]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/kern_exit.c
xnu-4570.1.46.tar.gz
[apple/xnu.git] / bsd / kern / kern_exit.c
index 3d17f687c108126913aff2c2d25c116e0d81a69e..b8f5def79fd308f346eed191d51d1dbae0a4759e 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2011 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2016 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
@@ -74,6 +74,7 @@
  
 #include <machine/reg.h>
 #include <machine/psl.h>
+#include <stdatomic.h>
 
 #include "compat_43.h"
 
 #include <sys/malloc.h>
 #include <sys/resourcevar.h>
 #include <sys/ptrace.h>
+#include <sys/proc_info.h>
+#include <sys/reason.h>
+#include <sys/_types/_timeval64.h>
 #include <sys/user.h>
 #include <sys/aio_kern.h>
 #include <sys/sysproto.h>
 #include <sys/signalvar.h>
 #include <sys/kdebug.h>
-#include <sys/filedesc.h>      /* fdfree */
-#if SYSV_SHM
-#include <sys/shm_internal.h>  /* shmexit */
-#endif
-#include <sys/acct.h>          /* acct_process */
+#include <sys/filedesc.h> /* fdfree */
+#include <sys/acct.h> /* acct_process */
+#include <sys/codesign.h>
+#include <sys/event.h> /* kevent_proc_copy_uptrs */
+#include <sys/sdt.h>
 
 #include <security/audit/audit.h>
 #include <bsm/audit_kevents.h>
 
 #include <mach/mach_types.h>
+#include <mach/task.h>
+#include <mach/thread_act.h>
 
+#include <kern/exc_resource.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>
+#include <kern/policy_internal.h>
+#include <kern/exc_guard.h>
 
-#if VM_PRESSURE_EVENTS
-#include <kern/vm_pressure.h>
-#endif
+#include <vm/vm_protos.h>
 
+#include <pexpert/pexpert.h>
+
+#if SYSV_SHM
+#include <sys/shm_internal.h>  /* shmexit */
+#endif /* SYSV_SHM */
+#if CONFIG_PERSONAS
+#include <sys/persona.h>
+#endif /* CONFIG_PERSONAS */
 #if CONFIG_MEMORYSTATUS
 #include <sys/kern_memorystatus.h>
-#endif
-
+#endif /* CONFIG_MEMORYSTATUS */
 #if CONFIG_DTRACE
 /* Do not include dtrace.h, it redefines kmem_[alloc/free] */
-extern void (*dtrace_fasttrap_exit_ptr)(proc_t);
-extern void (*dtrace_helpers_cleanup)(proc_t);
-extern void dtrace_lazy_dofs_destroy(proc_t);
-
+void dtrace_proc_exit(proc_t p);
 #include <sys/dtrace_ptss.h>
-#endif
-
+#endif /* CONFIG_DTRACE */
 #if CONFIG_MACF
-#include <security/mac.h>
+#include <security/mac_framework.h>
+#include <security/mac_mach_internal.h>
 #include <sys/syscall.h>
-#endif
+#endif /* CONFIG_MACF */
 
-#include <mach/mach_types.h>
-#include <mach/task.h>
-#include <mach/thread_act.h>
-
-#include <vm/vm_protos.h>
-
-#include <sys/sdt.h>
-
-extern boolean_t init_task_died;
-extern char init_task_failure_data[];
 void proc_prepareexit(proc_t p, int rv, boolean_t perf_notify);
+void gather_populate_corpse_crashinfo(proc_t p, task_t corpse_task,
+               mach_exception_data_type_t code, mach_exception_data_type_t subcode,
+               uint64_t *udata_buffer, int num_udata, void *reason);
+mach_exception_data_type_t proc_encode_exit_exception_code(proc_t p);
 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 reparentedtoinit, int locked, int droplock);
+static void populate_corpse_crashinfo(proc_t p, task_t corpse_task,
+               struct rusage_superset *rup, mach_exception_data_type_t code,
+               mach_exception_data_type_t subcode, uint64_t *udata_buffer,
+               int num_udata, os_reason_t reason);
+static void proc_update_corpse_exception_codes(proc_t p, mach_exception_data_type_t *code, mach_exception_data_type_t *subcode);
+extern int proc_pidpathinfo_internal(proc_t p, uint64_t arg, char *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);
+extern void task_coalition_ids(task_t task, uint64_t ids[COALITION_NUM_TYPES]);
+extern uint64_t        get_task_phys_footprint_limit(task_t);
+int proc_list_uptrs(void *p, uint64_t *udata_buffer, int size);
+extern uint64_t task_corpse_get_crashed_thread_id(task_t corpse_task);
+
 
