aboutsummaryrefslogtreecommitdiff
path: root/sys/kern
diff options
context:
space:
mode:
Diffstat (limited to 'sys/kern')
-rw-r--r--sys/kern/device_if.m10
-rw-r--r--sys/kern/init_sysent.c2
-rw-r--r--sys/kern/kern_exit.c2
-rw-r--r--sys/kern/kern_jail.c15
-rw-r--r--sys/kern/kern_prot.c56
-rw-r--r--sys/kern/kern_racct.c307
-rw-r--r--sys/kern/kern_time.c5
-rw-r--r--sys/kern/sys_timerfd.c1
-rw-r--r--sys/kern/syscalls.c2
-rw-r--r--sys/kern/syscalls.conf1
-rw-r--r--sys/kern/syscalls.master5
-rw-r--r--sys/kern/systrace_args.c8
-rw-r--r--sys/kern/vfs_default.c1
-rw-r--r--sys/kern/vfs_syscalls.c19
-rw-r--r--sys/kern/vfs_vnops.c5
15 files changed, 260 insertions, 179 deletions
diff --git a/sys/kern/device_if.m b/sys/kern/device_if.m
index c02e5a46f326..ed94b2ccbe1b 100644
--- a/sys/kern/device_if.m
+++ b/sys/kern/device_if.m
@@ -49,25 +49,25 @@ HEADER {
CODE {
static int null_shutdown(device_t dev)
{
- return 0;
+ return (0);
}
static int null_suspend(device_t dev)
{
- return 0;
+ return (0);
}
static int null_resume(device_t dev)
{
- return 0;
+ return (0);
}
static int null_quiesce(device_t dev)
{
- return 0;
+ return (0);
}
- static void * null_register(device_t dev)
+ static void *null_register(device_t dev)
{
return NULL;
}
diff --git a/sys/kern/init_sysent.c b/sys/kern/init_sysent.c
index 91792430d24c..c0a5479c9634 100644
--- a/sys/kern/init_sysent.c
+++ b/sys/kern/init_sysent.c
@@ -67,7 +67,7 @@
/* The casts are bogus but will do for now. */
struct sysent sysent[] = {
{ .sy_narg = 0, .sy_call = (sy_call_t *)nosys, .sy_auevent = AUE_NULL, .sy_flags = 0, .sy_thrcnt = SY_THR_STATIC }, /* 0 = syscall */
- { .sy_narg = AS(exit_args), .sy_call = (sy_call_t *)sys_exit, .sy_auevent = AUE_EXIT, .sy_flags = SYF_CAPENABLED, .sy_thrcnt = SY_THR_STATIC }, /* 1 = exit */
+ { .sy_narg = AS(_exit_args), .sy_call = (sy_call_t *)sys__exit, .sy_auevent = AUE_EXIT, .sy_flags = SYF_CAPENABLED, .sy_thrcnt = SY_THR_STATIC }, /* 1 = _exit */
{ .sy_narg = 0, .sy_call = (sy_call_t *)sys_fork, .sy_auevent = AUE_FORK, .sy_flags = SYF_CAPENABLED, .sy_thrcnt = SY_THR_STATIC }, /* 2 = fork */
{ .sy_narg = AS(read_args), .sy_call = (sy_call_t *)sys_read, .sy_auevent = AUE_READ, .sy_flags = SYF_CAPENABLED, .sy_thrcnt = SY_THR_STATIC }, /* 3 = read */
{ .sy_narg = AS(write_args), .sy_call = (sy_call_t *)sys_write, .sy_auevent = AUE_WRITE, .sy_flags = SYF_CAPENABLED, .sy_thrcnt = SY_THR_STATIC }, /* 4 = write */
diff --git a/sys/kern/kern_exit.c b/sys/kern/kern_exit.c
index 54e3044ab093..a32b5a1b3354 100644
--- a/sys/kern/kern_exit.c
+++ b/sys/kern/kern_exit.c
@@ -202,7 +202,7 @@ exit_onexit(struct proc *p)
* exit -- death of process.
