Document Actions
Revision History
MINT - Mesh Interconnection Network Technology.
Scalable Distributed Wireless VLAN Switch.
Hybrid Router/Switch Architecture.
Complete Virtual Ethernet Implementation for
homo/heterogeneous polymorphous mesh networks.
Infinet Wireless Ltd.
www.infinetwireless.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ATTENTION!
Starting from 1.22.0 firmware version, "switch" is partially incompatible with
other firmware versions.
It is highly recommended to perform firmware upgrade for units working in switch mode.
Compatibility for MINT protocol and routing is not disturbed.
"Over The Air Firmware Upgrade" feature also can be used.
Revision history:
06.11.2007 V1.26.0
~~~~~~~~~~~~~~~~~~~~
1. When upgrading unit with existing RMA configuration to MINT, "mint" will be automatically
started in "MASTER" mode (except units with LCCPE flag, which will work only in "SLAVE" mode)
2. New option [-]dhcp is added to MACF command
When "dhcp" option is turned on, macf will automatically learn addresses leased by a local
DHCP server (if started at the same unit).
These records are temporary and will not be saved into the config.
They remain active as long as corresponding DHCP leases are active.
3. New MINT option: -[no]replicate [$ACL]
This option helps to "isolate" users located behind remote nodes from each other at link layer
with minimal configuration overhead.
If MINT node has "mint -noreplicate" option configured, any client traffic coming from remote
node's eth port via wireless link cannot be forwarded back into wireless interface.
It is still possible, however, for an external wired switch connected to eth0 interface
to forward such traffic back to a MINT node.
By default, such replication is turned off (mint -replicate)
Optional ACL of type "num" (acl add $ISOLATE num N1 N2 ...) can be used, listing switch group
numbers which need to be filtered (by default command applies to all groups).
Please note that this option applies only to a trafic which originated from eth0 port of the
remote node and doesn't affect wireless connectivity between MINT nodes themselves
4. "autobitrate" didn't work correctly in "Long" mode - fixed
5. Roaming profiles with different channel sizes (full/half/quarter) may work incorrectly
if "-band ...." parameter was omitted at one of the profiles - fixed
6. New switch option : "switch group Y in-trunk X"
If trunk-group X is configured at the node (usually a BS) which should distribute several
VLAN trunks to the different remote nodes, such end nodes should use "in-trunk" option
to explicitly specify which trunk group number it corresponds to.
For example, if end node has auto-created group 100 which is a member of trunk-group 5
(VLAN100 was converted at the BS into a switch group 100), then switch config should look like:
switch group 100 in-trunk 5
It allows to create several non-overlapping trunk groups with (potentially) identical
VLAN numbers inside the same MINT network segment.
7. Syntax for switch group configuration extended. New option "xvlan XX"
( where XX - VLAN number or a name of vlan ACL).
In comparison to common "vlan XX" option, extended version allows to process also untagged frames.
For example,
switch list MYNET numrange add 100 200 300
switch group 10 xvlan MYNET
switch group 10 trunk on
Group 10 will process frames tagged with VLAN ID 100,200,300 and also untagged frames.
Untagged frames in this case will be marked with corresponding group number (10) when forwarded
into MINT network, tagged frames - with group number equal to VLAN tag (100, 200 or 300)
switch list MYNET numrange add 100 200 300
switch group 20 vlan MYNET
switch group 20 trunk off
Group 20 will process (and tag with group # 20) only frames tagged with VLANs listed in MYNET
8. Switch didn't untag frames sent to the IP addresses of the MINT node -fixed
9. Maximum number of vlanX router interfaces increased to 100
10. "arp proxyall [$ACL]" command now has an option to limit the scope of the command
to only networks listed in the ACL
11. Point-to-point mode performance improved
16.10.2007 V1.25.12
~~~~~~~~~~~~~~~~~~~~
1. When filters for IPSTAT ("ipstat add") are configured, egress packets were not counted in the
statistics - fixed
2. "mint addnode" without explicitly configured node key blocks neighbour assosiation
even if the global key was specified in MINT config - fixed
October 09, 2007 V1.25.11
~~~~~~~~~~~~~~~~~~~~
1. macf and rpfilter filters wrongly blocked DHCP queries. Corrected
2. mint map output is changed and advanced, including color indication of problem links.
3. autobitrate parameter didn`t work in LongRange mode, corrected.