 /*
  * Things which should have prototypes in headers, but don't
@@ -167,6 +184,7 @@ int waitidcontinue(int result);
 kern_return_t sys_perf_notify(thread_t thread, int pid);
 kern_return_t task_exception_notify(exception_type_t exception,
        mach_exception_data_type_t code, mach_exception_data_type_t subcode);
+kern_return_t task_violated_guard(mach_exception_code_t, mach_exception_subcode_t, void *);
 void   delay(int);
 void gather_rusage_info(proc_t p, rusage_info_current *ru, int flavor);
 
@@ -209,7 +227,7 @@ 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)));
@@ -222,13 +240,445 @@ copyoutsiginfo(user_siginfo_t *native, boolean_t is64, user_addr_t uaddr)
        }
 }
 
+void gather_populate_corpse_crashinfo(proc_t p, task_t corpse_task,
+               mach_exception_data_type_t code, mach_exception_data_type_t subcode,
+               uint64_t *udata_buffer, int num_udata, void *reason)
+{
+       struct rusage_superset rup;
+
+       gather_rusage_info(p, &rup.ri, RUSAGE_INFO_CURRENT);
+       rup.ri.ri_phys_footprint = 0;
+       populate_corpse_crashinfo(p, corpse_task, &rup, code, subcode,
+                       udata_buffer, num_udata, reason);
+}
+
+static void proc_update_corpse_exception_codes(proc_t p, mach_exception_data_type_t *code, mach_exception_data_type_t *subcode)
+{
+       mach_exception_data_type_t code_update = *code;
+       mach_exception_data_type_t subcode_update = *subcode;
+       if (p->p_exit_reason == OS_REASON_NULL) {
+               return;
+       }
+
+       switch (p->p_exit_reason->osr_namespace) {
+               case OS_REASON_JETSAM:
+                       if (p->p_exit_reason->osr_code == JETSAM_REASON_MEMORY_PERPROCESSLIMIT) {
+                               /* Update the code with EXC_RESOURCE code for high memory watermark */
+                               EXC_RESOURCE_ENCODE_TYPE(code_update, RESOURCE_TYPE_MEMORY);
+                               EXC_RESOURCE_ENCODE_FLAVOR(code_update, FLAVOR_HIGH_WATERMARK);
+                               EXC_RESOURCE_HWM_ENCODE_LIMIT(code_update, ((get_task_phys_footprint_limit(p->task)) >> 20));
+                               subcode_update = 0;
+                               break;
+                       }
+
+                       break;
+               default:
+                       break;
+       }
+
+       *code = code_update;
+       *subcode = subcode_update;
+       return;
+}
+
+mach_exception_data_type_t proc_encode_exit_exception_code(proc_t p)
+{
+       uint64_t subcode = 0;
+
+       if (p->p_exit_reason == OS_REASON_NULL) {
+               return 0;
+       }
+
+       /* Embed first 32 bits of osr_namespace and osr_code in exception code */
+       ENCODE_OSR_NAMESPACE_TO_MACH_EXCEPTION_CODE(subcode, p->p_exit_reason->osr_namespace);
+       ENCODE_OSR_CODE_TO_MACH_EXCEPTION_CODE(subcode, p->p_exit_reason->osr_code);
+       return (mach_exception_data_type_t)subcode;
+}
+
+static void
+populate_corpse_crashinfo(proc_t p, task_t corpse_task, struct rusage_superset *rup,
+               mach_exception_data_type_t code, mach_exception_data_type_t subcode,
+               uint64_t *udata_buffer, int num_udata, os_reason_t reason)
+{
+       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 = task_corpse_get_crashed_thread_id(corpse_task);
+       unsigned int pflags = 0;
+       uint64_t max_footprint_mb;
+       uint64_t max_footprint;
+       void *crash_info_ptr = task_get_corpseinfo(corpse_task);
+
+#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)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, exc_codes, sizeof(exc_codes));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PID, sizeof(p->p_pid), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_pid, sizeof(p->p_pid));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PPID, sizeof(p->p_ppid), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_ppid, sizeof(p->p_ppid));
+       }
+
+       /* Don't include the crashed thread ID if there's an exit reason that indicates it's irrelevant */
+       if ((p->p_exit_reason == OS_REASON_NULL) || !(p->p_exit_reason->osr_flags & OS_REASON_FLAG_NO_CRASHED_TID)) {
+               if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_CRASHED_THREADID, sizeof(uint64_t), &uaddr)) {
+                       kcdata_memcpy(crash_info_ptr, uaddr, &crashed_threadid, sizeof(uint64_t));
+               }
+       }
+
+       if (KERN_SUCCESS ==
+           kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_BSDINFOWITHUNIQID, sizeof(struct proc_uniqidentifierinfo), &uaddr)) {
+               proc_piduniqidentifierinfo(p, &p_uniqidinfo);
+               kcdata_memcpy(crash_info_ptr, uaddr, &p_uniqidinfo, sizeof(struct proc_uniqidentifierinfo));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_RUSAGE_INFO, sizeof(rusage_info_current), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &rup->ri, sizeof(rusage_info_current));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_CSFLAGS, sizeof(p->p_csflags), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_csflags, sizeof(p->p_csflags));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_NAME, sizeof(p->p_comm), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_comm, 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;
+               kcdata_memcpy(crash_info_ptr, uaddr, &t64, sizeof(t64));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_USERSTACK, sizeof(p->user_stack), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->user_stack, sizeof(p->user_stack));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_ARGSLEN, sizeof(p->p_argslen), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_argslen, sizeof(p->p_argslen));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_ARGC, sizeof(p->p_argc), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_argc, sizeof(p->p_argc));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_PROC_PATH, MAXPATHLEN, &uaddr)) {
