/*
* Process debugging system call.
*/
int
sys_ptrace(struct lwp *l, const struct sys_ptrace_args *uap, register_t *retval)
{
/* {
syscallarg(int) req;
syscallarg(pid_t) pid;
syscallarg(void *) addr;
syscallarg(int) data;
} */
struct proc *p = l->l_proc;
struct lwp *lt;
struct proc *t; /* target process */
struct uio uio;
struct iovec iov;
struct ptrace_io_desc piod;
struct ptrace_lwpinfo pl;
struct vmspace *vm;
int error, write, tmp, req, pheld;
int signo;
ksiginfo_t ksi;
#ifdef COREDUMP
char *path;
#endif
error = 0;
req = SCARG(uap, req);
/*
* If attaching or detaching, we need to get a write hold on the
* proclist lock so that we can re-parent the target process.
*/
mutex_enter(proc_lock);
/* "A foolish consistency..." XXX */
if (req == PT_TRACE_ME) {
t = p;
mutex_enter(t->p_lock);
} else {
/* Find the process we're supposed to be operating on. */
if ((t = p_find(SCARG(uap, pid), PFIND_LOCKED)) == NULL) {
mutex_exit(proc_lock);
return (ESRCH);
}
/* XXX-elad */
mutex_enter(t->p_lock);
error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
t, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
if (error) {
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
return (ESRCH);
}
}
/*
* Grab a reference on the process to prevent it from execing or
* exiting.
*/
if (!rw_tryenter(&t->p_reflock, RW_READER)) {
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
return EBUSY;
}
/* Make sure we can operate on it. */
switch (req) {
case PT_TRACE_ME:
/* Saying that you're being traced is always legal. */
break;
case PT_ATTACH:
/*
* You can't attach to a process if:
* (1) it's the process that's doing the attaching,
*/
if (t->p_pid == p->p_pid) {
error = EINVAL;
break;
}
/*
* (2) it's a system process
*/
if (t->p_flag & PK_SYSTEM) {
error = EPERM;
break;
}
/*
* (3) it's already being traced, or
*/
if (ISSET(t->p_slflag, PSL_TRACED)) {
error = EBUSY;
break;
}
/*
* (4) the tracer is chrooted, and its root directory is
* not at or above the root directory of the tracee
*/
mutex_exit(t->p_lock); /* XXXSMP */
tmp = proc_isunder(t, l);
mutex_enter(t->p_lock); /* XXXSMP */
if (!tmp) {
error = EPERM;
break;
}
break;
case PT_READ_I:
case PT_READ_D:
case PT_WRITE_I:
case PT_WRITE_D:
case PT_IO:
#ifdef PT_GETREGS
case PT_GETREGS:
#endif
#ifdef PT_SETREGS
case PT_SETREGS:
#endif
#ifdef PT_GETFPREGS
case PT_GETFPREGS:
#endif
#ifdef PT_SETFPREGS
case PT_SETFPREGS:
#endif
#ifdef __HAVE_PTRACE_MACHDEP
PTRACE_MACHDEP_REQUEST_CASES
#endif
/*
* You can't read/write the memory or registers of a process
* if the tracer is chrooted, and its root directory is not at
* or above the root directory of the tracee.
*/
mutex_exit(t->p_lock); /* XXXSMP */
tmp = proc_isunder(t, l);
mutex_enter(t->p_lock); /* XXXSMP */
if (!tmp) {
error = EPERM;
break;
}
/*FALLTHROUGH*/
case PT_CONTINUE:
case PT_KILL:
case PT_DETACH:
case PT_LWPINFO:
case PT_SYSCALL:
#ifdef COREDUMP
case PT_DUMPCORE:
#endif
#ifdef PT_STEP
case PT_STEP:
#endif
/*
* You can't do what you want to the process if:
* (1) It's not being traced at all,
*/
if (!ISSET(t->p_slflag, PSL_TRACED)) {
error = EPERM;
break;
}
/*
* (2) it's being traced by procfs (which has
* different signal delivery semantics),
*/
if (ISSET(t->p_slflag, PSL_FSTRACE)) {
uprintf("file system traced\n");
error = EBUSY;
break;
}
/*
* (3) it's not being traced by _you_, or
*/
if (t->p_pptr != p) {
uprintf("parent %d != %d\n", t->p_pptr->p_pid, p->p_pid);
error = EBUSY;
break;
}
/*
* (4) it's not currently stopped.