4. On the Gateworks AVILA (IXP42X) platform didn`t work Over The Air Update function, corrected.
5. Models that support CES could reset with unstable radio channel, corrected.
October 02, 2007 V1.25.9
~~~~~~~~~~~~~~~~~~~
1. Packets that were outgoing through vlanX interfaces wasn`t get under ipfw/qm/ipstat rules. Corrected.
September 29, 2007 V1.25.8
~~~~~~~~~~~~~~~~~~~
1. Error that blocked local broadcast traffic passing in some types of network topologies corrected.
September 28, 2007 V1.25.7
~~~~~~~~~~~~~~~~~~~
1. Now switch diagnostic messages more precisely reflects emerging events meaning.
2. Small errors correction
September 21, 2007 V1.25.6
~~~~~~~~~~~~~~~~~~~
1. Inaccuracies in switch operation corrected.
The only way to tag local (intranet) traffic for sending to wired segment - is using local-tag
2. rf ptp parameter is not automatically set on master/slave links.
When it was automatically set it lowers transmitting speed and leads to wrong CES devices operation
at simultaneous E1 and data flows transmission.
For models that support CES:
1. Commands added
ces eth media
ces eth bwlimit
ces eth stat
September 19, 2007 V1.25.5
~~~~~~~~~~~~~~~~~~~
1. Inaccuracies in switch operation while transmitting local (intranet) traffic corrected.
Admin group packets were wrongly tagged (VLAN ID) by one of client switch group rules.
Now local device`s traffic (and traffic inside MINT network) always comes in wired segment not tagged.
If in this case you need tagging then you have to set needed
VLAN ID by "sw local-tag XX" command.
September 17, 2007 V1.25.4
~~~~~~~~~~~~~~~~~~~
1. "lt - Link Test" command added for performing radio channel quality tests.
Testing protocol is active so you need to upgrade firmware on all devices.
2. Small corrections in different subsystems.
30.08.2007 V1.25.1
~~~~~~~~~~~~~~~~~~~
1. Bug in "rpfilter" fixed
25.08.2007 V1.25.0
~~~~~~~~~~~~~~~~~~~
1. ipfw: Bugfix in APR packets handling behaviour.
Fileter expression syntax now supports "arp" keyword as "proto" field (equal to "ether=806")
2. ipfw/qm didn't work with tunnel interfaces tunX - fixed
3. "qm stat" output now provides more extensive statistics
4. "autobitrate" now works more precisely on a heavy loaded link
5. MINT network is now more stable in a complex topology with large number of network loops
6. Maximum number of open telnet sessions to the device is increased to 6
7. Basic support of SNMP Trap Groups added
8. Network-critical priority 0 (zero) added to QoS map.
Now it looks as follows:
QM_PRIO_NETCRIT 0
QM_PRIO_VOICE 1
QM_PRIO_RT1 2
QM_PRIO_VIDEO 3
QM_PRIO_RT2 4
QM_PRIO_QOS1 5
QM_PRIO_QOS2 6
QM_PRIO_QOS3 7
QM_PRIO_QOS4 8
QM_PRIO_BUSINESS1 9
QM_PRIO_BUSINESS2 10
QM_PRIO_BUSINESS3 11
QM_PRIO_BUSINESS4 12
QM_PRIO_BUSINESS5 13
QM_PRIO_BUSINESS6 14
QM_PRIO_BUSINESS7 15
QM_PRIO_BUSINESS8 16
9. Other minor fixes and optimizations
27.07.2007 V1.24.4
~~~~~~~~~~~~~~~~~~~
1. Important:
"rpfilter" clause of "ipfw" blocks ARP packets - fixed
2. Under heavy traffic flood (several times exceeding link capacity),
traffic from strict low-priority queues was served altogether with high-priority traffic - fixed
July 24, 2007 V1.24.2
~~~~~~~~~~~~~~~~~~~~
1. Old implementation of RIP protocol is included for backward compatibility.
2. Devices configured as "-type slave" may drop association frequently - fixed.
3. "mint -mode " paramater incorrectly displayed when using SNMP access.
4. MINT monitor sometimes does not display current signal levels.
5. Neighbour search/select process in roaming is improved.
6. Major extensions to "qm"/"ipfw" logic:
Option "pass" is added, allowing to process the actions specified in the rule, but to continue lookup for further matching rules.
For example:
ipfw add 1 pass log tcp ....
ipfw add 2 reject tcp ....
If packet is matched rule 1, it will be logged, but flitering/matching will continue from rule 2.
Additional frame/packet classification options include:
[vlan=N] [dot1p=N] [swg=N] [ether=X] [dscp=N|tos=N],
allowing to match based on 802.1q VLAN ID, 802.1p priority,
MINT switching group number (SWitchGroup), Ethernet frame type (EtherType),
as well as ip_tos field against particular DSCP or IP_Precedence value.