+               char *buf = (char *) kalloc(MAXPATHLEN);
+               if (buf != NULL) {
+                       bzero(buf, MAXPATHLEN);
+                       proc_pidpathinfo_internal(p, 0, buf, MAXPATHLEN, &retval);
+                       kcdata_memcpy(crash_info_ptr, uaddr, buf, MAXPATHLEN);
+                       kfree(buf, MAXPATHLEN);
+               }
+       }
+
+       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)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &pflags, sizeof(pflags));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_UID, sizeof(p->p_uid), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_uid, sizeof(p->p_uid));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_GID, sizeof(p->p_gid), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_gid, 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)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &cputype, sizeof(cpu_type_t));
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_MEMORY_LIMIT, sizeof(max_footprint_mb), &uaddr)) {
+               max_footprint = get_task_phys_footprint_limit(p->task);
+               max_footprint_mb = max_footprint >> 20;
+               kcdata_memcpy(crash_info_ptr, uaddr, &max_footprint_mb, sizeof(max_footprint_mb));
+       }
+
+       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)) {
+                       kcdata_memcpy(crash_info_ptr, uaddr, &pwqinfo, sizeof(struct proc_workqueueinfo));
+               }
+       }
+
+       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, TASK_CRASHINFO_RESPONSIBLE_PID, sizeof(p->p_responsible_pid), &uaddr)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &p->p_responsible_pid, sizeof(p->p_responsible_pid));
+       }
+
+#if CONFIG_COALITIONS
+       if (KERN_SUCCESS == kcdata_get_memory_addr_for_array(crash_info_ptr, TASK_CRASHINFO_COALITION_ID, sizeof(uint64_t), COALITION_NUM_TYPES, &uaddr)) {
+               uint64_t coalition_ids[COALITION_NUM_TYPES];
+               task_coalition_ids(p->task, coalition_ids);
+               kcdata_memcpy(crash_info_ptr, uaddr, coalition_ids, sizeof(coalition_ids));
+       }
+#endif /* CONFIG_COALITIONS */
+
+#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)) {
+               kcdata_memcpy(crash_info_ptr, uaddr, &memstat_dirty_flags, sizeof(memstat_dirty_flags));
+       }
+#endif
+
+       if (p->p_exit_reason != OS_REASON_NULL && reason == OS_REASON_NULL) {
+               reason = p->p_exit_reason;
+       }
+       if (reason != OS_REASON_NULL) {
+               if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, EXIT_REASON_SNAPSHOT, sizeof(struct exit_reason_snapshot), &uaddr)) {
+                       struct exit_reason_snapshot ers = {
+                               .ers_namespace = reason->osr_namespace,
+                               .ers_code = reason->osr_code,
+                               .ers_flags = reason->osr_flags
+                       };
+
+                       kcdata_memcpy(crash_info_ptr, uaddr, &ers, sizeof(ers));
+               }
+
+               if (reason->osr_kcd_buf != 0) {
+                       uint32_t reason_buf_size = kcdata_memory_get_used_bytes(&reason->osr_kcd_descriptor);
+                       assert(reason_buf_size != 0);
+
+                       if (KERN_SUCCESS == kcdata_get_memory_addr(crash_info_ptr, KCDATA_TYPE_NESTED_KCDATA, reason_buf_size, &uaddr)) {
+                               kcdata_memcpy(crash_info_ptr, uaddr, reason->osr_kcd_buf, reason_buf_size);
+                       }
+               }
+       }
+
+       if (num_udata > 0) {
+               if (KERN_SUCCESS == kcdata_get_memory_addr_for_array(crash_info_ptr, TASK_CRASHINFO_UDATA_PTRS,
+                                       sizeof(uint64_t), num_udata, &uaddr)) {
+                       kcdata_memcpy(crash_info_ptr, uaddr, udata_buffer, sizeof(uint64_t) * num_udata);
+               }
+       }
+}
+
+/*
+ * We only parse exit reason kcdata blobs for launchd when it dies
+ * and we're going to panic.
+ *
+ * Meant to be called immediately before panicking.
+ */
+char *
+launchd_exit_reason_get_string_desc(os_reason_t exit_reason)
+{
+       kcdata_iter_t iter;
+
+       if (exit_reason == OS_REASON_NULL || exit_reason->osr_kcd_buf == NULL ||
+                       exit_reason->osr_bufsize == 0) {
+               return NULL;
+       }
+
+       iter = kcdata_iter(exit_reason->osr_kcd_buf, exit_reason->osr_bufsize);
+       if (!kcdata_iter_valid(iter)) {
+#if DEBUG || DEVELOPMENT
+               printf("launchd exit reason has invalid exit reason buffer\n");
+#endif
+               return NULL;
+       }
+
+       if (kcdata_iter_type(iter) != KCDATA_BUFFER_BEGIN_OS_REASON) {
+#if DEBUG || DEVELOPMENT
+               printf("launchd exit reason buffer type mismatch, expected %d got %d\n",
+                       KCDATA_BUFFER_BEGIN_OS_REASON, kcdata_iter_type(iter));
+#endif
+               return NULL;
+       }
+
+       iter = kcdata_iter_find_type(iter, EXIT_REASON_USER_DESC);
+       if (!kcdata_iter_valid(iter)) {
+               return NULL;
+       }
+
+       return (char *)kcdata_iter_payload(iter);
+}
+
+static __attribute__((noinline)) void
+launchd_crashed_panic(proc_t p, int rv)
+{
+       char *launchd_exit_reason_desc = launchd_exit_reason_get_string_desc(p->p_exit_reason);
+
