[illumos-gate] 17962 klm: refresh_nlm_rpc() handle re-init is missing CLSET_BINDSRCADDR -- Toomas Soome <tsoome⊙mc>

Heh... now that I have 10Gbit in a good chunk of work infra, I'm noticing some of the sommerfeld-reported phenomenon.

Doesn't help my switch is pushing same-LAN rtt to 300-400 usec

i'm looking forward to those changes...

so the tsq thing needs a bit of work and a lot of testing, but testing of the prototype right now across a wider set of drivers would be helpful (basically: which drivers do tx ring reclaim promptly enough to keep things moving). And it needs some careful review of the assumptions I made.

I've only tested it so far on e1000g and virtio; I have a test machine with an i40e that I need to set up so I can try it on a machine I can crash w/o breaking the home network.

i've got an ixgbe NIC on my home server I could try it with... and we have systems with i40e and mlxcx in the lab we could at least do a quick 'load and run iperf3' or such to comare before/after

sommerfeld: hopefully soon I can try it on 100G (Chelsio T6), I just have some things to get off my plate first

jbk: note that the limiter is disabled by default in the patch I published; you have to turn it on with mdb -kw. The outstanding-byte-count is maintained regardless of the state of the limiter.

do any of you have an easy to boot x86, and are willing to pastebin me the stack the first time we call `boot_alloc`?

there you go: pasteboard.co/g24RD912KSPh.png -- sorry, its console image:D

no that's fine, thanks

[illumos-gate] 17887 loader.efi: panic() should show stack trace -- Toomas Soome <tsoome⊙mc>