]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/kern_exit.c
xnu-3247.10.11.tar.gz
[apple/xnu.git] / bsd / kern / kern_exit.c
index 6d04dd6a9d576e2bcb685a7c26df64fdefbc8c60..fca7ab3293c0d0ec34269589e654d2b0ef8d2827 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2011, 2015 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
@@ -94,6 +94,8 @@
 #include <sys/malloc.h>
 #include <sys/resourcevar.h>
 #include <sys/ptrace.h>
+#include <sys/proc_info.h>
+#include <sys/_types/_timeval64.h>
 #include <sys/user.h>
 #include <sys/aio_kern.h>
 #include <sys/sysproto.h>
 #include <kern/kern_types.h>
 #include <kern/kalloc.h>
 #include <kern/task.h>
+#include <corpses/task_corpse.h>
 #include <kern/thread.h>
 #include <kern/thread_call.h>
 #include <kern/sched_prim.h>
 #include <kern/assert.h>
 #include <sys/codesign.h>
 
+#if VM_PRESSURE_EVENTS
+#include <kern/vm_pressure.h>
+#endif
+
+#if CONFIG_MEMORYSTATUS
+#include <sys/kern_memorystatus.h>
+#endif
+
 #if CONFIG_DTRACE
 /* Do not include dtrace.h, it redefines kmem_[alloc/free] */
 extern void (*dtrace_fasttrap_exit_ptr)(proc_t);
@@ -137,34 +148,41 @@ extern void dtrace_lazy_dofs_destroy(proc_t);
 #include <mach/task.h>
 #include <mach/thread_act.h>
 
+#include <vm/vm_protos.h>
+
 #include <sys/sdt.h>
 
-extern char init_task_failure_data[];
-void proc_prepareexit(proc_t p, int rv);
+extern boolean_t init_task_died;
+void proc_prepareexit(proc_t p, int rv, boolean_t perf_notify);
 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);
+static int reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock);
+static void populate_corpse_crashinfo(proc_t p, void *crash_info_ptr, struct rusage_superset *rup, mach_exception_data_type_t code, mach_exception_data_type_t subcode);
+extern int proc_pidpathinfo(proc_t p, uint64_t arg, user_addr_t buffer, uint32_t buffersize, int32_t *retval);
+
+static __attribute__((noinline)) void launchd_crashed_panic(proc_t p, int rv);
+extern void proc_piduniqidentifierinfo(proc_t p, struct proc_uniqidentifierinfo *p_uniqidinfo);
+
 
 /*
  * Things which should have prototypes in headers, but don't
  */
-void   *get_bsduthreadarg(thread_t);
 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(thread_t thread, int pid);
-kern_return_t abnormal_exit_notify(mach_exception_data_type_t code, 
-               mach_exception_data_type_t subcode);
+kern_return_t task_exception_notify(exception_type_t exception,
+       mach_exception_data_type_t code, mach_exception_data_type_t subcode);
 void   delay(int);
-                       
+void gather_rusage_info(proc_t p, rusage_info_current *ru, int flavor);
+
 /*
  * NOTE: Source and target may *NOT* overlap!
  * XXX Should share code with bsd/dev/ppc/unix_signal.c
  */
-static void
+void
 siginfo_user_to_user32(user_siginfo_t *in, user32_siginfo_t *out)
 {
        out->si_signo   = in->si_signo;
@@ -179,7 +197,7 @@ siginfo_user_to_user32(user_siginfo_t *in, user32_siginfo_t *out)
        out->si_band    = in->si_band;                  /* range reduction */
 }
 
-static void
+void
 siginfo_user_to_user64(user_siginfo_t *in, user64_siginfo_t *out)
 {
        out->si_signo   = in->si_signo;
@@ -194,6 +212,188 @@ siginfo_user_to_user64(user_siginfo_t *in, user64_siginfo_t *out)
        out->si_band    = in->si_band;                  /* range reduction */
 }
 