+       if (p->p_exit_reason == OS_REASON_NULL) {
+               printf("pid 1 exited -- no exit reason available -- (signal %d, exit %d)\n",
+                       WTERMSIG(rv), WEXITSTATUS(rv));
+       } else {
+               printf("pid 1 exited -- exit reason namespace %d subcode 0x%llx, description %s\n",
+                       p->p_exit_reason->osr_namespace, p->p_exit_reason->osr_code, launchd_exit_reason_desc ?
+                       launchd_exit_reason_desc : "none");
+       }
+
+       const char *launchd_crashed_prefix_str;
+
+       if (strnstr(p->p_name, "preinit", sizeof(p->p_name))) {
+               launchd_crashed_prefix_str = "LTE preinit process exited";
+       } else {
+               launchd_crashed_prefix_str = "initproc exited";
+       }
+
+#if (DEVELOPMENT || DEBUG) && CONFIG_COREDUMP
+       /*
+        * 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 /* (DEVELOPMENT || DEBUG) && CONFIG_COREDUMP */
+
+       sync(p, (void *)NULL, (int *)NULL);
+
+       if (p->p_exit_reason == OS_REASON_NULL) {
+               panic_with_options(0, NULL, DEBUGGER_OPTION_INITPROC_PANIC, "%s -- no exit reason available -- (signal %d, exit status %d %s)",
+                               launchd_crashed_prefix_str, WTERMSIG(rv), WEXITSTATUS(rv), ((p->p_csflags & CS_KILLED) ? "CS_KILLED" : ""));
+       } else {
+               panic_with_options(0, NULL, DEBUGGER_OPTION_INITPROC_PANIC, "%s %s -- exit reason namespace %d subcode 0x%llx description: %." LAUNCHD_PANIC_REASON_STRING_MAXLEN "s",
+                               ((p->p_csflags & CS_KILLED) ? "CS_KILLED" : ""),
+                               launchd_crashed_prefix_str, p->p_exit_reason->osr_namespace, p->p_exit_reason->osr_code,
+                               launchd_exit_reason_desc ? launchd_exit_reason_desc : "none");
+       }
+}
+
+#define OS_REASON_IFLAG_USER_FAULT 0x1
+
+#define OS_REASON_TOTAL_USER_FAULTS_PER_PROC  5
+
+static int
+abort_with_payload_internal(proc_t p,
+               uint32_t reason_namespace, uint64_t reason_code,
+               user_addr_t payload, uint32_t payload_size,
+               user_addr_t reason_string, uint64_t reason_flags,
+               uint32_t internal_flags)
+{
+       os_reason_t exit_reason = OS_REASON_NULL;
+       kern_return_t kr = KERN_SUCCESS;
+
+       if (internal_flags & OS_REASON_IFLAG_USER_FAULT) {
+               uint32_t old_value = atomic_load_explicit(&p->p_user_faults,
+                               memory_order_relaxed);
+               for (;;) {
+                       if (old_value >= OS_REASON_TOTAL_USER_FAULTS_PER_PROC) {
+                               return EQFULL;
+                       }
+                       // this reloads the value in old_value
+                       if (atomic_compare_exchange_strong_explicit(&p->p_user_faults,
+                                       &old_value, old_value + 1, memory_order_relaxed,
+                                       memory_order_relaxed)) {
+                               break;
+                       }
+               }
+       }
+
+       KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE,
+                                       p->p_pid, reason_namespace,
+                                       reason_code, 0, 0);
+
+       exit_reason = build_userspace_exit_reason(reason_namespace, reason_code,
+                       payload, payload_size, reason_string, reason_flags);
+
+       if (internal_flags & OS_REASON_IFLAG_USER_FAULT) {
+               mach_exception_code_t code = 0;
+
+               EXC_GUARD_ENCODE_TYPE(code, GUARD_TYPE_USER); /* simulated EXC_GUARD */
+               EXC_GUARD_ENCODE_FLAVOR(code, 0);
+               EXC_GUARD_ENCODE_TARGET(code, reason_namespace);
+
+               if (exit_reason == OS_REASON_NULL) {
+                       kr = KERN_RESOURCE_SHORTAGE;
+               } else {
+                       kr = task_violated_guard(code, reason_code, exit_reason);
+               }
+               os_reason_free(exit_reason);
+       } else {
+               /*
+                * We use SIGABRT (rather than calling exit directly from here) so that
+                * the debugger can catch abort_with_{reason,payload} calls.
+                */
+               psignal_try_thread_with_reason(p, current_thread(), SIGABRT, exit_reason);
+       }
+
+       switch (kr) {
+       case KERN_SUCCESS:
+               return 0;
+       case KERN_NOT_SUPPORTED:
+               return ENOTSUP;
+       case KERN_INVALID_ARGUMENT:
+               return EINVAL;
+       case KERN_RESOURCE_SHORTAGE:
+       default:
+               return EBUSY;
+       }
+}
+
+int
+abort_with_payload(struct proc *cur_proc, struct abort_with_payload_args *args,
+                               __unused void *retval)
+{
+       abort_with_payload_internal(cur_proc, args->reason_namespace,
+                       args->reason_code, args->payload, args->payload_size,
+                       args->reason_string, args->reason_flags, 0);
+
+       return 0;
+}
+
+int
+os_fault_with_payload(struct proc *cur_proc,
+               struct os_fault_with_payload_args *args, __unused int *retval)
+{
+       return abort_with_payload_internal(cur_proc, args->reason_namespace,
+                       args->reason_code, args->payload, args->payload_size,
+                       args->reason_string, args->reason_flags, OS_REASON_IFLAG_USER_FAULT);
+}
+
+
 /*
  * exit --
  *     Death of process.
  */
+__attribute__((noreturn))
 void
 exit(proc_t p, struct exit_args *uap, int *retval)
 {
+       p->p_xhighbits = ((uint32_t)(uap->rval) & 0xFF000000) >> 24;
        exit1(p, W_EXITCODE(uap->rval, 0), retval);
 
