int
recvit32(struct lwp *l, int s, struct netbsd32_msghdr *mp, struct iovec *iov, void *namelenp, register_t *retsize)
{
struct uio auio;
int i, len, error, iovlen;
struct mbuf *from = 0, *control = 0;
struct socket *so;
struct proc *p;
struct iovec *ktriov = NULL;
p = l->l_proc;
/* fd_getsock() will use the descriptor for us */
if ((error = fd_getsock(s, &so)) != 0)
return (error);
auio.uio_iov = iov;
auio.uio_iovcnt = mp->msg_iovlen;
auio.uio_rw = UIO_READ;
auio.uio_vmspace = l->l_proc->p_vmspace;
auio.uio_offset = 0; /* XXX */
auio.uio_resid = 0;
for (i = 0; i < mp->msg_iovlen; i++, iov++) {
#if 0
/* cannot happen iov_len is unsigned */
if (iov->iov_len < 0) {
error = EINVAL;
goto out1;
}
#endif
/*
* Reads return ssize_t because -1 is returned on error.
* Therefore we must restrict the length to SSIZE_MAX to
* avoid garbage return values.
*/
auio.uio_resid += iov->iov_len;
if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
error = EINVAL;
goto out1;
}
}
if (ktrpoint(KTR_GENIO)) {
iovlen = auio.uio_iovcnt * sizeof(struct iovec);
ktriov = (struct iovec *)malloc(iovlen, M_TEMP, M_WAITOK);
memcpy((void *)ktriov, (void *)auio.uio_iov, iovlen);
}
len = auio.uio_resid;
error = (*so->so_receive)(so, &from, &auio, NULL,
NETBSD32PTR64(mp->msg_control) ? &control : NULL,
&mp->msg_flags);
if (error) {
if (auio.uio_resid != len && (error == ERESTART ||
error == EINTR || error == EWOULDBLOCK))
error = 0;
}
if (ktriov != NULL) {
ktrgeniov(s, UIO_READ, ktriov, len - auio.uio_resid, error);
FREE(ktriov, M_TEMP);
}
if (error)
goto out;
*retsize = len - auio.uio_resid;
if (NETBSD32PTR64(mp->msg_name)) {
len = mp->msg_namelen;
if (len <= 0 || from == 0)
len = 0;
else {
if (len > from->m_len)
len = from->m_len;
/* else if len < from->m_len ??? */
error = copyout(mtod(from, void *),
(void *)NETBSD32PTR64(mp->msg_name),
(unsigned)len);
if (error)
goto out;
}
mp->msg_namelen = len;
if (namelenp &&
(error = copyout((void *)&len, namelenp, sizeof(int))))
goto out;
}
static int
do_sys_sendmsg_so(struct lwp *l, int s, struct socket *so, file_t *fp,
struct msghdr *mp, int flags, register_t *retsize)
{
struct iovec aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
struct mbuf *to, *control;
struct uio auio;
size_t len, iovsz;
int i, error;
ktrkuser("msghdr", mp, sizeof *mp);
/* If the caller passed us stuff in mbufs, we must free them. */
to = (mp->msg_flags & MSG_NAMEMBUF) ? mp->msg_name : NULL;
control = (mp->msg_flags & MSG_CONTROLMBUF) ? mp->msg_control : NULL;
iovsz = mp->msg_iovlen * sizeof(struct iovec);
if (mp->msg_flags & MSG_IOVUSRSPACE) {
if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
error = EMSGSIZE;
goto bad;
}
iov = kmem_alloc(iovsz, KM_SLEEP);
}
if (mp->msg_iovlen != 0) {
error = copyin(mp->msg_iov, iov, iovsz);
if (error)
goto bad;
}
mp->msg_iov = iov;
}
auio.uio_iov = mp->msg_iov;
auio.uio_iovcnt = mp->msg_iovlen;
auio.uio_rw = UIO_WRITE;
auio.uio_offset = 0; /* XXX */
auio.uio_resid = 0;
KASSERT(l == curlwp);
auio.uio_vmspace = l->l_proc->p_vmspace;
for (i = 0, tiov = mp->msg_iov; i < mp->msg_iovlen; i++, tiov++) {
/*
* Writes return ssize_t because -1 is returned on error.
* Therefore, we must restrict the length to SSIZE_MAX to
* avoid garbage return values.
