/*
* Initialize all of the debugging state in a port.
* Insert the port into a global list of all allocated ports.
*/
void
ipc_port_init_debug(
ipc_port_t port,
uintptr_t *callstack,
unsigned int callstack_max)
{
unsigned int i;
port->ip_thread = current_thread();
port->ip_timetrack = port_timestamp++;
for (i = 0; i < callstack_max; ++i)
port->ip_callstack[i] = callstack[i];
for (i = 0; i < IP_NSPARES; ++i)
port->ip_spares[i] = 0;
#ifdef MACH_BSD
task_t task = current_task();
if (task != TASK_NULL) {
struct proc* proc = (struct proc*) get_bsdtask_info(task);
if (proc)
port->ip_spares[0] = proc_pid(proc);
}
#endif /* MACH_BSD */
#if 0
lck_spin_lock(&port_alloc_queue_lock);
++port_count;
if (port_count_warning > 0 && port_count >= port_count_warning)
assert(port_count < port_count_warning);
queue_enter(&port_alloc_queue, port, ipc_port_t, ip_port_links);
lck_spin_unlock(&port_alloc_queue_lock);
#endif
}
/*
* pid_for_task
*
* Find the BSD process ID for the Mach task associated with the given Mach port
* name
*
* Parameters: args User argument descriptor (see below)
*
* Indirect parameters: args->t Mach port name
* args->pid Process ID (returned value; see below)
*
* Returns: KERL_SUCCESS Success
* KERN_FAILURE Not success
*
* Implicit returns: args->pid Process ID
*
*/
kern_return_t
pid_for_task(
struct pid_for_task_args *args)
{
mach_port_name_t t = args->t;
user_addr_t pid_addr = args->pid;
proc_t p;
task_t t1;
int pid = -1;
kern_return_t err = KERN_SUCCESS;
AUDIT_MACH_SYSCALL_ENTER(AUE_PIDFORTASK);
AUDIT_ARG(mach_port1, t);
t1 = port_name_to_task(t);
if (t1 == TASK_NULL) {
err = KERN_FAILURE;
goto pftout;
} else {
p = get_bsdtask_info(t1);
if (p) {
pid = proc_pid(p);
err = KERN_SUCCESS;
} else {
err = KERN_FAILURE;
}
}
task_deallocate(t1);
pftout:
AUDIT_ARG(pid, pid);
(void) copyout((char *) &pid, pid_addr, sizeof(int));
AUDIT_MACH_SYSCALL_EXIT(err);
return(err);
}
bool IOHIDEventSystemUserClient::
initWithTask(task_t owningTask, void * /* security_id */, UInt32 /* type */)
{
bool result = false;
OSObject* entitlement = copyClientEntitlement(owningTask, kIOHIDSystemUserAccessServiceEntitlement);
if (entitlement) {
result = (entitlement == kOSBooleanTrue);
entitlement->release();
}
if (!result) {
proc_t process;
process = (proc_t)get_bsdtask_info(owningTask);
char name[255];
bzero(name, sizeof(name));
proc_name(proc_pid(process), name, sizeof(name));
HIDLogError("%s is not entitled", name);
goto exit;
}
result = super::init();
require_action(result, exit, HIDLogError("failed"));
exit:
return result;
}
int
mac_exc_action_check_exception_send(struct task *victim_task, struct exception_action *action)
{
int error = 0;
struct proc *p = get_bsdtask_info(victim_task);
struct label *bsd_label = NULL;
struct label *label = NULL;
if (p != NULL) {
// Create a label from the still existing bsd process...
label = bsd_label = mac_exc_action_label_alloc();
MAC_PERFORM(exc_action_label_update, p, bsd_label);
} else {
// ... otherwise use the crash label on the task.
