RE: 回覆: 回覆: [phosphor-host-ipmid] Questions about IPMI (OEM) command development

[Chris Chen (TPI)]

Hi Vernon,

Got it. Thank you.

Regards,
Chris Chen

-----Original Message-----
From: Vernon Mauery <vernon.mauery at linux.intel.com> 
Sent: Wednesday, September 15, 2021 10:35 AM
To: Chris Chen (TPI) <Chris.Chen3 at flex.com>
Cc: openbmc at lists.ozlabs.org
Subject: Re: 回覆: 回覆: [phosphor-host-ipmid] Questions about IPMI (OEM) command development

On 11-Sep-2021 03:56 AM, Chris Chen (TPI) wrote:
>Hi Vernon,
>
>Thanks for your detailed explanation, I got a lot of concepts from it and realized how should I carry on next.
>
>However, I'm not sure whether my understanding of the filter is correct, so would like to confirm with you further.
>
>
>  *   If I for now have 3 IPMI provider libraries, such as the original library in "phosphor-host-ipmid", "intel-ipmi-oem" and "my own library", and they all have their own filter with whitelist. These 3 filters will all be registered. Then, when a command comes in, this command will be passed to each filter and check if the command exists in their whitelist. Am I correct?

There is an option to disable the built-in phosphor-host-ipmid filter. 
The one that ships with intel-ipmi-oem was added by Intel to replace the built-in filter for Intel builds. If the intel-ipmi-oem filter does not work for you, but you still want the rest of the provider commands, you might consider adding a patch to your layer that removes the filter so you can create your own.

The filters must all return OK (0) for the command to get executed. They will each be run in priority order, but if any of them return non-zero, then the command will not get executed.

>  *   If yes in above, that means I just need to add the command's NetFn/CommandID in one of these 3 whitelists, the command still will be passed to a specific provider once it exists in one of the whitelists? Or I should add that command's NetFn/Command to the whitelist of its dedicated library?

You would need to make sure that all of the filters are returning OK or the command will not get executed.

>Regards,
>Chris Chen
>
>________________________________
>寄件者: Vernon Mauery <vernon.mauery at linux.intel.com>
>寄件日期: 2021年9月10日 上午 04:15
>收件者: Chris Chen (TPI) <Chris.Chen3 at flex.com>
>副本: openbmc at lists.ozlabs.org <openbmc at lists.ozlabs.org>
>主旨: Re: 回覆: [phosphor-host-ipmid] Questions about IPMI (OEM) command 
>development
>
>On 06-Sep-2021 09:40 AM, Chris Chen (TPI) wrote:
>>Hi all,
>>
>>I guess this mail was got overwhelmed by bunches of emails, so please let me resend it and hope someone could help me. Another, I add one more question marked #5 below.
>
>Thanks for being persistent :)
>
>>
>>Hi all,
>>
>>I have studied "phosphor-host-ipmid" package for a couple of weeks and got understanding that it has implemented some mandatory and optional commands already. And I also knew that I'm able to leverage "intel-ipmi-oem" package to get more command handlers. But there're some questions raised in my mind, I hope someone could give me some suggestions.
>>
>>  1.  What's the normal development procedure in case I want to make our own OEM commands? I mean do you guys encourage me to create a new "xxx-ipmi-oem" package or to leverage "intel-ipmi-oem" and then add our own commands in it?
>
>First some common language:
>1) xxx-ipmi-oem is an IPMI provider library. It doesn't need to be 
>called xxx-ipmi-oem, but that is just a convenient naming convention.
>2) whitelist (prefer allowlist) is one kind of filter. Filters are 
>registered like command handlers, with a priority. Filters can do more 
>than just block or allow the commands coming in. The filter is passed 
>the request and can do anything it wants before passing it on. Logging, 
>mangling, blocking, etc.
>
>It depends on a lot of things. Maybe yes, create a new repo? More 
>discussion below.
>
>>  2.  I for now added "intel-ipmi-oem" package only for using its "<Chassis>:<Set Front Panel Enables>" command because I figured out the same command in the "phosphor-host-ipmid" is not able to match the d-bus interface in the "x86-power-control". I'm not sure if I will need other commands in the "intel-ipmi-oem", but what if I don't need other commands, how to remove them? to use the whitelist mechanism?
>
>There is not a method to remove them, but you can filter them as you 
>suggested. Also, if you want to add a different implementation (from a 
>different IPMI provider library) at a higher priority, only the higher 
>priority implementation will be used.
>
>>  3.  Is it okay to add more than 1 xxx-ipmi-ome packages? If yes, how to manage the whitelist between these 2 OEM packages?
>
>You can have as many IPMI provider libraries on the BMC as needed.
>
>The filtering mechanism does not discriminate between the provider, it 
>is used at execution time when a command comes in. Every command is 
>passed to each filter
>
>>  4.  Am I able to use "intel-ipmi-oem" on a platform that is using AMD CPU?
>
>IANAL, but, yes. It is open source and the license does not make that 
>restriction. It might look kind of weird though :)
>
>If you like all the IPMI functions implemented by the intel-ipmi-oem, 
>it is fine to use it as is. If there are things you want to change, 
>there are different ways to go about it.
>
>Here are a couple of scenarios with some options:
>1) You like the behavior of an IPMI provider library, but want some 
>changes. Talk with the owner and see if you can come to some consensus 
>that works for everyone. This is the most opensource, 
>community-building, good-will behavior of all. Submit a change and see 
>if you can get it merged.
>
>2) You like all the IPMI commands in an IPMI provider library except 
>for one or two. If you want to own the work to keep a patch up to date, 
>it is possible to just add a bbappend and a patch to modify the code 
>prior to building. This is okay, but maybe not as ideal. It may end up 
>with a smaller image size than adding a new provider library to 
>overload the command implementation.
>
>3) You like all the IPMI commands in an IPMI provider library except 
>for one or two. You can create a new IPMI provider library for your 
>organization that can filter out and/or provide overloads for any 
>existing implementation of IPMI commands from any number of other IPMI 
>providers.
>
>4) You only like some bits of another IPMI provider. You may be best 
>off to just copy those bits into a new IPMI provider library of your 
>own, add your own other OEM commands and/or implementations and carry on.
>
>Really, none of the 'solutions' are ideal, but we are trying to make 
>trade-offs here that allow individual contributing organizations 
>freedom to implement OEM commands as they wish without any one 
>organization being a gatekeeper.
>
>--Vernon
>
>Legal Disclaimer :
>The information contained in this message may be privileged and confidential.
>It is intended to be read only by the individual or entity to whom it 
>is addressed or by their designee. If the reader of this message is not 
>the intended recipient, you are on notice that any distribution of this 
>message, in any form, is strictly prohibited. If you have received this 
>message in error, please immediately notify the sender and delete or destroy any copy of this message!