freebsd-dev/sys/kern/kern_resource.c

644 lines
15 KiB
C
Raw Normal View History

1994-05-24 10:09:53 +00:00
/*-
* Copyright (c) 1982, 1986, 1991, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)kern_resource.c 8.5 (Berkeley) 1/21/94
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
* $Id: kern_resource.c,v 1.47 1999/04/27 12:21:07 phk Exp $
1994-05-24 10:09:53 +00:00
*/
#include "opt_compat.h"
#include "opt_rlimit.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sysproto.h>
1994-05-24 10:09:53 +00:00
#include <sys/kernel.h>
#include <sys/file.h>
#include <sys/resourcevar.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/vm_prot.h>
#include <sys/lock.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
1994-05-24 10:09:53 +00:00
1998-02-09 06:11:36 +00:00
static int donice __P((struct proc *curp, struct proc *chgp, int n));
/* dosetrlimit non-static: Needed by SysVR4 emulator */
int dosetrlimit __P((struct proc *p, u_int which, struct rlimit *limp));
1994-05-24 10:09:53 +00:00
/*
* Resource controls and accounting.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct getpriority_args {
int which;
int who;
};
#endif
int
getpriority(curp, uap)
1994-05-24 10:09:53 +00:00
struct proc *curp;
register struct getpriority_args *uap;
{
register struct proc *p;
1994-05-24 10:09:53 +00:00
register int low = PRIO_MAX + 1;
switch (uap->which) {
case PRIO_PROCESS:
if (uap->who == 0)
p = curp;
else
p = pfind(uap->who);
if (p == 0)
break;
low = p->p_nice;
break;
case PRIO_PGRP: {
register struct pgrp *pg;
if (uap->who == 0)
pg = curp->p_pgrp;
else if ((pg = pgfind(uap->who)) == NULL)
break;
for (p = pg->pg_members.lh_first; p != 0;
p = p->p_pglist.le_next) {
1994-05-24 10:09:53 +00:00
if (p->p_nice < low)
low = p->p_nice;
}
break;
}
case PRIO_USER:
if (uap->who == 0)
uap->who = curp->p_ucred->cr_uid;
for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
1994-05-24 10:09:53 +00:00
if (p->p_ucred->cr_uid == uap->who &&
p->p_nice < low)
low = p->p_nice;
break;
default:
return (EINVAL);
}
if (low == PRIO_MAX + 1)
return (ESRCH);
curp->p_retval[0] = low;
1994-05-24 10:09:53 +00:00
return (0);
}
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct setpriority_args {
int which;
int who;
int prio;
};
#endif
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
setpriority(curp, uap)
1994-05-24 10:09:53 +00:00
struct proc *curp;
register struct setpriority_args *uap;
{
register struct proc *p;
int found = 0, error = 0;
switch (uap->which) {
case PRIO_PROCESS:
if (uap->who == 0)
p = curp;
else
p = pfind(uap->who);
if (p == 0)
break;
error = donice(curp, p, uap->prio);
found++;
break;
case PRIO_PGRP: {
register struct pgrp *pg;
1995-05-30 08:16:23 +00:00
1994-05-24 10:09:53 +00:00
if (uap->who == 0)
pg = curp->p_pgrp;
else if ((pg = pgfind(uap->who)) == NULL)
break;
for (p = pg->pg_members.lh_first; p != 0;
p = p->p_pglist.le_next) {
1994-05-24 10:09:53 +00:00
error = donice(curp, p, uap->prio);
found++;
}
break;
}
case PRIO_USER:
if (uap->who == 0)
uap->who = curp->p_ucred->cr_uid;
for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
1994-05-24 10:09:53 +00:00
if (p->p_ucred->cr_uid == uap->who) {
error = donice(curp, p, uap->prio);
found++;
}
break;
default:
return (EINVAL);
}
if (found == 0)
return (ESRCH);
return (error);
}
1998-02-09 06:11:36 +00:00
static int
1994-05-24 10:09:53 +00:00
donice(curp, chgp, n)
register struct proc *curp, *chgp;
register int n;
{
register struct pcred *pcred = curp->p_cred;
if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
pcred->p_ruid != chgp->p_ucred->cr_uid)
return (EPERM);
if (n > PRIO_MAX)
n = PRIO_MAX;
if (n < PRIO_MIN)
n = PRIO_MIN;
if (n < chgp->p_nice && suser(curp))
1994-05-24 10:09:53 +00:00
return (EACCES);
chgp->p_nice = n;
(void)resetpriority(chgp);
return (0);
}
/* rtprio system call */
#ifndef _SYS_SYSPROTO_H_
