]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/kern_exit.c
xnu-1504.9.17.tar.gz
[apple/xnu.git] / bsd / kern / kern_exit.c
index c866e46c232ee573d5a11ad130c7fa0ffa87e01c..83f7c60dd3f6de9dae563efb4890c94f422073ce 100644 (file)
@@ -1,14 +1,19 @@
 /*
- * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
- * compliance with the License. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ * 
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
  * 
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
@@ -18,7 +23,7 @@
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */
 /*
  *
  *     @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
  */
+/*
+ * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
+ * support for mandatory and extensible security protections.  This notice
+ * is included in support of clause 2.2 (b) of the Apple Public License,
+ * Version 2.0.
+ */
  
 #include <machine/reg.h>
 #include <machine/psl.h>
@@ -70,6 +81,7 @@
 #include <sys/systm.h>
 #include <sys/ioctl.h>
 #include <sys/proc_internal.h>
+#include <sys/proc.h>
 #include <sys/kauth.h>
 #include <sys/tty.h>
 #include <sys/time.h>
 #include <sys/aio_kern.h>
 #include <sys/sysproto.h>
 #include <sys/signalvar.h>
+#include <sys/kdebug.h>
 #include <sys/filedesc.h>      /* fdfree */
+#if SYSV_SHM
 #include <sys/shm_internal.h>  /* shmexit */
+#endif
 #include <sys/acct.h>          /* acct_process */
-#include <machine/spl.h>
 
-#include <bsm/audit_kernel.h>
+#include <security/audit/audit.h>
 #include <bsm/audit_kevents.h>
 
 #include <mach/mach_types.h>
 #include <kern/kalloc.h>
 #include <kern/task.h>
 #include <kern/thread.h>
+#include <kern/thread_call.h>
 #include <kern/sched_prim.h>
 #include <kern/assert.h>
-#if KTRACE   
-#include <sys/ktrace.h>
+#if CONFIG_DTRACE
+/* Do not include dtrace.h, it redefines kmem_[alloc/free] */
+extern void (*dtrace_fasttrap_exit_ptr)(proc_t);
+extern void (*dtrace_helpers_cleanup)(proc_t);
+extern void dtrace_lazy_dofs_destroy(proc_t);
+
+#include <sys/dtrace_ptss.h>
+#endif
+
+#if CONFIG_MACF
+#include <security/mac.h>
+#include <sys/syscall.h>
 #endif
 
 #include <mach/mach_types.h>
 #include <mach/task.h>
 #include <mach/thread_act.h>
-#include <mach/mach_traps.h>   /* init_process */
+
+#include <sys/sdt.h>
 
 extern char init_task_failure_data[];
-int exit1(struct proc *, int, int *);
-void proc_prepareexit(struct proc *p);
-void vfork_exit(struct proc *p, int rv);
-void vproc_exit(struct proc *p);
-__private_extern__ void munge_rusage(struct rusage *a_rusage_p, struct user_rusage *a_user_rusage_p);
+void proc_prepareexit(proc_t p, int rv);
+void vfork_exit(proc_t p, int rv);
+void vproc_exit(proc_t p);
+__private_extern__ void munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p);
+__private_extern__ void munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p);
+static int reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int droplock);
 
 /*
  * Things which should have prototypes in headers, but don't
  */
-void   unix_syscall_return(int);
 void   *get_bsduthreadarg(thread_t);
-void   proc_exit(struct proc *p);
+void   proc_exit(proc_t p);
 int    wait1continue(int result);
 int    waitidcontinue(int result);
 int    *get_bsduthreadrval(thread_t);
-kern_return_t  sys_perf_notify(struct task *task, exception_data_t code,
-                       mach_msg_type_number_t codeCnt);
-
+kern_return_t sys_perf_notify(thread_t thread, int pid);
+kern_return_t abnormal_exit_notify(mach_exception_data_type_t code, 
+               mach_exception_data_type_t subcode);
+void workqueue_exit(struct proc *);
+void   delay(int);
+                       
 /*
  * NOTE: Source and target may *NOT* overlap!
  * XXX Should share code with bsd/dev/ppc/unix_signal.c
  */
 static void
-siginfo_64to32(user_siginfo_t *in, siginfo_t *out)
+siginfo_user_to_user32(user_siginfo_t *in, user32_siginfo_t *out)
+{
+       out->si_signo   = in->si_signo;
+       out->si_errno   = in->si_errno;
+       out->si_code    = in->si_code;
+       out->si_pid     = in->si_pid;
+       out->si_uid     = in->si_uid;
+       out->si_status  = in->si_status;
+       out->si_addr    = CAST_DOWN_EXPLICIT(user32_addr_t,in->si_addr);
+       /* following cast works for sival_int because of padding */
+       out->si_value.sival_ptr = CAST_DOWN_EXPLICIT(user32_addr_t,in->si_value.sival_ptr);
+       out->si_band    = in->si_band;                  /* range reduction */
+}
+
+static void
+siginfo_user_to_user64(user_siginfo_t *in, user64_siginfo_t *out)
 {
        out->si_signo   = in->si_signo;
        out->si_errno   = in->si_errno;
@@ -143,11 +187,10 @@ siginfo_64to32(user_siginfo_t *in, siginfo_t *out)
        out->si_pid     = in->si_pid;
        out->si_uid     = in->si_uid;
        out->si_status  = in->si_status;
-       out->si_addr    = CAST_DOWN(void *,in->si_addr);
+       out->si_addr    = in->si_addr;
        /* following cast works for sival_int because of padding */
-       out->si_value.sival_ptr = CAST_DOWN(void *,in->si_value.sival_ptr);
+       out->si_value.sival_ptr = in->si_value.sival_ptr;
        out->si_band    = in->si_band;                  /* range reduction */
-       out->pad[0]     = in->pad[0];                   /* mcontext.ss.r1 */
 }
 
 /*
@@ -155,12 +198,11 @@ siginfo_64to32(user_siginfo_t *in, siginfo_t *out)
  *     Death of process.
  */
 void
-exit(struct proc *p, struct exit_args *uap, int *retval)
+exit(proc_t p, struct exit_args *uap, int *retval)
 {
        exit1(p, W_EXITCODE(uap->rval, 0), retval);
 
        /* drop funnel before we return */
-       thread_funnel_set(kernel_flock, FALSE);
        thread_exception_return();
        /* NOTREACHED */
        while (TRUE)
@@ -174,11 +216,10 @@ exit(struct proc *p, struct exit_args *uap, int *retval)
  * status and rusage for wait().  Check for child processes and orphan them.
  */
 int
-exit1(struct proc *p, int rv, int *retval)
+exit1(proc_t p, int rv, int *retval)
 {
        thread_t self = current_thread();
        struct task *task = p->task;
-       register int s;
        struct uthread *ut;
 
        /*
@@ -190,40 +231,60 @@ exit1(struct proc *p, int rv, int *retval)
         ut = get_bsdthread_info(self);
         if (ut->uu_flag & UT_VFORK) {
                        vfork_exit(p, rv);
-                       vfork_return(self, p->p_pptr, p , retval);
+                       vfork_return(p , retval, p->p_pid);
                        unix_syscall_return(0);
                        /* NOT REACHED */
         }
-       AUDIT_SYSCALL_EXIT(0, p, ut); /* Exit is always successfull */
-        signal_lock(p);
+
+       /* 
+        * The parameter list of audit_syscall_exit() was augmented to
+        * take the Darwin syscall number as the first parameter,
+        * which is currently required by mac_audit_postselect().
+        */
+
+       /* 
+        * The BSM token contains two components: an exit status as passed
+        * to exit(), and a return value to indicate what sort of exit it 
+        * was.  The exit status is WEXITSTATUS(rv), but it's not clear
+        * what the return value is.
+        */
+       AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
+       AUDIT_SYSCALL_EXIT(SYS_exit, p, ut, 0); /* Exit is always successfull */
+
+       DTRACE_PROC1(exit, int, CLD_EXITED);
+
+        proc_lock(p);
        while (p->exit_thread != self) {
                if (sig_try_locked(p) <= 0) {
                        if (get_threadtask(self) != task) {
-                                signal_unlock(p);
+                               proc_unlock(p);
                                return(0);
                         }
-                       signal_unlock(p);
+                       proc_unlock(p);
                        thread_terminate(self);
-                       thread_funnel_set(kernel_flock, FALSE);
                        thread_exception_return();
                        /* NOTREACHED */
                }
                sig_lock_to_exit(p);
        }
-        signal_unlock(p);
-       if (p->p_pid == 1) {
+       if (p == initproc) {
+               proc_unlock(p);
                printf("pid 1 exited (signal %d, exit %d)",
                    WTERMSIG(rv), WEXITSTATUS(rv));
-               panic("init died\nState at Last Exception:\n\n%s", 
+               panic("%s died\nState at Last Exception:\n\n%s", 
+                                                       (p->p_comm[0] != '\0' ?
+                                                               p->p_comm :
+                                                               "launchd"),
                                                        init_task_failure_data);
        }
 
-       s = splsched();
-       p->p_flag |= P_WEXIT;
-       splx(s);
-       proc_prepareexit(p);
+       p->p_lflag |= P_LEXIT;
        p->p_xstat = rv;
 
+       proc_unlock(p);
+
+       proc_prepareexit(p, rv);
+
        /* task terminate will call proc_terminate and that cleans it up */
        task_terminate_internal(task);
 
