Bring in the most recent version of ipfw and dummynet, developed
and tested over the past two months in the ipfw3-head branch. This
also happens to be the same code available in the Linux and Windows
ports of ipfw and dummynet.
The major enhancement is a completely restructured version of
dummynet, with support for different packet scheduling algorithms
(loadable at runtime), faster queue/pipe lookup, and a much cleaner
internal architecture and kernel/userland ABI which simplifies
future extensions.
In addition to the existing schedulers (FIFO and WF2Q+), we include
a Deficit Round Robin (DRR or RR for brevity) scheduler, and a new,
very fast version of WF2Q+ called QFQ.
Some test code is also present (in sys/netinet/ipfw/test) that
lets you build and test schedulers in userland.
Also, we have added a compatibility layer that understands requests
from the RELENG_7 and RELENG_8 versions of the /sbin/ipfw binaries,
and replies correctly (at least, it does its best; sometimes you
just cannot tell who sent the request and how to answer).
The compatibility layer should make it possible to MFC this code in a
relatively short time.
Some minor glitches (e.g. handling of ipfw set enable/disable,
and a workaround for a bug in RELENG_7's /sbin/ipfw) will be
fixed with separate commits.
CREDITS:
This work has been partly supported by the ONELAB2 project, and
mostly developed by Riccardo Panicucci and myself.
The code for the qfq scheduler is mostly from Fabio Checconi,
and Marta Carbone and Francesco Magno have helped with testing,
debugging and some bug fixes.
2010-03-02 17:40:48 +00:00
|
|
|
/*
|
|
|
|
* Copyright (c) 2010 Riccardo Panicucci, Universita` di Pisa
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* $FreeBSD$
|
|
|
|
*/
|
|
|
|
|
|
|
|
#ifdef _KERNEL
|
|
|
|
#include <sys/malloc.h>
|
|
|
|
#include <sys/socket.h>
|
|
|
|
#include <sys/socketvar.h>
|
|
|
|
#include <sys/kernel.h>
|
|
|
|
#include <sys/mbuf.h>
|
|
|
|
#include <sys/module.h>
|
|
|
|
#include <net/if.h> /* IFNAMSIZ */
|
|
|
|
#include <netinet/in.h>
|
|
|
|
#include <netinet/ip_var.h> /* ipfw_rule_ref */
|
|
|
|
#include <netinet/ip_fw.h> /* flow_id */
|
|
|
|
#include <netinet/ip_dummynet.h>
|
|
|
|
#include <netinet/ipfw/dn_heap.h>
|
|
|
|
#include <netinet/ipfw/ip_dn_private.h>
|
|
|
|
#include <netinet/ipfw/dn_sched.h>
|
|
|
|
#else
|
|
|
|
#include <dn_test.h>
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#define DN_SCHED_RR 3 // XXX Where?
|
|
|
|
|
|
|
|
struct rr_queue {
|
|
|
|
struct dn_queue q; /* Standard queue */
|
|
|
|
int status; /* 1: queue is in the list */
|
|
|
|
int credit; /* Number of bytes to transmit */
|
|
|
|
int quantum; /* quantum * C */
|
|
|
|
struct rr_queue *qnext; /* */
|
|
|
|
};
|
|
|
|
|
|
|
|
/* struct rr_schk contains global config parameters
|
|
|
|
* and is right after dn_schk
|
|
|
|
*/
|
|
|
|
struct rr_schk {
|
|
|
|
int min_q; /* Min quantum */
|
|
|
|
int max_q; /* Max quantum */
|
|
|
|
int q_bytes; /* Bytes per quantum */
|
|
|
|
};
|
|
|
|
|
|
|
|
/* per-instance round robin list, right after dn_sch_inst */
|
|
|
|
struct rr_si {
|
|
|
|
struct rr_queue *head, *tail; /* Pointer to current queue */
|
|
|
|
};
|
|
|
|
|
|
|
|
/* Append a queue to the rr list */
|
|
|
|
static inline void
|
|
|
|
rr_append(struct rr_queue *q, struct rr_si *si)
|
|
|
|
{
|
|
|
|
q->status = 1; /* mark as in-rr_list */
|
|
|
|
q->credit = q->quantum; /* initialize credit */
|
|
|
|
|
|
|
|
/* append to the tail */
|
|
|
|
if (si->head == NULL)
|
|
|
|
si->head = q;
|
|
|
|
else
|
|
|
|
si->tail->qnext = q;
|
|
|
|
si->tail = q; /* advance the tail pointer */
|
|
|
|
q->qnext = si->head; /* make it circular */
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Remove the head queue from circular list. */
|
|
|
|
static inline void
|
|
|
|
rr_remove_head(struct rr_si *si)
|
|
|
|
{
|
|
|
|
if (si->head == NULL)
|
|
|
|
return; /* empty queue */
|
|
|
|
si->head->status = 0;
|
2010-04-19 16:17:30 +00:00
|
|
|
|
Bring in the most recent version of ipfw and dummynet, developed
and tested over the past two months in the ipfw3-head branch. This
also happens to be the same code available in the Linux and Windows
ports of ipfw and dummynet.
