freebsd-dev/sbin/pfctl/pfctl_altq.c

1544 lines
38 KiB
C
Raw Normal View History

/* $OpenBSD: pfctl_altq.c,v 1.93 2007/10/15 02:16:35 deraadt Exp $ */
/*
* Copyright (c) 2002
* Sony Computer Science Laboratories Inc.
* Copyright (c) 2002, 2003 Henning Brauer <henning@openbsd.org>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
2004-03-16 17:24:06 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
#define PFIOC_USE_LATEST
#include <sys/types.h>
#include <sys/ioctl.h>
#include <sys/socket.h>
#include <net/if.h>
#include <netinet/in.h>
#include <net/pfvar.h>
#include <err.h>
#include <errno.h>
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
#include <inttypes.h>
#include <limits.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <net/altq/altq.h>
#include <net/altq/altq_cbq.h>
#include <net/altq/altq_codel.h>
#include <net/altq/altq_priq.h>
#include <net/altq/altq_hfsc.h>
#include <net/altq/altq_fairq.h>
#include "pfctl_parser.h"
#include "pfctl.h"
#define is_sc_null(sc) (((sc) == NULL) || ((sc)->m1 == 0 && (sc)->m2 == 0))
static TAILQ_HEAD(altqs, pf_altq) altqs = TAILQ_HEAD_INITIALIZER(altqs);
static LIST_HEAD(gen_sc, segment) rtsc, lssc;
struct pf_altq *qname_to_pfaltq(const char *, const char *);
u_int32_t qname_to_qid(const char *);
static int eval_pfqueue_cbq(struct pfctl *, struct pf_altq *);
static int cbq_compute_idletime(struct pfctl *, struct pf_altq *);
static int check_commit_cbq(int, int, struct pf_altq *);
static int print_cbq_opts(const struct pf_altq *);
static int print_codel_opts(const struct pf_altq *,
const struct node_queue_opt *);
static int eval_pfqueue_priq(struct pfctl *, struct pf_altq *);
static int check_commit_priq(int, int, struct pf_altq *);
static int print_priq_opts(const struct pf_altq *);
static int eval_pfqueue_hfsc(struct pfctl *, struct pf_altq *);
static int check_commit_hfsc(int, int, struct pf_altq *);
static int print_hfsc_opts(const struct pf_altq *,
const struct node_queue_opt *);
static int eval_pfqueue_fairq(struct pfctl *, struct pf_altq *);
static int print_fairq_opts(const struct pf_altq *,
const struct node_queue_opt *);
static int check_commit_fairq(int, int, struct pf_altq *);
static void gsc_add_sc(struct gen_sc *, struct service_curve *);
static int is_gsc_under_sc(struct gen_sc *,
struct service_curve *);
static void gsc_destroy(struct gen_sc *);
static struct segment *gsc_getentry(struct gen_sc *, double);
static int gsc_add_seg(struct gen_sc *, double, double, double,
double);
static double sc_x2y(struct service_curve *, double);
#ifdef __FreeBSD__
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
u_int64_t getifspeed(int, char *);
#else
u_int32_t getifspeed(char *);
#endif
u_long getifmtu(char *);
int eval_queue_opts(struct pf_altq *, struct node_queue_opt *,
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
u_int64_t);
u_int64_t eval_bwspec(struct node_queue_bw *, u_int64_t);
void print_hfsc_sc(const char *, u_int, u_int, u_int,
const struct node_hfsc_sc *);
void print_fairq_sc(const char *, u_int, u_int, u_int,
const struct node_fairq_sc *);
void
pfaltq_store(struct pf_altq *a)
{
struct pf_altq *altq;
if ((altq = malloc(sizeof(*altq))) == NULL)
err(1, "malloc");
memcpy(altq, a, sizeof(struct pf_altq));
TAILQ_INSERT_TAIL(&altqs, altq, entries);
}
struct pf_altq *
pfaltq_lookup(const char *ifname)
{
struct pf_altq *altq;
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(ifname, altq->ifname, IFNAMSIZ) == 0 &&
altq->qname[0] == 0)
return (altq);
}
return (NULL);
}
struct pf_altq *
qname_to_pfaltq(const char *qname, const char *ifname)
{
struct pf_altq *altq;
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(ifname, altq->ifname, IFNAMSIZ) == 0 &&
strncmp(qname, altq->qname, PF_QNAME_SIZE) == 0)
return (altq);
}
return (NULL);
}
u_int32_t
qname_to_qid(const char *qname)
{
struct pf_altq *altq;
/*
* We guarantee that same named queues on different interfaces
* have the same qid, so we do NOT need to limit matching on
* one interface!
