#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sysctl.h>
#include <sys/kernel.h>
#include <sys/file_internal.h>
#include <sys/resourcevar.h>
#include <sys/malloc.h>
#include <sys/proc_internal.h>
#include <sys/kauth.h>
#include <machine/spl.h>
#include <sys/mount_internal.h>
#include <sys/sysproto.h>
#include <bsm/audit_kernel.h>
#include <machine/vmparam.h>
#include <mach/mach_types.h>
#include <mach/time_value.h>
#include <mach/task.h>
#include <mach/task_info.h>
#include <mach/vm_map.h>
#include <mach/mach_vm.h>
#include <mach/thread_act.h>
#include <kern/lock.h>
#include <kern/thread.h>
#include <kern/task.h>
#include <kern/clock.h>
#include <netinet/in.h>
#include <sys/socketvar.h>
#include <vm/vm_map.h>
int donice(struct proc *curp, struct proc *chgp, int n);
int dosetrlimit(struct proc *p, u_int which, struct rlimit *limp);
static int do_background_thread(struct proc *curp, int priority);
rlim_t maxdmap = MAXDSIZ;
rlim_t maxsmap = MAXSSIZ - PAGE_SIZE;
int maxfilesperproc = OPEN_MAX;
SYSCTL_INT( _kern, KERN_MAXPROCPERUID, maxprocperuid, CTLFLAG_RW,
&maxprocperuid, 0, "Maximum processes allowed per userid" );
SYSCTL_INT( _kern, KERN_MAXFILESPERPROC, maxfilesperproc, CTLFLAG_RW,
&maxfilesperproc, 0, "Maximum files allowed open per process" );
struct puser_nice_args {
proc_t curp;
int prio;
id_t who;
int * foundp;
int * errorp;
};
static int puser_donice_callback(proc_t p, void * arg);
struct ppgrp_nice_args {
proc_t curp;
int prio;
int * foundp;
int * errorp;
};
static int ppgrp_donice_callback(proc_t p, void * arg);
int
getpriority(struct proc *curp, struct getpriority_args *uap, register_t *retval)
{
struct proc *p;
int low = PRIO_MAX + 1;
kauth_cred_t my_cred;
if (uap->who > 0x7fffffff)
return (EINVAL);
switch (uap->which) {
case PRIO_PROCESS:
if (uap->who == 0) {
p = curp;
low = p->p_nice;
} else {
p = proc_find(uap->who);
if (p == 0)
break;
low = p->p_nice;
proc_rele(p);
}
break;
case PRIO_PGRP: {
struct pgrp *pg = PGRP_NULL;
if (uap->who == 0) {
pg = proc_pgrp(curp);
} else if ((pg = pgfind(uap->who)) == PGRP_NULL) {
break;
}
pgrp_lock(pg);
for (p = pg->pg_members.lh_first; p != 0; p = p->p_pglist.le_next) {
if (p->p_nice < low)
low = p->p_nice;
}
pgrp_unlock(pg);
pg_rele(pg);
break;
}
case PRIO_USER:
if (uap->who == 0)
uap->who = kauth_cred_getuid(kauth_cred_get());
proc_list_lock();
for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
my_cred = kauth_cred_proc_ref(p);
if (kauth_cred_getuid(my_cred) == uap->who &&
p->p_nice < low)
low = p->p_nice;
kauth_cred_unref(&my_cred);
}
proc_list_unlock();
break;
case PRIO_DARWIN_THREAD: {
thread_t thread;
struct uthread *ut;
if (uap->who != 0) {
return (EINVAL);
}
thread = current_thread();
ut = get_bsdthread_info(thread);
low = 0;
if ( (ut->uu_flag & UT_BACKGROUND) != 0 ) {
low = 1;
}
break;
}
default:
return (EINVAL);
}
if (low == PRIO_MAX + 1)
return (ESRCH);
*retval = low;