The same options are applicable to QoS manager (QM) configuration.
For example:
qm add 1 ch1 pass vlan=5 dscp=51 tcp ....
qm add 2 ch2 tcp ....
If rule 1 is matched, classification will continue, but all the options specified (vlan=5; dscp=51) will be immediately applied to a frame/packet.
New properties are also added to QM channel configuration:
vlan=N , dot1p=N , dscp=N
These parameters allow DSCP labels and/or 802.1p/802.1q manipulations, tagging, re-tagging, clearing or even removing tags.
Numerical value of "-2" removes the specified option from the channel configuration.
Value of "-1" "resets" the field:
- sets DSCP field (allowed range 0-63) to zero.
- sets 802.1p priority field (allowed range 0-7) to zero, AND,
if 802.1q VLAN ID field doesn't exist (or doesn't specified in the channel properties), removes the whole 802.1q/802.1p header.
- removes VLAN ID (allowed range 0-4095), completely with the 802.1p/802.1q header (even if 802.1p filed exists).
Zero value sets the specified field to "0".
Now, with "pass" option, QM allows to create complex and flexible traffic classification, prioritization and tagging rulesets.
For example:
# Channel 1 resets both DSCP tags and 802.1p priorities
qm ch1 dscp=0 dot1p=-1
# Channel 2 sets local flow priority to QM_PRIO_BUSINESS1 and packet field DSCP to 31
qm ch2 pri=9 dscp=31
# Channel 3 sets local flow priority to QM_PRIO_VIDEO and packet field DSCP to 11
qm ch3 pri=3 dscp=11
# Channel 2 sets local flow priority to QM_PRIO_BUSINESS8 and packet field DSCP to 51
qm ch4 pri=16 dscp=51
# Ruleset. Passing all the traffic from untrusted source via channel 1, to remove/reset all the priorities
qm add ch1 pass all from 0/0 to 0/0
# Some flows (TCP from the given subnet) goes into channel 2
qm add ch2 tcp from X.X.X.0/24 to 0/0
# Video (UDP flows from given senders) goes to channel 3
qm add ch3 udp from X.X.X.0/24 PORT to 0/0
# Remaining traffic will be treated "best-effort" by default.
# This rule assigns it to channel 4, to set up some "default priority" as specified
qm add ch4 all from 0/0 to 0/0
Note: Actual prioritization inside MINT network is based on "pri=N" settings, not on ToS/DSCP/802.1p!
DSCP tags are passed transparently through the MINT cloud, in bridge or router mode.
802.1p tags can be passed only in bridge mode.
It is possible to set/reset dot1p/vlan and/or dscp values when frame/packet leaves MINT network, if required.
For example:
# Channel 25 sets 802.1p priority of 5.
# Note: if VLAN header didn't exist before, it will be added automatically with 802.1q VLAN set to zero.
qm ch25 dot1p=5
# Channel 26 sets 802.1p priority of 7 and 802.1q VLAN - to 4
# Note: if VLAN header didn't exist before, it will be added automatically.
qm ch26 vlan=7 dot1p=4
# Place all the IP packets leaving MINT network via eth0 interface into channel 25, if they were labeled DSCP 11
qm addout eth0 ch25 dscp11 from 0/0 to 0/0
# Place all the IP packets leaving MINT network via eth0 interface into channel 26, if they were labeled DSCP 13
qm addout eth0 ch26 dscp13 from 0/0 to 0/0
and so on...
June 30, 2007 V1.23.1
~~~~~~~~~~~~~~~~~~~~
1. "Poll" mode bug fixed.
2. Device with DFS feature but with "-roaming disable" configured was unable to connect - fixed
June 21, 2007 V1.23.0
~~~~~~~~~~~~~~~~~~~~
1. MINT Protocol compatibility with earlier versions of MINT protocol restored
(1.18 and below)
2. Error fixed in marker access decreasing maximum attainable speed.
June 15, 2007 V1.22.18
~~~~~~~~~~~~~~~~~~~~
1. NAT Module worked improperly with the automatic appointment of public addresses obtained through DHCP.
2. NAT module created many false entries in the links table while fragmented packets processing, that mistake led to wrong exceed of the maxlinks Nat limit.
3. In the absence of GPS coordinates in configuration, "rf distance" parameter in the "auto" mode is taken to the nearest hub (to the base stations, for example).
09.06.2007 V1.22.17
~~~~~~~~~~~~~~~~~~~~
1. When installing firmware MINT over RMA, mint automatically runs on all radio interfaces of a device with default settings.
2. "mint addnode" command added with "-note STRING" parameter for comments.
