/*
* Destroy a process structure that resulted from a call to forkproc(), but
* which must be returned to the system because of a subsequent failure
* preventing it from becoming active.
*
* Parameters: p The incomplete process from forkproc()
*
* Returns: (void)
*
* Note: This function should only be used in an error handler following
* a call to forkproc().
*
* Operations occur in reverse order of those in forkproc().
*/
void
forkproc_free(proc_t p)
{
/* We held signal and a transition locks; drop them */
proc_signalend(p, 0);
proc_transend(p, 0);
/*
* If we have our own copy of the resource limits structure, we
* need to free it. If it's a shared copy, we need to drop our
* reference on it.
*/
proc_limitdrop(p, 0);
p->p_limit = NULL;
#if SYSV_SHM
/* Need to drop references to the shared memory segment(s), if any */
if (p->vm_shm) {
/*
* Use shmexec(): we have no address space, so no mappings
*
* XXX Yes, the routine is badly named.
*/
shmexec(p);
}
#endif
/* Need to undo the effects of the fdcopy(), if any */
fdfree(p);
/*
* Drop the reference on a text vnode pointer, if any
* XXX This code is broken in forkproc(); see <rdar://4256419>;
* XXX if anyone ever uses this field, we will be extremely unhappy.
*/
if (p->p_textvp) {
vnode_rele(p->p_textvp);
p->p_textvp = NULL;
}
/* Stop the profiling clock */
stopprofclock(p);
/* Release the credential reference */
kauth_cred_unref(&p->p_ucred);
proc_list_lock();
/* Decrement the count of processes in the system */
nprocs--;
proc_list_unlock();
thread_call_free(p->p_rcall);
/* Free allocated memory */
FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
proc_checkdeadrefs(p);
FREE_ZONE(p, sizeof *p, M_PROC);
}
static int
sd_callback2(proc_t p, void * args)
{
struct sd_iterargs * sd = (struct sd_iterargs *)args;
int signo = sd->signo;
int setsdstate = sd->setsdstate;
int countproc = sd->countproc;
proc_lock(p);
p->p_shutdownstate = setsdstate;
if (p->p_stat != SZOMB) {
proc_unlock(p);
if (countproc != 0) {
proc_list_lock();
p->p_listflag |= P_LIST_EXITCOUNT;
proc_shutdown_exitcount++;
proc_list_unlock();
}
psignal(p, signo);
if (countproc != 0)
sd->activecount++;
} else
proc_unlock(p);
return(PROC_RETURNED);
}
/*
* cloneproc
*
* Description: Create a new process from a specified process.
*
* Parameters: parent_task The parent task to be cloned, or
* TASK_NULL is task characteristics
* are not to be inherited
* be cloned, or TASK_NULL if the new
* task is not to inherit the VM
* characteristics of the parent
* parent_proc The parent process to be cloned
* inherit_memory True if the child is to inherit
* memory from the parent; if this is
* non-NULL, then the parent_task must
* also be non-NULL
* memstat_internal Whether to track the process in the
* jetsam priority list (if configured)
*
* Returns: !NULL pointer to new child thread
* NULL Failure (unspecified)
*
* Note: On return newly created child process has signal lock held
* to block delivery of signal to it if called with lock set.
* fork() code needs to explicity remove this lock before
* signals can be delivered
*
* In the case of bootstrap, this function can be called from
* bsd_utaskbootstrap() in order to bootstrap the first process;
* the net effect is to provide a uthread structure for the
* kernel process associated with the kernel task.
*
* XXX: Tristating using the value parent_task as the major key
* and inherit_memory as the minor key is something we should
* refactor later; we owe the current semantics, ultimately,
* to the semantics of task_create_internal. For now, we will
* live with this being somewhat awkward.
*/
thread_t
cloneproc(task_t parent_task, coalition_t *parent_coalitions, proc_t parent_proc, int inherit_memory, int memstat_internal)
{
#if !CONFIG_MEMORYSTATUS
#pragma unused(memstat_internal)
#endif
task_t child_task;
proc_t child_proc;
thread_t child_thread = NULL;
if ((child_proc = forkproc(parent_proc)) == NULL) {
/* Failed to allocate new process */
goto bad;
}
child_thread = fork_create_child(parent_task, parent_coalitions, child_proc, inherit_memory, (parent_task == TASK_NULL) ? FALSE : (parent_proc->p_flag & P_LP64));
if (child_thread == NULL) {
/*
* Failed to create thread; now we must deconstruct the new
* process previously obtained from forkproc().
