/*
* vfork_return
*
* Description: "Return" to parent vfork thread() following execve/_exit;
* this is done by reassociating the parent process structure
* with the task, thread, and uthread.
*
* Parameters: child_proc Child process
* retval System call return value array
* rval Return value to present to parent
*
* Returns: void
*
* Note: The caller resumes or exits the parent, as appropriate, after
* callling this function.
*/
void
vfork_return(proc_t child_proc, int32_t *retval, int rval)
{
proc_t parent_proc = child_proc->p_pptr;
thread_t parent_thread = (thread_t)current_thread();
uthread_t parent_uthread = (uthread_t)get_bsdthread_info(parent_thread);
act_thread_catt(parent_uthread->uu_userstate);
/* end vfork in parent */
proc_vfork_end(parent_proc);
/* REPATRIATE PARENT TASK, THREAD, UTHREAD */
parent_uthread->uu_userstate = 0;
parent_uthread->uu_flag &= ~UT_VFORK;
/* restore thread-set-id state */
if (parent_uthread->uu_flag & UT_WASSETUID) {
parent_uthread->uu_flag |= UT_SETUID;
parent_uthread->uu_flag &= UT_WASSETUID;
}
parent_uthread->uu_proc = 0;
parent_uthread->uu_sigmask = parent_uthread->uu_vforkmask;
child_proc->p_lflag &= ~P_LINVFORK;
child_proc->p_vforkact = (void *)0;
thread_set_parent(parent_thread, rval);
if (retval) {
retval[0] = rval;
retval[1] = 0; /* mark parent */
}
return;
}
void
bsd_utaskbootstrap(void)
{
thread_t thread;
struct uthread *ut;
/*
* Clone the bootstrap process from the kernel process, without
* inheriting either task characteristics or memory from the kernel;
*/
thread = cloneproc(TASK_NULL, COALITION_NULL, kernproc, FALSE, TRUE);
/* Hold the reference as it will be dropped during shutdown */
initproc = proc_find(1);
#if __PROC_INTERNAL_DEBUG
if (initproc == PROC_NULL)
panic("bsd_utaskbootstrap: initproc not set\n");
#endif
/*
* Since we aren't going back out the normal way to our parent,
* we have to drop the transition locks explicitly.
*/
proc_signalend(initproc, 0);
proc_transend(initproc, 0);
ut = (struct uthread *)get_bsdthread_info(thread);
ut->uu_sigmask = 0;
act_set_astbsd(thread);
proc_clear_return_wait(initproc, thread);
}
struct label *
mac_thread_get_threadlabel(struct thread *thread)
{
struct uthread *uthread = get_bsdthread_info(thread);
return (mac_thread_get_uthreadlabel(uthread));
}
void
bsdinit_task(void)
{
proc_t p = current_proc();
struct uthread *ut;
thread_t thread;
process_name("init", p);
ux_handler_init();
thread = current_thread();
(void) host_set_exception_ports(host_priv_self(),
EXC_MASK_ALL & ~(EXC_MASK_RPC_ALERT),//pilotfish (shark) needs this port
(mach_port_t) ux_exception_port,
EXCEPTION_DEFAULT| MACH_EXCEPTION_CODES,
0);
ut = (uthread_t)get_bsdthread_info(thread);
bsd_init_task = get_threadtask(thread);
init_task_died = FALSE;
#if CONFIG_MACF
mac_cred_label_associate_user(p->p_ucred);
#endif
load_init_program(p);
lock_trace = 1;
}
/*
* vfork
*
* Description: vfork system call
*
* Parameters: void [no arguments]
*
* Retval: 0 (to child process)
* !0 pid of child (to parent process)
* -1 error (see "Returns:")
*
* Returns: EAGAIN Administrative limit reached
* EINVAL vfork() called during vfork()
* ENOMEM Failed to allocate new process
*
* Note: After a successful call to this function, the parent process
* has its task, thread, and uthread lent to the child process,
* and control is returned to the caller; if this function is
* invoked as a system call, the return is to user space, and
* is effectively running on the child process.
*
* Subsequent calls that operate on process state are permitted,
* though discouraged, and will operate on the child process; any
* operations on the task, thread, or uthread will result in
* changes in the parent state, and, if inheritable, the child
* state, when a task, thread, and uthread are realized for the
* child process at execve() time, will also be effected. Given
* this, it's recemmended that people use the posix_spawn() call
* instead.
*
* BLOCK DIAGRAM OF VFORK
*
* Before:
*
* ,----------------. ,-------------.
* | | task | |
* | parent_thread | ------> | parent_task |
* | | <.list. | |
* `----------------' `-------------'
* uthread | ^ bsd_info | ^
* v | vc_thread v | task
* ,----------------. ,-------------.
* | | | |
* | parent_uthread | <.list. | parent_proc | <-- current_proc()
* | | | |
* `----------------' `-------------'
* uu_proc |
* v
* NULL
*
* After:
*
* ,----------------. ,-------------.
* | | task | |
* ,----> | parent_thread | ------> | parent_task |
* | | | <.list. | |
* | `----------------' `-------------'
* | uthread | ^ bsd_info | ^
* | v | vc_thread v | task
* | ,----------------. ,-------------.
* | | | | |
* | | parent_uthread | <.list. | parent_proc |
* | | | | |
* | `----------------' `-------------'
* | uu_proc | . list
* | v v
* | ,----------------.
* `----- | |
* p_vforkact | child_proc | <-- current_proc()
* | |
* `----------------'
*/
int
vfork(proc_t parent_proc, __unused struct vfork_args *uap, int32_t *retval)
{
thread_t child_thread;
int err;
if ((err = fork1(parent_proc, &child_thread, PROC_CREATE_VFORK, NULL)) != 0) {
retval[1] = 0;
} else {
uthread_t ut = get_bsdthread_info(current_thread());
proc_t child_proc = ut->uu_proc;
retval[0] = child_proc->p_pid;
retval[1] = 1; /* flag child return for user space */
/*
* Drop the signal lock on the child which was taken on our
* behalf by forkproc()/cloneproc() to prevent signals being
* received by the child in a partially constructed state.
*/
proc_signalend(child_proc, 0);
proc_transend(child_proc, 0);
proc_knote(parent_proc, NOTE_FORK | child_proc->p_pid);
DTRACE_PROC1(create, proc_t, child_proc);
ut->uu_flag &= ~UT_VFORKING;
}
return (err);
}
int *
get_bsduthreadrval(thread_t th)
{
struct uthread *ut;
ut = get_bsdthread_info(th);
return(&ut->uu_rval[0]);
}
/*
* dtrace_CRED() can be called from probe context. We cannot simply call kauth_cred_get() since
* that function may try to resolve a lazy credential binding, which entails taking the proc_lock.
*/
cred_t *
dtrace_CRED(void)
{
struct uthread *uthread = get_bsdthread_info(current_thread());
if (uthread == NULL)
return NULL;
else
return uthread->uu_ucred; /* May return NOCRED which is defined to be 0 */
}
static void
_sleep_continue(void)
{
register struct proc *p;
register thread_t self = current_act();
struct uthread * ut;
int sig, catch;
int error = 0;
ut = get_bsdthread_info(self);
catch = ut->uu_pri & PCATCH;
static void
_sleep_continue( __unused void *parameter, wait_result_t wresult)
{
struct proc *p = current_proc();
thread_t self = current_thread();
struct uthread * ut;
int sig, catch;
int error = 0;
int dropmutex, spinmutex;
ut = get_bsdthread_info(self);
catch = ut->uu_pri & PCATCH;
void
afs_osi_fullSigRestore()
{
#ifndef AFS_DARWIN80_ENV
struct uthread *user_thread = (struct uthread *)get_bsdthread_info(current_act());
/* Protect original sigmask */
if (user_thread->uu_oldmask) {
/* Restore original sigmask */
user_thread->uu_sigmask = user_thread->uu_oldmask;
/* Clear the oldmask */
user_thread->uu_oldmask = (sigset_t)0;
}
#endif
}
void *
get_bsduthreadarg(thread_t th)
{
void *arg_ptr;
struct uthread *ut;
ut = get_bsdthread_info(th);
if (ml_thread_is64bit(th) == TRUE)
arg_ptr = (void *)saved_state64(find_user_regs(th));
else
arg_ptr = (void *)(ut->uu_arg);
return(arg_ptr);
}
void
afs_osi_fullSigMask()
{
#ifndef AFS_DARWIN80_ENV
struct uthread *user_thread = (struct uthread *)get_bsdthread_info(current_act());
/* Protect original sigmask */
if (!user_thread->uu_oldmask) {
/* Back up current sigmask */
user_thread->uu_oldmask = user_thread->uu_sigmask;
/* Mask all signals */
user_thread->uu_sigmask = ~(sigset_t)0;
}
#endif
}
int throttle_get_io_policy(struct uthread **ut)
{
int policy = IOPOL_DEFAULT;
proc_t p = current_proc();
*ut = get_bsdthread_info(current_thread());
if (p != NULL)
policy = p->p_iopol_disk;
if (*ut != NULL) {
// the I/O policy of the thread overrides that of the process
// unless the I/O policy of the thread is default
if ((*ut)->uu_iopol_disk != IOPOL_DEFAULT)
policy = (*ut)->uu_iopol_disk;
}
return policy;
}
void throttle_lowpri_io(boolean_t ok_to_sleep)
{
int i;
int max_try_num;
struct uthread *ut;
struct _throttle_io_info_t *info;
ut = get_bsdthread_info(current_thread());
if ((ut->uu_lowpri_window == 0) || (ut->uu_throttle_info == NULL))
goto done;
info = ut->uu_throttle_info;
KERNEL_DEBUG((FSDBG_CODE(DBG_FSRW, 97)) | DBG_FUNC_START,
ut->uu_lowpri_window, ok_to_sleep, 0, 0, 0);
if (ok_to_sleep == TRUE) {
max_try_num = lowpri_max_waiting_msecs / LOWPRI_SLEEP_INTERVAL * MAX(1, info->numthreads_throttling);
for (i=0; i<max_try_num; i++) {
if (throttle_io_will_be_throttled_internal(ut->uu_lowpri_window, info)) {
IOSleep(LOWPRI_SLEEP_INTERVAL);
DEBUG_ALLOC_THROTTLE_INFO("sleeping because of info = %p\n", info, info );
} else {
break;
}
}
}
KERNEL_DEBUG((FSDBG_CODE(DBG_FSRW, 97)) | DBG_FUNC_END,
ut->uu_lowpri_window, i*5, 0, 0, 0);
SInt32 oldValue;
oldValue = OSDecrementAtomic(&info->numthreads_throttling);
if (oldValue <= 0) {
panic("%s: numthreads negative", __func__);
}
done:
ut->uu_lowpri_window = 0;
if (ut->uu_throttle_info)
throttle_info_rel(ut->uu_throttle_info);
ut->uu_throttle_info = NULL;
}
static uint64_t
machtrace_getarg(void *arg, dtrace_id_t id, void *parg, int argno, int aframes)
{
#pragma unused(arg,id,parg,aframes) /* __APPLE__ */
uint64_t val = 0;
syscall_arg_t *stack = (syscall_arg_t *)NULL;
uthread_t uthread = (uthread_t)get_bsdthread_info(current_thread());
if (uthread)
stack = (syscall_arg_t *)uthread->t_dtrace_syscall_args;
if (!stack)
return(0);
DTRACE_CPUFLAG_SET(CPU_DTRACE_NOFAULT);
/* dtrace_probe arguments arg0 .. arg4 are 64bits wide */
val = (uint64_t)*(stack+argno);
DTRACE_CPUFLAG_CLEAR(CPU_DTRACE_NOFAULT);
return (val);
}
/*
* 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 */
}
}
/*
* Convert a pathname into a pointer to a locked inode.