+static int
+copyoutsiginfo(user_siginfo_t *native, boolean_t is64, user_addr_t uaddr)
+{
+       if (is64) {
+               user64_siginfo_t sinfo64;
+                               
+               bzero(&sinfo64, sizeof (sinfo64));
+               siginfo_user_to_user64(native, &sinfo64);
+               return (copyout(&sinfo64, uaddr, sizeof (sinfo64)));
+       } else {
+               user32_siginfo_t sinfo32;
+
+               bzero(&sinfo32, sizeof (sinfo32));
+               siginfo_user_to_user32(native, &sinfo32);
+               return (copyout(&sinfo32, uaddr, sizeof (sinfo32)));
+       }
+}
+
+static void populate_corpse_crashinfo(proc_t p, void *crash_info_ptr, struct rusage_superset *rup, mach_exception_data_type_t code, mach_exception_data_type_t subcode)
+{
+       mach_vm_address_t uaddr = 0;
+       mach_exception_data_type_t exc_codes[EXCEPTION_CODE_MAX];
+       exc_codes[0] = code;
+       exc_codes[1] = subcode;
+       cpu_type_t cputype;
+       struct proc_uniqidentifierinfo p_uniqidinfo;
+       struct proc_workqueueinfo pwqinfo;
+       int retval = 0;
+       uint64_t crashed_threadid = thread_tid(current_thread());
+       unsigned int pflags = 0;
+
+#if CONFIG_MEMORYSTATUS
+       int memstat_dirty_flags = 0;
+#endif
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_EXCEPTION_CODES, sizeof(exc_codes), &uaddr)) {
+               copyout(exc_codes, uaddr, sizeof(exc_codes));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PID, sizeof(p->p_pid), &uaddr)) {
+               copyout(&p->p_pid, uaddr, sizeof(p->p_pid));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PPID, sizeof(p->p_ppid), &uaddr)) {
+               copyout(&p->p_ppid, uaddr, sizeof(p->p_ppid));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_CRASHED_THREADID, sizeof(uint64_t), &uaddr)) {
+               copyout(&crashed_threadid, uaddr, sizeof(uint64_t));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_RUSAGE, sizeof(struct rusage), &uaddr)) {
+               copyout(&rup->ru, uaddr, sizeof(struct rusage));
+       }
+
+       if (KERN_SUCCESS ==
+           kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_BSDINFOWITHUNIQID, sizeof(struct proc_uniqidentifierinfo), &uaddr)) {
+               proc_piduniqidentifierinfo(p, &p_uniqidinfo);
+               copyout(&p_uniqidinfo, uaddr, sizeof(struct proc_uniqidentifierinfo));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_RUSAGE_INFO, sizeof(rusage_info_current), &uaddr)) {
+               copyout(&rup->ri, uaddr, sizeof(rusage_info_current));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_CSFLAGS, sizeof(p->p_csflags), &uaddr)) {
+               copyout(&p->p_csflags, uaddr, sizeof(p->p_csflags));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_NAME, sizeof(p->p_comm), &uaddr)) {
+               copyout(&p->p_comm, uaddr, sizeof(p->p_comm));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_STARTTIME, sizeof(p->p_start), &uaddr)) {
+               struct timeval64 t64;
+               t64.tv_sec = (int64_t)p->p_start.tv_sec;
+               t64.tv_usec = (int64_t)p->p_start.tv_usec;
+               copyout(&t64, uaddr, sizeof(t64));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_USERSTACK, sizeof(p->user_stack), &uaddr)) {
+               copyout(&p->user_stack, uaddr, sizeof(p->user_stack));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_ARGSLEN, sizeof(p->p_argslen), &uaddr)) {
+               copyout(&p->p_argslen, uaddr, sizeof(p->p_argslen));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_ARGC, sizeof(p->p_argc), &uaddr)) {
+               copyout(&p->p_argc, uaddr, sizeof(p->p_argc));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_PATH, MAXPATHLEN, &uaddr)) {
+               proc_pidpathinfo(p, 0, uaddr, MAXPATHLEN, &retval);
+       }
+
+       pflags = p->p_flag & (P_LP64 | P_SUGID);
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_FLAGS, sizeof(pflags), &uaddr)) {
+               copyout(&pflags, uaddr, sizeof(pflags));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_UID, sizeof(p->p_uid), &uaddr)) {
+               copyout(&p->p_uid, uaddr, sizeof(p->p_uid));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_GID, sizeof(p->p_gid), &uaddr)) {
+               copyout(&p->p_gid, uaddr, sizeof(p->p_gid));
+       }
+
+       cputype = cpu_type() & ~CPU_ARCH_MASK;
+       if (IS_64BIT_PROCESS(p))
+               cputype |= CPU_ARCH_ABI64;
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_CPUTYPE, sizeof(cpu_type_t), &uaddr)) {
+               copyout(&cputype, uaddr, sizeof(cpu_type_t));
+       }
+
+       bzero(&pwqinfo, sizeof(struct proc_workqueueinfo));
+       retval = fill_procworkqueue(p, &pwqinfo);
+       if (retval == 0) {
+               if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_WORKQUEUEINFO, sizeof(struct proc_workqueueinfo), &uaddr)) {
+                       copyout(&pwqinfo, uaddr, sizeof(struct proc_workqueueinfo));
+               }
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_RESPONSIBLE_PID, sizeof(p->p_responsible_pid), &uaddr)) {
+               copyout(&p->p_responsible_pid, uaddr, sizeof(p->p_responsible_pid));
+       }
+
+#if CONFIG_MEMORYSTATUS
+       memstat_dirty_flags = memorystatus_dirty_get(p);
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_DIRTY_FLAGS, sizeof(memstat_dirty_flags), &uaddr)) {
+               copyout(&memstat_dirty_flags, uaddr, sizeof(memstat_dirty_flags));
+       }
+#endif
+
+}
+
+static __attribute__((noinline)) void
+launchd_crashed_panic(proc_t p, int rv)
+{
+       printf("pid 1 exited (signal %d, exit %d)\n",
+           WTERMSIG(rv), WEXITSTATUS(rv));
+
+#if (DEVELOPMENT || DEBUG)
+       /*
+        * For debugging purposes, generate a core file of initproc before
+        * panicking. Leave at least 300 MB free on the root volume, and ignore
+        * the process's corefile ulimit. fsync() the file to ensure it lands on disk
+        * before the panic hits.
+        */
+
+       int             err;
+       uint64_t        coredump_start = mach_absolute_time();
+       uint64_t        coredump_end;
+       clock_sec_t     tv_sec;
+       clock_usec_t    tv_usec;
+       uint32_t        tv_msec;
+
+       err = coredump(p, 300, COREDUMP_IGNORE_ULIMIT | COREDUMP_FULLFSYNC);
+
+       coredump_end = mach_absolute_time();
+
+       absolutetime_to_microtime(coredump_end - coredump_start, &tv_sec, &tv_usec);
+
+       tv_msec = tv_usec / 1000;
+
+       if (err != 0) {
+               printf("Failed to generate initproc core file: error %d, took %d.%03d seconds\n",
+                      err, (uint32_t)tv_sec, tv_msec);
+       } else {
+               printf("Generated initproc core file in %d.%03d seconds\n",
+                      (uint32_t)tv_sec, tv_msec);
+       }
+#endif
+
+       sync(p, (void *)NULL, (int *)NULL);
+
+       panic_plain("%s exited (signal %d, exit status %d %s)", (p->p_name[0] != '\0' ? p->p_name : "initproc"), WTERMSIG(rv),
+                   WEXITSTATUS(rv), ((p->p_csflags & CS_KILLED) ? "CS_KILLED" : ""));
+}
+
 /*
  * exit --
  *     Death of process.
@@ -203,7 +403,6 @@ exit(proc_t p, struct exit_args *uap, int *retval)
 {
        exit1(p, W_EXITCODE(uap->rval, 0), retval);
 
-       /* drop funnel before we return */
        thread_exception_return();
        /* NOTREACHED */
        while (TRUE)
@@ -218,10 +417,18 @@ exit(proc_t p, struct exit_args *uap, int *retval)
  */
 int
 exit1(proc_t p, int rv, int *retval)
+{
+       return exit1_internal(p, rv, retval, TRUE, TRUE, 0);
+}
+
+int
+exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, boolean_t perf_notify,
+              int jetsam_flags)
 {
        thread_t self = current_thread();
        struct task *task = p->task;
        struct uthread *ut;
+       int error = 0;
 
        /*
         * If a thread in this task has already
@@ -231,10 +438,14 @@ exit1(proc_t p, int rv, int *retval)
 
         ut = get_bsdthread_info(self);
         if (ut->uu_flag & UT_VFORK) {
-                       vfork_exit(p, rv);
-                       vfork_return(p , retval, p->p_pid);
-                       unix_syscall_return(0);
-                       /* NOT REACHED */
+               if (!thread_can_terminate) {
+                       return EINVAL;
+               }
+
+               vfork_exit(p, rv);
+               vfork_return(p , retval, p->p_pid);
+               unix_syscall_return(0);
+               /* NOT REACHED */
         }
 
        /* 
@@ -254,8 +465,31 @@ exit1(proc_t p, int rv, int *retval)
 
        DTRACE_PROC1(exit, int, CLD_EXITED);
 