*/
int
-sys_exit(struct thread *td, struct exit_args *uap)
+sys__exit(struct thread *td, struct _exit_args *uap)
{
exit1(td, uap->rval, 0);
diff --git a/sys/kern/kern_jail.c b/sys/kern/kern_jail.c
index 7ef1d19f0ea8..7c9a15ae18f3 100644
--- a/sys/kern/kern_jail.c
+++ b/sys/kern/kern_jail.c
@@ -115,8 +115,11 @@ struct prison prison0 = {
#else
.pr_flags = PR_HOST|_PR_IP_SADDRSEL,
#endif
- .pr_allow = PR_ALLOW_ALL_STATIC,
+ .pr_allow = PR_ALLOW_PRISON0,
};
+_Static_assert((PR_ALLOW_PRISON0 & ~PR_ALLOW_ALL_STATIC) == 0,
+ "Bits enabled in PR_ALLOW_PRISON0 that are not statically reserved");
+
MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
struct bool_flags {
@@ -232,6 +235,9 @@ static struct bool_flags pr_flag_allow[NBBY * NBPW] = {
{"allow.adjtime", "allow.noadjtime", PR_ALLOW_ADJTIME},
{"allow.settime", "allow.nosettime", PR_ALLOW_SETTIME},
{"allow.routing", "allow.norouting", PR_ALLOW_ROUTING},
+ {"allow.unprivileged_parent_tampering",
+ "allow.nounprivileged_parent_tampering",
+ PR_ALLOW_UNPRIV_PARENT_TAMPER},
};
static unsigned pr_allow_all = PR_ALLOW_ALL_STATIC;
const size_t pr_flag_allow_size = sizeof(pr_flag_allow);
@@ -4006,6 +4012,7 @@ prison_priv_check(struct ucred *cred, int priv)
case PRIV_DEBUG_DIFFCRED:
case PRIV_DEBUG_SUGID:
case PRIV_DEBUG_UNPRIV:
+ case PRIV_DEBUG_DIFFJAIL:
/*
* Allow jail to set various resource limits and login
@@ -4043,8 +4050,10 @@ prison_priv_check(struct ucred *cred, int priv)
*/
case PRIV_SCHED_DIFFCRED:
case PRIV_SCHED_CPUSET:
+ case PRIV_SCHED_DIFFJAIL:
case PRIV_SIGNAL_DIFFCRED:
case PRIV_SIGNAL_SUGID:
+ case PRIV_SIGNAL_DIFFJAIL:
/*
* Allow jailed processes to write to sysctls marked as jail
@@ -4688,6 +4697,10 @@ SYSCTL_JAIL_PARAM(_allow, read_msgbuf, CTLTYPE_INT | CTLFLAG_RW,
"B", "Jail may read the kernel message buffer");
SYSCTL_JAIL_PARAM(_allow, unprivileged_proc_debug, CTLTYPE_INT | CTLFLAG_RW,
"B", "Unprivileged processes may use process debugging facilities");
+SYSCTL_JAIL_PARAM(_allow, unprivileged_parent_tampering,
+ CTLTYPE_INT | CTLFLAG_RW, "B",
+ "Unprivileged parent jail processes may tamper with same-uid processes"
+ " (signal/debug/cpuset)");
SYSCTL_JAIL_PARAM(_allow, suser, CTLTYPE_INT | CTLFLAG_RW,
"B", "Processes in jail with uid 0 have privilege");
#ifdef VIMAGE
diff --git a/sys/kern/kern_prot.c b/sys/kern/kern_prot.c
index bbb622547598..2cd5b7069023 100644
--- a/sys/kern/kern_prot.c
+++ b/sys/kern/kern_prot.c
@@ -1914,6 +1914,38 @@ cr_canseejailproc(struct ucred *u1, struct ucred *u2)
}
/*
+ * Determine if u1 can tamper with the subject specified by u2, if they are in
+ * different jails and 'unprivileged_parent_tampering' jail policy allows it.
+ *
+ * May be called if u1 and u2 are in the same jail, but it is expected that the
+ * caller has already done a prison_check() prior to calling it.
+ *
+ * Returns: 0 for permitted, EPERM otherwise
+ */
+static int
+cr_can_tamper_with_subjail(struct ucred *u1, struct ucred *u2, int priv)
+{
+
+ MPASS(prison_check(u1, u2) == 0);
+ if (u1->cr_prison == u2->cr_prison)
+ return (0);
+
+ if (priv_check_cred(u1, priv) == 0)
+ return (0);
+
+ /*
+ * Jails do not maintain a distinct UID space, so process visibility is
+ * all that would control an unprivileged process' ability to tamper
+ * with a process in a subjail by default if we did not have the
+ * allow.unprivileged_parent_tampering knob to restrict it by default.
