freebsd-dev/sys/netinet/ipfw/dn_sched.h

192 lines
6.7 KiB
C
Raw Normal View History

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, Luigi Rizzo, 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.
*/
/*
* The API to write a packet scheduling algorithm for dummynet.
*
* $FreeBSD$
*/
#ifndef _DN_SCHED_H
#define _DN_SCHED_H
#define DN_MULTIQUEUE 0x01
/*
* Descriptor for a scheduling algorithm.
* Contains all function pointers for a given scheduler
* This is typically created when a module is loaded, and stored
* in a global list of schedulers.
*/
struct dn_alg {
uint32_t type; /* the scheduler type */
const char *name; /* scheduler name */
uint32_t flags; /* DN_MULTIQUEUE if supports multiple queues */
/*
* The following define the size of 3 optional data structures
* that may need to be allocated at runtime, and are appended
* to each of the base data structures: scheduler, sched.inst,
* and queue. We don't have a per-flowset structure.
*/
/* + parameters attached to the template, e.g.
* default queue sizes, weights, quantum size, and so on;
*/
size_t schk_datalen;
/* + per-instance parameters, such as timestamps,
* containers for queues, etc;
*/
size_t si_datalen;
size_t q_datalen; /* per-queue parameters (e.g. S,F) */
/*
* Methods implemented by the scheduler:
* enqueue enqueue packet 'm' on scheduler 's', queue 'q'.
* q is NULL for !MULTIQUEUE.
* Return 0 on success, 1 on drop (packet consumed anyways).
2010-03-21 15:52:55 +00:00
* Note that q should be interpreted only as a hint
* on the flow that the mbuf belongs to: while a
* scheduler will normally enqueue m into q, it is ok
* to leave q alone and put the mbuf elsewhere.
* This function is called in two cases:
* - when a new packet arrives to the scheduler;
* - when a scheduler is reconfigured. In this case the
* call is issued by the new_queue callback, with a
* non empty queue (q) and m pointing to the first
* mbuf in the queue. For this reason, the function
* should internally check for (m != q->mq.head)
* before calling dn_enqueue().
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
*
* dequeue Called when scheduler instance 's' can
* dequeue a packet. Return NULL if none are available.
* XXX what about non work-conserving ?
*
* config called on 'sched X config ...', normally writes
* in the area of size sch_arg
*
* destroy called on 'sched delete', frees everything
* in sch_arg (other parts are handled by more specific
* functions)
*
* new_sched called when a new instance is created, e.g.
* to create the local queue for !MULTIQUEUE, set V or
* copy parameters for WFQ, and so on.
*
* free_sched called when deleting an instance, cleans
* extra data in the per-instance area.
*
* new_fsk called when a flowset is linked to a scheduler,
* e.g. to validate parameters such as weights etc.
* free_fsk when a flowset is unlinked from a scheduler.
* (probably unnecessary)
*
* new_queue called to set the per-queue parameters,
* e.g. S and F, adjust sum of weights in the parent, etc.
2010-03-21 15:52:55 +00:00
*
* The new_queue callback is normally called from when
* creating a new queue. In some cases (such as a
* scheduler change or reconfiguration) it can be called
* with a non empty queue. In this case, the queue
* In case of non empty queue, the new_queue callback could
* need to call the enqueue function. In this case,
* the callback should eventually call enqueue() passing
* as m the first element in the queue.
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
*
* free_queue actions related to a queue removal, e.g. undo
* all the above. If the queue has data in it, also remove
* from the scheduler. This can e.g. happen during a reconfigure.
*/
int (*enqueue)(struct dn_sch_inst *, struct dn_queue *,
struct mbuf *);
struct mbuf * (*dequeue)(struct dn_sch_inst *);
int (*config)(struct dn_schk *);
int (*destroy)(struct dn_schk*);
int (*new_sched)(struct dn_sch_inst *);
int (*free_sched)(struct dn_sch_inst *);
int (*new_fsk)(struct dn_fsk *f);
int (*free_fsk)(struct dn_fsk *f);
int (*new_queue)(struct dn_queue *q);
int (*free_queue)(struct dn_queue *q);
/* run-time fields */
int ref_count; /* XXX number of instances in the system */
SLIST_ENTRY(dn_alg) next; /* Next scheduler in the list */
};
/* MSVC does not support initializers so we need this ugly macro */
#ifdef _WIN32
#define _SI(fld)
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
#else
#define _SI(fld) fld
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
#endif
/*
* Additionally, dummynet exports some functions and macros
* to be used by schedulers:
*/
void dn_free_pkts(struct mbuf *mnext);
int dn_enqueue(struct dn_queue *q, struct mbuf* m, int drop);
/* bound a variable between min and max */
int ipdn_bound_var(int *v, int dflt, int lo, int hi, const char *msg);
/*
* Extract the head of a queue, update stats. Must be the very last
* thing done on a dequeue as the queue itself may go away.
*/
static __inline struct mbuf*
dn_dequeue(struct dn_queue *q)
{
struct mbuf *m = q->mq.head;
if (m == NULL)
return NULL;
q->mq.head = m->m_nextpkt;
/* Update stats for the queue */
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
q->ni.length--;
q->ni.len_bytes -= m->m_pkthdr.len;
if (q->_si) {
q->_si->ni.length--;
q->_si->ni.len_bytes -= m->m_pkthdr.len;
}
if (q->ni.length == 0) /* queue is now idle */
q->q_time = dn_cfg.curr_time;
return m;
}
int dn_sched_modevent(module_t mod, int cmd, void *arg);
#define DECLARE_DNSCHED_MODULE(name, dnsched) \
static moduledata_t name##_mod = { \
#name, dn_sched_modevent, dnsched \
}; \
DECLARE_MODULE(name, name##_mod, \
SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY); \
MODULE_DEPEND(name, dummynet, 3, 3, 3);
#endif /* _DN_SCHED_H */