int
procfs_doprocctl(PFS_FILL_ARGS)
{
int error;
struct namemap *nm;
if (uio == NULL || uio->uio_rw != UIO_WRITE)
return (EOPNOTSUPP);
/*
* Map signal names into signal generation
* or debug control. Unknown commands and/or signals
* return EOPNOTSUPP.
*
* Sending a signal while the process is being debugged
* also has the side effect of letting the target continue
* to run. There is no way to single-step a signal delivery.
*/
error = EOPNOTSUPP;
sbuf_trim(sb);
sbuf_finish(sb);
nm = findname(ctlnames, sbuf_data(sb), sbuf_len(sb));
if (nm) {
printf("procfs: got a %s command\n", sbuf_data(sb));
error = procfs_control(td, p, nm->nm_val);
} else {
nm = findname(signames, sbuf_data(sb), sbuf_len(sb));
if (nm) {
printf("procfs: got a sig%s\n", sbuf_data(sb));
PROC_LOCK(p);
/* This is very broken XXXKSE: */
if (TRACE_WAIT_P(td->td_proc, p)) {
p->p_xstat = nm->nm_val;
#ifdef FIX_SSTEP
/* XXXKSE: */
FIX_SSTEP(FIRST_THREAD_IN_PROC(p));
#endif
/* XXXKSE: */
p->p_flag &= ~P_STOPPED_SIG;
PROC_SLOCK(p);
thread_unsuspend(p);
PROC_SUNLOCK(p);
} else
psignal(p, nm->nm_val);
PROC_UNLOCK(p);
error = 0;
}
}
return (error);
}
static void
poll_idle(void)
{
struct thread *td = curthread;
struct rtprio rtp;
rtp.prio = RTP_PRIO_MAX; /* lowest priority */
rtp.type = RTP_PRIO_IDLE;
PROC_SLOCK(td->td_proc);
rtp_to_pri(&rtp, td);
PROC_SUNLOCK(td->td_proc);
for (;;) {
if (poll_in_idle_loop && poll_handlers > 0) {
idlepoll_sleeping = 0;
ether_poll(poll_each_burst);
thread_lock(td);
mi_switch(SW_VOL, NULL);
thread_unlock(td);
} else {
idlepoll_sleeping = 1;
tsleep(&idlepoll_sleeping, 0, "pollid", hz * 3);
}
}
}
int
kern_thr_exit(struct thread *td)
{
struct proc *p;
p = td->td_proc;
rw_wlock(&tidhash_lock);
PROC_LOCK(p);
if (p->p_numthreads != 1) {
racct_sub(p, RACCT_NTHR, 1);
LIST_REMOVE(td, td_hash);
rw_wunlock(&tidhash_lock);
tdsigcleanup(td);
umtx_thread_exit(td);
PROC_SLOCK(p);
thread_stopped(p);
thread_exit();
/* NOTREACHED */
}
/*
* Ignore attempts to shut down last thread in the proc. This
* will actually call _exit(2) in the usermode trampoline when
* it returns.
*/
PROC_UNLOCK(p);
rw_wunlock(&tidhash_lock);
return (0);
}
void
kthread_exit(void)
{
struct proc *p;
p = curthread->td_proc;
/* A module may be waiting for us to exit. */
wakeup(curthread);
/*
* The last exiting thread in a kernel process must tear down
* the whole process.
*/
rw_wlock(&tidhash_lock);
PROC_LOCK(p);
if (p->p_numthreads == 1) {
PROC_UNLOCK(p);
rw_wunlock(&tidhash_lock);
kproc_exit(0);
}
LIST_REMOVE(curthread, td_hash);
rw_wunlock(&tidhash_lock);
umtx_thread_exit(curthread);
PROC_SLOCK(p);
thread_exit();
}
int
kern_thr_exit(struct thread *td)
{
struct proc *p;
p = td->td_proc;
/*
* If all of the threads in a process call this routine to
* exit (e.g. all threads call pthread_exit()), exactly one
* thread should return to the caller to terminate the process
* instead of the thread.
*
* Checking p_numthreads alone is not sufficient since threads
* might be committed to terminating while the PROC_LOCK is
* dropped in either ptracestop() or while removing this thread
* from the tidhash. Instead, the p_pendingexits field holds
* the count of threads in either of those states and a thread
* is considered the "last" thread if all of the other threads
* in a process are already terminating.
*/
PROC_LOCK(p);
if (p->p_numthreads == p->p_pendingexits + 1) {
/*
* Ignore attempts to shut down last thread in the
* proc. This will actually call _exit(2) in the
* usermode trampoline when it returns.
*/
PROC_UNLOCK(p);
return (0);
}
p->p_pendingexits++;
td->td_dbgflags |= TDB_EXIT;
if (p->p_flag & P_TRACED && p->p_flag2 & P2_LWP_EVENTS)
ptracestop(td, SIGTRAP);
PROC_UNLOCK(p);
tidhash_remove(td);
PROC_LOCK(p);
p->p_pendingexits--;
/*
* The check above should prevent all other threads from this
* process from exiting while the PROC_LOCK is dropped, so
* there must be at least one other thread other than the
* current thread.
