]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/kern_fork.c
xnu-1228.tar.gz
[apple/xnu.git] / bsd / kern / kern_fork.c
index 02f190c42bb9d75712069365f666b7c532ee8424..fa43edd2eb8b309d31c1761cd861735bb88ffc2b 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_fork.c 8.8 (Berkeley) 2/14/95
  */
+/*
+ * NOTICE: This file was modified by McAfee Research in 2004 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.
+ */
+/*
+ * 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 <kern/assert.h>
 #include <sys/param.h>
 #include <sys/vnode_internal.h>
 #include <sys/file_internal.h>
 #include <sys/acct.h>
-#if KTRACE
-#include <sys/ktrace.h>
+#include <sys/codesign.h>
+#include <sys/sysproto.h>
+#if CONFIG_DTRACE
+/* Do not include dtrace.h, it redefines kmem_[alloc/free] */
+extern void dtrace_fasttrap_fork(proc_t, proc_t);
+extern void (*dtrace_helpers_fork)(proc_t, proc_t);
+extern void dtrace_lazy_dofs_duplicate(proc_t, proc_t);
+
+#include <sys/dtrace_ptss.h>
 #endif
 
 #include <bsm/audit_kernel.h>
 #include <kern/kalloc.h>
 #include <kern/mach_param.h>
 #include <kern/task.h>
+#include <kern/thread_call.h>
 #include <kern/zalloc.h>
 
 #include <machine/spl.h>
 
-#include <vm/vm_protos.h>       // for vm_map_commpage64
+#if CONFIG_MACF
+#include <security/mac.h>
+#include <security/mac_mach_internal.h>
+#endif
+
+#include <vm/vm_map.h>
+#include <vm/vm_protos.h>
+#include <vm/vm_shared_region.h>
+
+#include <sys/shm_internal.h>  /* for shmfork() */
+#include <mach/task.h>         /* for thread_create() */
+#include <mach/thread_act.h>   /* for thread_resume() */
 
-thread_t cloneproc(struct proc *, int); 
-struct proc * forkproc(struct proc *, int);
-thread_t procdup(struct proc *child, struct proc *parent);
+#include <sys/sdt.h>
+
+/* XXX routines which should have Mach prototypes, but don't */
+void thread_set_parent(thread_t parent, int pid);
+extern void act_thread_catt(void *ctx);
+void thread_set_child(thread_t child, int pid);
+void *act_thread_csave(void);
+
+
+thread_t cloneproc(proc_t, int); 
+proc_t forkproc(proc_t, int);
+void forkproc_free(proc_t, int);
+thread_t procdup(proc_t parent, proc_t child);
+thread_t fork_create_child(task_t parent_task, proc_t child, int inherit_memory, int is64bit);
 
 #define        DOFORK  0x1     /* fork() system call */
 #define        DOVFORK 0x2     /* vfork() system call */
-static int fork1(struct proc *, long, register_t *);
 
-/*
- * fork system call.
- */
-int
-fork(struct proc *p, __unused void *uap, register_t *retval)
-{
-       return (fork1(p, (long)DOFORK, retval));
-}
 
 /*
- * vfork system call
+ * vfork
+ *
+ * Description:        vfork system call
+ *
+ * Parameters: void                    [no arguments]
+ *
+ * Retval:     0                       (to child process)
+ *             !0                      pid of child (to parent process)
+ *             -1                      error (see "Returns:")
+ *
+ * Returns:    EAGAIN                  Administrative limit reached
+ *             EINVAL                  vfork() caled during vfork()
+ *             ENOMEM                  Failed to allocate new process
+ *
+ * Note:       After a successful call to this function, the parent process
+ *             has its task, thread, and uthread lent to the child process,
+ *             and control is returned to the caller; if this function is
+ *             invoked as a system call, the return is to user space, and
+ *             is effectively running on the child process.
+ *
+ *             Subsequent calls that operate on process state are permitted,
+ *             though discouraged, and will operate on the child process; any
+ *             operations on the task, thread, or uthread will result in
+ *             changes in the parent state, and, if inheritable, the child
+ *             state, when a task, thread, and uthread are realized for the
+ *             child process at execve() time, will also be effected.  Given
+ *             this, it's recemmended that people use the posix_spawn() call
+ *             instead.
  */
 int
-vfork(struct proc *p, void *uap, register_t *retval)
+vfork(proc_t parent, __unused struct vfork_args *uap, register_t *retval)
 {
-       register struct proc * newproc;
-       register uid_t uid;
+       proc_t child;
+       uid_t uid;
        thread_t cur_act = (thread_t)current_thread();
        int count;
-       task_t t;
        uthread_t ut;
+#if CONFIG_MACF
+       int err;
+#endif
 
        /*
         * Although process entries are dynamically created, we still keep
@@ -129,50 +196,90 @@ vfork(struct proc *p, void *uap, register_t *retval)
         * 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");
                retval[1] = 0;
                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.
+        * 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 && count > p->p_rlimit[RLIMIT_NPROC].rlim_cur) {
+       if (uid != 0 &&
+           (rlim_t)count > parent->p_rlimit[RLIMIT_NPROC].rlim_cur) {
                (void)chgproccnt(uid, -1);
                return (EAGAIN);
        }
 
-       ut = (struct uthread *)get_bsdthread_info(cur_act);
+       ut = (uthread_t)get_bsdthread_info(cur_act);
        if (ut->uu_flag & UT_VFORK) {
-               printf("vfork called recursively by %s\n", p->p_comm);
+               printf("vfork called recursively by %s\n", parent->p_comm);
                (void)chgproccnt(uid, -1);
                return (EINVAL);
        }
-       p->p_flag  |= P_VFORK;
-       p->p_vforkcnt++;
+
+#if CONFIG_MACF
+       /*
+        * Determine if MAC policies applied to the process will allow
+        * it to fork.
+        */
+       err = mac_proc_check_fork(parent);
+       if (err  != 0) {
+               (void)chgproccnt(uid, -1);
+               return (err);
+       }
+#endif
+
+       proc_lock(parent);
+       parent->p_lflag  |= P_LVFORK;
+       parent->p_vforkcnt++;
+       proc_unlock(parent);
 