+       /* mark process is going to exit and pull out of DBG/disk throttle */
+       /* TODO: This should be done after becoming exit thread */
+       proc_set_task_policy(p->task, THREAD_NULL, TASK_POLICY_ATTRIBUTE,
+                            TASK_POLICY_TERMINATED, TASK_POLICY_ENABLE);
+
         proc_lock(p);
-       proc_transstart(p, 1);
+       error = proc_transstart(p, 1, ((jetsam_flags & P_JETSAM_VNODE) ? 1 : 0));
+       if (error == EDEADLK) {
+               /* Temp: If deadlock error, then it implies multithreaded exec is
+                * in progress. Instread of letting exit continue and 
+                * corrupting the freed memory, let the exit thread
+                * return. This will save corruption in remote case.
+                */
+               proc_unlock(p);
+               if (current_proc() == p){
+                       if (p->exit_thread == self)
+                               printf("exit_thread failed to exit, leaving process %s[%d] in unkillable limbo\n",
+                                      p->p_comm, p->p_pid);
+                       thread_exception_return();
+               } else {
+                       /* external termination like jetsam */
+                       return(error);
+               }
+       }
+
        while (p->exit_thread != self) {
                if (sig_try_locked(p) <= 0) {
                        proc_transend(p, 1);
@@ -264,30 +498,30 @@ exit1(proc_t p, int rv, int *retval)
                                return(0);
                         }
                        proc_unlock(p);
+                       
                        thread_terminate(self);
+                       if (!thread_can_terminate) {
+                               return 0;
+                       }
+
                        thread_exception_return();
                        /* NOTREACHED */
                }
                sig_lock_to_exit(p);
        }
-       if (p == initproc) {
-               proc_unlock(p);
-               printf("pid 1 exited (signal %d, exit %d)",
-                   WTERMSIG(rv), WEXITSTATUS(rv));
-               panic("%s died\nState at Last Exception:\n\n%s", 
-                                                       (p->p_comm[0] != '\0' ?
-                                                               p->p_comm :
-                                                               "launchd"),
-                                                       init_task_failure_data);
-       }
 
+       if (p == initproc && current_proc() == p) {
+               init_task_died = TRUE;
+       }
+       
        p->p_lflag |= P_LEXIT;
        p->p_xstat = rv;
+       p->p_lflag |= jetsam_flags;
 
        proc_transend(p, 1);
        proc_unlock(p);
 
-       proc_prepareexit(p, rv);
+       proc_prepareexit(p, rv, perf_notify);
 
        /* Last thread to terminate will call proc_exit() */
        task_terminate_internal(task);
@@ -296,12 +530,21 @@ exit1(proc_t p, int rv, int *retval)
 }
 
 void
-proc_prepareexit(proc_t p, int rv) 
+proc_prepareexit(proc_t p, int rv, boolean_t perf_notify
 {
-       mach_exception_data_type_t code, subcode;
+       mach_exception_data_type_t code = 0, subcode = 0;
+
        struct uthread *ut;
        thread_t self = current_thread();
        ut = get_bsdthread_info(self);
+       struct rusage_superset *rup;
+       int kr = 0;
+       int create_corpse = FALSE;
+
+       if (p == initproc) {
+               launchd_crashed_panic(p, rv);
+               /* NOTREACHED */
+       }
 
        /* If a core should be generated, notify crash reporter */
        if (hassigprop(WTERMSIG(rv), SA_CORE) || ((p->p_csflags & CS_KILLED) != 0)) {
@@ -323,13 +566,63 @@ proc_prepareexit(proc_t p, int rv)
                        ((ut->uu_exception & 0x0f) << 20) | 
                        ((int)ut->uu_code & 0xfffff);
                subcode = ut->uu_subcode;
-               (void) abnormal_exit_notify(code, subcode);
+
+               kr = task_exception_notify(EXC_CRASH, code, subcode);
+
+               /* Nobody handled EXC_CRASH?? remember to make corpse */
+               if (kr != 0) {
+                       create_corpse = TRUE;
+               }
        }
 
 skipcheck:
-       /* Notify the perf server */
-       (void)sys_perf_notify(self, p->p_pid);
+       /* Notify the perf server? */
+       if (perf_notify) {
+               (void)sys_perf_notify(self, p->p_pid);
+       }
+
+
+       /* stash the usage into corpse data if making_corpse == true */
+       if (create_corpse == TRUE) {
+               kr = task_mark_corpse(current_task());
+               if (kr != KERN_SUCCESS) {
+                       if (kr == KERN_NO_SPACE) {
+                               printf("Process[%d] has no vm space for corpse info.\n", p->p_pid);
+                       } else if (kr == KERN_NOT_SUPPORTED) {
+                               printf("Process[%d] was destined to be corpse. But corpse is disabled by config.\n", p->p_pid);
+                       } else {
+                               printf("Process[%d] crashed: %s. Too many corpses being created.\n", p->p_pid, p->p_comm);
+                       }
+                       create_corpse = FALSE;
+               } else {
+                       /* XXX: <rdar://problem/20491659> Need to sync ATM buffer before crash */
+                       kr = task_send_trace_memory(current_task(), p->p_pid, p->p_uniqueid);
+               }
+       }
+
+       /*
+        * Before this process becomes a zombie, stash resource usage
+        * stats in the proc for external observers to query
+        * via proc_pid_rusage().
+        *
+        * If the zombie allocation fails, just punt the stats.
+        */
+       MALLOC_ZONE(rup, struct rusage_superset *,
+                       sizeof (*rup), M_ZOMBIE, M_WAITOK);
+       if (rup != NULL) {
+               gather_rusage_info(p, &rup->ri, RUSAGE_INFO_CURRENT);
+               rup->ri.ri_phys_footprint = 0;
+               rup->ri.ri_proc_exit_abstime = mach_absolute_time();
 
+               /*
+                * Make the rusage_info visible to external observers
+                * only after it has been completely filled in.
+                */
+               p->p_ru = rup;
+       }
+       if (create_corpse) {
+               populate_corpse_crashinfo(p, task_get_corpseinfo(current_task()), rup, code, subcode);
+       }
        /*
         * Remove proc from allproc queue and from pidhash chain.
         * Need to do this before we do anything that can block.
@@ -339,6 +632,10 @@ skipcheck:
 
        proc_list_lock();
 
+#if CONFIG_MEMORYSTATUS
+       memorystatus_remove(p, TRUE);
+#endif
+
        LIST_REMOVE(p, p_list);
        LIST_INSERT_HEAD(&zombproc, p, p_list); /* Place onto zombproc. */
        /* will not be visible via proc_find */
@@ -373,11 +670,12 @@ proc_exit(proc_t p)
        struct uthread * uth;
        pid_t pid;
        int exitval;
+       int knote_hint;
 
-       uth = (struct uthread *)get_bsdthread_info(current_thread());
+       uth = current_uthread();
 
        proc_lock(p);
-       proc_transstart(p, 1);
+       proc_transstart(p, 1, 0);
        if( !(p->p_lflag & P_LEXIT)) {
                /*
                 * This can happen if a thread_terminate() occurs
@@ -386,7 +684,7 @@ proc_exit(proc_t p)
                p->p_lflag |= P_LEXIT;
                proc_transend(p, 1);
                proc_unlock(p);
-               proc_prepareexit(p, 0); 
+               proc_prepareexit(p, 0, TRUE);   
                (void) task_terminate_internal(task);
                proc_lock(p);
        } else {
@@ -410,8 +708,9 @@ proc_exit(proc_t p)
        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);
+       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON, 
+               BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXIT) | DBG_FUNC_START,
+               pid, exitval, 0, 0, 0);
 