label = get_task_crash_label(victim_task);
}
if (label == NULL) {
MAC_MACH_UNEXPECTED("mac_exc_action_check_exception_send: no exc_action label for proc %p", p);
return EPERM;
}
MAC_CHECK(exc_action_check_exception_send, label, action, action->label);
if (bsd_label != NULL) {
mac_exc_action_label_free(bsd_label);
}
return (error);
}
IOReturn RootDomainUserClient::secureSleepSystemOptions(
const void *inOptions,
IOByteCount inOptionsSize,
uint32_t *returnCode)
{
int local_priv = 0;
int admin_priv = 0;
IOReturn ret = kIOReturnNotPrivileged;
OSDictionary *unserializedOptions = NULL;
OSString *unserializeErrorString = NULL;
ret = clientHasPrivilege(fOwningTask, kIOClientPrivilegeLocalUser);
local_priv = (kIOReturnSuccess == ret);
ret = clientHasPrivilege(fOwningTask, kIOClientPrivilegeAdministrator);
admin_priv = (kIOReturnSuccess == ret);
if (inOptions)
{
unserializedOptions = OSDynamicCast( OSDictionary,
OSUnserializeXML((const char *)inOptions, inOptionsSize, &unserializeErrorString));
if (!unserializedOptions) {
IOLog("IOPMRootDomain SleepSystem unserialization failure: %s\n",
unserializeErrorString ? unserializeErrorString->getCStringNoCopy() : "Unknown");
}
}
if ( (local_priv || admin_priv) && fOwner )
{
proc_t p;
p = (proc_t)get_bsdtask_info(fOwningTask);
if (p) {
fOwner->setProperty("SleepRequestedByPID", proc_pid(p), 32);
}
if (unserializedOptions)
{
// Publish Sleep Options in registry under root_domain
fOwner->setProperty( kRootDomainSleepOptionsKey, unserializedOptions);
*returnCode = fOwner->sleepSystemOptions( unserializedOptions );
unserializedOptions->release();
} else {
// No options
// Clear any pre-existing options
fOwner->removeProperty( kRootDomainSleepOptionsKey );
*returnCode = fOwner->sleepSystemOptions( NULL );
}
} else {
*returnCode = kIOReturnNotPrivileged;
}
return kIOReturnSuccess;
}
bool IOHIDLibUserClient::initWithTask(task_t owningTask, void * /* security_id */, UInt32 /* type */)
{
if (!super::init())
return false;
if (IOUserClient::clientHasPrivilege(owningTask, kIOClientPrivilegeAdministrator) != kIOReturnSuccess) {
// Preparing for extended data. Set a temporary key.
setProperty(kIOHIDLibClientExtendedData, true);
}
fClient = owningTask;
task_reference (fClient);
proc_t p = (proc_t)get_bsdtask_info(fClient);
fPid = proc_pid(p);
fQueueMap = OSArray::withCapacity(4);
if (!fQueueMap)
return false;
return true;
}
/*
* vfork_return
*
* Description: "Return" to parent vfork thread() following execve/_exit;
* this is done by reassociating the parent process structure
* with the task, thread, and uthread.
*
* Refer to the ASCII art above vfork() to figure out the
* state we're undoing.
*
* Parameters: child_proc Child process
* retval System call return value array
* rval Return value to present to parent
*
* Returns: void
*
* Notes: The caller resumes or exits the parent, as appropriate, after
* calling this function.
*/
void
vfork_return(proc_t child_proc, int32_t *retval, int rval)
{
task_t parent_task = get_threadtask(child_proc->p_vforkact);
proc_t parent_proc = get_bsdtask_info(parent_task);
thread_t th = current_thread();
uthread_t uth = get_bsdthread_info(th);
act_thread_catt(uth->uu_userstate);
/* clear vfork state in parent proc structure */
proc_vfork_end(parent_proc);
/* REPATRIATE PARENT TASK, THREAD, UTHREAD */
uth->uu_userstate = 0;
uth->uu_flag &= ~UT_VFORK;
/* restore thread-set-id state */
if (uth->uu_flag & UT_WASSETUID) {
uth->uu_flag |= UT_SETUID;
uth->uu_flag &= UT_WASSETUID;
}
uth->uu_proc = 0;
uth->uu_sigmask = uth->uu_vforkmask;
proc_lock(child_proc);
child_proc->p_lflag &= ~P_LINVFORK;
child_proc->p_vforkact = 0;
proc_unlock(child_proc);
thread_set_parent(th, rval);
if (retval) {
retval[0] = rval;
retval[1] = 0; /* mark parent */
}
}
/*
* fork1
*
* Description: common code used by all new process creation other than the
* bootstrap of the initial process on the system
*
* Parameters: parent_proc parent process of the process being
* child_threadp pointer to location to receive the
* Mach thread_t of the child process
* breated
* kind kind of creation being requested
*
* Notes: Permissable values for 'kind':
*
* PROC_CREATE_FORK Create a complete process which will
* return actively running in both the
* parent and the child; the child copies
* the parent address space.
* PROC_CREATE_SPAWN Create a complete process which will
* return actively running in the parent
* only after returning actively running
* in the child; the child address space
* is newly created by an image activator,
* after which the child is run.
