = sched_tick;
new_thread->pri_shift = sched_pri_shift;
compute_priority(new_thread, FALSE);
new_thread->active = TRUE;
*out_thread = new_thread;
{
long dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4;
kdbg_trace_data(parent_task->bsd_info, &dbg_arg2);
KERNEL_DEBUG_CONSTANT(
TRACEDBG_CODE(DBG_TRACE_DATA, 1) | DBG_FUNC_NONE,
(vm_address_t)(uintptr_t)thread_tid(new_thread), dbg_arg2, 0, 0, 0);
kdbg_trace_string(parent_task->bsd_info,
&dbg_arg1, &dbg_arg2, &dbg_arg3, &dbg_arg4);
KERNEL_DEBUG_CONSTANT(
TRACEDBG_CODE(DBG_TRACE_STRING, 1) | DBG_FUNC_NONE,
dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4, 0);
}
DTRACE_PROC1(lwp__create, thread_t, *out_thread);
return (KERN_SUCCESS);
}