*/
if (t->p_stat != SSTOP || !t->p_waited /* XXXSMP */) {
uprintf("stat %d flag %d\n", t->p_stat,
!t->p_waited);
error = EBUSY;
break;
}
break;
default: /* It was not a legal request. */
error = EINVAL;
break;
}
if (error == 0)
error = kauth_authorize_process(l->l_cred,
KAUTH_PROCESS_PTRACE, t, KAUTH_ARG(req),
NULL, NULL);
if (error != 0) {
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
rw_exit(&t->p_reflock);
return error;
}
/* Do single-step fixup if needed. */
FIX_SSTEP(t);
/*
* XXX NJWLWP
*
* The entire ptrace interface needs work to be useful to a
* process with multiple LWPs. For the moment, we'll kluge
* this; memory access will be fine, but register access will
* be weird.
*/
lt = LIST_FIRST(&t->p_lwps);
KASSERT(lt != NULL);
lwp_addref(lt);
/*
* Which locks do we need held? XXX Ugly.
*/
switch (req) {
#ifdef PT_STEP
case PT_STEP:
#endif
case PT_CONTINUE:
case PT_DETACH:
case PT_KILL:
case PT_SYSCALL:
case PT_ATTACH:
case PT_TRACE_ME:
pheld = 1;
break;
default:
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
pheld = 0;
break;
}
/* Now do the operation. */
write = 0;
*retval = 0;
tmp = 0;
switch (req) {
case PT_TRACE_ME:
/* Just set the trace flag. */
SET(t->p_slflag, PSL_TRACED);
t->p_opptr = t->p_pptr;
break;
case PT_WRITE_I: /* XXX no separate I and D spaces */
case PT_WRITE_D:
#if defined(__HAVE_RAS)
/*
* Can't write to a RAS
*/
if (ras_lookup(t, SCARG(uap, addr)) != (void *)-1) {
error = EACCES;
break;
}
#endif
write = 1;
tmp = SCARG(uap, data);
/* FALLTHROUGH */
case PT_READ_I: /* XXX no separate I and D spaces */
case PT_READ_D:
/* write = 0 done above. */
iov.iov_base = (void *)&tmp;
iov.iov_len = sizeof(tmp);
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = (off_t)(unsigned long)SCARG(uap, addr);
uio.uio_resid = sizeof(tmp);
uio.uio_rw = write ? UIO_WRITE : UIO_READ;
UIO_SETUP_SYSSPACE(&uio);
error = process_domem(l, lt, &uio);
if (!write)
*retval = tmp;
break;
case PT_IO:
error = copyin(SCARG(uap, addr), &piod, sizeof(piod));
if (error)
break;
switch (piod.piod_op) {
case PIOD_READ_D:
case PIOD_READ_I:
uio.uio_rw = UIO_READ;
break;
case PIOD_WRITE_D:
case PIOD_WRITE_I:
/*
* Can't write to a RAS
*/
if (ras_lookup(t, SCARG(uap, addr)) != (void *)-1) {
return (EACCES);
}
uio.uio_rw = UIO_WRITE;
break;
default:
error = EINVAL;
break;
}
if (error)
break;
error = proc_vmspace_getref(l->l_proc, &vm);
if (error)
break;
iov.iov_base = piod.piod_addr;
iov.iov_len = piod.piod_len;
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = (off_t)(unsigned long)piod.piod_offs;
uio.uio_resid = piod.piod_len;
uio.uio_vmspace = vm;
error = process_domem(l, lt, &uio);
piod.piod_len -= uio.uio_resid;
(void) copyout(&piod, SCARG(uap, addr), sizeof(piod));
uvmspace_free(vm);
break;
#ifdef COREDUMP
case PT_DUMPCORE:
if ((path = SCARG(uap, addr)) != NULL) {
char *dst;
int len = SCARG(uap, data);
if (len < 0 || len >= MAXPATHLEN) {
