static int
fork_norfproc(struct thread *td, int flags)
{
int error;
struct proc *p1;
KASSERT((flags & RFPROC) == 0,
("fork_norfproc called with RFPROC set"));
p1 = td->td_proc;
if (((p1->p_flag & (P_HADTHREADS|P_SYSTEM)) == P_HADTHREADS) &&
(flags & (RFCFDG | RFFDG))) {
PROC_LOCK(p1);
if (thread_single(p1, SINGLE_BOUNDARY)) {
PROC_UNLOCK(p1);
return (ERESTART);
}
PROC_UNLOCK(p1);
}
error = vm_forkproc(td, NULL, NULL, NULL, flags);
if (error)
goto fail;
/*
* Close all file descriptors.
*/
if (flags & RFCFDG) {
struct filedesc *fdtmp;
fdtmp = fdinit(td->td_proc->p_fd, false);
fdescfree(td);
p1->p_fd = fdtmp;
}
/*
* Unshare file descriptors (from parent).
*/
if (flags & RFFDG)
fdunshare(td);
fail:
if (((p1->p_flag & (P_HADTHREADS|P_SYSTEM)) == P_HADTHREADS) &&
(flags & (RFCFDG | RFFDG))) {
PROC_LOCK(p1);
thread_single_end(p1, SINGLE_BOUNDARY);
PROC_UNLOCK(p1);
}
return (error);
}
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);
}
int
fork1(struct proc *p1, int exitsig, int flags, void *stack, size_t stacksize,
void (*func)(void *), void *arg, register_t *retval,
struct proc **rnewprocp)
{
struct proc *p2;
uid_t uid;
struct vmspace *vm;
int count;
vaddr_t uaddr;
int s;
extern void endtsleep(void *);
extern void realitexpire(void *);
/*
* Although process entries are dynamically created, we still keep
* a global limit on the maximum number we will create. We reserve
* the last 5 processes to root. The variable nprocs is the current
* number of processes, maxproc is the limit.
*/
uid = p1->p_cred->p_ruid;
if ((nprocs >= maxproc - 5 && uid != 0) || nprocs >= maxproc) {
static struct timeval lasttfm;
if (ratecheck(&lasttfm, &fork_tfmrate))
tablefull("proc");
return (EAGAIN);
}
nprocs++;
/*
* Increment the count of procs running with this uid. Don't allow
* a nonprivileged user to exceed their current limit.
*/
count = chgproccnt(uid, 1);
if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) {
(void)chgproccnt(uid, -1);
nprocs--;
return (EAGAIN);
}
uaddr = uvm_km_alloc1(kernel_map, USPACE, USPACE_ALIGN, 1);
if (uaddr == 0) {
chgproccnt(uid, -1);
nprocs--;
return (ENOMEM);
}
/*
* From now on, we're committed to the fork and cannot fail.
*/
/* Allocate new proc. */
p2 = pool_get(&proc_pool, PR_WAITOK);
p2->p_stat = SIDL; /* protect against others */
p2->p_exitsig = exitsig;
p2->p_forw = p2->p_back = NULL;
#ifdef RTHREADS
if (flags & FORK_THREAD) {
atomic_setbits_int(&p2->p_flag, P_THREAD);
p2->p_p = p1->p_p;
TAILQ_INSERT_TAIL(&p2->p_p->ps_threads, p2, p_thr_link);
} else {
process_new(p2, p1);
}
#else
process_new(p2, p1);
#endif
/*
* 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.
*/
bzero(&p2->p_startzero,
(unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero));
bcopy(&p1->p_startcopy, &p2->p_startcopy,
(unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
/*
* Initialize the timeouts.
*/
timeout_set(&p2->p_sleep_to, endtsleep, p2);
timeout_set(&p2->p_realit_to, realitexpire, p2);
#if defined(__HAVE_CPUINFO)
p2->p_cpu = p1->p_cpu;
#endif
/*
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
* The p_stats and p_sigacts substructs are set in vm_fork.
*/
p2->p_flag = 0;
p2->p_emul = p1->p_emul;
if (p1->p_flag & P_PROFIL)
startprofclock(p2);
atomic_setbits_int(&p2->p_flag, p1->p_flag & (P_SUGID | P_SUGIDEXEC));
if (flags & FORK_PTRACE)
atomic_setbits_int(&p2->p_flag, p1->p_flag & P_TRACED);
#ifdef RTHREADS
if (flags & FORK_THREAD) {
/* nothing */
} else
#endif
{
p2->p_p->ps_cred = pool_get(&pcred_pool, PR_WAITOK);
bcopy(p1->p_p->ps_cred, p2->p_p->ps_cred, sizeof(*p2->p_p->ps_cred));
p2->p_p->ps_cred->p_refcnt = 1;
crhold(p1->p_ucred);
}
TAILQ_INIT(&p2->p_selects);
/* 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_CLEANFILES)
p2->p_fd = fdinit(p1);
else if (flags & FORK_SHAREFILES)
p2->p_fd = fdshare(p1);
else
p2->p_fd = fdcopy(p1);
/*
* If ps_limit is still copy-on-write, bump refcnt,
* otherwise get a copy that won't be modified.