*/
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(qname, altq->qname, PF_QNAME_SIZE) == 0)
return (altq->qid);
}
return (0);
}
void
2008-12-10 20:59:26 +00:00
print_altq(const struct pf_altq *a, unsigned int level,
struct node_queue_bw *bw, struct node_queue_opt *qopts)
{
if (a->qname[0] != 0) {
print_queue(a, level, bw, 1, qopts);
return;
}
#ifdef __FreeBSD__
if (a->local_flags & PFALTQ_FLAG_IF_REMOVED)
printf("INACTIVE ");
#endif
printf("altq on %s ", a->ifname);
switch (a->scheduler) {
case ALTQT_CBQ:
if (!print_cbq_opts(a))
printf("cbq ");
break;
case ALTQT_PRIQ:
if (!print_priq_opts(a))
printf("priq ");
break;
case ALTQT_HFSC:
if (!print_hfsc_opts(a, qopts))
printf("hfsc ");
break;
case ALTQT_FAIRQ:
if (!print_fairq_opts(a, qopts))
printf("fairq ");
break;
case ALTQT_CODEL:
if (!print_codel_opts(a, qopts))
printf("codel ");
break;
}
if (bw != NULL && bw->bw_percent > 0) {
if (bw->bw_percent < 100)
printf("bandwidth %u%% ", bw->bw_percent);
} else
printf("bandwidth %s ", rate2str((double)a->ifbandwidth));
if (a->qlimit != DEFAULT_QLIMIT)
printf("qlimit %u ", a->qlimit);
printf("tbrsize %u ", a->tbrsize);
}
void
2008-12-10 20:59:26 +00:00
print_queue(const struct pf_altq *a, unsigned int level,
struct node_queue_bw *bw, int print_interface,
struct node_queue_opt *qopts)
{
2008-12-10 20:59:26 +00:00
unsigned int i;
#ifdef __FreeBSD__
if (a->local_flags & PFALTQ_FLAG_IF_REMOVED)
printf("INACTIVE ");
#endif
printf("queue ");
for (i = 0; i < level; ++i)
printf(" ");
printf("%s ", a->qname);
if (print_interface)
printf("on %s ", a->ifname);
if (a->scheduler == ALTQT_CBQ || a->scheduler == ALTQT_HFSC ||
a->scheduler == ALTQT_FAIRQ) {
if (bw != NULL && bw->bw_percent > 0) {
if (bw->bw_percent < 100)
printf("bandwidth %u%% ", bw->bw_percent);
} else
printf("bandwidth %s ", rate2str((double)a->bandwidth));
}
if (a->priority != DEFAULT_PRIORITY)
printf("priority %u ", a->priority);
if (a->qlimit != DEFAULT_QLIMIT)
printf("qlimit %u ", a->qlimit);
switch (a->scheduler) {
case ALTQT_CBQ:
print_cbq_opts(a);
break;
case ALTQT_PRIQ:
print_priq_opts(a);
break;
case ALTQT_HFSC:
print_hfsc_opts(a, qopts);
break;
case ALTQT_FAIRQ:
print_fairq_opts(a, qopts);
break;
}
}
/*
* eval_pfaltq computes the discipline parameters.
*/
int
eval_pfaltq(struct pfctl *pf, struct pf_altq *pa, struct node_queue_bw *bw,
struct node_queue_opt *opts)
{
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
u_int64_t rate;
u_int size, errors = 0;
if (bw->bw_absolute > 0)
pa->ifbandwidth = bw->bw_absolute;
else
#ifdef __FreeBSD__
if ((rate = getifspeed(pf->dev, pa->ifname)) == 0) {
#else
if ((rate = getifspeed(pa->ifname)) == 0) {
#endif
fprintf(stderr, "interface %s does not know its bandwidth, "
"please specify an absolute bandwidth\n",
pa->ifname);
errors++;
} else if ((pa->ifbandwidth = eval_bwspec(bw, rate)) == 0)
pa->ifbandwidth = rate;
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
/*
* Limit bandwidth to UINT_MAX for schedulers that aren't 64-bit ready.
*/
if ((pa->scheduler != ALTQT_HFSC) && (pa->ifbandwidth > UINT_MAX)) {
pa->ifbandwidth = UINT_MAX;
warnx("interface %s bandwidth limited to %" PRIu64 " bps "
"because selected scheduler is 32-bit limited\n", pa->ifname,
pa->ifbandwidth);
}
errors += eval_queue_opts(pa, opts, pa->ifbandwidth);
/* if tbrsize is not specified, use heuristics */
if (pa->tbrsize == 0) {
rate = pa->ifbandwidth;
if (rate <= 1 * 1000 * 1000)
size = 1;
else if (rate <= 10 * 1000 * 1000)
size = 4;
else if (rate <= 200 * 1000 * 1000)
size = 8;
else if (rate <= 2500 * 1000 * 1000ULL)
size = 24;
else
size = 128;
size = size * getifmtu(pa->ifname);
pa->tbrsize = size;
}
return (errors);
}
/*
* check_commit_altq does consistency check for each interface
*/
int
check_commit_altq(int dev, int opts)
{
struct pf_altq *altq;
int error = 0;
/* call the discipline check for each interface. */
TAILQ_FOREACH(altq, &altqs, entries) {
if (altq->qname[0] == 0) {
switch (altq->scheduler) {
case ALTQT_CBQ:
error = check_commit_cbq(dev, opts, altq);
break;
case ALTQT_PRIQ:
error = check_commit_priq(dev, opts, altq);
break;
case ALTQT_HFSC:
error = check_commit_hfsc(dev, opts, altq);
break;
case ALTQT_FAIRQ:
error = check_commit_fairq(dev, opts, altq);
break;
default:
break;
}
}
}
return (error);
}
/*
* eval_pfqueue computes the queue parameters.