 #if CONFIG_DTRACE
        /*
@@ -445,12 +744,12 @@ proc_exit(proc_t 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);
-
        nspace_proc_exit(p);
 
+#if VM_PRESSURE_EVENTS
+       vm_pressure_proc_cleanup(p);
+#endif
+
        /*
         * need to cancel async IO requests that can be cancelled and wait for those
         * already active.  MAY BLOCK!
@@ -458,8 +757,13 @@ proc_exit(proc_t p)
        
        proc_refdrain(p);
 
+       /* if any pending cpu limits action, clear it */
+       task_clear_cpuusage(p->task, TRUE);
+
        workqueue_mark_exiting(p);
        workqueue_exit(p);
+       kqueue_dealloc(p->p_wqkqueue);
+       p->p_wqkqueue = NULL;
 
        _aio_exit( p );
 
@@ -477,7 +781,7 @@ proc_exit(proc_t p)
                 * 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);
+               throttle_lowpri_io(0);
        }
 
 #if SYSV_SHM
@@ -500,9 +804,9 @@ proc_exit(proc_t p)
                if (sessp->s_ttyvp != NULLVP) {
                        struct vnode *ttyvp;
                        int ttyvid;
+                       int cttyflag = 0;
                        struct vfs_context context;
-                       struct tty * tp;
-
+                       struct tty *tp;
 
                        /*
                         * Controlling process.
@@ -510,55 +814,72 @@ proc_exit(proc_t p)
                         * drain controlling terminal
                         * and revoke access to controlling terminal.
                         */
+                       session_lock(sessp);
                        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)) {
+                               /*
+                                * We're going to SIGHUP the foreground process
+                                * group. It can't change from this point on
+                                * until the revoke is complete.
+                                * The process group changes under both the tty
+                                * lock and proc_list_lock but we need only one
+                                */
+                               tty_lock(tp);
+                               ttysetpgrphup(tp);
+                               tty_unlock(tp);
+
+                               tty_pgsignal(tp, SIGHUP, 1);
 
-                                       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);
                        }
-                       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)
-                        */
+                       cttyflag = sessp->s_flags & S_CTTYREF;
+                       sessp->s_flags &= ~S_CTTYREF;
+                       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);
+                               if (cttyflag) {
+                                       /*
+                                        * Release the extra usecount taken in cttyopen.
+                                        * usecount should be released after VNOP_REVOKE is called.
+                                        * This usecount was taken to ensure that
+                                        * the VNOP_REVOKE results in a close to
+                                        * the tty since cttyclose is a no-op.
+                                        */
+                                       vnode_rele(ttyvp);
+                               }
+                               vnode_put(ttyvp);
+                               kauth_cred_unref(&context.vc_ucred);
+                               ttyvp = NULLVP;
+                       }
+                       if (tp) {
+                               /*
+                                * This is cleared even if not set. This is also done in
+                                * spec_close to ensure that the flag is cleared.
+                                */
+                               tty_lock(tp);
+                               ttyclrpgrphup(tp);
+                               tty_unlock(tp);
+
+                               ttyfree(tp);
+                       }
                }
-               
                session_lock(sessp);
                sessp->s_leader = NULL;
                session_unlock(sessp);
@@ -584,7 +905,7 @@ proc_exit(proc_t p)
        /* 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;
+               int reparentedtoinit = (q->p_listflag & P_LIST_DEADPARENT) ? 1 : 0;
                if (q->p_stat == SZOMB) {
                        if (p != q->p_pptr)
                                panic("parent child linkage broken");
@@ -602,14 +923,15 @@ proc_exit(proc_t p)
                         * 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);
+                       (void)reap_child_locked(p, q, 1, reparentedtoinit, 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) {
+                               struct proc *opp;
+
                                /*
                                 * Take a reference on the child process to
                                 * ensure it doesn't exit and disappear between
@@ -620,8 +942,26 @@ proc_exit(proc_t p)
                                        continue;
 
                                proc_list_unlock();
+
+                               opp = proc_find(q->p_oppid);
+                               if (opp != PROC_NULL) {
+                                       proc_list_lock();
+                                       q->p_oppid = 0;
+                                       proc_list_unlock();
+                                       proc_reparentlocked(q, opp, 0, 0);
+                                       proc_rele(opp);
+                               } else {
+                                       /* original parent exited while traced */
+                                       proc_list_lock();
+                                       q->p_listflag |= P_LIST_DEADPARENT;
+                                       q->p_oppid = 0;
+                                       proc_list_unlock();
+                                       proc_reparentlocked(q, initproc, 0, 0);
+                               }
+
                                proc_lock(q);
                                q->p_lflag &= ~P_LTRACED;
+
                                if (q->sigwait_thread) {
                                        thread_t thread = q->sigwait_thread;
 
@@ -643,6 +983,9 @@ proc_exit(proc_t p)
                                psignal(q, SIGKILL);
                                proc_list_lock();
                                proc_rele_locked(q);
+                       } else {
+                               q->p_listflag |= P_LIST_DEADPARENT;
+                               proc_reparentlocked(q, initproc, 0, 1);
                        }
                }
        }
@@ -664,12 +1007,11 @@ proc_exit(proc_t p)
         * 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) {
            calcru(p, &p->p_stats->p_ru.ru_utime, &p->p_stats->p_ru.ru_stime, NULL);
-           *p->p_ru = p->p_stats->p_ru;
+           p->p_ru->ru = p->p_stats->p_ru;
 
-           ruadd(p->p_ru, &p->p_stats->p_cru);
+           ruadd(&(p->p_ru->ru), &p->p_stats->p_cru);
        }
 
        /*
@@ -716,6 +1058,7 @@ proc_exit(proc_t p)
        proc_limitdrop(p, 1);
        p->p_limit = NULL;
 
+       vm_purgeable_disown(p->task);
 
        /*
         * Finish up by terminating the task
@@ -725,8 +1068,8 @@ proc_exit(proc_t p)
        p->task = TASK_NULL;
        set_bsdtask_info(task, NULL);
 
-       /* exit status will be seen  by parent process */
-       proc_knote(p, NOTE_EXIT | (p->p_xstat & 0xffff));
+       knote_hint = NOTE_EXIT | (p->p_xstat & 0xffff);
+       proc_knote(p, knote_hint);
 
        /* mark the thread as the one that is doing proc_exit
         * no need to hold proc lock in uthread_free
@@ -738,6 +1081,8 @@ proc_exit(proc_t p)
        pp = proc_parent(p);
        if (pp->p_flag & P_NOCLDWAIT) {
 