error = EINVAL;
break;
}
dst = malloc(len + 1, M_TEMP, M_WAITOK);
if ((error = copyin(path, dst, len)) != 0) {
free(dst, M_TEMP);
break;
}
path = dst;
path[len] = '\0';
}
error = coredump(lt, path);
if (path)
free(path, M_TEMP);
break;
#endif
#ifdef PT_STEP
case PT_STEP:
/*
* From the 4.4BSD PRM:
* "Execution continues as in request PT_CONTINUE; however
* as soon as possible after execution of at least one
* instruction, execution stops again. [ ... ]"
*/
#endif
case PT_CONTINUE:
case PT_SYSCALL:
case PT_DETACH:
if (req == PT_SYSCALL) {
if (!ISSET(t->p_slflag, PSL_SYSCALL)) {
SET(t->p_slflag, PSL_SYSCALL);
#ifdef __HAVE_SYSCALL_INTERN
(*t->p_emul->e_syscall_intern)(t);
#endif
}
} else {
if (ISSET(t->p_slflag, PSL_SYSCALL)) {
CLR(t->p_slflag, PSL_SYSCALL);
#ifdef __HAVE_SYSCALL_INTERN
(*t->p_emul->e_syscall_intern)(t);
#endif
}
}
p->p_trace_enabled = trace_is_enabled(p);
/*
* From the 4.4BSD PRM:
* "The data argument is taken as a signal number and the
* child's execution continues at location addr as if it
* incurred that signal. Normally the signal number will
* be either 0 to indicate that the signal that caused the
* stop should be ignored, or that value fetched out of
* the process's image indicating which signal caused
* the stop. If addr is (int *)1 then execution continues
* from where it stopped."
*/
/* Check that the data is a valid signal number or zero. */
if (SCARG(uap, data) < 0 || SCARG(uap, data) >= NSIG) {
error = EINVAL;
break;
}
uvm_lwp_hold(lt);
/* If the address parameter is not (int *)1, set the pc. */
if ((int *)SCARG(uap, addr) != (int *)1)
if ((error = process_set_pc(lt, SCARG(uap, addr))) != 0) {
uvm_lwp_rele(lt);
break;
}
#ifdef PT_STEP
/*
* Arrange for a single-step, if that's requested and possible.
*/
error = process_sstep(lt, req == PT_STEP);
if (error) {
uvm_lwp_rele(lt);
break;
}
#endif
uvm_lwp_rele(lt);
if (req == PT_DETACH) {
CLR(t->p_slflag, PSL_TRACED|PSL_FSTRACE|PSL_SYSCALL);
/* give process back to original parent or init */
if (t->p_opptr != t->p_pptr) {
struct proc *pp = t->p_opptr;
proc_reparent(t, pp ? pp : initproc);
}
/* not being traced any more */
t->p_opptr = NULL;
}
signo = SCARG(uap, data);
sendsig:
/* Finally, deliver the requested signal (or none). */
if (t->p_stat == SSTOP) {
/*
* Unstop the process. If it needs to take a
* signal, make all efforts to ensure that at
* an LWP runs to see it.
*/
t->p_xstat = signo;
proc_unstop(t);
} else if (signo != 0) {
KSI_INIT_EMPTY(&ksi);
ksi.ksi_signo = signo;
kpsignal2(t, &ksi);
}
break;
case PT_KILL:
/* just send the process a KILL signal. */
signo = SIGKILL;
goto sendsig; /* in PT_CONTINUE, above. */
case PT_ATTACH:
/*
* Go ahead and set the trace flag.
* Save the old parent (it's reset in
* _DETACH, and also in kern_exit.c:wait4()
* Reparent the process so that the tracing
* proc gets to see all the action.
* Stop the target.