+ */
+ if (prison_allow(u2, PR_ALLOW_UNPRIV_PARENT_TAMPER))
+ return (0);
+
+ return (EPERM);
+}
+
+/*
* Helper for cr_cansee*() functions to abide by system-wide security.bsd.see_*
* policies. Determines if u1 "can see" u2 according to these policies.
* Returns: 0 for permitted, ESRCH otherwise
@@ -2062,6 +2094,19 @@ cr_cansignal(struct ucred *cred, struct proc *proc, int signum)
return (error);
}
+ /*
+ * At this point, the target may be in a different jail than the
+ * subject -- the subject must be in a parent jail to the target,
+ * whether it is prison0 or a subordinate of prison0 that has
+ * children. Additional privileges are required to allow this, as
+ * whether the creds are truly equivalent or not must be determined on
+ * a case-by-case basis.
+ */
+ error = cr_can_tamper_with_subjail(cred, proc->p_ucred,
+ PRIV_SIGNAL_DIFFJAIL);
+ if (error)
+ return (error);
+
return (0);
}
@@ -2138,6 +2183,12 @@ p_cansched(struct thread *td, struct proc *p)
if (error)
return (error);
}
+
+ error = cr_can_tamper_with_subjail(td->td_ucred, p->p_ucred,
+ PRIV_SCHED_DIFFJAIL);
+ if (error)
+ return (error);
+
return (0);
}
@@ -2258,6 +2309,11 @@ p_candebug(struct thread *td, struct proc *p)
return (error);
}
+ error = cr_can_tamper_with_subjail(td->td_ucred, p->p_ucred,
+ PRIV_DEBUG_DIFFJAIL);
+ if (error)
+ return (error);
+
/* Can't trace init when securelevel > 0. */
if (p == initproc) {
error = securelevel_gt(td->td_ucred, 0);
diff --git a/sys/kern/kern_racct.c b/sys/kern/kern_racct.c
index 7ee3b9e2048a..7351e9cb6313 100644
--- a/sys/kern/kern_racct.c
+++ b/sys/kern/kern_racct.c
@@ -96,6 +96,13 @@ static void racct_sub_cred_locked(struct ucred *cred, int resource,
uint64_t amount);
static void racct_add_cred_locked(struct ucred *cred, int resource,
uint64_t amount);
+static int racct_set_locked(struct proc *p, int resource, uint64_t amount,
+ int force);
+static void racct_updatepcpu_locked(struct proc *p);
+static void racct_updatepcpu_racct_locked(struct racct *racct);
+static void racct_updatepcpu_containers(void);
+static void racct_settime_locked(struct proc *p, bool exit);
+static void racct_zeropcpu_locked(struct proc *p);
SDT_PROVIDER_DEFINE(racct);
SDT_PROBE_DEFINE3(racct, , rusage, add,
@@ -308,68 +315,6 @@ fixpt_t ccpu_exp[] = {
#define CCPU_EXP_MAX 110
-/*
- * This function is analogical to the getpcpu() function in the ps(1) command.
- * They should both calculate in the same way so that the racct %cpu
- * calculations are consistent with the values shown by the ps(1) tool.
- * The calculations are more complex in the 4BSD scheduler because of the value
- * of the ccpu variable. In ULE it is defined to be zero which saves us some
- * work.
- */
-static uint64_t
-racct_getpcpu(struct proc *p, u_int pcpu)
-{
- u_int swtime;
-#ifdef SCHED_4BSD
- fixpt_t pctcpu, pctcpu_next;
-#endif
- fixpt_t p_pctcpu;
- struct thread *td;
-
- ASSERT_RACCT_ENABLED();
- KASSERT((p->p_flag & P_IDLEPROC) == 0,
- ("racct_getpcpu: idle process %p", p));
-
- swtime = (ticks - p->p_swtick) / hz;
-
- /*
- * For short-lived processes, the sched_pctcpu() returns small
- * values even for cpu intensive processes. Therefore we use
- * our own estimate in this case.