*
* The FOLLOW flag is set when symbolic links are to be followed
* when they occur at the end of the name translation process.
* Symbolic links are always followed for all other pathname
* components other than the last.
*
* The segflg defines whether the name is to be copied from user
* space or kernel space.
*
* Overall outline of namei:
*
* copy in name
* get starting directory
* while (!done && !error) {
* call lookup to search path.
* if symbolic link, massage name in buffer and continue
* }
*
* Returns: 0 Success
* ENOENT No such file or directory
* ELOOP Too many levels of symbolic links
* ENAMETOOLONG Filename too long
* copyinstr:EFAULT Bad address
* copyinstr:ENAMETOOLONG Filename too long
* lookup:EBADF Bad file descriptor
* lookup:EROFS
* lookup:EACCES
* lookup:EPERM
* lookup:ERECYCLE vnode was recycled from underneath us in lookup.
* This means we should re-drive lookup from this point.
* lookup: ???
* VNOP_READLINK:???
*/
int
namei(struct nameidata *ndp)
{
struct filedesc *fdp; /* pointer to file descriptor state */
struct vnode *dp; /* the directory we are searching */
struct vnode *usedvp = ndp->ni_dvp; /* store pointer to vp in case we must loop due to
heavy vnode pressure */
u_long cnpflags = ndp->ni_cnd.cn_flags; /* store in case we have to restore after loop */
int error;
struct componentname *cnp = &ndp->ni_cnd;
vfs_context_t ctx = cnp->cn_context;
proc_t p = vfs_context_proc(ctx);
#if CONFIG_AUDIT
/* XXX ut should be from context */
uthread_t ut = (struct uthread *)get_bsdthread_info(current_thread());
#endif
fdp = p->p_fd;
#if DIAGNOSTIC
if (!vfs_context_ucred(ctx) || !p)
panic ("namei: bad cred/proc");
if (cnp->cn_nameiop & (~OPMASK))
panic ("namei: nameiop contaminated with flags");
if (cnp->cn_flags & OPMASK)
panic ("namei: flags contaminated with nameiops");
#endif
/*
* A compound VNOP found something that needs further processing:
* either a trigger vnode, a covered directory, or a symlink.
*/
if (ndp->ni_flag & NAMEI_CONTLOOKUP) {
int rdonly, vbusyflags, keep_going, wantparent;
rdonly = cnp->cn_flags & RDONLY;
vbusyflags = ((cnp->cn_flags & CN_NBMOUNTLOOK) != 0) ? LK_NOWAIT : 0;
keep_going = 0;
wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
ndp->ni_flag &= ~(NAMEI_CONTLOOKUP);
error = lookup_handle_found_vnode(ndp, &ndp->ni_cnd, rdonly, vbusyflags,
&keep_going, ndp->ni_ncgeneration, wantparent, 0, ctx);
if (error)
goto out_drop;
if (keep_going) {
if ((cnp->cn_flags & ISSYMLINK) == 0) {
panic("We need to keep going on a continued lookup, but for vp type %d (tag %d)\n", ndp->ni_vp->v_type, ndp->ni_vp->v_tag);
}
goto continue_symlink;
}
return 0;
}
vnode_recycled:
/*
* Get a buffer for the name to be translated, and copy the
* name into the buffer.
*/
if ((cnp->cn_flags & HASBUF) == 0) {
cnp->cn_pnbuf = ndp->ni_pathbuf;
cnp->cn_pnlen = PATHBUFLEN;
}
#if LP64_DEBUG
if ((UIO_SEG_IS_USER_SPACE(ndp->ni_segflg) == 0)
&& (ndp->ni_segflg != UIO_SYSSPACE)
&& (ndp->ni_segflg != UIO_SYSSPACE32)) {
panic("%s :%d - invalid ni_segflg\n", __FILE__, __LINE__);
}
#endif /* LP64_DEBUG */
retry_copy:
if (UIO_SEG_IS_USER_SPACE(ndp->ni_segflg)) {
error = copyinstr(ndp->ni_dirp, cnp->cn_pnbuf,
cnp->cn_pnlen, (size_t *)&ndp->ni_pathlen);
} else {
error = copystr(CAST_DOWN(void *, ndp->ni_dirp), cnp->cn_pnbuf,
cnp->cn_pnlen, (size_t *)&ndp->ni_pathlen);
}
if (error == ENAMETOOLONG && !(cnp->cn_flags & HASBUF)) {
MALLOC_ZONE(cnp->cn_pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
if (cnp->cn_pnbuf == NULL) {
error = ENOMEM;
goto error_out;
}
cnp->cn_flags |= HASBUF;
cnp->cn_pnlen = MAXPATHLEN;
goto retry_copy;
}
if (error)
goto error_out;
#if CONFIG_VOLFS
/*
* Check for legacy volfs style pathnames.
*
* For compatibility reasons we currently allow these paths,
* but future versions of the OS may not support them.
*/
if (ndp->ni_pathlen >= VOLFS_MIN_PATH_LEN &&
cnp->cn_pnbuf[0] == '/' &&
cnp->cn_pnbuf[1] == '.' &&
cnp->cn_pnbuf[2] == 'v' &&
cnp->cn_pnbuf[3] == 'o' &&
cnp->cn_pnbuf[4] == 'l' &&
cnp->cn_pnbuf[5] == '/' ) {
char * realpath;
int realpath_err;
/* Attempt to resolve a legacy volfs style pathname. */
MALLOC_ZONE(realpath, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
if (realpath) {
/*
* We only error out on the ENAMETOOLONG cases where we know that
* vfs_getrealpath translation succeeded but the path could not fit into
* MAXPATHLEN characters. In other failure cases, we may be dealing with a path
* that legitimately looks like /.vol/1234/567 and is not meant to be translated
*/
if ((realpath_err= vfs_getrealpath(&cnp->cn_pnbuf[6], realpath, MAXPATHLEN, ctx))) {
FREE_ZONE(realpath, MAXPATHLEN, M_NAMEI);
if (realpath_err == ENOSPC || realpath_err == ENAMETOOLONG){
error = ENAMETOOLONG;
goto error_out;
}
} else {
if (cnp->cn_flags & HASBUF) {
FREE_ZONE(cnp->cn_pnbuf, cnp->cn_pnlen, M_NAMEI);
}
cnp->cn_pnbuf = realpath;
cnp->cn_pnlen = MAXPATHLEN;
ndp->ni_pathlen = strlen(realpath) + 1;
cnp->cn_flags |= HASBUF | CN_VOLFSPATH;
}
}
}
#endif /* CONFIG_VOLFS */
#if CONFIG_AUDIT
/* If we are auditing the kernel pathname, save the user pathname */
if (cnp->cn_flags & AUDITVNPATH1)
AUDIT_ARG(upath, ut->uu_cdir, cnp->cn_pnbuf, ARG_UPATH1);
if (cnp->cn_flags & AUDITVNPATH2)
AUDIT_ARG(upath, ut->uu_cdir, cnp->cn_pnbuf, ARG_UPATH2);
#endif /* CONFIG_AUDIT */
/*
* Do not allow empty pathnames
*/
if (*cnp->cn_pnbuf == '\0') {
error = ENOENT;
goto error_out;
}
ndp->ni_loopcnt = 0;
/*
* determine the starting point for the translation.