*/
t->p_opptr = t->p_pptr;
if (t->p_pptr != p) {
struct proc *parent = t->p_pptr;
if (parent->p_lock < t->p_lock) {
if (!mutex_tryenter(parent->p_lock)) {
mutex_exit(t->p_lock);
mutex_enter(parent->p_lock);
}
} else if (parent->p_lock > t->p_lock) {
mutex_enter(parent->p_lock);
}
parent->p_slflag |= PSL_CHTRACED;
proc_reparent(t, p);
if (parent->p_lock != t->p_lock)
mutex_exit(parent->p_lock);
}
SET(t->p_slflag, PSL_TRACED);
signo = SIGSTOP;
goto sendsig;
case PT_LWPINFO:
if (SCARG(uap, data) != sizeof(pl)) {
error = EINVAL;
break;
}
error = copyin(SCARG(uap, addr), &pl, sizeof(pl));
if (error)
break;
tmp = pl.pl_lwpid;
lwp_delref(lt);
mutex_enter(t->p_lock);
if (tmp == 0)
lt = LIST_FIRST(&t->p_lwps);
else {
lt = lwp_find(t, tmp);
if (lt == NULL) {
mutex_exit(t->p_lock);
error = ESRCH;
break;
}
lt = LIST_NEXT(lt, l_sibling);
}
while (lt != NULL && lt->l_stat == LSZOMB)
lt = LIST_NEXT(lt, l_sibling);
pl.pl_lwpid = 0;
pl.pl_event = 0;
if (lt) {
lwp_addref(lt);
pl.pl_lwpid = lt->l_lid;
if (lt->l_lid == t->p_sigctx.ps_lwp)
pl.pl_event = PL_EVENT_SIGNAL;
}
mutex_exit(t->p_lock);
error = copyout(&pl, SCARG(uap, addr), sizeof(pl));
break;
#ifdef PT_SETREGS
case PT_SETREGS:
write = 1;
#endif
#ifdef PT_GETREGS
case PT_GETREGS:
/* write = 0 done above. */
#endif
#if defined(PT_SETREGS) || defined(PT_GETREGS)
tmp = SCARG(uap, data);
if (tmp != 0 && t->p_nlwps > 1) {
lwp_delref(lt);
mutex_enter(t->p_lock);
lt = lwp_find(t, tmp);
if (lt == NULL) {
mutex_exit(t->p_lock);
error = ESRCH;
break;
}
lwp_addref(lt);
mutex_exit(t->p_lock);
}
if (!process_validregs(lt))
error = EINVAL;
else {
error = proc_vmspace_getref(l->l_proc, &vm);
if (error)
break;
iov.iov_base = SCARG(uap, addr);
iov.iov_len = sizeof(struct reg);
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = 0;
uio.uio_resid = sizeof(struct reg);
uio.uio_rw = write ? UIO_WRITE : UIO_READ;
uio.uio_vmspace = vm;
error = process_doregs(l, lt, &uio);
uvmspace_free(vm);
}
break;
#endif
#ifdef PT_SETFPREGS
case PT_SETFPREGS:
write = 1;
#endif
#ifdef PT_GETFPREGS
case PT_GETFPREGS:
/* write = 0 done above. */
#endif
#if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
tmp = SCARG(uap, data);
if (tmp != 0 && t->p_nlwps > 1) {
lwp_delref(lt);
mutex_enter(t->p_lock);
lt = lwp_find(t, tmp);
if (lt == NULL) {
mutex_exit(t->p_lock);
error = ESRCH;
break;
}
lwp_addref(lt);
mutex_exit(t->p_lock);
}
if (!process_validfpregs(lt))
error = EINVAL;
else {
error = proc_vmspace_getref(l->l_proc, &vm);
if (error)
break;
iov.iov_base = SCARG(uap, addr);
iov.iov_len = sizeof(struct fpreg);
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = 0;
uio.uio_resid = sizeof(struct fpreg);
uio.uio_rw = write ? UIO_WRITE : UIO_READ;
uio.uio_vmspace = vm;
error = process_dofpregs(l, lt, &uio);
uvmspace_free(vm);
}
break;
#endif
#ifdef __HAVE_PTRACE_MACHDEP
PTRACE_MACHDEP_REQUEST_CASES
error = ptrace_machdep_dorequest(l, lt,
req, SCARG(uap, addr), SCARG(uap, data));
break;
#endif
}
if (pheld) {
mutex_exit(t->p_lock);
mutex_exit(proc_lock);
}
if (lt != NULL)
lwp_delref(lt);
rw_exit(&t->p_reflock);
return error;
}
/*
* General fork call. Note that another LWP in the process may call exec()
* or exit() while we are forking. It's safe to continue here, because
* neither operation will complete until all LWPs have exited the process.