- */
- if (swtime < RACCT_PCPU_SECS)
- return (pcpu);
-
- p_pctcpu = 0;
- FOREACH_THREAD_IN_PROC(p, td) {
- thread_lock(td);
-#ifdef SCHED_4BSD
- pctcpu = sched_pctcpu(td);
- /* Count also the yet unfinished second. */
- pctcpu_next = (pctcpu * ccpu_exp[1]) >> FSHIFT;
- pctcpu_next += sched_pctcpu_delta(td);
- p_pctcpu += max(pctcpu, pctcpu_next);
-#else
- /*
- * In ULE the %cpu statistics are updated on every
- * sched_pctcpu() call. So special calculations to
- * account for the latest (unfinished) second are
- * not needed.
- */
- p_pctcpu += sched_pctcpu(td);
-#endif
- thread_unlock(td);
- }
-
-#ifdef SCHED_4BSD
- if (swtime <= CCPU_EXP_MAX)
- return ((100 * (uint64_t)p_pctcpu * 1000000) /
- (FSCALE - ccpu_exp[swtime]));
-#endif
-
- return ((100 * (uint64_t)p_pctcpu * 1000000) / FSCALE);
-}
-
static void
racct_add_racct(struct racct *dest, const struct racct *src)
{
@@ -499,19 +444,6 @@ racct_adjust_resource(struct racct *racct, int resource,
("%s: resource %d usage < 0", __func__, resource));
racct->r_resources[resource] = 0;
}
-
- /*
- * There are some cases where the racct %cpu resource would grow
- * beyond 100% per core. For example in racct_proc_exit() we add
- * the process %cpu usage to the ucred racct containers. If too
- * many processes terminated in a short time span, the ucred %cpu
- * resource could grow too much. Also, the 4BSD scheduler sometimes
- * returns for a thread more than 100% cpu usage. So we set a sane
- * boundary here to 100% * the maximum number of CPUs.
- */
- if ((resource == RACCT_PCTCPU) &&
- (racct->r_resources[RACCT_PCTCPU] > 100 * 1000000 * (int64_t)MAXCPU))
- racct->r_resources[RACCT_PCTCPU] = 100 * 1000000 * (int64_t)MAXCPU;
}
static int
@@ -635,10 +567,44 @@ racct_add_buf(struct proc *p, const struct buf *bp, int is_write)
RACCT_UNLOCK();
}
+static void
+racct_settime_locked(struct proc *p, bool exit)
+{
+ struct thread *td;
+ struct timeval wallclock;
+ uint64_t runtime;
+
+ ASSERT_RACCT_ENABLED();
+ RACCT_LOCK_ASSERT();
+ PROC_LOCK_ASSERT(p, MA_OWNED);
+
+ if (exit) {
+ /*
+ * proc_reap() has already calculated rux
+ * and added crux to rux.
+ */
+ runtime = cputick2usec(p->p_rux.rux_runtime -
+ p->p_crux.rux_runtime);
+ } else {
+ PROC_STATLOCK(p);
+ FOREACH_THREAD_IN_PROC(p, td)
+ ruxagg(p, td);
+ PROC_STATUNLOCK(p);
+ runtime = cputick2usec(p->p_rux.rux_runtime);
+ }
+ microuptime(&wallclock);
+ timevalsub(&wallclock, &p->p_stats->p_start);
+
+ racct_set_locked(p, RACCT_CPU, runtime, 0);
+ racct_set_locked(p, RACCT_WALLCLOCK,
+ (uint64_t)wallclock.tv_sec * 1000000 +
+ wallclock.tv_usec, 0);
+}
+
static int
racct_set_locked(struct proc *p, int resource, uint64_t amount, int force)
{
- int64_t old_amount, decayed_amount, diff_proc, diff_cred;
+ int64_t old_amount, diff_proc, diff_cred;
#ifdef RCTL
int error;
#endif
@@ -655,17 +621,7 @@ racct_set_locked(struct proc *p, int resource, uint64_t amount, int force)
* The diffs may be negative.
*/
diff_proc = amount - old_amount;
- if (resource == RACCT_PCTCPU) {
- /*
- * Resources in per-credential racct containers may decay.
- * If this is the case, we need to calculate the difference
- * between the new amount and the proportional value of the
- * old amount that has decayed in the ucred racct containers.