The major enhancement is a completely restructured version of
dummynet, with support for different packet scheduling algorithms
(loadable at runtime), faster queue/pipe lookup, and a much cleaner
internal architecture and kernel/userland ABI which simplifies
future extensions.
In addition to the existing schedulers (FIFO and WF2Q+), we include
a Deficit Round Robin (DRR or RR for brevity) scheduler, and a new,
very fast version of WF2Q+ called QFQ.
Some test code is also present (in sys/netinet/ipfw/test) that
lets you build and test schedulers in userland.
Also, we have added a compatibility layer that understands requests
from the RELENG_7 and RELENG_8 versions of the /sbin/ipfw binaries,
and replies correctly (at least, it does its best; sometimes you
just cannot tell who sent the request and how to answer).
The compatibility layer should make it possible to MFC this code in a
relatively short time.
Some minor glitches (e.g. handling of ipfw set enable/disable,
and a workaround for a bug in RELENG_7's /sbin/ipfw) will be
fixed with separate commits.
CREDITS:
This work has been partly supported by the ONELAB2 project, and
mostly developed by Riccardo Panicucci and myself.
The code for the qfq scheduler is mostly from Fabio Checconi,
and Marta Carbone and Francesco Magno have helped with testing,
debugging and some bug fixes.
2010-03-02 17:40:48 +00:00
|
|
|
if (si->head == si->tail) {
|
|
|
|
si->head = si->tail = NULL;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
si->head = si->head->qnext;
|
|
|
|
si->tail->qnext = si->head;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Remove a queue from circular list.
|
|
|
|
* XXX see if ti can be merge with remove_queue()
|
|
|
|
*/
|
|
|
|
static inline void
|
|
|
|
remove_queue_q(struct rr_queue *q, struct rr_si *si)
|
|
|
|
{
|
|
|
|
struct rr_queue *prev;
|
|
|
|
|
|
|
|
if (q->status != 1)
|
|
|
|
return;
|
|
|
|
if (q == si->head) {
|
|
|
|
rr_remove_head(si);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (prev = si->head; prev; prev = prev->qnext) {
|
|
|
|
if (prev->qnext != q)
|
|
|
|
continue;
|
|
|
|
prev->qnext = q->qnext;
|
|
|
|
if (q == si->tail)
|
|
|
|
si->tail = prev;
|
|
|
|
q->status = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
next_pointer(struct rr_si *si)
|
|
|
|
{
|
|
|
|
if (si->head == NULL)
|
|
|
|
return; /* empty queue */
|
|
|
|
|
|
|
|
si->head = si->head->qnext;
|
|
|
|
si->tail = si->tail->qnext;
|
|
|
|
}
|
|
|
|
|
2010-04-19 16:17:30 +00:00
|
|
|
static int
|
Bring in the most recent version of ipfw and dummynet, developed
and tested over the past two months in the ipfw3-head branch. This
also happens to be the same code available in the Linux and Windows
ports of ipfw and dummynet.
The major enhancement is a completely restructured version of
dummynet, with support for different packet scheduling algorithms
(loadable at runtime), faster queue/pipe lookup, and a much cleaner
internal architecture and kernel/userland ABI which simplifies
future extensions.
In addition to the existing schedulers (FIFO and WF2Q+), we include
a Deficit Round Robin (DRR or RR for brevity) scheduler, and a new,
very fast version of WF2Q+ called QFQ.
Some test code is also present (in sys/netinet/ipfw/test) that
lets you build and test schedulers in userland.