+               if (p->p_ru != NULL) {
+                       proc_lock(pp);
 #if 3839178
                /*
                 * If the parent is ignoring SIGCHLD, then POSIX requires
@@ -752,19 +1097,18 @@ proc_exit(proc_t p)
                 * zombie to init.  If we were unable to allocate a
                 * zombie structure, this information is lost.
                 */
-               if (p->p_ru != NULL) {
-                       proc_lock(pp);
-                       ruadd(&pp->p_stats->p_cru, p->p_ru);
+                       ruadd(&pp->p_stats->p_cru, &p->p_ru->ru);
+#endif /* !3839178 */
+                       update_rusage_info_child(&pp->p_stats->ri_child, &p->p_ru->ri);
                        proc_unlock(pp);
                }
-#endif /* !3839178 */
-
+               
                /* kernel can reap this one, no need to move it to launchd */
                proc_list_lock();
                p->p_listflag |= P_LIST_DEADPARENT;
                proc_list_unlock();
        }
-       if ((p->p_listflag & P_LIST_DEADPARENT) == 0) {
+       if ((p->p_listflag & P_LIST_DEADPARENT) == 0 || p->p_oppid) {
                if (pp != initproc) {
                        proc_lock(pp);
                        pp->si_pid = p->p_pid;
@@ -783,8 +1127,9 @@ proc_exit(proc_t p)
                 * 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);
+               KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
+                       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
@@ -806,8 +1151,9 @@ proc_exit(proc_t p)
                 *  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);
+               KERNEL_DEBUG_CONSTANT_IST(KDEBUG_COMMON,
+                       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);
@@ -825,7 +1171,7 @@ proc_exit(proc_t p)
                 * 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);
+               (void)reap_child_locked(pp, p, 1, 0, 1, 1);
                /* list lock dropped by reap_child_locked */
        }
        if (uth->uu_lowpri_window) {
@@ -835,7 +1181,7 @@ proc_exit(proc_t p)
                 * 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);
+               throttle_lowpri_io(0);
        }
 
        proc_rele(pp);
@@ -860,7 +1206,7 @@ proc_exit(proc_t p)
  *             1                       Process was reaped
  */
 static int
-reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int droplock)
+reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoinit, int locked, int droplock)
 {
        proc_t trace_parent = PROC_NULL;        /* Traced parent process, if tracing */
 
@@ -875,44 +1221,69 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int d
         * 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
+        *      reparentedtoinit         -> came to launchd by reparenting
         */
-       if (child->p_oppid && (trace_parent = proc_find(child->p_oppid)) 
-                       && !((trace_parent == initproc) && (child->p_lflag & P_LIST_DEADPARENT))) {
+       if (child->p_oppid) {
+               int knote_hint;
+               pid_t oppid;
+
                proc_lock(child);
+               oppid = child->p_oppid;
                child->p_oppid = 0;
+               knote_hint = NOTE_EXIT | (child->p_xstat & 0xffff);
                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 = kauth_cred_getruid(child->p_ucred);
-                       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))
+
+               if ((trace_parent = proc_find(oppid))
+                       && !((trace_parent == initproc) && reparentedtoinit)) {
+                               
+                       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 = kauth_cred_getruid(child->p_ucred);
+                               proc_unlock(trace_parent);
+                       }
+                       proc_reparentlocked(child, trace_parent, 1, 0);
+                       
+                       /* resend knote to original parent (and others) after reparenting */
+                       proc_knote(child, knote_hint);
+                       
+                       psignal(trace_parent, SIGCHLD);
                        proc_list_lock();
-               return (0);
-       }
-       
-       if (trace_parent != PROC_NULL) {
-               proc_rele(trace_parent);
+                       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 we can't reparent (e.g. the original parent exited while child was being debugged, or
+                * original parent is the same as the debugger currently exiting), we still need to satisfy
+                * the knote lifecycle for other observers on the system. While the debugger was attached,
+                * the NOTE_EXIT would not have been broadcast during initial child termination.
+                */
+               proc_knote(child, knote_hint);
+
+               if (trace_parent != PROC_NULL) {
+                       proc_rele(trace_parent);
+               }
        }
        
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wdeprecated-declarations"
        proc_knote(child, NOTE_REAP);
+#pragma clang diagnostic pop
+
        proc_knote_drain(child);
 
        child->p_xstat = 0;
@@ -928,7 +1299,8 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int d
                 */
                if (!(parent->p_flag & P_NOCLDWAIT))
 #endif /* 3839178 */
-                       ruadd(&parent->p_stats->p_cru, child->p_ru);
+                       ruadd(&parent->p_stats->p_cru, &child->p_ru->ru);
+               update_rusage_info_child(&parent->p_stats->ri_child, &child->p_ru->ri);
                proc_unlock(parent);
                FREE_ZONE(child->p_ru, sizeof *child->p_ru, M_ZOMBIE);
                child->p_ru = NULL;
@@ -944,13 +1316,8 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int d
         * and refernce is dropped after these calls down below
         * (locking protection is provided by list lock held in chgproccnt)
         */
-       (void)chgproccnt(kauth_cred_getruid(child->p_ucred), -1);
 
-#if CONFIG_LCTX
-       ALLLCTX_LOCK;
-       leavelctx(child);
-       ALLLCTX_UNLOCK;
-#endif
+       (void)chgproccnt(kauth_cred_getruid(child->p_ucred), -1);
 
        /*
         * Free up credentials.
@@ -983,6 +1350,14 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int d
        proc_checkdeadrefs(child);
        nprocs--;
 
+       if (deadparent) {
+               /*
+                * If a child zombie is being reaped because its parent
+                * is exiting, make sure we update the list flag
+                */
+               child->p_listflag |= P_LIST_DEADPARENT;
+       }
+
        proc_list_unlock();
 
 #if CONFIG_FINE_LOCK_GROUPS
@@ -1013,19 +1388,24 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int locked, int d
 int
 wait1continue(int result)
 {
-       void *vt;
+       proc_t p;
        thread_t thread;
+       uthread_t uth;
+       struct _wait4_data *wait4_data;
+       struct wait4_nocancel_args *uap;
        int *retval;
-       proc_t p;
 
        if (result)
                return(result);
 
        p = current_proc();
        thread = current_thread();
-       vt = get_bsduthreadarg(thread);
-       retval = get_bsduthreadrval(thread);
-       return(wait4(p, (struct wait4_args *)vt, retval));
+       uth = (struct uthread *)get_bsdthread_info(thread);
+
+       wait4_data = &uth->uu_kevent.uu_wait4_data;
+       uap = wait4_data->args;
+       retval = wait4_data->retval;
+       return(wait4_nocancel(p, uap, retval));
 }
 
 int
@@ -1042,6 +1422,8 @@ wait4_nocancel(proc_t q, struct wait4_nocancel_args *uap, int32_t *retval)
        int sibling_count;
        proc_t p;
        int status, error;
+       uthread_t uth;
+       struct _wait4_data *wait4_data;
 
        AUDIT_ARG(pid, uap->pid);
 