return (0);
}
static int
puser_donice_callback(proc_t p, void * arg)
{
int error, n;
struct puser_nice_args * pun = (struct puser_nice_args *)arg;
kauth_cred_t my_cred;
my_cred = kauth_cred_proc_ref(p);
if (kauth_cred_getuid(my_cred) == pun->who) {
error = donice(pun->curp, p, pun->prio);
if (pun->errorp != NULL)
*pun->errorp = error;
if (pun->foundp != NULL) {
n = *pun->foundp;
*pun->foundp = n+1;
}
}
kauth_cred_unref(&my_cred);
return(PROC_RETURNED);
}
static int
ppgrp_donice_callback(proc_t p, void * arg)
{
int error;
struct ppgrp_nice_args * pun = (struct ppgrp_nice_args *)arg;
int n;
error = donice(pun->curp, p, pun->prio);
if (pun->errorp != NULL)
*pun->errorp = error;
if (pun->foundp!= NULL) {
n = *pun->foundp;
*pun->foundp = n+1;
}
return(PROC_RETURNED);
}
int
setpriority(struct proc *curp, struct setpriority_args *uap, __unused register_t *retval)
{
struct proc *p;
int found = 0, error = 0;
int refheld = 0;
AUDIT_ARG(cmd, uap->which);
AUDIT_ARG(owner, uap->who, 0);
AUDIT_ARG(value, uap->prio);
if (uap->who > 0x7fffffff)
return (EINVAL);
switch (uap->which) {
case PRIO_PROCESS:
if (uap->who == 0)
p = curp;
else {
p = proc_find(uap->who);
if (p == 0)
break;
refheld = 1;
}
error = donice(curp, p, uap->prio);
found++;
if (refheld != 0)
proc_rele(p);
break;
case PRIO_PGRP: {
struct pgrp *pg = PGRP_NULL;
struct ppgrp_nice_args ppgrp;
if (uap->who == 0) {
pg = proc_pgrp(curp);
} else if ((pg = pgfind(uap->who)) == PGRP_NULL)
break;
ppgrp.curp = curp;
ppgrp.prio = uap->prio;
ppgrp.foundp = &found;
ppgrp.errorp = &error;
pgrp_iterate(pg, PGRP_DROPREF, ppgrp_donice_callback, (void *)&ppgrp, NULL, NULL);
break;
}
case PRIO_USER: {
struct puser_nice_args punice;
if (uap->who == 0)
uap->who = kauth_cred_getuid(kauth_cred_get());
punice.curp = curp;
punice.prio = uap->prio;
punice.who = uap->who;
punice.foundp = &found;
error = 0;
punice.errorp = &error;
proc_iterate(PROC_ALLPROCLIST, puser_donice_callback, (void *)&punice, NULL, NULL);
break;
}
case PRIO_DARWIN_THREAD: {
if (uap->who != 0) {
return (EINVAL);
}
error = do_background_thread(curp, uap->prio);
found++;
break;
}
default:
return (EINVAL);
}
if (found == 0)
return (ESRCH);
return (error);
}
int
donice(struct proc *curp, struct proc *chgp, int n)
{
int error = 0;
kauth_cred_t ucred;
kauth_cred_t my_cred;
ucred = kauth_cred_proc_ref(curp);
my_cred = kauth_cred_proc_ref(chgp);
if (suser(ucred, NULL) && ucred->cr_ruid &&
kauth_cred_getuid(ucred) != kauth_cred_getuid(my_cred) &&
ucred->cr_ruid != kauth_cred_getuid(my_cred)) {
error = EPERM;
goto out;
}
if (n > PRIO_MAX)
n = PRIO_MAX;
if (n < PRIO_MIN)
n = PRIO_MIN;
if (n < chgp->p_nice && suser(ucred, &curp->p_acflag)) {
error = EACCES;
goto out;
}
#if CONFIG_MACF
error = mac_proc_check_sched(curp, chgp);
if (error)
goto out;
#endif
proc_lock(chgp);
chgp->p_nice = n;
proc_unlock(chgp);
(void)resetpriority(chgp);
out:
kauth_cred_unref(&ucred);