*/
int
fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize,
void (*func)(void *), void *arg, register_t *retval,
struct proc **rnewprocp)
{
struct proc *p1, *p2, *parent;
struct plimit *p1_lim;
uid_t uid;
struct lwp *l2;
int count;
vaddr_t uaddr;
int tnprocs;
int tracefork;
int error = 0;
p1 = l1->l_proc;
uid = kauth_cred_getuid(l1->l_cred);
tnprocs = atomic_inc_uint_nv(&nprocs);
/*
* Although process entries are dynamically created, we still keep
* a global limit on the maximum number we will create.
*/
if (__predict_false(tnprocs >= maxproc))
error = -1;
else
error = kauth_authorize_process(l1->l_cred,
KAUTH_PROCESS_FORK, p1, KAUTH_ARG(tnprocs), NULL, NULL);
if (error) {
static struct timeval lasttfm;
atomic_dec_uint(&nprocs);
if (ratecheck(&lasttfm, &fork_tfmrate))
tablefull("proc", "increase kern.maxproc or NPROC");
if (forkfsleep)
kpause("forkmx", false, forkfsleep, NULL);
return EAGAIN;
}
/*
* Enforce limits.
*/
count = chgproccnt(uid, 1);
if (__predict_false(count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) {
if (kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_RLIMIT,
p1, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS),
&p1->p_rlimit[RLIMIT_NPROC], KAUTH_ARG(RLIMIT_NPROC)) != 0) {
(void)chgproccnt(uid, -1);
atomic_dec_uint(&nprocs);
if (forkfsleep)
kpause("forkulim", false, forkfsleep, NULL);
return EAGAIN;
}
}
/*
* Allocate virtual address space for the U-area now, while it
* is still easy to abort the fork operation if we're out of
* kernel virtual address space.
*/
uaddr = uvm_uarea_alloc();
if (__predict_false(uaddr == 0)) {
(void)chgproccnt(uid, -1);
atomic_dec_uint(&nprocs);
return ENOMEM;
}
/*
* We are now committed to the fork. From here on, we may
* block on resources, but resource allocation may NOT fail.
*/
/* Allocate new proc. */
p2 = proc_alloc();
/*
* Make a proc table entry for the new process.
* Start by zeroing the section of proc that is zero-initialized,
* then copy the section that is copied directly from the parent.
*/
memset(&p2->p_startzero, 0,
(unsigned) ((char *)&p2->p_endzero - (char *)&p2->p_startzero));
memcpy(&p2->p_startcopy, &p1->p_startcopy,
(unsigned) ((char *)&p2->p_endcopy - (char *)&p2->p_startcopy));
TAILQ_INIT(&p2->p_sigpend.sp_info);
LIST_INIT(&p2->p_lwps);
LIST_INIT(&p2->p_sigwaiters);
/*
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
* Inherit flags we want to keep. The flags related to SIGCHLD
* handling are important in order to keep a consistent behaviour
* for the child after the fork. If we are a 32-bit process, the
* child will be too.
*/
p2->p_flag =
p1->p_flag & (PK_SUGID | PK_NOCLDWAIT | PK_CLDSIGIGN | PK_32);
p2->p_emul = p1->p_emul;
p2->p_execsw = p1->p_execsw;
if (flags & FORK_SYSTEM) {
/*
* Mark it as a system process. Set P_NOCLDWAIT so that
* children are reparented to init(8) when they exit.
* init(8) can easily wait them out for us.
*/
p2->p_flag |= (PK_SYSTEM | PK_NOCLDWAIT);
}
mutex_init(&p2->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
mutex_init(&p2->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
rw_init(&p2->p_reflock);
cv_init(&p2->p_waitcv, "wait");
cv_init(&p2->p_lwpcv, "lwpwait");
/*
* Share a lock between the processes if they are to share signal
* state: we must synchronize access to it.