        /* The newly created process comes with signal lock held */
-       newproc = (struct proc *)forkproc(p,1);
-
-       AUDIT_ARG(pid, newproc->p_pid);
-
-       LIST_INSERT_AFTER(p, newproc, p_pglist);
-       newproc->p_pptr = p;
-       newproc->task = p->task;
-       LIST_INSERT_HEAD(&p->p_children, newproc, p_sibling);
-       LIST_INIT(&newproc->p_children);
-       LIST_INSERT_HEAD(&allproc, newproc, p_list);
-       LIST_INSERT_HEAD(PIDHASH(newproc->p_pid), newproc, p_hash);
-       TAILQ_INIT(& newproc->p_evlist);
-       newproc->p_stat = SRUN;
-       newproc->p_flag  |= P_INVFORK;
-       newproc->p_vforkact = cur_act;
+       if ((child = forkproc(parent,1)) == NULL) {
+               /* Failed to allocate new process */
+               (void)chgproccnt(uid, -1);
+               /*
+                * XXX kludgy, but necessary without a full flags audit...
+                * XXX these are inherited by the child, which depends on
+                * XXX P_VFORK being set.
+                */
+               proc_lock(parent);
+               parent->p_lflag &= ~P_LVFORK;
+               parent->p_vforkcnt--;
+               proc_unlock(parent);
+               return (ENOMEM);
+       }
+
+#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->p_ucred, child);
+#endif
+
+       AUDIT_ARG(pid, child->p_pid);
+
+       child->task = parent->task;
+
+       /* make child visible */
+       pinsertchild(parent, child);
+
+       child->p_lflag  |= P_LINVFORK;
+       child->p_vforkact = cur_act;
+       child->p_stat = SRUN;
 
        ut->uu_flag |= UT_VFORK;
-       ut->uu_proc = newproc;
+       ut->uu_proc = child;
        ut->uu_userstate = (void *)act_thread_csave();
        ut->uu_vforkmask = ut->uu_sigmask;
 
@@ -182,47 +289,66 @@ vfork(struct proc *p, void *uap, register_t *retval)
                ut->uu_flag &= ~UT_SETUID;
        }
        
-       thread_set_child(cur_act, newproc->p_pid);
+       thread_set_child(cur_act, child->p_pid);
 
-       microtime(&newproc->p_stats->p_start);
-       newproc->p_acflag = AFORK;
+       microtime(&child->p_start);
+       microtime(&child->p_stats->p_start); /* for compat sake */
+       child->p_acflag = AFORK;
 
        /*
         * 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).
         */
-       newproc->p_flag |= P_PPWAIT;
+       child->p_lflag |= P_LPPWAIT;
 
        /* drop the signal lock on the child */
-       signal_unlock(newproc);
+       proc_signalend(child, 0);
+       proc_transend(child, 0);
 
-       retval[0] = newproc->p_pid;
-       retval[1] = 1;                  /* mark child */
+       retval[0] = child->p_pid;
+       retval[1] = 1;                  /* flag child return for user space */
+
+       DTRACE_PROC1(create, proc_t, child);
 
        return (0);
 }
 
 /*
- * Return to parent vfork ehread()
+ * vfork_return
+ *
+ * Description:        "Return" to parent vfork thread() following execve/_exit;
+ *             this is done by reassociating the parent process structure
+ *             with the task, thread, and uthread.
+ *
+ * Parameters: child                   Child process
+ *             retval                  System call return value array
+ *             rval                    Return value to present to parent
+ *
+ * Returns:    void
+ *
+ * Note:       The caller resumes or exits the parent, as appropriate, after
+ *             callling this function.
  */
 void
-vfork_return(__unused thread_t th_act, struct proc *p, struct proc *p2,
-       register_t *retval)
+vfork_return(proc_t child, register_t *retval, int rval)
 {
+       proc_t parent = child->p_pptr;
        thread_t cur_act = (thread_t)current_thread();
        uthread_t ut;
        
-       ut = (struct uthread *)get_bsdthread_info(cur_act);
+       ut = (uthread_t)get_bsdthread_info(cur_act);
 
        act_thread_catt(ut->uu_userstate);
 
        /* Make sure only one at this time */
-       p->p_vforkcnt--;
-       if (p->p_vforkcnt <0)
+       proc_lock(parent);
+       parent->p_vforkcnt--;
+       if (parent->p_vforkcnt <0)
                panic("vfork cnt is -ve");
-       if (p->p_vforkcnt <=0)
-               p->p_flag  &= ~P_VFORK;
+       if (parent->p_vforkcnt <=0)
+               parent->p_lflag  &= ~P_LVFORK;
+       proc_unlock(parent);
        ut->uu_userstate = 0;
        ut->uu_flag &= ~UT_VFORK;
        /* restore thread-set-id state */
@@ -232,69 +358,196 @@ vfork_return(__unused thread_t th_act, struct proc *p, struct proc *p2,
        }
        ut->uu_proc = 0;
        ut->uu_sigmask = ut->uu_vforkmask;
-       p2->p_flag  &= ~P_INVFORK;
-       p2->p_vforkact = (void *)0;
+       child->p_lflag  &= ~P_LINVFORK;
+       child->p_vforkact = (void *)0;
 
