NetBSD/sys/kern/kern_time.c

802 lines
20 KiB
C

/* $NetBSD: kern_time.c,v 1.61 2002/01/31 00:13:08 simonb Exp $ */
/*-
* Copyright (c) 2000 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Christopher G. Demetriou.
*
* 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 NetBSD
* Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation 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 NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
*/
/*
* Copyright (c) 1982, 1986, 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* 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_time.c 8.4 (Berkeley) 5/26/95
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: kern_time.c,v 1.61 2002/01/31 00:13:08 simonb Exp $");
#include "fs_nfs.h"
#include "opt_nfs.h"
#include "opt_nfsserver.h"
#include <sys/param.h>
#include <sys/resourcevar.h>
#include <sys/kernel.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/vnode.h>
#include <sys/signalvar.h>
#include <sys/syslog.h>
#include <sys/mount.h>
#include <sys/syscallargs.h>
#include <uvm/uvm_extern.h>
#if defined(NFS) || defined(NFSSERVER)
#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
#include <nfs/nfs_var.h>
#endif
#include <machine/cpu.h>
/*
* Time of day and interval timer support.
*
* These routines provide the kernel entry points to get and set
* the time-of-day and per-process interval timers. Subroutines
* here provide support for adding and subtracting timeval structures
* and decrementing interval timers, optionally reloading the interval
* timers when they expire.
*/
/* This function is used by clock_settime and settimeofday */
int
settime(tv)
struct timeval *tv;
{
struct timeval delta;
struct cpu_info *ci;
int s;
/* WHAT DO WE DO ABOUT PENDING REAL-TIME TIMEOUTS??? */
s = splclock();
timersub(tv, &time, &delta);
if ((delta.tv_sec < 0 || delta.tv_usec < 0) && securelevel > 1) {
splx(s);
return (EPERM);
}
#ifdef notyet
if ((delta.tv_sec < 86400) && securelevel > 0) {
splx(s);
return (EPERM);
}
#endif
time = *tv;
(void) spllowersoftclock();
timeradd(&boottime, &delta, &boottime);
/*
* XXXSMP
* This is wrong. We should traverse a list of all
* CPUs and add the delta to the runtime of those
* CPUs which have a process on them.
*/
ci = curcpu();
timeradd(&ci->ci_schedstate.spc_runtime, &delta,
&ci->ci_schedstate.spc_runtime);
# if (defined(NFS) && !defined (NFS_V2_ONLY)) || defined(NFSSERVER)
nqnfs_lease_updatetime(delta.tv_sec);
# endif
splx(s);
resettodr();
return (0);
}
/* ARGSUSED */
int
sys_clock_gettime(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_clock_gettime_args /* {
syscallarg(clockid_t) clock_id;
syscallarg(struct timespec *) tp;
} */ *uap = v;
clockid_t clock_id;
struct timeval atv;
struct timespec ats;
int s;
clock_id = SCARG(uap, clock_id);
switch (clock_id) {
case CLOCK_REALTIME:
microtime(&atv);
TIMEVAL_TO_TIMESPEC(&atv,&ats);
break;
case CLOCK_MONOTONIC:
/* XXX "hz" granularity */
s = splclock();
atv = mono_time;
splx(s);
TIMEVAL_TO_TIMESPEC(&atv,&ats);
break;
default:
return (EINVAL);
}
return copyout(&ats, SCARG(uap, tp), sizeof(ats));
}
/* ARGSUSED */
int
sys_clock_settime(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_clock_settime_args /* {
syscallarg(clockid_t) clock_id;
syscallarg(const struct timespec *) tp;
} */ *uap = v;
int error;
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
return (error);
return (clock_settime1(SCARG(uap, clock_id), SCARG(uap, tp)));
}
int
clock_settime1(clock_id, tp)
clockid_t clock_id;
const struct timespec *tp;
{
struct timespec ats;
struct timeval atv;
int error;
if ((error = copyin(tp, &ats, sizeof(ats))) != 0)
return (error);
switch (clock_id) {
case CLOCK_REALTIME:
TIMESPEC_TO_TIMEVAL(&atv, &ats);
if ((error = settime(&atv)) != 0)
return (error);
break;
case CLOCK_MONOTONIC:
return (EINVAL); /* read-only clock */
default:
return (EINVAL);
}
return 0;
}
int
sys_clock_getres(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_clock_getres_args /* {
syscallarg(clockid_t) clock_id;
syscallarg(struct timespec *) tp;
} */ *uap = v;
clockid_t clock_id;
struct timespec ts;
int error = 0;
clock_id = SCARG(uap, clock_id);