Also, we have added a compatibility layer that understands requests
from the RELENG_7 and RELENG_8 versions of the /sbin/ipfw binaries,
and replies correctly (at least, it does its best; sometimes you
just cannot tell who sent the request and how to answer).
The compatibility layer should make it possible to MFC this code in a
relatively short time.
Some minor glitches (e.g. handling of ipfw set enable/disable,
and a workaround for a bug in RELENG_7's /sbin/ipfw) will be
fixed with separate commits.
CREDITS:
This work has been partly supported by the ONELAB2 project, and
mostly developed by Riccardo Panicucci and myself.
The code for the qfq scheduler is mostly from Fabio Checconi,
and Marta Carbone and Francesco Magno have helped with testing,
debugging and some bug fixes.
2010-03-02 17:40:48 +00:00
|
|
|
rr_enqueue(struct dn_sch_inst *_si, struct dn_queue *q, struct mbuf *m)
|
|
|
|
{
|
|
|
|
struct rr_si *si;
|
|
|
|
struct rr_queue *rrq;
|
|
|
|
|
|
|
|
if (m != q->mq.head) {
|
|
|
|
if (dn_enqueue(q, m, 0)) /* packet was dropped */
|
|
|
|
return 1;
|
|
|
|
if (m != q->mq.head)
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-04-19 16:17:30 +00:00
|
|
|
/* If reach this point, queue q was idle */
|
Bring in the most recent version of ipfw and dummynet, developed
and tested over the past two months in the ipfw3-head branch. This
also happens to be the same code available in the Linux and Windows
ports of ipfw and dummynet.
The major enhancement is a completely restructured version of
dummynet, with support for different packet scheduling algorithms
(loadable at runtime), faster queue/pipe lookup, and a much cleaner
internal architecture and kernel/userland ABI which simplifies
future extensions.
In addition to the existing schedulers (FIFO and WF2Q+), we include
a Deficit Round Robin (DRR or RR for brevity) scheduler, and a new,
very fast version of WF2Q+ called QFQ.
Some test code is also present (in sys/netinet/ipfw/test) that
lets you build and test schedulers in userland.
Also, we have added a compatibility layer that understands requests
from the RELENG_7 and RELENG_8 versions of the /sbin/ipfw binaries,
and replies correctly (at least, it does its best; sometimes you
just cannot tell who sent the request and how to answer).
The compatibility layer should make it possible to MFC this code in a
relatively short time.
Some minor glitches (e.g. handling of ipfw set enable/disable,
and a workaround for a bug in RELENG_7's /sbin/ipfw) will be
fixed with separate commits.
CREDITS:
This work has been partly supported by the ONELAB2 project, and
mostly developed by Riccardo Panicucci and myself.
The code for the qfq scheduler is mostly from Fabio Checconi,
and Marta Carbone and Francesco Magno have helped with testing,
debugging and some bug fixes.
2010-03-02 17:40:48 +00:00
|
|
|
si = (struct rr_si *)(_si + 1);
|
|
|
|
rrq = (struct rr_queue *)q;
|
|
|
|
|
|
|
|
if (rrq->status == 1) /* Queue is already in the queue list */
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Insert the queue in the queue list */
|
|
|
|
rr_append(rrq, si);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct mbuf *
|
|
|
|
rr_dequeue(struct dn_sch_inst *_si)
|
|
|
|
{
|
|
|
|
/* Access scheduler instance private data */
|
|
|
|
struct rr_si *si = (struct rr_si *)(_si + 1);
|
|
|
|
struct rr_queue *rrq;
|
|
|
|
uint64_t len;
|
|
|
|
|
|
|
|
while ( (rrq = si->head) ) {
|
|
|
|
struct mbuf *m = rrq->q.mq.head;
