freebsd-skq/sys/netinet/ip_dummynet.h

286 lines
8.9 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) 1998-2010 Luigi Rizzo, Universita` di Pisa
* Portions Copyright (c) 2000 Akamba Corp.
* 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.
*
1999-08-28 01:08:13 +00:00
* $FreeBSD$
*/
#ifndef _IP_DUMMYNET_H
#define _IP_DUMMYNET_H
Import Dummynet AQM version 0.2.1 (CoDel, FQ-CoDel, PIE and FQ-PIE). Centre for Advanced Internet Architectures Implementing AQM in FreeBSD * Overview <http://caia.swin.edu.au/freebsd/aqm/index.html> * Articles, Papers and Presentations <http://caia.swin.edu.au/freebsd/aqm/papers.html> * Patches and Tools <http://caia.swin.edu.au/freebsd/aqm/downloads.html> Overview Recent years have seen a resurgence of interest in better managing the depth of bottleneck queues in routers, switches and other places that get congested. Solutions include transport protocol enhancements at the end-hosts (such as delay-based or hybrid congestion control schemes) and active queue management (AQM) schemes applied within bottleneck queues. The notion of AQM has been around since at least the late 1990s (e.g. RFC 2309). In recent years the proliferation of oversized buffers in all sorts of network devices (aka bufferbloat) has stimulated keen community interest in four new AQM schemes -- CoDel, FQ-CoDel, PIE and FQ-PIE. The IETF AQM working group is looking to document these schemes, and independent implementations are a corner-stone of the IETF's process for confirming the clarity of publicly available protocol descriptions. While significant development work on all three schemes has occured in the Linux kernel, there is very little in FreeBSD. Project Goals This project began in late 2015, and aims to design and implement functionally-correct versions of CoDel, FQ-CoDel, PIE and FQ_PIE in FreeBSD (with code BSD-licensed as much as practical). We have chosen to do this as extensions to FreeBSD's ipfw/dummynet firewall and traffic shaper. Implementation of these AQM schemes in FreeBSD will: * Demonstrate whether the publicly available documentation is sufficient to enable independent, functionally equivalent implementations * Provide a broader suite of AQM options for sections the networking community that rely on FreeBSD platforms Program Members: * Rasool Al Saadi (developer) * Grenville Armitage (project lead) Acknowledgements: This project has been made possible in part by a gift from the Comcast Innovation Fund. Submitted by: Rasool Al-Saadi <ralsaadi@swin.edu.au> X-No objection: core MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D6388
2016-05-26 21:40:13 +00:00
#define NEW_AQM
/*
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
* Definition of the kernel-userland API for dummynet.
*
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
* Setsockopt() and getsockopt() pass a batch of objects, each
* of them starting with a "struct dn_id" which should fully identify
* the object and its relation with others in the sequence.
* The first object in each request should have
* type= DN_CMD_*, id = DN_API_VERSION.
* For other objects, type and subtype specify the object, len indicates
* the total length including the header, and 'id' identifies the specific
* object.
*
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
* Most objects are numbered with an identifier in the range 1..65535.
* DN_MAX_ID indicates the first value outside the range.
2000-06-08 09:45:23 +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
#define DN_API_VERSION 12500000
#define DN_MAX_ID 0x10000
2000-06-08 09:45:23 +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
struct dn_id {
uint16_t len; /* total obj len including this header */
uint8_t type;
uint8_t subtype;
uint32_t id; /* generic id */
};
/*
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
* These values are in the type field of struct dn_id.
