jbk: do you know why the libmlsvc dmod isn't packaged?

no

jbk: So I'd start with looking at the bridge config space and seeing what's in the slot capabilities register. That's one way that the LED can be advertised.

Another way is to see if they have the Native PCIe Enclosure Management Extended Capability.

At least those are the "standard" PCIe based ways to do it. But otherwise, I'd probably go through what all they expose in IPMI and/or redfish.

Just in case the old-timers weren't feeling old enough... techhub.social/@SolarisDiaspora/114683151979332159

well ahead of you

rmustacc: also.. on this box are a few interesting devices that seem to report '-1' for # of lanes (e.g. '60/5/0 PCIe Gen 1x4294967295 pcieb9 Unknown device: 0x352c') -- any suggestions on digging into that? presumably we probably want to handle that better

lest someone thing they have a really, really, really high speed link :)

you're probably getting PCI_INVAL* from a cfgspace read

I don't remember which functions on x86 are at the bottom of config space reads though

jbk: Could you run pcieadm save-cfgspace -a, zip that up, and send it to me along with a prtconf -v output?

It's unlikely an invalid read in this case given my memory.

Maybe also an smbios dump just so I know what I'm looking at.

I could swear it was a symptom when I'd broken that stuff

... nothing to say I didn't break more than that stuff