*/
int
eval_pfqueue(struct pfctl *pf, struct pf_altq *pa, struct node_queue_bw *bw,
struct node_queue_opt *opts)
{
/* should be merged with expand_queue */
struct pf_altq *if_pa, *parent, *altq;
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
u_int64_t bwsum;
int error = 0;
/* find the corresponding interface and copy fields used by queues */
if ((if_pa = pfaltq_lookup(pa->ifname)) == NULL) {
fprintf(stderr, "altq not defined on %s\n", pa->ifname);
return (1);
}
pa->scheduler = if_pa->scheduler;
pa->ifbandwidth = if_pa->ifbandwidth;
if (qname_to_pfaltq(pa->qname, pa->ifname) != NULL) {
fprintf(stderr, "queue %s already exists on interface %s\n",
pa->qname, pa->ifname);
return (1);
}
pa->qid = qname_to_qid(pa->qname);
parent = NULL;
if (pa->parent[0] != 0) {
parent = qname_to_pfaltq(pa->parent, pa->ifname);
if (parent == NULL) {
fprintf(stderr, "parent %s not found for %s\n",
pa->parent, pa->qname);
return (1);
}
pa->parent_qid = parent->qid;
}
if (pa->qlimit == 0)
pa->qlimit = DEFAULT_QLIMIT;
if (pa->scheduler == ALTQT_CBQ || pa->scheduler == ALTQT_HFSC ||
pa->scheduler == ALTQT_FAIRQ) {
pa->bandwidth = eval_bwspec(bw,
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
parent == NULL ? pa->ifbandwidth : parent->bandwidth);
if (pa->bandwidth > pa->ifbandwidth) {
fprintf(stderr, "bandwidth for %s higher than "
"interface\n", pa->qname);
return (1);
}
/* check the sum of the child bandwidth is under parent's */
if (parent != NULL) {
if (pa->bandwidth > parent->bandwidth) {
warnx("bandwidth for %s higher than parent",
pa->qname);
return (1);
}
bwsum = 0;
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname,
IFNAMSIZ) == 0 &&
altq->qname[0] != 0 &&
strncmp(altq->parent, pa->parent,
PF_QNAME_SIZE) == 0)
bwsum += altq->bandwidth;
}
bwsum += pa->bandwidth;
if (bwsum > parent->bandwidth) {
warnx("the sum of the child bandwidth higher"
" than parent \"%s\"", parent->qname);
}
}
}
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
if (eval_queue_opts(pa, opts,
parent == NULL ? pa->ifbandwidth : parent->bandwidth))
return (1);
switch (pa->scheduler) {
case ALTQT_CBQ:
error = eval_pfqueue_cbq(pf, pa);
break;
case ALTQT_PRIQ:
error = eval_pfqueue_priq(pf, pa);
break;
case ALTQT_HFSC:
error = eval_pfqueue_hfsc(pf, pa);
break;
case ALTQT_FAIRQ:
error = eval_pfqueue_fairq(pf, pa);
break;
default:
break;
}
return (error);
}
/*
* CBQ support functions
*/
#define RM_FILTER_GAIN 5 /* log2 of gain, e.g., 5 => 31/32 */
#define RM_NS_PER_SEC (1000000000)
static int
eval_pfqueue_cbq(struct pfctl *pf, struct pf_altq *pa)
{
struct cbq_opts *opts;
u_int ifmtu;
if (pa->priority >= CBQ_MAXPRI) {
warnx("priority out of range: max %d", CBQ_MAXPRI - 1);
return (-1);
}
ifmtu = getifmtu(pa->ifname);
opts = &pa->pq_u.cbq_opts;
if (opts->pktsize == 0) { /* use default */
opts->pktsize = ifmtu;
if (opts->pktsize > MCLBYTES) /* do what TCP does */
opts->pktsize &= ~MCLBYTES;
} else if (opts->pktsize > ifmtu)
opts->pktsize = ifmtu;
if (opts->maxpktsize == 0) /* use default */
opts->maxpktsize = ifmtu;
else if (opts->maxpktsize > ifmtu)
opts->pktsize = ifmtu;
if (opts->pktsize > opts->maxpktsize)
opts->pktsize = opts->maxpktsize;
if (pa->parent[0] == 0)
opts->flags |= (CBQCLF_ROOTCLASS | CBQCLF_WRR);
cbq_compute_idletime(pf, pa);
return (0);
}
/*
* compute ns_per_byte, maxidle, minidle, and offtime
*/
static int
cbq_compute_idletime(struct pfctl *pf, struct pf_altq *pa)
{
struct cbq_opts *opts;
double maxidle_s, maxidle, minidle;
double offtime, nsPerByte, ifnsPerByte, ptime, cptime;
double z, g, f, gton, gtom;
u_int minburst, maxburst;
opts = &pa->pq_u.cbq_opts;
ifnsPerByte = (1.0 / (double)pa->ifbandwidth) * RM_NS_PER_SEC * 8;
minburst = opts->minburst;
maxburst = opts->maxburst;
if (pa->bandwidth == 0)
f = 0.0001; /* small enough? */
else
f = ((double) pa->bandwidth / (double) pa->ifbandwidth);
nsPerByte = ifnsPerByte / f;
ptime = (double)opts->pktsize * ifnsPerByte;
cptime = ptime * (1.0 - f) / f;
if (nsPerByte * (double)opts->maxpktsize > (double)INT_MAX) {
/*
* this causes integer overflow in kernel!
* (bandwidth < 6Kbps when max_pkt_size=1500)
*/
if (pa->bandwidth != 0 && (pf->opts & PF_OPT_QUIET) == 0) {
warnx("queue bandwidth must be larger than %s",
rate2str(ifnsPerByte * (double)opts->maxpktsize /
(double)INT_MAX * (double)pa->ifbandwidth));
fprintf(stderr, "cbq: queue %s is too slow!\n",
pa->qname);
}
nsPerByte = (double)(INT_MAX / opts->maxpktsize);
}
if (maxburst == 0) { /* use default */
if (cptime > 10.0 * 1000000)
maxburst = 4;
else
maxburst = 16;
}
if (minburst == 0) /* use default */
minburst = 2;
if (minburst > maxburst)
minburst = maxburst;
z = (double)(1 << RM_FILTER_GAIN);
g = (1.0 - 1.0 / z);
gton = pow(g, (double)maxburst);
gtom = pow(g, (double)(minburst-1));
maxidle = ((1.0 / f - 1.0) * ((1.0 - gton) / gton));
maxidle_s = (1.0 - g);
if (maxidle > maxidle_s)
maxidle = ptime * maxidle;
else
maxidle = ptime * maxidle_s;
offtime = cptime * (1.0 + 1.0/(1.0 - g) * (1.0 - gtom) / gtom);
