hm. is my reading of the sources correct? my use case: a VNIC in a global zone configured to advertise an ULA ipv6 prefix to the etherstub, but I don't see any RA packet. I checked the in.ndpd/main.c and saw this branch, github.com/illumos/illumos-gate/blo…md-inet/usr.lib/in.ndpd/main.c#L496 . Running /usr/lib/inet/in.ndpd -d -t shows me exactly that: "if_process: interface

not yet present nfs_gz0"

As a result, VNIC is configured to act as a router, but skipped on each event to advertise a prefix

gemelen: what does ifconfig -a show for that interface? from a very quick look at the source, it needs to be 'UP' and the 'NOLOCAL' flag should _not_ be present on it

jbk: You didn't file a bug yet for the show-devs issue you had right?

I'm writing something up but didn't want to dup it.

Ah, never mind. Keith wrote a bug for it. illumos.org/issues/16377

→ BUG 16377: pcieadm show-devs output incorrect for dev with no current-width property (New)

not yet

OK. I'll just use that one for this.

The note I left on there more or less gets that understood. This is mostly a call generator and confusing. But the link is down on all of them, so there's not going to be useful information there.

jbk: the ways I've created the vnic resulted in two logical interfaces, nfs_gz0 which wasn't UP and nfs_gz0:1 in UP. but doing `ifconfig nfs_gz0 inet6 up` broke connectivity. anyway, that was a direction for me

btw, the odd i40e speed issue I mentioned last week got weirder. i40e <-> adtran ONT: 85+% packet loss in ping at 10G. i40e -> switch -> adtran ONT: 0% packet loss at 2.5G, 5G, and 10G. updating the i40e "NVM" firmware had no effect (other than changing the boot message from "i40e0: The driver for the device detected an older version of the NVM image (1.7) than expected (1.9)." i40e0: The driver for the device detected a newer version of the

NVM image (1.15) than expected (1.10)".to " But Sonic says a switch in the middle isn't a supportable configuration.

oops, that got mangled by a paste.

the supportable configs (direct cable router <-> ONT) either don't work or are at reduced speed, while the apparently-working config is deemed not supportable by the ISP.

if I had to guess, the 'not supported' is probably just 'we expect to see 1 MAC on the other end of the pipe'

but I don't think we're doing anything that'd matter in the i40e driver

well maybe flow control or fec?

those can sometimes be a problem

if you've not messed with those yet

jbk: flow control is off as is fec; if I'm not mistaken fec is for speeds over 10G

jbk: support guy said "you'd periodically have to call in to have us clear the mac filter" but wasn't specific about that.

I wish he'd said more, I don't see how that would happen

if the switch started a DHCP request I can see that mucking things up.

(I have it configured as L2-only on the relevant ports; L3 only on the out-of-band mgt port)

i guess i40e's internal switching doesn't count :)

i can't recall about mellanox, and i've not looked at chelsio at all, but I do wonder if any other NICs do anything similar (basically geared towards a system hosting VMs)

well, I can think of some people using chelsio nics on a system geared to hosting VMs

though that doesn't answer your question

yes, but that doesn't mean they have an entire switch, complete with the ability to do things like run LLDP on chip

but right now, IIRC, vnics basically just maybe get a TX and RX ring from the HW

though there's probably more that could be done by the NIC -- but is it common enough where it'd make sense to maybe have some internal API

jbk/richlowe: tech is here, showed me the aforementioned mac filter entries and it's just (ipv4/v6 <-> mac) pairs (a limited number of them).

so the meltdowns they are concerned with involved .. less careful configs, no doubt.

yeah, if you have apple hardware (especially), their mac randomizer can cause real trouble for things like that.

it protects your privacy behind a veil of having to call IT and ask specifically if they can look at all the entries that are you every so often.

I ask my family to turn it off for the "at home" network location so my `mac-check` script can detect bogons/new-devices more easily.

I also have NAT because VZ is too gougy and doesn't let static-IPv4 customers ALSO have IPv6. :upside-down-face:

[illumos-gate] 17456 duplicate bool definition in qede NIC driver -- Kyle Simpson <kyle⊙oc>

[illumos-gate] 17471 remove the rest of wireless USB -- Richard Lowe <richlowe⊙rn>

[illumos-gate] 17482 pmap could label the comm page -- Andy Fiddaman <illumos⊙fn>