*/
forkproc_free(child_proc);
goto bad;
}
child_task = get_threadtask(child_thread);
if (parent_proc->p_flag & P_LP64) {
task_set_64bit(child_task, TRUE);
OSBitOrAtomic(P_LP64, (UInt32 *)&child_proc->p_flag);
} else {
task_set_64bit(child_task, FALSE);
OSBitAndAtomic(~((uint32_t)P_LP64), (UInt32 *)&child_proc->p_flag);
}
#if CONFIG_MEMORYSTATUS
if (memstat_internal) {
proc_list_lock();
child_proc->p_memstat_state |= P_MEMSTAT_INTERNAL;
proc_list_unlock();
}
#endif
/* make child visible */
pinsertchild(parent_proc, child_proc);
/*
* Make child runnable, set start time.
*/
child_proc->p_stat = SRUN;
bad:
return(child_thread);
}
void
ux_handler_init(void)
{
thread_t thread = THREAD_NULL;
ux_exception_port = MACH_PORT_NULL;
(void) kernel_thread_start((thread_continue_t)ux_handler, NULL, &thread);
thread_deallocate(thread);
proc_list_lock();
if (ux_exception_port == MACH_PORT_NULL) {
(void)msleep(&ux_exception_port, proc_list_mlock, 0, "ux_handler_wait", 0);
}
proc_list_unlock();
}
/*
* fork1
*
* Description: common code used by all new process creation other than the
* bootstrap of the initial process on the system
*
* Parameters: parent_proc parent process of the process being
* child_threadp pointer to location to receive the
* Mach thread_t of the child process
* breated
* kind kind of creation being requested
*
* Notes: Permissable values for 'kind':
*
* PROC_CREATE_FORK Create a complete process which will
* return actively running in both the
* parent and the child; the child copies
* the parent address space.
* PROC_CREATE_SPAWN Create a complete process which will
* return actively running in the parent
* only after returning actively running
* in the child; the child address space
* is newly created by an image activator,
* after which the child is run.
* PROC_CREATE_VFORK Creates a partial process which will
* borrow the parent task, thread, and
* uthread to return running in the child;
* the child address space and other parts
* are lazily created at execve() time, or
* the child is terminated, and the parent
* does not actively run until that
* happens.
*
* At first it may seem strange that we return the child thread
* address rather than process structure, since the process is
* the only part guaranteed to be "new"; however, since we do
* not actualy adjust other references between Mach and BSD (see
* the block diagram above the implementation of vfork()), this
* is the only method which guarantees us the ability to get
* back to the other information.
*/
int
fork1(proc_t parent_proc, thread_t *child_threadp, int kind)
{
thread_t parent_thread = (thread_t)current_thread();
uthread_t parent_uthread = (uthread_t)get_bsdthread_info(parent_thread);
proc_t child_proc = NULL; /* set in switch, but compiler... */
thread_t child_thread = NULL;
uid_t uid;
int count;
int err = 0;
int spawn = 0;
/*
* Although process entries are dynamically created, we still keep
* a global limit on the maximum number we will create. Don't allow
* a nonprivileged user to use the last process; don't let root
* exceed the limit. The variable nprocs is the current number of
* processes, maxproc is the limit.
*/
uid = kauth_cred_get()->cr_ruid;
proc_list_lock();
if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) {
proc_list_unlock();
tablefull("proc");
return (EAGAIN);
}
proc_list_unlock();
/*
* Increment the count of procs running with this uid. Don't allow
* a nonprivileged user to exceed their current limit, which is
* always less than what an rlim_t can hold.
* (locking protection is provided by list lock held in chgproccnt)
*/
count = chgproccnt(uid, 1);
if (uid != 0 &&
(rlim_t)count > parent_proc->p_rlimit[RLIMIT_NPROC].rlim_cur) {
err = EAGAIN;
goto bad;
}
#if CONFIG_MACF
/*
* Determine if MAC policies applied to the process will allow
* it to fork. This is an advisory-only check.
*/
err = mac_proc_check_fork(parent_proc);
if (err != 0) {
goto bad;
}
#endif
switch(kind) {
case PROC_CREATE_VFORK:
/*
* Prevent a vfork while we are in vfork(); we should
* also likely preventing a fork here as well, and this
* check should then be outside the switch statement,
* since the proc struct contents will copy from the
* child and the tash/thread/uthread from the parent in
* that case. We do not support vfork() in vfork()
* because we don't have to; the same non-requirement
* is true of both fork() and posix_spawn() and any
* call other than execve() amd _exit(), but we've
* been historically lenient, so we continue to be so
* (for now).