*/
if (flags & FORK_SHARESIGS) {
p2->p_lock = p1->p_lock;
mutex_obj_hold(p1->p_lock);
} else
p2->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
kauth_proc_fork(p1, p2);
p2->p_raslist = NULL;
#if defined(__HAVE_RAS)
ras_fork(p1, p2);
#endif
/* bump references to the text vnode (for procfs) */
p2->p_textvp = p1->p_textvp;
if (p2->p_textvp)
vref(p2->p_textvp);
if (flags & FORK_SHAREFILES)
fd_share(p2);
else if (flags & FORK_CLEANFILES)
p2->p_fd = fd_init(NULL);
else
p2->p_fd = fd_copy();
/* XXX racy */
p2->p_mqueue_cnt = p1->p_mqueue_cnt;
if (flags & FORK_SHARECWD)
cwdshare(p2);
else
p2->p_cwdi = cwdinit();
/*
* Note: p_limit (rlimit stuff) is copy-on-write, so normally
* we just need increase pl_refcnt.
*/
p1_lim = p1->p_limit;
if (!p1_lim->pl_writeable) {
lim_addref(p1_lim);
p2->p_limit = p1_lim;
} else {
p2->p_limit = lim_copy(p1_lim);
}
if (flags & FORK_PPWAIT) {
/* Mark ourselves as waiting for a child. */
l1->l_pflag |= LP_VFORKWAIT;
p2->p_lflag = PL_PPWAIT;
p2->p_vforklwp = l1;
} else {
p2->p_lflag = 0;
}
p2->p_sflag = 0;
p2->p_slflag = 0;
parent = (flags & FORK_NOWAIT) ? initproc : p1;
p2->p_pptr = parent;
p2->p_ppid = parent->p_pid;
LIST_INIT(&p2->p_children);
p2->p_aio = NULL;
#ifdef KTRACE
/*
* Copy traceflag and tracefile if enabled.
* If not inherited, these were zeroed above.
*/
if (p1->p_traceflag & KTRFAC_INHERIT) {
mutex_enter(&ktrace_lock);
p2->p_traceflag = p1->p_traceflag;
if ((p2->p_tracep = p1->p_tracep) != NULL)
ktradref(p2);
mutex_exit(&ktrace_lock);
}
#endif
/*
* Create signal actions for the child process.
*/
p2->p_sigacts = sigactsinit(p1, flags & FORK_SHARESIGS);
mutex_enter(p1->p_lock);
p2->p_sflag |=
(p1->p_sflag & (PS_STOPFORK | PS_STOPEXEC | PS_NOCLDSTOP));
sched_proc_fork(p1, p2);
mutex_exit(p1->p_lock);
p2->p_stflag = p1->p_stflag;
/*
* p_stats.
* Copy parts of p_stats, and zero out the rest.
*/
p2->p_stats = pstatscopy(p1->p_stats);
/*
* Set up the new process address space.
*/
uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? true : false);
/*
* Finish creating the child process.
* It will return through a different path later.
*/
lwp_create(l1, p2, uaddr, (flags & FORK_PPWAIT) ? LWP_VFORK : 0,
stack, stacksize, (func != NULL) ? func : child_return, arg, &l2,
l1->l_class);
/*
* Inherit l_private from the parent.
* Note that we cannot use lwp_setprivate() here since that
* also sets the CPU TLS register, which is incorrect if the
* process has changed that without letting the kernel know.
*/
l2->l_private = l1->l_private;
/*
* If emulation has a process fork hook, call it now.
*/
if (p2->p_emul->e_proc_fork)
(*p2->p_emul->e_proc_fork)(p2, l1, flags);
/*
* ...and finally, any other random fork hooks that subsystems
* might have registered.
*/
doforkhooks(p2, p1);
SDT_PROBE(proc,,,create, p2, p1, flags, 0, 0);
/*
* It's now safe for the scheduler and other processes to see the
* child process.
*/
mutex_enter(proc_lock);
if (p1->p_session->s_ttyvp != NULL && p1->p_lflag & PL_CONTROLT)
p2->p_lflag |= PL_CONTROLT;
LIST_INSERT_HEAD(&parent->p_children, p2, p_sibling);
p2->p_exitsig = exitsig; /* signal for parent on exit */
/*
* We don't want to tracefork vfork()ed processes because they
* will not receive the SIGTRAP until it is too late.