- */
- decayed_amount = old_amount * RACCT_DECAY_FACTOR / FSCALE;
- diff_cred = amount - decayed_amount;
- } else
- diff_cred = diff_proc;
+ diff_cred = diff_proc;
#ifdef notyet
KASSERT(diff_proc >= 0 || RACCT_CAN_DROP(resource),
("%s: usage of non-droppable resource %d dropping", __func__,
@@ -908,8 +864,6 @@ racct_proc_fork(struct proc *parent, struct proc *child)
goto out;
#endif
- /* Init process cpu time. */
- child->p_prev_runtime = 0;
child->p_throttled = 0;
/*
@@ -964,37 +918,16 @@ racct_proc_fork_done(struct proc *child)
void
racct_proc_exit(struct proc *p)
{
- struct timeval wallclock;
- uint64_t pct_estimate, pct, runtime;
int i;
if (!racct_enable)
return;
PROC_LOCK(p);
- /*
- * We don't need to calculate rux, proc_reap() has already done this.
- */
- runtime = cputick2usec(p->p_rux.rux_runtime);
-#ifdef notyet
- KASSERT(runtime >= p->p_prev_runtime, ("runtime < p_prev_runtime"));
-#else
- if (runtime < p->p_prev_runtime)
- runtime = p->p_prev_runtime;
-#endif
- microuptime(&wallclock);
- timevalsub(&wallclock, &p->p_stats->p_start);
- if (wallclock.tv_sec > 0 || wallclock.tv_usec > 0) {
- pct_estimate = (1000000 * runtime * 100) /
- ((uint64_t)wallclock.tv_sec * 1000000 +
- wallclock.tv_usec);
- } else
- pct_estimate = 0;
- pct = racct_getpcpu(p, pct_estimate);
-
RACCT_LOCK();
- racct_set_locked(p, RACCT_CPU, runtime, 0);
- racct_add_cred_locked(p->p_ucred, RACCT_PCTCPU, pct);
+
+ racct_settime_locked(p, true);
+ racct_zeropcpu_locked(p);
KASSERT(p->p_racct->r_resources[RACCT_RSS] == 0,
("process reaped with %ju allocated for RSS\n",
@@ -1068,6 +1001,10 @@ racct_move(struct racct *dest, struct racct *src)
RACCT_LOCK();
racct_add_racct(dest, src);
racct_sub_racct(src, src);
+ dest->r_runtime = src->r_runtime;
+ dest->r_time = src->r_time;
+ src->r_runtime = 0;
+ timevalsub(&src->r_time, &src->r_time);
RACCT_UNLOCK();
}
@@ -1170,8 +1107,6 @@ racct_proc_wakeup(struct proc *p)
static void
racct_decay_callback(struct racct *racct, void *dummy1, void *dummy2)
{
- int64_t r_old, r_new;
-
ASSERT_RACCT_ENABLED();
RACCT_LOCK_ASSERT();
@@ -1181,15 +1116,6 @@ racct_decay_callback(struct racct *racct, void *dummy1, void *dummy2)
rctl_throttle_decay(racct, RACCT_READIOPS);
rctl_throttle_decay(racct, RACCT_WRITEIOPS);
#endif
-
- r_old = racct->r_resources[RACCT_PCTCPU];
-
- /* If there is nothing to decay, just exit. */
- if (r_old <= 0)
- return;
-
- r_new = r_old * RACCT_DECAY_FACTOR / FSCALE;
- racct->r_resources[RACCT_PCTCPU] = r_new;
}
static void
@@ -1221,15 +1147,105 @@ racct_decay(void)
}
static void
+racct_updatepcpu_racct_locked(struct racct *racct)
+{
+ struct timeval diff;
+ uint64_t elapsed;
+ uint64_t runtime;
+ uint64_t newpcpu;
+ uint64_t oldpcpu;
+
+ ASSERT_RACCT_ENABLED();
+ RACCT_LOCK_ASSERT();
+
+ /* Difference between now and previously-recorded time. */
+ microuptime(&diff);
+ timevalsub(&diff, &racct->r_time);
+ elapsed = (uint64_t)diff.tv_sec * 1000000 + diff.tv_usec;
+
+ /* Difference between current and previously-recorded runtime. */
+ runtime = racct->r_resources[RACCT_CPU] - racct->r_runtime;
+
+ newpcpu = runtime * 100 * 1000000 / elapsed;
+ oldpcpu = racct->r_resources[RACCT_PCTCPU];
+ /*
+ * This calculation is equivalent to
+ * (1 - 0.3) * newpcpu + 0.3 * oldpcpu
+ * where RACCT_DECAY_FACTOR = 0.3 * FSCALE.