*
* <rdar://6640521> Probably a source of random panics
*/
if (parent_uthread->uu_flag & UT_VFORK) {
printf("fork1 called within vfork by %s\n", parent_proc->p_comm);
err = EINVAL;
goto bad;
}
/*
* Flag us in progress; if we chose to support vfork() in
* vfork(), we would chain our parent at this point (in
* effect, a stack push). We don't, since we actually want
* to disallow everything not specified in the standard
*/
proc_vfork_begin(parent_proc);
/* The newly created process comes with signal lock held */
if ((child_proc = forkproc(parent_proc)) == NULL) {
/* Failed to allocate new process */
proc_vfork_end(parent_proc);
err = ENOMEM;
goto bad;
}
// XXX BEGIN: wants to move to be common code (and safe)
#if CONFIG_MACF
/*
* allow policies to associate the credential/label that
* we referenced from the parent ... with the child
* JMM - this really isn't safe, as we can drop that
* association without informing the policy in other
* situations (keep long enough to get policies changed)
*/
mac_cred_label_associate_fork(child_proc->p_ucred, child_proc);
#endif
/*
* Propogate change of PID - may get new cred if auditing.
*
* NOTE: This has no effect in the vfork case, since
* child_proc->task != current_task(), but we duplicate it
* because this is probably, ultimately, wrong, since we
* will be running in the "child" which is the parent task
* with the wrong token until we get to the execve() or
* _exit() call; a lot of "undefined" can happen before
* that.
*
* <rdar://6640530> disallow everything but exeve()/_exit()?
*/
set_security_token(child_proc);
AUDIT_ARG(pid, child_proc->p_pid);
AUDIT_SESSION_PROCNEW(child_proc->p_ucred);
// XXX END: wants to move to be common code (and safe)
/*
* BORROW PARENT TASK, THREAD, UTHREAD FOR CHILD
*
* Note: this is where we would "push" state instead of setting
* it for nested vfork() support (see proc_vfork_end() for
* description if issues here).
*/
child_proc->task = parent_proc->task;
child_proc->p_lflag |= P_LINVFORK;
child_proc->p_vforkact = parent_thread;
child_proc->p_stat = SRUN;
parent_uthread->uu_flag |= UT_VFORK;
parent_uthread->uu_proc = child_proc;
parent_uthread->uu_userstate = (void *)act_thread_csave();
parent_uthread->uu_vforkmask = parent_uthread->uu_sigmask;
/* temporarily drop thread-set-id state */
if (parent_uthread->uu_flag & UT_SETUID) {
parent_uthread->uu_flag |= UT_WASSETUID;
parent_uthread->uu_flag &= ~UT_SETUID;
}
/* blow thread state information */
/* XXX is this actually necessary, given syscall return? */
thread_set_child(parent_thread, child_proc->p_pid);
child_proc->p_acflag = AFORK; /* forked but not exec'ed */
/*
* Preserve synchronization semantics of vfork. If
* waiting for child to exec or exit, set P_PPWAIT
* on child, and sleep on our proc (in case of exit).
*/
child_proc->p_lflag |= P_LPPWAIT;
pinsertchild(parent_proc, child_proc); /* set visible */
break;
case PROC_CREATE_SPAWN:
/*
* A spawned process differs from a forked process in that
* the spawned process does not carry around the parents
* baggage with regard to address space copying, dtrace,
* and so on.
*/
spawn = 1;
/* FALLSTHROUGH */
case PROC_CREATE_FORK:
/*
* When we clone the parent process, we are going to inherit
* its task attributes and memory, since when we fork, we
* will, in effect, create a duplicate of it, with only minor
* differences. Contrarily, spawned processes do not inherit.
*/
if ((child_thread = cloneproc(parent_proc->task, parent_proc, spawn ? FALSE : TRUE)) == NULL) {
/* Failed to create thread */
err = EAGAIN;
goto bad;
}
/* copy current thread state into the child thread (only for fork) */
if (!spawn) {
thread_dup(child_thread);
}
/* child_proc = child_thread->task->proc; */
child_proc = (proc_t)(get_bsdtask_info(get_threadtask(child_thread)));
// XXX BEGIN: wants to move to be common code (and safe)
#if CONFIG_MACF
/*
* allow policies to associate the credential/label that
* we referenced from the parent ... with the child
* JMM - this really isn't safe, as we can drop that
* association without informing the policy in other
* situations (keep long enough to get policies changed)
*/
mac_cred_label_associate_fork(child_proc->p_ucred, child_proc);
#endif
/*
* Propogate change of PID - may get new cred if auditing.
*
* NOTE: This has no effect in the vfork case, since
* child_proc->task != current_task(), but we duplicate it
* because this is probably, ultimately, wrong, since we
* will be running in the "child" which is the parent task
* with the wrong token until we get to the execve() or
* _exit() call; a lot of "undefined" can happen before
* that.