struct rtprio_args {
int function;
pid_t pid;
struct rtprio *rtp;
};
#endif
/*
* Set realtime priority
*/
/* ARGSUSED */
int
rtprio(curp, uap)
struct proc *curp;
register struct rtprio_args *uap;
{
register struct proc *p;
register struct pcred *pcred = curp->p_cred;
struct rtprio rtp;
int error;
error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
if (error)
return (error);
if (uap->pid == 0)
p = curp;
else
p = pfind(uap->pid);
if (p == 0)
return (ESRCH);
switch (uap->function) {
case RTP_LOOKUP:
return (copyout(&p->p_rtprio, uap->rtp, sizeof(struct rtprio)));
case RTP_SET:
if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
pcred->pc_ucred->cr_uid != p->p_ucred->cr_uid &&
pcred->p_ruid != p->p_ucred->cr_uid)
return (EPERM);
/* disallow setting rtprio in most cases if not superuser */
if (suser(curp)) {
/* can't set someone else's */
if (uap->pid)
1995-05-30 08:16:23 +00:00
return (EPERM);
/* can't set realtime priority */
/*
* Realtime priority has to be restricted for reasons which should be
* obvious. However, for idle priority, there is a potential for
* system deadlock if an idleprio process gains a lock on a resource
* that other processes need (and the idleprio process can't run
* due to a CPU-bound normal process). Fix me! XXX
*/
#if 0
if (RTP_PRIO_IS_REALTIME(rtp.type))
#endif
if (rtp.type != RTP_PRIO_NORMAL)
return (EPERM);
}
switch (rtp.type) {
#ifdef RTP_PRIO_FIFO
case RTP_PRIO_FIFO:
#endif
case RTP_PRIO_REALTIME:
case RTP_PRIO_NORMAL:
case RTP_PRIO_IDLE:
if (rtp.prio > RTP_PRIO_MAX)
return (EINVAL);
p->p_rtprio = rtp;
return (0);
default:
return (EINVAL);
}
1995-05-30 08:16:23 +00:00
default:
return (EINVAL);
}
}
1994-05-24 10:09:53 +00:00
#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
#ifndef _SYS_SYSPROTO_H_
struct osetrlimit_args {
1994-05-24 10:09:53 +00:00
u_int which;
struct orlimit *rlp;
1994-05-24 10:09:53 +00:00
};
#endif
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
osetrlimit(p, uap)
1994-05-24 10:09:53 +00:00
struct proc *p;
register struct osetrlimit_args *uap;
1994-05-24 10:09:53 +00:00
{
struct orlimit olim;
struct rlimit lim;
int error;
if ((error =
copyin((caddr_t)uap->rlp, (caddr_t)&olim, sizeof(struct orlimit))))
1994-05-24 10:09:53 +00:00
return (error);
lim.rlim_cur = olim.rlim_cur;
lim.rlim_max = olim.rlim_max;
return (dosetrlimit(p, uap->which, &lim));
}
#ifndef _SYS_SYSPROTO_H_
struct ogetrlimit_args {
1994-05-24 10:09:53 +00:00
u_int which;
struct orlimit *rlp;
};
#endif
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
ogetrlimit(p, uap)
1994-05-24 10:09:53 +00:00
struct proc *p;
register struct ogetrlimit_args *uap;
1994-05-24 10:09:53 +00:00
{
struct orlimit olim;
if (uap->which >= RLIM_NLIMITS)
return (EINVAL);
olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur;
if (olim.rlim_cur == -1)
olim.rlim_cur = 0x7fffffff;
olim.rlim_max = p->p_rlimit[uap->which].rlim_max;
if (olim.rlim_max == -1)
olim.rlim_max = 0x7fffffff;
return (copyout((caddr_t)&olim, (caddr_t)uap->rlp, sizeof(olim)));
}
#endif /* COMPAT_43 || COMPAT_SUNOS */
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct __setrlimit_args {
u_int which;
struct rlimit *rlp;
1994-05-24 10:09:53 +00:00
};
#endif
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
setrlimit(p, uap)
1994-05-24 10:09:53 +00:00
struct proc *p;
register struct __setrlimit_args *uap;
{
struct rlimit alim;
int error;
if ((error =
copyin((caddr_t)uap->rlp, (caddr_t)&alim, sizeof (struct rlimit))))
1994-05-24 10:09:53 +00:00
return (error);
return (dosetrlimit(p, uap->which, &alim));
}
int
1994-05-24 10:09:53 +00:00
dosetrlimit(p, which, limp)
struct proc *p;
u_int which;
struct rlimit *limp;
{
register struct rlimit *alimp;
int error;
if (which >= RLIM_NLIMITS)
return (EINVAL);
alimp = &p->p_rlimit[which];
/*
* Preserve historical bugs by treating negative limits as unsigned.