* PROC_CREATE_VFORK Creates a partial process which will
* borrow the parent task, thread, and
* uthread to return running in the child;
* the child address space and other parts
* are lazily created at execve() time, or
* the child is terminated, and the parent
* does not actively run until that
* happens.
*
* At first it may seem strange that we return the child thread
* address rather than process structure, since the process is
* the only part guaranteed to be "new"; however, since we do
* not actualy adjust other references between Mach and BSD (see
* the block diagram above the implementation of vfork()), this
* is the only method which guarantees us the ability to get
* back to the other information.
*/
int
fork1(proc_t parent_proc, thread_t *child_threadp, int kind)
{
thread_t parent_thread = (thread_t)current_thread();
uthread_t parent_uthread = (uthread_t)get_bsdthread_info(parent_thread);
proc_t child_proc = NULL; /* set in switch, but compiler... */
thread_t child_thread = NULL;
uid_t uid;
int count;
int err = 0;
int spawn = 0;
/*
* Although process entries are dynamically created, we still keep
* a global limit on the maximum number we will create. Don't allow
* a nonprivileged user to use the last process; don't let root
* exceed the limit. The variable nprocs is the current number of
* processes, maxproc is the limit.
*/
uid = kauth_cred_get()->cr_ruid;
proc_list_lock();
if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) {
proc_list_unlock();
tablefull("proc");
return (EAGAIN);
}
proc_list_unlock();
/*
* Increment the count of procs running with this uid. Don't allow
* a nonprivileged user to exceed their current limit, which is
* always less than what an rlim_t can hold.
* (locking protection is provided by list lock held in chgproccnt)
*/
count = chgproccnt(uid, 1);
if (uid != 0 &&
(rlim_t)count > parent_proc->p_rlimit[RLIMIT_NPROC].rlim_cur) {
err = EAGAIN;
goto bad;
}
#if CONFIG_MACF
/*
* Determine if MAC policies applied to the process will allow
* it to fork. This is an advisory-only check.
*/
err = mac_proc_check_fork(parent_proc);
if (err != 0) {
goto bad;
}
#endif
switch(kind) {
case PROC_CREATE_VFORK:
/*
* Prevent a vfork while we are in vfork(); we should
* also likely preventing a fork here as well, and this
* check should then be outside the switch statement,
* since the proc struct contents will copy from the
* child and the tash/thread/uthread from the parent in
* that case. We do not support vfork() in vfork()
* because we don't have to; the same non-requirement
* is true of both fork() and posix_spawn() and any
* call other than execve() amd _exit(), but we've
* been historically lenient, so we continue to be so
* (for now).
*
* <rdar://6640521> Probably a source of random panics
*/
if (parent_uthread->uu_flag & UT_VFORK) {
printf("fork1 called within vfork by %s\n", parent_proc->p_comm);
err = EINVAL;
goto bad;
}
/*
* Flag us in progress; if we chose to support vfork() in
* vfork(), we would chain our parent at this point (in
* effect, a stack push). We don't, since we actually want
* to disallow everything not specified in the standard
*/
proc_vfork_begin(parent_proc);
/* The newly created process comes with signal lock held */
if ((child_proc = forkproc(parent_proc)) == NULL) {
/* Failed to allocate new process */
proc_vfork_end(parent_proc);
err = ENOMEM;
goto bad;
}
// XXX BEGIN: wants to move to be common code (and safe)
#if CONFIG_MACF
/*
* allow policies to associate the credential/label that
* we referenced from the parent ... with the child
* JMM - this really isn't safe, as we can drop that
* association without informing the policy in other
* situations (keep long enough to get policies changed)
*/
mac_cred_label_associate_fork(child_proc->p_ucred, child_proc);
#endif
/*
* Propogate change of PID - may get new cred if auditing.
*
* NOTE: This has no effect in the vfork case, since
* child_proc->task != current_task(), but we duplicate it
* because this is probably, ultimately, wrong, since we
* will be running in the "child" which is the parent task
* with the wrong token until we get to the execve() or
* _exit() call; a lot of "undefined" can happen before
* that.
*
* <rdar://6640530> disallow everything but exeve()/_exit()?