*/
if ((ndp->ni_rootdir = fdp->fd_rdir) == NULLVP) {
if ( !(fdp->fd_flags & FD_CHROOT))
ndp->ni_rootdir = rootvnode;
}
cnp->cn_nameptr = cnp->cn_pnbuf;
ndp->ni_usedvp = NULLVP;
if (*(cnp->cn_nameptr) == '/') {
while (*(cnp->cn_nameptr) == '/') {
cnp->cn_nameptr++;
ndp->ni_pathlen--;
}
dp = ndp->ni_rootdir;
} else if (cnp->cn_flags & USEDVP) {
dp = ndp->ni_dvp;
ndp->ni_usedvp = dp;
} else
dp = vfs_context_cwd(ctx);
if (dp == NULLVP || (dp->v_lflag & VL_DEAD)) {
error = ENOENT;
goto error_out;
}
ndp->ni_dvp = NULLVP;
ndp->ni_vp = NULLVP;
for (;;) {
ndp->ni_startdir = dp;
if ( (error = lookup(ndp)) ) {
goto error_out;
}
/*
* Check for symbolic link
*/
if ((cnp->cn_flags & ISSYMLINK) == 0) {
return (0);
}
continue_symlink:
/* Gives us a new path to process, and a starting dir */
error = lookup_handle_symlink(ndp, &dp, ctx);
if (error != 0) {
break;
}
}
/*
* only come here if we fail to handle a SYMLINK...
* if either ni_dvp or ni_vp is non-NULL, then
* we need to drop the iocount that was picked
* up in the lookup routine
*/
out_drop:
if (ndp->ni_dvp)
vnode_put(ndp->ni_dvp);
if (ndp->ni_vp)
vnode_put(ndp->ni_vp);
error_out:
if ( (cnp->cn_flags & HASBUF) ) {
cnp->cn_flags &= ~HASBUF;
FREE_ZONE(cnp->cn_pnbuf, cnp->cn_pnlen, M_NAMEI);
}
cnp->cn_pnbuf = NULL;
ndp->ni_vp = NULLVP;
ndp->ni_dvp = NULLVP;
if (error == ERECYCLE){
/* vnode was recycled underneath us. re-drive lookup to start at
the beginning again, since recycling invalidated last lookup*/
ndp->ni_cnd.cn_flags = cnpflags;
ndp->ni_dvp = usedvp;
goto vnode_recycled;
}
return (error);
}
/*
* 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_syscall_return(int error)
{
thread_t thread;
struct uthread *uthread;
struct proc *p;
unsigned int code;
struct sysent *callp;
thread = current_thread();
uthread = get_bsdthread_info(thread);
pal_register_cache_state(thread, DIRTY);
p = current_proc();
if (proc_is64bit(p)) {
x86_saved_state64_t *regs;
regs = saved_state64(find_user_regs(thread));
code = uthread->syscall_code;
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
#if CONFIG_DTRACE
if (callp->sy_call == dtrace_systrace_syscall)
dtrace_systrace_syscall_return( code, error, uthread->uu_rval );
#endif /* CONFIG_DTRACE */
AUDIT_SYSCALL_EXIT(code, p, uthread, error);
if (error == ERESTART) {
/*
* repeat the syscall
*/
pal_syscall_restart( thread, find_user_regs(thread) );
}
else if (error != EJUSTRETURN) {
if (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_syscall_return: error=%d retval=(%llu,%llu)\n",
error, regs->rax, regs->rdx);
} else {
x86_saved_state32_t *regs;
regs = saved_state32(find_user_regs(thread));
regs->efl &= ~(EFL_CF);
code = uthread->syscall_code;
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
#if CONFIG_DTRACE
if (callp->sy_call == dtrace_systrace_syscall)
dtrace_systrace_syscall_return( code, error, uthread->uu_rval );
#endif /* CONFIG_DTRACE */
AUDIT_SYSCALL_EXIT(code, p, uthread, error);
if (error == ERESTART) {
pal_syscall_restart( thread, find_user_regs(thread) );
}
else if (error != EJUSTRETURN) {
if (error) {
regs->eax = error;
regs->efl |= EFL_CF; /* carry bit */
} else { /* (not error) */
regs->eax = uthread->uu_rval[0];
regs->edx = uthread->uu_rval[1];
}
}
DEBUG_KPRINT_SYSCALL_UNIX(
"unix_syscall_return: error=%d retval=(%u,%u)\n",
error, regs->eax, regs->edx);
}
uthread->uu_flag &= ~UT_NOTCANCELPT;
if (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 (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 */
}
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 */
}
/* ARGSUSED */
int
sigreturn(
struct proc * p,
struct sigreturn_args * uap,
__unused int *retval)
{
union {
user_ucontext32_t uc32;
#if defined(__arm64__)
user_ucontext64_t uc64;
#endif
} uctx;
union {
mcontext32_t mc32;
#if defined(__arm64__)
mcontext64_t mc64;
#endif
} mctx;
struct sigacts *ps = p->p_sigacts;
int error, sigmask = 0, onstack = 0;
thread_t th_act;
struct uthread *ut;
uint32_t sigreturn_validation;
user_addr_t token_uctx;
kern_return_t kr;
th_act = current_thread();
ut = (struct uthread *) get_bsdthread_info(th_act);
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
error = sigreturn_copyin_ctx64(&uctx.uc64, &mctx.mc64, uap->uctx);
if (error != 0) {
return error;
}
onstack = uctx.uc64.uc_onstack;
sigmask = uctx.uc64.uc_sigmask;
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
error = sigreturn_copyin_ctx32(&uctx.uc32, &mctx.mc32, uap->uctx);
if (error != 0) {
return error;
}
onstack = uctx.uc32.uc_onstack;
sigmask = uctx.uc32.uc_sigmask;
}
if ((onstack & 01))
p->p_sigacts->ps_sigstk.ss_flags |= SA_ONSTACK;
else
p->p_sigacts->ps_sigstk.ss_flags &= ~SA_ONSTACK;
ut->uu_sigmask = sigmask & ~sigcantmask;
if (ut->uu_siglist & ~ut->uu_sigmask)
signal_setast(current_thread());
sigreturn_validation = atomic_load_explicit(
&ps->ps_sigreturn_validation, memory_order_relaxed);
token_uctx = uap->uctx;
kr = machine_thread_siguctx_pointer_convert_to_user(th_act, &token_uctx);
assert(kr == KERN_SUCCESS);
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
user64_addr_t token;
token = (user64_addr_t)token_uctx ^ (user64_addr_t)ps->ps_sigreturn_token;
if ((user64_addr_t)uap->token != token) {
#if DEVELOPMENT || DEBUG
printf("process %s[%d] sigreturn token mismatch: received 0x%llx expected 0x%llx\n",
p->p_comm, p->p_pid, (user64_addr_t)uap->token, token);
#endif /* DEVELOPMENT || DEBUG */
if (sigreturn_validation != PS_SIGRETURN_VALIDATION_DISABLED) {
return EINVAL;
}
}
error = sigreturn_set_state64(th_act, &mctx.mc64);
if (error != 0) {
#if DEVELOPMENT || DEBUG
printf("process %s[%d] sigreturn set_state64 error %d\n",
p->p_comm, p->p_pid, error);
#endif /* DEVELOPMENT || DEBUG */
return error;
}
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
user32_addr_t token;
token = (user32_addr_t)token_uctx ^ (user32_addr_t)ps->ps_sigreturn_token;
if ((user32_addr_t)uap->token != token) {
#if DEVELOPMENT || DEBUG
printf("process %s[%d] sigreturn token mismatch: received 0x%x expected 0x%x\n",
p->p_comm, p->p_pid, (user32_addr_t)uap->token, token);
#endif /* DEVELOPMENT || DEBUG */
if (sigreturn_validation != PS_SIGRETURN_VALIDATION_DISABLED) {
return EINVAL;
}
}
error = sigreturn_set_state32(th_act, &mctx.mc32);
if (error != 0) {
#if DEVELOPMENT || DEBUG
printf("process %s[%d] sigreturn sigreturn_set_state32 error %d\n",
p->p_comm, p->p_pid, error);
#endif /* DEVELOPMENT || DEBUG */
return error;
}
}
return (EJUSTRETURN);
}
/*
* Send an interrupt to process.
*
*/
void
sendsig(
struct proc * p,
user_addr_t catcher,
int sig,
int mask,
__unused uint32_t code
)
{
union {
struct ts32 {
arm_thread_state_t ss;
} ts32;
#if defined(__arm64__)
struct ts64 {
arm_thread_state64_t ss;
} ts64;
#endif
} ts;
union {
struct user_sigframe32 uf32;
#if defined(__arm64__)
struct user_sigframe64 uf64;
#endif
} user_frame;
user_siginfo_t sinfo;
user_addr_t sp = 0, trampact;
struct sigacts *ps = p->p_sigacts;
int oonstack, infostyle;
thread_t th_act;
struct uthread *ut;
user_size_t stack_size = 0;
user_addr_t p_uctx, token_uctx;
kern_return_t kr;
th_act = current_thread();
ut = get_bsdthread_info(th_act);
bzero(&ts, sizeof(ts));
bzero(&user_frame, sizeof(user_frame));
if (p->p_sigacts->ps_siginfo & sigmask(sig))
infostyle = UC_FLAVOR;
else
infostyle = UC_TRAD;
trampact = ps->ps_trampact[sig];
oonstack = ps->ps_sigstk.ss_flags & SA_ONSTACK;
/*
* Get sundry thread state.