@@ -231,16 +292,39 @@ exit1(struct proc *p, int rv, int *retval)
 }
 
 void
-proc_prepareexit(struct proc *p
+proc_prepareexit(proc_t p, int rv
 {
+       mach_exception_data_type_t code, subcode;
        struct uthread *ut;
-       exception_data_t        code[EXCEPTION_CODE_MAX];
        thread_t self = current_thread();
+       ut = get_bsdthread_info(self);
+
+       /* If a core should be generated, notify crash reporter */
+       if (hassigprop(WTERMSIG(rv), SA_CORE)) {
+               /* 
+                * Workaround for processes checking up on PT_DENY_ATTACH:
+                * should be backed out post-Leopard (details in 5431025).
+                */
+               if ((SIGSEGV == WTERMSIG(rv)) && 
+                               (p->p_pptr->p_lflag & P_LNOATTACH)) {
+                       goto skipcheck;
+               }
+
+               /*
+                * Crash Reporter looks for the signal value, original exception
+                * type, and low 20 bits of the original code in code[0] 
+                * (8, 4, and 20 bits respectively). code[1] is unmodified. 
+                */
+               code = ((WTERMSIG(rv) & 0xff) << 24) |
+                       ((ut->uu_exception & 0x0f) << 20) | 
+                       ((int)ut->uu_code & 0xfffff);
+               subcode = ut->uu_subcode;
+               (void) abnormal_exit_notify(code, subcode);
+       }
 
-       code[0] = (exception_data_t)0xFF000001;         /* Set terminate code */
-       code[1] = (exception_data_t)p->p_pid;           /* Pass out the pid */
+skipcheck:
        /* Notify the perf server */
-       (void)sys_perf_notify(p->task, (exception_data_t)&code, 2);
+       (void)sys_perf_notify(self, p->p_pid);
 
        /*
         * Remove proc from allproc queue and from pidhash chain.
@@ -248,46 +332,97 @@ proc_prepareexit(struct proc *p)
         * Not doing causes things like mount() find this on allproc
         * in partially cleaned state.
         */
+
+       proc_list_lock();
+
        LIST_REMOVE(p, p_list);
        LIST_INSERT_HEAD(&zombproc, p, p_list); /* Place onto zombproc. */
-       LIST_REMOVE(p, p_hash);
+       /* will not be visible via proc_find */
+       p->p_listflag |= P_LIST_EXITED;
+
+       proc_list_unlock();
+
 
 #ifdef PGINPROF
        vmsizmon();
 #endif
        /*
         * If parent is waiting for us to exit or exec,
-        * P_PPWAIT is set; we will wakeup the parent below.
+        * P_LPPWAIT is set; we will wakeup the parent below.
         */
-       p->p_flag &= ~(P_TRACED | P_PPWAIT);
+       proc_lock(p);
+       p->p_lflag &= ~(P_LTRACED | P_LPPWAIT);
        p->p_sigignore = ~(sigcantmask);
-       p->p_siglist = 0;
-       ut = get_bsdthread_info(self);
        ut->uu_siglist = 0;
-       untimeout(realitexpire, (caddr_t)p->p_pid);
+       proc_unlock(p);
 }
 
 void 
-proc_exit(struct proc *p)
+proc_exit(proc_t p)
 {
-       register struct proc *q, *nq, *pp;
+       proc_t q;
+       proc_t pp;
        struct task *task = p->task;
-       register int s;
-       boolean_t funnel_state;
+       vnode_t tvp = NULLVP;
+       struct pgrp * pg;
+       struct session *sessp;
+       struct uthread * uth;
+       pid_t pid;
+       int exitval;
 
        /* This can happen if thread_terminate of the single thread
         * process 
         */
 
-       funnel_state = thread_funnel_set(kernel_flock, TRUE);
-       if( !(p->p_flag & P_WEXIT)) {
-               s = splsched();
-               p->p_flag |= P_WEXIT;
-               splx(s);
-               proc_prepareexit(p);    
+       uth = (struct uthread *)get_bsdthread_info(current_thread());
+
+       proc_lock(p);
+       if( !(p->p_lflag & P_LEXIT)) {
+               p->p_lflag |= P_LEXIT;
+               proc_unlock(p);
+               proc_prepareexit(p, 0); 
+               proc_lock(p);
        }
 
        p->p_lflag |= P_LPEXIT;
+       proc_unlock(p);
+       pid = p->p_pid;
+       exitval = p->p_xstat;
+       KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_START,
+                                             pid, exitval, 0, 0, 0);
+
+#if CONFIG_DTRACE
+       /*
+        * Free any outstanding lazy dof entries. It is imperative we
+        * always call dtrace_lazy_dofs_destroy, 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 cleanup. The lazy dof faulting will then
+        * install new helpers which will never be cleaned up, and leak.
+        */
+       dtrace_lazy_dofs_destroy(p);
+
+       /*
+        * Clean up any DTrace helper actions or probes for the process.
+        */
+       if (p->p_dtrace_helpers != NULL) {
+               (*dtrace_helpers_cleanup)(p);
+       }
+
+       /*
+        * Clean up any DTrace probes associated with this process.
+        */
+       /*
+        * APPLE NOTE: We release ptss pages/entries in dtrace_fasttrap_exit_ptr(),
+        * call this after dtrace_helpers_cleanup()
+        */
+       proc_lock(p);
+       if (p->p_dtrace_probes && dtrace_fasttrap_exit_ptr) {
+               (*dtrace_fasttrap_exit_ptr)(p);
+       }
+       proc_unlock(p);
+#endif
+
        /* XXX Zombie allocation may fail, in which case stats get lost */
        MALLOC_ZONE(p->p_ru, struct rusage *,
                        sizeof (*p->p_ru), M_ZOMBIE, M_WAITOK);
@@ -297,13 +432,9 @@ proc_exit(struct proc *p)
         * already active.  MAY BLOCK!
         */
        
-       p->p_lflag |= P_LREFDRAIN;
-       while (p->p_internalref) {
-               p->p_lflag |= P_LREFDRAINWAIT;
-               msleep(&p->p_internalref, (lck_mtx_t *)0, 0, "proc_refdrain", 0) ;
-       }
-       p->p_lflag &= ~P_LREFDRAIN;
-       p->p_lflag |= P_LREFDEAD;
+       proc_refdrain(p);
+
+       workqueue_exit(p);
 
        _aio_exit( p );
 
@@ -313,18 +444,40 @@ proc_exit(struct proc *p)
         */
        fdfree(p);
 
+       if (uth->uu_lowpri_window) {
+               /*
+                * task is marked as a low priority I/O type
+                * and the I/O we issued while in flushing files on close
+                * collided with normal I/O operations...
+                * no need to throttle this thread since its going away
+                * but we do need to update our bookeeping w/r to throttled threads
+                */
+               throttle_lowpri_io(FALSE);
+       }
+
+#if SYSV_SHM
        /* Close ref SYSV Shared memory*/
        if (p->vm_shm)
                shmexit(p);
+#endif
+#if SYSV_SEM
        /* Release SYSV semaphores */
        semexit(p);
+#endif
        
-       if (SESS_LEADER(p)) {
-               register struct session *sp = p->p_session;
+#if PSYNCH
+       pth_proc_hashdelete(p);
+#endif /* PSYNCH */
+
+       sessp = proc_session(p);
+       if (SESS_LEADER(p, sessp)) {
 
-               if (sp->s_ttyvp) {
+               if (sessp->s_ttyvp != NULLVP) {
                        struct vnode *ttyvp;
+                       int ttyvid;
                        struct vfs_context context;
+                       struct tty * tp;
+
 
                        /*
                         * Controlling process.
@@ -332,109 +485,151 @@ proc_exit(struct proc *p)
                         * drain controlling terminal
                         * and revoke access to controlling terminal.
                         */
-                       if (sp->s_ttyp->t_session == sp) {
-                               if (sp->s_ttyp->t_pgrp)
-                                       pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
-                               (void) ttywait(sp->s_ttyp);
-                               /*
-                                * The tty could have been revoked
-                                * if we blocked.
-                                */
-                               context.vc_proc = p;
-                               context.vc_ucred = p->p_ucred;
-                               if (sp->s_ttyvp)
-                                       VNOP_REVOKE(sp->s_ttyvp, REVOKEALL, &context);
+                       tp = SESSION_TP(sessp);
+
+                       if ((tp != TTY_NULL) && (tp->t_session == sessp)) {
+                               tty_pgsignal(tp, SIGHUP, 1);
+
+                               session_lock(sessp);
+                               /* reget potentially tp due to revocation */
+                               tp = SESSION_TP(sessp);
+                               ttyvp = sessp->s_ttyvp;
+                               ttyvid = sessp->s_ttyvid;
+                               sessp->s_ttyvp = NULLVP;
+                               sessp->s_ttyvid = 0;
+                               sessp->s_ttyp = TTY_NULL;
+                               sessp->s_ttypgrpid = NO_PID;
+                               session_unlock(sessp);
+
+                               if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) {
+
+                                       if (tp != TTY_NULL) {
+                                               tty_lock(tp);
+                                               (void) ttywait(tp);
+                                               tty_unlock(tp);
+                                       }
+                                       context.vc_thread = proc_thread(p); /* XXX */
+                                       context.vc_ucred = kauth_cred_proc_ref(p);
+                                       VNOP_REVOKE(ttyvp, REVOKEALL, &context);
+                                       vnode_put(ttyvp);
+                                       kauth_cred_unref(&context.vc_ucred);
+                               }
+                       } else {
+                               session_lock(sessp);
+                               /* reget potentially tp due to revocation */
+                               tp = SESSION_TP(sessp);
+                               ttyvp = sessp->s_ttyvp;
+                               sessp->s_ttyvp = NULLVP;
+                               sessp->s_ttyvid = 0;
+                               sessp->s_ttyp = TTY_NULL;
+                               sessp->s_ttypgrpid = NO_PID;
+                               session_unlock(sessp);
                        }
-                       ttyvp = sp->s_ttyvp;
-                       sp->s_ttyvp = NULL;
-                       if (ttyvp) {
+                       if (ttyvp)
                                vnode_rele(ttyvp);
-                       }
                        /*
                         * s_ttyp is not zero'd; we use this to indicate
                         * that the session once had a controlling terminal.
                         * (for logging and informational purposes)
                         */
                }
-               sp->s_leader = NULL;
+               
+               session_lock(sessp);
+               sessp->s_leader = NULL;
+               session_unlock(sessp);
        }
+       session_rele(sessp);
+
+       pg = proc_pgrp(p);
+       fixjobc(p, pg, 0);
+       pg_rele(pg);
 
-       fixjobc(p, p->p_pgrp, 0);
        p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
        (void)acct_process(p);
 