*/
set_security_token(child_proc);
AUDIT_ARG(pid, child_proc->p_pid);
AUDIT_SESSION_PROCNEW(child_proc->p_ucred);
// XXX END: wants to move to be common code (and safe)
/*
* BORROW PARENT TASK, THREAD, UTHREAD FOR CHILD
*
* Note: this is where we would "push" state instead of setting
* it for nested vfork() support (see proc_vfork_end() for
* description if issues here).
*/
child_proc->task = parent_proc->task;
child_proc->p_lflag |= P_LINVFORK;
child_proc->p_vforkact = parent_thread;
child_proc->p_stat = SRUN;
parent_uthread->uu_flag |= UT_VFORK;
parent_uthread->uu_proc = child_proc;
parent_uthread->uu_userstate = (void *)act_thread_csave();
parent_uthread->uu_vforkmask = parent_uthread->uu_sigmask;
/* temporarily drop thread-set-id state */
if (parent_uthread->uu_flag & UT_SETUID) {
parent_uthread->uu_flag |= UT_WASSETUID;
parent_uthread->uu_flag &= ~UT_SETUID;
}
/* blow thread state information */
/* XXX is this actually necessary, given syscall return? */
thread_set_child(parent_thread, child_proc->p_pid);
child_proc->p_acflag = AFORK; /* forked but not exec'ed */
/*
* Preserve synchronization semantics of vfork. If
* waiting for child to exec or exit, set P_PPWAIT
* on child, and sleep on our proc (in case of exit).
*/
child_proc->p_lflag |= P_LPPWAIT;
pinsertchild(parent_proc, child_proc); /* set visible */
break;
case PROC_CREATE_SPAWN:
/*
* A spawned process differs from a forked process in that
* the spawned process does not carry around the parents
* baggage with regard to address space copying, dtrace,
* and so on.
*/
spawn = 1;
/* FALLSTHROUGH */
case PROC_CREATE_FORK:
/*
* When we clone the parent process, we are going to inherit
* its task attributes and memory, since when we fork, we
* will, in effect, create a duplicate of it, with only minor
* differences. Contrarily, spawned processes do not inherit.
*/
if ((child_thread = cloneproc(parent_proc->task, parent_proc, spawn ? FALSE : TRUE)) == NULL) {
/* Failed to create thread */
err = EAGAIN;
goto bad;
}
/* copy current thread state into the child thread (only for fork) */
if (!spawn) {
thread_dup(child_thread);
}
/* child_proc = child_thread->task->proc; */
child_proc = (proc_t)(get_bsdtask_info(get_threadtask(child_thread)));
// XXX BEGIN: wants to move to be common code (and safe)
#if CONFIG_MACF
/*
* allow policies to associate the credential/label that
* we referenced from the parent ... with the child
* JMM - this really isn't safe, as we can drop that
* association without informing the policy in other
* situations (keep long enough to get policies changed)
*/
mac_cred_label_associate_fork(child_proc->p_ucred, child_proc);
#endif
/*
* Propogate change of PID - may get new cred if auditing.
*
* NOTE: This has no effect in the vfork case, since
* child_proc->task != current_task(), but we duplicate it
* because this is probably, ultimately, wrong, since we
* will be running in the "child" which is the parent task
* with the wrong token until we get to the execve() or
* _exit() call; a lot of "undefined" can happen before
* that.
*
* <rdar://6640530> disallow everything but exeve()/_exit()?
*/
set_security_token(child_proc);
AUDIT_ARG(pid, child_proc->p_pid);
AUDIT_SESSION_PROCNEW(child_proc->p_ucred);
// XXX END: wants to move to be common code (and safe)
/*
* Blow thread state information; this is what gives the child
* process its "return" value from a fork() call.
*
* Note: this should probably move to fork() proper, since it
* is not relevent to spawn, and the value won't matter
* until we resume the child there. If you are in here
* refactoring code, consider doing this at the same time.
*/
thread_set_child(child_thread, child_proc->p_pid);
child_proc->p_acflag = AFORK; /* forked but not exec'ed */
// <rdar://6598155> dtrace code cleanup needed
#if CONFIG_DTRACE
/*
* This code applies to new processes who are copying the task
* and thread state and address spaces of their parent process.
*/
if (!spawn) {
// <rdar://6598155> call dtrace specific function here instead of all this...
/*
* APPLE NOTE: Solaris does a sprlock() and drops the
* proc_lock here. We're cheating a bit and only taking
* the p_dtrace_sprlock lock. A full sprlock would
* task_suspend the parent.