3. A number of inaccuracies is eliminated in roaming and airupdate mechanisms.
May 29, 2007 V1.22.12
~~~~~~~~~~~~~~~~~~~~
1. Module rip is deleted from CPE devices software.
2. Connection between two radios of a single device is not possible.
24.05.2007 V1.22.11
~~~~~~~~~~~~~~~~~~~~
1. SLAVE node having connections via pseudo-rf interface (prf) and working in roaming enable mode didnІt connect with roaming leader through the radio interface, fixed.
2. Speed response to power down of a neighbor node is raised.
3. SLAVE node after connecting to the network enters eco mode, reducing service packages network load.
4. New SNMP Traps added.
17.05.2007 V1.22.8
~~~~~~~~~~~~~~~~~~~
1. mint IFNAME trace MAC command added
2. Fixed error with the previous version of MINT with wrong working after the protocol restart (mint restart)
11.05.2007 V1.22.3
~~~~~~~~~~~~~~~~~~~
1. SNMP (snmpd) module bug fixed (caused CPE reboot).
2. Imported configuration protection from incorrect data and open quotes improved.
3. Public beta-version
07.05.2007 V1.22.2
~~~~~~~~~~~~~~~~~~~
1. AudioMonitor support added
mint IFNAME monitor -audio [full] [-mac MAC]
This command starts working immediately and can be saved in
the configuration which allows using it even after unit's reboot.
To turn AudioMonitor off:
mint IFNAME monitor -noaudio
In "full" mode, audio monitoring is performed according to a maximal
receiving level among neighbors and candidates.
If "full" parameter is not specified, a an arithmetical mean of receive and transmit
signal is calculated (only for neighbors). The neighbor for which this value
is maximal will be indicated in audio monitor.
06.05.2007 V1.22.1
~~~~~~~~~~~~~~~~~~~
1. Broadcast traffic optimization for point-to-point links.
2. Switch nodes search mechanism error fixed.
3. Minor errors in "ipfw/qm" rules fixed.
4. Unconditional 802.1p support added
5. "sys [no]iptos" option removed.
Imstead, the following command added:
qm option [-]voice [-]dot1p [-]tos [-]strict
to manage automatical prioritization in specified modes.
"strict" option means that "Strict Priority" policy is applied
to all queues, otherwise (by default) "Weighted Fair Queuing" policy
is used.
02.05.2007 V1.22.0
~~~~~~~~~~~~~~~~~~~
1. New algorithm for broadcast/multicast traffic transmitting over MINT network.
Now broadcast traffic is only spread across those nodes which belong to the
corresponding switching zone and nodes which form optimal paths to deliver
data for this zone (incompatible with versions below 1.22.0)
2. Transparent packets prioritization support in MINT network.
Supported by using channels management in "qm" command.
Administrator can put streams into different channels based on
"qm/ipfw" rules as well as "tos" and "dscp" fields.
qm ch1 pri=12
qm add ch1 all from x/x to y/y
qm add ch1 dscp31 all from a to b
qm add ch1 dscp42
Each channel can be assigned a priority (1..16)
Once assigned, a priority will be automatically recognized by every node
inside MINT network.
Priority scheme looks as follows:
QM_PRIO_VOICE 1
QM_PRIO_RT1 2
QM_PRIO_VIDEO 3
QM_PRIO_RT2 4
QM_PRIO_QOS1 5
QM_PRIO_QOS2 6
QM_PRIO_QOS3 7
QM_PRIO_QOS4 8
QM_PRIO_BUSINESS1 9
QM_PRIO_BUSINESS2 10
QM_PRIO_BUSINESS3 11
QM_PRIO_BUSINESS4 12
QM_PRIO_BUSINESS5 13
QM_PRIO_BUSINESS6 14
QM_PRIO_BUSINESS7 15
QM_PRIO_BUSINESS8 16
Priorities "1" and "2" are additionally processed as "voice".
Packets from which the priority is not clearly defined will be sent via
common queue with "Best Effort".
Priorities might be automatically shifted by one or two levels up in
"Emergency Level" mode.
3. New firmware features an updated version of autobitrate mechanism which provides
with increased link reliability and stability (esp for mobile clients).
4. Additional roaming profiles parameters are added:
[-type {master|mesh|slave}]
[-key XXX] [-nodeid N] [-netid N]
5. "Over-the-air Update" mechanism upgrade. Automatic upgrade now is performed
with no interruption of regular data flows.
6. "dhcpd" module error fixed which caused CPE rebooting.
7. "arip" module error fixed which occured when having several aliases on one
interface.