*/
if (proc_is64bit_data(p)) {
#ifdef __arm64__
if (sendsig_get_state64(th_act, &ts.ts64.ss, &user_frame.uf64.mctx) != 0) {
goto bad2;
}
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
if (sendsig_get_state32(th_act, &ts.ts32.ss, &user_frame.uf32.mctx) != 0) {
goto bad2;
}
}
/*
* Figure out where our new stack lives.
*/
if ((ps->ps_flags & SAS_ALTSTACK) && !oonstack &&
(ps->ps_sigonstack & sigmask(sig))) {
sp = ps->ps_sigstk.ss_sp;
sp += ps->ps_sigstk.ss_size;
stack_size = ps->ps_sigstk.ss_size;
ps->ps_sigstk.ss_flags |= SA_ONSTACK;
} else {
/*
* Get stack pointer, and allocate enough space
* for signal handler data.
*/
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
sp = CAST_USER_ADDR_T(ts.ts64.ss.sp);
sp = (sp - sizeof(user_frame.uf64) - C_64_REDZONE_LEN) & ~0xf; /* Make sure to align to 16 bytes and respect red zone */
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
sp = CAST_USER_ADDR_T(ts.ts32.ss.sp);
sp -= sizeof(user_frame.uf32);
#if defined(__arm__) && (__BIGGEST_ALIGNMENT__ > 4)
sp &= ~0xf; /* Make sure to align to 16 bytes for armv7k */
#endif
}
}
proc_unlock(p);
/*
* Fill in ucontext (points to mcontext, i.e. thread states).
*/
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
sendsig_fill_uctx64(&user_frame.uf64.uctx, oonstack, mask, sp, (user64_size_t)stack_size,
(user64_addr_t)&((struct user_sigframe64*)sp)->mctx);
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
sendsig_fill_uctx32(&user_frame.uf32.uctx, oonstack, mask, sp, (user32_size_t)stack_size,
(user32_addr_t)&((struct user_sigframe32*)sp)->mctx);
}
/*
* Setup siginfo.
*/
bzero((caddr_t) & sinfo, sizeof(sinfo));
sinfo.si_signo = sig;
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
sinfo.si_addr = ts.ts64.ss.pc;
sinfo.pad[0] = ts.ts64.ss.sp;
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
sinfo.si_addr = ts.ts32.ss.pc;
sinfo.pad[0] = ts.ts32.ss.sp;
}
switch (sig) {
case SIGILL:
#ifdef BER_XXX
if (mctx.ss.srr1 & (1 << (31 - SRR1_PRG_ILL_INS_BIT)))
sinfo.si_code = ILL_ILLOPC;
else if (mctx.ss.srr1 & (1 << (31 - SRR1_PRG_PRV_INS_BIT)))
sinfo.si_code = ILL_PRVOPC;
else if (mctx.ss.srr1 & (1 << (31 - SRR1_PRG_TRAP_BIT)))
sinfo.si_code = ILL_ILLTRP;
else
sinfo.si_code = ILL_NOOP;
#else
sinfo.si_code = ILL_ILLTRP;
#endif
break;
case SIGFPE:
break;
case SIGBUS:
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
sinfo.si_addr = user_frame.uf64.mctx.es.far;
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
sinfo.si_addr = user_frame.uf32.mctx.es.far;
}
sinfo.si_code = BUS_ADRALN;
break;
case SIGSEGV:
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
sinfo.si_addr = user_frame.uf64.mctx.es.far;
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
sinfo.si_addr = user_frame.uf32.mctx.es.far;
}
#ifdef BER_XXX
/* First check in srr1 and then in dsisr */
if (mctx.ss.srr1 & (1 << (31 - DSISR_PROT_BIT)))
sinfo.si_code = SEGV_ACCERR;
else if (mctx.es.dsisr & (1 << (31 - DSISR_PROT_BIT)))
sinfo.si_code = SEGV_ACCERR;
else
sinfo.si_code = SEGV_MAPERR;
#else
sinfo.si_code = SEGV_ACCERR;
#endif
break;
default:
{
int status_and_exitcode;
/*
* All other signals need to fill out a minimum set of
* information for the siginfo structure passed into
* the signal handler, if SA_SIGINFO was specified.
*
* p->si_status actually contains both the status and
* the exit code; we save it off in its own variable
* for later breakdown.
*/
proc_lock(p);
sinfo.si_pid = p->si_pid;
p->si_pid = 0;
status_and_exitcode = p->si_status;
p->si_status = 0;
sinfo.si_uid = p->si_uid;
p->si_uid = 0;
sinfo.si_code = p->si_code;
p->si_code = 0;
proc_unlock(p);
if (sinfo.si_code == CLD_EXITED) {
if (WIFEXITED(status_and_exitcode))
sinfo.si_code = CLD_EXITED;
else if (WIFSIGNALED(status_and_exitcode)) {
if (WCOREDUMP(status_and_exitcode)) {
sinfo.si_code = CLD_DUMPED;
status_and_exitcode = W_EXITCODE(status_and_exitcode,status_and_exitcode);
} else {
sinfo.si_code = CLD_KILLED;
status_and_exitcode = W_EXITCODE(status_and_exitcode,status_and_exitcode);
}
}
}
/*
* The recorded status contains the exit code and the
* signal information, but the information to be passed
* in the siginfo to the handler is supposed to only
* contain the status, so we have to shift it out.
*/
sinfo.si_status = (WEXITSTATUS(status_and_exitcode) & 0x00FFFFFF) | (((uint32_t)(p->p_xhighbits) << 24) & 0xFF000000);
p->p_xhighbits = 0;
break;
}
}
#if CONFIG_DTRACE
sendsig_do_dtrace(ut, &sinfo, sig, catcher);
#endif /* CONFIG_DTRACE */
/*
* Copy signal-handling frame out to user space, set thread state.
*/
if (proc_is64bit_data(p)) {
#if defined(__arm64__)
user64_addr_t token;
/*
* mctx filled in when we get state. uctx filled in by
* sendsig_fill_uctx64(). We fill in the sinfo now.
*/
siginfo_user_to_user64(&sinfo, &user_frame.uf64.sinfo);
p_uctx = (user_addr_t)&((struct user_sigframe64*)sp)->uctx;
/*
* Generate the validation token for sigreturn
*/
token_uctx = p_uctx;
kr = machine_thread_siguctx_pointer_convert_to_user(th_act, &token_uctx);
assert(kr == KERN_SUCCESS);
token = (user64_addr_t)token_uctx ^ (user64_addr_t)ps->ps_sigreturn_token;
if (copyout(&user_frame.uf64, sp, sizeof(user_frame.uf64)) != 0) {
goto bad;
}
if (sendsig_set_thread_state64(&ts.ts64.ss,
catcher, infostyle, sig, (user64_addr_t)&((struct user_sigframe64*)sp)->sinfo,
(user64_addr_t)p_uctx, token, trampact, sp, th_act) != KERN_SUCCESS)
goto bad;
#else
panic("Shouldn't have 64-bit thread states on a 32-bit kernel.");
#endif
} else {
user32_addr_t token;
/*
* mctx filled in when we get state. uctx filled in by
* sendsig_fill_uctx32(). We fill in the sinfo, *pointer*
* to uctx and token now.
*/
siginfo_user_to_user32(&sinfo, &user_frame.uf32.sinfo);
p_uctx = (user_addr_t)&((struct user_sigframe32*)sp)->uctx;
/*
* Generate the validation token for sigreturn
*/
token_uctx = (user_addr_t)p_uctx;
kr = machine_thread_siguctx_pointer_convert_to_user(th_act, &token_uctx);
assert(kr == KERN_SUCCESS);
token = (user32_addr_t)token_uctx ^ (user32_addr_t)ps->ps_sigreturn_token;
user_frame.uf32.puctx = (user32_addr_t)p_uctx;
user_frame.uf32.token = token;
if (copyout(&user_frame.uf32, sp, sizeof(user_frame.uf32)) != 0) {
goto bad;
}
if (sendsig_set_thread_state32(&ts.ts32.ss,
CAST_DOWN_EXPLICIT(user32_addr_t, catcher), infostyle, sig, (user32_addr_t)&((struct user_sigframe32*)sp)->sinfo,
CAST_DOWN_EXPLICIT(user32_addr_t, trampact), CAST_DOWN_EXPLICIT(user32_addr_t, sp), th_act) != KERN_SUCCESS)
goto bad;
}
proc_lock(p);
return;
bad:
proc_lock(p);
bad2:
SIGACTION(p, SIGILL) = SIG_DFL;
sig = sigmask(SIGILL);
p->p_sigignore &= ~sig;
p->p_sigcatch &= ~sig;
ut->uu_sigmask &= ~sig;
/* sendsig is called with signal lock held */
proc_unlock(p);
psignal_locked(p, SIGILL);
proc_lock(p);
}
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);
}
/*
* Search a pathname.
* This is a very central and rather complicated routine.
*
* The pathname is pointed to by ni_ptr and is of length ni_pathlen.
* The starting directory is taken from ni_startdir. The pathname is
* descended until done, or a symbolic link is encountered. The variable
* ni_more is clear if the path is completed; it is set to one if a
* symbolic link needing interpretation is encountered.
*
* The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
* whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it, the parent directory is returned
* locked. If flag has WANTPARENT or'ed into it, the parent directory is
* returned unlocked. Otherwise the parent directory is not returned. If
* the target of the pathname exists and LOCKLEAF is or'ed into the flag
* the target is returned locked, otherwise it is returned unlocked.