* To preserve the ABI, never rearrange the list or delete
* entries with the exception of DN_LAST
*/
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
enum {
DN_NONE = 0,
DN_LINK = 1,
DN_FS,
DN_SCH,
DN_SCH_I,
DN_QUEUE,
DN_DELAY_LINE,
DN_PROFILE,
DN_FLOW, /* struct dn_flow */
DN_TEXT, /* opaque text is the object */
DN_CMD_CONFIG = 0x80, /* objects follow */
DN_CMD_DELETE, /* subtype + list of entries */
DN_CMD_GET, /* subtype + list of entries */
DN_CMD_FLUSH,
/* for compatibility with FreeBSD 7.2/8 */
DN_COMPAT_PIPE,
DN_COMPAT_QUEUE,
DN_GET_COMPAT,
/* special commands for emulation of sysctl variables */
DN_SYSCTL_GET,
DN_SYSCTL_SET,
Import Dummynet AQM version 0.2.1 (CoDel, FQ-CoDel, PIE and FQ-PIE). Centre for Advanced Internet Architectures Implementing AQM in FreeBSD * Overview <http://caia.swin.edu.au/freebsd/aqm/index.html> * Articles, Papers and Presentations <http://caia.swin.edu.au/freebsd/aqm/papers.html> * Patches and Tools <http://caia.swin.edu.au/freebsd/aqm/downloads.html> Overview Recent years have seen a resurgence of interest in better managing the depth of bottleneck queues in routers, switches and other places that get congested. Solutions include transport protocol enhancements at the end-hosts (such as delay-based or hybrid congestion control schemes) and active queue management (AQM) schemes applied within bottleneck queues. The notion of AQM has been around since at least the late 1990s (e.g. RFC 2309). In recent years the proliferation of oversized buffers in all sorts of network devices (aka bufferbloat) has stimulated keen community interest in four new AQM schemes -- CoDel, FQ-CoDel, PIE and FQ-PIE. The IETF AQM working group is looking to document these schemes, and independent implementations are a corner-stone of the IETF's process for confirming the clarity of publicly available protocol descriptions. While significant development work on all three schemes has occured in the Linux kernel, there is very little in FreeBSD. Project Goals This project began in late 2015, and aims to design and implement functionally-correct versions of CoDel, FQ-CoDel, PIE and FQ_PIE in FreeBSD (with code BSD-licensed as much as practical). We have chosen to do this as extensions to FreeBSD's ipfw/dummynet firewall and traffic shaper. Implementation of these AQM schemes in FreeBSD will: * Demonstrate whether the publicly available documentation is sufficient to enable independent, functionally equivalent implementations * Provide a broader suite of AQM options for sections the networking community that rely on FreeBSD platforms Program Members: * Rasool Al Saadi (developer) * Grenville Armitage (project lead) Acknowledgements: This project has been made possible in part by a gift from the Comcast Innovation Fund. Submitted by: Rasool Al-Saadi <ralsaadi@swin.edu.au> X-No objection: core MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D6388
2016-05-26 21:40:13 +00:00
#ifdef NEW_AQM
/* subtypes used for setting/getting extra parameters.
* these subtypes used with IP_DUMMYNET3 command (get)
* and DN_TEXT (set). */
DN_AQM_PARAMS, /* AQM extra params */
DN_SCH_PARAMS, /* scheduler extra params */
#endif
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
DN_LAST,
};
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
enum { /* subtype for schedulers, flowset and the like */
DN_SCHED_UNKNOWN = 0,
DN_SCHED_FIFO = 1,
DN_SCHED_WF2QP = 2,
/* others are in individual modules */
};
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
enum { /* user flags */
DN_HAVE_MASK = 0x0001, /* fs or sched has a mask */
DN_NOERROR = 0x0002, /* do not report errors */
DN_QHT_HASH = 0x0004, /* qht is a hash table */
DN_QSIZE_BYTES = 0x0008, /* queue size is in bytes */
DN_HAS_PROFILE = 0x0010, /* a link has a profile */
DN_IS_RED = 0x0020,
DN_IS_GENTLE_RED= 0x0040,
DN_IS_ECN = 0x0080,
Import Dummynet AQM version 0.2.1 (CoDel, FQ-CoDel, PIE and FQ-PIE). Centre for Advanced Internet Architectures Implementing AQM in FreeBSD * Overview <http://caia.swin.edu.au/freebsd/aqm/index.html> * Articles, Papers and Presentations <http://caia.swin.edu.au/freebsd/aqm/papers.html> * Patches and Tools <http://caia.swin.edu.au/freebsd/aqm/downloads.html> Overview Recent years have seen a resurgence of interest in better managing the depth of bottleneck queues in routers, switches and other places that get congested. Solutions include transport protocol enhancements at the end-hosts (such as delay-based or hybrid congestion control schemes) and active queue management (AQM) schemes applied within bottleneck queues. The notion of AQM has been around since at least the late 1990s (e.g. RFC 2309). In recent years the proliferation of oversized buffers in all sorts of network devices (aka bufferbloat) has stimulated keen community interest in four new AQM schemes -- CoDel, FQ-CoDel, PIE and FQ-PIE. The IETF AQM working group is looking to document these schemes, and independent implementations are a corner-stone of the IETF's process for confirming the clarity of publicly available protocol descriptions. While significant development work on all three schemes has occured in the Linux kernel, there is very little in FreeBSD. Project Goals This project began in late 2015, and aims to design and implement functionally-correct versions of CoDel, FQ-CoDel, PIE and FQ_PIE in FreeBSD (with code BSD-licensed as much as practical). We have chosen to do this as extensions to FreeBSD's ipfw/dummynet firewall and traffic shaper. Implementation of these AQM schemes in FreeBSD will: * Demonstrate whether the publicly available documentation is sufficient to enable independent, functionally equivalent implementations * Provide a broader suite of AQM options for sections the networking community that rely on FreeBSD platforms Program Members: * Rasool Al Saadi (developer) * Grenville Armitage (project lead) Acknowledgements: This project has been made possible in part by a gift from the Comcast Innovation Fund. Submitted by: Rasool Al-Saadi <ralsaadi@swin.edu.au> X-No objection: core MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D6388
2016-05-26 21:40:13 +00:00
#ifdef NEW_AQM
DN_IS_AQM = 0x0100, /* AQMs: e.g Codel & PIE */
#endif
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
DN_PIPE_CMD = 0x1000, /* pipe config... */
};
2000-06-08 09:45:23 +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
* link template.