8. Bug fixed - some telnet clients incorrectly imported/exported unit's configuration
Scalable Distributed Wireless VLAN Switch.
Hybrid Router/Switch Architecture.
Complete Virtual Ethernet Implementation for
homo/heterogeneous polymorphous mesh networks.
Infinet Wireless Ltd.
www.infinetwireless.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ATTENTION!
Starting from 1.22.0 firmware version, "switch" is partially incompatible with
other firmware versions.
It is highly recommended to perform firmware upgrade for units working in switch mode.
Compatibility for MINT protocol and routing is not disturbed.
"Over The Air Firmware Upgrade" feature also can be used.
Revision history:
06.11.2007 V1.26.0
~~~~~~~~~~~~~~~~~~~~
1. When upgrading unit with existing RMA configuration to MINT, "mint" will be automatically
started in "MASTER" mode (except units with LCCPE flag, which will work only in "SLAVE" mode)
2. New option [-]dhcp is added to MACF command
When "dhcp" option is turned on, macf will automatically learn addresses leased by a local
DHCP server (if started at the same unit).
These records are temporary and will not be saved into the config.
They remain active as long as corresponding DHCP leases are active.
3. New MINT option: -[no]replicate [$ACL]
This option helps to "isolate" users located behind remote nodes from each other at link layer
with minimal configuration overhead.
If MINT node has "mint -noreplicate" option configured, any client traffic coming from remote
node's eth port via wireless link cannot be forwarded back into wireless interface.
It is still possible, however, for an external wired switch connected to eth0 interface
to forward such traffic back to a MINT node.
By default, such replication is turned off (mint -replicate)
Optional ACL of type "num" (acl add $ISOLATE num N1 N2 ...) can be used, listing switch group
numbers which need to be filtered (by default command applies to all groups).
Please note that this option applies only to a trafic which originated from eth0 port of the
remote node and doesn't affect wireless connectivity between MINT nodes themselves
4. "autobitrate" didn't work correctly in "Long" mode - fixed
5. Roaming profiles with different channel sizes (full/half/quarter) may work incorrectly
if "-band ...." parameter was omitted at one of the profiles - fixed
6. New switch option : "switch group Y in-trunk X"
If trunk-group X is configured at the node (usually a BS) which should distribute several
VLAN trunks to the different remote nodes, such end nodes should use "in-trunk" option
to explicitly specify which trunk group number it corresponds to.
For example, if end node has auto-created group 100 which is a member of trunk-group 5
(VLAN100 was converted at the BS into a switch group 100), then switch config should look like:
switch group 100 in-trunk 5
It allows to create several non-overlapping trunk groups with (potentially) identical
VLAN numbers inside the same MINT network segment.
7. Syntax for switch group configuration extended. New option "xvlan XX"
( where XX - VLAN number or a name of vlan ACL).
In comparison to common "vlan XX" option, extended version allows to process also untagged frames.
For example,
switch list MYNET numrange add 100 200 300
switch group 10 xvlan MYNET
switch group 10 trunk on
Group 10 will process frames tagged with VLAN ID 100,200,300 and also untagged frames.
Untagged frames in this case will be marked with corresponding group number (10) when forwarded
into MINT network, tagged frames - with group number equal to VLAN tag (100, 200 or 300)
switch list MYNET numrange add 100 200 300
switch group 20 vlan MYNET
switch group 20 trunk off
Group 20 will process (and tag with group # 20) only frames tagged with VLANs listed in MYNET
8. Switch didn't untag frames sent to the IP addresses of the MINT node -fixed
9. Maximum number of vlanX router interfaces increased to 100
10. "arp proxyall [$ACL]" command now has an option to limit the scope of the command
to only networks listed in the ACL
11. Point-to-point mode performance improved
16.10.2007 V1.25.12
~~~~~~~~~~~~~~~~~~~~
1. When filters for IPSTAT ("ipstat add") are configured, egress packets were not counted in the
statistics - fixed
2. "mint addnode" without explicitly configured node key blocks neighbour assosiation
even if the global key was specified in MINT config - fixed
October 09, 2007 V1.25.11
~~~~~~~~~~~~~~~~~~~~
1. macf and rpfilter filters wrongly blocked DHCP queries. Corrected
2. mint map output is changed and advanced, including color indication of problem links.
3. autobitrate parameter didn`t work in LongRange mode, corrected.
