hrm...

i'll need to create a test case when i get a chance..

but it looks like signalfd() doesn't record the source pid

or doesn't deliver it

rmustacc: are you satisfied with my explenation as to why no viona.conf was used for code.illumos.org/c/illumos-gate/+/2980

→ CODE REVIEW 2980: 15464 viona should copy tx buffers by default (NEW) | illumos.org/issues/15464

sjorge: I'll circle back there. I won't be able to test something with bnxe until tomorrow and will probably need some help from pmooney to get things rigged up.

[illumos-gate] 15825 nvidia driver bloat exhausts heaptext -- Hans Rosenfeld <rosenfeld⊙gho>

rmustacc: doesn't that imply then you can't poll w/ a single queue? Since the poll entry point is only documented using rings (which requires it's own interrupt), if you have to share it with anything else (e.g. link events, other NIC events), you can't do that

[illumos-gate] 15834 Update Intel microcode to 20230808 -- Dan McDonald <danmcd⊙mi>

If there is no other way to control interrupt generation, yes. But that's not the case even for e1000g.

Almost every NIC these days has a separate way of saying whether or not a queue should generate interrupts.

So even if the MSI-X is shared between multiple causes the

*that's fine.

Does anyone know if there are any plans for an illumos presence at FOSDEM next year?

I'm not sure that this has been usually planned this far in advance, but I would expect various folks to be attending again.

jbk: If you share what device you're looking at, I can try to help look at something and be a little more useful potentially.

rmustacc That's good to hear. It is quite far off, but I'm planning to attend and I was wondering if any help was needed. Do you happen to know who the right persons might be to talk to about that (closer to the date)?

There's a regular group who goes I think. I don't remember who all that is. andyf probably can share who does more of the planning.

dlyund: we usually plan fosdem in october

Thanks rmustacc and Woodstock :-). I'll check with andyf in October if I don't hear something sooner.

I think Woodstock often does a lot of the organizing as well among others.

rmustacc: both pre and post change the function should basically always return true for bxne

dlyund: we don't plan far ahead, as evident last year we did not get a table :( but some of us were present and just hung out

I would expect, even if we don't get a table for some of us to be at FOSDEM next year as wel

I should really read scrollback :D I did not see woodstock's reply until I wrote the rest

Yeah, I'll try and keep my calendar clear for 2024 after missing FOSDEM this year

sjorge then I shall look forward to meeting some of you! :D

A table could be fun though ;-) .