+ */
+ racct->r_resources[RACCT_PCTCPU] = ((FSCALE - RACCT_DECAY_FACTOR) *
+ newpcpu + RACCT_DECAY_FACTOR * oldpcpu) / FSCALE;
+ if (racct->r_resources[RACCT_PCTCPU] >
+ 100 * 1000000 * (uint64_t)mp_ncpus)
+ racct->r_resources[RACCT_PCTCPU] = 100 * 1000000 *
+ (uint64_t)mp_ncpus;
+
+ /* Record current times. */
+ racct->r_runtime = racct->r_resources[RACCT_CPU];
+ timevaladd(&racct->r_time, &diff);
+}
+
+static void
+racct_zeropcpu_locked(struct proc *p)
+{
+ ASSERT_RACCT_ENABLED();
+ PROC_LOCK_ASSERT(p, MA_OWNED);
+
+ p->p_racct->r_resources[RACCT_PCTCPU] = 0;
+}
+
+static void
+racct_updatepcpu_locked(struct proc *p)
+{
+ ASSERT_RACCT_ENABLED();
+ PROC_LOCK_ASSERT(p, MA_OWNED);
+
+ racct_updatepcpu_racct_locked(p->p_racct);
+}
+
+static void
+racct_updatepcpu_pre(void)
+{
+
+ RACCT_LOCK();
+}
+
+static void
+racct_updatepcpu_post(void)
+{
+
+ RACCT_UNLOCK();
+}
+
+static void
+racct_updatepcpu_racct_callback(struct racct *racct, void *dummy1, void *dummy2)
+{
+ racct_updatepcpu_racct_locked(racct);
+}
+
+static void
+racct_updatepcpu_containers(void)
+{
+ ASSERT_RACCT_ENABLED();
+
+ ui_racct_foreach(racct_updatepcpu_racct_callback, racct_updatepcpu_pre,
+ racct_updatepcpu_post, NULL, NULL);
+ loginclass_racct_foreach(racct_updatepcpu_racct_callback, racct_updatepcpu_pre,
+ racct_updatepcpu_post, NULL, NULL);
+ prison_racct_foreach(racct_updatepcpu_racct_callback, racct_updatepcpu_pre,
+ racct_updatepcpu_post, NULL, NULL);
+}
+
+static void
racctd(void)
{
- struct thread *td;
struct proc *p;
- struct timeval wallclock;
- uint64_t pct, pct_estimate, runtime;
+ struct proc *idle;
ASSERT_RACCT_ENABLED();
+ idle = STAILQ_FIRST(&cpuhead)->pc_idlethread->td_proc;
+
for (;;) {
racct_decay();
@@ -1237,36 +1253,16 @@ racctd(void)
FOREACH_PROC_IN_SYSTEM(p) {
PROC_LOCK(p);
+ if (p == idle) {
+ PROC_UNLOCK(p);
+ continue;
+ }
if (p->p_state != PRS_NORMAL ||
(p->p_flag & P_IDLEPROC) != 0) {
- if (p->p_state == PRS_ZOMBIE)
- racct_set(p, RACCT_PCTCPU, 0);
PROC_UNLOCK(p);
continue;
}
- microuptime(&wallclock);
- timevalsub(&wallclock, &p->p_stats->p_start);
- PROC_STATLOCK(p);
- FOREACH_THREAD_IN_PROC(p, td)
- ruxagg(p, td);
- runtime = cputick2usec(p->p_rux.rux_runtime);
- PROC_STATUNLOCK(p);
-#ifdef notyet
- KASSERT(runtime >= p->p_prev_runtime,
- ("runtime < p_prev_runtime"));
-#else
- if (runtime < p->p_prev_runtime)
- runtime = p->p_prev_runtime;
-#endif
- p->p_prev_runtime = runtime;
- if (wallclock.tv_sec > 0 || wallclock.tv_usec > 0) {
- pct_estimate = (1000000 * runtime * 100) /
- ((uint64_t)wallclock.tv_sec * 1000000 +
- wallclock.tv_usec);
- } else
- pct_estimate = 0;
- pct = racct_getpcpu(p, pct_estimate);
RACCT_LOCK();
#ifdef RCTL
rctl_throttle_decay(p->p_racct, RACCT_READBPS);
@@ -1274,11 +1270,8 @@ racctd(void)