4. On the Gateworks AVILA (IXP42X) platform didn`t work Over The Air Update function, corrected.
5. Models that support CES could reset with unstable radio channel, corrected.
October 02, 2007 V1.25.9
~~~~~~~~~~~~~~~~~~~
1. Packets that were outgoing through vlanX interfaces wasn`t get under ipfw/qm/ipstat rules. Corrected.
September 29, 2007 V1.25.8
~~~~~~~~~~~~~~~~~~~
1. Error that blocked local broadcast traffic passing in some types of network topologies corrected.
September 28, 2007 V1.25.7
~~~~~~~~~~~~~~~~~~~
1. Now switch diagnostic messages more precisely reflects emerging events meaning.
2. Small errors correction
September 21, 2007 V1.25.6
~~~~~~~~~~~~~~~~~~~
1. Inaccuracies in switch operation corrected.
The only way to tag local (intranet) traffic for sending to wired segment - is using local-tag
2. rf ptp parameter is not automatically set on master/slave links.
When it was automatically set it lowers transmitting speed and leads to wrong CES devices operation
at simultaneous E1 and data flows transmission.
For models that support CES:
1. Commands added
ces eth media
ces eth bwlimit
ces eth stat
September 19, 2007 V1.25.5
~~~~~~~~~~~~~~~~~~~
1. Inaccuracies in switch operation while transmitting local (intranet) traffic corrected.
Admin group packets were wrongly tagged (VLAN ID) by one of client switch group rules.
Now local device`s traffic (and traffic inside MINT network) always comes in wired segment not tagged.
If in this case you need tagging then you have to set needed
VLAN ID by "sw local-tag XX" command.
September 17, 2007 V1.25.4
~~~~~~~~~~~~~~~~~~~
1. "lt - Link Test" command added for performing radio channel quality tests.
Testing protocol is active so you need to upgrade firmware on all devices.
2. Small corrections in different subsystems.
30.08.2007 V1.25.1
~~~~~~~~~~~~~~~~~~~
1. Bug in "rpfilter" fixed
25.08.2007 V1.25.0
~~~~~~~~~~~~~~~~~~~
1. ipfw: Bugfix in APR packets handling behaviour.
Fileter expression syntax now supports "arp" keyword as "proto" field (equal to "ether=806")
2. ipfw/qm didn't work with tunnel interfaces tunX - fixed
3. "qm stat" output now provides more extensive statistics
4. "autobitrate" now works more precisely on a heavy loaded link
5. MINT network is now more stable in a complex topology with large number of network loops
6. Maximum number of open telnet sessions to the device is increased to 6
7. Basic support of SNMP Trap Groups added
8. Network-critical priority 0 (zero) added to QoS map.
Now it looks as follows:
QM_PRIO_NETCRIT 0
QM_PRIO_VOICE 1
QM_PRIO_RT1 2
QM_PRIO_VIDEO 3
QM_PRIO_RT2 4
QM_PRIO_QOS1 5
QM_PRIO_QOS2 6
QM_PRIO_QOS3 7
QM_PRIO_QOS4 8
QM_PRIO_BUSINESS1 9
QM_PRIO_BUSINESS2 10
QM_PRIO_BUSINESS3 11
QM_PRIO_BUSINESS4 12
QM_PRIO_BUSINESS5 13
QM_PRIO_BUSINESS6 14
QM_PRIO_BUSINESS7 15
QM_PRIO_BUSINESS8 16
9. Other minor fixes and optimizations
27.07.2007 V1.24.4
~~~~~~~~~~~~~~~~~~~
1. Important:
"rpfilter" clause of "ipfw" blocks ARP packets - fixed
2. Under heavy traffic flood (several times exceeding link capacity),
traffic from strict low-priority queues was served altogether with high-priority traffic - fixed
July 24, 2007 V1.24.2
~~~~~~~~~~~~~~~~~~~~
1. Old implementation of RIP protocol is included for backward compatibility.
2. Devices configured as "-type slave" may drop association frequently - fixed.
3. "mint -mode " paramater incorrectly displayed when using SNMP access.
4. MINT monitor sometimes does not display current signal levels.
5. Neighbour search/select process in roaming is improved.
6. Major extensions to "qm"/"ipfw" logic:
Option "pass" is added, allowing to process the actions specified in the rule, but to continue lookup for further matching rules.
For example:
ipfw add 1 pass log tcp ....
ipfw add 2 reject tcp ....
If packet is matched rule 1, it will be logged, but flitering/matching will continue from rule 2.
Additional frame/packet classification options include:
[vlan=N] [dot1p=N] [swg=N] [ether=X] [dscp=N|tos=N],
allowing to match based on 802.1q VLAN ID, 802.1p priority,
MINT switching group number (SWitchGroup), Ethernet frame type (EtherType),
as well as ip_tos field against particular DSCP or IP_Precedence value.