*/
tracefork = (p1->p_slflag & (PSL_TRACEFORK|PSL_TRACED)) ==
(PSL_TRACEFORK|PSL_TRACED) && (flags && FORK_PPWAIT) == 0;
if (tracefork) {
p2->p_slflag |= PSL_TRACED;
p2->p_opptr = p2->p_pptr;
if (p2->p_pptr != p1->p_pptr) {
struct proc *parent1 = p2->p_pptr;
if (parent1->p_lock < p2->p_lock) {
if (!mutex_tryenter(parent1->p_lock)) {
mutex_exit(p2->p_lock);
mutex_enter(parent1->p_lock);
}
} else if (parent1->p_lock > p2->p_lock) {
mutex_enter(parent1->p_lock);
}
parent1->p_slflag |= PSL_CHTRACED;
proc_reparent(p2, p1->p_pptr);
if (parent1->p_lock != p2->p_lock)
mutex_exit(parent1->p_lock);
}
/*
* Set ptrace status.
*/
p1->p_fpid = p2->p_pid;
p2->p_fpid = p1->p_pid;
}
LIST_INSERT_AFTER(p1, p2, p_pglist);
LIST_INSERT_HEAD(&allproc, p2, p_list);
p2->p_trace_enabled = trace_is_enabled(p2);
#ifdef __HAVE_SYSCALL_INTERN
(*p2->p_emul->e_syscall_intern)(p2);
#endif
/*
* Update stats now that we know the fork was successful.
*/
uvmexp.forks++;
if (flags & FORK_PPWAIT)
uvmexp.forks_ppwait++;
if (flags & FORK_SHAREVM)
uvmexp.forks_sharevm++;
/*
* Pass a pointer to the new process to the caller.
*/
if (rnewprocp != NULL)
*rnewprocp = p2;
if (ktrpoint(KTR_EMUL))
p2->p_traceflag |= KTRFAC_TRC_EMUL;
/*
* Notify any interested parties about the new process.
*/
if (!SLIST_EMPTY(&p1->p_klist)) {
mutex_exit(proc_lock);
KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
mutex_enter(proc_lock);
}
/*
* Make child runnable, set start time, and add to run queue except
* if the parent requested the child to start in SSTOP state.
*/
mutex_enter(p2->p_lock);
/*
* Start profiling.
*/
if ((p2->p_stflag & PST_PROFIL) != 0) {
mutex_spin_enter(&p2->p_stmutex);
startprofclock(p2);
mutex_spin_exit(&p2->p_stmutex);
}
getmicrotime(&p2->p_stats->p_start);
p2->p_acflag = AFORK;
lwp_lock(l2);
KASSERT(p2->p_nrlwps == 1);
if (p2->p_sflag & PS_STOPFORK) {
struct schedstate_percpu *spc = &l2->l_cpu->ci_schedstate;
p2->p_nrlwps = 0;
p2->p_stat = SSTOP;
p2->p_waited = 0;
p1->p_nstopchild++;
l2->l_stat = LSSTOP;
KASSERT(l2->l_wchan == NULL);
lwp_unlock_to(l2, spc->spc_lwplock);
} else {
p2->p_nrlwps = 1;
p2->p_stat = SACTIVE;
l2->l_stat = LSRUN;
sched_enqueue(l2, false);
lwp_unlock(l2);
}
/*
* Return child pid to parent process,
* marking us as parent via retval[1].
*/
if (retval != NULL) {
retval[0] = p2->p_pid;
retval[1] = 0;
}
mutex_exit(p2->p_lock);
/*
* Preserve synchronization semantics of vfork. If waiting for
* child to exec or exit, sleep until it clears LP_VFORKWAIT.
*/
#if 0
while (l1->l_pflag & LP_VFORKWAIT) {
cv_wait(&l1->l_waitcv, proc_lock);
}
#else
while (p2->p_lflag & PL_PPWAIT)
cv_wait(&p1->p_waitcv, proc_lock);
#endif
/*
* Let the parent know that we are tracing its child.
*/
if (tracefork) {
ksiginfo_t ksi;
KSI_INIT_EMPTY(&ksi);
ksi.ksi_signo = SIGTRAP;
ksi.ksi_lid = l1->l_lid;
kpsignal(p1, &ksi, NULL);
}
mutex_exit(proc_lock);
return 0;
}