minidle = -((double)opts->maxpktsize * (double)nsPerByte);
/* scale parameters */
maxidle = ((maxidle * 8.0) / nsPerByte) *
pow(2.0, (double)RM_FILTER_GAIN);
offtime = (offtime * 8.0) / nsPerByte *
pow(2.0, (double)RM_FILTER_GAIN);
minidle = ((minidle * 8.0) / nsPerByte) *
pow(2.0, (double)RM_FILTER_GAIN);
maxidle = maxidle / 1000.0;
offtime = offtime / 1000.0;
minidle = minidle / 1000.0;
opts->minburst = minburst;
opts->maxburst = maxburst;
opts->ns_per_byte = (u_int)nsPerByte;
opts->maxidle = (u_int)fabs(maxidle);
opts->minidle = (int)minidle;
opts->offtime = (u_int)fabs(offtime);
return (0);
}
static int
check_commit_cbq(int dev, int opts, struct pf_altq *pa)
{
struct pf_altq *altq;
int root_class, default_class;
int error = 0;
/*
* check if cbq has one root queue and one default queue
* for this interface
*/
root_class = default_class = 0;
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
if (altq->pq_u.cbq_opts.flags & CBQCLF_ROOTCLASS)
root_class++;
if (altq->pq_u.cbq_opts.flags & CBQCLF_DEFCLASS)
default_class++;
}
if (root_class != 1) {
warnx("should have one root queue on %s", pa->ifname);
error++;
}
if (default_class != 1) {
warnx("should have one default queue on %s", pa->ifname);
error++;
}
return (error);
}
static int
print_cbq_opts(const struct pf_altq *a)
{
const struct cbq_opts *opts;
opts = &a->pq_u.cbq_opts;
if (opts->flags) {
printf("cbq(");
if (opts->flags & CBQCLF_RED)
printf(" red");
if (opts->flags & CBQCLF_ECN)
printf(" ecn");
if (opts->flags & CBQCLF_RIO)
printf(" rio");
if (opts->flags & CBQCLF_CODEL)
printf(" codel");
if (opts->flags & CBQCLF_CLEARDSCP)
printf(" cleardscp");
if (opts->flags & CBQCLF_FLOWVALVE)
printf(" flowvalve");
if (opts->flags & CBQCLF_BORROW)
printf(" borrow");
if (opts->flags & CBQCLF_WRR)
printf(" wrr");
if (opts->flags & CBQCLF_EFFICIENT)
printf(" efficient");
if (opts->flags & CBQCLF_ROOTCLASS)
printf(" root");
if (opts->flags & CBQCLF_DEFCLASS)
printf(" default");
printf(" ) ");
return (1);
} else
return (0);
}
/*
* PRIQ support functions
*/
static int
eval_pfqueue_priq(struct pfctl *pf, struct pf_altq *pa)
{
struct pf_altq *altq;
if (pa->priority >= PRIQ_MAXPRI) {
warnx("priority out of range: max %d", PRIQ_MAXPRI - 1);
return (-1);
}
/* the priority should be unique for the interface */
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) == 0 &&
altq->qname[0] != 0 && altq->priority == pa->priority) {
warnx("%s and %s have the same priority",
altq->qname, pa->qname);
return (-1);
}
}
return (0);
}
static int
check_commit_priq(int dev, int opts, struct pf_altq *pa)
{
struct pf_altq *altq;
int default_class;
int error = 0;
/*
* check if priq has one default class for this interface
*/
default_class = 0;
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
if (altq->pq_u.priq_opts.flags & PRCF_DEFAULTCLASS)
default_class++;
}
if (default_class != 1) {
warnx("should have one default queue on %s", pa->ifname);
error++;
}
return (error);
}
static int
print_priq_opts(const struct pf_altq *a)
{
const struct priq_opts *opts;
opts = &a->pq_u.priq_opts;
if (opts->flags) {
printf("priq(");
if (opts->flags & PRCF_RED)
printf(" red");
if (opts->flags & PRCF_ECN)
printf(" ecn");
if (opts->flags & PRCF_RIO)
printf(" rio");
if (opts->flags & PRCF_CODEL)
printf(" codel");
if (opts->flags & PRCF_CLEARDSCP)
printf(" cleardscp");
if (opts->flags & PRCF_DEFAULTCLASS)
printf(" default");
printf(" ) ");
return (1);
} else
return (0);
}
/*
* HFSC support functions
*/
static int
eval_pfqueue_hfsc(struct pfctl *pf, struct pf_altq *pa)
{
struct pf_altq *altq, *parent;
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
struct hfsc_opts_v1 *opts;
struct service_curve sc;
opts = &pa->pq_u.hfsc_opts;
if (pa->parent[0] == 0) {
/* root queue */
opts->lssc_m1 = pa->ifbandwidth;
opts->lssc_m2 = pa->ifbandwidth;
opts->lssc_d = 0;
return (0);
}
LIST_INIT(&rtsc);
LIST_INIT(&lssc);
/* if link_share is not specified, use bandwidth */
if (opts->lssc_m2 == 0)
opts->lssc_m2 = pa->bandwidth;
if ((opts->rtsc_m1 > 0 && opts->rtsc_m2 == 0) ||
(opts->lssc_m1 > 0 && opts->lssc_m2 == 0) ||
(opts->ulsc_m1 > 0 && opts->ulsc_m2 == 0)) {
warnx("m2 is zero for %s", pa->qname);
return (-1);
}
if ((opts->rtsc_m1 < opts->rtsc_m2 && opts->rtsc_m1 != 0) ||
(opts->lssc_m1 < opts->lssc_m2 && opts->lssc_m1 != 0) ||
(opts->ulsc_m1 < opts->ulsc_m2 && opts->ulsc_m1 != 0)) {
warnx("m1 must be zero for convex curve: %s", pa->qname);
return (-1);
}
/*
* admission control:
* for the real-time service curve, the sum of the service curves
* should not exceed 80% of the interface bandwidth. 20% is reserved
* not to over-commit the actual interface bandwidth.
* for the linkshare service curve, the sum of the child service
* curve should not exceed the parent service curve.
* for the upper-limit service curve, the assigned bandwidth should
* be smaller than the interface bandwidth, and the upper-limit should
* be larger than the real-time service curve when both are defined.