The same options are applicable to QoS manager (QM) configuration.
For example:
qm add 1 ch1 pass vlan=5 dscp=51 tcp ....
qm add 2 ch2 tcp ....
If rule 1 is matched, classification will continue, but all the options specified (vlan=5; dscp=51) will be immediately applied to a frame/packet.
New properties are also added to QM channel configuration:
vlan=N , dot1p=N , dscp=N
These parameters allow DSCP labels and/or 802.1p/802.1q manipulations, tagging, re-tagging, clearing or even removing tags.
Numerical value of "-2" removes the specified option from the channel configuration.
Value of "-1" "resets" the field:
- sets DSCP field (allowed range 0-63) to zero.
- sets 802.1p priority field (allowed range 0-7) to zero, AND,
if 802.1q VLAN ID field doesn't exist (or doesn't specified in the channel properties), removes the whole 802.1q/802.1p header.
- removes VLAN ID (allowed range 0-4095), completely with the 802.1p/802.1q header (even if 802.1p filed exists).
Zero value sets the specified field to "0".
Now, with "pass" option, QM allows to create complex and flexible traffic classification, prioritization and tagging rulesets.
For example:
# Channel 1 resets both DSCP tags and 802.1p priorities
qm ch1 dscp=0 dot1p=-1
# Channel 2 sets local flow priority to QM_PRIO_BUSINESS1 and packet field DSCP to 31
qm ch2 pri=9 dscp=31
# Channel 3 sets local flow priority to QM_PRIO_VIDEO and packet field DSCP to 11
qm ch3 pri=3 dscp=11
# Channel 2 sets local flow priority to QM_PRIO_BUSINESS8 and packet field DSCP to 51
qm ch4 pri=16 dscp=51
# Ruleset. Passing all the traffic from untrusted source via channel 1, to remove/reset all the priorities
qm add ch1 pass all from 0/0 to 0/0
# Some flows (TCP from the given subnet) goes into channel 2
qm add ch2 tcp from X.X.X.0/24 to 0/0
# Video (UDP flows from given senders) goes to channel 3
qm add ch3 udp from X.X.X.0/24 PORT to 0/0
# Remaining traffic will be treated "best-effort" by default.
# This rule assigns it to channel 4, to set up some "default priority" as specified
qm add ch4 all from 0/0 to 0/0
Note: Actual prioritization inside MINT network is based on "pri=N" settings, not on ToS/DSCP/802.1p!
DSCP tags are passed transparently through the MINT cloud, in bridge or router mode.
802.1p tags can be passed only in bridge mode.
It is possible to set/reset dot1p/vlan and/or dscp values when frame/packet leaves MINT network, if required.
For example:
# Channel 25 sets 802.1p priority of 5.
# Note: if VLAN header didn't exist before, it will be added automatically with 802.1q VLAN set to zero.
qm ch25 dot1p=5
# Channel 26 sets 802.1p priority of 7 and 802.1q VLAN - to 4
# Note: if VLAN header didn't exist before, it will be added automatically.
qm ch26 vlan=7 dot1p=4
# Place all the IP packets leaving MINT network via eth0 interface into channel 25, if they were labeled DSCP 11
qm addout eth0 ch25 dscp11 from 0/0 to 0/0
# Place all the IP packets leaving MINT network via eth0 interface into channel 26, if they were labeled DSCP 13
qm addout eth0 ch26 dscp13 from 0/0 to 0/0
and so on...
June 30, 2007 V1.23.1
~~~~~~~~~~~~~~~~~~~~
1. "Poll" mode bug fixed.
2. Device with DFS feature but with "-roaming disable" configured was unable to connect - fixed
June 21, 2007 V1.23.0
~~~~~~~~~~~~~~~~~~~~
1. MINT Protocol compatibility with earlier versions of MINT protocol restored
(1.18 and below)
2. Error fixed in marker access decreasing maximum attainable speed.
June 15, 2007 V1.22.18
~~~~~~~~~~~~~~~~~~~~
1. NAT Module worked improperly with the automatic appointment of public addresses obtained through DHCP.
2. NAT module created many false entries in the links table while fragmented packets processing, that mistake led to wrong exceed of the maxlinks Nat limit.
3. In the absence of GPS coordinates in configuration, "rf distance" parameter in the "auto" mode is taken to the nearest hub (to the base stations, for example).
09.06.2007 V1.22.17
~~~~~~~~~~~~~~~~~~~~
1. When installing firmware MINT over RMA, mint automatically runs on all radio interfaces of a device with default settings.
2. "mint addnode" command added with "-note STRING" parameter for comments.