*
* <rdar://6640530> disallow everything but exeve()/_exit()?
*/
set_security_token(child_proc);
AUDIT_ARG(pid, child_proc->p_pid);
AUDIT_SESSION_PROCNEW(child_proc->p_ucred);
// XXX END: wants to move to be common code (and safe)
/*
* Blow thread state information; this is what gives the child
* process its "return" value from a fork() call.
*
* Note: this should probably move to fork() proper, since it
* is not relevent to spawn, and the value won't matter
* until we resume the child there. If you are in here
* refactoring code, consider doing this at the same time.
*/
thread_set_child(child_thread, child_proc->p_pid);
child_proc->p_acflag = AFORK; /* forked but not exec'ed */
// <rdar://6598155> dtrace code cleanup needed
#if CONFIG_DTRACE
/*
* This code applies to new processes who are copying the task
* and thread state and address spaces of their parent process.
*/
if (!spawn) {
// <rdar://6598155> call dtrace specific function here instead of all this...
/*
* APPLE NOTE: Solaris does a sprlock() and drops the
* proc_lock here. We're cheating a bit and only taking
* the p_dtrace_sprlock lock. A full sprlock would
* task_suspend the parent.
*/
lck_mtx_lock(&parent_proc->p_dtrace_sprlock);
/*
* Remove all DTrace tracepoints from the child process. We
* need to do this _before_ duplicating USDT providers since
* any associated probes may be immediately enabled.
*/
if (parent_proc->p_dtrace_count > 0) {
dtrace_fasttrap_fork(parent_proc, child_proc);
}
lck_mtx_unlock(&parent_proc->p_dtrace_sprlock);
/*
* Duplicate any lazy dof(s). This must be done while NOT
* holding the parent sprlock! Lock ordering is
* dtrace_dof_mode_lock, then sprlock. It is imperative we
* always call dtrace_lazy_dofs_duplicate, rather than null
* check and call if !NULL. If we NULL test, during lazy dof
* faulting we can race with the faulting code and proceed
* from here to beyond the helpers copy. The lazy dof
* faulting will then fail to copy the helpers to the child
* process.
*/
dtrace_lazy_dofs_duplicate(parent_proc, child_proc);
/*
* Duplicate any helper actions and providers. The SFORKING
* we set above informs the code to enable USDT probes that
* sprlock() may fail because the child is being forked.
*/
/*
* APPLE NOTE: As best I can tell, Apple's sprlock() equivalent
* never fails to find the child. We do not set SFORKING.
*/
if (parent_proc->p_dtrace_helpers != NULL && dtrace_helpers_fork) {
(*dtrace_helpers_fork)(parent_proc, child_proc);
}
}
#endif /* CONFIG_DTRACE */
break;
default:
panic("fork1 called with unknown kind %d", kind);
break;
}
/* return the thread pointer to the caller */
*child_threadp = child_thread;
bad:
/*
* In the error case, we return a 0 value for the returned pid (but
* it is ignored in the trampoline due to the error return); this
* is probably not necessary.
*/
if (err) {
(void)chgproccnt(uid, -1);
}
return (err);
}
/*
* 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);
}
static void
proc_shutdown(void)
{
vfs_context_t ctx = vfs_context_current();
struct proc *p, *self;
int delayterm = 0;
struct sd_filterargs sfargs;
struct sd_iterargs sdargs;
int error = 0;
struct timespec ts;
/*
* Kill as many procs as we can. (Except ourself...)
