/*
* 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;
}
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);
}