-#if KTRACE
-       /* 
-        * release trace file
-        */
-       p->p_traceflag = 0;     /* don't trace the vnode_put() */
-       if (p->p_tracep) {
-               struct vnode *tvp = p->p_tracep;
-               p->p_tracep = NULL;
-               vnode_rele(tvp);
+       proc_list_lock();
+
+       if ((p->p_listflag & P_LIST_EXITCOUNT) == P_LIST_EXITCOUNT) {
+               p->p_listflag &= ~P_LIST_EXITCOUNT;
+               proc_shutdown_exitcount--;
+               if (proc_shutdown_exitcount == 0)
+                       wakeup(&proc_shutdown_exitcount);
        }
-#endif
 
-       while (q = p->p_children.lh_first) {
-               proc_reparent(q, initproc);
-               /*
-                * Traced processes are killed
-                * since their existence means someone is messing up.
-                */
-               if (q->p_flag & P_TRACED) {
-                       q->p_flag &= ~P_TRACED;
-                       if (q->sigwait_thread) {
-                               /*
-                                * The sigwait_thread could be stopped at a
-                                * breakpoint. Wake it up to kill.
-                                * Need to do this as it could be a thread which is not
-                                * the first thread in the task. So any attempts to kill
-                                * the process would result into a deadlock on q->sigwait.
-                                */
-                               thread_resume((thread_t)q->sigwait_thread);
-                               clear_wait(q->sigwait_thread, THREAD_INTERRUPTED);
-                               threadsignal((thread_t)q->sigwait_thread, SIGKILL, 0);
+       /* wait till parentrefs are dropped and grant no more */
+       proc_childdrainstart(p);
+       while ((q = p->p_children.lh_first) != NULL) {
+               q->p_listflag |= P_LIST_DEADPARENT;
+               if (q->p_stat == SZOMB) {
+                       if (p != q->p_pptr)
+                               panic("parent child linkage broken");
+                       /* check for sysctl zomb lookup */
+                       while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
+                               msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
+                       }
+                       q->p_listflag |= P_LIST_WAITING;
+                       /*
+                        * This is a named reference and it is not granted
+                        * if the reap is already in progress. So we get
+                        * the reference here exclusively and their can be
+                        * no waiters. So there is no need for a wakeup
+                        * after we are done. AlsO  the reap frees the structure
+                        * and the proc struct cannot be used for wakeups as well. 
+                        * It is safe to use q here as this is system reap
+                        */
+                       (void)reap_child_locked(p, q, 1, 1, 0);
+               } else {
+                       proc_reparentlocked(q, initproc, 0, 1);
+                       /*
+                       * Traced processes are killed
+                       * since their existence means someone is messing up.
+                       */
+                       if (q->p_lflag & P_LTRACED) {
+                               proc_list_unlock();
+                               proc_lock(q);
+                               q->p_lflag &= ~P_LTRACED;
+                               if (q->sigwait_thread) {
+                                       proc_unlock(q);
+                                       /*
+                                       * The sigwait_thread could be stopped at a
+                                       * breakpoint. Wake it up to kill.
+                                       * Need to do this as it could be a thread which is not
+                                       * the first thread in the task. So any attempts to kill
+                                       * the process would result into a deadlock on q->sigwait.
+                                       */
+                                       thread_resume((thread_t)q->sigwait_thread);
+                                       clear_wait(q->sigwait_thread, THREAD_INTERRUPTED);
+                                       threadsignal((thread_t)q->sigwait_thread, SIGKILL, 0);
+                               } else
+                                       proc_unlock(q);
+                               psignal(q, SIGKILL);
+                               proc_list_lock();
                        }
-                       psignal(q, SIGKILL);
                }
        }
 
+       proc_childdrainend(p);
+       proc_list_unlock();
+
+       /*
+        * Release reference to text vnode
+        */
+       tvp = p->p_textvp;
+       p->p_textvp = NULL;
+       if (tvp != NULLVP) {
+               vnode_rele(tvp);
+       }
+
        /*
         * Save exit status and final rusage info, adding in child rusage
         * info and self times.  If we were unable to allocate a zombie
         * structure, this information is lost.
         */
+       /* No need for locking here as no one than this thread can access this */
        if (p->p_ru != NULL) {
            *p->p_ru = p->p_stats->p_ru;
 
-           timerclear(&p->p_ru->ru_utime);
-           timerclear(&p->p_ru->ru_stime);
-
-           if (task) {
-               task_basic_info_data_t tinfo;
-               task_thread_times_info_data_t ttimesinfo;
-               int task_info_stuff, task_ttimes_stuff;
-               struct timeval ut,st;
-
-               task_info_stuff = TASK_BASIC_INFO_COUNT;
-               task_info(task, TASK_BASIC_INFO,
-                         (task_info_t)&tinfo, &task_info_stuff);
-               p->p_ru->ru_utime.tv_sec = tinfo.user_time.seconds;
-               p->p_ru->ru_utime.tv_usec = tinfo.user_time.microseconds;
-               p->p_ru->ru_stime.tv_sec = tinfo.system_time.seconds;
-               p->p_ru->ru_stime.tv_usec = tinfo.system_time.microseconds;
-
-               task_ttimes_stuff = TASK_THREAD_TIMES_INFO_COUNT;
-               task_info(task, TASK_THREAD_TIMES_INFO,
-                         (task_info_t)&ttimesinfo, &task_ttimes_stuff);
-
-               ut.tv_sec = ttimesinfo.user_time.seconds;
-               ut.tv_usec = ttimesinfo.user_time.microseconds;
-               st.tv_sec = ttimesinfo.system_time.seconds;
-               st.tv_usec = ttimesinfo.system_time.microseconds;
-               timeradd(&ut,&p->p_ru->ru_utime,&p->p_ru->ru_utime);
-               timeradd(&st,&p->p_ru->ru_stime,&p->p_ru->ru_stime);
-           }
+           calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
 
            ruadd(p->p_ru, &p->p_stats->p_cru);
        }
@@ -455,71 +650,162 @@ proc_exit(struct proc *p)
                }
        }
 
+       proc_spinlock(p);
+       if (thread_call_cancel(p->p_rcall))
+               p->p_ractive--;
+
+       while (p->p_ractive > 0) {
+               proc_spinunlock(p);
+               
+               delay(1);
+
+               proc_spinlock(p);
+       }
+       proc_spinunlock(p);
+
+       thread_call_free(p->p_rcall);
+       p->p_rcall = NULL;
+
        /*
         * Other substructures are freed from wait().
         */
-       FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_SUBPROC);
+       FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
        p->p_stats = NULL;
 
-       FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SUBPROC);
+       FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
        p->p_sigacts = NULL;
 
-       if (--p->p_limit->p_refcnt == 0)
-               FREE_ZONE(p->p_limit, sizeof *p->p_limit, M_SUBPROC);
+       proc_limitdrop(p, 1);
        p->p_limit = NULL;
 
+
        /*
         * Finish up by terminating the task
         * and halt this thread (only if a
         * member of the task exiting).
         */
        p->task = TASK_NULL;
-       //task->proc = NULL;
        set_bsdtask_info(task, NULL);
 
-       KNOTE(&p->p_klist, NOTE_EXIT);
+       proc_knote(p, NOTE_EXIT);
 
+       /* mark the thread as the one that is doing proc_exit
+        * no need to hold proc lock in uthread_free
+        */
+       uth->uu_flag |= UT_PROCEXIT;
        /*
         * Notify parent that we're gone.
         */
-       if (p->p_pptr->p_flag & P_NOCLDWAIT) {
-               struct proc *opp = p->p_pptr;
+       pp = proc_parent(p);
+       if (pp->p_flag & P_NOCLDWAIT) {
 
+#if 3839178
+               /*
+                * If the parent is ignoring SIGCHLD, then POSIX requires
+                * us to not add the resource usage to the parent process -
+                * we are only going to hand it off to init to get reaped.
+                * We should contest the standard in this case on the basis
+                * of RLIMIT_CPU.
+                */
+#else  /* !3839178 */
                /*
                 * Add child resource usage to parent before giving
                 * zombie to init.  If we were unable to allocate a
                 * zombie structure, this information is lost.
                 */
-               if (p->p_ru != NULL)
-                       ruadd(&p->p_pptr->p_stats->p_cru, p->p_ru);
+               if (p->p_ru != NULL) {
+                       proc_lock(pp);
+                       ruadd(&pp->p_stats->p_cru, p->p_ru);
+                       proc_unlock(pp);
+               }
+#endif /* !3839178 */
 
-               proc_reparent(p, initproc);
-               /* If there are no more children wakeup parent */
-               if (LIST_EMPTY(&opp->p_children))
-                       wakeup((caddr_t)opp);
-       }
-       /* should be fine as parent proc would be initproc */
-       pp = p->p_pptr;
-       if (pp != initproc) {
-               pp->si_pid = p->p_pid;
-               pp->si_status = p->p_xstat;
-               pp->si_code = CLD_EXITED;
-               pp->si_uid = p->p_ucred->cr_ruid;
+               /* kernel can reap this one, no need to move it to launchd */
+               proc_list_lock();
+               p->p_listflag |= P_LIST_DEADPARENT;
+               proc_list_unlock();
        }
-       /* mark as a zombie */
-       p->p_stat = SZOMB;
+       if ((p->p_listflag & P_LIST_DEADPARENT) == 0) {
+               if (pp != initproc) {
+                       proc_lock(pp);
+                       pp->si_pid = p->p_pid;
+                       pp->si_status = p->p_xstat;
+                       pp->si_code = CLD_EXITED;
+                       /*
+                        * p_ucred usage is safe as it is an exiting process
+                        * and reference is dropped in reap
+                        */
+                       pp->si_uid = p->p_ucred->cr_ruid;
+                       proc_unlock(pp);
+               }
+               /* mark as a zombie */
+               /* No need to take proc lock as all refs are drained and
+                * no one except parent (reaping ) can look at this.
+                * The write is to an int and is coherent. Also parent is
+                *  keyed off of list lock for reaping
+                */
+               KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END,
+                                             pid, exitval, 0, 0, 0);
+               p->p_stat = SZOMB;
+               /* 
+                * The current process can be reaped so, no one
+                * can depend on this
+                */
+
+               psignal(pp, SIGCHLD);
+       
+               /* and now wakeup the parent */
+               proc_list_lock();
+               wakeup((caddr_t)pp);
+               proc_list_unlock();
+       } else {
+               /* should be fine as parent proc would be initproc */
+               /* mark as a zombie */
+               /* No need to take proc lock as all refs are drained and
+                * no one except parent (reaping ) can look at this.
+                * The write is to an int and is coherent. Also parent is
+                *  keyed off of list lock for reaping
+                */
+               proc_list_lock();
+               KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_END,
+                                             pid, exitval, 0, 0, 0);
+               /* check for sysctl zomb lookup */
+               while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
+                       msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
+               }
+               /* safe to use p as this is a system reap */
+               p->p_stat = SZOMB;
+               p->p_listflag |= P_LIST_WAITING;
 