@@ -1074,6 +1456,8 @@ loop1:
 
 
                if (p->p_stat == SZOMB) {
+                       int reparentedtoinit = (p->p_listflag & P_LIST_DEADPARENT) ? 1 : 0;
+
                        proc_list_unlock();
 #if CONFIG_MACF
                        if ((error = mac_proc_check_wait(q, p)) != 0)
@@ -1095,14 +1479,14 @@ loop1:
                                } else {
                                        if (IS_64BIT_PROCESS(q)) {
                                                struct user64_rusage    my_rusage;
-                                               munge_user64_rusage(p->p_ru, &my_rusage);
+                                               munge_user64_rusage(&p->p_ru->ru, &my_rusage);
                                                error = copyout((caddr_t)&my_rusage,
                                                        uap->rusage,
                                                        sizeof (my_rusage));
                                        }
                                        else {
                                                struct user32_rusage    my_rusage;
-                                               munge_user32_rusage(p->p_ru, &my_rusage);
+                                               munge_user32_rusage(&p->p_ru->ru, &my_rusage);
                                                error = copyout((caddr_t)&my_rusage,
                                                        uap->rusage,
                                                        sizeof (my_rusage));
@@ -1120,7 +1504,7 @@ loop1:
                         */
                        if ( sibling_count == 0 ) {
                                int mask = sigmask(SIGCHLD);
-                               uthread_t uth = (struct uthread *)get_bsdthread_info(current_thread());
+                               uth = current_uthread();
 
                                if ( (uth->uu_sigmask & mask) != 0 ) {
                                        /* we are blocking SIGCHLD signals.  clear any pending SIGCHLD.
@@ -1134,7 +1518,7 @@ loop1:
                        }
                        
                        /* Clean up */
-                       (void)reap_child_locked(q, p, 0, 0, 0);
+                       (void)reap_child_locked(q, p, 0, reparentedtoinit, 0, 0);
 
                        return (0);
                }
@@ -1197,6 +1581,12 @@ loop1:
                return (0);
        }
 
+       /* Save arguments for continuation. Backing storage is in uthread->uu_arg, and will not be deallocated */
+       uth = current_uthread();
+       wait4_data = &uth->uu_kevent.uu_wait4_data;
+       wait4_data->args = uap;
+       wait4_data->retval = retval;
+
        if ((error = msleep0((caddr_t)q, proc_list_mlock, PWAIT | PCATCH | PDROP, "wait", 0, wait1continue)))
                return (error);
 
@@ -1209,21 +1599,34 @@ out:
        return (error);
 }
 
+#if DEBUG
+#define ASSERT_LCK_MTX_OWNED(lock)     \
+                               lck_mtx_assert(lock, LCK_MTX_ASSERT_OWNED)
+#else
+#define ASSERT_LCK_MTX_OWNED(lock)     /* nothing */
+#endif
 
 int
 waitidcontinue(int result)
 {
-       void *vt;
+       proc_t p;
        thread_t thread;
+       uthread_t uth;
+       struct _waitid_data *waitid_data;
+       struct waitid_nocancel_args *uap;
        int *retval;
 
        if (result)
-               return(result);
+               return (result);
 
+       p = current_proc();
        thread = current_thread();
-       vt = get_bsduthreadarg(thread);
-       retval = get_bsduthreadrval(thread);
-       return(waitid(current_proc(), (struct waitid_args *)vt, retval));
+       uth = (struct uthread *)get_bsdthread_info(thread);
+
+       waitid_data = &uth->uu_kevent.uu_waitid_data;
+       uap = waitid_data->args;
+       retval = waitid_data->retval;
+       return(waitid_nocancel(p, uap, retval));
 }
 
 /*
@@ -1232,7 +1635,7 @@ waitidcontinue(int result)
  *
  * Parameters: uap->idtype             one of P_PID, P_PGID, P_ALL
  *             uap->id                 pid_t or gid_t or ignored
- *             uap->infop              Address of signinfo_t struct in
+ *             uap->infop              Address of siginfo_t struct in
  *                                     user space into which to return status
  *             uap->options            flag values
  *
@@ -1243,33 +1646,26 @@ int
 waitid(proc_t q, struct waitid_args *uap, int32_t *retval)
 {
        __pthread_testcancel(1);
-       return(waitid_nocancel(q, (struct waitid_nocancel_args *)uap, retval));
+       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)
+waitid_nocancel(proc_t q, struct waitid_nocancel_args *uap,
+       __unused int32_t *retval)
 {
-       user_siginfo_t  collect64;      /* siginfo data to return to caller */
-
+       user_siginfo_t  siginfo;        /* siginfo data to return to caller */
+       boolean_t caller64 = IS_64BIT_PROCESS(q);
        int nfound;
        proc_t p;
        int error;
+       uthread_t uth;
+       struct _waitid_data *waitid_data;
 
-       /*
-        * 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))
+       if (uap->options == 0 ||
+           (uap->options & ~(WNOHANG|WNOWAIT|WCONTINUED|WSTOPPED|WEXITED)))
                return (EINVAL);        /* bits set that aren't recognized */
 
-       /*
-        * Overly critical options checking, per POSIX
-        */
-       switch(uap->idtype) {
+       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)
@@ -1284,7 +1680,8 @@ loop:
 loop1:
        nfound = 0;
        for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
-               switch(uap->idtype) {
+
+               switch (uap->idtype) {
                case P_PID:     /* child with process ID equal to... */
                        if (p->p_pid != (pid_t)uap->id)
                                continue;
@@ -1304,68 +1701,44 @@ loop1:
                 * the single return for waited process guarantee.
                 */
                if (p->p_listflag & P_LIST_WAITING) {
-                       (void)msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitidcoll", 0);
+                       (void) msleep(&p->p_stat, proc_list_mlock,
+                               PWAIT, "waitidcoll", 0);
                        goto loop1;
                }
                p->p_listflag |= P_LIST_WAITING;                /* mark busy */
 
                nfound++;
 
-               /*
-                * Types of processes we are interested in
-                *
-                * XXX Don't know what to do for WCONTINUED?!?
-                */
-               switch(p->p_stat) {
+               bzero(&siginfo, sizeof (siginfo));
+
+               switch (p->p_stat) {
                case SZOMB:             /* Exited */
                        if (!(uap->options & WEXITED))
                                break;
-
-                       /* drop the lock and the thread is going to return */
                        proc_list_unlock();
+#if CONFIG_MACF
+                       if ((error = mac_proc_check_wait(q, p)) != 0)
+                               goto out;
+#endif
+                       siginfo.si_signo = SIGCHLD;
+                       siginfo.si_pid = p->p_pid;
+                       siginfo.si_status = WEXITSTATUS(p->p_xstat);
+                       if (WIFSIGNALED(p->p_xstat)) {
+                               siginfo.si_code = WCOREDUMP(p->p_xstat) ?
+                                       CLD_DUMPED : CLD_KILLED;
+                       } else
+                               siginfo.si_code = CLD_EXITED;
 