*/
self = (struct proc *)current_proc();
/*
* Signal the init with SIGTERM so that he does not launch
* new processes
*/
p = proc_find(1);
if (p && p != self) {
psignal(p, SIGTERM);
}
proc_rele(p);
printf("Killing all processes ");
sigterm_loop:
/*
* send SIGTERM to those procs interested in catching one
*/
sfargs.delayterm = delayterm;
sfargs.shutdownstate = 0;
sdargs.signo = SIGTERM;
sdargs.setsdstate = 1;
sdargs.countproc = 1;
sdargs.activecount = 0;
error = 0;
/* post a SIGTERM to all that catch SIGTERM and not marked for delay */
proc_rebootscan(sd_callback1, (void *)&sdargs, sd_filt1, (void *)&sfargs);
if (sdargs.activecount != 0 && proc_shutdown_exitcount!= 0) {
proc_list_lock();
if (proc_shutdown_exitcount != 0) {
/*
* now wait for up to 30 seconds to allow those procs catching SIGTERM
* to digest it
* as soon as these procs have exited, we'll continue on to the next step
*/
ts.tv_sec = 30;
ts.tv_nsec = 0;
error = msleep(&proc_shutdown_exitcount, proc_list_mlock, PWAIT, "shutdownwait", &ts);
if (error != 0) {
for (p = allproc.lh_first; p; p = p->p_list.le_next) {
if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT)
p->p_listflag &= ~P_LIST_EXITCOUNT;
}
for (p = zombproc.lh_first; p; p = p->p_list.le_next) {
if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT)
p->p_listflag &= ~P_LIST_EXITCOUNT;
}
}
}
proc_list_unlock();
}
if (error == ETIMEDOUT) {
/*
* log the names of the unresponsive tasks
*/
proc_list_lock();
for (p = allproc.lh_first; p; p = p->p_list.le_next) {
if (p->p_shutdownstate == 1) {
printf("%s[%d]: didn't act on SIGTERM\n", p->p_comm, p->p_pid);
sd_log(ctx, "%s[%d]: didn't act on SIGTERM\n", p->p_comm, p->p_pid);
}
}
proc_list_unlock();
delay_for_interval(1000 * 5, 1000 * 1000);
}
/*
* send a SIGKILL to all the procs still hanging around
*/
sfargs.delayterm = delayterm;
sfargs.shutdownstate = 2;
sdargs.signo = SIGKILL;
sdargs.setsdstate = 2;
sdargs.countproc = 1;
sdargs.activecount = 0;
/* post a SIGKILL to all that catch SIGTERM and not marked for delay */
proc_rebootscan(sd_callback2, (void *)&sdargs, sd_filt2, (void *)&sfargs);
if (sdargs.activecount != 0 && proc_shutdown_exitcount!= 0) {
proc_list_lock();
if (proc_shutdown_exitcount != 0) {
/*
* wait for up to 60 seconds to allow these procs to exit normally
*
* History: The delay interval was changed from 100 to 200
* for NFS requests in particular.
*/
ts.tv_sec = 60;
ts.tv_nsec = 0;
error = msleep(&proc_shutdown_exitcount, proc_list_mlock, PWAIT, "shutdownwait", &ts);
if (error != 0) {
for (p = allproc.lh_first; p; p = p->p_list.le_next) {
if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT)
p->p_listflag &= ~P_LIST_EXITCOUNT;
}
for (p = zombproc.lh_first; p; p = p->p_list.le_next) {
if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT)
p->p_listflag &= ~P_LIST_EXITCOUNT;
}
}
}
proc_list_unlock();
}
/*
* if we still have procs that haven't exited, then brute force 'em
*/
sfargs.delayterm = delayterm;
sfargs.shutdownstate = 3;
sdargs.signo = 0;
sdargs.setsdstate = 3;
sdargs.countproc = 0;
sdargs.activecount = 0;
/* post a SIGTERM to all that catch SIGTERM and not marked for delay */
proc_rebootscan(sd_callback3, (void *)&sdargs, sd_filt2, (void *)&sfargs);
printf("\n");
/* Now start the termination of processes that are marked for delayed termn */
if (delayterm == 0) {
delayterm = 1;
goto sigterm_loop;
}
sd_closelog(ctx);
/* drop the ref on initproc */
proc_rele(initproc);
printf("continuing\n");
}
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");
}
int
ptrace(struct proc *p, struct ptrace_args *uap, int32_t *retval)
{
struct proc *t = current_proc(); /* target process */
task_t task;
thread_t th_act;
struct uthread *ut;
int tr_sigexc = 0;
int error = 0;
int stopped = 0;
AUDIT_ARG(cmd, uap->req);
AUDIT_ARG(pid, uap->pid);
AUDIT_ARG(addr, uap->addr);
AUDIT_ARG(value32, uap->data);
if (uap->req == PT_DENY_ATTACH) {
proc_lock(p);
if (ISSET(p->p_lflag, P_LTRACED)) {
proc_unlock(p);
KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_FRCEXIT) | DBG_FUNC_NONE,
p->p_pid, W_EXITCODE(ENOTSUP, 0), 4, 0, 0);
exit1(p, W_EXITCODE(ENOTSUP, 0), retval);
thread_exception_return();
/* NOTREACHED */
}
SET(p->p_lflag, P_LNOATTACH);
proc_unlock(p);
return(0);
}
if (uap->req == PT_FORCEQUOTA) {
if (kauth_cred_issuser(kauth_cred_get())) {
OSBitOrAtomic(P_FORCEQUOTA, &t->p_flag);
return (0);
} else
return (EPERM);
}
/*
* Intercept and deal with "please trace me" request.
*/
if (uap->req == PT_TRACE_ME) {
retry_trace_me:;
proc_t pproc = proc_parent(p);
if (pproc == NULL)
return (EINVAL);
#if CONFIG_MACF
/*
* NB: Cannot call kauth_authorize_process(..., KAUTH_PROCESS_CANTRACE, ...)