-       psignal(pp, SIGCHLD);
+               /*
+                * This is a named reference and it is not granted
+                * if the reap is already in progress. So we get
+                * the reference here exclusively and their can be
+                * no waiters. So there is no need for a wakeup
+                * after we are done. AlsO  the reap frees the structure
+                * and the proc struct cannot be used for wakeups as well. 
+                * It is safe to use p here as this is system reap
+                */
+               (void)reap_child_locked(pp, p, 1, 1, 1);
+               /* list lock dropped by reap_child_locked */
+       }
+       if (uth->uu_lowpri_window) {
+               /*
+                * task is marked as a low priority I/O type and we've
+                * somehow picked up another throttle during exit processing...
+                * no need to throttle this thread since its going away
+                * but we do need to update our bookeeping w/r to throttled threads
+                */
+               throttle_lowpri_io(FALSE);
+       }
 
-       /* and now wakeup the parent */
-       wakeup((caddr_t)p->p_pptr);
+       proc_rele(pp);
 
-       (void) thread_funnel_set(kernel_flock, funnel_state);
 }
 
 
 /*
- * reap_child_process
+ * reap_child_locked
  *
  * Description:        Given a process from which all status information needed
  *             has already been extracted, if the process is a ptrace
@@ -527,81 +813,161 @@ proc_exit(struct proc *p)
  *             parent, else recover all resources remaining associated
  *             with it.
  *
- * Parameters: struct proc *parent     Parent of process being reaped
- *             struct proc *child      Process to reap
+ * Parameters: proc_t parent           Parent of process being reaped
+ *             proc_t child            Process to reap
  *
  * Returns:    0                       Process was not reaped because it
  *                                     came from an attach
  *             1                       Process was reaped
  */
 static int
-reap_child_process(struct proc *parent, struct proc *child)
+reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int droplock)
 {
-       struct proc *trace_parent;      /* Traced parent process, if tracing */
-       struct vnode *tvp;              /* Traced vnode pointer, if used */
+       proc_t trace_parent = PROC_NULL;        /* Traced parent process, if tracing */
 
+       if (locked == 1)
+               proc_list_unlock();
+       
        /*
         * If we got the child via a ptrace 'attach',
         * we need to give it back to the old parent.
+        *
+        * Exception: someone who has been reparented to launchd before being
+        * ptraced can simply be reaped, refer to radar 5677288
+        *      p_oppid                  -> ptraced
+        *      trace_parent == initproc -> away from launchd
+        *      P_LIST_DEADPARENT        -> came to launchd by reparenting
         */
-       if (child->p_oppid && (trace_parent = pfind(child->p_oppid))) {
+       if (child->p_oppid && (trace_parent = proc_find(child->p_oppid)) 
+                       && !((trace_parent == initproc) && (child->p_lflag & P_LIST_DEADPARENT))) {
+               proc_lock(child);
                child->p_oppid = 0;
-               proc_reparent(child, trace_parent);
+               proc_unlock(child);
                if (trace_parent != initproc) {
+                       /* 
+                        * proc internal fileds  and p_ucred usage safe 
+                        * here as child is dead and is not reaped or 
+                        * reparented yet 
+                        */
+                       proc_lock(trace_parent);
                        trace_parent->si_pid = child->p_pid;
                        trace_parent->si_status = child->p_xstat;
                        trace_parent->si_code = CLD_CONTINUED;
                        trace_parent->si_uid = child->p_ucred->cr_ruid;
+                       proc_unlock(trace_parent);
                }
+               proc_reparentlocked(child, trace_parent, 1, 0);
                psignal(trace_parent, SIGCHLD);
+               proc_list_lock();
                wakeup((caddr_t)trace_parent);
+               child->p_listflag &= ~P_LIST_WAITING;
+               wakeup(&child->p_stat);
+               proc_list_unlock();
+               proc_rele(trace_parent);
+               if ((locked == 1) && (droplock == 0))
+                       proc_list_lock();
                return (0);
        }
+       
+       if (trace_parent != PROC_NULL) {
+               proc_rele(trace_parent);
+       }
+       
+       proc_knote(child, NOTE_REAP);
+       proc_knote_drain(child);
+
        child->p_xstat = 0;
        if (child->p_ru) {
-               ruadd(&parent->p_stats->p_cru, child->p_ru);
+               proc_lock(parent);
+#if 3839178
+               /*
+                * If the parent is ignoring SIGCHLD, then POSIX requires
+                * us to not add the resource usage to the parent process -
+                * we are only going to hand it off to init to get reaped.
+                * We should contest the standard in this case on the basis
+                * of RLIMIT_CPU.
+                */
+               if (!(parent->p_flag & P_NOCLDWAIT))
+#endif /* 3839178 */
+                       ruadd(&parent->p_stats->p_cru, child->p_ru);
+               proc_unlock(parent);
                FREE_ZONE(child->p_ru, sizeof *child->p_ru, M_ZOMBIE);
                child->p_ru = NULL;
        } else {
                printf("Warning : lost p_ru for %s\n", child->p_comm);
        }
 
+       AUDIT_SESSION_PROCEXIT(child->p_ucred);
+
        /*
         * Decrement the count of procs running with this uid.
+        * p_ucred usage is safe here as it is an exited process.
+        * and refernce is dropped after these calls down below
+        * (locking protection is provided by list lock held in chgproccnt)
         */
        (void)chgproccnt(child->p_ucred->cr_ruid, -1);
 
-       /*
-        * Free up credentials.
-        */
-       if (child->p_ucred != NOCRED) {
-               kauth_cred_t ucr = child->p_ucred;
-                       child->p_ucred = NOCRED;
-                       kauth_cred_rele(ucr);
-               }
+#if CONFIG_LCTX
+       ALLLCTX_LOCK;
+       leavelctx(child);
+       ALLLCTX_UNLOCK;
+#endif
 
        /*
-        * Release reference to text vnode
+        * Free up credentials.
         */
-       tvp = child->p_textvp;
-       child->p_textvp = NULL;
-       if (tvp) {
-               vnode_rele(tvp);
+       if (IS_VALID_CRED(child->p_ucred)) {
+               kauth_cred_unref(&child->p_ucred);
        }
+
+       /*  XXXX Note NOT SAFE TO USE p_ucred from this point onwards */
+
        /*
         * Finally finished with old proc entry.
         * Unlink it from its process group and free it.
         */
        leavepgrp(child);
+
+       proc_list_lock();
        LIST_REMOVE(child, p_list);     /* off zombproc */
+       parent->p_childrencnt--;
        LIST_REMOVE(child, p_sibling);
-       child->p_lflag &= ~P_LWAITING;
+       /* If there are no more children wakeup parent */
+       if ((deadparent != 0) && (LIST_EMPTY(&parent->p_children)))
+               wakeup((caddr_t)parent);        /* with list lock held */
+       child->p_listflag &= ~P_LIST_WAITING;
        wakeup(&child->p_stat);
 
+       /* Take it out of process hash */
+       LIST_REMOVE(child, p_hash);
+       child->p_listflag &= ~P_LIST_INHASH;
+       proc_checkdeadrefs(child);
+       nprocs--;
+
+       proc_list_unlock();
+
+#ifdef CONFIG_EMBEDDED
        lck_mtx_destroy(&child->p_mlock, proc_lck_grp);
        lck_mtx_destroy(&child->p_fdmlock, proc_lck_grp);
+#if CONFIG_DTRACE
+       lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
+#endif
+       lck_spin_destroy(&child->p_slock, proc_lck_grp);
+
+#else  
+       lck_mtx_destroy(&child->p_mlock, proc_mlock_grp);
+       lck_mtx_destroy(&child->p_fdmlock, proc_fdmlock_grp);
+#if CONFIG_DTRACE
+       lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
+#endif
+       lck_spin_destroy(&child->p_slock, proc_slock_grp);
+#endif
+       workqueue_destroy_lock(child);
+
        FREE_ZONE(child, sizeof *child, M_PROC);
-       nprocs--;
+       if ((locked == 1) && (droplock == 0))
+               proc_list_lock();
+
        return (1);
 }
 
@@ -612,7 +978,7 @@ wait1continue(int result)
        void *vt;
        thread_t thread;
        int *retval;
-       struct proc *p;
+       proc_t p;
 
        if (result)
                return(result);
@@ -621,85 +987,129 @@ wait1continue(int result)
        thread = current_thread();
        vt = get_bsduthreadarg(thread);
        retval = get_bsduthreadrval(thread);
-       return(wait4((struct proc *)p, (struct wait4_args *)vt, retval));
+       return(wait4(p, (struct wait4_args *)vt, retval));
+}
+
+int
+wait4(proc_t q, struct wait4_args *uap, int32_t *retval)
+{
+       __pthread_testcancel(1);
+       return(wait4_nocancel(q, (struct wait4_nocancel_args *)uap, retval));
 }
 
 int
-wait4(struct proc *q, struct wait4_args *uap, register_t *retval)
+wait4_nocancel(proc_t q, struct wait4_nocancel_args *uap, int32_t *retval)
 {
-       register int nfound;
-       register struct proc *p;
+       int nfound;
+       int sibling_count;
+       proc_t p;
        int status, error;
 