* (If PL_SHAREMOD is clear, the structure is shared
* copy-on-write.)
*/
#ifdef RTHREADS
if (flags & FORK_THREAD) {
/* nothing */
} else
#endif
{
if (p1->p_p->ps_limit->p_lflags & PL_SHAREMOD)
p2->p_p->ps_limit = limcopy(p1->p_p->ps_limit);
else {
p2->p_p->ps_limit = p1->p_p->ps_limit;
p2->p_p->ps_limit->p_refcnt++;
}
}
if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
atomic_setbits_int(&p2->p_flag, P_CONTROLT);
if (flags & FORK_PPWAIT)
atomic_setbits_int(&p2->p_flag, P_PPWAIT);
p2->p_pptr = p1;
if (flags & FORK_NOZOMBIE)
atomic_setbits_int(&p2->p_flag, P_NOZOMBIE);
LIST_INIT(&p2->p_children);
#ifdef KTRACE
/*
* Copy traceflag and tracefile if enabled.
* If not inherited, these were zeroed above.
*/
if (p1->p_traceflag & KTRFAC_INHERIT) {
p2->p_traceflag = p1->p_traceflag;
if ((p2->p_tracep = p1->p_tracep) != NULL)
VREF(p2->p_tracep);
}
#endif
/*
* set priority of child to be that of parent
* XXX should move p_estcpu into the region of struct proc which gets
* copied.
*/
scheduler_fork_hook(p1, p2);
/*
* Create signal actions for the child process.
*/
if (flags & FORK_SIGHAND)
sigactsshare(p1, p2);
else
p2->p_sigacts = sigactsinit(p1);
/*
* If emulation has process fork hook, call it now.
*/
if (p2->p_emul->e_proc_fork)
(*p2->p_emul->e_proc_fork)(p2, p1);
p2->p_addr = (struct user *)uaddr;
/*
* Finish creating the child process. It will return through a
* different path later.
*/
uvm_fork(p1, p2, ((flags & FORK_SHAREVM) ? TRUE : FALSE), stack,
stacksize, func ? func : child_return, arg ? arg : p2);
timeout_set(&p2->p_stats->p_virt_to, virttimer_trampoline, p2);
timeout_set(&p2->p_stats->p_prof_to, proftimer_trampoline, p2);
vm = p2->p_vmspace;
if (flags & FORK_FORK) {
forkstat.cntfork++;
forkstat.sizfork += vm->vm_dsize + vm->vm_ssize;
} else if (flags & FORK_VFORK) {
forkstat.cntvfork++;
forkstat.sizvfork += vm->vm_dsize + vm->vm_ssize;
} else if (flags & FORK_RFORK) {
forkstat.cntrfork++;
forkstat.sizrfork += vm->vm_dsize + vm->vm_ssize;
} else {
forkstat.cntkthread++;
forkstat.sizkthread += vm->vm_dsize + vm->vm_ssize;
}
/* Find an unused pid satisfying 1 <= lastpid <= PID_MAX */
do {
lastpid = 1 + (randompid ? arc4random() : lastpid) % PID_MAX;
} while (pidtaken(lastpid));
p2->p_pid = lastpid;
LIST_INSERT_HEAD(&allproc, p2, p_list);
LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
LIST_INSERT_HEAD(&p1->p_children, p2, p_sibling);
LIST_INSERT_AFTER(p1, p2, p_pglist);
if (p2->p_flag & P_TRACED) {
p2->p_oppid = p1->p_pid;
if (p2->p_pptr != p1->p_pptr)
proc_reparent(p2, p1->p_pptr);
/*
* Set ptrace status.
*/
if (flags & FORK_FORK) {
p2->p_ptstat = malloc(sizeof(*p2->p_ptstat),
M_SUBPROC, M_WAITOK);
p1->p_ptstat->pe_report_event = PTRACE_FORK;
p2->p_ptstat->pe_report_event = PTRACE_FORK;
p1->p_ptstat->pe_other_pid = p2->p_pid;
p2->p_ptstat->pe_other_pid = p1->p_pid;
}
}
#if NSYSTRACE > 0
if (ISSET(p1->p_flag, P_SYSTRACE))
systrace_fork(p1, p2);
#endif
/*
* Make child runnable, set start time, and add to run queue.
*/
SCHED_LOCK(s);
getmicrotime(&p2->p_stats->p_start);
p2->p_acflag = AFORK;
p2->p_stat = SRUN;
setrunqueue(p2);
SCHED_UNLOCK(s);
/*
* Notify any interested parties about the new process.
*/
KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
/*
* Update stats now that we know the fork was successfull.
*/
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;
/*
* 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).
*/
if (flags & FORK_PPWAIT)
while (p2->p_flag & P_PPWAIT)
tsleep(p1, PWAIT, "ppwait", 0);
/*
* If we're tracing the child, alert the parent too.
*/
if ((flags & FORK_PTRACE) && (p1->p_flag & P_TRACED))
psignal(p1, SIGTRAP);
/*
* Return child pid to parent process,
* marking us as parent via retval[1].
*/
if (retval != NULL) {
retval[0] = p2->p_pid;
retval[1] = 0;
}
return (0);
}