*/
if (limp->rlim_cur < 0)
limp->rlim_cur = RLIM_INFINITY;
if (limp->rlim_max < 0)
limp->rlim_max = RLIM_INFINITY;
1995-05-30 08:16:23 +00:00
if (limp->rlim_cur > alimp->rlim_max ||
1994-05-24 10:09:53 +00:00
limp->rlim_max > alimp->rlim_max)
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
if ((error = suser_xxx(0, p, PRISON_ROOT)))
1994-05-24 10:09:53 +00:00
return (error);
if (limp->rlim_cur > limp->rlim_max)
limp->rlim_cur = limp->rlim_max;
if (p->p_limit->p_refcnt > 1 &&
(p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
p->p_limit->p_refcnt--;
p->p_limit = limcopy(p->p_limit);
alimp = &p->p_rlimit[which];
}
switch (which) {
case RLIMIT_CPU:
if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000)
p->p_limit->p_cpulimit = RLIM_INFINITY;
else
p->p_limit->p_cpulimit =
(rlim_t)1000000 * limp->rlim_cur;
break;
1994-05-24 10:09:53 +00:00
case RLIMIT_DATA:
if (limp->rlim_cur > MAXDSIZ)
limp->rlim_cur = MAXDSIZ;
if (limp->rlim_max > MAXDSIZ)
limp->rlim_max = MAXDSIZ;
1994-05-24 10:09:53 +00:00
break;
case RLIMIT_STACK:
if (limp->rlim_cur > MAXSSIZ)
limp->rlim_cur = MAXSSIZ;
if (limp->rlim_max > MAXSSIZ)
limp->rlim_max = MAXSSIZ;
1994-05-24 10:09:53 +00:00
/*
* Stack is allocated to the max at exec time with only
* "rlim_cur" bytes accessible. If stack limit is going
* up make more accessible, if going down make inaccessible.
*/
if (limp->rlim_cur != alimp->rlim_cur) {
vm_offset_t addr;
vm_size_t size;
vm_prot_t prot;
if (limp->rlim_cur > alimp->rlim_cur) {
1994-05-24 10:09:53 +00:00
prot = VM_PROT_ALL;
size = limp->rlim_cur - alimp->rlim_cur;
addr = USRSTACK - limp->rlim_cur;
} else {
prot = VM_PROT_NONE;
size = alimp->rlim_cur - limp->rlim_cur;
addr = USRSTACK - alimp->rlim_cur;
}
addr = trunc_page(addr);
size = round_page(size);
(void) vm_map_protect(&p->p_vmspace->vm_map,
addr, addr+size, prot, FALSE);
}
break;
case RLIMIT_NOFILE:
if (limp->rlim_cur > maxfilesperproc)
limp->rlim_cur = maxfilesperproc;
if (limp->rlim_max > maxfilesperproc)
limp->rlim_max = maxfilesperproc;
1994-05-24 10:09:53 +00:00
break;
case RLIMIT_NPROC:
if (limp->rlim_cur > maxprocperuid)
limp->rlim_cur = maxprocperuid;
if (limp->rlim_max > maxprocperuid)
limp->rlim_max = maxprocperuid;
1994-05-24 10:09:53 +00:00
break;
}
*alimp = *limp;
return (0);
}
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct __getrlimit_args {
u_int which;
struct rlimit *rlp;
};
#endif
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
getrlimit(p, uap)
1994-05-24 10:09:53 +00:00
struct proc *p;
register struct __getrlimit_args *uap;
{
if (uap->which >= RLIM_NLIMITS)
return (EINVAL);
return (copyout((caddr_t)&p->p_rlimit[uap->which], (caddr_t)uap->rlp,
sizeof (struct rlimit)));
}
/*
* Transform the running time and tick information in proc p into user,
* system, and interrupt time usage.