*/
auio.uio_resid += tiov->iov_len;
if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
error = EINVAL;
goto bad;
}
}
if (mp->msg_name && to == NULL) {
error = sockargs(&to, mp->msg_name, mp->msg_namelen,
MT_SONAME);
if (error)
goto bad;
}
if (mp->msg_control) {
if (mp->msg_controllen < CMSG_ALIGN(sizeof(struct cmsghdr))) {
error = EINVAL;
goto bad;
}
if (control == NULL) {
error = sockargs(&control, mp->msg_control,
mp->msg_controllen, MT_CONTROL);
if (error)
goto bad;
}
}
if (ktrpoint(KTR_GENIO) && iovsz > 0) {
ktriov = kmem_alloc(iovsz, KM_SLEEP);
memcpy(ktriov, auio.uio_iov, iovsz);
}
if (mp->msg_name)
MCLAIM(to, so->so_mowner);
if (mp->msg_control)
MCLAIM(control, so->so_mowner);
len = auio.uio_resid;
error = (*so->so_send)(so, to, &auio, NULL, control, flags, l);
/* Protocol is responsible for freeing 'control' */
control = NULL;
if (error) {
if (auio.uio_resid != len && (error == ERESTART ||
error == EINTR || error == EWOULDBLOCK))
error = 0;
if (error == EPIPE && (fp->f_flag & FNOSIGPIPE) == 0 &&
(flags & MSG_NOSIGNAL) == 0) {
mutex_enter(proc_lock);
psignal(l->l_proc, SIGPIPE);
mutex_exit(proc_lock);
}
}
if (error == 0)
*retsize = len - auio.uio_resid;
bad:
if (ktriov != NULL) {
ktrgeniov(s, UIO_WRITE, ktriov, *retsize, error);
kmem_free(ktriov, iovsz);
}
if (iov != aiov)
kmem_free(iov, iovsz);
if (to)
m_freem(to);
if (control)
m_freem(control);
return error;
}
static int
do_sys_recvmsg_so(struct lwp *l, int s, struct socket *so, struct msghdr *mp,
struct mbuf **from, struct mbuf **control, register_t *retsize)
{
struct iovec aiov[UIO_SMALLIOV], *iov = aiov, *tiov, *ktriov = NULL;
struct uio auio;
size_t len, iovsz;
int i, error;
ktrkuser("msghdr", mp, sizeof *mp);
*from = NULL;
if (control != NULL)
*control = NULL;
iovsz = mp->msg_iovlen * sizeof(struct iovec);
if (mp->msg_flags & MSG_IOVUSRSPACE) {
if ((unsigned int)mp->msg_iovlen > UIO_SMALLIOV) {
if ((unsigned int)mp->msg_iovlen > IOV_MAX) {
error = EMSGSIZE;
goto out;
}
iov = kmem_alloc(iovsz, KM_SLEEP);
}
if (mp->msg_iovlen != 0) {
error = copyin(mp->msg_iov, iov, iovsz);
if (error)
goto out;
}
auio.uio_iov = iov;
} else
auio.uio_iov = mp->msg_iov;
auio.uio_iovcnt = mp->msg_iovlen;
auio.uio_rw = UIO_READ;
auio.uio_offset = 0; /* XXX */
auio.uio_resid = 0;
KASSERT(l == curlwp);
auio.uio_vmspace = l->l_proc->p_vmspace;
tiov = auio.uio_iov;
for (i = 0; i < mp->msg_iovlen; i++, tiov++) {
/*
* Reads return ssize_t because -1 is returned on error.
* Therefore we must restrict the length to SSIZE_MAX to
* avoid garbage return values.
*/
auio.uio_resid += tiov->iov_len;
if (tiov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
error = EINVAL;
goto out;
}
}
if (ktrpoint(KTR_GENIO) && iovsz > 0) {
ktriov = kmem_alloc(iovsz, KM_SLEEP);
memcpy(ktriov, auio.uio_iov, iovsz);
}
len = auio.uio_resid;
mp->msg_flags &= MSG_USERFLAGS;
error = (*so->so_receive)(so, from, &auio, NULL, control,
&mp->msg_flags);
len -= auio.uio_resid;
*retsize = len;
if (error != 0 && len != 0
&& (error == ERESTART || error == EINTR || error == EWOULDBLOCK))
/* Some data transferred */
error = 0;
if (ktriov != NULL) {
ktrgeniov(s, UIO_READ, ktriov, len, error);
kmem_free(ktriov, iovsz);
}
if (error != 0) {
m_freem(*from);
*from = NULL;
if (control != NULL) {
free_control_mbuf(l, *control, *control);
*control = NULL;
}
}
out:
if (iov != aiov)
kmem_free(iov, iovsz);
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;
}