-       thread_set_parent(cur_act, p2->p_pid);
+       thread_set_parent(cur_act, rval);
 
        if (retval) {
-               retval[0] = p2->p_pid;
+               retval[0] = rval;
                retval[1] = 0;                  /* mark parent */
        }
 
        return;
 }
 
+
+/*
+ * fork_create_child
+ *
+ * Description:        Common operations associated with the creation of a child
+ *             process
+ *
+ * Parameters: parent_task             parent task
+ *             child                   child process
+ *             inherit_memory          TRUE, if the parents address space is
+ *                                     to be inherited by the child
+ *             is64bit                 TRUE, if the child being created will
+ *                                     be associated with a 64 bit process
+ *                                     rather than a 32 bit process
+ *
+ * Note:       This code is called in the fork() case, from the execve() call
+ *             graph, if implementing an execve() following a vfork(), from
+ *             the posix_spawn() call graph (which implicitly includes a
+ *             vfork() equivalent call, and in the system bootstrap case.
+ *
+ *             It creates a new task and thread (and as a side effect of the
+ *             thread creation, a uthread), which is then associated with the
+ *             process 'child'.  If the parent process address space is to
+ *             be inherited, then a flag indicates that the newly created
+ *             task should inherit this from the child task.
+ *
+ *             As a special concession to bootstrapping the initial process
+ *             in the system, it's possible for 'parent_task' to be TASK_NULL;
+ *             in this case, 'inherit_memory' MUST be FALSE.
+ */
 thread_t
-procdup(struct proc *child, struct proc *parent)
+fork_create_child(task_t parent_task, proc_t child, int inherit_memory, int is64bit)
 {
-       thread_t                thread;
-       task_t                  task;
-       kern_return_t   result;
+       thread_t        child_thread = NULL;
+       task_t          child_task;
+       kern_return_t   result;
+
+       /* Create a new task for the child process */
+       result = task_create_internal(parent_task,
+                                       inherit_memory,
+                                       is64bit,
+                                       &child_task);
+       if (result != KERN_SUCCESS) {
+               printf("execve: task_create_internal failed.  Code: %d\n", result);
+               goto bad;
+       }
 
-       if (parent->task == kernel_task)
-               result = task_create_internal(TASK_NULL, FALSE, &task);
+       /* Set the child task to the new task */
+       child->task = child_task;
+
+       /* Set child task proc to child proc */
+       set_bsdtask_info(child_task, child);
+
+       /* Propagate CPU limit timer from parent */
+       if (timerisset(&child->p_rlim_cpu))
+               task_vtimer_set(child_task, TASK_VTIMER_RLIM);
+
+       /* Set/clear 64 bit vm_map flag */
+       if (is64bit)
+               vm_map_set_64bit(get_task_map(child_task));
        else
-               result = task_create_internal(parent->task, TRUE, &task);
-       if (result != KERN_SUCCESS)
-           printf("fork/procdup: task_create failed. Code: 0x%x\n", result);
-       child->task = task;
-       /* task->proc = child; */
-       set_bsdtask_info(task, child);
-       if (parent->p_flag & P_LP64) {
-               task_set_64bit(task, TRUE);
-               child->p_flag |= P_LP64;
-#ifdef __PPC__
-                /* LP64todo - clean up this hacked mapping of commpage */
-               pmap_map_sharedpage(task, get_map_pmap(get_task_map(task)));
-                vm_map_commpage64(get_task_map(task));
-#endif /* __PPC__ */
-       } else {
-               task_set_64bit(task, FALSE);
-               child->p_flag &= ~P_LP64;
-       }
+               vm_map_set_32bit(get_task_map(child_task));
+
+#if CONFIG_MACF
+       /* Update task for MAC framework */
+       /* valid to use p_ucred as child is still not running ... */
+       mac_task_label_update_cred(child->p_ucred, child_task);
+#endif
+
+       /* Set child scheduler priority if nice value inherited from parent */
        if (child->p_nice != 0)
                resetpriority(child);
-               
-       result = thread_create(task, &thread);
-       if (result != KERN_SUCCESS)
-           printf("fork/procdup: thread_create failed. Code: 0x%x\n", result);
 
-       return(thread);
+       /* Create a new thread for the child process */
+       result = thread_create(child_task, &child_thread);
+       if (result != KERN_SUCCESS) {
+               printf("execve: thread_create failed. Code: %d\n", result);
+               task_deallocate(child_task);
+               child_task = NULL;
+       }
+bad:
+       thread_yield_internal(1);
+
+       return(child_thread);
+}
+
+
+/*
+ * procdup
+ *
+ * Description:        Givben a parent process, provide a duplicate task and thread
+ *             for a child process of that parent.
+ *
+ * Parameters: parent                  Parent process to use as the template
+ *             child                   Child process to duplicate into
+ *
+ * Returns:    !NULL                   Child process thread pointer
+ *             NULL                    Failure (unspecified)
+ *
+ * Note:       Most of the heavy lifting is done by fork_create_child(); this
+ *             function exists more or less to deal with the 64 bit commpage,
+ *             which requires explicit inheritance, the x86 commpage, which
+ *             should not need explicit mapping any more, but apparently does,
+ *             and to be variant for the bootstrap process.
+ *
+ *             There is a special case where the system is being bootstraped,
+ *             where this function will be called from cloneproc(), called in
+ *             turn from bsd_utaskbootstrap().  In this case, we are acting
+ *             to create a task and thread (and uthread) for the benefit of
+ *             the kernel process - the first process in the system (PID 0).
+ *
+ *             In that specific case, we will *not* pass a parent task, since
+ *             there is *not* parent task present to pass.
+ *
+ * XXX:                This function should go away; the variance can moved into
+ * XXX:                cloneproc(), and the 64bit commpage code can be moved into
+ * XXX:                fork_create_child(), after the x86 commpage inheritance is
+ * XXX:                corrected.
+ */
+thread_t
+procdup(proc_t parent, proc_t child)
+{
+       thread_t                child_thread;
+       task_t                  child_task;
+
+       if (parent->task == kernel_task)
+               child_thread = fork_create_child(TASK_NULL, child, FALSE, FALSE);
+       else
+               child_thread = fork_create_child(parent->task, child, TRUE, (parent->p_flag & P_LP64));
+
+       if (child_thread != NULL) {
+               child_task = get_threadtask(child_thread);
+               if (parent->p_flag & P_LP64) {
+                       task_set_64bit(child_task, TRUE);
+                       OSBitOrAtomic(P_LP64, (UInt32 *)&child->p_flag);
+#ifdef __ppc__
+                       /* LP64todo - clean up hacked mapping of commpage */
+                       /* 
+                        * PPC51: ppc64 is limited to 51-bit addresses.
+                        * Memory above that limit is handled specially at
+                        * the pmap level.
+                        */
+                        pmap_map_sharedpage(child_task, get_map_pmap(get_task_map(child_task)));
+#endif /* __ppc__ */
+               } else {
+                       task_set_64bit(child_task, FALSE);
+                       OSBitAndAtomic(~((uint32_t)P_LP64), (UInt32 *)&child->p_flag);
+               }
+       }
+
+       return(child_thread);
 }
 