*/
KASSERT(p->p_numthreads > 1, ("too few threads"));
racct_sub(p, RACCT_NTHR, 1);
tdsigcleanup(td);
umtx_thread_exit(td);
PROC_SLOCK(p);
thread_stopped(p);
thread_exit();
/* NOTREACHED */
}
static void
racctd(void)
{
struct thread *td;
struct proc *p;
struct timeval wallclock;
uint64_t runtime;
for (;;) {
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
if (p->p_state != PRS_NORMAL)
continue;
if (p->p_flag & P_SYSTEM)
continue;
microuptime(&wallclock);
timevalsub(&wallclock, &p->p_stats->p_start);
PROC_LOCK(p);
PROC_SLOCK(p);
FOREACH_THREAD_IN_PROC(p, td) {
ruxagg(p, td);
thread_lock(td);
thread_unlock(td);
}
runtime = cputick2usec(p->p_rux.rux_runtime);
PROC_SUNLOCK(p);
#ifdef notyet
KASSERT(runtime >= p->p_prev_runtime,
("runtime < p_prev_runtime"));
#else
if (runtime < p->p_prev_runtime)
runtime = p->p_prev_runtime;
#endif
p->p_prev_runtime = runtime;
mtx_lock(&racct_lock);
racct_set_locked(p, RACCT_CPU, runtime);
racct_set_locked(p, RACCT_WALLCLOCK,
wallclock.tv_sec * 1000000 + wallclock.tv_usec);
mtx_unlock(&racct_lock);
PROC_UNLOCK(p);
}
sx_sunlock(&allproc_lock);
pause("-", hz);
}
int
sys_thr_exit(struct thread *td, struct thr_exit_args *uap)
/* long *state */
{
struct proc *p;
p = td->td_proc;
/* Signal userland that it can free the stack. */
if ((void *)uap->state != NULL) {
suword_lwpid(uap->state, 1);
kern_umtx_wake(td, uap->state, INT_MAX, 0);
}
rw_wlock(&tidhash_lock);
PROC_LOCK(p);
if (p->p_numthreads != 1) {
racct_sub(p, RACCT_NTHR, 1);
LIST_REMOVE(td, td_hash);
rw_wunlock(&tidhash_lock);
tdsigcleanup(td);
PROC_SLOCK(p);
thread_stopped(p);
thread_exit();
/* NOTREACHED */
}
/*
* Ignore attempts to shut down last thread in the proc. This
* will actually call _exit(2) in the usermode trampoline when
* it returns.
*/
PROC_UNLOCK(p);
rw_wunlock(&tidhash_lock);
return (0);
}
int
procfs_doprocstatus(PFS_FILL_ARGS)
{
struct session *sess;
struct thread *tdfirst;
struct tty *tp;
struct ucred *cr;
const char *wmesg;
char *pc;
char *sep;
int pid, ppid, pgid, sid;
int i;
pid = p->p_pid;
PROC_LOCK(p);
ppid = p->p_pptr ? p->p_pptr->p_pid : 0;
pgid = p->p_pgrp->pg_id;
sess = p->p_pgrp->pg_session;
SESS_LOCK(sess);
sid = sess->s_leader ? sess->s_leader->p_pid : 0;
/* comm pid ppid pgid sid tty ctty,sldr start ut st wmsg
euid ruid rgid,egid,groups[1 .. ngroups]
*/
pc = p->p_comm;
do {
if (*pc < 33 || *pc > 126 || *pc == '\\')
sbuf_printf(sb, "\\%03o", *pc);
else
sbuf_putc(sb, *pc);
} while (*++pc);
sbuf_printf(sb, " %d %d %d %d ", pid, ppid, pgid, sid);
if ((p->p_flag & P_CONTROLT) && (tp = sess->s_ttyp))
sbuf_printf(sb, "%s ", devtoname(tp->t_dev));
else
sbuf_printf(sb, "- ");
sep = "";
if (sess->s_ttyvp) {
sbuf_printf(sb, "%sctty", sep);
sep = ",";
}
if (SESS_LEADER(p)) {
sbuf_printf(sb, "%ssldr", sep);
sep = ",";
}
SESS_UNLOCK(sess);
if (*sep != ',') {
sbuf_printf(sb, "noflags");
}
tdfirst = FIRST_THREAD_IN_PROC(p);
thread_lock(tdfirst);
if (tdfirst->td_wchan != NULL) {
KASSERT(tdfirst->td_wmesg != NULL,
("wchan %p has no wmesg", tdfirst->td_wchan));
wmesg = tdfirst->td_wmesg;
} else
wmesg = "nochan";
thread_unlock(tdfirst);
if (p->p_flag & P_INMEM) {
struct timeval start, ut, st;
PROC_SLOCK(p);
calcru(p, &ut, &st);
PROC_SUNLOCK(p);
start = p->p_stats->p_start;
timevaladd(&start, &boottime);
sbuf_printf(sb, " %jd,%ld %jd,%ld %jd,%ld",
(intmax_t)start.tv_sec, start.tv_usec,
(intmax_t)ut.tv_sec, ut.tv_usec,
(intmax_t)st.tv_sec, st.tv_usec);
} else
sbuf_printf(sb, " -1,-1 -1,-1 -1,-1");
sbuf_printf(sb, " %s", wmesg);
cr = p->p_ucred;
sbuf_printf(sb, " %lu %lu %lu",
(u_long)cr->cr_uid,
(u_long)cr->cr_ruid,
(u_long)cr->cr_rgid);
/* egid (cr->cr_svgid) is equal to cr_ngroups[0]
see also getegid(2) in /sys/kern/kern_prot.c */
for (i = 0; i < cr->cr_ngroups; i++) {
sbuf_printf(sb, ",%lu", (u_long)cr->cr_groups[i]);
}
if (jailed(cr)) {
mtx_lock(&cr->cr_prison->pr_mtx);
sbuf_printf(sb, " %s",
prison_name(td->td_ucred->cr_prison, cr->cr_prison));
mtx_unlock(&cr->cr_prison->pr_mtx);
} else {
sbuf_printf(sb, " -");
}
PROC_UNLOCK(p);
sbuf_printf(sb, "\n");
return (0);
}
/*
* Exit: deallocate address space and other resources, change proc state to
* zombie, and unlink proc from allproc and parent's lists. Save exit status
* and rusage for wait(). Check for child processes and orphan them.