* When creating or renaming and LOCKPARENT is specified, the target may not
* be ".". When deleting and LOCKPARENT is specified, the target may be ".".
*
* Overall outline of lookup:
*
* dirloop:
* identify next component of name at ndp->ni_ptr
* handle degenerate case where name is null string
* if .. and crossing mount points and on mounted filesys, find parent
* call VNOP_LOOKUP routine for next component name
* directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
* component vnode returned in ni_vp (if it exists), locked.
* if result vnode is mounted on and crossing mount points,
* find mounted on vnode
* if more components of name, do next level at dirloop
* return the answer in ni_vp, locked if LOCKLEAF set
* if LOCKPARENT set, return locked parent in ni_dvp
* if WANTPARENT set, return unlocked parent in ni_dvp
*
* Returns: 0 Success
* ENOENT No such file or directory
* EBADF Bad file descriptor
* ENOTDIR Not a directory
* EROFS Read-only file system [CREATE]
* EISDIR Is a directory [CREATE]
* cache_lookup_path:ERECYCLE (vnode was recycled from underneath us, redrive lookup again)
* vnode_authorize:EROFS
* vnode_authorize:EACCES
* vnode_authorize:EPERM
* vnode_authorize:???
* VNOP_LOOKUP:ENOENT No such file or directory
* VNOP_LOOKUP:EJUSTRETURN Restart system call (INTERNAL)
* VNOP_LOOKUP:???
* VFS_ROOT:ENOTSUP
* VFS_ROOT:ENOENT
* VFS_ROOT:???
*/
int
lookup(struct nameidata *ndp)
{
char *cp; /* pointer into pathname argument */
vnode_t tdp; /* saved dp */
vnode_t dp; /* the directory we are searching */
int docache = 1; /* == 0 do not cache last component */
int wantparent; /* 1 => wantparent or lockparent flag */
int rdonly; /* lookup read-only flag bit */
int dp_authorized = 0;
int error = 0;
struct componentname *cnp = &ndp->ni_cnd;
vfs_context_t ctx = cnp->cn_context;
int vbusyflags = 0;
int nc_generation = 0;
vnode_t last_dp = NULLVP;
int keep_going;
int atroot;
/*
* Setup: break out flag bits into variables.
*/
if (cnp->cn_flags & (NOCACHE | DOWHITEOUT)) {
if ((cnp->cn_flags & NOCACHE) || (cnp->cn_nameiop == DELETE))
docache = 0;
}
wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
rdonly = cnp->cn_flags & RDONLY;
cnp->cn_flags &= ~ISSYMLINK;
cnp->cn_consume = 0;
dp = ndp->ni_startdir;
ndp->ni_startdir = NULLVP;
if ((cnp->cn_flags & CN_NBMOUNTLOOK) != 0)
vbusyflags = LK_NOWAIT;
cp = cnp->cn_nameptr;
if (*cp == '\0') {
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_vp = dp;
error = lookup_handle_emptyname(ndp, cnp, wantparent);
if (error) {
goto bad;
}
return 0;
}
dirloop:
atroot = 0;
ndp->ni_vp = NULLVP;
if ( (error = cache_lookup_path(ndp, cnp, dp, ctx, &dp_authorized, last_dp)) ) {
dp = NULLVP;
goto bad;
}
if ((cnp->cn_flags & ISLASTCN)) {
if (docache)
cnp->cn_flags |= MAKEENTRY;
} else
cnp->cn_flags |= MAKEENTRY;
dp = ndp->ni_dvp;
if (ndp->ni_vp != NULLVP) {
/*
* cache_lookup_path returned a non-NULL ni_vp then,
* we're guaranteed that the dp is a VDIR, it's
* been authorized, and vp is not ".."
*
* make sure we don't try to enter the name back into
* the cache if this vp is purged before we get to that
* check since we won't have serialized behind whatever
* activity is occurring in the FS that caused the purge
*/
if (dp != NULLVP)
nc_generation = dp->v_nc_generation - 1;
goto returned_from_lookup_path;
}
/*
* Handle "..": two special cases.
* 1. If at root directory (e.g. after chroot)
* or at absolute root directory
* then ignore it so can't get out.
* 2. If this vnode is the root of a mounted
* filesystem, then replace it with the
* vnode which was mounted on so we take the
* .. in the other file system.
*/
if ( (cnp->cn_flags & ISDOTDOT) ) {
for (;;) {
if (dp == ndp->ni_rootdir || dp == rootvnode) {
ndp->ni_dvp = dp;
ndp->ni_vp = dp;
/*
* we're pinned at the root
* we've already got one reference on 'dp'
* courtesy of cache_lookup_path... take
* another one for the ".."
* if we fail to get the new reference, we'll
* drop our original down in 'bad'
*/
if ( (vnode_get(dp)) ) {
error = ENOENT;
goto bad;
}
atroot = 1;
goto returned_from_lookup_path;
}
if ((dp->v_flag & VROOT) == 0 ||
(cnp->cn_flags & NOCROSSMOUNT))
break;
if (dp->v_mount == NULL) { /* forced umount */
error = EBADF;
goto bad;
}
tdp = dp;
dp = tdp->v_mount->mnt_vnodecovered;
vnode_put(tdp);
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_dvp = dp;
dp_authorized = 0;
}
}
/*
* We now have a segment name to search for, and a directory to search.
*/
unionlookup:
ndp->ni_vp = NULLVP;
if (dp->v_type != VDIR) {
error = ENOTDIR;
goto lookup_error;
}
if ( (cnp->cn_flags & DONOTAUTH) != DONOTAUTH ) {
error = lookup_authorize_search(dp, cnp, dp_authorized, ctx);
if (error) {
goto lookup_error;
}
}
/*
* Now that we've authorized a lookup, can bail out if the filesystem
* will be doing a batched operation. Return an iocount on dvp.
*/
#if NAMEDRSRCFORK
if ((cnp->cn_flags & ISLASTCN) && namei_compound_available(dp, ndp) && !(cnp->cn_flags & CN_WANTSRSRCFORK)) {
#else
if ((cnp->cn_flags & ISLASTCN) && namei_compound_available(dp, ndp)) {
#endif /* NAMEDRSRCFORK */
ndp->ni_flag |= NAMEI_UNFINISHED;
ndp->ni_ncgeneration = dp->v_nc_generation;
return 0;
}
nc_generation = dp->v_nc_generation;
error = VNOP_LOOKUP(dp, &ndp->ni_vp, cnp, ctx);
if ( error ) {
lookup_error:
if ((error == ENOENT) &&
(dp->v_flag & VROOT) && (dp->v_mount != NULL) &&
(dp->v_mount->mnt_flag & MNT_UNION)) {
#if CONFIG_VFS_FUNNEL
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(dp, NULL);
}
#endif /* CONFIG_VFS_FUNNEL */
tdp = dp;
dp = tdp->v_mount->mnt_vnodecovered;
vnode_put(tdp);
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_dvp = dp;
dp_authorized = 0;
goto unionlookup;
}
if (error != EJUSTRETURN)
goto bad;
if (ndp->ni_vp != NULLVP)
panic("leaf should be empty");
error = lookup_validate_creation_path(ndp);
if (error)
goto bad;
/*
* We return with ni_vp NULL to indicate that the entry
* doesn't currently exist, leaving a pointer to the
* referenced directory vnode in ndp->ni_dvp.
*/
if (cnp->cn_flags & SAVESTART) {
if ( (vnode_get(ndp->ni_dvp)) ) {
error = ENOENT;
goto bad;
}
ndp->ni_startdir = ndp->ni_dvp;
}
if (!wantparent)
vnode_put(ndp->ni_dvp);
if (kdebug_enable)
kdebug_lookup(ndp->ni_dvp, cnp);
return (0);
}
returned_from_lookup_path:
/* We'll always have an iocount on ni_vp when this finishes. */
error = lookup_handle_found_vnode(ndp, cnp, rdonly, vbusyflags, &keep_going, nc_generation, wantparent, atroot, ctx);
if (error != 0) {
goto bad2;
}
if (keep_going) {
dp = ndp->ni_vp;
/* namei() will handle symlinks */
if ((dp->v_type == VLNK) &&
((cnp->cn_flags & FOLLOW) || (ndp->ni_flag & NAMEI_TRAILINGSLASH) || *ndp->ni_next == '/')) {
return 0;
}
/*
* Otherwise, there's more path to process.
* cache_lookup_path is now responsible for dropping io ref on dp
* when it is called again in the dirloop. This ensures we hold
* a ref on dp until we complete the next round of lookup.
*/
last_dp = dp;
goto dirloop;
}
return (0);
bad2:
#if CONFIG_VFS_FUNNEL
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(ndp->ni_dvp, NULL);
}
#endif /* CONFIG_VFS_FUNNEL */
if (ndp->ni_dvp)
vnode_put(ndp->ni_dvp);
vnode_put(ndp->ni_vp);
ndp->ni_vp = NULLVP;
if (kdebug_enable)
kdebug_lookup(dp, cnp);
return (error);
bad:
#if CONFIG_VFS_FUNNEL
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(ndp->ni_dvp, NULL);
}
#endif /* CONFIG_VFS_FUNNEL */
if (dp)
vnode_put(dp);
ndp->ni_vp = NULLVP;
if (kdebug_enable)
kdebug_lookup(dp, cnp);
return (error);
}
int
lookup_validate_creation_path(struct nameidata *ndp)
{
struct componentname *cnp = &ndp->ni_cnd;
/*
* If creating and at end of pathname, then can consider
* allowing file to be created.