 
-static int
-fork1(p1, flags, retval)
-       struct proc *p1;
-       long flags;
-       register_t *retval;
+/*
+ * fork
+ *
+ * Description:        fork system call.
+ *
+ * Parameters: parent                  Parent process to fork
+ *             uap (void)              [unused]
+ *             retval                  Return value
+ *
+ * Returns:    0                       Success
+ *             EAGAIN                  Resource unavailable, try again
+ */
+int
+fork(proc_t parent, __unused struct fork_args *uap, register_t *retval)
 {
-       register struct proc *p2;
-       register uid_t uid;
+       proc_t child;
+       uid_t uid;
        thread_t newth;
        int count;
        task_t t;
+#if CONFIG_MACF
+       int err;
+#endif
 
        /*
         * Although process entries are dynamically created, we still keep
@@ -304,47 +557,123 @@ fork1(p1, flags, retval)
         * 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");
                retval[1] = 0;
                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.
+        * 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 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) {
+       if (uid != 0 &&
+           (rlim_t)count > parent->p_rlimit[RLIMIT_NPROC].rlim_cur) {
                (void)chgproccnt(uid, -1);
                return (EAGAIN);
        }
 
+#if CONFIG_MACF
+       /*
+        * Determine if MAC policies applied to the process will allow
+        * it to fork.
+        */
+       err = mac_proc_check_fork(parent);
+       if (err != 0) {
+               (void)chgproccnt(uid, -1);
+               return (err);
+       }
+#endif
+
        /* The newly created process comes with signal lock held */
-       newth = cloneproc(p1, 1);
+       if ((newth = cloneproc(parent, 1)) == NULL) {
+               /* Failed to create thread */
+               (void)chgproccnt(uid, -1);
+               return (EAGAIN);
+       }
+
        thread_dup(newth);
-       /* p2 = newth->task->proc; */
-       p2 = (struct proc *)(get_bsdtask_info(get_threadtask(newth)));
-       set_security_token(p2);         /* propagate change of PID */
+       /* child = newth->task->proc; */
+       child = (proc_t)(get_bsdtask_info(get_threadtask(newth)));
+
+#if CONFIG_MACF
+       /* inform policies of new process sharing this cred/label */
+       /* safe to use p_ucred here since child is not running */
+       /* JMM - unsafe to assume the association will stay - as */
+       /*        there are other ways it can be dropped without */
+       /*        informing the policies. */
+       mac_cred_label_associate_fork(child->p_ucred, child);
+#endif
+
+       /* propogate change of PID - may get new cred if auditing */
+       set_security_token(child);
 
-       AUDIT_ARG(pid, p2->p_pid);
+       AUDIT_ARG(pid, child->p_pid);
 
-       thread_set_child(newth, p2->p_pid);
+       thread_set_child(newth, child->p_pid);
 
-       microtime(&p2->p_stats->p_start);
-       p2->p_acflag = AFORK;
+       microtime(&child->p_start);
+       microtime(&child->p_stats->p_start);    /* for compat sake */
+       child->p_acflag = AFORK;
 
+#if CONFIG_DTRACE
        /*
-        * 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).
+        * 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->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->p_dtrace_count > 0) {
+               dtrace_fasttrap_fork(parent, child);
+       }
+
+       lck_mtx_unlock(&parent->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, child);
+       
+       /*
+        * 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 (flags == DOVFORK)
-               p2->p_flag |= P_PPWAIT;
+       if (parent->p_dtrace_helpers != NULL && dtrace_helpers_fork) {
+               (*dtrace_helpers_fork)(parent, child);
+       }
+
+#endif
+
        /* drop the signal lock on the child */
-       signal_unlock(p2);
+       proc_signalend(child, 0);
+       proc_transend(child, 0);
 
-       (void) thread_resume(newth);
+       /* "Return" to the child */
+       (void)thread_resume(newth);
 
         /* drop the extra references we got during the creation */
         if ((t = (task_t)get_threadtask(newth)) != NULL) {
@@ -352,80 +681,221 @@ fork1(p1, flags, retval)
         }
         thread_deallocate(newth);
 
-       KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
+       proc_knote(parent, NOTE_FORK | child->p_pid);
 
-       while (p2->p_flag & P_PPWAIT)
-               tsleep(p1, PWAIT, "ppwait", 0);
+       retval[0] = child->p_pid;
+       retval[1] = 0;                  /* flag parent */
 
-       retval[0] = p2->p_pid;
-       retval[1] = 0;                  /* mark parent */
+       DTRACE_PROC1(create, proc_t, child);
 
        return (0);
 }
 
 /*
- * cloneproc()
+ * cloneproc
+ *
+ * Description: Create a new process from a specified process.
+ *
+ * Parameters: parent                  The parent process of the process to
+ *                                     be cloned
+ *             lock                    Whether or not the signal lock was held
+ *                                     when calling cloneproc().
+ *
+ * Returns:    !NULL                   pointer to new child thread
+ *             NULL                    Failure (unspecified)
  *
- * Create a new process from a specified process.
- * 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
+ * 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.  This results
+ *             in a side effect in procdup(), which is why the code is more
+ *             complicated at the top of that function.
  */
 thread_t
-cloneproc(p1, lock)
-       register struct proc *p1;
-       register int lock;
+cloneproc(proc_t parent, int lock)
 {
-       register struct proc *p2;
-       thread_t th;
+       proc_t child;
+       thread_t th = NULL;
 