+       AUDIT_ARG(pid, uap->pid);
+
        if (uap->pid == 0)
-               uap->pid = -q->p_pgid;
+               uap->pid = -q->p_pgrpid;
 
 loop:
+       proc_list_lock();
+loop1:
        nfound = 0;
+       sibling_count = 0;
+
        for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
+               if ( p->p_sibling.le_next != 0 )
+                       sibling_count++;
                if (uap->pid != WAIT_ANY &&
                    p->p_pid != uap->pid &&
-                   p->p_pgid != -(uap->pid))
+                   p->p_pgrpid != -(uap->pid))
                        continue;
+
                nfound++;
 
                /* XXX This is racy because we don't get the lock!!!! */
 
-               if (p->p_lflag & P_LWAITING) {
-                       (void)tsleep(&p->p_stat, PWAIT, "waitcoll", 0);
-                       goto loop;
+               if (p->p_listflag & P_LIST_WAITING) {
+                       (void)msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
+                       goto loop1;
                }
-               p->p_lflag |= P_LWAITING;   /* only allow single thread to wait() */
+               p->p_listflag |= P_LIST_WAITING;   /* only allow single thread to wait() */
+
 
                if (p->p_stat == SZOMB) {
+                       proc_list_unlock();
+#if CONFIG_MACF
+                       if ((error = mac_proc_check_wait(q, p)) != 0)
+                               goto out;
+#endif
                        retval[0] = p->p_pid;
                        if (uap->status) {
-                               status = p->p_xstat;    /* convert to int */
+                               /* Legacy apps expect only 8 bits of status */
+                               status = 0xffff & p->p_xstat;   /* convert to int */
                                error = copyout((caddr_t)&status,
                                                        uap->status,
                                                    sizeof(status));
-                               if (error) {
-                                       p->p_lflag &= ~P_LWAITING;
-                                       wakeup(&p->p_stat);
-                                       return (error);
-                               }
+                               if (error) 
+                                       goto out;
                        }
                        if (uap->rusage) {
                                if (p->p_ru == NULL) {
                                        error = ENOMEM;
                                } else {
                                        if (IS_64BIT_PROCESS(q)) {
-                                               struct user_rusage      my_rusage;
-                                               munge_rusage(p->p_ru, &my_rusage);
+                                               struct user64_rusage    my_rusage;
+                                               munge_user64_rusage(p->p_ru, &my_rusage);
                                                error = copyout((caddr_t)&my_rusage,
                                                        uap->rusage,
                                                        sizeof (my_rusage));
                                        }
                                        else {
-                                               error = copyout((caddr_t)p->p_ru,
+                                               struct user32_rusage    my_rusage;
+                                               munge_user32_rusage(p->p_ru, &my_rusage);
+                                               error = copyout((caddr_t)&my_rusage,
                                                        uap->rusage,
-                                                       sizeof (struct rusage));
+                                                       sizeof (my_rusage));
                                        }
                                }
                                /* information unavailable? */
-                               if (error) {
-                                       p->p_lflag &= ~P_LWAITING;
-                                       wakeup(&p->p_stat);
-                                       return (error);
-                               }
+                               if (error) 
+                                       goto out;
                        }
 
-                       /* Clean up */
-                       if (!reap_child_process(q, p)) {
-                               p->p_lflag &= ~P_LWAITING;
-                               wakeup(&p->p_stat);
+                       /* Conformance change for 6577252.
+                        * When SIGCHLD is blocked and wait() returns because the status
+                        * of a child process is available and there are no other 
+                        * children processes, then any pending SIGCHLD signal is cleared.
+                        */
+                       if ( sibling_count == 0 ) {
+                               int mask = sigmask(SIGCHLD);
+                               uthread_t uth = (struct uthread *)get_bsdthread_info(current_thread());
+
+                               if ( (uth->uu_sigmask & mask) != 0 ) {
+                                       /* we are blocking SIGCHLD signals.  clear any pending SIGCHLD.
+                                        * This locking looks funny but it is protecting access to the 
+                                        * thread via p_uthlist.
+                                        */
+                                       proc_lock(q); 
+                                       uth->uu_siglist &= ~mask;       /* clear pending signal */
+                                       proc_unlock(q);
+                               }
                        }
+                       
+                       /* Clean up */
+                       (void)reap_child_locked(q, p, 0, 0, 0);
 
                        return (0);
                }
-               if (p->p_stat == SSTOP && (p->p_flag & P_WAITED) == 0 &&
-                   (p->p_flag & P_TRACED || uap->options & WUNTRACED)) {
-                       p->p_flag |= P_WAITED;
+               if (p->p_stat == SSTOP && (p->p_lflag & P_LWAITED) == 0 &&
+                   (p->p_lflag & P_LTRACED || uap->options & WUNTRACED)) {
+                       proc_list_unlock();
+#if CONFIG_MACF
+                       if ((error = mac_proc_check_wait(q, p)) != 0)
+                               goto out;
+#endif
+                       proc_lock(p);
+                       p->p_lflag |= P_LWAITED;
+                       proc_unlock(p);
                        retval[0] = p->p_pid;
                        if (uap->status) {
                                status = W_STOPCODE(p->p_xstat);
@@ -708,25 +1118,57 @@ loop:
                                    sizeof(status));
                        } else
                                error = 0;
-                       p->p_lflag &= ~P_LWAITING;
-                       wakeup(&p->p_stat);
-                       return (error);
+                       goto out;
+               }
+               /*
+                * If we are waiting for continued processses, and this
+                * process was continued
+                */
+               if ((uap->options & WCONTINUED) &&
+                   (p->p_flag & P_CONTINUED)) {
+                       proc_list_unlock();
+#if CONFIG_MACF
+                       if ((error = mac_proc_check_wait(q, p)) != 0)
+                               goto out;
+#endif
+
+                       /* Prevent other process for waiting for this event */
+                       OSBitAndAtomic(~((uint32_t)P_CONTINUED), &p->p_flag);
+                       retval[0] = p->p_pid;
+                       if (uap->status) {
+                               status = W_STOPCODE(SIGCONT);
+                               error = copyout((caddr_t)&status,
+                                       uap->status,
+                                   sizeof(status));
+                       } else
+                               error = 0;
+                       goto out;
                }
-               p->p_lflag &= ~P_LWAITING;
+               p->p_listflag &= ~P_LIST_WAITING;
                wakeup(&p->p_stat);
        }
-       if (nfound == 0)
+       /* list lock is held when we get here any which way */
+       if (nfound == 0) {
+               proc_list_unlock();
                return (ECHILD);
+       }
 
        if (uap->options & WNOHANG) {
                retval[0] = 0;
+               proc_list_unlock();
                return (0);
        }
 
-       if ((error = tsleep0((caddr_t)q, PWAIT | PCATCH, "wait", 0, wait1continue)))
+       if ((error = msleep0((caddr_t)q, proc_list_mlock, PWAIT | PCATCH | PDROP, "wait", 0, wait1continue)))
                return (error);
 
        goto loop;
+out:
+       proc_list_lock();
+       p->p_listflag &= ~P_LIST_WAITING;
+       wakeup(&p->p_stat);
+       proc_list_unlock();
+       return (error);
 }
 
 
@@ -736,16 +1178,14 @@ waitidcontinue(int result)
        void *vt;
        thread_t thread;
        int *retval;
-       struct proc *p;
 
        if (result)
                return(result);
 
-       p = current_proc();
        thread = current_thread();
        vt = get_bsduthreadarg(thread);
        retval = get_bsduthreadrval(thread);
-       return(waitid((struct proc *)p, (struct waitid_args *)vt, retval));
+       return(waitid(current_proc(), (struct waitid_args *)vt, retval));
 }
 
 /*
@@ -762,15 +1202,48 @@ waitidcontinue(int result)
  *             !0                      Error returning status to user space
  */
 int
-waitid(struct proc *q, struct waitid_args *uap, register_t *retval)
+waitid(proc_t q, struct waitid_args *uap, int32_t *retval)
+{
+       __pthread_testcancel(1);
+       return(waitid_nocancel(q, (struct waitid_nocancel_args *)uap, retval));
+}
+
+int
+waitid_nocancel(proc_t q, struct waitid_nocancel_args *uap, __unused int32_t *retval)
 {
        user_siginfo_t  collect64;      /* siginfo data to return to caller */
 
-       register int nfound;
-       register struct proc *p;
+       int nfound;
+       proc_t p;
        int error;
 
+       /*
+        * Forced validation of options for T.waitpid 21; should be a TSD!
+        * This will pass the test, but note that we have more bits than the
+        * standard specifies that we will allow in, in this case.  The test
+        * passes because they light all the bits, not just the ones we allow,
+        * and so the following check returns EINVAL like the test wants.
+        */
+       if (((uap->options & (WNOHANG|WNOWAIT|WCONTINUED|WUNTRACED|WSTOPPED|WEXITED)) != uap->options) ||
+           (uap->options == 0))
+               return (EINVAL);        /* bits set that aren't recognized */
+
+       /*
+        * Overly critical options checking, per POSIX
+        */
+       switch(uap->idtype) {
+       case P_PID:     /* child with process ID equal to... */
+       case P_PGID:    /* child with process group ID equal to... */
+               if (((int)uap->id) < 0)
+                       return (EINVAL);
+               break;
+       case P_ALL:     /* any child */
+               break;
+       }
+
 loop:
+       proc_list_lock();
+loop1:
        nfound = 0;
        for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
                switch(uap->idtype) {
@@ -779,7 +1252,7 @@ loop:
                                continue;
                        break;
                case P_PGID:    /* child with process group ID equal to... */
-                       if (p->p_pgid != (pid_t)uap->id)
+                       if (p->p_pgrpid != (pid_t)uap->id)
                                continue;
                        break;
                case P_ALL:     /* any child */
@@ -792,11 +1265,11 @@ loop:
                 * Wait collision; go to sleep and restart; used to maintain
                 * the single return for waited process guarantee.
                 */
-               if (p->p_lflag & P_LWAITING) {
-                       (void)tsleep(&p->p_stat, PWAIT, "waitidcoll", 0);
-                       goto loop;
+               if (p->p_listflag & P_LIST_WAITING) {
+                       (void)msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitidcoll", 0);
+                       goto loop1;
                }
-               p->p_lflag |= P_LWAITING;               /* mark busy */
+               p->p_listflag |= P_LIST_WAITING;                /* mark busy */
 
                nfound++;
 
