Gargoyle Forum

Hi all,

my goal is to have two different Wifi network, one without limitation and the second one with QoS enable to limit upload and download bandwidth.
I setup an Openwrt router with some pkgs (wshaper & bandwidthd) to gain the funcionalities desired, it works but with some limitation due to high resources required by bandwidthd.

Network setup:
WAN -> PPPoe 10Mb download / 1Mb upload

lan1 -> 192.168.1.1 DHCP 192.168.1.100-200
Wifi1 -> lan1 -> "unlimited"

lan2 -> 192.168.2.1 DHCP 192.168.2.100-200
Wifi2 -> lan2 -> "limited" @ 1Mb download & 0,2Mb upload

All wifi connected devices (phones and tablets) can easily switch between wifi networks to get limited or not.

I setup a Gargoyle installation, added the second wifi network editing /etc/config/wirless, network, dhcp and firewall and is corretcly running.
As of someone else already found, QoS and bandwidth monitor doesn't shows any data; If I'm not in wrong, Gargoyle's QoS works only with one "bridge-lan" interface.

Now I reverted to a "simple" setup with one Wifi, created Quota & BWM for 2 different IPs range: 192.168.1.10-20 "unlimted" - 192.168.1.50-150 "limited". Devices that needs "unlimited" connection have a IP (10-20) reserved in DHCP table.
It works flawlessy, but I would like a simpler setup: now, every time that a "new" clients needs a unlimited connection (or if an old one needs to be limited), "someone" has to login in web interface make the appropriate modifications to the DHCP table.

My question is:
Is possible to "tie" DHCP making it serving 192.168.1.10-100 IPs to wifi1's clients and 192.168.1.101-200 IPs to wifi2's clients (I don't matter if both ranges are in the same subnet) ?

This has not been done in Gargoyle. I have speculated that something like this might be possible and wrote about it here. But to my knowledge it has not yet been done.

http://www.gargoyle-router.com/phpbb/vi ... 967#p25847

You actually don't need to split the subnets to do this. It's possible to use the --physdev-in iptables match to differentiate which bridge interface a packet comes in on in iptables.

It may be possible to then get around the problem of matching outbound traffic (you will get errors if you try using the --physdev-out), you can use the physdev-in match to create an ipset of all ips in the guest network, and then match on that.

The problem I see with that approach is how to have different IP address assignments for the guest vs non-guest Wifi. With the approach I suggested you can use dnsmasq to do this. With your approach do you think it could be done? I could not find a way but that does not mean there isn't one.

I think having two ranges would be the best way to match the quota and QoS system allowing selections based on guest vs non-guest.

pbix: My suggestion would be to modify the qos, quotas, etc to have an explicit option in the interface to match the guest network.

In order to have separate subnets extensive modifications to the implementations of these features would be required anyway. I expect that adding a new type of match based on wireless network and implemented using --physdev would be much easier than modifying these capabilities to properly handle separate subnets. Do you disagree?

One added advantage of doing it that way, is that instead of having to know something about networking and subnets to implement the match, people could just select "match guest network" from a dropdown, or enable an appropriately named checkbox rather than having to manually enter the proper subnet and a mask they want to match.

Well my idea was to use different address ranges in the same subnet so no modifications would be needed to any pages. That's the whole point really.

You know better than anyone as to how much work it would be to make special selections for guest networks on the QoS, Quota, DHCP, Bandwidth display pages and consider the implications to all the plugins. Making such features is great for people who use guest networks but a complication for those who do not. Such compromises are what software design is about.

+1 for Eric

But I'm not a developer