#join #omnios

as there been some talk about Intel nics for a couple of days, but nothing (yet) today.. I have a ixgbe / Intel 82599 that I've tried for a couple of days to accept an unsupported SFP+ transceiver. I've ad-hoc modified the /kernel/drv/ixgbe.conf for several times now, but it wont apply the allow_unsupported_sfp = 1 which I from a thread dated in November 2016 saw hints about a less restrictive

declaration together with unit-address only.. the driver won't apply that; In afterlight I would have gone for a chelsio from what I've read.. but ebay found this nic for my motherboard.. I feel puzzled about the name= and parent= construct - in my case it seeems parent is pci8086 which doesn't feel right to me :/

My setup pastebin.pl/view/8843f039 - in my interpretation it should then be rewritten as: name = "pciex8086,10fb" parent = "/pci@0,0/pci8086,\<pci8086\>6f08@3" unit-address = "2" ?

you should be able to just allow it globally.. that might be easier than trying to get the notation to apply it to just one NIC

That was my thought, as I will never care about supported gbics at home (and the other ixgbe controller which is built-in is 10GBASE-T anyway).. I restored the file and ran: echo "allow_unsupported_sfp = 1;" >> /kernel/drv/ixgbe.conf but it dont care: "Sep 29 16:35:45 lunar ixgbe: [ID 611667 kern.warning] WARNING: ixgbe2: Unsupported SFP+ module detected. Please replace it with a supported SFP+ module

per Intel documentation, or bypass this check"

what distro?

omnios

you might need to do 'bootadm update-archive'.. but it will require a reboot regardless after making the change

(techincally there are ways around doing a reboot, but it can get complicated)

(while not 100% true, generally the .conf files are used when the driver is loaded, so changes made after that won't take effect until the driver is reloaded)

I forgot about the 'bootadm update-archive v' which contained a ` changed /kernel/drv/ixgbe.conf

but still, no change after a reboot. And no worries about reboot here.. this is my 4th capable Linux-hypervisor that ends up becoming a illumos instead very recently, and I lost my unix beard anyway when family member this weekend rebooted my server as a first resort (when it should've been the last).. :-)

jbk: I tried to ppt it to a bhyve FreeBSD guest but their driver responds with similar message with the similar method.. so I might have to get that Intel gbic instead.. Thanks for your suggestion anyhow!

the driver source for most platforms (including illumos) derives from code provided by Intel, so I'm guessing that check is probably in that common code that everyone is using

[illumos-gate] 15006 Implement pre-auth integrity for fksmbd -- Gordon Ross <gwr⊙rc>

yeah, but through websearches I saw suggests that this Intel driver is a disaster apparently with different teams / code bases for each platform and that both the illumos/FreeBSD versions was inferior (and I've been foolish to think that Intel always was a stable choice)

[illumos-gate] 13923 zpool list -v output is misaligned -- Brian Behlendorf <behlendorf1⊙lg>

[illumos-gate] 15020 HBA_LoadLibrary needs to close config file when done -- Rick McNeal <rick.mcneal⊙nc>

I even tried a ppt to a Linux bhyve guest.. same same.. while at the same time a DAC works in the same port.. so time for some more ebay shopping :)

hrm.. is it me, or would requesting the results of a query be paged into groups of X suggested you want to process the results in groups of X?

and not say 'gather all the results, and then return all of them'

which would seem to defeat a major point of paginating results