kauth_cred_unref(&my_cred);
return (error);
}
static int
do_background_thread(struct proc *curp, int priority)
{
int i;
thread_t thread;
struct uthread *ut;
thread_precedence_policy_data_t policy;
struct filedesc *fdp;
struct fileproc *fp;
thread = current_thread();
ut = get_bsdthread_info(thread);
if ( (priority & PRIO_DARWIN_BG) == 0 ) {
if ( (ut->uu_flag & UT_BACKGROUND) == 0 ) {
return(0);
}
ut->uu_flag &= ~UT_BACKGROUND;
ut->uu_iopol_disk = IOPOL_NORMAL;
policy.importance = 0;
thread_policy_set( thread, THREAD_PRECEDENCE_POLICY,
(thread_policy_t)&policy,
THREAD_PRECEDENCE_POLICY_COUNT );
proc_fdlock(curp);
fdp = curp->p_fd;
for ( i = 0; i < fdp->fd_nfiles; i++ ) {
struct socket *sockp;
fp = fdp->fd_ofiles[ i ];
if ( fp == NULL || (fdp->fd_ofileflags[ i ] & UF_RESERVED) != 0 ||
fp->f_fglob->fg_type != DTYPE_SOCKET ) {
continue;
}
sockp = (struct socket *)fp->f_fglob->fg_data;
if ( sockp->so_background_thread != thread ) {
continue;
}
sockp->so_traffic_mgt_flags &= ~TRAFFIC_MGT_SO_BACKGROUND;
sockp->so_background_thread = NULL;
}
proc_fdunlock(curp);
return(0);
}
if ( (ut->uu_flag & UT_BACKGROUND) != 0 ) {
return(0);
}
ut->uu_flag |= UT_BACKGROUND;
ut->uu_iopol_disk = IOPOL_THROTTLE;
policy.importance = INT_MIN;
thread_policy_set( thread, THREAD_PRECEDENCE_POLICY,
(thread_policy_t)&policy,
THREAD_PRECEDENCE_POLICY_COUNT );
return(0);
}
int
setrlimit(struct proc *p, struct setrlimit_args *uap, __unused register_t *retval)
{
struct rlimit alim;
int error;
if ((error = copyin(uap->rlp, (caddr_t)&alim,
sizeof (struct rlimit))))
return (error);
return (dosetrlimit(p, uap->which, &alim));
}
int
dosetrlimit(struct proc *p, u_int which, struct rlimit *limp)
{
struct rlimit *alimp;
int error;
kern_return_t kr;
int posix = (which & _RLIMIT_POSIX_FLAG) ? 1 : 0;
which &= ~_RLIMIT_POSIX_FLAG;
if (which >= RLIM_NLIMITS)
return (EINVAL);
alimp = &p->p_rlimit[which];
if (limp->rlim_cur > limp->rlim_max)
return EINVAL;
if (limp->rlim_cur > alimp->rlim_max ||
limp->rlim_max > alimp->rlim_max)
if ((error = suser(kauth_cred_get(), &p->p_acflag))) {
return (error);
}
proc_limitblock(p);
if ((error = proc_limitreplace(p)) != 0) {
proc_limitunblock(p);
return(error);
}
alimp = &p->p_rlimit[which];
switch (which) {
case RLIMIT_CPU:
if (limp->rlim_cur == RLIM_INFINITY) {
task_vtimer_clear(p->task, TASK_VTIMER_RLIM);
timerclear(&p->p_rlim_cpu);
}
else {
task_absolutetime_info_data_t tinfo;
mach_msg_type_number_t count;
struct timeval ttv, tv;
count = TASK_ABSOLUTETIME_INFO_COUNT;
task_info(p->task, TASK_ABSOLUTETIME_INFO,
(task_info_t)&tinfo, &count);
absolutetime_to_microtime(tinfo.total_user + tinfo.total_system,
(uint32_t *) &ttv.tv_sec, (uint32_t *) &ttv.tv_usec);
tv.tv_sec = (limp->rlim_cur > __INT_MAX__ ? __INT_MAX__ : limp->rlim_cur);
tv.tv_usec = 0;
timersub(&tv, &ttv, &p->p_rlim_cpu);
timerclear(&tv);
if (timercmp(&p->p_rlim_cpu, &tv, >))