*/
void
exit1(struct thread *td, int rv)
{
struct proc *p, *nq, *q;
struct vnode *vtmp;
struct vnode *ttyvp = NULL;
struct plimit *plim;
mtx_assert(&Giant, MA_NOTOWNED);
p = td->td_proc;
/*
* XXX in case we're rebooting we just let init die in order to
* work around an unsolved stack overflow seen very late during
* shutdown on sparc64 when the gmirror worker process exists.
*/
if (p == initproc && rebooting == 0) {
printf("init died (signal %d, exit %d)\n",
WTERMSIG(rv), WEXITSTATUS(rv));
panic("Going nowhere without my init!");
}
/*
* MUST abort all other threads before proceeding past here.
*/
PROC_LOCK(p);
while (p->p_flag & P_HADTHREADS) {
/*
* First check if some other thread got here before us.
* If so, act appropriately: exit or suspend.
*/
thread_suspend_check(0);
/*
* Kill off the other threads. This requires
* some co-operation from other parts of the kernel
* so it may not be instantaneous. With this state set
* any thread entering the kernel from userspace will
* thread_exit() in trap(). Any thread attempting to
* sleep will return immediately with EINTR or EWOULDBLOCK
* which will hopefully force them to back out to userland
* freeing resources as they go. Any thread attempting
* to return to userland will thread_exit() from userret().
* thread_exit() will unsuspend us when the last of the
* other threads exits.
* If there is already a thread singler after resumption,
* calling thread_single will fail; in that case, we just
* re-check all suspension request, the thread should
* either be suspended there or exit.
*/
if (!thread_single(SINGLE_EXIT))
break;
/*
* All other activity in this process is now stopped.
* Threading support has been turned off.
*/
}
KASSERT(p->p_numthreads == 1,
("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
racct_sub(p, RACCT_NTHR, 1);
/*
* Wakeup anyone in procfs' PIOCWAIT. They should have a hold
* on our vmspace, so we should block below until they have
* released their reference to us. Note that if they have
* requested S_EXIT stops we will block here until they ack
* via PIOCCONT.
*/
_STOPEVENT(p, S_EXIT, rv);
/*
* Ignore any pending request to stop due to a stop signal.
* Once P_WEXIT is set, future requests will be ignored as
* well.
*/
p->p_flag &= ~P_STOPPED_SIG;
KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
/*
* Note that we are exiting and do another wakeup of anyone in
* PIOCWAIT in case they aren't listening for S_EXIT stops or
* decided to wait again after we told them we are exiting.
*/
p->p_flag |= P_WEXIT;
wakeup(&p->p_stype);
/*
* Wait for any processes that have a hold on our vmspace to
* release their reference.
*/
while (p->p_lock > 0)
msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
p->p_xstat = rv; /* Let event handler change exit status */
PROC_UNLOCK(p);
/* Drain the limit callout while we don't have the proc locked */
callout_drain(&p->p_limco);
#ifdef AUDIT
/*
* The Sun BSM exit token contains two components: an exit status as
* passed to exit(), and a return value to indicate what sort of exit
* it was. The exit status is WEXITSTATUS(rv), but it's not clear
* what the return value is.
*/
AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
AUDIT_SYSCALL_EXIT(0, td);
#endif
/* Are we a task leader? */
if (p == p->p_leader) {
mtx_lock(&ppeers_lock);
q = p->p_peers;
while (q != NULL) {
PROC_LOCK(q);
kern_psignal(q, SIGKILL);
PROC_UNLOCK(q);
q = q->p_peers;
}
while (p->p_peers != NULL)
msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
mtx_unlock(&ppeers_lock);
}
/*
* Check if any loadable modules need anything done at process exit.
* E.g. SYSV IPC stuff
* XXX what if one of these generates an error?
*/
EVENTHANDLER_INVOKE(process_exit, p);
/*
* If parent is waiting for us to exit or exec,
* P_PPWAIT is set; we will wakeup the parent below.
*/
PROC_LOCK(p);
rv = p->p_xstat; /* Event handler could change exit status */
stopprofclock(p);
p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
/*
* Stop the real interval timer. If the handler is currently
* executing, prevent it from rearming itself and let it finish.
*/
if (timevalisset(&p->p_realtimer.it_value) &&
callout_stop(&p->p_itcallout) == 0) {
timevalclear(&p->p_realtimer.it_interval);
msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
KASSERT(!timevalisset(&p->p_realtimer.it_value),
("realtime timer is still armed"));
}
PROC_UNLOCK(p);
/*
* Reset any sigio structures pointing to us as a result of
* F_SETOWN with our pid.
*/
funsetownlst(&p->p_sigiolst);
/*
* If this process has an nlminfo data area (for lockd), release it
*/
if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
(*nlminfo_release_p)(p);
/*
* Close open files and release open-file table.
* This may block!
*/
fdescfree(td);
/*
* If this thread tickled GEOM, we need to wait for the giggling to
* stop before we return to userland
*/
if (td->td_pflags & TDP_GEOM)
g_waitidle();
/*
* Remove ourself from our leader's peer list and wake our leader.