*/
parent = qname_to_pfaltq(pa->parent, pa->ifname);
if (parent == NULL)
errx(1, "parent %s not found for %s", pa->parent, pa->qname);
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
/* if the class has a real-time service curve, add it. */
if (opts->rtsc_m2 != 0 && altq->pq_u.hfsc_opts.rtsc_m2 != 0) {
sc.m1 = altq->pq_u.hfsc_opts.rtsc_m1;
sc.d = altq->pq_u.hfsc_opts.rtsc_d;
sc.m2 = altq->pq_u.hfsc_opts.rtsc_m2;
gsc_add_sc(&rtsc, &sc);
}
if (strncmp(altq->parent, pa->parent, PF_QNAME_SIZE) != 0)
continue;
/* if the class has a linkshare service curve, add it. */
if (opts->lssc_m2 != 0 && altq->pq_u.hfsc_opts.lssc_m2 != 0) {
sc.m1 = altq->pq_u.hfsc_opts.lssc_m1;
sc.d = altq->pq_u.hfsc_opts.lssc_d;
sc.m2 = altq->pq_u.hfsc_opts.lssc_m2;
gsc_add_sc(&lssc, &sc);
}
}
/* check the real-time service curve. reserve 20% of interface bw */
if (opts->rtsc_m2 != 0) {
/* add this queue to the sum */
sc.m1 = opts->rtsc_m1;
sc.d = opts->rtsc_d;
sc.m2 = opts->rtsc_m2;
gsc_add_sc(&rtsc, &sc);
/* compare the sum with 80% of the interface */
sc.m1 = 0;
sc.d = 0;
sc.m2 = pa->ifbandwidth / 100 * 80;
if (!is_gsc_under_sc(&rtsc, &sc)) {
warnx("real-time sc exceeds 80%% of the interface "
"bandwidth (%s)", rate2str((double)sc.m2));
goto err_ret;
}
}
/* check the linkshare service curve. */
if (opts->lssc_m2 != 0) {
/* add this queue to the child sum */
sc.m1 = opts->lssc_m1;
sc.d = opts->lssc_d;
sc.m2 = opts->lssc_m2;
gsc_add_sc(&lssc, &sc);
/* compare the sum of the children with parent's sc */
sc.m1 = parent->pq_u.hfsc_opts.lssc_m1;
sc.d = parent->pq_u.hfsc_opts.lssc_d;
sc.m2 = parent->pq_u.hfsc_opts.lssc_m2;
if (!is_gsc_under_sc(&lssc, &sc)) {
warnx("linkshare sc exceeds parent's sc");
goto err_ret;
}
}
/* check the upper-limit service curve. */
if (opts->ulsc_m2 != 0) {
if (opts->ulsc_m1 > pa->ifbandwidth ||
opts->ulsc_m2 > pa->ifbandwidth) {
warnx("upper-limit larger than interface bandwidth");
goto err_ret;
}
if (opts->rtsc_m2 != 0 && opts->rtsc_m2 > opts->ulsc_m2) {
warnx("upper-limit sc smaller than real-time sc");
goto err_ret;
}
}
gsc_destroy(&rtsc);
gsc_destroy(&lssc);
return (0);
err_ret:
gsc_destroy(&rtsc);
gsc_destroy(&lssc);
return (-1);
}
/*
* FAIRQ support functions
*/
static int
eval_pfqueue_fairq(struct pfctl *pf __unused, struct pf_altq *pa)
{
struct pf_altq *altq, *parent;
struct fairq_opts *opts;
struct service_curve sc;
opts = &pa->pq_u.fairq_opts;
if (pa->parent[0] == 0) {
/* root queue */
opts->lssc_m1 = pa->ifbandwidth;
opts->lssc_m2 = pa->ifbandwidth;
opts->lssc_d = 0;
return (0);
}
LIST_INIT(&lssc);
/* if link_share is not specified, use bandwidth */
if (opts->lssc_m2 == 0)
opts->lssc_m2 = pa->bandwidth;
/*
* admission control:
* for the real-time service curve, the sum of the service curves
* should not exceed 80% of the interface bandwidth. 20% is reserved
* not to over-commit the actual interface bandwidth.
* for the link-sharing service curve, the sum of the child service
* curve should not exceed the parent service curve.
* for the upper-limit service curve, the assigned bandwidth should
* be smaller than the interface bandwidth, and the upper-limit should
* be larger than the real-time service curve when both are defined.
*/
parent = qname_to_pfaltq(pa->parent, pa->ifname);
if (parent == NULL)
errx(1, "parent %s not found for %s", pa->parent, pa->qname);
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
if (strncmp(altq->parent, pa->parent, PF_QNAME_SIZE) != 0)
continue;
/* if the class has a link-sharing service curve, add it. */
if (opts->lssc_m2 != 0 && altq->pq_u.fairq_opts.lssc_m2 != 0) {
sc.m1 = altq->pq_u.fairq_opts.lssc_m1;
sc.d = altq->pq_u.fairq_opts.lssc_d;
sc.m2 = altq->pq_u.fairq_opts.lssc_m2;
gsc_add_sc(&lssc, &sc);
}
}
/* check the link-sharing service curve. */
if (opts->lssc_m2 != 0) {
sc.m1 = parent->pq_u.fairq_opts.lssc_m1;
sc.d = parent->pq_u.fairq_opts.lssc_d;
sc.m2 = parent->pq_u.fairq_opts.lssc_m2;
if (!is_gsc_under_sc(&lssc, &sc)) {
warnx("link-sharing sc exceeds parent's sc");
goto err_ret;
}
}
gsc_destroy(&lssc);
return (0);
err_ret:
gsc_destroy(&lssc);
return (-1);
}
static int
check_commit_hfsc(int dev, int opts, struct pf_altq *pa)
{
struct pf_altq *altq, *def = NULL;
int default_class;
int error = 0;
/* check if hfsc has one default queue for this interface */
default_class = 0;
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
if (altq->parent[0] == 0) /* dummy root */
continue;
if (altq->pq_u.hfsc_opts.flags & HFCF_DEFAULTCLASS) {
default_class++;
def = altq;
}
}
if (default_class != 1) {
warnx("should have one default queue on %s", pa->ifname);
return (1);
}
/* make sure the default queue is a leaf */
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
if (strncmp(altq->parent, def->qname, PF_QNAME_SIZE) == 0) {
warnx("default queue is not a leaf");
error++;
}
}
return (error);
}
static int