-       p2 = (struct proc *)forkproc(p1,lock);
+       if ((child = forkproc(parent,lock)) == NULL) {
+               /* Failed to allocate new process */
+               goto bad;
+       }
 
+       if ((th = procdup(parent, child)) == NULL) {
+               /*
+                * Failed to create thread; now we must deconstruct the new
+                * process previously obtained from forkproc().
+                */
+               forkproc_free(child, lock);
+               goto bad;
+       }
 
-       th = procdup(p2, p1);   /* child, parent */
+       /* make child visible */
+       pinsertchild(parent, child);
 
-       LIST_INSERT_AFTER(p1, p2, p_pglist);
-       p2->p_pptr = p1;
-       LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling);
-       LIST_INIT(&p2->p_children);
-       LIST_INSERT_HEAD(&allproc, p2, p_list);
-       LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
-       TAILQ_INIT(&p2->p_evlist);
        /*
         * Make child runnable, set start time.
         */
-       p2->p_stat = SRUN;
+       child->p_stat = SRUN;
 
+bad:
        return(th);
 }
 
-struct proc *
-forkproc(p1, lock)
-       register struct proc *p1;
-       register int lock;
+/*
+ * 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()
+ *             lock                    Whether or not the signal lock was held
+ *                                     when calling forkproc().
+ *
+ * Returns:    (void)
+ *
+ * Note:       This function should only be used in an error handler following
+ *             a call to forkproc().  The 'lock' paramenter should be the same
+ *             as the lock parameter passed to forkproc().
+ *
+ *             Operations occur in reverse order of those in forkproc().
+ */
+void
+forkproc_free(proc_t p, int lock)
 {
-       register struct proc *p2, *newproc;
-       static int nextpid = 0, pidchecked = 0;
-
-       /* Allocate new proc. */
-       MALLOC_ZONE(newproc, struct proc *,
-                       sizeof *newproc, M_PROC, M_WAITOK);
-       if (newproc == NULL)
-               panic("forkproc: M_PROC zone exhausted");
-       MALLOC_ZONE(newproc->p_stats, struct pstats *,
-                       sizeof *newproc->p_stats, M_SUBPROC, M_WAITOK);
-       if (newproc->p_stats == NULL)
-               panic("forkproc: M_SUBPROC zone exhausted (p_stats)");
-       MALLOC_ZONE(newproc->p_sigacts, struct sigacts *,
-                       sizeof *newproc->p_sigacts, M_SUBPROC, M_WAITOK);
-       if (newproc->p_sigacts == NULL)
-               panic("forkproc: M_SUBPROC zone exhausted (p_sigacts)");
+
+       /* Drop the signal lock, if it was held */
+       if (lock) {
+               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);
 
        /*
-        * Find an unused process ID.  We remember a range of unused IDs
-        * ready to use (from nextpid+1 through pidchecked-1).
+        * 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);
+}
+
+
+/*
+ * forkproc
+ *
+ * Description:        Create a new process structure, given a parent process
+ *             structure.
+ *
+ * Parameters: parent                  The parent process
+ *             lock                    If the signal lock should be taken on
+ *                                     the newly created 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, int lock)
+{
+       struct proc *  child;   /* Our new process */
+       static int nextpid = 0, pidwrap = 0;
+       int error = 0;
+       struct session *sessp;
+       uthread_t uth_parent = (uthread_t)get_bsdthread_info(current_thread());
+
+       MALLOC_ZONE(child, proc_t , sizeof *child, M_PROC, M_WAITOK);
+       if (child == NULL) {
+               printf("forkproc: M_PROC zone exhausted\n");
+               goto bad;
+       }
+       /* zero it out as we need to insert in hash */
+       bzero(child, sizeof *child);
+
+       MALLOC_ZONE(child->p_stats, struct pstats *,
+                       sizeof *child->p_stats, M_PSTATS, M_WAITOK);
+       if (child->p_stats == NULL) {
+               printf("forkproc: M_SUBPROC zone exhausted (p_stats)\n");
+               FREE_ZONE(child, sizeof *child, M_PROC);
+               child = NULL;
+               goto bad;
+       }
+       MALLOC_ZONE(child->p_sigacts, struct sigacts *,
+                       sizeof *child->p_sigacts, M_SIGACTS, M_WAITOK);
+       if (child->p_sigacts == NULL) {
+               printf("forkproc: M_SUBPROC zone exhausted (p_sigacts)\n");
+               FREE_ZONE(child->p_stats, sizeof *child->p_stats, M_PSTATS);
+               FREE_ZONE(child, sizeof *child, M_PROC);
+               child = NULL;
+               goto bad;
+       }
+       child->p_rcall = thread_call_allocate((thread_call_func_t)realitexpire, child);
+       if (child->p_rcall == NULL) {
+               FREE_ZONE(child->p_sigacts, sizeof *child->p_sigacts, M_SIGACTS);
+               FREE_ZONE(child->p_stats, sizeof *child->p_stats, M_PSTATS);
+               FREE_ZONE(child, sizeof *child, M_PROC);
+               child = NULL;
+               goto bad;
+       }
+
+
+       /*
+        * Find an unused PID.  
+        */
+
+       proc_list_lock();
+
        nextpid++;
 retry:
        /*
@@ -435,59 +905,52 @@ retry:
         */
        if (nextpid >= PID_MAX) {
                nextpid = 100;
-               pidchecked = 0;
+               pidwrap = 1;
        }
-       if (nextpid >= pidchecked) {
-               int doingzomb = 0;
+       if (pidwrap != 0) {
 