        thread_exception_return();
@@ -252,6 +702,16 @@ exit1(proc_t p, int rv, int *retval)
 int
 exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, boolean_t perf_notify,
               int jetsam_flags)
+{
+       return exit_with_reason(p, rv, retval, thread_can_terminate, perf_notify, jetsam_flags, OS_REASON_NULL);
+}
+
+/*
+ * NOTE: exit_with_reason drops a reference on the passed exit_reason
+ */
+int
+exit_with_reason(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, boolean_t perf_notify,
+               int jetsam_flags, struct os_reason *exit_reason)
 {
        thread_t self = current_thread();
        struct task *task = p->task;
@@ -265,7 +725,9 @@ exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, bo
         */
 
         ut = get_bsdthread_info(self);
-        if (ut->uu_flag & UT_VFORK) {
+        if ((p == current_proc()) &&
+            (ut->uu_flag & UT_VFORK)) {
+               os_reason_free(exit_reason);
                if (!thread_can_terminate) {
                        return EINVAL;
                }
@@ -289,38 +751,48 @@ exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, bo
         * what the return value is.
         */
        AUDIT_ARG(exit, WEXITSTATUS(rv), 0);
+       /*
+        * TODO: what to audit here when jetsam calls exit and the uthread,
+        * 'ut' does not belong to the proc, 'p'.
+        */
        AUDIT_SYSCALL_EXIT(SYS_exit, p, ut, 0); /* Exit is always successfull */
 