check_commit_fairq(int dev __unused, int opts __unused, struct pf_altq *pa)
{
struct pf_altq *altq, *def = NULL;
int default_class;
int error = 0;
/* check if fairq has one default queue for this interface */
default_class = 0;
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
if (altq->pq_u.fairq_opts.flags & FARF_DEFAULTCLASS) {
default_class++;
def = altq;
}
}
if (default_class != 1) {
warnx("should have one default queue on %s", pa->ifname);
return (1);
}
/* make sure the default queue is a leaf */
TAILQ_FOREACH(altq, &altqs, entries) {
if (strncmp(altq->ifname, pa->ifname, IFNAMSIZ) != 0)
continue;
if (altq->qname[0] == 0) /* this is for interface */
continue;
if (strncmp(altq->parent, def->qname, PF_QNAME_SIZE) == 0) {
warnx("default queue is not a leaf");
error++;
}
}
return (error);
}
static int
print_hfsc_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
{
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
const struct hfsc_opts_v1 *opts;
const struct node_hfsc_sc *rtsc, *lssc, *ulsc;
opts = &a->pq_u.hfsc_opts;
if (qopts == NULL)
rtsc = lssc = ulsc = NULL;
else {
rtsc = &qopts->data.hfsc_opts.realtime;
lssc = &qopts->data.hfsc_opts.linkshare;
ulsc = &qopts->data.hfsc_opts.upperlimit;
}
if (opts->flags || opts->rtsc_m2 != 0 || opts->ulsc_m2 != 0 ||
(opts->lssc_m2 != 0 && (opts->lssc_m2 != a->bandwidth ||
opts->lssc_d != 0))) {
printf("hfsc(");
if (opts->flags & HFCF_RED)
printf(" red");
if (opts->flags & HFCF_ECN)
printf(" ecn");
if (opts->flags & HFCF_RIO)
printf(" rio");
if (opts->flags & HFCF_CODEL)
printf(" codel");
if (opts->flags & HFCF_CLEARDSCP)
printf(" cleardscp");
if (opts->flags & HFCF_DEFAULTCLASS)
printf(" default");
if (opts->rtsc_m2 != 0)
print_hfsc_sc("realtime", opts->rtsc_m1, opts->rtsc_d,
opts->rtsc_m2, rtsc);
if (opts->lssc_m2 != 0 && (opts->lssc_m2 != a->bandwidth ||
opts->lssc_d != 0))
print_hfsc_sc("linkshare", opts->lssc_m1, opts->lssc_d,
opts->lssc_m2, lssc);
if (opts->ulsc_m2 != 0)
print_hfsc_sc("upperlimit", opts->ulsc_m1, opts->ulsc_d,
opts->ulsc_m2, ulsc);
printf(" ) ");
return (1);
} else
return (0);
}
static int
print_codel_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
{
const struct codel_opts *opts;
opts = &a->pq_u.codel_opts;
if (opts->target || opts->interval || opts->ecn) {
printf("codel(");
if (opts->target)
printf(" target %d", opts->target);
if (opts->interval)
printf(" interval %d", opts->interval);
if (opts->ecn)
printf("ecn");
printf(" ) ");
return (1);
}
return (0);
}
static int
print_fairq_opts(const struct pf_altq *a, const struct node_queue_opt *qopts)
{
const struct fairq_opts *opts;
const struct node_fairq_sc *loc_lssc;
opts = &a->pq_u.fairq_opts;
if (qopts == NULL)
loc_lssc = NULL;
else
loc_lssc = &qopts->data.fairq_opts.linkshare;
if (opts->flags ||
(opts->lssc_m2 != 0 && (opts->lssc_m2 != a->bandwidth ||
opts->lssc_d != 0))) {
printf("fairq(");
if (opts->flags & FARF_RED)
printf(" red");
if (opts->flags & FARF_ECN)
printf(" ecn");
if (opts->flags & FARF_RIO)
printf(" rio");
if (opts->flags & FARF_CODEL)
printf(" codel");
if (opts->flags & FARF_CLEARDSCP)
printf(" cleardscp");
if (opts->flags & FARF_DEFAULTCLASS)
printf(" default");
if (opts->lssc_m2 != 0 && (opts->lssc_m2 != a->bandwidth ||
opts->lssc_d != 0))
print_fairq_sc("linkshare", opts->lssc_m1, opts->lssc_d,
opts->lssc_m2, loc_lssc);
printf(" ) ");
return (1);
} else
return (0);
}
/*
* admission control using generalized service curve
*/
/* add a new service curve to a generalized service curve */
static void
gsc_add_sc(struct gen_sc *gsc, struct service_curve *sc)
{
if (is_sc_null(sc))
return;
if (sc->d != 0)
gsc_add_seg(gsc, 0.0, 0.0, (double)sc->d, (double)sc->m1);
gsc_add_seg(gsc, (double)sc->d, 0.0, INFINITY, (double)sc->m2);
}
/*
* check whether all points of a generalized service curve have
* their y-coordinates no larger than a given two-piece linear
* service curve.
*/
static int
is_gsc_under_sc(struct gen_sc *gsc, struct service_curve *sc)
{
struct segment *s, *last, *end;
double y;
if (is_sc_null(sc)) {
if (LIST_EMPTY(gsc))
return (1);
LIST_FOREACH(s, gsc, _next) {
if (s->m != 0)
return (0);
}
return (1);
}
/*
* gsc has a dummy entry at the end with x = INFINITY.
* loop through up to this dummy entry.
*/
end = gsc_getentry(gsc, INFINITY);
if (end == NULL)
return (1);
last = NULL;
for (s = LIST_FIRST(gsc); s != end; s = LIST_NEXT(s, _next)) {
if (s->y > sc_x2y(sc, s->x))
return (0);
last = s;
}
/* last now holds the real last segment */
if (last == NULL)
return (1);
if (last->m > sc->m2)
return (0);
if (last->x < sc->d && last->m > sc->m1) {
y = last->y + (sc->d - last->x) * last->m;
if (y > sc_x2y(sc, sc->d))
return (0);
}
return (1);
}
static void
gsc_destroy(struct gen_sc *gsc)
{
struct segment *s;
while ((s = LIST_FIRST(gsc)) != NULL) {
LIST_REMOVE(s, _next);
free(s);
}
}
/*
* return a segment entry starting at x.