*/
mtx_lock(&ppeers_lock);
if (p->p_leader->p_peers) {
q = p->p_leader;
while (q->p_peers != p)
q = q->p_peers;
q->p_peers = p->p_peers;
wakeup(p->p_leader);
}
mtx_unlock(&ppeers_lock);
vmspace_exit(td);
sx_xlock(&proctree_lock);
if (SESS_LEADER(p)) {
struct session *sp = p->p_session;
struct tty *tp;
/*
* s_ttyp is not zero'd; we use this to indicate that
* the session once had a controlling terminal. (for
* logging and informational purposes)
*/
SESS_LOCK(sp);
ttyvp = sp->s_ttyvp;
tp = sp->s_ttyp;
sp->s_ttyvp = NULL;
sp->s_ttydp = NULL;
sp->s_leader = NULL;
SESS_UNLOCK(sp);
/*
* Signal foreground pgrp and revoke access to
* controlling terminal if it has not been revoked
* already.
*
* Because the TTY may have been revoked in the mean
* time and could already have a new session associated
* with it, make sure we don't send a SIGHUP to a
* foreground process group that does not belong to this
* session.
*/
if (tp != NULL) {
tty_lock(tp);
if (tp->t_session == sp)
tty_signal_pgrp(tp, SIGHUP);
tty_unlock(tp);
}
if (ttyvp != NULL) {
sx_xunlock(&proctree_lock);
if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
VOP_REVOKE(ttyvp, REVOKEALL);
VOP_UNLOCK(ttyvp, 0);
}
sx_xlock(&proctree_lock);
}
}
fixjobc(p, p->p_pgrp, 0);
sx_xunlock(&proctree_lock);
(void)acct_process(td);
/* Release the TTY now we've unlocked everything. */
if (ttyvp != NULL)
vrele(ttyvp);
#ifdef KTRACE
ktrprocexit(td);
#endif
/*
* Release reference to text vnode
*/
if ((vtmp = p->p_textvp) != NULL) {
p->p_textvp = NULL;
vrele(vtmp);
}
/*
* Release our limits structure.
*/
plim = p->p_limit;
p->p_limit = NULL;
lim_free(plim);
tidhash_remove(td);
/*
* Remove proc from allproc queue and pidhash chain.
* Place onto zombproc. Unlink from parent's child list.
*/
sx_xlock(&allproc_lock);
LIST_REMOVE(p, p_list);
LIST_INSERT_HEAD(&zombproc, p, p_list);
LIST_REMOVE(p, p_hash);
sx_xunlock(&allproc_lock);
/*
* Call machine-dependent code to release any
* machine-dependent resources other than the address space.
* The address space is released by "vmspace_exitfree(p)" in
* vm_waitproc().
*/
cpu_exit(td);
WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
/*
* Reparent all of our children to init.
*/
sx_xlock(&proctree_lock);
q = LIST_FIRST(&p->p_children);
if (q != NULL) /* only need this if any child is S_ZOMB */
wakeup(initproc);
for (; q != NULL; q = nq) {
nq = LIST_NEXT(q, p_sibling);
PROC_LOCK(q);
proc_reparent(q, initproc);
q->p_sigparent = SIGCHLD;
/*
* Traced processes are killed
* since their existence means someone is screwing up.
*/
if (q->p_flag & P_TRACED) {
struct thread *temp;
/*
* Since q was found on our children list, the
* proc_reparent() call moved q to the orphan
* list due to present P_TRACED flag. Clear
* orphan link for q now while q is locked.
*/
clear_orphan(q);
q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
FOREACH_THREAD_IN_PROC(q, temp)
temp->td_dbgflags &= ~TDB_SUSPEND;
kern_psignal(q, SIGKILL);
}
PROC_UNLOCK(q);
}
/*
* Also get rid of our orphans.
*/
while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
PROC_LOCK(q);
clear_orphan(q);
PROC_UNLOCK(q);
}
/* Save exit status. */
PROC_LOCK(p);
p->p_xthread = td;
/* Tell the prison that we are gone. */
prison_proc_free(p->p_ucred->cr_prison);
#ifdef KDTRACE_HOOKS
/*
* Tell the DTrace fasttrap provider about the exit if it
* has declared an interest.
*/
if (dtrace_fasttrap_exit)
dtrace_fasttrap_exit(p);
#endif
/*
* Notify interested parties of our demise.
*/
KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
#ifdef KDTRACE_HOOKS
int reason = CLD_EXITED;
if (WCOREDUMP(rv))
reason = CLD_DUMPED;
else if (WIFSIGNALED(rv))
reason = CLD_KILLED;
SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
#endif
/*
* Just delete all entries in the p_klist. At this point we won't
* report any more events, and there are nasty race conditions that
* can beat us if we don't.
*/
knlist_clear(&p->p_klist, 1);
/*
* If this is a process with a descriptor, we may not need to deliver
* a signal to the parent. proctree_lock is held over
* procdesc_exit() to serialize concurrent calls to close() and
* exit().
*/
if (p->p_procdesc == NULL || procdesc_exit(p)) {
/*
* Notify parent that we're gone. If parent has the
* PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
* notify process 1 instead (and hope it will handle this
* situation).
*/
PROC_LOCK(p->p_pptr);
mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
if (p->p_pptr->p_sigacts->ps_flag &
(PS_NOCLDWAIT | PS_CLDSIGIGN)) {
struct proc *pp;
mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
pp = p->p_pptr;
PROC_UNLOCK(pp);
proc_reparent(p, initproc);
p->p_sigparent = SIGCHLD;
PROC_LOCK(p->p_pptr);
/*
* Notify parent, so in case he was wait(2)ing or
* executing waitpid(2) with our pid, he will
* continue.
