struct tmpfs_node *
tmpfs_node_get(struct tmpfs_mount *mp)
{
if (atomic_inc_uint_nv(&mp->tm_nodes_cnt) >= mp->tm_nodes_max) {
atomic_dec_uint(&mp->tm_nodes_cnt);
return NULL;
}
if (!tmpfs_mem_incr(mp, sizeof(struct tmpfs_node))) {
atomic_dec_uint(&mp->tm_nodes_cnt);
return NULL;
}
return pool_get(&tmpfs_node_pool, PR_WAITOK);
}
/*
* data contains amount of time to sleep, in milliseconds
*/
static int
filt_timerattach(struct knote *kn)
{
callout_t *calloutp;
struct kqueue *kq;
int tticks;
tticks = mstohz(kn->kn_sdata);
/* if the supplied value is under our resolution, use 1 tick */
if (tticks == 0) {
if (kn->kn_sdata == 0)
return EINVAL;
tticks = 1;
}
if (atomic_inc_uint_nv(&kq_ncallouts) >= kq_calloutmax ||
(calloutp = kmem_alloc(sizeof(*calloutp), KM_NOSLEEP)) == NULL) {
atomic_dec_uint(&kq_ncallouts);
return ENOMEM;
}
callout_init(calloutp, CALLOUT_MPSAFE);
kq = kn->kn_kq;
mutex_spin_enter(&kq->kq_lock);
kn->kn_flags |= EV_CLEAR; /* automatically set */
kn->kn_hook = calloutp;
mutex_spin_exit(&kq->kq_lock);
callout_reset(calloutp, tticks, filt_timerexpire, kn);
return (0);
}
static int
tegra_cpufreq_freq_helper(SYSCTLFN_ARGS)
{
struct sysctlnode node;
int fq, oldfq = 0, error;
uint64_t xc;
node = *rnode;
node.sysctl_data = &fq;
fq = cpufreq_get_rate();
if (rnode->sysctl_num == cpufreq_node_target)
oldfq = fq;
error = sysctl_lookup(SYSCTLFN_CALL(&node));
if (error || newp == NULL)
return error;
if (fq == oldfq || rnode->sysctl_num != cpufreq_node_target)
return 0;
if (atomic_cas_uint(&cpufreq_busy, 0, 1) != 0)
return EBUSY;
error = cpufreq_set_rate(fq);
if (error == 0) {
xc = xc_broadcast(0, tegra_cpufreq_post, NULL, NULL);
xc_wait(xc);
pmf_event_inject(NULL, PMFE_SPEED_CHANGED);
}
atomic_dec_uint(&cpufreq_busy);
return error;
}
void
tmpfs_node_put(struct tmpfs_mount *mp, struct tmpfs_node *tn)
{
atomic_dec_uint(&mp->tm_nodes_cnt);
tmpfs_mem_decr(mp, sizeof(struct tmpfs_node));
pool_put(&tmpfs_node_pool, tn);
}
/*
* The clean design of if_clone and autoconf(9) makes that part
* really straightforward. The second argument of config_detach
* means neither QUIET nor FORCED.
*/
static int
tap_clone_destroy(struct ifnet *ifp)
{
struct tap_softc *sc = ifp->if_softc;
int error = tap_clone_destroyer(sc->sc_dev);
if (error == 0)
atomic_dec_uint(&tap_count);
return error;
}
static void
filt_timerdetach(struct knote *kn)
{
callout_t *calloutp;
calloutp = (callout_t *)kn->kn_hook;
callout_halt(calloutp, NULL);
callout_destroy(calloutp);
kmem_free(calloutp, sizeof(*calloutp));
atomic_dec_uint(&kq_ncallouts);
}
static int
mpls_clone_destroy(struct ifnet *ifp)
{
int s;
bpf_detach(ifp);
s = splnet();
if_detach(ifp);
splx(s);
free(ifp->if_softc, M_DEVBUF);
atomic_dec_uint(&mpls_count);
return 0;
}
void
iommulib_nex_close(dev_info_t *rdip)
{
iommulib_unit_t *unitp;
const char *driver;
int instance;
uint32_t unitid;
iommulib_nex_t *nexp;
const char *f = "iommulib_nex_close";
ASSERT(IOMMU_USED(rdip));
unitp = DEVI(rdip)->devi_iommulib_handle;
mutex_enter(&iommulib_lock);
mutex_enter(&unitp->ilu_lock);
nexp = (iommulib_nex_t *)unitp->ilu_nex;
DEVI(rdip)->devi_iommulib_handle = NULL;
unitid = unitp->ilu_unitid;
driver = ddi_driver_name(unitp->ilu_dip);
instance = ddi_get_instance(unitp->ilu_dip);
unitp->ilu_ref--;
