Redfish: Disable/enable out of band IPMI

[Joseph Reynolds]

On 12/12/19 1:30 AM, Carol Wang wrote:
> Hi,
>
> Looking to implement IPMI Enable / Disable as part of Redfish's 
> Manager Network
> Protocol.
> https://redfish.dmtf.org/schemas/ManagerNetworkProtocol.v1_5_0.json

Carol, that is good.  We are also looking to implement this protocol, 
plus the ability to control other services such as Avahi service 
discovery, and SSH access to the BMC shell.  I hope we can work together.

Note that we have discussed this work in the following places:
- https://github.com/openbmc/openbmc/wiki/Security-working-group
- https://lists.ozlabs.org/pipermail/openbmc/2019-September/018379.html
- https://github.com/ibm-openbmc/dev/issues/612

With that said, I don't have an answer to your question.  Can someone 
who knows systemd help us determine the best way to disable/stop and 
re-enable/start the [phosphor-ipmi-net][2] service?  I think a D-Bus 
interface to represent the status of the service and be able to control 
it is expected.

I can help with the bmcweb implementation, if needed.

[2]: 
https://github.com/openbmc/openbmc/blob/master/meta-phosphor/recipes-phosphor/ipmi/phosphor-ipmi-net_git.bb

- Joseph
> Two ways to get/set the current status of net IPMI:
> 1. Add an interface in phosphor-dbus-interface to indicate the status 
> of net IPMI.
> Have a daemon to monitor the status, if the status is changed, then 
> enable or
> disable the net IPMI service and socket.
> 2. Check the net IPMI socket state by getData()[1] in bmcweb. If the 
> state is
> "running" or "listening", the net IPMI status is true, otherwise, the 
> status is
> false. Then bmcweb can enable or disable the service and socket.
>
> Wondering if anyone has any thoughts on this feature, which way is better.
> If add interface, in which daemon this interface should be implemented?
>
> [1] 
> https://github.com/openbmc/bmcweb/blob/master/redfish-core/lib/network_protocol.hpp#L190
>
> Thanks,