*/
lck_mtx_lock(&parent_proc->p_dtrace_sprlock);
/*
* Remove all DTrace tracepoints from the child process. We
* need to do this _before_ duplicating USDT providers since
* any associated probes may be immediately enabled.
*/
if (parent_proc->p_dtrace_count > 0) {
dtrace_fasttrap_fork(parent_proc, child_proc);
}
lck_mtx_unlock(&parent_proc->p_dtrace_sprlock);
/*
* Duplicate any lazy dof(s). This must be done while NOT
* holding the parent sprlock! Lock ordering is
* dtrace_dof_mode_lock, then sprlock. It is imperative we
* always call dtrace_lazy_dofs_duplicate, rather than null
* check and call if !NULL. If we NULL test, during lazy dof
* faulting we can race with the faulting code and proceed
* from here to beyond the helpers copy. The lazy dof
* faulting will then fail to copy the helpers to the child
* process.
*/
dtrace_lazy_dofs_duplicate(parent_proc, child_proc);
/*
* Duplicate any helper actions and providers. The SFORKING
* we set above informs the code to enable USDT probes that
* sprlock() may fail because the child is being forked.
*/
/*
* APPLE NOTE: As best I can tell, Apple's sprlock() equivalent
* never fails to find the child. We do not set SFORKING.
*/
if (parent_proc->p_dtrace_helpers != NULL && dtrace_helpers_fork) {
(*dtrace_helpers_fork)(parent_proc, child_proc);
}
}
#endif /* CONFIG_DTRACE */
break;
default:
panic("fork1 called with unknown kind %d", kind);
break;
}
/* return the thread pointer to the caller */
*child_threadp = child_thread;
bad:
/*
* In the error case, we return a 0 value for the returned pid (but
* it is ignored in the trampoline due to the error return); this
* is probably not necessary.
*/
if (err) {
(void)chgproccnt(uid, -1);
}
return (err);
}
/*
* This routine frees all the BSD context in uthread except the credential.
* It does not free the uthread structure as well
*/
void
uthread_cleanup(task_t task, void *uthread, void * bsd_info)
{
struct _select *sel;
uthread_t uth = (uthread_t)uthread;
proc_t p = (proc_t)bsd_info;
if (uth->uu_lowpri_window || uth->uu_throttle_info) {
/*
* task is marked as a low priority I/O type
* and we've somehow managed to not dismiss the throttle
* through the normal exit paths back to user space...
* no need to throttle this thread since its going away
* but we do need to update our bookeeping w/r to throttled threads
*
* Calling this routine will clean up any throttle info reference
* still inuse by the thread.
*/
throttle_lowpri_io(FALSE);
}
/*
* Per-thread audit state should never last beyond system
* call return. Since we don't audit the thread creation/
* removal, the thread state pointer should never be
* non-NULL when we get here.
*/
assert(uth->uu_ar == NULL);
sel = &uth->uu_select;
/* cleanup the select bit space */
if (sel->nbytes) {
FREE(sel->ibits, M_TEMP);
FREE(sel->obits, M_TEMP);
sel->nbytes = 0;
}
if (uth->uu_cdir) {
vnode_rele(uth->uu_cdir);
uth->uu_cdir = NULLVP;
}
if (uth->uu_allocsize && uth->uu_wqset){
kfree(uth->uu_wqset, uth->uu_allocsize);
sel->count = 0;
uth->uu_allocsize = 0;
uth->uu_wqset = 0;
sel->wql = 0;
}
if(uth->pth_name != NULL)
{
kfree(uth->pth_name, MAXTHREADNAMESIZE);
uth->pth_name = 0;
}
if ((task != kernel_task) && p) {
if (((uth->uu_flag & UT_VFORK) == UT_VFORK) && (uth->uu_proc != PROC_NULL)) {
vfork_exit_internal(uth->uu_proc, 0, 1);
}
/*
* Remove the thread from the process list and
* transfer [appropriate] pending signals to the process.
*/
if (get_bsdtask_info(task) == p) {
proc_lock(p);
TAILQ_REMOVE(&p->p_uthlist, uth, uu_list);
p->p_siglist |= (uth->uu_siglist & execmask & (~p->p_sigignore | sigcantmask));
proc_unlock(p);
}
#if CONFIG_DTRACE
struct dtrace_ptss_page_entry *tmpptr = uth->t_dtrace_scratch;
uth->t_dtrace_scratch = NULL;
if (tmpptr != NULL) {
dtrace_ptss_release_entry(p, tmpptr);
}
#endif
}
}
void *
uthread_alloc(task_t task, thread_t thread, int noinherit)
{
proc_t p;
uthread_t uth;
uthread_t uth_parent;
void *ut;
if (!uthread_zone_inited)
uthread_zone_init();
ut = (void *)zalloc(uthread_zone);
bzero(ut, sizeof(struct uthread));
p = (proc_t) get_bsdtask_info(task);
uth = (uthread_t)ut;
/*
* Thread inherits credential from the creating thread, if both
* are in the same task.