*/
wakeup(pp);
} else
mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
if (p->p_pptr == initproc)
kern_psignal(p->p_pptr, SIGCHLD);
else if (p->p_sigparent != 0) {
if (p->p_sigparent == SIGCHLD)
childproc_exited(p);
else /* LINUX thread */
kern_psignal(p->p_pptr, p->p_sigparent);
}
} else
PROC_LOCK(p->p_pptr);
sx_xunlock(&proctree_lock);
/*
* The state PRS_ZOMBIE prevents other proesses from sending
* signal to the process, to avoid memory leak, we free memory
* for signal queue at the time when the state is set.
*/
sigqueue_flush(&p->p_sigqueue);
sigqueue_flush(&td->td_sigqueue);
/*
* We have to wait until after acquiring all locks before
* changing p_state. We need to avoid all possible context
* switches (including ones from blocking on a mutex) while
* marked as a zombie. We also have to set the zombie state
* before we release the parent process' proc lock to avoid
* a lost wakeup. So, we first call wakeup, then we grab the
* sched lock, update the state, and release the parent process'
* proc lock.
*/
wakeup(p->p_pptr);
cv_broadcast(&p->p_pwait);
sched_exit(p->p_pptr, td);
PROC_SLOCK(p);
p->p_state = PRS_ZOMBIE;
PROC_UNLOCK(p->p_pptr);
/*
* Hopefully no one will try to deliver a signal to the process this
* late in the game.
*/
knlist_destroy(&p->p_klist);
/*
* Save our children's rusage information in our exit rusage.
*/
ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
/*
* Make sure the scheduler takes this thread out of its tables etc.
* This will also release this thread's reference to the ucred.
* Other thread parts to release include pcb bits and such.
*/
thread_exit();
}
static int
procfs_control(struct thread *td, struct proc *p, int op)
{
int error = 0;
struct thread *temp;
/*
* Attach - attaches the target process for debugging
* by the calling process.
*/
if (op == PROCFS_CTL_ATTACH) {
sx_xlock(&proctree_lock);
PROC_LOCK(p);
if ((error = p_candebug(td, p)) != 0)
goto out;
if (p->p_flag & P_TRACED) {
error = EBUSY;
goto out;
}
/* Can't trace yourself! */
if (p->p_pid == td->td_proc->p_pid) {
error = EINVAL;
goto out;
}
/*
* 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.
*/
p->p_flag |= P_TRACED;
faultin(p);
p->p_xstat = 0; /* XXX ? */
if (p->p_pptr != td->td_proc) {
p->p_oppid = p->p_pptr->p_pid;
proc_reparent(p, td->td_proc);
}
psignal(p, SIGSTOP);
out:
PROC_UNLOCK(p);
sx_xunlock(&proctree_lock);
return (error);
}
/*
* Authorization check: rely on normal debugging protection, except
* allow processes to disengage debugging on a process onto which
* they have previously attached, but no longer have permission to
* debug.
*/
PROC_LOCK(p);
if (op != PROCFS_CTL_DETACH &&
((error = p_candebug(td, p)))) {
PROC_UNLOCK(p);
return (error);
}
/*
* Target process must be stopped, owned by (td) and
* be set up for tracing (P_TRACED flag set).
* Allow DETACH to take place at any time for sanity.
* Allow WAIT any time, of course.
*/
switch (op) {
case PROCFS_CTL_DETACH:
case PROCFS_CTL_WAIT:
break;
default:
if (!TRACE_WAIT_P(td->td_proc, p)) {
PROC_UNLOCK(p);
return (EBUSY);
}
}
#ifdef FIX_SSTEP
/*
* do single-step fixup if needed
*/
FIX_SSTEP(FIRST_THREAD_IN_PROC(p));
#endif
/*
* Don't deliver any signal by default.
* To continue with a signal, just send
* the signal name to the ctl file
*/
p->p_xstat = 0;
switch (op) {
/*
* Detach. Cleans up the target process, reparent it if possible
* and set it running once more.
*/
case PROCFS_CTL_DETACH:
/* if not being traced, then this is a painless no-op */
if ((p->p_flag & P_TRACED) == 0) {
PROC_UNLOCK(p);
return (0);
}
/* not being traced any more */
p->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
/* remove pending SIGTRAP, else the process will die */
sigqueue_delete_proc(p, SIGTRAP);
FOREACH_THREAD_IN_PROC(p, temp)
temp->td_dbgflags &= ~TDB_SUSPEND;
PROC_UNLOCK(p);
/* give process back to original parent */
sx_xlock(&proctree_lock);
if (p->p_oppid != p->p_pptr->p_pid) {
struct proc *pp;
pp = pfind(p->p_oppid);
PROC_LOCK(p);
if (pp) {
PROC_UNLOCK(pp);
proc_reparent(p, pp);
}
} else
PROC_LOCK(p);
p->p_oppid = 0;
p->p_flag &= ~P_WAITED; /* XXX ? */
sx_xunlock(&proctree_lock);
wakeup(td->td_proc); /* XXX for CTL_WAIT below ? */
break;
/*
* Step. Let the target process execute a single instruction.
* What does it mean to single step a threaded program?
*/
case PROCFS_CTL_STEP:
error = proc_sstep(FIRST_THREAD_IN_PROC(p));
if (error) {
PROC_UNLOCK(p);
return (error);
}
break;
/*
* Run. Let the target process continue running until a breakpoint
* or some other trap.
*/
case PROCFS_CTL_RUN:
p->p_flag &= ~P_STOPPED_SIG; /* this uses SIGSTOP */
break;
/*
* Wait for the target process to stop.
