freebsd-nq/sys/kern/kern_syscalls.c
Mateusz Guzik 79ca7cbf09 Avoid calls to syscall_thread_enter/exit for statically defined syscalls
The entire mechanism is rarely used and is quite not performant due to
atomci ops on the syscall table. It also has added overhead for completely
unrelated syscalls.

Reduce it by avoiding the func calls if possible (which consistutes vast
majority of cases).

Provides about 3% syscall rate speed up for getuid on Broadwell.
2018-05-07 22:29:32 +00:00

257 lines
6.5 KiB
C

/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 1999 Assar Westerlund
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/sx.h>
#include <sys/syscall.h>
#include <sys/sysent.h>
#include <sys/sysproto.h>
#include <sys/systm.h>
#include <machine/atomic.h>
/*
* Acts like "nosys" but can be identified in sysent for dynamic call
* number assignment for a limited number of calls.
*
* Place holder for system call slots reserved for loadable modules.
*/
int
lkmnosys(struct thread *td, struct nosys_args *args)
{
return (nosys(td, args));
}
int
lkmressys(struct thread *td, struct nosys_args *args)
{
return (nosys(td, args));
}
static void
syscall_thread_drain(struct sysent *se)
{
u_int32_t cnt, oldcnt;
do {
oldcnt = se->sy_thrcnt;
KASSERT((oldcnt & SY_THR_STATIC) == 0,
("drain on static syscall"));
cnt = oldcnt | SY_THR_DRAINING;
} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
while (atomic_cmpset_32(&se->sy_thrcnt, SY_THR_DRAINING,
SY_THR_ABSENT) == 0)
pause("scdrn", hz/2);
}
int
_syscall_thread_enter(struct thread *td, struct sysent *se)
{
u_int32_t cnt, oldcnt;
do {
oldcnt = se->sy_thrcnt;
if ((oldcnt & (SY_THR_DRAINING | SY_THR_ABSENT)) != 0)
return (ENOSYS);
cnt = oldcnt + SY_THR_INCR;
} while (atomic_cmpset_acq_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
return (0);
}
void
_syscall_thread_exit(struct thread *td, struct sysent *se)
{
u_int32_t cnt, oldcnt;
do {
oldcnt = se->sy_thrcnt;
cnt = oldcnt - SY_THR_INCR;
} while (atomic_cmpset_rel_32(&se->sy_thrcnt, oldcnt, cnt) == 0);
}
int
kern_syscall_register(struct sysent *sysents, int *offset,
struct sysent *new_sysent, struct sysent *old_sysent, int flags)
{
int i;
if ((flags & ~SY_THR_STATIC) != 0)
return (EINVAL);
if (*offset == NO_SYSCALL) {
for (i = 1; i < SYS_MAXSYSCALL; ++i)
if (sysents[i].sy_call == (sy_call_t *)lkmnosys)
break;
if (i == SYS_MAXSYSCALL)
return (ENFILE);
*offset = i;
} else if (*offset < 0 || *offset >= SYS_MAXSYSCALL)
return (EINVAL);
else if (sysents[*offset].sy_call != (sy_call_t *)lkmnosys &&
sysents[*offset].sy_call != (sy_call_t *)lkmressys)
return (EEXIST);
KASSERT(sysents[*offset].sy_thrcnt == SY_THR_ABSENT,
("dynamic syscall is not protected"));
*old_sysent = sysents[*offset];
new_sysent->sy_thrcnt = SY_THR_ABSENT;
sysents[*offset] = *new_sysent;
atomic_store_rel_32(&sysents[*offset].sy_thrcnt, flags);
return (0);
}
int
kern_syscall_deregister(struct sysent *sysents, int offset,
const struct sysent *old_sysent)
{
struct sysent *se;
if (offset == 0)
return (0); /* XXX? */
se = &sysents[offset];
if ((se->sy_thrcnt & SY_THR_STATIC) != 0)
return (EINVAL);
syscall_thread_drain(se);
sysent[offset] = *old_sysent;
return (0);
}
int
syscall_module_handler(struct module *mod, int what, void *arg)
{
return (kern_syscall_module_handler(sysent, mod, what, arg));
}
int
kern_syscall_module_handler(struct sysent *sysents, struct module *mod,
int what, void *arg)
{
struct syscall_module_data *data = arg;
modspecific_t ms;
int error;
switch (what) {
case MOD_LOAD:
error = kern_syscall_register(sysents, data->offset,
data->new_sysent, &data->old_sysent, data->flags);
if (error) {
/* Leave a mark so we know to safely unload below. */
data->offset = NULL;
return (error);
}
ms.intval = *data->offset;
MOD_XLOCK;
module_setspecific(mod, &ms);
MOD_XUNLOCK;
if (data->chainevh)
error = data->chainevh(mod, what, data->chainarg);
return (error);
case MOD_UNLOAD:
/*
* MOD_LOAD failed, so just return without calling the
* chained handler since we didn't pass along the MOD_LOAD
* event.
*/
if (data->offset == NULL)
return (0);
if (data->chainevh) {
error = data->chainevh(mod, what, data->chainarg);
if (error)
return error;
}
error = kern_syscall_deregister(sysents, *data->offset,
&data->old_sysent);
return (error);
default:
if (data->chainevh)
return (data->chainevh(mod, what, data->chainarg));
return (EOPNOTSUPP);
}
/* NOTREACHED */
}
int
syscall_helper_register(struct syscall_helper_data *sd, int flags)
{
return (kern_syscall_helper_register(sysent, sd, flags));
}
int
kern_syscall_helper_register(struct sysent *sysents,
struct syscall_helper_data *sd, int flags)
{
struct syscall_helper_data *sd1;
int error;
for (sd1 = sd; sd1->syscall_no != NO_SYSCALL; sd1++) {
error = kern_syscall_register(sysents, &sd1->syscall_no,
&sd1->new_sysent, &sd1->old_sysent, flags);
if (error != 0) {
kern_syscall_helper_unregister(sysents, sd);
return (error);
}
sd1->registered = 1;
}
return (0);
}
int
syscall_helper_unregister(struct syscall_helper_data *sd)
{
return (kern_syscall_helper_unregister(sysent, sd));
}
int
kern_syscall_helper_unregister(struct sysent *sysents,
struct syscall_helper_data *sd)
{
struct syscall_helper_data *sd1;
for (sd1 = sd; sd1->registered != 0; sd1++) {
kern_syscall_deregister(sysents, sd1->syscall_no,
&sd1->old_sysent);
sd1->registered = 0;
}
return (0);
}