*/
if (cnp->cn_flags & RDONLY) {
return EROFS;
}
if ((cnp->cn_flags & ISLASTCN) && (ndp->ni_flag & NAMEI_TRAILINGSLASH) && !(cnp->cn_flags & WILLBEDIR)) {
return ENOENT;
}
return 0;
}
/*
* Modifies only ni_vp. Always returns with ni_vp still valid (iocount held).
*/
int
lookup_traverse_mountpoints(struct nameidata *ndp, struct componentname *cnp, vnode_t dp,
int vbusyflags, vfs_context_t ctx)
{
mount_t mp;
vnode_t tdp;
int error = 0;
uthread_t uth;
uint32_t depth = 0;
int dont_cache_mp = 0;
vnode_t mounted_on_dp;
int current_mount_generation = 0;
mounted_on_dp = dp;
current_mount_generation = mount_generation;
while ((dp->v_type == VDIR) && dp->v_mountedhere &&
((cnp->cn_flags & NOCROSSMOUNT) == 0)) {
#if CONFIG_TRIGGERS
/*
* For a trigger vnode, call its resolver when crossing its mount (if requested)
*/
if (dp->v_resolve) {
(void) vnode_trigger_resolve(dp, ndp, ctx);
}
#endif
vnode_lock(dp);
if ((dp->v_type == VDIR) && (mp = dp->v_mountedhere)) {
mp->mnt_crossref++;
vnode_unlock(dp);
if (vfs_busy(mp, vbusyflags)) {
mount_dropcrossref(mp, dp, 0);
if (vbusyflags == LK_NOWAIT) {
error = ENOENT;
goto out;
}
continue;
}
/*
* XXX - if this is the last component of the
* pathname, and it's either not a lookup operation
* or the NOTRIGGER flag is set for the operation,
* set a uthread flag to let VFS_ROOT() for autofs
* know it shouldn't trigger a mount.
*/
uth = (struct uthread *)get_bsdthread_info(current_thread());
if ((cnp->cn_flags & ISLASTCN) &&
(cnp->cn_nameiop != LOOKUP ||
(cnp->cn_flags & NOTRIGGER))) {
uth->uu_notrigger = 1;
dont_cache_mp = 1;
}
error = VFS_ROOT(mp, &tdp, ctx);
/* XXX - clear the uthread flag */
uth->uu_notrigger = 0;
mount_dropcrossref(mp, dp, 0);
vfs_unbusy(mp);
if (error) {
goto out;
}
vnode_put(dp);
ndp->ni_vp = dp = tdp;
depth++;
#if CONFIG_TRIGGERS
/*
* Check if root dir is a trigger vnode
*/
if (dp->v_resolve) {
error = vnode_trigger_resolve(dp, ndp, ctx);
if (error) {
goto out;
}
}
#endif
} else {
vnode_unlock(dp);
break;
}
}
if (depth && !dont_cache_mp) {
mp = mounted_on_dp->v_mountedhere;
if (mp) {
mount_lock_spin(mp);
mp->mnt_realrootvp_vid = dp->v_id;
mp->mnt_realrootvp = dp;
mp->mnt_generation = current_mount_generation;
mount_unlock(mp);
}
}
return 0;
out:
return error;
}
/*
* forkproc
*
* Description: Create a new process structure, given a parent process
* structure.
*
* Parameters: parent_proc The parent process
*
* Returns: !NULL The new process structure
* NULL Error (insufficient free memory)
*
* Note: When successful, the newly created process structure is
* partially initialized; if a caller needs to deconstruct the
* returned structure, they must call forkproc_free() to do so.
*/
proc_t
forkproc(proc_t parent_proc)
{
proc_t child_proc; /* Our new process */
static int nextpid = 0, pidwrap = 0, nextpidversion = 0;
int error = 0;
struct session *sessp;
uthread_t parent_uthread = (uthread_t)get_bsdthread_info(current_thread());
MALLOC_ZONE(child_proc, proc_t , sizeof *child_proc, M_PROC, M_WAITOK);
if (child_proc == NULL) {
printf("forkproc: M_PROC zone exhausted\n");
goto bad;
}
/* zero it out as we need to insert in hash */
bzero(child_proc, sizeof *child_proc);
MALLOC_ZONE(child_proc->p_stats, struct pstats *,
sizeof *child_proc->p_stats, M_PSTATS, M_WAITOK);
if (child_proc->p_stats == NULL) {
printf("forkproc: M_SUBPROC zone exhausted (p_stats)\n");
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
MALLOC_ZONE(child_proc->p_sigacts, struct sigacts *,
sizeof *child_proc->p_sigacts, M_SIGACTS, M_WAITOK);
if (child_proc->p_sigacts == NULL) {
printf("forkproc: M_SUBPROC zone exhausted (p_sigacts)\n");
FREE_ZONE(child_proc->p_stats, sizeof *child_proc->p_stats, M_PSTATS);
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
/* allocate a callout for use by interval timers */
child_proc->p_rcall = thread_call_allocate((thread_call_func_t)realitexpire, child_proc);
if (child_proc->p_rcall == NULL) {
FREE_ZONE(child_proc->p_sigacts, sizeof *child_proc->p_sigacts, M_SIGACTS);
FREE_ZONE(child_proc->p_stats, sizeof *child_proc->p_stats, M_PSTATS);
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
/*
* Find an unused PID.
*/
proc_list_lock();
nextpid++;
retry:
/*
* If the process ID prototype has wrapped around,
* restart somewhat above 0, as the low-numbered procs
* tend to include daemons that don't exit.
*/
if (nextpid >= PID_MAX) {
nextpid = 100;
pidwrap = 1;
}
if (pidwrap != 0) {
/* if the pid stays in hash both for zombie and runniing state */
if (pfind_locked(nextpid) != PROC_NULL) {
nextpid++;
goto retry;
}
if (pgfind_internal(nextpid) != PGRP_NULL) {
nextpid++;
goto retry;
}
if (session_find_internal(nextpid) != SESSION_NULL) {
nextpid++;
goto retry;
}
}
nprocs++;
child_proc->p_pid = nextpid;
child_proc->p_idversion = nextpidversion++;
#if 1
if (child_proc->p_pid != 0) {
if (pfind_locked(child_proc->p_pid) != PROC_NULL)
panic("proc in the list already\n");
}
#endif
/* Insert in the hash */
child_proc->p_listflag |= (P_LIST_INHASH | P_LIST_INCREATE);
LIST_INSERT_HEAD(PIDHASH(child_proc->p_pid), child_proc, p_hash);
proc_list_unlock();
/*
* We've identified the PID we are going to use; initialize the new
* process structure.
*/
child_proc->p_stat = SIDL;
child_proc->p_pgrpid = PGRPID_DEAD;
/*
* The zero'ing of the proc was at the allocation time due to need
* for insertion to hash. Copy the section that is to be copied
* directly from the parent.
*/
bcopy(&parent_proc->p_startcopy, &child_proc->p_startcopy,
(unsigned) ((caddr_t)&child_proc->p_endcopy - (caddr_t)&child_proc->p_startcopy));
/*
* Some flags are inherited from the parent.
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
* The p_stats and p_sigacts substructs are set in vm_fork.
*/
child_proc->p_flag = (parent_proc->p_flag & (P_LP64 | P_TRANSLATED | P_AFFINITY));
if (parent_proc->p_flag & P_PROFIL)
startprofclock(child_proc);
/*
* Note that if the current thread has an assumed identity, this
* credential will be granted to the new process.
*/
child_proc->p_ucred = kauth_cred_get_with_ref();
#ifdef CONFIG_EMBEDDED
lck_mtx_init(&child_proc->p_mlock, proc_lck_grp, proc_lck_attr);
lck_mtx_init(&child_proc->p_fdmlock, proc_lck_grp, proc_lck_attr);
#if CONFIG_DTRACE
lck_mtx_init(&child_proc->p_dtrace_sprlock, proc_lck_grp, proc_lck_attr);
#endif
lck_spin_init(&child_proc->p_slock, proc_lck_grp, proc_lck_attr);
#else /* !CONFIG_EMBEDDED */
lck_mtx_init(&child_proc->p_mlock, proc_mlock_grp, proc_lck_attr);
lck_mtx_init(&child_proc->p_fdmlock, proc_fdmlock_grp, proc_lck_attr);
#if CONFIG_DTRACE
lck_mtx_init(&child_proc->p_dtrace_sprlock, proc_lck_grp, proc_lck_attr);
#endif
lck_spin_init(&child_proc->p_slock, proc_slock_grp, proc_lck_attr);
#endif /* !CONFIG_EMBEDDED */
klist_init(&child_proc->p_klist);
if (child_proc->p_textvp != NULLVP) {
/* bump references to the text vnode */
/* Need to hold iocount across the ref call */
if (vnode_getwithref(child_proc->p_textvp) == 0) {
error = vnode_ref(child_proc->p_textvp);
vnode_put(child_proc->p_textvp);
if (error != 0)
child_proc->p_textvp = NULLVP;
}
}
/*
* Copy the parents per process open file table to the child; if
* there is a per-thread current working directory, set the childs
* per-process current working directory to that instead of the
* parents.
