From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============5384712766215748344==" MIME-Version: 1.0 From: Denis Kenzior Subject: Re: [PATCH 12/13] ap: add support for DHCPv4 server Date: Tue, 20 Oct 2020 13:51:39 -0500 Message-ID: In-Reply-To: List-Id: To: iwd@lists.01.org --===============5384712766215748344== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Hi James, On 10/20/20 1:41 PM, James Prestwood wrote: > On Tue, 2020-10-20 at 13:28 -0500, Denis Kenzior wrote: >> Hi James, >> >> On 10/20/20 1:02 PM, James Prestwood wrote: >>> The DHCP server can be enabled by including an [IPv4] group >>> in the AP provisioning file and including at least a subnet >>> mask. >> >> Any chance we can add this code to the current implementation without >> changing >> the API just yet? I think it is safe to assume that if the address >> is set prior >> to AP starting on a given interface, then we should not start our own >> DHCP, and >> if the address isn't set, then start DHCP server. If the address is >> provided by >> provisioning, then start our own DHCP as well, overriding whatever we >> picked. > = > Sure, so keep the 'psk' DBus argument and remove [Security].Passphrase? > It seems weird to have both since we would be ignoring one. I don't know yet. I don't think removing the simple 'Start' method which j= ust = provides an SSID and PSK is a good idea. It is useful for quick and dirty = setups. Provisioning files might be for more advanced users and might need = their own dedicated Start method. So something like KnownNetwork(s) but fo= r = AccessPointConfiguration(s). > = > But do we really want to make the decision to use DHCP based on if the > address is set? It really seems like the user would want to decide that > explicitly with a config file. OTOH if all no DHCP options are required > (using defaults) we would need some way of knowing to start DHCP or > not. I just don't see a correlation between the address being set and > the user wanting/not wanting a DHCP server. Well, the user has to first switch the device into AP mode or create the = required interface first somehow. So before they actually hit .Start, they= have = a chance to configure the IP or not. If the IP is configured, then we assu= me = the interface is managed by someone else. If it isn't, we take ownership. I suppose we can add an explicit setting in main.conf somewhere, or in the = profile itself. But I'd keep things simple for now. >> >> I think we may need Andrew's opinion on whether these should be part >> of >> ap_config. P2P might not really care about setting up these >> variables and would >> prefer for ap.c to just figure things out. > = > Sure, but they aren't required anyways so p2p doesn't have to include > them and defaults will be used (once netmask is fixed). The ap_config > structure was just a convenient storage object for these but we could > define something else too. Yes, true. But not much sense in actually storing them long-term if that's= the = case. Generate them / load from settings, give them to dhcp-server and for= get. Regards, -Denis --===============5384712766215748344==--