* since that assumes the process being checked is the current process
* when, in this case, it is the current process's parent.
* Most of the other checks in cantrace() don't apply either.
*/
if ((error = mac_proc_check_debug(pproc, p)) == 0) {
#endif
proc_lock(p);
/* Make sure the process wasn't re-parented. */
if (p->p_ppid != pproc->p_pid) {
proc_unlock(p);
proc_rele(pproc);
goto retry_trace_me;
}
SET(p->p_lflag, P_LTRACED);
/* Non-attached case, our tracer is our parent. */
p->p_oppid = p->p_ppid;
proc_unlock(p);
/* Child and parent will have to be able to run modified code. */
cs_allow_invalid(p);
cs_allow_invalid(pproc);
#if CONFIG_MACF
}
#endif
proc_rele(pproc);
return (error);
}
if (uap->req == PT_SIGEXC) {
proc_lock(p);
if (ISSET(p->p_lflag, P_LTRACED)) {
SET(p->p_lflag, P_LSIGEXC);
proc_unlock(p);
return(0);
} else {
proc_unlock(p);
return(EINVAL);
}
}
/*
* We do not want ptrace to do anything with kernel or launchd
*/
if (uap->pid < 2) {
return(EPERM);
}
/*
* Locate victim, and make sure it is traceable.
*/
if ((t = proc_find(uap->pid)) == NULL)
return (ESRCH);
AUDIT_ARG(process, t);
task = t->task;
if (uap->req == PT_ATTACHEXC) {
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wdeprecated-declarations"
uap->req = PT_ATTACH;
tr_sigexc = 1;
}
if (uap->req == PT_ATTACH) {
#pragma clang diagnostic pop
int err;
if ( kauth_authorize_process(proc_ucred(p), KAUTH_PROCESS_CANTRACE,
t, (uintptr_t)&err, 0, 0) == 0 ) {
/* it's OK to attach */
proc_lock(t);
SET(t->p_lflag, P_LTRACED);
if (tr_sigexc)
SET(t->p_lflag, P_LSIGEXC);
t->p_oppid = t->p_ppid;
/* Check whether child and parent are allowed to run modified
* code (they'll have to) */
proc_unlock(t);
cs_allow_invalid(t);
cs_allow_invalid(p);
if (t->p_pptr != p)
proc_reparentlocked(t, p, 1, 0);
proc_lock(t);
if (get_task_userstop(task) > 0 ) {
stopped = 1;
}
t->p_xstat = 0;
proc_unlock(t);
psignal(t, SIGSTOP);
/*
* If the process was stopped, wake up and run through
* issignal() again to properly connect to the tracing
* process.
*/
if (stopped)
task_resume(task);
error = 0;
goto out;
}
else {
/* not allowed to attach, proper error code returned by kauth_authorize_process */
if (ISSET(t->p_lflag, P_LNOATTACH)) {
psignal(p, SIGSEGV);
}
error = err;
goto out;
}
}
/*
* You can't do what you want to the process if:
* (1) It's not being traced at all,
*/
proc_lock(t);
if (!ISSET(t->p_lflag, P_LTRACED)) {
proc_unlock(t);
error = EPERM;
goto out;
}
/*
* (2) it's not being traced by _you_, or
*/
if (t->p_pptr != p) {
proc_unlock(t);
error = EBUSY;
goto out;
}
/*
* (3) it's not currently stopped.
*/
if (t->p_stat != SSTOP) {
proc_unlock(t);
error = EBUSY;
goto out;
}
/*
* Mach version of ptrace executes request directly here,
* thus simplifying the interaction of ptrace and signals.