*
* XXX may fail to copy descriptors to child
*/
child_proc->p_fd = fdcopy(parent_proc, parent_uthread->uu_cdir);
#if SYSV_SHM
if (parent_proc->vm_shm) {
/* XXX may fail to attach shm to child */
(void)shmfork(parent_proc, child_proc);
}
#endif
/*
* inherit the limit structure to child
*/
proc_limitfork(parent_proc, child_proc);
if (child_proc->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
uint64_t rlim_cur = child_proc->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur;
child_proc->p_rlim_cpu.tv_sec = (rlim_cur > __INT_MAX__) ? __INT_MAX__ : rlim_cur;
}
/* Intialize new process stats, including start time */
/* <rdar://6640543> non-zeroed portion contains garbage AFAICT */
bzero(&child_proc->p_stats->pstat_startzero,
(unsigned) ((caddr_t)&child_proc->p_stats->pstat_endzero -
(caddr_t)&child_proc->p_stats->pstat_startzero));
bzero(&child_proc->p_stats->user_p_prof, sizeof(struct user_uprof));
microtime(&child_proc->p_start);
child_proc->p_stats->p_start = child_proc->p_start; /* for compat */
if (parent_proc->p_sigacts != NULL)
(void)memcpy(child_proc->p_sigacts,
parent_proc->p_sigacts, sizeof *child_proc->p_sigacts);
else
(void)memset(child_proc->p_sigacts, 0, sizeof *child_proc->p_sigacts);
sessp = proc_session(parent_proc);
if (sessp->s_ttyvp != NULL && parent_proc->p_flag & P_CONTROLT)
OSBitOrAtomic(P_CONTROLT, &child_proc->p_flag);
session_rele(sessp);
/*
* block all signals to reach the process.
* no transition race should be occuring with the child yet,
* but indicate that the process is in (the creation) transition.
*/
proc_signalstart(child_proc, 0);
proc_transstart(child_proc, 0);
child_proc->p_pcaction = (parent_proc->p_pcaction) & P_PCMAX;
TAILQ_INIT(&child_proc->p_uthlist);
TAILQ_INIT(&child_proc->p_aio_activeq);
TAILQ_INIT(&child_proc->p_aio_doneq);
/* Inherit the parent flags for code sign */
child_proc->p_csflags = parent_proc->p_csflags;
/*
* All processes have work queue locks; cleaned up by
* reap_child_locked()
*/
workqueue_init_lock(child_proc);
/*
* Copy work queue information
*
* Note: This should probably only happen in the case where we are
* creating a child that is a copy of the parent; since this
* routine is called in the non-duplication case of vfork()
* or posix_spawn(), then this information should likely not
* be duplicated.
*
* <rdar://6640553> Work queue pointers that no longer point to code
*/
child_proc->p_wqthread = parent_proc->p_wqthread;
child_proc->p_threadstart = parent_proc->p_threadstart;
child_proc->p_pthsize = parent_proc->p_pthsize;
child_proc->p_targconc = parent_proc->p_targconc;
if ((parent_proc->p_lflag & P_LREGISTER) != 0) {
child_proc->p_lflag |= P_LREGISTER;
}
child_proc->p_dispatchqueue_offset = parent_proc->p_dispatchqueue_offset;
#if PSYNCH
pth_proc_hashinit(child_proc);
#endif /* PSYNCH */
#if CONFIG_LCTX
child_proc->p_lctx = NULL;
/* Add new process to login context (if any). */
if (parent_proc->p_lctx != NULL) {
/*
* <rdar://6640564> This should probably be delayed in the
* vfork() or posix_spawn() cases.
*/
LCTX_LOCK(parent_proc->p_lctx);
enterlctx(child_proc, parent_proc->p_lctx, 0);
}
#endif
bad:
return(child_proc);
}
void
mach_call_munger64(x86_saved_state_t *state)
{
int call_number;
int argc;
mach_call_t mach_call;
struct mach_call_args args = { 0, 0, 0, 0, 0, 0, 0, 0, 0 };
x86_saved_state64_t *regs;
#if PROC_REF_DEBUG
struct uthread *ut = get_bsdthread_info(current_thread());
uthread_reset_proc_refcount(ut);
#endif
assert(is_saved_state64(state));
regs = saved_state64(state);
call_number = (int)(regs->rax & SYSCALL_NUMBER_MASK);
DEBUG_KPRINT_SYSCALL_MACH(
"mach_call_munger64: code=%d(%s)\n",
call_number, mach_syscall_name_table[call_number]);
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
MACHDBG_CODE(DBG_MACH_EXCP_SC,(call_number)) | DBG_FUNC_START,
regs->rdi, regs->rsi, regs->rdx, regs->r10, 0);
if (call_number < 0 || call_number >= mach_trap_count) {
i386_exception(EXC_SYSCALL, regs->rax, 1);
/* NOTREACHED */
}
mach_call = (mach_call_t)mach_trap_table[call_number].mach_trap_function;
if (mach_call == (mach_call_t)kern_invalid) {
i386_exception(EXC_SYSCALL, regs->rax, 1);
/* NOTREACHED */
}
argc = mach_trap_table[call_number].mach_trap_arg_count;
if (argc) {
int args_in_regs = MIN(6, argc);
memcpy(&args.arg1, ®s->rdi, args_in_regs * sizeof(syscall_arg_t));
if (argc > 6) {
int copyin_count;
assert(argc <= 9);
copyin_count = (argc - 6) * (int)sizeof(syscall_arg_t);
if (copyin((user_addr_t)(regs->isf.rsp + sizeof(user_addr_t)), (char *)&args.arg7, copyin_count)) {
regs->rax = KERN_INVALID_ARGUMENT;
thread_exception_return();
/* NOTREACHED */
}
}
}
#ifdef MACH_BSD
mach_kauth_cred_uthread_update();
#endif
regs->rax = (uint64_t)mach_call((void *)&args);
DEBUG_KPRINT_SYSCALL_MACH( "mach_call_munger64: retval=0x%llx\n", regs->rax);
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
MACHDBG_CODE(DBG_MACH_EXCP_SC,(call_number)) | DBG_FUNC_END,
regs->rax, 0, 0, 0, 0);
throttle_lowpri_io(1);
#if PROC_REF_DEBUG
if (__improbable(uthread_get_proc_refcount(ut) != 0)) {
panic("system call returned with uu_proc_refcount != 0");
}
#endif
thread_exception_return();
/* NOTREACHED */
}
/*
* 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);
}
/*
* Routine: task_for_pid
* Purpose:
* Get the task port for another "process", named by its
* process ID on the same host as "target_task".
*
* Only permitted to privileged processes, or processes
* with the same user ID.
*
* XXX This should be a BSD system call, not a Mach trap!!!
*/
kern_return_t
task_for_pid(
struct task_for_pid_args *args)
{
mach_port_name_t target_tport = args->target_tport;
int pid = args->pid;
user_addr_t task_addr = args->t;
struct uthread *uthread;
proc_t p = PROC_NULL;
task_t t1 = TASK_NULL;
mach_port_name_t tret = MACH_PORT_NULL;
ipc_port_t tfpport;
void * sright;
int error = 0;
AUDIT_MACH_SYSCALL_ENTER(AUE_TASKFORPID);
AUDIT_ARG(pid, pid);
AUDIT_ARG(mach_port1, target_tport);
#if defined(SECURE_KERNEL)
if (0 == pid) {
(void ) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t));
AUDIT_MACH_SYSCALL_EXIT(KERN_FAILURE);
return(KERN_FAILURE);
}
#endif
t1 = port_name_to_task(target_tport);
if (t1 == TASK_NULL) {
(void) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t));
AUDIT_MACH_SYSCALL_EXIT(KERN_FAILURE);
return(KERN_FAILURE);
}
/*
* Delayed binding of thread credential to process credential, if we
* are not running with an explicitly set thread credential.
*/
uthread = get_bsdthread_info(current_thread());
kauth_cred_uthread_update(uthread, current_proc());
p = proc_find(pid);
AUDIT_ARG(process, p);
if (!(task_for_pid_posix_check(p))) {
error = KERN_FAILURE;
goto tfpout;
}
if (p->task != TASK_NULL) {
/* If we aren't root and target's task access port is set... */
if (!kauth_cred_issuser(kauth_cred_get()) &&
p != current_proc() &&
(task_get_task_access_port(p->task, &tfpport) == 0) &&
(tfpport != IPC_PORT_NULL)) {
if (tfpport == IPC_PORT_DEAD) {
error = KERN_PROTECTION_FAILURE;
goto tfpout;
}
/* Call up to the task access server */
error = check_task_access(tfpport, proc_selfpid(), kauth_getgid(), pid);
if (error != MACH_MSG_SUCCESS) {
if (error == MACH_RCV_INTERRUPTED)
error = KERN_ABORTED;
else
error = KERN_FAILURE;
goto tfpout;
}
}
#if CONFIG_MACF
error = mac_proc_check_get_task(kauth_cred_get(), p);
if (error) {
error = KERN_FAILURE;
goto tfpout;
}
#endif
/* Grant task port access */
task_reference(p->task);
sright = (void *) convert_task_to_port(p->task);
tret = ipc_port_copyout_send(
sright,
get_task_ipcspace(current_task()));
}
error = KERN_SUCCESS;
tfpout:
task_deallocate(t1);
AUDIT_ARG(mach_port2, tret);
(void) copyout((char *) &tret, task_addr, sizeof(mach_port_name_t));
if (p != PROC_NULL)
proc_rele(p);
AUDIT_MACH_SYSCALL_EXIT(error);
return(error);
}
int
ptrace(struct proc *p, struct ptrace_args *uap, register_t *retval)
{
struct proc *t = current_proc(); /* target process */
task_t task;
thread_t th_act;
struct uthread *ut;
int tr_sigexc = 0;
int error = 0;
int stopped = 0;
AUDIT_ARG(cmd, uap->req);
AUDIT_ARG(pid, uap->pid);
AUDIT_ARG(addr, uap->addr);
AUDIT_ARG(value, uap->data);
if (uap->req == PT_DENY_ATTACH) {
proc_lock(p);
if (ISSET(p->p_lflag, P_LTRACED)) {
proc_unlock(p);
exit1(p, W_EXITCODE(ENOTSUP, 0), retval);
/* drop funnel before we return */
thread_exception_return();
/* NOTREACHED */
}
SET(p->p_lflag, P_LNOATTACH);
proc_unlock(p);
return(0);
}
if (uap->req == PT_FORCEQUOTA) {
if (is_suser()) {
OSBitOrAtomic(P_FORCEQUOTA, (UInt32 *)&t->p_flag);
return (0);
} else
return (EPERM);
}
/*
* Intercept and deal with "please trace me" request.