task_vtimer_set(p->task, TASK_VTIMER_RLIM);
else {
task_vtimer_clear(p->task, TASK_VTIMER_RLIM);
timerclear(&p->p_rlim_cpu);
psignal(p, SIGXCPU);
}
}
break;
case RLIMIT_DATA:
if (limp->rlim_cur > maxdmap)
limp->rlim_cur = maxdmap;
if (limp->rlim_max > maxdmap)
limp->rlim_max = maxdmap;
break;
case RLIMIT_STACK:
if (limp->rlim_cur > maxsmap ||
limp->rlim_max > maxsmap) {
if (posix) {
error = EINVAL;
goto out;
}
else {
if (limp->rlim_cur > maxsmap)
limp->rlim_cur = maxsmap;
if (limp->rlim_max > maxsmap)
limp->rlim_max = maxsmap;
}
}
if (limp->rlim_cur > alimp->rlim_cur) {
user_addr_t addr;
user_size_t size;
size = round_page_64(limp->rlim_cur);
size -= round_page_64(alimp->rlim_cur);
#if STACK_GROWTH_UP
addr = p->user_stack + round_page_64(alimp->rlim_cur);
#else
addr = p->user_stack - round_page_64(limp->rlim_cur);
#endif
kr = mach_vm_protect(current_map(),
addr, size,
FALSE, VM_PROT_DEFAULT);
if (kr != KERN_SUCCESS) {
error = EINVAL;
goto out;
}
} else if (limp->rlim_cur < alimp->rlim_cur) {
user_addr_t addr;
user_size_t size;
user_addr_t cur_sp;
cur_sp = thread_adjuserstack(current_thread(),
0);
#if STACK_GROWTH_UP
if (cur_sp >= p->user_stack &&
cur_sp < (p->user_stack +
round_page_64(alimp->rlim_cur))) {
if (cur_sp >= (p->user_stack +
round_page_64(limp->rlim_cur))) {
error = EINVAL;
goto out;
}
} else {
error = EINVAL;
goto out;
}
#else
if (cur_sp <= p->user_stack &&
cur_sp > (p->user_stack -
round_page_64(alimp->rlim_cur))) {
if (cur_sp <= (p->user_stack -
round_page_64(limp->rlim_cur))) {
error = EINVAL;
goto out;
}
} else {
error = EINVAL;
goto out;
}
#endif
size = round_page_64(alimp->rlim_cur);
size -= round_page_64(limp->rlim_cur);
#if STACK_GROWTH_UP
addr = p->user_stack + round_page_64(limp->rlim_cur);
#else
addr = p->user_stack - round_page_64(alimp->rlim_cur);
#endif
kr = mach_vm_protect(current_map(),
addr, size,
FALSE, VM_PROT_NONE);
if (kr != KERN_SUCCESS) {
error = EINVAL;
goto out;
}
} else {
}
break;
case RLIMIT_NOFILE:
if ( is_suser() ) {
if (limp->rlim_cur != alimp->rlim_cur &&
limp->rlim_cur > (rlim_t)maxfiles) {
if (posix) {
error = EINVAL;
goto out;
}
limp->rlim_cur = maxfiles;
}
if (limp->rlim_max != alimp->rlim_max &&
limp->rlim_max > (rlim_t)maxfiles)
limp->rlim_max = maxfiles;
}
else {
if (limp->rlim_cur != alimp->rlim_cur &&
limp->rlim_cur > (rlim_t)maxfilesperproc) {
if (posix) {
error = EINVAL;
goto out;
}
limp->rlim_cur = maxfilesperproc;
}
if (limp->rlim_max != alimp->rlim_max &&
limp->rlim_max > (rlim_t)maxfilesperproc)
limp->rlim_max = maxfilesperproc;
}
break;
case RLIMIT_NPROC:
if ( is_suser() ) {
if (limp->rlim_cur > (rlim_t)maxproc)
limp->rlim_cur = maxproc;
if (limp->rlim_max > (rlim_t)maxproc)
limp->rlim_max = maxproc;
}
else {
if (limp->rlim_cur > (rlim_t)maxprocperuid)
limp->rlim_cur = maxprocperuid;
if (limp->rlim_max > (rlim_t)maxprocperuid)
limp->rlim_max = maxprocperuid;
}
break;