|
|
|
|
if ( m == NULL) {
|
|
|
|
/* empty queue, remove from list */
|
|
|
|
rr_remove_head(si);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
len = m->m_pkthdr.len;
|
|
|
|
|
|
|
|
if (len > rrq->credit) {
|
|
|
|
/* Packet too big */
|
|
|
|
rrq->credit += rrq->quantum;
|
|
|
|
/* Try next queue */
|
|
|
|
next_pointer(si);
|
|
|
|
} else {
|
|
|
|
rrq->credit -= len;
|
|
|
|
return dn_dequeue(&rrq->q);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* no packet to dequeue*/
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
rr_config(struct dn_schk *_schk)
|
|
|
|
{
|
|
|
|
struct rr_schk *schk = (struct rr_schk *)(_schk + 1);
|
|
|
|
ND("called");
|
|
|
|
|
|
|
|
/* use reasonable quantums (64..2k bytes, default 1500) */
|
|
|
|
schk->min_q = 64;
|
|
|
|
schk->max_q = 2048;
|
|
|
|
schk->q_bytes = 1500; /* quantum */
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
rr_new_sched(struct dn_sch_inst *_si)
|
|
|
|
{
|
|
|
|
struct rr_si *si = (struct rr_si *)(_si + 1);
|
|
|
|
|
|
|
|
ND("called");
|
|
|
|
si->head = si->tail = NULL;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
rr_free_sched(struct dn_sch_inst *_si)
|
|
|
|
{
|
|
|
|
ND("called");
|
|
|
|
/* Nothing to do? */
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
rr_new_fsk(struct dn_fsk *fs)
|
|
|
|
{
|
|
|
|
struct rr_schk *schk = (struct rr_schk *)(fs->sched + 1);
|
|
|
|
/* par[0] is the weight, par[1] is the quantum step */
|
|
|
|
ipdn_bound_var(&fs->fs.par[0], 1,
|
|
|
|
1, 65536, "RR weight");
|
|
|
|
ipdn_bound_var(&fs->fs.par[1], schk->q_bytes,
|
|
|
|
schk->min_q, schk->max_q, "RR quantum");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
rr_new_queue(struct dn_queue *_q)
|
|
|
|
{
|
|
|
|
struct rr_queue *q = (struct rr_queue *)_q;
|
|
|
|
|
|
|
|
_q->ni.oid.subtype = DN_SCHED_RR;
|
|
|
|
|
|
|
|
q->quantum = _q->fs->fs.par[0] * _q->fs->fs.par[1];
|
|
|
|
ND("called, q->quantum %d", q->quantum);
|
|
|
|
q->credit = q->quantum;
|
|
|
|
q->status = 0;
|
|
|
|
|
|
|
|
if (_q->mq.head != NULL) {
|
|
|
|
/* Queue NOT empty, insert in the queue list */
|
|
|
|
rr_append(q, (struct rr_si *)(_q->_si + 1));
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
rr_free_queue(struct dn_queue *_q)
|
|
|
|
{
|
|
|
|
struct rr_queue *q = (struct rr_queue *)_q;
|
|
|
|
|
|
|
|
ND("called");
|
|
|
|
if (q->status == 1) {
|
|
|
|
struct rr_si *si = (struct rr_si *)(_q->_si + 1);
|
|
|
|
remove_queue_q(q, si);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* RR scheduler descriptor
|
|
|
|
* contains the type of the scheduler, the name, the size of the
|
|
|
|
* structures and function pointers.
|
|
|
|
*/
|
|
|
|
static struct dn_alg rr_desc = {
|
|
|
|
_SI( .type = ) DN_SCHED_RR,
|
|
|
|
_SI( .name = ) "RR",
|
|
|
|
_SI( .flags = ) DN_MULTIQUEUE,
|
|
|
|
|
|
|
|
_SI( .schk_datalen = ) 0,
|
|
|
|
_SI( .si_datalen = ) sizeof(struct rr_si),
|
|
|
|
_SI( .q_datalen = ) sizeof(struct rr_queue) - sizeof(struct dn_queue),
|
|
|
|
|
|
|
|
_SI( .enqueue = ) rr_enqueue,
|
|
|
|
_SI( .dequeue = ) rr_dequeue,
|
|
|
|
|
|
|
|
_SI( .config = ) rr_config,
|
|
|
|
_SI( .destroy = ) NULL,
|
|
|
|
_SI( .new_sched = ) rr_new_sched,
|
|
|
|
_SI( .free_sched = ) rr_free_sched,
|
|
|
|
_SI( .new_fsk = ) rr_new_fsk,
|
|
|
|
_SI( .free_fsk = ) NULL,
|
|
|
|
_SI( .new_queue = ) rr_new_queue,
|
|
|
|
_SI( .free_queue = ) rr_free_queue,
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
DECLARE_DNSCHED_MODULE(dn_rr, &rr_desc);
|