-                       /* Collect "siginfo" information for caller */
-                       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 = WEXITSTATUS(p->p_xstat);
-                       collect64.si_band = 0;
-
-                       if (IS_64BIT_PROCESS(p)) {
-                               user64_siginfo_t sinfo64;
-                               
-                               siginfo_user_to_user64(&collect64, &sinfo64);
-                               
-                               error = copyout((caddr_t)&sinfo64,
-                                       uap->infop,
-                                       sizeof(sinfo64));
-                       } else {
-                               user32_siginfo_t sinfo32;
-                               
-                               siginfo_user_to_user32(&collect64, &sinfo32);
-                               
-                               error = copyout((caddr_t)&sinfo32,
-                                                               uap->infop,
-                                                               sizeof(sinfo32));
-                       }
-                       /* information unavailable? */
-                       if (error) 
+                       if ((error = copyoutsiginfo(&siginfo,
+                           caller64, uap->infop)) != 0)
                                goto out;
 
                        /* Prevent other process for waiting for this event? */
                        if (!(uap->options & WNOWAIT)) {
-                               /* Clean up */
-                               (void)reap_child_locked(q, p, 0, 0, 0);
-                       } else {
-                               proc_list_lock();
-                               p->p_listflag &= ~P_LIST_WAITING;
-                               proc_list_unlock();
+                               (void) reap_child_locked(q, p, 0, 0, 0, 0);
+                               return (0);
                        }
-
-                       return (0);
+                       goto out;
 
                case SSTOP:             /* Stopped */
                        /*
@@ -1381,41 +1754,18 @@ loop1:
                         */
                        if ((p->p_lflag & P_LWAITED) != 0)
                                break;
-
-                       /* drop the lock and the thread is going to return */
                        proc_list_unlock();
+#if CONFIG_MACF
+                       if ((error = mac_proc_check_wait(q, p)) != 0)
+                               goto out;
+#endif
+                       siginfo.si_signo = SIGCHLD;
+                       siginfo.si_pid = p->p_pid;
+                       siginfo.si_status = p->p_xstat; /* signal number */
+                       siginfo.si_code = CLD_STOPPED;
 
-                       /* Collect "siginfo" information for caller */
-                       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)) {
-                               user64_siginfo_t sinfo64;
-                               
-                               siginfo_user_to_user64(&collect64, &sinfo64);
-                               
-                               error = copyout((caddr_t)&sinfo64,
-                                                               uap->infop,
-                                                               sizeof(sinfo64));
-                       } else {
-                               user32_siginfo_t sinfo32;
-                               
-                               siginfo_user_to_user32(&collect64, &sinfo32);
-                               
-                               error = copyout((caddr_t)&sinfo32,
-                                                               uap->infop,
-                                                               sizeof(sinfo32));
-                       }
-                       /* information unavailable? */
-                       if (error)
+                       if ((error = copyoutsiginfo(&siginfo,
+                           caller64, uap->infop)) != 0)
                                goto out;
 
                        /* Prevent other process for waiting for this event? */
@@ -1424,12 +1774,9 @@ loop1:
                                p->p_lflag |= P_LWAITED;
                                proc_unlock(p);
                        }
-
-                       error = 0;
                        goto out;
 
-               default:                /* All others */
-                                       /* ...meaning Continued */
+               default:                /* All other states => Continued */
                        if (!(uap->options & WCONTINUED))
                                break;
 
@@ -1440,60 +1787,40 @@ loop1:
                         */
                        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 */
+#if CONFIG_MACF
+                       if ((error = mac_proc_check_wait(q, p)) != 0)
+                               goto out;
+#endif
+                       siginfo.si_signo = SIGCHLD;
+                       siginfo.si_code = CLD_CONTINUED;
                        proc_lock(p);
-                       collect64.si_signo = SIGCHLD;
-                       collect64.si_code = CLD_CONTINUED;
-                       collect64.si_errno = 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;
+                       siginfo.si_pid = p->p_contproc;
+                       siginfo.si_status = p->p_xstat;
                        proc_unlock(p);
 
-                       if (IS_64BIT_PROCESS(p)) {
-                               user64_siginfo_t sinfo64;
-                               
-                               siginfo_user_to_user64(&collect64, &sinfo64);
-                               
-                               error = copyout((caddr_t)&sinfo64,
-                                                               uap->infop,
-                                                               sizeof(sinfo64));
-                       } else {
-                               user32_siginfo_t sinfo32;
-                               
-                               siginfo_user_to_user32(&collect64, &sinfo32);
-                               
-                               error = copyout((caddr_t)&sinfo32,
-                                                               uap->infop,
-                                                               sizeof(sinfo32));
-                       }
-                       /* information unavailable? */
-                       if (error)
+                       if ((error = copyoutsiginfo(&siginfo,
+                           caller64, uap->infop)) != 0)
                                goto out;
 
                        /* Prevent other process for waiting for this event? */
                        if (!(uap->options & WNOWAIT)) {
-                               OSBitAndAtomic(~((uint32_t)P_CONTINUED), &p->p_flag);
+                               OSBitAndAtomic(~((uint32_t)P_CONTINUED),
+                                   &p->p_flag);
                        }
-
-                       error = 0;
                        goto out;
                }
-               /* LIST LOCK IS HELD HERE */
+               ASSERT_LCK_MTX_OWNED(proc_list_mlock);
+
                /* Not a process we are interested in; go on to next child */
-               
+
                p->p_listflag &= ~P_LIST_WAITING;
                wakeup(&p->p_stat);
        }
+       ASSERT_LCK_MTX_OWNED(proc_list_mlock);
 
-       /* list lock is always held */
        /* No child processes that could possibly satisfy the request? */
+
        if (nfound == 0) {
                proc_list_unlock();
                return (ECHILD);
@@ -1501,10 +1828,30 @@ loop1:
 
        if (uap->options & WNOHANG) {
                proc_list_unlock();
+#if CONFIG_MACF
+               if ((error = mac_proc_check_wait(q, p)) != 0)
+                       return (error);
+#endif
+               /*
+                * The state of the siginfo structure in this case
+                * is undefined.  Some implementations bzero it, some
+                * (like here) leave it untouched for efficiency.
+                *
+                * Thus the most portable check for "no matching pid with
+                * WNOHANG" is to store a zero into si_pid before
+                * invocation, then check for a non-zero value afterwards.
+                */
                return (0);
        }
 
-       if ((error = msleep0((caddr_t)q, proc_list_mlock, PWAIT | PCATCH | PDROP, "waitid", 0, waitidcontinue)))
+       /* Save arguments for continuation. Backing storage is in uthread->uu_arg, and will not be deallocated */
+       uth = current_uthread();
+       waitid_data = &uth->uu_kevent.uu_waitid_data;
+       waitid_data->args = uap;
+       waitid_data->retval = retval;
+
+       if ((error = msleep0(q, proc_list_mlock,
+           PWAIT | PCATCH | PDROP, "waitid", 0, waitidcontinue)) != 0)
                return (error);
 
        goto loop;
@@ -1622,6 +1969,10 @@ vfork_exit_internal(proc_t p, int rv, int forceexit)
 
        proc_list_lock();
 