2000-06-08 09:45:23 +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
struct dn_link {
struct dn_id oid;
/*
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
* Userland sets bw and delay in bits/s and milliseconds.
* The kernel converts this back and forth to bits/tick and ticks.
* XXX what about burst ?
*/
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
int32_t link_nr;
int bandwidth; /* bit/s or bits/tick. */
int delay; /* ms and ticks */
uint64_t burst; /* scaled. bits*Hz XXX */
};
2000-06-08 09:45:23 +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
* A flowset, which is a template for flows. Contains parameters
* from the command line: id, target scheduler, queue sizes, plr,
* flow masks, buckets for the flow hash, and possibly scheduler-
* specific parameters (weight, quantum and so on).
*/
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
struct dn_fs {
struct dn_id oid;
uint32_t fs_nr; /* the flowset number */
uint32_t flags; /* userland flags */
int qsize; /* queue size in slots or bytes */
int32_t plr; /* PLR, pkt loss rate (2^31-1 means 100%) */
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
uint32_t buckets; /* buckets used for the queue hash table */
2000-06-08 09:45:23 +00:00
struct ipfw_flow_id flow_mask;
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
uint32_t sched_nr; /* the scheduler we attach to */
/* generic scheduler parameters. Leave them at -1 if unset.
* Now we use 0: weight, 1: lmax, 2: priority
*/
int par[4];
/* RED/GRED parameters.
* weight and probabilities are in the range 0..1 represented
* in fixed point arithmetic with SCALE_RED decimal bits.
*/
#define SCALE_RED 16
#define SCALE(x) ( (x) << SCALE_RED )
#define SCALE_VAL(x) ( (x) >> SCALE_RED )
#define SCALE_MUL(x,y) ( ( (x) * (y) ) >> SCALE_RED )
int w_q ; /* queue weight (scaled) */
int max_th ; /* maximum threshold for queue (scaled) */
int min_th ; /* minimum threshold for queue (scaled) */
int max_p ; /* maximum value for p_b (scaled) */
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
};
/*
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
* dn_flow collects flow_id and stats for queues and scheduler
* instances, and is used to pass these info to userland.
* oid.type/oid.subtype describe the object, oid.id is number
* of the parent object.
*/
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
struct dn_flow {
struct dn_id oid;
struct ipfw_flow_id fid;
uint64_t tot_pkts; /* statistics counters */
uint64_t tot_bytes;
2012-08-07 07:52:25 +00:00
uint32_t length; /* Queue length, in packets */
uint32_t len_bytes; /* Queue length, in bytes */
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
uint32_t drops;
};
2000-06-08 09:45:23 +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
* Scheduler template, mostly indicating the name, number,
* sched_mask and buckets.
*/
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
struct dn_sch {
struct dn_id oid;
uint32_t sched_nr; /* N, scheduler number */
uint32_t buckets; /* number of buckets for the instances */
uint32_t flags; /* have_mask, ... */
char name[16]; /* null terminated */
/* mask to select the appropriate scheduler instance */
struct ipfw_flow_id sched_mask; /* M */
};
2000-06-08 09:45:23 +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
/* A delay profile is attached to a link.