3. A number of inaccuracies is eliminated in roaming and airupdate mechanisms.
May 29, 2007 V1.22.12
~~~~~~~~~~~~~~~~~~~~
1. Module rip is deleted from CPE devices software.
2. Connection between two radios of a single device is not possible.
24.05.2007 V1.22.11
~~~~~~~~~~~~~~~~~~~~
1. SLAVE node having connections via pseudo-rf interface (prf) and working in roaming enable mode didnІt connect with roaming leader through the radio interface, fixed.
2. Speed response to power down of a neighbor node is raised.
3. SLAVE node after connecting to the network enters eco mode, reducing service packages network load.
4. New SNMP Traps added.
17.05.2007 V1.22.8
~~~~~~~~~~~~~~~~~~~
1. mint IFNAME trace MAC command added
2. Fixed error with the previous version of MINT with wrong working after the protocol restart (mint restart)
11.05.2007 V1.22.3
~~~~~~~~~~~~~~~~~~~
1. SNMP (snmpd) module bug fixed (caused CPE reboot).
2. Imported configuration protection from incorrect data and open quotes improved.
3. Public beta-version
07.05.2007 V1.22.2
~~~~~~~~~~~~~~~~~~~
1. AudioMonitor support added
mint IFNAME monitor -audio [full] [-mac MAC]
This command starts working immediately and can be saved in
the configuration which allows using it even after unit's reboot.
To turn AudioMonitor off:
mint IFNAME monitor -noaudio
In "full" mode, audio monitoring is performed according to a maximal
receiving level among neighbors and candidates.
If "full" parameter is not specified, a an arithmetical mean of receive and transmit
signal is calculated (only for neighbors). The neighbor for which this value
is maximal will be indicated in audio monitor.
06.05.2007 V1.22.1
~~~~~~~~~~~~~~~~~~~
1. Broadcast traffic optimization for point-to-point links.
2. Switch nodes search mechanism error fixed.
3. Minor errors in "ipfw/qm" rules fixed.
4. Unconditional 802.1p support added
5. "sys [no]iptos" option removed.
Imstead, the following command added:
qm option [-]voice [-]dot1p [-]tos [-]strict
to manage automatical prioritization in specified modes.
"strict" option means that "Strict Priority" policy is applied
to all queues, otherwise (by default) "Weighted Fair Queuing" policy
is used.
02.05.2007 V1.22.0
~~~~~~~~~~~~~~~~~~~
1. New algorithm for broadcast/multicast traffic transmitting over MINT network.
Now broadcast traffic is only spread across those nodes which belong to the
corresponding switching zone and nodes which form optimal paths to deliver
data for this zone (incompatible with versions below 1.22.0)
2. Transparent packets prioritization support in MINT network.
Supported by using channels management in "qm" command.
Administrator can put streams into different channels based on
"qm/ipfw" rules as well as "tos" and "dscp" fields.
qm ch1 pri=12
qm add ch1 all from x/x to y/y
qm add ch1 dscp31 all from a to b
qm add ch1 dscp42
Each channel can be assigned a priority (1..16)
Once assigned, a priority will be automatically recognized by every node
inside MINT network.
Priority scheme looks as follows:
QM_PRIO_VOICE 1
QM_PRIO_RT1 2
QM_PRIO_VIDEO 3
QM_PRIO_RT2 4
QM_PRIO_QOS1 5
QM_PRIO_QOS2 6
QM_PRIO_QOS3 7
QM_PRIO_QOS4 8
QM_PRIO_BUSINESS1 9
QM_PRIO_BUSINESS2 10
QM_PRIO_BUSINESS3 11
QM_PRIO_BUSINESS4 12
QM_PRIO_BUSINESS5 13
QM_PRIO_BUSINESS6 14
QM_PRIO_BUSINESS7 15
QM_PRIO_BUSINESS8 16
Priorities "1" and "2" are additionally processed as "voice".
Packets from which the priority is not clearly defined will be sent via
common queue with "Best Effort".
Priorities might be automatically shifted by one or two levels up in
"Emergency Level" mode.
3. New firmware features an updated version of autobitrate mechanism which provides
with increased link reliability and stability (esp for mobile clients).
4. Additional roaming profiles parameters are added:
[-type {master|mesh|slave}]
[-key XXX] [-nodeid N] [-netid N]
5. "Over-the-air Update" mechanism upgrade. Automatic upgrade now is performed
with no interruption of regular data flows.
6. "dhcpd" module error fixed which caused CPE rebooting.
7. "arip" module error fixed which occured when having several aliases on one
interface.
8. Bug fixed - some telnet clients incorrectly imported/exported unit's configuration