@@ -810,41 +1283,48 @@ loop:
                        if (!(uap->options & WEXITED))
                                break;
 
+                       /* drop the lock and the thread is going to return */
+                       proc_list_unlock();
+
                        /* Collect "siginfo" information for caller */
-                       collect64.si_signo = 0;
+                       collect64.si_signo = SIGCHLD;
                        collect64.si_code = 0;
                        collect64.si_errno = 0;
                        collect64.si_pid = 0;
                        collect64.si_uid = 0;
                        collect64.si_addr  = 0;
-                       collect64.si_status = p->p_xstat;
+                       collect64.si_status = WEXITSTATUS(p->p_xstat);
                        collect64.si_band = 0;
 
                        if (IS_64BIT_PROCESS(p)) {
-                               error = copyout((caddr_t)&collect64,
+                               user64_siginfo_t sinfo64;
+                               
+                               siginfo_user_to_user64(&collect64, &sinfo64);
+                               
+                               error = copyout((caddr_t)&sinfo64,
                                        uap->infop,
-                                       sizeof(collect64));
+                                       sizeof(sinfo64));
                        } else {
-                               siginfo_t collect;
-                               siginfo_64to32(&collect64,&collect);
-                               error = copyout((caddr_t)&collect,
-                                       uap->infop,
-                                       sizeof(collect));
+                               user32_siginfo_t sinfo32;
+                               
+                               siginfo_user_to_user32(&collect64, &sinfo32);
+                               
+                               error = copyout((caddr_t)&sinfo32,
+                                                               uap->infop,
+                                                               sizeof(sinfo32));
                        }
                        /* information unavailable? */
-                       if (error) {
-                               p->p_lflag &= ~P_LWAITING;
-                               wakeup(&p->p_stat);
-                               return (error);
-                       }
+                       if (error) 
+                               goto out;
 
                        /* Prevent other process for waiting for this event? */
                        if (!(uap->options & WNOWAIT)) {
                                /* Clean up */
-                               if (!reap_child_process(q, p)) {
-                                       p->p_lflag &= ~P_LWAITING;
-                                       wakeup(&p->p_stat);
-                               }
+                               (void)reap_child_locked(q, p, 0, 0, 0);
+                       } else {
+                               proc_list_lock();
+                               p->p_listflag &= ~P_LIST_WAITING;
+                               proc_list_unlock();
                        }
 
                        return (0);
@@ -861,53 +1341,54 @@ loop:
                         * If someone has already waited it, we lost a race
                         * to be the one to return status.
                         */
-                       if ((p->p_flag & P_WAITED) != 0)
+                       if ((p->p_lflag & P_LWAITED) != 0)
                                break;
 
-                       /*
-                        * If this is not a traced process, and they haven't
-                        * indicated an interest in untraced processes, then
-                        * ignore this one.
-                        */
-                       if (!(p->p_flag & P_TRACED) && !(uap->options & WUNTRACED))
-                               break;
+                       /* drop the lock and the thread is going to return */
+                       proc_list_unlock();
 
                        /* Collect "siginfo" information for caller */
-                       collect64.si_signo = 0;
+                       collect64.si_signo = SIGCHLD;
                        collect64.si_code = 0;
                        collect64.si_errno = 0;
                        collect64.si_pid = 0;
                        collect64.si_uid = 0;
                        collect64.si_addr  = 0;
+                       proc_lock(p);
                        collect64.si_status = p->p_xstat;
+                       proc_unlock(p);
                        collect64.si_band = 0;
 
                        if (IS_64BIT_PROCESS(p)) {
-                               error = copyout((caddr_t)&collect64,
-                                       uap->infop,
-                                       sizeof(collect64));
+                               user64_siginfo_t sinfo64;
+                               
+                               siginfo_user_to_user64(&collect64, &sinfo64);
+                               
+                               error = copyout((caddr_t)&sinfo64,
+                                                               uap->infop,
+                                                               sizeof(sinfo64));
                        } else {
-                               siginfo_t collect;
-                               siginfo_64to32(&collect64,&collect);
-                               error = copyout((caddr_t)&collect,
-                                       uap->infop,
-                                       sizeof(collect));
+                               user32_siginfo_t sinfo32;
+                               
+                               siginfo_user_to_user32(&collect64, &sinfo32);
+                               
+                               error = copyout((caddr_t)&sinfo32,
+                                                               uap->infop,
+                                                               sizeof(sinfo32));
                        }
                        /* information unavailable? */
-                       if (error) {
-                               p->p_lflag &= ~P_LWAITING;
-                               wakeup(&p->p_stat);
-                               return (error);
-                       }
+                       if (error)
+                               goto out;
 
                        /* Prevent other process for waiting for this event? */
                        if (!(uap->options & WNOWAIT)) {
-                               p->p_flag |= P_WAITED;
+                               proc_lock(p);
+                               p->p_lflag |= P_LWAITED;
+                               proc_unlock(p);
                        }
 
-                       p->p_lflag &= ~P_LWAITING;
-                       wakeup(&p->p_stat);
-                       return (0);
+                       error = 0;
+                       goto out;
 
                default:                /* All others */
                                        /* ...meaning Continued */
@@ -922,119 +1403,124 @@ loop:
                        if ((p->p_flag & P_CONTINUED) == 0)
                                break;
 
+                       /* drop the lock and the thread is going to return */
+                       proc_list_unlock();
+
                        /* Collect "siginfo" information for caller */
-                       collect64.si_signo = 0;
-                       collect64.si_code = 0;
+                       proc_lock(p);
+                       collect64.si_signo = SIGCHLD;
+                       collect64.si_code = CLD_CONTINUED;
                        collect64.si_errno = 0;
-                       collect64.si_pid = 0;
+                       collect64.si_pid = p->p_contproc;
                        collect64.si_uid = 0;
                        collect64.si_addr  = 0;
                        collect64.si_status = p->p_xstat;
                        collect64.si_band = 0;
+                       proc_unlock(p);
 
                        if (IS_64BIT_PROCESS(p)) {
-                               error = copyout((caddr_t)&collect64,
-                                       uap->infop,
-                                       sizeof(collect64));
+                               user64_siginfo_t sinfo64;
+                               
+                               siginfo_user_to_user64(&collect64, &sinfo64);
+                               
+                               error = copyout((caddr_t)&sinfo64,
+                                                               uap->infop,
+                                                               sizeof(sinfo64));
                        } else {
-                               siginfo_t collect;
-                               siginfo_64to32(&collect64,&collect);
-                               error = copyout((caddr_t)&collect,
-                                       uap->infop,
-                                       sizeof(collect));
+                               user32_siginfo_t sinfo32;
+                               
+                               siginfo_user_to_user32(&collect64, &sinfo32);
+                               
+                               error = copyout((caddr_t)&sinfo32,
+                                                               uap->infop,
+                                                               sizeof(sinfo32));
                        }
                        /* information unavailable? */
-                       if (error) {
-                               p->p_lflag &= ~P_LWAITING;
-                               wakeup(&p->p_stat);
-                               return (error);
-                       }
+                       if (error)
+                               goto out;
 
                        /* Prevent other process for waiting for this event? */
                        if (!(uap->options & WNOWAIT)) {
-                               p->p_flag &= ~P_CONTINUED;
+                               OSBitAndAtomic(~((uint32_t)P_CONTINUED), &p->p_flag);
                        }
 
-                       p->p_lflag &= ~P_LWAITING;
-                       wakeup(&p->p_stat);
-                       return (0);
-
-                       break;
+                       error = 0;
+                       goto out;
                }
-
-
+               /* LIST LOCK IS HELD HERE */
                /* Not a process we are interested in; go on to next child */
-               p->p_lflag &= ~P_LWAITING;
+               
+               p->p_listflag &= ~P_LIST_WAITING;
                wakeup(&p->p_stat);
        }
 
+       /* list lock is always held */
        /* No child processes that could possibly satisfy the request? */
-       if (nfound == 0)
+       if (nfound == 0) {
+               proc_list_unlock();
                return (ECHILD);
+       }
 
        if (uap->options & WNOHANG) {
-               retval[0] = 0;
+               proc_list_unlock();
                return (0);
        }
 
-       if ((error = tsleep0((caddr_t)q, PWAIT | PCATCH, "waitid", 0, waitidcontinue)))
+       if ((error = msleep0((caddr_t)q, proc_list_mlock, PWAIT | PCATCH | PDROP, "waitid", 0, waitidcontinue)))
                return (error);
 
        goto loop;
+out:
+       proc_list_lock();
+       p->p_listflag &= ~P_LIST_WAITING;
+       wakeup(&p->p_stat);
+       proc_list_unlock();
+       return (error);
 }
 
 /*
  * make process 'parent' the new parent of process 'child'.
  */
 void
-proc_reparent(struct proc *child, struct proc *parent)
+proc_reparentlocked(proc_t child, proc_t parent, int cansignal, int locked)
 {
+       proc_t oldparent = PROC_NULL;
 
        if (child->p_pptr == parent)
                return;
 
+       if (locked == 0)
+               proc_list_lock();
+
+       oldparent = child->p_pptr;
+#if __PROC_INTERNAL_DEBUG
+       if (oldparent == PROC_NULL)
+               panic("proc_reparent: process %p does not have a parent\n", child);
+#endif
+
        LIST_REMOVE(child, p_sibling);
+#if __PROC_INTERNAL_DEBUG
+       if (oldparent->p_childrencnt == 0)
+               panic("process children count already 0\n");
+#endif
+       oldparent->p_childrencnt--;
+#if __PROC_INTERNAL_DEBUG1
+       if (oldparent->p_childrencnt < 0)
+               panic("process children count -ve\n");
+#endif
        LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
+       parent->p_childrencnt++;        
        child->p_pptr = parent;
+       child->p_ppid = parent->p_pid;
 