* if gsc has no entry starting at x, a new entry is created at x.
*/
static struct segment *
gsc_getentry(struct gen_sc *gsc, double x)
{
struct segment *new, *prev, *s;
prev = NULL;
LIST_FOREACH(s, gsc, _next) {
if (s->x == x)
return (s); /* matching entry found */
else if (s->x < x)
prev = s;
else
break;
}
/* we have to create a new entry */
if ((new = calloc(1, sizeof(struct segment))) == NULL)
return (NULL);
new->x = x;
if (x == INFINITY || s == NULL)
new->d = 0;
else if (s->x == INFINITY)
new->d = INFINITY;
else
new->d = s->x - x;
if (prev == NULL) {
/* insert the new entry at the head of the list */
new->y = 0;
new->m = 0;
LIST_INSERT_HEAD(gsc, new, _next);
} else {
/*
* the start point intersects with the segment pointed by
* prev. divide prev into 2 segments
*/
if (x == INFINITY) {
prev->d = INFINITY;
if (prev->m == 0)
new->y = prev->y;
else
new->y = INFINITY;
} else {
prev->d = x - prev->x;
new->y = prev->d * prev->m + prev->y;
}
new->m = prev->m;
LIST_INSERT_AFTER(prev, new, _next);
}
return (new);
}
/* add a segment to a generalized service curve */
static int
gsc_add_seg(struct gen_sc *gsc, double x, double y, double d, double m)
{
struct segment *start, *end, *s;
double x2;
if (d == INFINITY)
x2 = INFINITY;
else
x2 = x + d;
start = gsc_getentry(gsc, x);
end = gsc_getentry(gsc, x2);
if (start == NULL || end == NULL)
return (-1);
for (s = start; s != end; s = LIST_NEXT(s, _next)) {
s->m += m;
s->y += y + (s->x - x) * m;
}
end = gsc_getentry(gsc, INFINITY);
for (; s != end; s = LIST_NEXT(s, _next)) {
s->y += m * d;
}
return (0);
}
/* get y-projection of a service curve */
static double
sc_x2y(struct service_curve *sc, double x)
{
double y;
if (x <= (double)sc->d)
/* y belongs to the 1st segment */
y = x * (double)sc->m1;
else
/* y belongs to the 2nd segment */
y = (double)sc->d * (double)sc->m1
+ (x - (double)sc->d) * (double)sc->m2;
return (y);
}
/*
* misc utilities
*/
#define R2S_BUFS 8
#define RATESTR_MAX 16
char *
rate2str(double rate)
{
char *buf;
static char r2sbuf[R2S_BUFS][RATESTR_MAX]; /* ring bufer */
static int idx = 0;
int i;
static const char unit[] = " KMG";
buf = r2sbuf[idx++];
if (idx == R2S_BUFS)
idx = 0;
for (i = 0; rate >= 1000 && i <= 3; i++)
rate /= 1000;
if ((int)(rate * 100) % 100)
snprintf(buf, RATESTR_MAX, "%.2f%cb", rate, unit[i]);
else
snprintf(buf, RATESTR_MAX, "%d%cb", (int)rate, unit[i]);
return (buf);
}
#ifdef __FreeBSD__
/*
* XXX
* FreeBSD does not have SIOCGIFDATA.
* To emulate this, DIOCGIFSPEED ioctl added to pf.
*/
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
u_int64_t
getifspeed(int pfdev, char *ifname)
{
struct pf_ifspeed io;
bzero(&io, sizeof io);
if (strlcpy(io.ifname, ifname, IFNAMSIZ) >=
sizeof(io.ifname))
errx(1, "getifspeed: strlcpy");
if (ioctl(pfdev, DIOCGIFSPEED, &io) == -1)
err(1, "DIOCGIFSPEED");
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
return (io.baudrate);
}
#else
u_int32_t
getifspeed(char *ifname)
{
int s;
struct ifreq ifr;
struct if_data ifrdat;
if ((s = socket(get_socket_domain(), SOCK_DGRAM, 0)) < 0)
err(1, "socket");
bzero(&ifr, sizeof(ifr));
if (strlcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name)) >=
sizeof(ifr.ifr_name))
errx(1, "getifspeed: strlcpy");
ifr.ifr_data = (caddr_t)&ifrdat;
if (ioctl(s, SIOCGIFDATA, (caddr_t)&ifr) == -1)
err(1, "SIOCGIFDATA");
if (close(s))
err(1, "close");
return ((u_int32_t)ifrdat.ifi_baudrate);
}
#endif
u_long
getifmtu(char *ifname)
{
int s;
struct ifreq ifr;
if ((s = socket(get_socket_domain(), SOCK_DGRAM, 0)) < 0)
err(1, "socket");
bzero(&ifr, sizeof(ifr));
if (strlcpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name)) >=
sizeof(ifr.ifr_name))
errx(1, "getifmtu: strlcpy");
if (ioctl(s, SIOCGIFMTU, (caddr_t)&ifr) == -1)
#ifdef __FreeBSD__
ifr.ifr_mtu = 1500;
#else
err(1, "SIOCGIFMTU");
#endif
if (close(s))
err(1, "close");
if (ifr.ifr_mtu > 0)
return (ifr.ifr_mtu);
else {
warnx("could not get mtu for %s, assuming 1500", ifname);
return (1500);
}
}
int
eval_queue_opts(struct pf_altq *pa, struct node_queue_opt *opts,
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
u_int64_t ref_bw)
{
int errors = 0;
switch (pa->scheduler) {
case ALTQT_CBQ:
pa->pq_u.cbq_opts = opts->data.cbq_opts;
break;
case ALTQT_PRIQ:
pa->pq_u.priq_opts = opts->data.priq_opts;
break;
case ALTQT_HFSC:
pa->pq_u.hfsc_opts.flags = opts->data.hfsc_opts.flags;
if (opts->data.hfsc_opts.linkshare.used) {
pa->pq_u.hfsc_opts.lssc_m1 =
eval_bwspec(&opts->data.hfsc_opts.linkshare.m1,
ref_bw);
pa->pq_u.hfsc_opts.lssc_m2 =
eval_bwspec(&opts->data.hfsc_opts.linkshare.m2,
ref_bw);
pa->pq_u.hfsc_opts.lssc_d =