* If the target is not being traced then just wait
* to enter
*/
case PROCFS_CTL_WAIT:
if (p->p_flag & P_TRACED) {
while (error == 0 &&
(P_SHOULDSTOP(p)) &&
(p->p_flag & P_TRACED) &&
(p->p_pptr == td->td_proc))
error = msleep(p, &p->p_mtx,
PWAIT|PCATCH, "procfsx", 0);
if (error == 0 && !TRACE_WAIT_P(td->td_proc, p))
error = EBUSY;
} else {
while (error == 0 && P_SHOULDSTOP(p))
error = msleep(p, &p->p_mtx,
PWAIT|PCATCH, "procfs", 0);
}
PROC_UNLOCK(p);
return (error);
default:
panic("procfs_control");
}
PROC_SLOCK(p);
thread_unsuspend(p); /* If it can run, let it do so. */
PROC_SUNLOCK(p);
PROC_UNLOCK(p);
return (0);
}
static void
do_fork(struct thread *td, struct fork_req *fr, struct proc *p2, struct thread *td2,
struct vmspace *vm2, struct file *fp_procdesc)
{
struct proc *p1, *pptr;
int trypid;
struct filedesc *fd;
struct filedesc_to_leader *fdtol;
struct sigacts *newsigacts;
sx_assert(&proctree_lock, SX_SLOCKED);
sx_assert(&allproc_lock, SX_XLOCKED);
p1 = td->td_proc;
trypid = fork_findpid(fr->fr_flags);
sx_sunlock(&proctree_lock);
p2->p_state = PRS_NEW; /* protect against others */
p2->p_pid = trypid;
AUDIT_ARG_PID(p2->p_pid);
LIST_INSERT_HEAD(&allproc, p2, p_list);
allproc_gen++;
LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
tidhash_add(td2);
PROC_LOCK(p2);
PROC_LOCK(p1);
sx_xunlock(&allproc_lock);
bcopy(&p1->p_startcopy, &p2->p_startcopy,
__rangeof(struct proc, p_startcopy, p_endcopy));
pargs_hold(p2->p_args);
PROC_UNLOCK(p1);
bzero(&p2->p_startzero,
__rangeof(struct proc, p_startzero, p_endzero));
/* Tell the prison that we exist. */
prison_proc_hold(p2->p_ucred->cr_prison);
PROC_UNLOCK(p2);
/*
* Malloc things while we don't hold any locks.
*/
if (fr->fr_flags & RFSIGSHARE)
newsigacts = NULL;
else
newsigacts = sigacts_alloc();
/*
* Copy filedesc.
*/
if (fr->fr_flags & RFCFDG) {
fd = fdinit(p1->p_fd, false);
fdtol = NULL;
} else if (fr->fr_flags & RFFDG) {
fd = fdcopy(p1->p_fd);
fdtol = NULL;
} else {
fd = fdshare(p1->p_fd);
if (p1->p_fdtol == NULL)
p1->p_fdtol = filedesc_to_leader_alloc(NULL, NULL,
p1->p_leader);
if ((fr->fr_flags & RFTHREAD) != 0) {
/*
* Shared file descriptor table, and shared
* process leaders.
*/
fdtol = p1->p_fdtol;
FILEDESC_XLOCK(p1->p_fd);
fdtol->fdl_refcount++;
FILEDESC_XUNLOCK(p1->p_fd);
} else {
/*
* Shared file descriptor table, and different
* process leaders.
*/
fdtol = filedesc_to_leader_alloc(p1->p_fdtol,
p1->p_fd, p2);
}
}
/*
* 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.
*/
PROC_LOCK(p2);
PROC_LOCK(p1);
bzero(&td2->td_startzero,
__rangeof(struct thread, td_startzero, td_endzero));
bcopy(&td->td_startcopy, &td2->td_startcopy,
__rangeof(struct thread, td_startcopy, td_endcopy));
bcopy(&p2->p_comm, &td2->td_name, sizeof(td2->td_name));
td2->td_sigstk = td->td_sigstk;
td2->td_flags = TDF_INMEM;
td2->td_lend_user_pri = PRI_MAX;
#ifdef VIMAGE
td2->td_vnet = NULL;
td2->td_vnet_lpush = NULL;
#endif
/*
* Allow the scheduler to initialize the child.
*/
thread_lock(td);
sched_fork(td, td2);
thread_unlock(td);
/*
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
*/
p2->p_flag = P_INMEM;
p2->p_flag2 = p1->p_flag2 & (P2_NOTRACE | P2_NOTRACE_EXEC | P2_TRAPCAP);
p2->p_swtick = ticks;
if (p1->p_flag & P_PROFIL)
startprofclock(p2);
/*
* Whilst the proc lock is held, copy the VM domain data out
* using the VM domain method.
*/
vm_domain_policy_init(&p2->p_vm_dom_policy);
vm_domain_policy_localcopy(&p2->p_vm_dom_policy,
&p1->p_vm_dom_policy);
if (fr->fr_flags & RFSIGSHARE) {
p2->p_sigacts = sigacts_hold(p1->p_sigacts);
} else {
sigacts_copy(newsigacts, p1->p_sigacts);
p2->p_sigacts = newsigacts;
}
if (fr->fr_flags & RFTSIGZMB)
p2->p_sigparent = RFTSIGNUM(fr->fr_flags);
else if (fr->fr_flags & RFLINUXTHPN)
p2->p_sigparent = SIGUSR1;
else
p2->p_sigparent = SIGCHLD;
p2->p_textvp = p1->p_textvp;
p2->p_fd = fd;
p2->p_fdtol = fdtol;
if (p1->p_flag2 & P2_INHERIT_PROTECTED) {
p2->p_flag |= P_PROTECTED;
p2->p_flag2 |= P2_INHERIT_PROTECTED;
}
/*
* p_limit is copy-on-write. Bump its refcount.
