termination on hardware errors

termination on hardware errors

[Andrew Geissler]

We have a requirement from our manufacturing team that they be able to configure
the BMC to fail the boot of a system if any error log is generated that calls
out a piece of hardware. The idea being that if the firmware finds any issue
with the hardware, manufacturing wants the system to halt and provide a clear
indication of what that failure is so they can fix it prior to shipping the
system.

High level this will probably be a phosphor-settings field that phosphor-logging
code would query when it gets an error log that has a inventory path in it.
I'd like an external state to indicate when this event occurs. Maybe a new
BMC state like Quiesce? (Quiesce is used by host when it is in fail state
and has an easy mapping to the corresponding Redfish Quiesce)

This is just a feeler email to see if anyone else has an interest in this type
of function or if they have other ideas on how to fulfill this requirement.

Andrew

Re: termination on hardware errors

[Andrew Geissler]

For those interested, which measured by the response to this email, there
are tons of, here’s a link to the design doc :)

https://gerrit.openbmc-project.xyz/c/openbmc/docs/+/29603

> On Feb 18, 2020, at 10:24 AM, Andrew Geissler <geissonator@gmail.com> wrote:
> 
> We have a requirement from our manufacturing team that they be able to configure
> the BMC to fail the boot of a system if any error log is generated that calls
> out a piece of hardware. The idea being that if the firmware finds any issue
> with the hardware, manufacturing wants the system to halt and provide a clear
> indication of what that failure is so they can fix it prior to shipping the
> system.
> 
> High level this will probably be a phosphor-settings field that phosphor-logging
> code would query when it gets an error log that has a inventory path in it.
> I'd like an external state to indicate when this event occurs. Maybe a new
> BMC state like Quiesce? (Quiesce is used by host when it is in fail state
> and has an easy mapping to the corresponding Redfish Quiesce)
> 
> This is just a feeler email to see if anyone else has an interest in this type
> of function or if they have other ideas on how to fulfill this requirement.
> 
> Andrew