*/
/* proc lock is held here */
switch (uap->req) {
case PT_DETACH:
if (t->p_oppid != t->p_ppid) {
struct proc *pp;
proc_unlock(t);
pp = proc_find(t->p_oppid);
if (pp != PROC_NULL) {
proc_reparentlocked(t, pp, 1, 0);
proc_rele(pp);
} else {
/* original parent exited while traced */
proc_list_lock();
t->p_listflag |= P_LIST_DEADPARENT;
proc_list_unlock();
proc_reparentlocked(t, initproc, 1, 0);
}
proc_lock(t);
}
t->p_oppid = 0;
CLR(t->p_lflag, P_LTRACED);
CLR(t->p_lflag, P_LSIGEXC);
proc_unlock(t);
goto resume;
case PT_KILL:
/*
* Tell child process to kill itself after it
* is resumed by adding NSIG to p_cursig. [see issig]
*/
proc_unlock(t);
#if CONFIG_MACF
error = mac_proc_check_signal(p, t, SIGKILL);
if (0 != error)
goto resume;
#endif
psignal(t, SIGKILL);
goto resume;
case PT_STEP: /* single step the child */
case PT_CONTINUE: /* continue the child */
proc_unlock(t);
th_act = (thread_t)get_firstthread(task);
if (th_act == THREAD_NULL) {
error = EINVAL;
goto out;
}
/* force use of Mach SPIs (and task_for_pid security checks) to adjust PC */
if (uap->addr != (user_addr_t)1) {
error = ENOTSUP;
goto out;
}
if ((unsigned)uap->data >= NSIG) {
error = EINVAL;
goto out;
}
if (uap->data != 0) {
#if CONFIG_MACF
error = mac_proc_check_signal(p, t, uap->data);
if (0 != error)
goto out;
#endif
psignal(t, uap->data);
}
if (uap->req == PT_STEP) {
/*
* set trace bit
* we use sending SIGSTOP as a comparable security check.
*/
#if CONFIG_MACF
error = mac_proc_check_signal(p, t, SIGSTOP);
if (0 != error) {
goto out;
}
#endif
if (thread_setsinglestep(th_act, 1) != KERN_SUCCESS) {
error = ENOTSUP;
goto out;
}
} else {
/*
* clear trace bit if on
* we use sending SIGCONT as a comparable security check.
*/
#if CONFIG_MACF
error = mac_proc_check_signal(p, t, SIGCONT);
if (0 != error) {
goto out;
}
#endif
if (thread_setsinglestep(th_act, 0) != KERN_SUCCESS) {
error = ENOTSUP;
goto out;
}
}
resume:
proc_lock(t);
t->p_xstat = uap->data;
t->p_stat = SRUN;
if (t->sigwait) {
wakeup((caddr_t)&(t->sigwait));
proc_unlock(t);
if ((t->p_lflag & P_LSIGEXC) == 0) {
task_resume(task);
}
} else
proc_unlock(t);
break;
case PT_THUPDATE: {
proc_unlock(t);
if ((unsigned)uap->data >= NSIG) {
error = EINVAL;
goto out;
}
th_act = port_name_to_thread(CAST_MACH_PORT_TO_NAME(uap->addr));
if (th_act == THREAD_NULL) {
error = ESRCH;
goto out;
}
ut = (uthread_t)get_bsdthread_info(th_act);
if (uap->data)
ut->uu_siglist |= sigmask(uap->data);
proc_lock(t);
t->p_xstat = uap->data;
t->p_stat = SRUN;
proc_unlock(t);
thread_deallocate(th_act);
error = 0;
}
break;
default:
proc_unlock(t);
error = EINVAL;
goto out;
}
error = 0;
out:
proc_rele(t);
return(error);
}
static
void
ux_handler(void)
{
task_t self = current_task();
mach_port_name_t exc_port_name;
mach_port_name_t exc_set_name;
/* self->kernel_vm_space = TRUE; */
ux_handler_self = self;
/*
* Allocate a port set that we will receive on.
*/
if (mach_port_allocate(get_task_ipcspace(ux_handler_self), MACH_PORT_RIGHT_PORT_SET, &exc_set_name) != MACH_MSG_SUCCESS)
panic("ux_handler: port_set_allocate failed");
/*
* Allocate an exception port and use object_copyin to
* translate it to the global name. Put it into the set.