* Note that a profile, as any other object, cannot be longer than 2^16
*/
#define ED_MAX_SAMPLES_NO 1024
struct dn_profile {
struct dn_id oid;
/* fields to simulate a delay profile */
#define ED_MAX_NAME_LEN 32
char name[ED_MAX_NAME_LEN];
int link_nr;
int loss_level;
int bandwidth; // XXX use link bandwidth?
int samples_no; /* actual len of samples[] */
int samples[ED_MAX_SAMPLES_NO]; /* may be shorter */
};
Import Dummynet AQM version 0.2.1 (CoDel, FQ-CoDel, PIE and FQ-PIE). Centre for Advanced Internet Architectures Implementing AQM in FreeBSD * Overview <http://caia.swin.edu.au/freebsd/aqm/index.html> * Articles, Papers and Presentations <http://caia.swin.edu.au/freebsd/aqm/papers.html> * Patches and Tools <http://caia.swin.edu.au/freebsd/aqm/downloads.html> Overview Recent years have seen a resurgence of interest in better managing the depth of bottleneck queues in routers, switches and other places that get congested. Solutions include transport protocol enhancements at the end-hosts (such as delay-based or hybrid congestion control schemes) and active queue management (AQM) schemes applied within bottleneck queues. The notion of AQM has been around since at least the late 1990s (e.g. RFC 2309). In recent years the proliferation of oversized buffers in all sorts of network devices (aka bufferbloat) has stimulated keen community interest in four new AQM schemes -- CoDel, FQ-CoDel, PIE and FQ-PIE. The IETF AQM working group is looking to document these schemes, and independent implementations are a corner-stone of the IETF's process for confirming the clarity of publicly available protocol descriptions. While significant development work on all three schemes has occured in the Linux kernel, there is very little in FreeBSD. Project Goals This project began in late 2015, and aims to design and implement functionally-correct versions of CoDel, FQ-CoDel, PIE and FQ_PIE in FreeBSD (with code BSD-licensed as much as practical). We have chosen to do this as extensions to FreeBSD's ipfw/dummynet firewall and traffic shaper. Implementation of these AQM schemes in FreeBSD will: * Demonstrate whether the publicly available documentation is sufficient to enable independent, functionally equivalent implementations * Provide a broader suite of AQM options for sections the networking community that rely on FreeBSD platforms Program Members: * Rasool Al Saadi (developer) * Grenville Armitage (project lead) Acknowledgements: This project has been made possible in part by a gift from the Comcast Innovation Fund. Submitted by: Rasool Al-Saadi <ralsaadi@swin.edu.au> X-No objection: core MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D6388
2016-05-26 21:40:13 +00:00
#ifdef NEW_AQM
/* Extra parameters for AQM and scheduler.
* This struct is used to pass and retrieve parameters (configurations)
* to/from AQM and Scheduler.
*/
struct dn_extra_parms {
struct dn_id oid;
char name[16];
uint32_t nr;
#define DN_MAX_EXTRA_PARM 10
int64_t par[DN_MAX_EXTRA_PARM];
};
#endif
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
/*
* Overall structure of dummynet
In dummynet, packets are selected with the firewall rules, and passed
to two different objects: PIPE or QUEUE (bad name).
A QUEUE defines a classifier, which groups packets into flows
according to a 'mask', puts them into independent queues (one
per flow) with configurable size and queue management policy,
and passes flows to a scheduler:
(flow_mask|sched_mask) sched_mask
+---------+ weight Wx +-------------+
| |->-[flow]-->--| |-+
-->--| QUEUE x | ... | | |
| |->-[flow]-->--| SCHEDuler N | |
+---------+ | | |
... | +--[LINK N]-->--
+---------+ weight Wy | | +--[LINK N]-->--
| |->-[flow]-->--| | |
-->--| QUEUE y | ... | | |
| |->-[flow]-->--| | |
+---------+ +-------------+ |
+-------------+
Many QUEUE objects can connect to the same scheduler, each
QUEUE object can have its own set of parameters.
In turn, the SCHEDuler 'forks' multiple instances according
to a 'sched_mask', each instance manages its own set of queues
and transmits on a private instance of a configurable LINK.
A PIPE is a simplified version of the above, where there
is no flow_mask, and each scheduler instance handles a single queue.
The following data structures (visible from userland) describe
the objects used by dummynet:
+ dn_link, contains the main configuration parameters related
to delay and bandwidth;
+ dn_profile describes a delay profile;
+ dn_flow describes the flow status (flow id, statistics)
+ dn_sch describes a scheduler
+ dn_fs describes a flowset (msk, weight, queue parameters)
*
*/
#endif /* _IP_DUMMYNET_H */