*/
void
1994-05-24 10:09:53 +00:00
calcru(p, up, sp, ip)
struct proc *p;
struct timeval *up;
struct timeval *sp;
struct timeval *ip;
1994-05-24 10:09:53 +00:00
{
/* {user, system, interrupt, total} {ticks, usec}; previous tu: */
u_int64_t ut, uu, st, su, it, iu, tt, tu, ptu;
int s;
1994-05-24 10:09:53 +00:00
struct timeval tv;
/* XXX: why spl-protect ? worst case is an off-by-one report */
1994-05-24 10:09:53 +00:00
s = splstatclock();
ut = p->p_uticks;
st = p->p_sticks;
1994-05-24 10:09:53 +00:00
it = p->p_iticks;
splx(s);
tt = ut + st + it;
if (tt == 0) {
st = 1;
tt = 1;
1994-05-24 10:09:53 +00:00
}
tu = p->p_runtime;
#ifdef SMP
if (p->p_oncpu != 0xff) {
#else
if (p == curproc) {
#endif
1994-05-24 10:09:53 +00:00
/*
* Adjust for the current time slice. This is actually fairly
* important since the error here is on the order of a time
* quantum, which is much greater than the sampling error.
*/
microuptime(&tv);
tu += (tv.tv_usec - switchtime.tv_usec) +
(tv.tv_sec - switchtime.tv_sec) * (int64_t)1000000;
1994-05-24 10:09:53 +00:00
}
ptu = p->p_stats->p_uu + p->p_stats->p_su + p->p_stats->p_iu;
if (tu < ptu || (int64_t)tu < 0) {
/* XXX no %qd in kernel. Truncate. */
printf("calcru: negative time of %ld usec for pid %d (%s)\n",
(long)tu, p->p_pid, p->p_comm);
tu = ptu;
}
/* Subdivide tu. */
uu = (tu * ut) / tt;
su = (tu * st) / tt;
iu = tu - uu - su;
/* Enforce monotonicity. */
if (uu < p->p_stats->p_uu || su < p->p_stats->p_su ||
iu < p->p_stats->p_iu) {
if (uu < p->p_stats->p_uu)
uu = p->p_stats->p_uu;
else if (uu + p->p_stats->p_su + p->p_stats->p_iu > tu)
uu = tu - p->p_stats->p_su - p->p_stats->p_iu;
if (st == 0)
su = p->p_stats->p_su;
else {
su = ((tu - uu) * st) / (st + it);
if (su < p->p_stats->p_su)
su = p->p_stats->p_su;
else if (uu + su + p->p_stats->p_iu > tu)
su = tu - uu - p->p_stats->p_iu;
}
KASSERT(uu + su + p->p_stats->p_iu <= tu,
("calcru: monotonisation botch 1"));
iu = tu - uu - su;
KASSERT(iu >= p->p_stats->p_iu,
("calcru: monotonisation botch 2"));
}
p->p_stats->p_uu = uu;
p->p_stats->p_su = su;
p->p_stats->p_iu = iu;
up->tv_sec = uu / 1000000;
up->tv_usec = uu % 1000000;
sp->tv_sec = su / 1000000;
sp->tv_usec = su % 1000000;
1994-05-24 10:09:53 +00:00
if (ip != NULL) {
ip->tv_sec = iu / 1000000;
ip->tv_usec = iu % 1000000;
1994-05-24 10:09:53 +00:00
}
}
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct getrusage_args {
int who;
struct rusage *rusage;
};
#endif
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
getrusage(p, uap)
1994-05-24 10:09:53 +00:00
register struct proc *p;
register struct getrusage_args *uap;
{
register struct rusage *rup;
switch (uap->who) {
case RUSAGE_SELF:
rup = &p->p_stats->p_ru;
calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
break;
case RUSAGE_CHILDREN:
rup = &p->p_stats->p_cru;
break;
default:
return (EINVAL);
}
return (copyout((caddr_t)rup, (caddr_t)uap->rusage,
sizeof (struct rusage)));
}
void
1994-05-24 10:09:53 +00:00
ruadd(ru, ru2)
register struct rusage *ru, *ru2;
{
register long *ip, *ip2;
register int i;
timevaladd(&ru->ru_utime, &ru2->ru_utime);
timevaladd(&ru->ru_stime, &ru2->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++;
}
/*
* Make a copy of the plimit structure.
* We share these structures copy-on-write after fork,
* and copy when a limit is changed.
*/
struct plimit *
limcopy(lim)
struct plimit *lim;
{
register struct plimit *copy;
MALLOC(copy, struct plimit *, sizeof(struct plimit),
M_SUBPROC, M_WAITOK);
bcopy(lim->pl_rlimit, copy->pl_rlimit, sizeof(struct plimit));
1994-05-24 10:09:53 +00:00
copy->p_lflags = 0;
copy->p_refcnt = 1;
return (copy);
}