switch (clock_id) {
case CLOCK_REALTIME:
case CLOCK_MONOTONIC:
ts.tv_sec = 0;
ts.tv_nsec = 1000000000 / hz;
break;
default:
return (EINVAL);
}
if (SCARG(uap, tp))
error = copyout(&ts, SCARG(uap, tp), sizeof(ts));
return error;
}
/* ARGSUSED */
int
sys_nanosleep(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
static int nanowait;
struct sys_nanosleep_args/* {
syscallarg(struct timespec *) rqtp;
syscallarg(struct timespec *) rmtp;
} */ *uap = v;
struct timespec rqt;
struct timespec rmt;
struct timeval atv, utv;
int error, s, timo;
error = copyin((caddr_t)SCARG(uap, rqtp), (caddr_t)&rqt,
sizeof(struct timespec));
if (error)
return (error);
TIMESPEC_TO_TIMEVAL(&atv,&rqt)
if (itimerfix(&atv) || atv.tv_sec > 1000000000)
return (EINVAL);
s = splclock();
timeradd(&atv,&time,&atv);
timo = hzto(&atv);
/*
* Avoid inadvertantly sleeping forever
*/
if (timo == 0)
timo = 1;
splx(s);
error = tsleep(&nanowait, PWAIT | PCATCH, "nanosleep", timo);
if (error == ERESTART)
error = EINTR;
if (error == EWOULDBLOCK)
error = 0;
if (SCARG(uap, rmtp)) {
int error;
s = splclock();
utv = time;
splx(s);
timersub(&atv, &utv, &utv);
if (utv.tv_sec < 0)
timerclear(&utv);
TIMEVAL_TO_TIMESPEC(&utv,&rmt);
error = copyout((caddr_t)&rmt, (caddr_t)SCARG(uap,rmtp),
sizeof(rmt));
if (error)
return (error);
}
return error;
}
/* ARGSUSED */
int
sys_gettimeofday(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_gettimeofday_args /* {
syscallarg(struct timeval *) tp;
syscallarg(struct timezone *) tzp;
} */ *uap = v;
struct timeval atv;
int error = 0;
struct timezone tzfake;
if (SCARG(uap, tp)) {
microtime(&atv);
error = copyout(&atv, SCARG(uap, tp), sizeof(atv));
if (error)
return (error);
}
if (SCARG(uap, tzp)) {
/*
* NetBSD has no kernel notion of time zone, so we just
* fake up a timezone struct and return it if demanded.
*/
tzfake.tz_minuteswest = 0;
tzfake.tz_dsttime = 0;
error = copyout(&tzfake, SCARG(uap, tzp), sizeof(tzfake));
}
return (error);
}
/* ARGSUSED */
int
sys_settimeofday(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_settimeofday_args /* {
syscallarg(const struct timeval *) tv;
syscallarg(const struct timezone *) tzp;
} */ *uap = v;
int error;
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
return (error);
return settimeofday1(SCARG(uap, tv), SCARG(uap, tzp), p);
}
int
settimeofday1(utv, utzp, p)
const struct timeval *utv;
const struct timezone *utzp;
struct proc *p;
{
struct timeval atv;
struct timezone atz;
struct timeval *tv = NULL;
struct timezone *tzp = NULL;
int error;
/* Verify all parameters before changing time. */
if (utv) {
if ((error = copyin(utv, &atv, sizeof(atv))) != 0)
return (error);
tv = &atv;
}
/* XXX since we don't use tz, probably no point in doing copyin. */
if (utzp) {
if ((error = copyin(utzp, &atz, sizeof(atz))) != 0)
return (error);
tzp = &atz;
}
if (tv)
if ((error = settime(tv)) != 0)
return (error);
/*
* NetBSD has no kernel notion of time zone, and only an
* obsolete program would try to set it, so we log a warning.
*/
if (tzp)
log(LOG_WARNING, "pid %d attempted to set the "
"(obsolete) kernel time zone\n", p->p_pid);
return (0);
}
int tickdelta; /* current clock skew, us. per tick */
long timedelta; /* unapplied time correction, us. */
long bigadj = 1000000; /* use 10x skew above bigadj us. */
/* ARGSUSED */
int
sys_adjtime(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_adjtime_args /* {
syscallarg(const struct timeval *) delta;
syscallarg(struct timeval *) olddelta;
} */ *uap = v;
int error;
if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
return (error);
return adjtime1(SCARG(uap, delta), SCARG(uap, olddelta), p);
}
int
adjtime1(delta, olddelta, p)
const struct timeval *delta;
struct timeval *olddelta;
struct proc *p;
{
struct timeval atv;
struct timeval *oatv = NULL;
long ndelta, ntickdelta, odelta;
int error;
int s;
error = copyin(delta, &atv, sizeof(struct timeval));
if (error)
return (error);
if (olddelta != NULL) {
if (uvm_useracc((caddr_t)olddelta,
sizeof(struct timeval), B_WRITE) == FALSE)
return (EFAULT);
oatv = olddelta;
}
/*
* Compute the total correction and the rate at which to apply it.