case RLIMIT_MEMLOCK:
vm_map_set_user_wire_limit(current_map(), limp->rlim_cur);
break;
}
proc_lock(p);
*alimp = *limp;
proc_unlock(p);
error = 0;
out:
proc_limitunblock(p);
return (error);
}
int
getrlimit(struct proc *p, struct getrlimit_args *uap, __unused register_t *retval)
{
struct rlimit lim;
uap->which &= ~_RLIMIT_POSIX_FLAG;
if (uap->which >= RLIM_NLIMITS)
return (EINVAL);
proc_limitget(p, uap->which, &lim);
return (copyout((caddr_t)&lim,
uap->rlp, sizeof (struct rlimit)));
}
void
calcru(struct proc *p, struct timeval *up, struct timeval *sp, struct timeval *ip)
{
task_t task;
timerclear(up);
timerclear(sp);
if (ip != NULL)
timerclear(ip);
task = p->task;
if (task) {
task_basic_info_data_t tinfo;
task_thread_times_info_data_t ttimesinfo;
mach_msg_type_number_t task_info_stuff, task_ttimes_stuff;
struct timeval ut,st;
task_info_stuff = TASK_BASIC_INFO_COUNT;
task_info(task, TASK_BASIC_INFO,
(task_info_t)&tinfo, &task_info_stuff);
ut.tv_sec = tinfo.user_time.seconds;
ut.tv_usec = tinfo.user_time.microseconds;
st.tv_sec = tinfo.system_time.seconds;
st.tv_usec = tinfo.system_time.microseconds;
timeradd(&ut, up, up);
timeradd(&st, sp, sp);
task_ttimes_stuff = TASK_THREAD_TIMES_INFO_COUNT;
task_info(task, TASK_THREAD_TIMES_INFO,
(task_info_t)&ttimesinfo, &task_ttimes_stuff);
ut.tv_sec = ttimesinfo.user_time.seconds;
ut.tv_usec = ttimesinfo.user_time.microseconds;
st.tv_sec = ttimesinfo.system_time.seconds;
st.tv_usec = ttimesinfo.system_time.microseconds;
timeradd(&ut, up, up);
timeradd(&st, sp, sp);
}
}
__private_extern__ void munge_rusage(struct rusage *a_rusage_p, struct user_rusage *a_user_rusage_p);
int
getrusage(struct proc *p, struct getrusage_args *uap, __unused register_t *retval)
{
struct rusage *rup, rubuf;
struct user_rusage rubuf64;
size_t retsize = sizeof(rubuf);
caddr_t retbuf = (caddr_t)&rubuf;
struct timeval utime;
struct timeval stime;
switch (uap->who) {
case RUSAGE_SELF:
calcru(p, &utime, &stime, NULL);
proc_lock(p);
rup = &p->p_stats->p_ru;
rup->ru_utime = utime;
rup->ru_stime = stime;
rubuf = *rup;
proc_unlock(p);
break;
case RUSAGE_CHILDREN:
proc_lock(p);
rup = &p->p_stats->p_cru;
rubuf = *rup;
proc_unlock(p);
break;
default:
return (EINVAL);
}
if (IS_64BIT_PROCESS(p)) {
retsize = sizeof(rubuf64);
retbuf = (caddr_t)&rubuf64;
munge_rusage(&rubuf, &rubuf64);
}
return (copyout(retbuf, uap->rusage, retsize));
}
void
ruadd(struct rusage *ru, struct rusage *ru2)
{
long *ip, *ip2;
int i;
timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
if (ru->ru_maxrss < ru2->ru_maxrss)
ru->ru_maxrss = ru2->ru_maxrss;
ip = &ru->ru_first; ip2 = &ru2->ru_first;
for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
*ip++ += *ip2++;
}
void
proc_limitget(proc_t p, int which, struct rlimit * limp)
{
proc_list_lock();
limp->rlim_cur = p->p_rlimit[which].rlim_cur;
limp->rlim_max = p->p_rlimit[which].rlim_max;
proc_list_unlock();
}
void