opts->data.hfsc_opts.linkshare.d;
}
if (opts->data.hfsc_opts.realtime.used) {
pa->pq_u.hfsc_opts.rtsc_m1 =
eval_bwspec(&opts->data.hfsc_opts.realtime.m1,
ref_bw);
pa->pq_u.hfsc_opts.rtsc_m2 =
eval_bwspec(&opts->data.hfsc_opts.realtime.m2,
ref_bw);
pa->pq_u.hfsc_opts.rtsc_d =
opts->data.hfsc_opts.realtime.d;
}
if (opts->data.hfsc_opts.upperlimit.used) {
pa->pq_u.hfsc_opts.ulsc_m1 =
eval_bwspec(&opts->data.hfsc_opts.upperlimit.m1,
ref_bw);
pa->pq_u.hfsc_opts.ulsc_m2 =
eval_bwspec(&opts->data.hfsc_opts.upperlimit.m2,
ref_bw);
pa->pq_u.hfsc_opts.ulsc_d =
opts->data.hfsc_opts.upperlimit.d;
}
break;
case ALTQT_FAIRQ:
pa->pq_u.fairq_opts.flags = opts->data.fairq_opts.flags;
pa->pq_u.fairq_opts.nbuckets = opts->data.fairq_opts.nbuckets;
pa->pq_u.fairq_opts.hogs_m1 =
eval_bwspec(&opts->data.fairq_opts.hogs_bw, ref_bw);
if (opts->data.fairq_opts.linkshare.used) {
pa->pq_u.fairq_opts.lssc_m1 =
eval_bwspec(&opts->data.fairq_opts.linkshare.m1,
ref_bw);
pa->pq_u.fairq_opts.lssc_m2 =
eval_bwspec(&opts->data.fairq_opts.linkshare.m2,
ref_bw);
pa->pq_u.fairq_opts.lssc_d =
opts->data.fairq_opts.linkshare.d;
}
break;
case ALTQT_CODEL:
pa->pq_u.codel_opts.target = opts->data.codel_opts.target;
pa->pq_u.codel_opts.interval = opts->data.codel_opts.interval;
pa->pq_u.codel_opts.ecn = opts->data.codel_opts.ecn;
break;
default:
warnx("eval_queue_opts: unknown scheduler type %u",
opts->qtype);
errors++;
break;
}
return (errors);
}
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
/*
* If absolute bandwidth if set, return the lesser of that value and the
* reference bandwidth. Limiting to the reference bandwidth allows simple
* limiting of configured bandwidth parameters for schedulers that are
* 32-bit limited, as the root/interface bandwidth (top-level reference
* bandwidth) will be properly limited in that case.
*
* Otherwise, if the absolute bandwidth is not set, return given percentage
* of reference bandwidth.
*/
u_int64_t
eval_bwspec(struct node_queue_bw *bw, u_int64_t ref_bw)
{
if (bw->bw_absolute > 0)
Extended pf(4) ioctl interface and pfctl(8) to allow bandwidths of 2^32 bps or greater to be used. Prior to this, bandwidth parameters would simply wrap at the 2^32 boundary. The computations in the HFSC scheduler and token bucket regulator have been modified to operate correctly up to at least 100 Gbps. No other algorithms have been examined or modified for correct operation above 2^32 bps (some may have existing computation resolution or overflow issues at rates below that threshold). pfctl(8) will now limit non-HFSC bandwidth parameters to 2^32 - 1 before passing them to the kernel. The extensions to the pf(4) ioctl interface have been made in a backwards-compatible way by versioning affected data structures, supporting all versions in the kernel, and implementing macros that will cause existing code that consumes that interface to use version 0 without source modifications. If version 0 consumers of the interface are used against a new kernel that has had bandwidth parameters of 2^32 or greater configured by updated tools, such bandwidth parameters will be reported as 2^32 - 1 bps by those old consumers. All in-tree consumers of the pf(4) interface have been updated. To update out-of-tree consumers to the latest version of the interface, define PFIOC_USE_LATEST ahead of any includes and use the code of pfctl(8) as a guide for the ioctls of interest. PR: 211730 Reviewed by: jmallett, kp, loos MFC after: 2 weeks Relnotes: yes Sponsored by: RG Nets Differential Revision: https://reviews.freebsd.org/D16782
2018-08-22 19:38:48 +00:00
return (MIN(bw->bw_absolute, ref_bw));
if (bw->bw_percent > 0)
return (ref_bw / 100 * bw->bw_percent);
return (0);
}
void
print_hfsc_sc(const char *scname, u_int m1, u_int d, u_int m2,
const struct node_hfsc_sc *sc)
{
printf(" %s", scname);
if (d != 0) {
printf("(");
if (sc != NULL && sc->m1.bw_percent > 0)
printf("%u%%", sc->m1.bw_percent);
else
printf("%s", rate2str((double)m1));
printf(" %u", d);
}
if (sc != NULL && sc->m2.bw_percent > 0)
printf(" %u%%", sc->m2.bw_percent);
else
printf(" %s", rate2str((double)m2));
if (d != 0)
printf(")");
}
void
print_fairq_sc(const char *scname, u_int m1, u_int d, u_int m2,
const struct node_fairq_sc *sc)
{
printf(" %s", scname);
if (d != 0) {
printf("(");
if (sc != NULL && sc->m1.bw_percent > 0)
printf("%u%%", sc->m1.bw_percent);
else
printf("%s", rate2str((double)m1));
printf(" %u", d);
}
if (sc != NULL && sc->m2.bw_percent > 0)
printf(" %u%%", sc->m2.bw_percent);
else
printf(" %s", rate2str((double)m2));
if (d != 0)
printf(")");
}