+#if CONFIG_MEMORYSTATUS
+       memorystatus_remove(p, TRUE);
+#endif
+
        LIST_REMOVE(p, p_list);
        LIST_INSERT_HEAD(&zombproc, p, p_list); /* Place onto zombproc. */
        /* will not be visible via proc_find */
@@ -1668,11 +2019,11 @@ vproc_exit(proc_t p)
 #endif
        struct pgrp * pg;
        struct session *sessp;
+       struct rusage_superset *rup;
 
        /* 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);
-
+       MALLOC_ZONE(rup, struct rusage_superset *,
+                       sizeof (*rup), M_ZOMBIE, M_WAITOK);
 
        proc_refdrain(p);
 
@@ -1688,8 +2039,9 @@ vproc_exit(proc_t p)
                if (sessp->s_ttyvp != NULLVP) {
                        struct vnode *ttyvp;
                        int ttyvid;
+                       int cttyflag = 0;
                        struct vfs_context context;
-                       struct tty * tp;
+                       struct tty *tp;
 
                        /*
                         * Controlling process.
@@ -1697,54 +2049,72 @@ vproc_exit(proc_t p)
                         * drain controlling terminal
                         * and revoke access to controlling terminal.
                         */
+                       session_lock(sessp);
                        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);
+                               session_unlock(sessp);
+
                                /*
-                                * The tty could have been revoked
-                                * if we blocked.
+                                * We're going to SIGHUP the foreground process
+                                * group. It can't change from this point on
+                                * until the revoke is complete.
+                                * The process group changes under both the tty
+                                * lock and proc_list_lock but we need only one
                                 */
+                               tty_lock(tp);
+                               ttysetpgrphup(tp);
+                               tty_unlock(tp);
+
+                               tty_pgsignal(tp, SIGHUP, 1);
 
                                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);
+                       }
+                       cttyflag = sessp->s_flags & S_CTTYREF;
+                       sessp->s_flags &= ~S_CTTYREF;
+                       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)) {
+                               if (tp != TTY_NULL) {
+                                       tty_lock(tp);
+                                       (void) ttywait(tp);
+                                       tty_unlock(tp);
                                }
-                       } 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);
+                               context.vc_thread = proc_thread(p); /* XXX */
+                               context.vc_ucred = kauth_cred_proc_ref(p);
+                               VNOP_REVOKE(ttyvp, REVOKEALL, &context);
+                               if (cttyflag) {
+                                       /*
+                                        * Release the extra usecount taken in cttyopen.
+                                        * usecount should be released after VNOP_REVOKE is called.
+                                        * This usecount was taken to ensure that
+                                        * the VNOP_REVOKE results in a close to
+                                        * the tty since cttyclose is a no-op.
+                                        */
+                                       vnode_rele(ttyvp);
+                               }
+                               vnode_put(ttyvp);
+                               kauth_cred_unref(&context.vc_ucred);
+                               ttyvp = NULLVP;
                        }
-                       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)
-                        */
-               }
+                       if (tp) {
+                               /*
+                                * This is cleared even if not set. This is also done in
+                                * spec_close to ensure that the flag is cleared.
+                                */
+                               tty_lock(tp);
+                               ttyclrpgrphup(tp);
+                               tty_unlock(tp);
 
+                               ttyfree(tp);
+                       }
+               }
                session_lock(sessp);
                sessp->s_leader = NULL;
                session_unlock(sessp);
@@ -1760,7 +2130,6 @@ vproc_exit(proc_t p)
        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");
@@ -1778,17 +2147,36 @@ vproc_exit(proc_t p)
                         * 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);
+                       (void)reap_child_locked(p, q, 1, 0, 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) {
+                               struct proc *opp;
+
                                proc_list_unlock();
+
+                               opp = proc_find(q->p_oppid);
+                               if (opp != PROC_NULL) {
+                                       proc_list_lock();
+                                       q->p_oppid = 0;
+                                       proc_list_unlock();
+                                       proc_reparentlocked(q, opp, 0, 0);
+                                       proc_rele(opp);
+                               } else {
+                                       /* original parent exited while traced */
+                                       proc_list_lock();
+                                       q->p_listflag |= P_LIST_DEADPARENT;
+                                       q->p_oppid = 0;
+                                       proc_list_unlock();
+                                       proc_reparentlocked(q, initproc, 0, 0);
+                               }
+
                                proc_lock(q);
                                q->p_lflag &= ~P_LTRACED;
+
                                if (q->sigwait_thread) {
                                        thread_t thread = q->sigwait_thread;
 
@@ -1809,6 +2197,9 @@ vproc_exit(proc_t p)
 
                                psignal(q, SIGKILL);
                                proc_list_lock();
+                       } else {
+                               q->p_listflag |= P_LIST_DEADPARENT;
+                               proc_reparentlocked(q, initproc, 0, 1);
                        }
                }
        }
@@ -1830,26 +2221,25 @@ vproc_exit(proc_t p)
         * 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 (rup != NULL) {
+           rup->ru = p->p_stats->p_ru;
+           timerclear(&rup->ru.ru_utime);
+           timerclear(&rup->ru.ru_stime);
 
 #ifdef  FIXME
            if (task) {
-               task_basic_info_data_t tinfo;
+               mach_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_stuff = MACH_TASK_BASIC_INFO_COUNT;
+               task_info(task, MACH_TASK_BASIC_INFO,
                          &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;
+               p->p_ru->ru.ru_utime.tv_sec = tinfo.user_time.seconds;
+               p->p_ru->ru.ru_utime.tv_usec = tinfo.user_time.microseconds;
+               p->p_ru->ru.ru_stime.tv_sec = tinfo.system_time.seconds;
+               p->p_ru->ru.ru_stime.tv_usec = tinfo.system_time.microseconds;
 
                task_ttimes_stuff = TASK_THREAD_TIMES_INFO_COUNT;
                task_info(task, TASK_THREAD_TIMES_INFO,
@@ -1859,12 +2249,22 @@ vproc_exit(proc_t p)
                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);
+               timeradd(&ut,&p->p_ru->ru.ru_utime,&p->p_ru->ru.ru_utime);
+                       timeradd(&st,&p->p_ru->ru.ru_stime,&p->p_ru->ru.ru_stime);
            }
 #endif /* FIXME */
 
-           ruadd(p->p_ru, &p->p_stats->p_cru);
+           ruadd(&rup->ru, &p->p_stats->p_cru);
+
+               gather_rusage_info(p, &rup->ri, RUSAGE_INFO_CURRENT);
+               rup->ri.ri_phys_footprint = 0;
+               rup->ri.ri_proc_exit_abstime = mach_absolute_time();
+
+               /*
+                * Now that we have filled in the rusage info, make it
+                * visible to an external observer via proc_pid_rusage().
+                */
+               p->p_ru = rup;
        }
 
        /*
@@ -1956,7 +2356,7 @@ vproc_exit(proc_t p)
                 * 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);
+               (void)reap_child_locked(pp, p, 0, 0, 1, 1);
                /* list lock dropped by reap_child_locked */
        }
        proc_rele(pp);