*
* If the creating thread has no credential or is from another
* task we can leave the new thread credential NULL. If it needs
* one later, it will be lazily assigned from the task's process.
*/
uth_parent = (uthread_t)get_bsdthread_info(current_thread());
if ((noinherit == 0) && task == current_task() &&
uth_parent != NULL &&
IS_VALID_CRED(uth_parent->uu_ucred)) {
/*
* XXX The new thread is, in theory, being created in context
* XXX of parent thread, so a direct reference to the parent
* XXX is OK.
*/
kauth_cred_ref(uth_parent->uu_ucred);
uth->uu_ucred = uth_parent->uu_ucred;
/* the credential we just inherited is an assumed credential */
if (uth_parent->uu_flag & UT_SETUID)
uth->uu_flag |= UT_SETUID;
} else {
/* sometimes workqueue threads are created out task context */
if ((task != kernel_task) && (p != PROC_NULL))
uth->uu_ucred = kauth_cred_proc_ref(p);
else
uth->uu_ucred = NOCRED;
}
if ((task != kernel_task) && p) {
proc_lock(p);
if (noinherit != 0) {
/* workq threads will not inherit masks */
uth->uu_sigmask = ~workq_threadmask;
} else if (uth_parent) {
if (uth_parent->uu_flag & UT_SAS_OLDMASK)
uth->uu_sigmask = uth_parent->uu_oldmask;
else
uth->uu_sigmask = uth_parent->uu_sigmask;
}
uth->uu_context.vc_thread = thread;
TAILQ_INSERT_TAIL(&p->p_uthlist, uth, uu_list);
proc_unlock(p);
#if CONFIG_DTRACE
if (p->p_dtrace_ptss_pages != NULL) {
uth->t_dtrace_scratch = dtrace_ptss_claim_entry(p);
}
#endif
}
return (ut);
}
/*
* Function: unix_syscall
*
* Inputs: regs - pointer to i386 save area
*
* Outputs: none
*/
void
unix_syscall(x86_saved_state_t *state)
{
thread_t thread;
void *vt;
unsigned int code;
struct sysent *callp;
int error;
vm_offset_t params;
struct proc *p;
struct uthread *uthread;
x86_saved_state32_t *regs;
boolean_t is_vfork;
assert(is_saved_state32(state));
regs = saved_state32(state);
#if DEBUG
if (regs->eax == 0x800)
thread_exception_return();
#endif
thread = current_thread();
uthread = get_bsdthread_info(thread);
/* Get the approriate proc; may be different from task's for vfork() */
is_vfork = uthread->uu_flag & UT_VFORK;
if (__improbable(is_vfork != 0))
p = current_proc();
else
p = (struct proc *)get_bsdtask_info(current_task());
/* Verify that we are not being called from a task without a proc */
if (__improbable(p == NULL)) {
regs->eax = EPERM;
regs->efl |= EFL_CF;
task_terminate_internal(current_task());
thread_exception_return();
/* NOTREACHED */
}
code = regs->eax & I386_SYSCALL_NUMBER_MASK;
DEBUG_KPRINT_SYSCALL_UNIX("unix_syscall: code=%d(%s) eip=%u\n",
code, syscallnames[code >= NUM_SYSENT ? 63 : code], (uint32_t)regs->eip);
params = (vm_offset_t) (regs->uesp + sizeof (int));
regs->efl &= ~(EFL_CF);
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
if (__improbable(callp == sysent)) {
code = fuword(params);
params += sizeof(int);
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
}
vt = (void *)uthread->uu_arg;
if (callp->sy_arg_bytes != 0) {
#if CONFIG_REQUIRES_U32_MUNGING
sy_munge_t *mungerp;
#else
#error U32 syscalls on x86_64 kernel requires munging
#endif
uint32_t nargs;
assert((unsigned) callp->sy_arg_bytes <= sizeof (uthread->uu_arg));
nargs = callp->sy_arg_bytes;
error = copyin((user_addr_t) params, (char *) vt, nargs);
if (error) {
regs->eax = error;
regs->efl |= EFL_CF;
thread_exception_return();
/* NOTREACHED */
}
if (__probable(code != 180)) {
int *ip = (int *)vt;
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
*ip, *(ip+1), *(ip+2), *(ip+3), 0);
}
#if CONFIG_REQUIRES_U32_MUNGING
mungerp = callp->sy_arg_munge32;
if (mungerp != NULL)
(*mungerp)(vt);
#endif
} else
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
0, 0, 0, 0, 0);
/*
* Delayed binding of thread credential to process credential, if we
* are not running with an explicitly set thread credential.