*/
lim_fork(p1, p2);
thread_cow_get_proc(td2, p2);
pstats_fork(p1->p_stats, p2->p_stats);
PROC_UNLOCK(p1);
PROC_UNLOCK(p2);
/* Bump references to the text vnode (for procfs). */
if (p2->p_textvp)
vrefact(p2->p_textvp);
/*
* Set up linkage for kernel based threading.
*/
if ((fr->fr_flags & RFTHREAD) != 0) {
mtx_lock(&ppeers_lock);
p2->p_peers = p1->p_peers;
p1->p_peers = p2;
p2->p_leader = p1->p_leader;
mtx_unlock(&ppeers_lock);
PROC_LOCK(p1->p_leader);
if ((p1->p_leader->p_flag & P_WEXIT) != 0) {
PROC_UNLOCK(p1->p_leader);
/*
* The task leader is exiting, so process p1 is
* going to be killed shortly. Since p1 obviously
* isn't dead yet, we know that the leader is either
* sending SIGKILL's to all the processes in this
* task or is sleeping waiting for all the peers to
* exit. We let p1 complete the fork, but we need
* to go ahead and kill the new process p2 since
* the task leader may not get a chance to send
* SIGKILL to it. We leave it on the list so that
* the task leader will wait for this new process
* to commit suicide.
*/
PROC_LOCK(p2);
kern_psignal(p2, SIGKILL);
PROC_UNLOCK(p2);
} else
PROC_UNLOCK(p1->p_leader);
} else {
p2->p_peers = NULL;
p2->p_leader = p2;
}
sx_xlock(&proctree_lock);
PGRP_LOCK(p1->p_pgrp);
PROC_LOCK(p2);
PROC_LOCK(p1);
/*
* Preserve some more flags in subprocess. P_PROFIL has already
* been preserved.
*/
p2->p_flag |= p1->p_flag & P_SUGID;
td2->td_pflags |= (td->td_pflags & TDP_ALTSTACK) | TDP_FORKING;
SESS_LOCK(p1->p_session);
if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
p2->p_flag |= P_CONTROLT;
SESS_UNLOCK(p1->p_session);
if (fr->fr_flags & RFPPWAIT)
p2->p_flag |= P_PPWAIT;
p2->p_pgrp = p1->p_pgrp;
LIST_INSERT_AFTER(p1, p2, p_pglist);
PGRP_UNLOCK(p1->p_pgrp);
LIST_INIT(&p2->p_children);
LIST_INIT(&p2->p_orphans);
callout_init_mtx(&p2->p_itcallout, &p2->p_mtx, 0);
/*
* If PF_FORK is set, the child process inherits the
* procfs ioctl flags from its parent.
*/
if (p1->p_pfsflags & PF_FORK) {
p2->p_stops = p1->p_stops;
p2->p_pfsflags = p1->p_pfsflags;
}
/*
* This begins the section where we must prevent the parent
* from being swapped.
*/
_PHOLD(p1);
PROC_UNLOCK(p1);
/*
* Attach the new process to its parent.
*
* If RFNOWAIT is set, the newly created process becomes a child
* of init. This effectively disassociates the child from the
* parent.
*/
if ((fr->fr_flags & RFNOWAIT) != 0) {
pptr = p1->p_reaper;
p2->p_reaper = pptr;
} else {
p2->p_reaper = (p1->p_treeflag & P_TREE_REAPER) != 0 ?
p1 : p1->p_reaper;
pptr = p1;
}
p2->p_pptr = pptr;
LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
LIST_INIT(&p2->p_reaplist);
LIST_INSERT_HEAD(&p2->p_reaper->p_reaplist, p2, p_reapsibling);
if (p2->p_reaper == p1)
p2->p_reapsubtree = p2->p_pid;
sx_xunlock(&proctree_lock);
/* Inform accounting that we have forked. */
p2->p_acflag = AFORK;
PROC_UNLOCK(p2);
#ifdef KTRACE
ktrprocfork(p1, p2);
#endif
/*
* Finish creating the child process. It will return via a different
* execution path later. (ie: directly into user mode)
*/
vm_forkproc(td, p2, td2, vm2, fr->fr_flags);
if (fr->fr_flags == (RFFDG | RFPROC)) {
VM_CNT_INC(v_forks);
VM_CNT_ADD(v_forkpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
} else if (fr->fr_flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
VM_CNT_INC(v_vforks);
VM_CNT_ADD(v_vforkpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
} else if (p1 == &proc0) {
VM_CNT_INC(v_kthreads);
VM_CNT_ADD(v_kthreadpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
} else {
VM_CNT_INC(v_rforks);
VM_CNT_ADD(v_rforkpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
}
/*
* Associate the process descriptor with the process before anything
* can happen that might cause that process to need the descriptor.
* However, don't do this until after fork(2) can no longer fail.
*/
if (fr->fr_flags & RFPROCDESC)
procdesc_new(p2, fr->fr_pd_flags);
/*
* Both processes are set up, now check if any loadable modules want
* to adjust anything.
*/
EVENTHANDLER_INVOKE(process_fork, p1, p2, fr->fr_flags);
/*
* Set the child start time and mark the process as being complete.
*/
PROC_LOCK(p2);
PROC_LOCK(p1);
microuptime(&p2->p_stats->p_start);
PROC_SLOCK(p2);
p2->p_state = PRS_NORMAL;
PROC_SUNLOCK(p2);
#ifdef KDTRACE_HOOKS
/*
* Tell the DTrace fasttrap provider about the new process so that any
* tracepoints inherited from the parent can be removed. We have to do
* this only after p_state is PRS_NORMAL since the fasttrap module will
* use pfind() later on.