        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,
+       proc_set_task_policy(p->task, TASK_POLICY_ATTRIBUTE,
                             TASK_POLICY_TERMINATED, TASK_POLICY_ENABLE);
 
         proc_lock(p);
-       error = proc_transstart(p, 1, ((jetsam_flags & P_JETSAM_VNODE) ? 1 : 0));
+       error = proc_transstart(p, 1, (jetsam_flags ? 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.
+               /*
+                * If proc_transstart() returns EDEADLK, then another thread
+                * is either exec'ing or exiting. Return an error and allow
+                * the other thread to continue.
                 */
                proc_unlock(p);
+               os_reason_free(exit_reason);
                if (current_proc() == p){
-                       if (p->exit_thread == self)
+                       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);
+                       }
+
+                       if (thread_can_terminate) {
+                               thread_exception_return();
+                       }
                }
+
+               return error;
        }
 
        while (p->exit_thread != self) {
                if (sig_try_locked(p) <= 0) {
                        proc_transend(p, 1);
+                       os_reason_free(exit_reason);
+
                        if (get_threadtask(self) != task) {
                                proc_unlock(p);
                                return(0);
@@ -337,32 +809,16 @@ exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, bo
                }
                sig_lock_to_exit(p);
        }
-       if (p == initproc && current_proc() == p) {
-               proc_unlock(p);
-               printf("pid 1 exited (signal %d, exit %d)",
-                   WTERMSIG(rv), WEXITSTATUS(rv));
-#if (DEVELOPMENT || DEBUG)
-               int err;
-               /*
-                * 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.
-                */
-               if ((err = coredump(p, 300, 1)) != 0) {
-                       printf("Failed to generate initproc core file: error %d", err);
-               } else {
-                       printf("Generated initproc core file");
-                       sync(p, (void *)NULL, (int *)NULL);
-               }
-#endif
-               init_task_died = TRUE;
-               panic("%s died\nState at Last Exception:\n\n%s", 
-                                                       (p->p_comm[0] != '\0' ?
-                                                               p->p_comm :
-                                                               "launchd"),
-                                                       init_task_failure_data);
+
+       if (exit_reason != OS_REASON_NULL) {
+               KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_COMMIT) | DBG_FUNC_NONE,
+                                                       p->p_pid, exit_reason->osr_namespace,
+                                                       exit_reason->osr_code, 0, 0);
        }
 
+       assert(p->p_exit_reason == OS_REASON_NULL);
+       p->p_exit_reason = exit_reason;
+
        p->p_lflag |= P_LEXIT;
        p->p_xstat = rv;
        p->p_lflag |= jetsam_flags;
@@ -381,14 +837,35 @@ exit1_internal(proc_t p, int rv, int *retval, boolean_t thread_can_terminate, bo
 void
 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)) {
+       /*
+        * Generate a corefile/crashlog if:
+        *      The process doesn't have an exit reason that indicates no crash report should be created
+        *      AND any of the following are true:
+        *      - The process was terminated due to a fatal signal that generates a core
+        *      - The process was killed due to a code signing violation
+        *      - The process has an exit reason that indicates we should generate a crash report
+        *
+        * The first condition is necessary because abort_with_reason()/payload() use SIGABRT
+        * (which normally triggers a core) but may indicate that no crash report should be created.
+        */
+       if (!(PROC_HAS_EXITREASON(p) && (PROC_EXITREASON_FLAGS(p) & OS_REASON_FLAG_NO_CRASH_REPORT)) &&
+               (hassigprop(WTERMSIG(rv), SA_CORE) || ((p->p_csflags & CS_KILLED) != 0) ||
+               (PROC_HAS_EXITREASON(p) && (PROC_EXITREASON_FLAGS(p) &
+                                                       OS_REASON_FLAG_GENERATE_CRASH_REPORT)))) {
                /* 
                 * Workaround for processes checking up on PT_DENY_ATTACH:
                 * should be backed out post-Leopard (details in 5431025).
@@ -407,7 +884,13 @@ proc_prepareexit(proc_t p, int rv, boolean_t perf_notify)
                        ((ut->uu_exception & 0x0f) << 20) | 
                        ((int)ut->uu_code & 0xfffff);
                subcode = ut->uu_subcode;
-               (void) task_exception_notify(EXC_CRASH, 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:
@@ -416,6 +899,22 @@ skipcheck:
                (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(p->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;
+               }
+       }
+
        /*
         * Before this process becomes a zombie, stash resource usage
         * stats in the proc for external observers to query
@@ -436,7 +935,30 @@ skipcheck:
                 */
                p->p_ru = rup;
        }
+       if (create_corpse) {
+               int est_knotes = 0, num_knotes = 0;
+               uint64_t *buffer = NULL;
+               int buf_size = 0;
+
+               /* Get all the udata pointers from kqueue */
+               est_knotes = kevent_proc_copy_uptrs(p, NULL, 0);
+               if (est_knotes > 0) {
+                       buf_size = (est_knotes + 32) * sizeof(uint64_t);
+                       buffer = (uint64_t *) kalloc(buf_size);
+                       num_knotes = kevent_proc_copy_uptrs(p, buffer, buf_size);
+                       if (num_knotes > est_knotes + 32) {
+                               num_knotes = est_knotes + 32;
+                       }
+               }
 