*/
kauth_cred_uthread_update(uthread, p);
uthread->uu_rval[0] = 0;
uthread->uu_rval[1] = 0;
uthread->uu_flag |= UT_NOTCANCELPT;
uthread->syscall_code = code;
#ifdef JOE_DEBUG
uthread->uu_iocount = 0;
uthread->uu_vpindex = 0;
#endif
AUDIT_SYSCALL_ENTER(code, p, uthread);
error = (*(callp->sy_call))((void *) p, (void *) vt, &(uthread->uu_rval[0]));
AUDIT_SYSCALL_EXIT(code, p, uthread, error);
#ifdef JOE_DEBUG
if (uthread->uu_iocount)
printf("system call returned with uu_iocount != 0\n");
#endif
#if CONFIG_DTRACE
uthread->t_dtrace_errno = error;
#endif /* CONFIG_DTRACE */
if (__improbable(error == ERESTART)) {
/*
* Move the user's pc back to repeat the syscall:
* 5 bytes for a sysenter, or 2 for an int 8x.
* The SYSENTER_TF_CS covers single-stepping over a sysenter
* - see debug trap handler in idt.s/idt64.s
*/
pal_syscall_restart(thread, state);
}
else if (error != EJUSTRETURN) {
if (__improbable(error)) {
regs->eax = error;
regs->efl |= EFL_CF; /* carry bit */
} else { /* (not error) */
/*
* We split retval across two registers, in case the
* syscall had a 64-bit return value, in which case
* eax/edx matches the function call ABI.
*/
regs->eax = uthread->uu_rval[0];
regs->edx = uthread->uu_rval[1];
}
}
DEBUG_KPRINT_SYSCALL_UNIX(
"unix_syscall: error=%d retval=(%u,%u)\n",
error, regs->eax, regs->edx);
uthread->uu_flag &= ~UT_NOTCANCELPT;
if (__improbable(uthread->uu_lowpri_window)) {
/*
* task is marked as a low priority I/O type
* and the I/O we issued while in this system call
* collided with normal I/O operations... we'll
* delay in order to mitigate the impact of this
* task on the normal operation of the system
*/
throttle_lowpri_io(1);
}
if (__probable(code != 180))
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
error, uthread->uu_rval[0], uthread->uu_rval[1], p->p_pid, 0);
if (__improbable(!is_vfork && callp->sy_call == (sy_call_t *)execve && !error)) {
pal_execve_return(thread);
}
thread_exception_return();
/* NOTREACHED */
}
void
unix_syscall64(x86_saved_state_t *state)
{
thread_t thread;
void *vt;
unsigned int code;
struct sysent *callp;
int args_in_regs;
boolean_t args_start_at_rdi;
int error;
struct proc *p;
struct uthread *uthread;
x86_saved_state64_t *regs;
assert(is_saved_state64(state));
regs = saved_state64(state);
#if DEBUG
if (regs->rax == 0x2000800)
thread_exception_return();
#endif
thread = current_thread();
uthread = get_bsdthread_info(thread);
/* Get the approriate proc; may be different from task's for vfork() */
if (__probable(!(uthread->uu_flag & UT_VFORK)))
p = (struct proc *)get_bsdtask_info(current_task());
else
p = current_proc();
/* Verify that we are not being called from a task without a proc */
if (__improbable(p == NULL)) {
regs->rax = EPERM;
regs->isf.rflags |= EFL_CF;
task_terminate_internal(current_task());
thread_exception_return();
/* NOTREACHED */
}
code = regs->rax & SYSCALL_NUMBER_MASK;
DEBUG_KPRINT_SYSCALL_UNIX(
"unix_syscall64: code=%d(%s) rip=%llx\n",
code, syscallnames[code >= NUM_SYSENT ? 63 : code], regs->isf.rip);
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
vt = (void *)uthread->uu_arg;
if (__improbable(callp == sysent)) {
/*
* indirect system call... system call number
* passed as 'arg0'
*/
code = regs->rdi;
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