-               pidchecked = PID_MAX;
-               /*
-                * Scan the active and zombie procs to check whether this pid
-                * is in use.  Remember the lowest pid that's greater
-                * than nextpid, so we can avoid checking for a while.
-                */
-               p2 = allproc.lh_first;
-again:
-               for (; p2 != 0; p2 = p2->p_list.le_next) {
-                       while (p2->p_pid == nextpid ||
-                           p2->p_pgrp->pg_id == nextpid ||
-                               p2->p_session->s_sid == nextpid) {
-                               nextpid++;
-                               if (nextpid >= pidchecked)
-                                       goto retry;
-                       }
-                       if (p2->p_pid > nextpid && pidchecked > p2->p_pid)
-                               pidchecked = p2->p_pid;
-                       if (p2->p_pgrp && p2->p_pgrp->pg_id > nextpid && 
-                           pidchecked > p2->p_pgrp->pg_id)
-                               pidchecked = p2->p_pgrp->pg_id;
-                       if (p2->p_session->s_sid > nextpid &&
-                               pidchecked > p2->p_session->s_sid)
-                               pidchecked = p2->p_session->s_sid;
-               }
-               if (!doingzomb) {
-                       doingzomb = 1;
-                       p2 = zombproc.lh_first;
-                       goto again;
+               /* 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++;
-       p2 = newproc;
-       p2->p_stat = SIDL;
-       p2->p_shutdownstate = 0;
-       p2->p_pid = nextpid;
+       child->p_pid = nextpid;
+#if 1
+       if (child->p_pid != 0) {
+               if (pfind_locked(child->p_pid) != PROC_NULL)
+                       panic("proc in the list already\n");
+       }
+#endif
+       /* Insert in the hash */
+       child->p_listflag |= (P_LIST_INHASH | P_LIST_INCREATE);
+       LIST_INSERT_HEAD(PIDHASH(child->p_pid), child, p_hash);
+       proc_list_unlock();
+
 
        /*
-        * Make a proc table entry for the new process.
-        * Start by zeroing the section of proc that is zero-initialized,
-        * then copy the section that is copied directly from the parent.
+        * We've identified the PID we are going to use; initialize the new
+        * process structure.
         */
-       bzero(&p2->p_startzero,
-           (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
-       bcopy(&p1->p_startcopy, &p2->p_startcopy,
-           (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
-       p2->vm_shm = (void *)NULL; /* Make sure it is zero */
+       child->p_stat = SIDL;
+       child->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->p_startcopy, &child->p_startcopy,
+           (unsigned) ((caddr_t)&child->p_endcopy - (caddr_t)&child->p_startcopy));
 
        /*
         * Some flags are inherited from the parent.
@@ -495,101 +958,97 @@ again:
         * Increase reference counts on shared objects.
         * The p_stats and p_sigacts substructs are set in vm_fork.
         */
-       p2->p_flag = (p1->p_flag & (P_LP64 | P_CLASSIC | P_AFFINITY));
-       if (p1->p_flag & P_PROFIL)
-               startprofclock(p2);
+       child->p_flag = (parent->p_flag & (P_LP64 | P_TRANSLATED | P_AFFINITY));
+       if (parent->p_flag & P_PROFIL)
+               startprofclock(child);
        /*
         * Note that if the current thread has an assumed identity, this
         * credential will be granted to the new process.
         */
-       p2->p_ucred = kauth_cred_get_with_ref();
-
-       lck_mtx_init(&p2->p_mlock, proc_lck_grp, proc_lck_attr);
-       lck_mtx_init(&p2->p_fdmlock, proc_lck_grp, proc_lck_attr);
-       klist_init(&p2->p_klist);
+       child->p_ucred = kauth_cred_get_with_ref();
 
-       /* bump references to the text vnode */
-       p2->p_textvp = p1->p_textvp;
-       if (p2->p_textvp) {
-               vnode_rele(p2->p_textvp);
+       lck_mtx_init(&child->p_mlock, proc_lck_grp, proc_lck_attr);
+       lck_mtx_init(&child->p_fdmlock, proc_lck_grp, proc_lck_attr);
+#if CONFIG_DTRACE
+       lck_mtx_init(&child->p_dtrace_sprlock, proc_lck_grp, proc_lck_attr);
+#endif
+       lck_spin_init(&child->p_slock, proc_lck_grp, proc_lck_attr);
+       klist_init(&child->p_klist);
+
+       if (child->p_textvp != NULLVP) {
+               /* bump references to the text vnode */
+               /* Need to hold iocount across the ref call */
+               if (vnode_getwithref(child->p_textvp) == 0) {
+                       error = vnode_ref(child->p_textvp);
+                       vnode_put(child->p_textvp);
+                       if (error != 0)
+                               child->p_textvp = NULLVP;
+               }
        }
+
        /* XXX may fail to copy descriptors to child */
-       p2->p_fd = fdcopy(p1);
+       child->p_fd = fdcopy(parent, uth_parent->uu_cdir);
 
-       if (p1->vm_shm) {
+#if SYSV_SHM
+       if (parent->vm_shm) {
                /* XXX may fail to attach shm to child */
-               (void)shmfork(p1,p2);
+               (void)shmfork(parent,child);
        }
+#endif
        /*
-        * If p_limit is still copy-on-write, bump refcnt,
-        * otherwise get a copy that won't be modified.
-        * (If PL_SHAREMOD is clear, the structure is shared
-        * copy-on-write.)
+        * inherit the limit structure to child
         */
-       if (p1->p_limit->p_lflags & PL_SHAREMOD)
-               p2->p_limit = limcopy(p1->p_limit);
-       else {
-               p2->p_limit = p1->p_limit;
-               p2->p_limit->p_refcnt++;
+       proc_limitfork(parent, child);
+
+       if (child->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
+               uint64_t rlim_cur = child->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur;
+               child->p_rlim_cpu.tv_sec = (rlim_cur > __INT_MAX__) ? __INT_MAX__ : rlim_cur;
        }
 