-       if (initproc == parent && child->p_stat == SZOMB)
-               psignal(initproc, SIGCHLD);
-}
-
-/*
- *     Make the current process an "init" process, meaning
- *     that it doesn't have a parent, and that it won't be
- *     gunned down by kill(-1, 0).
- */
-kern_return_t
-init_process(__unused struct init_process_args *args)
-{
-       register struct proc *p = current_proc();
+       proc_list_unlock();
 
-       AUDIT_MACH_SYSCALL_ENTER(AUE_INITPROCESS);
-       if (suser(kauth_cred_get(), &p->p_acflag)) {
-               AUDIT_MACH_SYSCALL_EXIT(KERN_NO_ACCESS);
-               return(KERN_NO_ACCESS);
-       }
-
-       if (p->p_pid != 1 && p->p_pgid != p->p_pid)
-               enterpgrp(p, p->p_pid, 0);
-       p->p_flag |= P_SYSTEM;
-
-       /*
-        *      Take us out of the sibling chain, and
-        *      out of our parent's child chain.
-        */
-       LIST_REMOVE(p, p_sibling);
-       p->p_sibling.le_prev = NULL;
-       p->p_sibling.le_next = NULL;
-       p->p_pptr = kernproc;
-
-       AUDIT_MACH_SYSCALL_EXIT(KERN_SUCCESS);
-       return(KERN_SUCCESS);
+       if ((cansignal != 0) && (initproc == parent) && (child->p_stat == SZOMB))
+               psignal(initproc, SIGCHLD);
+       if (locked == 1)
+               proc_list_lock();
 }
 
-
 /*
  * Exit: deallocate address space and other resources, change proc state
  * to zombie, and unlink proc from allproc and parent's lists.  Save exit
@@ -1042,15 +1528,19 @@ init_process(__unused struct init_process_args *args)
  */
 
 void
-vfork_exit(struct proc *p, int rv)
+vfork_exit(proc_t p, int rv)
+{
+       vfork_exit_internal(p, rv, 0);
+}
+
+void
+vfork_exit_internal(proc_t p, int rv, int forceexit)
 {
        thread_t self = current_thread();
 #ifdef FIXME
        struct task *task = p->task;
 #endif
-       register int s;
        struct uthread *ut;
-       exception_data_t        code[EXCEPTION_CODE_MAX];
 
        /*
         * If a thread in this task has already
@@ -1059,39 +1549,31 @@ vfork_exit(struct proc *p, int rv)
         */
 
         ut = get_bsdthread_info(self);
-#ifdef FIXME
-        signal_lock(p);
-       while (p->exit_thread != self) {
-               if (sig_try_locked(p) <= 0) {
-                       if (get_threadtask(self) != task) {
-                                signal_unlock(p);
-                               return;
-                        }
-                       signal_unlock(p);
-                       thread_terminate(self);
-                       thread_funnel_set(kernel_flock, FALSE);
-                       thread_exception_return();
-                       /* NOTREACHED */
-               }
-               sig_lock_to_exit(p);
-       }
-        signal_unlock(p);
-       if (p->p_pid == 1) {
-               printf("pid 1 exited (signal %d, exit %d)",
-                   WTERMSIG(rv), WEXITSTATUS(rv));
-panic("init died\nState at Last Exception:\n\n%s", init_task_failure_data);
-       }
-#endif /* FIXME */
 
-       s = splsched();
-       p->p_flag |= P_WEXIT;
-       p->p_lflag |= P_LPEXIT;
-       splx(s);
 
-       code[0] = (exception_data_t)0xFF000001;         /* Set terminate code */
-       code[1] = (exception_data_t)p->p_pid;           /* Pass out the pid */
-       /* Notify the perf server */
-       (void)sys_perf_notify(p->task, (exception_data_t)&code, 2);
+       proc_lock(p);
+        if ((p->p_lflag & P_LPEXIT) == P_LPEXIT) {
+               /* 
+               * This happens when a parent exits/killed and vfork is in progress  
+               * other threads. But shutdown code for ex has already called exit1()
+               */
+               proc_unlock(p);
+               return;
+       }
+       p->p_lflag |= (P_LEXIT | P_LPEXIT);
+       proc_unlock(p);
+
+       if (forceexit == 0) {
+               /*
+                * parent of a vfork child has already called exit() and the 
+                * thread that has vfork in proress terminates. So there is no
+                * separate address space here and it has already been marked for
+                * termination. This was never covered before and could cause problems
+                * if we block here for outside code.
+                */
+               /* Notify the perf server */
+               (void)sys_perf_notify(self, p->p_pid);
+       }
 
        /*
         * Remove proc from allproc queue and from pidhash chain.
@@ -1099,49 +1581,77 @@ panic("init died\nState at Last Exception:\n\n%s", init_task_failure_data);
         * Not doing causes things like mount() find this on allproc
         * in partially cleaned state.
         */
+
+       proc_list_lock();
+
        LIST_REMOVE(p, p_list);
        LIST_INSERT_HEAD(&zombproc, p, p_list); /* Place onto zombproc. */
-       LIST_REMOVE(p, p_hash);
-       /*
-        * If parent is waiting for us to exit or exec,
-        * P_PPWAIT is set; we will wakeup the parent below.
-        */
-       p->p_flag &= ~(P_TRACED | P_PPWAIT);
-       p->p_sigignore = ~0;
-       p->p_siglist = 0;
+       /* will not be visible via proc_find */
+       p->p_listflag |= P_LIST_EXITED;
 
-       ut->uu_siglist = 0;
-       untimeout(realitexpire, (caddr_t)p->p_pid);
+       proc_list_unlock();
 
+       proc_lock(p);
        p->p_xstat = rv;
+       p->p_lflag &= ~(P_LTRACED | P_LPPWAIT);
+       p->p_sigignore = ~0;
+       proc_unlock(p);
+
+       proc_spinlock(p);
+       if (thread_call_cancel(p->p_rcall))
+               p->p_ractive--;
+
+       while (p->p_ractive > 0) {
+               proc_spinunlock(p);
+               
+               delay(1);
+
+               proc_spinlock(p);
+       }
+       proc_spinunlock(p);
+
+       thread_call_free(p->p_rcall);
+       p->p_rcall = NULL;
+
+       ut->uu_siglist = 0;
 
        vproc_exit(p);
 }
 
 void 
-vproc_exit(struct proc *p)
+vproc_exit(proc_t p)
 {
-       register struct proc *q, *nq, *pp;
+       proc_t q;
+       proc_t pp;
+       
+       vnode_t tvp;
 #ifdef FIXME
        struct task *task = p->task;
 #endif
+       struct pgrp * pg;
+       struct session *sessp;
 
        /* XXX Zombie allocation may fail, in which case stats get lost */
        MALLOC_ZONE(p->p_ru, struct rusage *,
                        sizeof (*p->p_ru), M_ZOMBIE, M_WAITOK);
 
+
+       proc_refdrain(p);
+
        /*
         * Close open files and release open-file table.
         * This may block!
         */
        fdfree(p);
 
-       if (SESS_LEADER(p)) {
-               register struct session *sp = p->p_session;
-
-               if (sp->s_ttyvp) {
+       sessp = proc_session(p);
+       if (SESS_LEADER(p, sessp)) {
+               
+               if (sessp->s_ttyvp != NULLVP) {
                        struct vnode *ttyvp;
+                       int ttyvid;
                        struct vfs_context context;
+                       struct tty * tp;
 
                        /*
                         * Controlling process.
@@ -1149,77 +1659,137 @@ vproc_exit(struct proc *p)
                         * drain controlling terminal
                         * and revoke access to controlling terminal.
                         */
-                       if (sp->s_ttyp->t_session == sp) {
-                               if (sp->s_ttyp->t_pgrp)
-                                       pgsignal(sp->s_ttyp->t_pgrp, SIGHUP, 1);
-                               (void) ttywait(sp->s_ttyp);
+                       tp = SESSION_TP(sessp);
+
+                       if ((tp != TTY_NULL) && (tp->t_session == sessp)) {
+                               tty_pgsignal(tp, SIGHUP, 1);
+                               tty_lock(tp);
+                               (void) ttywait(tp);
+                               tty_unlock(tp);
                                /*
                                 * The tty could have been revoked
                                 * if we blocked.
                                 */
-                               context.vc_proc = p;
-                               context.vc_ucred = p->p_ucred;
-                               if (sp->s_ttyvp)
-                                       VNOP_REVOKE(sp->s_ttyvp, REVOKEALL, &context);
+
+                               session_lock(sessp);
+                               /* reget in case of race */
+                               tp = SESSION_TP(sessp);
+                               ttyvp = sessp->s_ttyvp;
+                               ttyvid = sessp->s_ttyvid;
+                               sessp->s_ttyvp = NULL;
+                               sessp->s_ttyvid = 0;
+                               sessp->s_ttyp = TTY_NULL;
+                               sessp->s_ttypgrpid = NO_PID;
+                               session_unlock(sessp);
+
+                              if ((ttyvp != NULLVP) && (vnode_getwithvid(ttyvp, ttyvid) == 0)) {
+                                       context.vc_thread = proc_thread(p); /* XXX */
+                                       context.vc_ucred = kauth_cred_proc_ref(p);
+                                       VNOP_REVOKE(ttyvp, REVOKEALL, &context);
+                                       vnode_put(ttyvp);
+                                       kauth_cred_unref(&context.vc_ucred);
+                               }
+                       } else {
+                               session_lock(sessp);
+                               ttyvp = sessp->s_ttyvp;
+                               sessp->s_ttyvp = NULL;
+                               sessp->s_ttyvid = 0;
+                               sessp->s_ttyp = TTY_NULL;
+                               sessp->s_ttypgrpid = NO_PID;
+                               session_unlock(sessp);
                        }
-                       ttyvp = sp->s_ttyvp;
-                       sp->s_ttyvp = NULL;
-                       if (ttyvp) {
+                       if (ttyvp) 
                                vnode_rele(ttyvp);
-                       }
                        /*
                         * s_ttyp is not zero'd; we use this to indicate
                         * that the session once had a controlling terminal.
                         * (for logging and informational purposes)
                         */
                }
-               sp->s_leader = NULL;
+
+               session_lock(sessp);
+               sessp->s_leader = NULL;
+               session_unlock(sessp);
        }
+       session_rele(sessp);
 