* Round the adjustment down to a whole multiple of the per-tick
* delta, so that after some number of incremental changes in
* hardclock(), tickdelta will become zero, lest the correction
* overshoot and start taking us away from the desired final time.
*/
ndelta = atv.tv_sec * 1000000 + atv.tv_usec;
if (ndelta > bigadj || ndelta < -bigadj)
ntickdelta = 10 * tickadj;
else
ntickdelta = tickadj;
if (ndelta % ntickdelta)
ndelta = ndelta / ntickdelta * ntickdelta;
/*
* To make hardclock()'s job easier, make the per-tick delta negative
* if we want time to run slower; then hardclock can simply compute
* tick + tickdelta, and subtract tickdelta from timedelta.
*/
if (ndelta < 0)
ntickdelta = -ntickdelta;
s = splclock();
odelta = timedelta;
timedelta = ndelta;
tickdelta = ntickdelta;
splx(s);
if (olddelta) {
atv.tv_sec = odelta / 1000000;
atv.tv_usec = odelta % 1000000;
(void) copyout(&atv, olddelta, sizeof(struct timeval));
}
return (0);
}
/*
* Get value of an interval timer. The process virtual and
* profiling virtual time timers are kept in the p_stats area, since
* they can be swapped out. These are kept internally in the
* way they are specified externally: in time until they expire.
*
* The real time interval timer is kept in the process table slot
* for the process, and its value (it_value) is kept as an
* absolute time rather than as a delta, so that it is easy to keep
* periodic real-time signals from drifting.
*
* Virtual time timers are processed in the hardclock() routine of
* kern_clock.c. The real time timer is processed by a timeout
* routine, called from the softclock() routine. Since a callout
* may be delayed in real time due to interrupt processing in the system,
* it is possible for the real time timeout routine (realitexpire, given below),
* to be delayed in real time past when it is supposed to occur. It
* does not suffice, therefore, to reload the real timer .it_value from the
* real time timers .it_interval. Rather, we compute the next time in
* absolute time the timer should go off.
*/
/* ARGSUSED */
int
sys_getitimer(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_getitimer_args /* {
syscallarg(int) which;
syscallarg(struct itimerval *) itv;
} */ *uap = v;
int which = SCARG(uap, which);
struct itimerval aitv;
int s;
if ((u_int)which > ITIMER_PROF)
return (EINVAL);
s = splclock();
if (which == ITIMER_REAL) {
/*
* Convert from absolute to relative time in .it_value
* part of real time timer. If time for real time timer
* has passed return 0, else return difference between
* current time and time for the timer to go off.
*/
aitv = p->p_realtimer;
if (timerisset(&aitv.it_value)) {
if (timercmp(&aitv.it_value, &time, <))
timerclear(&aitv.it_value);
else
timersub(&aitv.it_value, &time, &aitv.it_value);
}
} else
aitv = p->p_stats->p_timer[which];
splx(s);
return (copyout(&aitv, SCARG(uap, itv), sizeof(struct itimerval)));
}
/* ARGSUSED */
int
sys_setitimer(p, v, retval)
struct proc *p;
void *v;
register_t *retval;
{
struct sys_setitimer_args /* {
syscallarg(int) which;
syscallarg(const struct itimerval *) itv;
syscallarg(struct itimerval *) oitv;
} */ *uap = v;
int which = SCARG(uap, which);
struct sys_getitimer_args getargs;
struct itimerval aitv;
const struct itimerval *itvp;
int s, error;
if ((u_int)which > ITIMER_PROF)
return (EINVAL);
itvp = SCARG(uap, itv);
if (itvp &&
(error = copyin(itvp, &aitv, sizeof(struct itimerval)) != 0))
return (error);
if (SCARG(uap, oitv) != NULL) {
SCARG(&getargs, which) = which;
SCARG(&getargs, itv) = SCARG(uap, oitv);
if ((error = sys_getitimer(p, &getargs, retval)) != 0)
return (error);
}
if (itvp == 0)
return (0);
if (itimerfix(&aitv.it_value) || itimerfix(&aitv.it_interval))
return (EINVAL);
s = splclock();
if (which == ITIMER_REAL) {
callout_stop(&p->p_realit_ch);
if (timerisset(&aitv.it_value)) {
/*
* Don't need to check hzto() return value, here.
* callout_reset() does it for us.
*/
timeradd(&aitv.it_value, &time, &aitv.it_value);
callout_reset(&p->p_realit_ch, hzto(&aitv.it_value),
realitexpire, p);
}
p->p_realtimer = aitv;
} else
p->p_stats->p_timer[which] = aitv;
splx(s);
return (0);
}
/*
* Real interval timer expired:
* send process whose timer expired an alarm signal.
* If time is not set up to reload, then just return.