rctl_throttle_decay(p->p_racct, RACCT_READIOPS);
rctl_throttle_decay(p->p_racct, RACCT_WRITEIOPS);
#endif
- racct_set_locked(p, RACCT_PCTCPU, pct, 1);
- racct_set_locked(p, RACCT_CPU, runtime, 0);
- racct_set_locked(p, RACCT_WALLCLOCK,
- (uint64_t)wallclock.tv_sec * 1000000 +
- wallclock.tv_usec, 0);
+ racct_settime_locked(p, false);
+ racct_updatepcpu_locked(p);
RACCT_UNLOCK();
PROC_UNLOCK(p);
}
@@ -1306,6 +1299,8 @@ racctd(void)
PROC_UNLOCK(p);
}
sx_sunlock(&allproc_lock);
+
+ racct_updatepcpu_containers();
pause("-", hz);
}
}
diff --git a/sys/kern/kern_time.c b/sys/kern/kern_time.c
index 9830e5093a3a..2a6f0989f6aa 100644
--- a/sys/kern/kern_time.c
+++ b/sys/kern/kern_time.c
@@ -571,7 +571,10 @@ kern_clock_nanosleep(struct thread *td, clockid_t clock_id, int flags,
td->td_rtcgen =
atomic_load_acq_int(&rtc_generation);
error = kern_clock_gettime(td, clock_id, &now);
- KASSERT(error == 0, ("kern_clock_gettime: %d", error));
+ if (error != 0) {
+ td->td_rtcgen = 0;
+ return (error);
+ }
timespecsub(&ts, &now, &ts);
}
if (ts.tv_sec < 0 || (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
diff --git a/sys/kern/sys_timerfd.c b/sys/kern/sys_timerfd.c
index ab7e048a2ab1..565ab3ad6ee6 100644
--- a/sys/kern/sys_timerfd.c
+++ b/sys/kern/sys_timerfd.c
@@ -206,7 +206,6 @@ retry:
mtx_unlock(&tfd->tfd_lock);
return (EAGAIN);
}
- td->td_rtcgen = atomic_load_acq_int(&rtc_generation);
error = mtx_sleep(&tfd->tfd_count, &tfd->tfd_lock,
PCATCH, "tfdrd", 0);
if (error == 0) {
diff --git a/sys/kern/syscalls.c b/sys/kern/syscalls.c
index 90a4f3a7dad8..09bf4d519927 100644
--- a/sys/kern/syscalls.c
+++ b/sys/kern/syscalls.c
@@ -6,7 +6,7 @@
const char *syscallnames[] = {
"syscall", /* 0 = syscall */
- "exit", /* 1 = exit */
+ "_exit", /* 1 = _exit */
"fork", /* 2 = fork */
"read", /* 3 = read */
"write", /* 4 = write */
diff --git a/sys/kern/syscalls.conf b/sys/kern/syscalls.conf
index a98d52659832..ae7bd1f87612 100644
--- a/sys/kern/syscalls.conf
+++ b/sys/kern/syscalls.conf
@@ -1,3 +1,4 @@
libsysmap="../../lib/libsys/syscalls.map"
libsys_h="../../lib/libsys/_libsys.h"
sysmk="../sys/syscall.mk"
+syshdr_extra="#define SYS_exit SYS__exit"
diff --git a/sys/kern/syscalls.master b/sys/kern/syscalls.master
index a8815afee866..53b5d3cbbba9 100644
--- a/sys/kern/syscalls.master
+++ b/sys/kern/syscalls.master
@@ -51,6 +51,7 @@
; SYSMUX syscall multiplexer. No prototype, argument struct, or
; handler is declared or used. Handled in MD syscall code.
; CAPENABLED syscall is allowed in capability mode
+; NORETURN the syscall does not return
;
; To support programmatic generation of both the default ABI and 32-bit compat
; (freebsd32) we impose a number of restrictions on the types of system calls.
@@ -124,8 +125,8 @@
...