*/
if (uap->req == PT_TRACE_ME) {
proc_lock(p);
SET(p->p_lflag, P_LTRACED);
/* Non-attached case, our tracer is our parent. */
p->p_oppid = p->p_ppid;
proc_unlock(p);
return(0);
}
if (uap->req == PT_SIGEXC) {
proc_lock(p);
if (ISSET(p->p_lflag, P_LTRACED)) {
SET(p->p_lflag, P_LSIGEXC);
proc_unlock(p);
return(0);
} else {
proc_unlock(p);
return(EINVAL);
}
}
/*
* We do not want ptrace to do anything with kernel or launchd
*/
if (uap->pid < 2) {
return(EPERM);
}
/*
* Locate victim, and make sure it is traceable.
*/
if ((t = proc_find(uap->pid)) == NULL)
return (ESRCH);
AUDIT_ARG(process, t);
task = t->task;
if (uap->req == PT_ATTACHEXC) {
uap->req = PT_ATTACH;
tr_sigexc = 1;
}
if (uap->req == PT_ATTACH) {
int err;
if ( kauth_authorize_process(proc_ucred(p), KAUTH_PROCESS_CANTRACE,
t, (uintptr_t)&err, 0, 0) == 0 ) {
/* it's OK to attach */
proc_lock(t);
SET(t->p_lflag, P_LTRACED);
if (tr_sigexc)
SET(t->p_lflag, P_LSIGEXC);
t->p_oppid = t->p_ppid;
proc_unlock(t);
if (t->p_pptr != p)
proc_reparentlocked(t, p, 1, 0);
proc_lock(t);
if (get_task_userstop(task) > 0 ) {
stopped = 1;
}
t->p_xstat = 0;
proc_unlock(t);
psignal(t, SIGSTOP);
/*
* If the process was stopped, wake up and run through
* issignal() again to properly connect to the tracing
* process.
*/
if (stopped)
task_resume(task);
error = 0;
goto out;
}
else {
/* not allowed to attach, proper error code returned by kauth_authorize_process */
if (ISSET(t->p_lflag, P_LNOATTACH)) {
psignal(p, SIGSEGV);
}
error = err;
goto out;
}
}
/*
* You can't do what you want to the process if:
* (1) It's not being traced at all,
*/
proc_lock(t);
if (!ISSET(t->p_lflag, P_LTRACED)) {
proc_unlock(t);
error = EPERM;
goto out;
}
/*
* (2) it's not being traced by _you_, or
*/
if (t->p_pptr != p) {
proc_unlock(t);
error = EBUSY;
goto out;
}
/*
* (3) it's not currently stopped.
*/
if (t->p_stat != SSTOP) {
proc_unlock(t);
error = EBUSY;
goto out;
}
/*
* Mach version of ptrace executes request directly here,
* thus simplifying the interaction of ptrace and signals.
*/
/* proc lock is held here */
switch (uap->req) {
case PT_DETACH:
if (t->p_oppid != t->p_ppid) {
struct proc *pp;
proc_unlock(t);
pp = proc_find(t->p_oppid);
proc_reparentlocked(t, pp ? pp : initproc, 1, 0);
if (pp != PROC_NULL)
proc_rele(pp);
proc_lock(t);
}
t->p_oppid = 0;
CLR(t->p_lflag, P_LTRACED);
CLR(t->p_lflag, P_LSIGEXC);
proc_unlock(t);
goto resume;
case PT_KILL:
/*
* Tell child process to kill itself after it
* is resumed by adding NSIG to p_cursig. [see issig]
*/
proc_unlock(t);
psignal(t, SIGKILL);
goto resume;
case PT_STEP: /* single step the child */
case PT_CONTINUE: /* continue the child */
proc_unlock(t);
th_act = (thread_t)get_firstthread(task);
if (th_act == THREAD_NULL) {
error = EINVAL;
goto out;
}
if (uap->addr != (user_addr_t)1) {
#if defined(ppc)
#define ALIGNED(addr,size) (((unsigned)(addr)&((size)-1))==0)
if (!ALIGNED((int)uap->addr, sizeof(int)))
return (ERESTART);
#undef ALIGNED
#endif
thread_setentrypoint(th_act, uap->addr);
}
if ((unsigned)uap->data >= NSIG) {
error = EINVAL;
goto out;
}
if (uap->data != 0) {
psignal(t, uap->data);
}
if (uap->req == PT_STEP) {
/*
* set trace bit
*/
if (thread_setsinglestep(th_act, 1) != KERN_SUCCESS) {
error = ENOTSUP;
goto out;
}
} else {
/*
* clear trace bit if on
*/
if (thread_setsinglestep(th_act, 0) != KERN_SUCCESS) {
error = ENOTSUP;
goto out;
}
}
resume:
proc_lock(t);
t->p_xstat = uap->data;
t->p_stat = SRUN;
if (t->sigwait) {
wakeup((caddr_t)&(t->sigwait));
proc_unlock(t);
if ((t->p_lflag & P_LSIGEXC) == 0) {
task_resume(task);
}
} else
proc_unlock(t);
break;
case PT_THUPDATE: {
proc_unlock(t);
if ((unsigned)uap->data >= NSIG) {
error = EINVAL;
goto out;
}
th_act = port_name_to_thread(CAST_DOWN(mach_port_name_t, uap->addr));
if (th_act == THREAD_NULL)
return (ESRCH);
ut = (uthread_t)get_bsdthread_info(th_act);
if (uap->data)
ut->uu_siglist |= sigmask(uap->data);
proc_lock(t);
t->p_xstat = uap->data;
t->p_stat = SRUN;
proc_unlock(t);
thread_deallocate(th_act);
error = 0;
}
break;
default:
proc_unlock(t);
error = EINVAL;
goto out;
}
error = 0;
out:
proc_rele(t);
return(error);
}
kern_return_t
task_name_for_pid(
struct task_name_for_pid_args *args)
{
mach_port_name_t target_tport = args->target_tport;
int pid = args->pid;
user_addr_t task_addr = args->t;
struct uthread *uthread;
proc_t p = PROC_NULL;
task_t t1;
mach_port_name_t tret;
void * sright;
int error = 0, refheld = 0;
kauth_cred_t target_cred;
AUDIT_MACH_SYSCALL_ENTER(AUE_TASKNAMEFORPID);
AUDIT_ARG(pid, pid);
AUDIT_ARG(mach_port1, target_tport);
t1 = port_name_to_task(target_tport);
if (t1 == TASK_NULL) {
(void) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t));
AUDIT_MACH_SYSCALL_EXIT(KERN_FAILURE);
return(KERN_FAILURE);
}
/*
* Delayed binding of thread credential to process credential, if we
* are not running with an explicitly set thread credential.
*/
uthread = get_bsdthread_info(current_thread());
kauth_cred_uthread_update(uthread, current_proc());
p = proc_find(pid);
AUDIT_ARG(process, p);
if (p != PROC_NULL) {
target_cred = kauth_cred_proc_ref(p);
refheld = 1;
if ((p->p_stat != SZOMB)
&& ((current_proc() == p)
|| kauth_cred_issuser(kauth_cred_get())
|| ((kauth_cred_getuid(target_cred) == kauth_cred_getuid(kauth_cred_get())) &&
((target_cred->cr_ruid == kauth_cred_get()->cr_ruid))))) {
if (p->task != TASK_NULL) {
task_reference(p->task);
#if CONFIG_MACF
error = mac_proc_check_get_task_name(kauth_cred_get(), p);
if (error) {
task_deallocate(p->task);
goto noperm;
}
#endif
sright = (void *)convert_task_name_to_port(p->task);
tret = ipc_port_copyout_send(sright,
get_task_ipcspace(current_task()));
} else
tret = MACH_PORT_NULL;
AUDIT_ARG(mach_port2, tret);
(void) copyout((char *)&tret, task_addr, sizeof(mach_port_name_t));
task_deallocate(t1);
error = KERN_SUCCESS;
goto tnfpout;
}
}
#if CONFIG_MACF
noperm:
#endif
task_deallocate(t1);
tret = MACH_PORT_NULL;
(void) copyout((char *) &tret, task_addr, sizeof(mach_port_name_t));
error = KERN_FAILURE;
tnfpout:
if (refheld != 0)
kauth_cred_unref(&target_cred);
if (p != PROC_NULL)
proc_rele(p);
AUDIT_MACH_SYSCALL_EXIT(error);
return(error);
}