* Else compute next time timer should go off which is > current time.
* This is where delay in processing this timeout causes multiple
* SIGALRM calls to be compressed into one.
*/
void
realitexpire(arg)
void *arg;
{
struct proc *p;
int s;
p = (struct proc *)arg;
psignal(p, SIGALRM);
if (!timerisset(&p->p_realtimer.it_interval)) {
timerclear(&p->p_realtimer.it_value);
return;
}
for (;;) {
s = splclock();
timeradd(&p->p_realtimer.it_value,
&p->p_realtimer.it_interval, &p->p_realtimer.it_value);
if (timercmp(&p->p_realtimer.it_value, &time, >)) {
/*
* Don't need to check hzto() return value, here.
* callout_reset() does it for us.
*/
callout_reset(&p->p_realit_ch,
hzto(&p->p_realtimer.it_value), realitexpire, p);
splx(s);
return;
}
splx(s);
}
}
/*
* Check that a proposed value to load into the .it_value or
* .it_interval part of an interval timer is acceptable, and
* fix it to have at least minimal value (i.e. if it is less
* than the resolution of the clock, round it up.)
*/
int
itimerfix(tv)
struct timeval *tv;
{
if (tv->tv_sec < 0 || tv->tv_usec < 0 || tv->tv_usec >= 1000000)
return (EINVAL);
if (tv->tv_sec == 0 && tv->tv_usec != 0 && tv->tv_usec < tick)
tv->tv_usec = tick;
return (0);
}
/*
* Decrement an interval timer by a specified number
* of microseconds, which must be less than a second,
* i.e. < 1000000. If the timer expires, then reload
* it. In this case, carry over (usec - old value) to
* reduce the value reloaded into the timer so that
* the timer does not drift. This routine assumes
* that it is called in a context where the timers
* on which it is operating cannot change in value.
*/
int
itimerdecr(itp, usec)
struct itimerval *itp;
int usec;
{
if (itp->it_value.tv_usec < usec) {
if (itp->it_value.tv_sec == 0) {
/* expired, and already in next interval */
usec -= itp->it_value.tv_usec;
goto expire;
}
itp->it_value.tv_usec += 1000000;
itp->it_value.tv_sec--;
}
itp->it_value.tv_usec -= usec;
usec = 0;
if (timerisset(&itp->it_value))
return (1);
/* expired, exactly at end of interval */
expire:
if (timerisset(&itp->it_interval)) {
itp->it_value = itp->it_interval;
itp->it_value.tv_usec -= usec;
if (itp->it_value.tv_usec < 0) {
itp->it_value.tv_usec += 1000000;
itp->it_value.tv_sec--;
}
} else
itp->it_value.tv_usec = 0; /* sec is already 0 */
return (0);
}
/*
* ratecheck(): simple time-based rate-limit checking. see ratecheck(9)
* for usage and rationale.
*/
int
ratecheck(lasttime, mininterval)
struct timeval *lasttime;
const struct timeval *mininterval;
{
struct timeval tv, delta;
int s, rv = 0;
s = splclock();
tv = mono_time;
splx(s);
timersub(&tv, lasttime, &delta);
/*
* check for 0,0 is so that the message will be seen at least once,
* even if interval is huge.
*/
if (timercmp(&delta, mininterval, >=) ||
(lasttime->tv_sec == 0 && lasttime->tv_usec == 0)) {
*lasttime = tv;
rv = 1;
}
return (rv);
}
/*
* ppsratecheck(): packets (or events) per second limitation.
*/
int
ppsratecheck(lasttime, curpps, maxpps)
struct timeval *lasttime;
int *curpps;
int maxpps; /* maximum pps allowed */
{
struct timeval tv, delta;
int s, rv;
s = splclock();
tv = mono_time;
splx(s);
timersub(&tv, lasttime, &delta);
/*
* check for 0,0 is so that the message will be seen at least once.
* if more than one second have passed since the last update of
* lasttime, reset the counter.
*
* we do increment *curpps even in *curpps < maxpps case, as some may
* try to use *curpps for stat purposes as well.
*/
if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
delta.tv_sec >= 1) {
*lasttime = tv;
*curpps = 0;
rv = 1;
} else if (maxpps < 0)
rv = 1;
else if (*curpps < maxpps)
rv = 1;
else
rv = 0;
#if 1 /*DIAGNOSTIC?*/
/* be careful about wrap-around */
if (*curpps + 1 > *curpps)
*curpps = *curpps + 1;
#else
/*
* assume that there's not too many calls to this function.
* not sure if the assumption holds, as it depends on *caller's*
* behavior, not the behavior of this function.
* IMHO it is wrong to make assumption on the caller's behavior,
* so the above #if is #if 1, not #ifdef DIAGNOSTIC.
*/
*curpps = *curpps + 1;
#endif
return (rv);
}