);
}
-1 AUE_EXIT STD|CAPENABLED {
- void exit(
+1 AUE_EXIT STD|CAPENABLED|NORETURN {
+ void _exit(
int rval
);
}
diff --git a/sys/kern/systrace_args.c b/sys/kern/systrace_args.c
index 467caa71f20d..4dfc63924da9 100644
--- a/sys/kern/systrace_args.c
+++ b/sys/kern/systrace_args.c
@@ -17,9 +17,9 @@ systrace_args(int sysnum, void *params, uint64_t *uarg, int *n_args)
*n_args = 0;
break;
}
- /* exit */
+ /* _exit */
case 1: {
- struct exit_args *p = params;
+ struct _exit_args *p = params;
iarg[a++] = p->rval; /* int */
*n_args = 1;
break;
@@ -3513,7 +3513,7 @@ systrace_entry_setargdesc(int sysnum, int ndx, char *desc, size_t descsz)
/* syscall */
case 0:
break;
- /* exit */
+ /* _exit */
case 1:
switch (ndx) {
case 0:
@@ -9380,7 +9380,7 @@ systrace_return_setargdesc(int sysnum, int ndx, char *desc, size_t descsz)
switch (sysnum) {
/* syscall */
case 0:
- /* exit */
+ /* _exit */
case 1:
if (ndx == 0 || ndx == 1)
p = "void";
diff --git a/sys/kern/vfs_default.c b/sys/kern/vfs_default.c
index fd6202a1424c..85f67731e1cc 100644
--- a/sys/kern/vfs_default.c
+++ b/sys/kern/vfs_default.c
@@ -457,6 +457,7 @@ vop_stdpathconf(struct vop_pathconf_args *ap)
case _PC_NAMEDATTR_ENABLED:
case _PC_HAS_NAMEDATTR:
case _PC_HAS_HIDDENSYSTEM:
+ case _PC_CLONE_BLKSIZE:
*ap->a_retval = 0;
return (0);
default:
diff --git a/sys/kern/vfs_syscalls.c b/sys/kern/vfs_syscalls.c
index 9704e9c160a8..b805e147bd62 100644
--- a/sys/kern/vfs_syscalls.c
+++ b/sys/kern/vfs_syscalls.c
@@ -5050,15 +5050,16 @@ kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
size_t retlen;
void *rl_rcookie, *rl_wcookie;
off_t inoff, outoff, savinoff, savoutoff;
- bool foffsets_locked;
+ bool foffsets_locked, foffsets_set;
infp = outfp = NULL;
rl_rcookie = rl_wcookie = NULL;
foffsets_locked = false;
+ foffsets_set = false;
error = 0;
retlen = 0;
- if (flags != 0) {
+ if ((flags & ~COPY_FILE_RANGE_USERFLAGS) != 0) {
error = EINVAL;
goto out;
}
@@ -5122,6 +5123,8 @@ kern_copy_file_range(struct thread *td, int infd, off_t *inoffp, int outfd,
}
foffset_lock_pair(infp1, &inoff, outfp1, &outoff, 0);
foffsets_locked = true;
+ } else {
+ foffsets_set = true;
}
savinoff = inoff;
savoutoff = outoff;
@@ -5180,11 +5183,12 @@ out:
vn_rangelock_unlock(invp, rl_rcookie);
if (rl_wcookie != NULL)
vn_rangelock_unlock(outvp, rl_wcookie);
+ if ((foffsets_locked || foffsets_set) &&
+ (error == EINTR || error == ERESTART)) {
+ inoff = savinoff;
+ outoff = savoutoff;
+ }
if (foffsets_locked) {
- if (error == EINTR || error == ERESTART) {
- inoff = savinoff;
- outoff = savoutoff;
- }
if (inoffp == NULL)
foffset_unlock(infp, inoff, 0);
else
@@ -5193,6 +5197,9 @@ out:
foffset_unlock(outfp, outoff, 0);
else
*outoffp = outoff;
+ } else if (foffsets_set) {
+ *inoffp = inoff;
+ *outoffp = outoff;
}
if (outfp != NULL)
fdrop(outfp, td);
diff --git a/sys/kern/vfs_vnops.c b/sys/kern/vfs_vnops.c
index 6451c9e07a60..93f87ddae4de 100644
--- a/sys/kern/vfs_vnops.c
+++ b/sys/kern/vfs_vnops.c
@@ -3443,6 +3443,11 @@ vn_generic_copy_file_range(struct vnode *invp, off_t *inoffp,
interrupted = 0;
dat = NULL;
+ if ((flags & COPY_FILE_RANGE_CLONE) != 0) {
+ error = ENOSYS;
+ goto out;
+ }
+
error = vn_lock(invp, LK_SHARED);
if (error != 0)
goto out;
generated by cgit v1.2.3 (git 2.25.1) at 2025-08-11 14:21:09 +0000