-       bzero(&p2->p_stats->pstat_startzero,
-           (unsigned) ((caddr_t)&p2->p_stats->pstat_endzero -
-           (caddr_t)&p2->p_stats->pstat_startzero));
-       bcopy(&p1->p_stats->pstat_startcopy, &p2->p_stats->pstat_startcopy,
-           ((caddr_t)&p2->p_stats->pstat_endcopy -
-            (caddr_t)&p2->p_stats->pstat_startcopy));
+       bzero(&child->p_stats->pstat_startzero,
+           (unsigned) ((caddr_t)&child->p_stats->pstat_endzero -
+           (caddr_t)&child->p_stats->pstat_startzero));
 
-       bzero(&p2->p_stats->user_p_prof, sizeof(struct user_uprof));
+       bzero(&child->p_stats->user_p_prof, sizeof(struct user_uprof));
 
-       if (p1->p_sigacts != NULL)
-               (void)memcpy(p2->p_sigacts,
-                               p1->p_sigacts, sizeof *p2->p_sigacts);
+       if (parent->p_sigacts != NULL)
+               (void)memcpy(child->p_sigacts,
+                               parent->p_sigacts, sizeof *child->p_sigacts);
        else
-               (void)memset(p2->p_sigacts, 0, sizeof *p2->p_sigacts);
-
-       if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
-               p2->p_flag |= P_CONTROLT;
+               (void)memset(child->p_sigacts, 0, sizeof *child->p_sigacts);
 
-       p2->p_argslen = p1->p_argslen;
-       p2->p_argc = p1->p_argc;
-       p2->p_xstat = 0;
-       p2->p_ru = NULL;
+       sessp = proc_session(parent);
+       if (sessp->s_ttyvp != NULL && parent->p_flag & P_CONTROLT)
+               OSBitOrAtomic(P_CONTROLT, (UInt32 *)&child->p_flag);
+       session_rele(sessp);
 
-       p2->p_debugger = 0;     /* don't inherit */
-       lockinit(&p2->signal_lock, PVM, "signal", 0, 0);
        /* block all signals to reach the process */
-       if (lock)
-               signal_lock(p2);
-       p2->sigwait = FALSE;
-       p2->sigwait_thread = NULL;
-       p2->exit_thread = NULL;
-       p2->user_stack = p1->user_stack;
-       p2->p_vforkcnt = 0;
-       p2->p_vforkact = 0;
-       p2->p_lflag  = 0;
-       p2->p_ladvflag  = 0;
-       p2->p_internalref  = 0;
-       TAILQ_INIT(&p2->p_uthlist);
-       TAILQ_INIT(&p2->aio_activeq);
-       TAILQ_INIT(&p2->aio_doneq);
-       p2->aio_active_count = 0;
-       p2->aio_done_count = 0;
-
-#if KTRACE
-       /*
-        * Copy traceflag and tracefile if enabled.
-        * If not inherited, these were zeroed above.
-        */
-       if (p1->p_traceflag&KTRFAC_INHERIT) {
-               p2->p_traceflag = p1->p_traceflag;
-               if ((p2->p_tracep = p1->p_tracep) != NULL) {
-                       vnode_ref(p2->p_tracep);
-               }
+       if (lock) {
+               proc_signalstart(child, 0);
+               proc_transstart(child, 0);
+       }
+
+       TAILQ_INIT(&child->p_uthlist);
+       TAILQ_INIT(&child->aio_activeq);
+       TAILQ_INIT(&child->aio_doneq);
+       /* Inherit the parent flags for code sign */
+       child->p_csflags = parent->p_csflags;
+       child->p_wqthread = parent->p_wqthread;
+       child->p_threadstart = parent->p_threadstart;
+       child->p_pthsize = parent->p_pthsize;
+       workqueue_init_lock(child);
+
+#if CONFIG_LCTX
+       child->p_lctx = NULL;
+       /* Add new process to login context (if any). */
+       if (parent->p_lctx != NULL) {
+               LCTX_LOCK(parent->p_lctx);
+               enterlctx(child, parent->p_lctx, 0);
        }
 #endif
-       return(p2);
 
+bad:
+       return(child);
 }
 
 void
@@ -604,12 +1063,36 @@ proc_unlock(proc_t p)
        lck_mtx_unlock(&p->p_mlock);
 }
 
+void
+proc_spinlock(proc_t p)
+{
+       lck_spin_lock(&p->p_slock);
+}
+
+void
+proc_spinunlock(proc_t p)
+{
+       lck_spin_unlock(&p->p_slock);
+}
+
+void 
+proc_list_lock(void)
+{
+       lck_mtx_lock(proc_list_mlock);
+}
+
+void 
+proc_list_unlock(void)
+{
+       lck_mtx_unlock(proc_list_mlock);
+}
+
 #include <kern/zalloc.h>
 
 struct zone    *uthread_zone;
-int uthread_zone_inited = 0;
+static int uthread_zone_inited = 0;
 
-void
+static void
 uthread_zone_init(void)
 {
        if (!uthread_zone_inited) {
@@ -622,12 +1105,12 @@ uthread_zone_init(void)
 }
 
 void *
-uthread_alloc(task_t task, thread_t thr_act )
+uthread_alloc(task_t task, thread_t thread)
 {
-       struct proc *p;
-       struct uthread *uth, *uth_parent;
+       proc_t p;
+       uthread_t uth;
+       uthread_t uth_parent;
        void *ut;
-       boolean_t funnel_state;
 
        if (!uthread_zone_inited)
                uthread_zone_init();
@@ -635,8 +1118,8 @@ uthread_alloc(task_t task, thread_t thr_act )
        ut = (void *)zalloc(uthread_zone);
        bzero(ut, sizeof(struct uthread));
 