*/
if ((fr->fr_flags & RFMEM) == 0 && dtrace_fasttrap_fork)
dtrace_fasttrap_fork(p1, p2);
#endif
/*
* Hold the process so that it cannot exit after we make it runnable,
* but before we wait for the debugger.
*/
_PHOLD(p2);
if (p1->p_ptevents & PTRACE_FORK) {
/*
* Arrange for debugger to receive the fork event.
*
* We can report PL_FLAG_FORKED regardless of
* P_FOLLOWFORK settings, but it does not make a sense
* for runaway child.
*/
td->td_dbgflags |= TDB_FORK;
td->td_dbg_forked = p2->p_pid;
td2->td_dbgflags |= TDB_STOPATFORK;
}
if (fr->fr_flags & RFPPWAIT) {
td->td_pflags |= TDP_RFPPWAIT;
td->td_rfppwait_p = p2;
td->td_dbgflags |= TDB_VFORK;
}
PROC_UNLOCK(p2);
/*
* Now can be swapped.
*/
_PRELE(p1);
PROC_UNLOCK(p1);
/*
* Tell any interested parties about the new process.
*/
knote_fork(p1->p_klist, p2->p_pid);
SDT_PROBE3(proc, , , create, p2, p1, fr->fr_flags);
if (fr->fr_flags & RFPROCDESC) {
procdesc_finit(p2->p_procdesc, fp_procdesc);
fdrop(fp_procdesc, td);
}
if ((fr->fr_flags & RFSTOPPED) == 0) {
/*
* If RFSTOPPED not requested, make child runnable and
* add to run queue.
*/
thread_lock(td2);
TD_SET_CAN_RUN(td2);
sched_add(td2, SRQ_BORING);
thread_unlock(td2);
if (fr->fr_pidp != NULL)
*fr->fr_pidp = p2->p_pid;
} else {
*fr->fr_procp = p2;
}
PROC_LOCK(p2);
/*
* Wait until debugger is attached to child.
*/
while (td2->td_proc == p2 && (td2->td_dbgflags & TDB_STOPATFORK) != 0)
cv_wait(&p2->p_dbgwait, &p2->p_mtx);
_PRELE(p2);
racct_proc_fork_done(p2);
PROC_UNLOCK(p2);
}
/*
* Process ioctls
*/
int
procfs_ioctl(PFS_IOCTL_ARGS)
{
struct procfs_status *ps;
#ifdef COMPAT_FREEBSD32
struct procfs_status32 *ps32;
#endif
int error, flags, sig;
#ifdef COMPAT_FREEBSD6
int ival;
#endif
KASSERT(p != NULL,
("%s() called without a process", __func__));
PROC_LOCK_ASSERT(p, MA_OWNED);
error = 0;
switch (cmd) {
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 1, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 1):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCBIS:
p->p_stops |= *(unsigned int *)data;
break;
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 2, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 2):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCBIC:
p->p_stops &= ~*(unsigned int *)data;
break;
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 3, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 3):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCSFL:
flags = *(unsigned int *)data;
if (flags & PF_ISUGID) {
/*
* XXXRW: Is this specific check required here, as
* p_candebug() should implement it, or other checks
* are missing.
*/
error = priv_check(td, PRIV_DEBUG_SUGID);
if (error)
break;
}
p->p_pfsflags = flags;
break;
case PIOCGFL:
*(unsigned int *)data = p->p_pfsflags;
break;
case PIOCWAIT:
while (p->p_step == 0 && (p->p_flag & P_WEXIT) == 0) {
/* sleep until p stops */
_PHOLD(p);
error = msleep(&p->p_stype, &p->p_mtx,
PWAIT|PCATCH, "pioctl", 0);
_PRELE(p);
if (error != 0)
break;
}
/* fall through to PIOCSTATUS */
case PIOCSTATUS:
ps = (struct procfs_status *)data;
ps->state = (p->p_step == 0);
ps->flags = 0; /* nope */
ps->events = p->p_stops;
ps->why = p->p_step ? p->p_stype : 0;
ps->val = p->p_step ? p->p_xstat : 0;
break;
#ifdef COMPAT_FREEBSD32
case PIOCWAIT32:
while (p->p_step == 0 && (p->p_flag & P_WEXIT) == 0) {
/* sleep until p stops */
_PHOLD(p);
error = msleep(&p->p_stype, &p->p_mtx,
PWAIT|PCATCH, "pioctl", 0);
_PRELE(p);
if (error != 0)
break;
}
/* fall through to PIOCSTATUS32 */
case PIOCSTATUS32:
ps32 = (struct procfs_status32 *)data;
ps32->state = (p->p_step == 0);
ps32->flags = 0; /* nope */
ps32->events = p->p_stops;
ps32->why = p->p_step ? p->p_stype : 0;
ps32->val = p->p_step ? p->p_xstat : 0;
break;
#endif
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 5, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 5):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCCONT:
if (p->p_step == 0)
break;
sig = *(unsigned int *)data;
if (sig != 0 && !_SIG_VALID(sig)) {
error = EINVAL;
break;
}
#if 0
p->p_step = 0;
if (P_SHOULDSTOP(p)) {
p->p_xstat = sig;
p->p_flag &= ~(P_STOPPED_TRACE|P_STOPPED_SIG);
PROC_SLOCK(p);
thread_unsuspend(p);
PROC_SUNLOCK(p);
} else if (sig)
kern_psignal(p, sig);
#else
if (sig)
kern_psignal(p, sig);
p->p_step = 0;
wakeup(&p->p_step);
#endif
break;
default:
error = (ENOTTY);
}
return (error);
}
