Feedback on Current OpenBMC and Proposing Some Improvements

[Alex Qiu]

Hi OpenBMC community,

It has been a while since Google has adopted the dynamic software stack of
OpenBMC, namely using entity-manager for FRU discovery, dbus-sensors for
sensor reading, and intel-ipmi-oem for IPMI command handling. We discovered
issues and limitations with this dynamic software stack along the way, so
I’m proposing some ideas on how OpenBMC may improve, which may lead to
detailed designs about it. Let me call it "Improvements" in this email per
say. I think the highlight of these ideas are: 1) having a robust framework
to handle hardware topology, and 2) having accommodations for code to
intervene on varieties of BMC tasks.


I'll split the content of this topic into two additional emails for easier
digestion: 1) Difficulties and Issue Examples; 2) "Improvements" Ideas. The
main discussion may still stay in this thread.


Since this is a big architectural change compared to the existing dynamic
software stack, I would like to hear feedback or review on the conceptual
ideas before we turn these ideas into more concrete designs or prototypes.
On the other hand, there is a high probability that I didn’t express my
idea well enough to understand, and there may be a language barrier to get
over. I’ll try to see if I can use some code to make a tiny prototype to
illustrate the ideas better at some point. Thank you!


- Alex Qiu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/openbmc/attachments/20200618/6b3b0302/attachment-0001.htm>