-       p = (struct proc *) get_bsdtask_info(task);
-       uth = (struct uthread *)ut;
+       p = (proc_t) get_bsdtask_info(task);
+       uth = (uthread_t)ut;
 
        /*
         * Thread inherits credential from the creating thread, if both
@@ -646,48 +1129,59 @@ uthread_alloc(task_t task, thread_t thr_act )
         * task we can leave the new thread credential NULL.  If it needs
         * one later, it will be lazily assigned from the task's process.
         */
-       uth_parent = (struct uthread *)get_bsdthread_info(current_thread());
-       if ((task == current_task()) && 
-           (uth_parent != NULL) &&
-           (uth_parent->uu_ucred != NOCRED)) {
+       uth_parent = (uthread_t)get_bsdthread_info(current_thread());
+       if (task == current_task() && 
+           uth_parent != NULL &&
+           IS_VALID_CRED(uth_parent->uu_ucred)) {
+               /*
+                * XXX The new thread is, in theory, being created in context
+                * XXX of parent thread, so a direct reference to the parent
+                * XXX is OK.
+                */
+               kauth_cred_ref(uth_parent->uu_ucred);
                uth->uu_ucred = uth_parent->uu_ucred;
-               kauth_cred_ref(uth->uu_ucred);
                /* the credential we just inherited is an assumed credential */
                if (uth_parent->uu_flag & UT_SETUID)
                        uth->uu_flag |= UT_SETUID;
        } else {
                uth->uu_ucred = NOCRED;
        }
+
        
-       if (task != kernel_task) {
+       if ((task != kernel_task) && p) {
                
-               funnel_state = thread_funnel_set(kernel_flock, TRUE);
+               proc_lock(p);
                if (uth_parent) {
                        if (uth_parent->uu_flag & UT_SAS_OLDMASK)
                                uth->uu_sigmask = uth_parent->uu_oldmask;
                        else
                                uth->uu_sigmask = uth_parent->uu_sigmask;
                }
-               uth->uu_act = thr_act;
-               //signal_lock(p);
-               if (p) {
-                       TAILQ_INSERT_TAIL(&p->p_uthlist, uth, uu_list);
+               uth->uu_context.vc_thread = thread;
+               TAILQ_INSERT_TAIL(&p->p_uthlist, uth, uu_list);
+               proc_unlock(p);
+
+#if CONFIG_DTRACE
+               if (p->p_dtrace_ptss_pages != NULL) {
+                       uth->t_dtrace_scratch = dtrace_ptss_claim_entry(p);
                }
-               //signal_unlock(p);
-               (void)thread_funnel_set(kernel_flock, funnel_state);
+#endif
        }
 
        return (ut);
 }
 
 
+/* 
+ * This routine frees all the BSD context in uthread except the credential.
+ * It does not free the uthread structure as well
+ */
 void
-uthread_free(task_t task, void *uthread, void * bsd_info)
+uthread_cleanup(task_t task, void *uthread, void * bsd_info)
 {
        struct _select *sel;
-       struct uthread *uth = (struct uthread *)uthread;
-       struct proc * p = (struct proc *)bsd_info;
-       boolean_t funnel_state;
+       uthread_t uth = (uthread_t)uthread;
+       proc_t p = (proc_t)bsd_info;
 
        /*
         * Per-thread audit state should never last beyond system
@@ -702,26 +1196,59 @@ uthread_free(task_t task, void *uthread, void * bsd_info)
        if (sel->nbytes) {
                FREE(sel->ibits, M_TEMP);
                FREE(sel->obits, M_TEMP);
+               sel->nbytes = 0;
        }
 
-       if (sel->allocsize && sel->wqset){
-               kfree(sel->wqset, sel->allocsize);
+       if (uth->uu_cdir) {
+               vnode_rele(uth->uu_cdir);
+               uth->uu_cdir = NULLVP;
+       }
+
+       if (uth->uu_allocsize && uth->uu_wqset){
+               kfree(uth->uu_wqset, uth->uu_allocsize);
                sel->count = 0;
-               sel->allocsize = 0;
-               sel->wqset = 0;
+               uth->uu_allocsize = 0;
+               uth->uu_wqset = 0;
                sel->wql = 0;
        }
 
-       if (uth->uu_ucred != NOCRED)
-               kauth_cred_rele(uth->uu_ucred);
 
        if ((task != kernel_task) && p) {
-               funnel_state = thread_funnel_set(kernel_flock, TRUE);
-               //signal_lock(p);
-               TAILQ_REMOVE(&p->p_uthlist, uth, uu_list);
-               //signal_unlock(p);
-               (void)thread_funnel_set(kernel_flock, funnel_state);
+
+               if (((uth->uu_flag & UT_VFORK) == UT_VFORK) && (uth->uu_proc != PROC_NULL))  {
+                       vfork_exit_internal(uth->uu_proc, 0, 1);
+               }
+               if (get_bsdtask_info(task) == p) { 
+                       proc_lock(p);
+                       TAILQ_REMOVE(&p->p_uthlist, uth, uu_list);
+                       proc_unlock(p);
+               }
+#if CONFIG_DTRACE
+               if (uth->t_dtrace_scratch != NULL) {
+                       dtrace_ptss_release_entry(p, uth->t_dtrace_scratch);
+               }
+#endif
        }
+}
+
+/* This routine releases the credential stored in uthread */
+void
+uthread_cred_free(void *uthread)
+{
+       uthread_t uth = (uthread_t)uthread;
+
+       /* and free the uthread itself */
+       if (IS_VALID_CRED(uth->uu_ucred)) {
+               kauth_cred_t oldcred = uth->uu_ucred;
+               uth->uu_ucred = NOCRED;
+               kauth_cred_unref(&oldcred);
+       }
+}
+
+/* This routine frees the uthread structure held in thread structure */
+void
+uthread_zone_free(void *uthread)
+{
        /* and free the uthread itself */
        zfree(uthread_zone, uthread);
 }