+               /* Update the code, subcode based on exit reason */
+               proc_update_corpse_exception_codes(p, &code, &subcode);
+               populate_corpse_crashinfo(p, p->task, rup,
+                               code, subcode, buffer, num_knotes, NULL);
+               if (buffer != NULL) {
+                       kfree(buffer, buf_size);
+               }
+       }
        /*
         * Remove proc from allproc queue and from pidhash chain.
         * Need to do this before we do anything that can block.
@@ -527,43 +1049,11 @@ proc_exit(proc_t p)
                pid, exitval, 0, 0, 0);
 
 #if CONFIG_DTRACE
-       /*
-        * Free any outstanding lazy dof entries. It is imperative we
-        * always call dtrace_lazy_dofs_destroy, rather than null check
-        * and call if !NULL. If we NULL test, during lazy dof faulting
-        * we can race with the faulting code and proceed from here to
-        * beyond the helpers cleanup. The lazy dof faulting will then
-        * install new helpers which will never be cleaned up, and leak.
-        */
-       dtrace_lazy_dofs_destroy(p);
-
-       /*
-        * Clean up any DTrace helper actions or probes for the process.
-        */
-       if (p->p_dtrace_helpers != NULL) {
-               (*dtrace_helpers_cleanup)(p);
-       }
-
-       /*
-        * Clean up any DTrace probes associated with this process.
-        */
-       /*
-        * APPLE NOTE: We release ptss pages/entries in dtrace_fasttrap_exit_ptr(),
-        * call this after dtrace_helpers_cleanup()
-        */
-       proc_lock(p);
-       if (p->p_dtrace_probes && dtrace_fasttrap_exit_ptr) {
-               (*dtrace_fasttrap_exit_ptr)(p);
-       }
-       proc_unlock(p);
+       dtrace_proc_exit(p);
 #endif
 
        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!
@@ -575,7 +1065,6 @@ proc_exit(proc_t p)
        task_clear_cpuusage(p->task, TRUE);
 
        workqueue_mark_exiting(p);
-       workqueue_exit(p);
 
        _aio_exit( p );
 
@@ -585,6 +1074,12 @@ proc_exit(proc_t p)
         */
        fdfree(p);
 
+       /*
+        * Once all the knotes, kqueues & workloops are destroyed, get rid of the
+        * workqueue.
+        */
+       workqueue_exit(p);
+
        if (uth->uu_lowpri_window) {
                /*
                 * task is marked as a low priority I/O type
@@ -787,7 +1282,7 @@ proc_exit(proc_t p)
                                        */
                                        thread_resume(thread);
                                        clear_wait(thread, THREAD_INTERRUPTED);
-                                       threadsignal(thread, SIGKILL, 0);
+                                       threadsignal(thread, SIGKILL, 0, TRUE);
                                } else {
                                        proc_unlock(q);
                                }
@@ -805,6 +1300,15 @@ proc_exit(proc_t p)
        proc_childdrainend(p);
        proc_list_unlock();
 
+#if CONFIG_MACF
+       /*
+        * Notify MAC policies that proc is dead.
+        * This should be replaced with proper label management
+        * (rdar://problem/32126399).
+        */
+       mac_proc_notify_exit(p);
+#endif
+
        /*
         * Release reference to text vnode
         */
@@ -842,21 +1346,7 @@ proc_exit(proc_t p)
                }
        }
 
-       proc_spinlock(p);
-       if (thread_call_cancel(p->p_rcall))
-               p->p_ractive--;
-
-       while (p->p_ractive > 0) {
-               proc_spinunlock(p);
-               
-               delay(1);
-
-               proc_spinlock(p);
-       }
-       proc_spinunlock(p);
-
-       thread_call_free(p->p_rcall);
-       p->p_rcall = NULL;
+       proc_free_realitimer(p);
 
        /*
         * Other substructures are freed from wait().
@@ -924,6 +1414,8 @@ proc_exit(proc_t p)
                if (pp != initproc) {
                        proc_lock(pp);
                        pp->si_pid = p->p_pid;
+                       pp->p_xhighbits = p->p_xhighbits;
+                       p->p_xhighbits = 0;
                        pp->si_status = p->p_xstat;
                        pp->si_code = CLD_EXITED;
                        /*
@@ -1128,13 +1620,16 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoi
         * and refernce is dropped after these calls down below
         * (locking protection is provided by list lock held in chgproccnt)
         */
+#if CONFIG_PERSONAS
+       /*
+        * persona_proc_drop calls chgproccnt(-1) on the persona uid,
+        * and (+1) on the child->p_ucred uid
+        */
+       persona_proc_drop(child);
+#endif
        (void)chgproccnt(kauth_cred_getruid(child->p_ucred), -1);
 
-#if CONFIG_LCTX
-       ALLLCTX_LOCK;
-       leavelctx(child);
-       ALLLCTX_UNLOCK;
-#endif
+       os_reason_free(child->p_exit_reason);
 
        /*
         * Free up credentials.
@@ -1179,6 +1674,7 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoi
 