-       fixjobc(p, p->p_pgrp, 0);
-       p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
+       pg = proc_pgrp(p);
+       fixjobc(p, pg, 0);
+       pg_rele(pg);
 
-#if KTRACE
-       /* 
-        * release trace file
-        */
-       p->p_traceflag = 0;     /* don't trace the vnode_rele() */
-       if (p->p_tracep) {
-               struct vnode *tvp = p->p_tracep;
-               p->p_tracep = NULL;
-               vnode_rele(tvp);
-       }
-#endif
+       p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
 
-       while (q = p->p_children.lh_first) {
-               proc_reparent(q, initproc);
-               /*
-                * Traced processes are killed
-                * since their existence means someone is messing up.
-                */
-               if (q->p_flag & P_TRACED) {
-                       q->p_flag &= ~P_TRACED;
-                       if (q->sigwait_thread) {
-                               /*
-                                * The sigwait_thread could be stopped at a
-                                * breakpoint. Wake it up to kill.
-                                * Need to do this as it could be a thread which is not
-                                * the first thread in the task. So any attempts to kill
-                                * the process would result into a deadlock on q->sigwait.
-                                */
-                               thread_resume((thread_t)q->sigwait_thread);
-                               clear_wait(q->sigwait_thread, THREAD_INTERRUPTED);
-                               threadsignal((thread_t)q->sigwait_thread, SIGKILL, 0);
+       proc_list_lock();
+       proc_childdrainstart(p);
+       while ((q = p->p_children.lh_first) != NULL) {
+               q->p_listflag |= P_LIST_DEADPARENT;
+               if (q->p_stat == SZOMB) {
+                       if (p != q->p_pptr)
+                               panic("parent child linkage broken");
+                       /* check for lookups by zomb sysctl */
+                       while ((q->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
+                               msleep(&q->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
+                       }
+                       q->p_listflag |= P_LIST_WAITING;
+                       /*
+                        * This is a named reference and it is not granted
+                        * if the reap is already in progress. So we get
+                        * the reference here exclusively and their can be
+                        * no waiters. So there is no need for a wakeup
+                        * after we are done. AlsO  the reap frees the structure
+                        * and the proc struct cannot be used for wakeups as well. 
+                        * It is safe to use q here as this is system reap
+                        */
+                       (void)reap_child_locked(p, q, 1, 1, 0);
+               } else {
+                       proc_reparentlocked(q, initproc, 0, 1);
+                       /*
+                       * Traced processes are killed
+                       * since their existence means someone is messing up.
+                       */
+                       if (q->p_lflag & P_LTRACED) {
+                               proc_list_unlock();
+                               proc_lock(q);
+                               q->p_lflag &= ~P_LTRACED;
+                               if (q->sigwait_thread) {
+                                       proc_unlock(q);
+                                       /*
+                                       * The sigwait_thread could be stopped at a
+                                       * breakpoint. Wake it up to kill.
+                                       * Need to do this as it could be a thread which is not
+                                       * the first thread in the task. So any attempts to kill
+                                       * the process would result into a deadlock on q->sigwait.
+                                       */
+                                       thread_resume((thread_t)q->sigwait_thread);
+                                       clear_wait(q->sigwait_thread, THREAD_INTERRUPTED);
+                                       threadsignal((thread_t)q->sigwait_thread, SIGKILL, 0);
+                               } else
+                                       proc_unlock(q);
+                                       
+                               psignal(q, SIGKILL);
+                               proc_list_lock();
                        }
-                       psignal(q, SIGKILL);
                }
        }
 
+       proc_childdrainend(p);
+       proc_list_unlock();
+
+       /*
+        * Release reference to text vnode
+        */
+       tvp = p->p_textvp;
+       p->p_textvp = NULL;
+       if (tvp != NULLVP) {
+               vnode_rele(tvp);
+       }
+
        /*
         * Save exit status and final rusage info, adding in child rusage
         * info and self times.  If we were unable to allocate a zombie
         * structure, this information is lost.
         */
+       /* No need for locking here as no one than this thread can access this */
        if (p->p_ru != NULL) {
            *p->p_ru = p->p_stats->p_ru;
            timerclear(&p->p_ru->ru_utime);
@@ -1275,14 +1845,13 @@ vproc_exit(struct proc *p)
        /*
         * Other substructures are freed from wait().
         */
-       FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_SUBPROC);
+       FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
        p->p_stats = NULL;
 
-       FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SUBPROC);
+       FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
        p->p_sigacts = NULL;
 
-       if (--p->p_limit->p_refcnt == 0)
-               FREE_ZONE(p->p_limit, sizeof *p->p_limit, M_SUBPROC);
+       proc_limitdrop(p, 1);
        p->p_limit = NULL;
 
        /*
@@ -1295,31 +1864,68 @@ vproc_exit(struct proc *p)
        /*
         * Notify parent that we're gone.
         */
-       pp = p->p_pptr;
-       if (pp != initproc) {
-               pp->si_pid = p->p_pid;
-               pp->si_status = p->p_xstat;
-               pp->si_code = CLD_EXITED;
-               pp->si_uid = p->p_ucred->cr_ruid;
-       }
-       /* mark as a zombie */
-       p->p_stat = SZOMB;
+       pp = proc_parent(p);
+       if ((p->p_listflag & P_LIST_DEADPARENT) == 0) {
+               if (pp != initproc) {
+                       proc_lock(pp);
+                       pp->si_pid = p->p_pid;
+                       pp->si_status = p->p_xstat;
+                       pp->si_code = CLD_EXITED;
+                       /*
+                        * p_ucred usage is safe as it is an exiting process
+                        * and reference is dropped in reap
+                        */
+                       pp->si_uid = p->p_ucred->cr_ruid;
+                       proc_unlock(pp);
+               }
+               /* mark as a zombie */
+               /* mark as a zombie */
+               /* No need to take proc lock as all refs are drained and
+                * no one except parent (reaping ) can look at this.
+                * The write is to an int and is coherent. Also parent is
+                *  keyed off of list lock for reaping
+                */
+               p->p_stat = SZOMB;
+
+               psignal(pp, SIGCHLD);
 
-       psignal(p->p_pptr, SIGCHLD);
+               /* and now wakeup the parent */
+               proc_list_lock();
+               wakeup((caddr_t)pp);
+               proc_list_unlock();
+       } else {
+               proc_list_lock();
+               /* check for lookups by zomb sysctl */
+               while ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) {
+                       msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0);
+               }
+               p->p_stat = SZOMB;
+               p->p_listflag |= P_LIST_WAITING;
 
-       /* and now wakeup the parent */
-       wakeup((caddr_t)p->p_pptr);
+               /*
+                * This is a named reference and it is not granted
+                * if the reap is already in progress. So we get
+                * the reference here exclusively and their can be
+                * no waiters. So there is no need for a wakeup
+                * after we are done. AlsO  the reap frees the structure
+                * and the proc struct cannot be used for wakeups as well. 
+                * It is safe to use p here as this is system reap
+                */
+               (void)reap_child_locked(pp, p, 0, 1, 1);
+               /* list lock dropped by reap_child_locked */
+       }
+       proc_rele(pp);
 }
 
                                                
 /*
  * munge_rusage
  *     LP64 support - long is 64 bits if we are dealing with a 64 bit user
- *     process.  We munge the kernel (32 bit) version of rusage into the
+ *     process.  We munge the kernel version of rusage into the
  *     64 bit version.
  */
 __private_extern__  void 
-munge_rusage(struct rusage *a_rusage_p, struct user_rusage *a_user_rusage_p)
+munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p)
 {
        /* timeval changes size, so utime and stime need special handling */
        a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec;
@@ -1345,3 +1951,32 @@ munge_rusage(struct rusage *a_rusage_p, struct user_rusage *a_user_rusage_p)
        a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw;
        a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw;
 }
+
+/* For a 64-bit kernel and 32-bit userspace, munging may be needed */
+__private_extern__  void 
+munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p)
+{
+       /* timeval changes size, so utime and stime need special handling */
+       a_user_rusage_p->ru_utime.tv_sec = a_rusage_p->ru_utime.tv_sec;
+       a_user_rusage_p->ru_utime.tv_usec = a_rusage_p->ru_utime.tv_usec;
+       a_user_rusage_p->ru_stime.tv_sec = a_rusage_p->ru_stime.tv_sec;
+       a_user_rusage_p->ru_stime.tv_usec = a_rusage_p->ru_stime.tv_usec;
+       /*
+        * everything else can be a direct assign. We currently ignore
+        * the loss of precision
+        */
+       a_user_rusage_p->ru_maxrss = a_rusage_p->ru_maxrss;
+       a_user_rusage_p->ru_ixrss = a_rusage_p->ru_ixrss;
+       a_user_rusage_p->ru_idrss = a_rusage_p->ru_idrss;
+       a_user_rusage_p->ru_isrss = a_rusage_p->ru_isrss;
+       a_user_rusage_p->ru_minflt = a_rusage_p->ru_minflt;
+       a_user_rusage_p->ru_majflt = a_rusage_p->ru_majflt;
+       a_user_rusage_p->ru_nswap = a_rusage_p->ru_nswap;
+       a_user_rusage_p->ru_inblock = a_rusage_p->ru_inblock;
+       a_user_rusage_p->ru_oublock = a_rusage_p->ru_oublock;
+       a_user_rusage_p->ru_msgsnd = a_rusage_p->ru_msgsnd;
+       a_user_rusage_p->ru_msgrcv = a_rusage_p->ru_msgrcv;
+       a_user_rusage_p->ru_nsignals = a_rusage_p->ru_nsignals;
+       a_user_rusage_p->ru_nvcsw = a_rusage_p->ru_nvcsw;
+       a_user_rusage_p->ru_nivcsw = a_rusage_p->ru_nivcsw;
+}