proc_limitdrop(proc_t p, int exiting)
{
struct plimit * freelim = NULL;
struct plimit * freeoldlim = NULL;
proc_list_lock();
if (--p->p_limit->pl_refcnt == 0) {
freelim = p->p_limit;
p->p_limit = NULL;
}
if ((exiting != 0) && (p->p_olimit != NULL) && (--p->p_olimit->pl_refcnt == 0)) {
freeoldlim = p->p_olimit;
p->p_olimit = NULL;
}
proc_list_unlock();
if (freelim != NULL)
FREE_ZONE(freelim, sizeof *p->p_limit, M_PLIMIT);
if (freeoldlim != NULL)
FREE_ZONE(freeoldlim, sizeof *p->p_olimit, M_PLIMIT);
}
void
proc_limitfork(proc_t parent, proc_t child)
{
proc_list_lock();
child->p_limit = parent->p_limit;
child->p_limit->pl_refcnt++;
child->p_olimit = NULL;
proc_list_unlock();
}
void
proc_limitblock(proc_t p)
{
proc_lock(p);
while (p->p_lflag & P_LLIMCHANGE) {
p->p_lflag |= P_LLIMWAIT;
msleep(&p->p_olimit, &p->p_mlock, 0, "proc_limitblock", NULL);
}
p->p_lflag |= P_LLIMCHANGE;
proc_unlock(p);
}
void
proc_limitunblock(proc_t p)
{
proc_lock(p);
p->p_lflag &= ~P_LLIMCHANGE;
if (p->p_lflag & P_LLIMWAIT) {
p->p_lflag &= ~P_LLIMWAIT;
wakeup(&p->p_olimit);
}
proc_unlock(p);
}
int
proc_limitreplace(proc_t p)
{
struct plimit *copy;
proc_list_lock();
if (p->p_limit->pl_refcnt == 1) {
proc_list_unlock();
return(0);
}
proc_list_unlock();
MALLOC_ZONE(copy, struct plimit *,
sizeof(struct plimit), M_PLIMIT, M_WAITOK);
if (copy == NULL) {
return(ENOMEM);
}
proc_list_lock();
bcopy(p->p_limit->pl_rlimit, copy->pl_rlimit,
sizeof(struct rlimit) * RLIM_NLIMITS);
copy->pl_refcnt = 1;
p->p_olimit = p->p_limit;
p->p_limit = copy;
proc_list_unlock();
return(0);
}
int
iopolicysys(__unused struct proc *p, __unused struct iopolicysys_args *uap, __unused register_t *retval)
{
int error = 0;
thread_t thread = THREAD_NULL;
int *policy;
struct uthread *ut = NULL;
struct _iopol_param_t iop_param;
if ((error = copyin(uap->arg, &iop_param, sizeof(iop_param))) != 0)
goto exit;
if (iop_param.iop_iotype != IOPOL_TYPE_DISK) {
error = EINVAL;
goto exit;
}
switch (iop_param.iop_scope) {
case IOPOL_SCOPE_PROCESS:
policy = &p->p_iopol_disk;
break;
case IOPOL_SCOPE_THREAD:
thread = current_thread();
ut = get_bsdthread_info(thread);
policy = &ut->uu_iopol_disk;
break;
default:
error = EINVAL;
goto exit;
}
switch(uap->cmd) {
case IOPOL_CMD_SET:
switch (iop_param.iop_policy) {
case IOPOL_DEFAULT:
case IOPOL_NORMAL:
case IOPOL_THROTTLE:
case IOPOL_PASSIVE:
proc_lock(p);
*policy = iop_param.iop_policy;
proc_unlock(p);
break;
default:
error = EINVAL;
goto exit;
}
break;
case IOPOL_CMD_GET:
switch (*policy) {
case IOPOL_DEFAULT:
case IOPOL_NORMAL:
case IOPOL_THROTTLE:
case IOPOL_PASSIVE:
iop_param.iop_policy = *policy;
break;
default: printf("%s: unknown I/O policy %d\n", __func__, *policy);
*policy = IOPOL_DEFAULT;
iop_param.iop_policy = *policy;
}
error = copyout((caddr_t)&iop_param, uap->arg, sizeof(iop_param));
break;
default:
error = EINVAL; break;
}
exit:
*retval = error;
return (error);
}