I thought I saw an update (or maybe just an issue filed) sometime relatively recently to update the intel core code for a NIC... but I can't recall if that was for i40e or a different intel nic

(if it was i40e, maybe that migt help?)

sommerfeld: at first glance, it appears the mechanims for controlling it are documented

so it should at least in theory be possible I think..

do we know who has what?

it'd be useful if rather than asking each time we made a note, but I know _I_ haven't.

we have some systems w/ i40e in our lab

my home system has igb and ixgbe as well

we also have come mlxcx cards

for the intel ones the specifics seem like they matter

though I used the emulated one in qemu for pci bringup, so that driver is presumably about as forgiving as possible

sommerfeld: Speed negotiation as in disable auto-neg or control what it advertises?

rmustacc: good question. preferably both. Not sure which one is more likely to avoid whatever is causing the link flaps.

as the other end is an 10gig ONT managed by my ISP I have extremely limited visibility into what it's doing.

I don't think it'd be too bad at some point.

@sommerfeld I have an i40e core code update branch.

I can't remember if it builds.

Lemme spin one and see what does/doesn't happen.

Will have an answer in <= 40mins. MUCH <= if a trainwreck.

The branch is in github.com/danmcd/illumos-gate/tree/i40e-common-update @sommerfeld

Shit, still needs work; blew up on build.

Hmmm... blew up not where I expected.

Oh that's right, smatch whines but gcc doesn't. No packages but yes driver.

@sommerfeld want a compile copy? Or you can just take from the aforementioned branch and build it yourself.

I think @tsoome_'s new smatch copes much better with it.

just grabbed the branch. (git remote add danmcd github.com/danmcd/illumos-gate.git; git fetch --all)

yes, the updated smatch can handle pragma packed better and is not complaining.

Thanks @sommerfeld --> I'll not force-push to that branch to keep it honest since someone's downstreaming it.

I want to JUST have it be a common-code update. Some reports say it might fix fenix illumos#13230

fenix` illumos#13230

fenix illumos 13230

BUG 13230: i40e has duplicate traffic when used with bhyve/snoop running (New)

↳ illumos.org/issues/13230

might need to extend its matching :)

My bad.

I'm used to internal TritonBot (nee JoyBot) ways.

heh..

[illumos-gate] 17813 mac softring polling does not enforce byte limit -- Ryan Zezeski <ryan⊙zc>

Thanks @rzezeski for that!

[illumos-gate] 17789 want transitional virtio drivers -- Andy Fiddaman <illumos⊙fn>

[illumos-gate] 17836 Re-enable RDSEED on Zen5 processors with new enough microcode -- Andy Fiddaman <illumos⊙fn>

andyf: nice nice nice

Yes. I owe you some work for braich, I know

sure, but it makes other things nicer immediately too :)