 #if CONFIG_FINE_LOCK_GROUPS
        lck_mtx_destroy(&child->p_mlock, proc_mlock_grp);
+       lck_mtx_destroy(&child->p_ucred_mlock, proc_ucred_mlock_grp);
        lck_mtx_destroy(&child->p_fdmlock, proc_fdmlock_grp);
 #if CONFIG_DTRACE
        lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
@@ -1186,13 +1682,13 @@ reap_child_locked(proc_t parent, proc_t child, int deadparent, int reparentedtoi
        lck_spin_destroy(&child->p_slock, proc_slock_grp);
 #else /* CONFIG_FINE_LOCK_GROUPS */
        lck_mtx_destroy(&child->p_mlock, proc_lck_grp);
+       lck_mtx_destroy(&child->p_ucred_mlock, proc_lck_grp);
        lck_mtx_destroy(&child->p_fdmlock, proc_lck_grp);
 #if CONFIG_DTRACE
        lck_mtx_destroy(&child->p_dtrace_sprlock, proc_lck_grp);
 #endif
        lck_spin_destroy(&child->p_slock, proc_lck_grp);
 #endif /* CONFIG_FINE_LOCK_GROUPS */
-       workqueue_destroy_lock(child);
 
        FREE_ZONE(child, sizeof *child, M_PROC);
        if ((locked == 1) && (droplock == 0))
@@ -1253,7 +1749,7 @@ loop1:
        nfound = 0;
        sibling_count = 0;
 
-       for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
+       PCHILDREN_FOREACH(q, p) {
                if ( p->p_sibling.le_next != 0 )
                        sibling_count++;
                if (uap->pid != WAIT_ANY &&
@@ -1496,8 +1992,8 @@ loop:
        proc_list_lock();
 loop1:
        nfound = 0;
-       for (p = q->p_children.lh_first; p != 0; p = p->p_sibling.le_next) {
 
+       PCHILDREN_FOREACH(q, p) {
                switch (uap->idtype) {
                case P_PID:     /* child with process ID equal to... */
                        if (p->p_pid != (pid_t)uap->id)
@@ -1539,7 +2035,8 @@ loop1:
 #endif
                        siginfo.si_signo = SIGCHLD;
                        siginfo.si_pid = p->p_pid;
-                       siginfo.si_status = WEXITSTATUS(p->p_xstat);
+                       siginfo.si_status = (WEXITSTATUS(p->p_xstat) & 0x00FFFFFF) | (((uint32_t)(p->p_xhighbits) << 24) & 0xFF000000);
+                       p->p_xhighbits = 0;
                        if (WIFSIGNALED(p->p_xstat)) {
                                siginfo.si_code = WCOREDUMP(p->p_xstat) ?
                                        CLD_DUMPED : CLD_KILLED;
@@ -1803,37 +2300,15 @@ vfork_exit_internal(proc_t p, int rv, int forceexit)
        p->p_sigignore = ~0;
        proc_unlock(p);
 
-       proc_spinlock(p);
-       if (thread_call_cancel(p->p_rcall))
-               p->p_ractive--;
-
-       while (p->p_ractive > 0) {
-               proc_spinunlock(p);
-               
-               delay(1);
-
-               proc_spinlock(p);
-       }
-       proc_spinunlock(p);
-
-       thread_call_free(p->p_rcall);
-       p->p_rcall = NULL;
-
        ut->uu_siglist = 0;
 
-       vproc_exit(p);
-}
+       /* begin vproc_exit */
 
-void 
-vproc_exit(proc_t p)
-{
        proc_t q;
        proc_t pp;
-       
+
        vnode_t tvp;
-#ifdef FIXME
-       struct task *task = p->task;
-#endif
+
        struct pgrp * pg;
        struct session *sessp;
        struct rusage_superset *rup;
@@ -2007,7 +2482,7 @@ vproc_exit(proc_t p)
                                        */
                                        thread_resume(thread);
                                        clear_wait(thread, THREAD_INTERRUPTED);
-                                       threadsignal(thread, SIGKILL, 0);
+                                       threadsignal(thread, SIGKILL, 0, TRUE);
                                } else {
                                        proc_unlock(q);
                                }
@@ -2104,6 +2579,8 @@ vproc_exit(proc_t p)
        pth_proc_hashdelete(p);
 #endif /* PSYNCH */
 
+       proc_free_realitimer(p);
+
        /*
         * Other substructures are freed from wait().
         */
@@ -2131,6 +2608,8 @@ vproc_exit(proc_t p)
                if (pp != initproc) {
                        proc_lock(pp);
                        pp->si_pid = p->p_pid;
+                       pp->p_xhighbits = p->p_xhighbits;
+                       p->p_xhighbits = 0;
                        pp->si_status = p->p_xstat;
                        pp->si_code = CLD_EXITED;
                        /*