args_start_at_rdi = FALSE;
args_in_regs = 5;
} else {
args_start_at_rdi = TRUE;
args_in_regs = 6;
}
if (callp->sy_narg != 0) {
assert(callp->sy_narg <= 8); /* size of uu_arg */
args_in_regs = MIN(args_in_regs, callp->sy_narg);
memcpy(vt, args_start_at_rdi ? ®s->rdi : ®s->rsi, args_in_regs * sizeof(syscall_arg_t));
if (code != 180) {
uint64_t *ip = (uint64_t *)vt;
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
(int)(*ip), (int)(*(ip+1)), (int)(*(ip+2)), (int)(*(ip+3)), 0);
}
if (__improbable(callp->sy_narg > args_in_regs)) {
int copyin_count;
copyin_count = (callp->sy_narg - args_in_regs) * sizeof(syscall_arg_t);
error = copyin((user_addr_t)(regs->isf.rsp + sizeof(user_addr_t)), (char *)&uthread->uu_arg[args_in_regs], copyin_count);
if (error) {
regs->rax = error;
regs->isf.rflags |= EFL_CF;
thread_exception_return();
/* NOTREACHED */
}
}
} else
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
0, 0, 0, 0, 0);
/*
* Delayed binding of thread credential to process credential, if we
* are not running with an explicitly set thread credential.
*/
kauth_cred_uthread_update(uthread, p);
uthread->uu_rval[0] = 0;
uthread->uu_rval[1] = 0;
uthread->uu_flag |= UT_NOTCANCELPT;
uthread->syscall_code = code;
#ifdef JOE_DEBUG
uthread->uu_iocount = 0;
uthread->uu_vpindex = 0;
#endif
AUDIT_SYSCALL_ENTER(code, p, uthread);
error = (*(callp->sy_call))((void *) p, vt, &(uthread->uu_rval[0]));
AUDIT_SYSCALL_EXIT(code, p, uthread, error);
#ifdef JOE_DEBUG
if (uthread->uu_iocount)
printf("system call returned with uu_iocount != 0\n");
#endif
#if CONFIG_DTRACE
uthread->t_dtrace_errno = error;
#endif /* CONFIG_DTRACE */
if (__improbable(error == ERESTART)) {
/*
* all system calls come through via the syscall instruction
* in 64 bit mode... its 2 bytes in length
* move the user's pc back to repeat the syscall:
*/
pal_syscall_restart( thread, state );
}
else if (error != EJUSTRETURN) {
if (__improbable(error)) {
regs->rax = error;
regs->isf.rflags |= EFL_CF; /* carry bit */
} else { /* (not error) */
switch (callp->sy_return_type) {
case _SYSCALL_RET_INT_T:
regs->rax = uthread->uu_rval[0];
regs->rdx = uthread->uu_rval[1];
break;
case _SYSCALL_RET_UINT_T:
regs->rax = ((u_int)uthread->uu_rval[0]);
regs->rdx = ((u_int)uthread->uu_rval[1]);
break;
case _SYSCALL_RET_OFF_T:
case _SYSCALL_RET_ADDR_T:
case _SYSCALL_RET_SIZE_T:
case _SYSCALL_RET_SSIZE_T:
case _SYSCALL_RET_UINT64_T:
regs->rax = *((uint64_t *)(&uthread->uu_rval[0]));
regs->rdx = 0;
break;
case _SYSCALL_RET_NONE:
break;
default:
panic("unix_syscall: unknown return type");
break;
}
regs->isf.rflags &= ~EFL_CF;
}
}
DEBUG_KPRINT_SYSCALL_UNIX(
"unix_syscall64: error=%d retval=(%llu,%llu)\n",
error, regs->rax, regs->rdx);
uthread->uu_flag &= ~UT_NOTCANCELPT;
if (__improbable(uthread->uu_lowpri_window)) {
/*
* task is marked as a low priority I/O type
* and the I/O we issued while in this system call
* collided with normal I/O operations... we'll
* delay in order to mitigate the impact of this
* task on the normal operation of the system
*/
throttle_lowpri_io(1);
}
if (__probable(code != 180))
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
error, uthread->uu_rval[0], uthread->uu_rval[1], p->p_pid, 0);
thread_exception_return();
/* NOTREACHED */
}