mutex_exit(&unitp->ilu_lock);
mutex_exit(&iommulib_lock);
atomic_dec_uint(&nexp->nex_ref);
if (iommulib_debug) {
char *buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
(void) ddi_pathname(rdip, buf);
cmn_err(CE_NOTE, "%s: %s%d: closing IOMMU for dip (%p), "
"unitid=%u rdip path = %s", f, driver, instance,
(void *)rdip, unitid, buf);
kmem_free(buf, MAXPATHLEN);
}
}
/*
* 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;
}
static int
linux_clone_nptl(struct lwp *l, const struct linux_sys_clone_args *uap, register_t *retval)
{
/* {
syscallarg(int) flags;
syscallarg(void *) stack;
syscallarg(void *) parent_tidptr;
syscallarg(void *) tls;
syscallarg(void *) child_tidptr;
} */
struct proc *p;
struct lwp *l2;
struct linux_emuldata *led;
void *parent_tidptr, *tls, *child_tidptr;
struct schedstate_percpu *spc;
vaddr_t uaddr;
lwpid_t lid;
int flags, tnprocs, error;
p = l->l_proc;
flags = SCARG(uap, flags);
parent_tidptr = SCARG(uap, parent_tidptr);
tls = SCARG(uap, tls);
child_tidptr = SCARG(uap, child_tidptr);
tnprocs = atomic_inc_uint_nv(&nprocs);
if (__predict_false(tnprocs >= maxproc) ||
kauth_authorize_process(l->l_cred, KAUTH_PROCESS_FORK, p,
KAUTH_ARG(tnprocs), NULL, NULL) != 0) {
atomic_dec_uint(&nprocs);
return EAGAIN;
}
uaddr = uvm_uarea_alloc();
if (__predict_false(uaddr == 0)) {
atomic_dec_uint(&nprocs);
return ENOMEM;
}
error = lwp_create(l, p, uaddr, LWP_DETACHED | LWP_PIDLID,
SCARG(uap, stack), 0, child_return, NULL, &l2, l->l_class);
if (__predict_false(error)) {
DPRINTF(("%s: lwp_create error=%d\n", __func__, error));
atomic_dec_uint(&nprocs);
uvm_uarea_free(uaddr);
return error;
}
lid = l2->l_lid;
/* LINUX_CLONE_CHILD_CLEARTID: clear TID in child's memory on exit() */
if (flags & LINUX_CLONE_CHILD_CLEARTID) {
led = l2->l_emuldata;
led->led_clear_tid = child_tidptr;
}
/* LINUX_CLONE_PARENT_SETTID: store child's TID in parent's memory */
if (flags & LINUX_CLONE_PARENT_SETTID) {
if ((error = copyout(&lid, parent_tidptr, sizeof(lid))) != 0)
printf("%s: LINUX_CLONE_PARENT_SETTID "
"failed (parent_tidptr = %p tid = %d error=%d)\n",
__func__, parent_tidptr, lid, error);
}
/* LINUX_CLONE_CHILD_SETTID: store child's TID in child's memory */
if (flags & LINUX_CLONE_CHILD_SETTID) {
if ((error = copyout(&lid, child_tidptr, sizeof(lid))) != 0)
printf("%s: LINUX_CLONE_CHILD_SETTID "
"failed (child_tidptr = %p, tid = %d error=%d)\n",
__func__, child_tidptr, lid, error);
}
if (flags & LINUX_CLONE_SETTLS) {
error = LINUX_LWP_SETPRIVATE(l2, tls);
if (error) {
DPRINTF(("%s: LINUX_LWP_SETPRIVATE %d\n", __func__,
error));
lwp_exit(l2);
return error;
}
}
/*
* Set the new LWP running, unless the process is stopping,
* then the LWP is created stopped.
*/
mutex_enter(p->p_lock);
lwp_lock(l2);
spc = &l2->l_cpu->ci_schedstate;
if ((l->l_flag & (LW_WREBOOT | LW_WSUSPEND | LW_WEXIT)) == 0) {
if (p->p_stat == SSTOP || (p->p_sflag & PS_STOPPING) != 0) {
KASSERT(l2->l_wchan == NULL);
l2->l_stat = LSSTOP;
p->p_nrlwps--;
lwp_unlock_to(l2, spc->spc_lwplock);
} else {
KASSERT(lwp_locked(l2, spc->spc_mutex));
l2->l_stat = LSRUN;
sched_enqueue(l2, false);
lwp_unlock(l2);
}
} else {
l2->l_stat = LSSUSPENDED;
p->p_nrlwps--;
lwp_unlock_to(l2, spc->spc_lwplock);
}
mutex_exit(p->p_lock);
retval[0] = lid;
retval[1] = 0;
return 0;
}