*/
if (mach_port_allocate(get_task_ipcspace(ux_handler_self), MACH_PORT_RIGHT_RECEIVE, &exc_port_name) != MACH_MSG_SUCCESS)
panic("ux_handler: port_allocate failed");
if (mach_port_move_member(get_task_ipcspace(ux_handler_self),
exc_port_name, exc_set_name) != MACH_MSG_SUCCESS)
panic("ux_handler: port_set_add failed");
if (ipc_object_copyin(get_task_ipcspace(self), exc_port_name,
MACH_MSG_TYPE_MAKE_SEND,
(void *) &ux_exception_port) != MACH_MSG_SUCCESS)
panic("ux_handler: object_copyin(ux_exception_port) failed");
proc_list_lock();
thread_wakeup(&ux_exception_port);
proc_list_unlock();
/* Message handling loop. */
for (;;) {
struct rep_msg {
mach_msg_header_t Head;
NDR_record_t NDR;
kern_return_t RetCode;
} rep_msg;
struct exc_msg {
mach_msg_header_t Head;
/* start of the kernel processed data */
mach_msg_body_t msgh_body;
mach_msg_port_descriptor_t thread;
mach_msg_port_descriptor_t task;
/* end of the kernel processed data */
NDR_record_t NDR;
exception_type_t exception;
mach_msg_type_number_t codeCnt;
mach_exception_data_t code;
/* some times RCV_TO_LARGE probs */
char pad[512];
} exc_msg;
mach_port_name_t reply_port;
kern_return_t result;
exc_msg.Head.msgh_local_port = CAST_MACH_NAME_TO_PORT(exc_set_name);
exc_msg.Head.msgh_size = sizeof (exc_msg);
#if 0
result = mach_msg_receive(&exc_msg.Head);
#else
result = mach_msg_receive(&exc_msg.Head, MACH_RCV_MSG,
sizeof (exc_msg), exc_set_name,
MACH_MSG_TIMEOUT_NONE, MACH_PORT_NULL,
0);
#endif
if (result == MACH_MSG_SUCCESS) {
reply_port = CAST_MACH_PORT_TO_NAME(exc_msg.Head.msgh_remote_port);
if (mach_exc_server(&exc_msg.Head, &rep_msg.Head)) {
result = mach_msg_send(&rep_msg.Head, MACH_SEND_MSG,
sizeof (rep_msg),MACH_MSG_TIMEOUT_NONE,MACH_PORT_NULL);
if (reply_port != 0 && result != MACH_MSG_SUCCESS)
mach_port_deallocate(get_task_ipcspace(ux_handler_self), reply_port);
}
}
else if (result == MACH_RCV_TOO_LARGE)
/* ignore oversized messages */;
else
panic("exception_handler");
}
}
int
fasttrap_pid_probe(arm_saved_state_t *regs)
{
proc_t *p = current_proc();
user_addr_t new_pc = 0;
fasttrap_bucket_t *bucket;
lck_mtx_t *pid_mtx;
fasttrap_tracepoint_t *tp, tp_local;
pid_t pid;
dtrace_icookie_t cookie;
uint_t is_enabled = 0;
int instr_size;
int was_simulated = 1, retire_tp = 1;
user_addr_t pc = regs->pc;
uthread_t uthread = (uthread_t) get_bsdthread_info(current_thread());
/*
* It's possible that a user (in a veritable orgy of bad planning)
* could redirect this thread's flow of control before it reached the
* return probe fasttrap. In this case we need to kill the process
* since it's in a unrecoverable state.
*/
if (uthread->t_dtrace_step) {
ASSERT(uthread->t_dtrace_on);
fasttrap_sigtrap(p, uthread, pc);
return (0);
}
/*
* Clear all user tracing flags.
*/
uthread->t_dtrace_ft = 0;
uthread->t_dtrace_pc = 0;
uthread->t_dtrace_npc = 0;
uthread->t_dtrace_scrpc = 0;
uthread->t_dtrace_astpc = 0;
/*
* Treat a child created by a call to vfork(2) as if it were its
* parent. We know that there's only one thread of control in such a
* process: this one.
*/
if (p->p_lflag & P_LINVFORK) {
proc_list_lock();
while (p->p_lflag & P_LINVFORK)
p = p->p_pptr;
proc_list_unlock();
}
pid = p->p_pid;
pid_mtx = &cpu_core[CPU->cpu_id].cpuc_pid_lock;
lck_mtx_lock(pid_mtx);
bucket = &fasttrap_tpoints.fth_table[FASTTRAP_TPOINTS_INDEX(pid,pc)];
/*
* Lookup the tracepoint that the process just hit.
*/
for (tp = bucket->ftb_data; tp != NULL; tp = tp->ftt_next) {
if (pid == tp->ftt_pid && pc == tp->ftt_pc &&
tp->ftt_proc->ftpc_acount != 0)
break;
}
/*
* If we couldn't find a matching tracepoint, either a tracepoint has
* been inserted without using the pid<pid> ioctl interface (see
* fasttrap_ioctl), or somehow we have mislaid this tracepoint.
*/
if (tp == NULL) {
lck_mtx_unlock(pid_mtx);
return (-1);
}
/* Default to always execute */
int condition_code = 0xE;
if (tp->ftt_thumb) {
uint32_t itstate = GETITSTATE(regs->cpsr);
if (itstate != 0) {
/* In IT block, make sure it's the last statement in the block */
if (ISLASTINIT(itstate)) {
condition_code = itstate >> 4;
} else {
printf("dtrace: fasttrap: Tried to trace instruction %08x at %08x but not at end of IT block\n",
(tp->ftt_thumb && dtrace_instr_size(tp->ftt_instr,tp->ftt_thumb) == 2) ? tp->ftt_instr1 : tp->ftt_instr, pc);
fasttrap_tracepoint_remove(p, tp);
lck_mtx_unlock(pid_mtx);
return (-1);
}
}
/*
* 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");
}