HEADS UP: Updated em/igb/ixgbe back in snapshots

dostrow

A quick question looking for advice here. We had just purchased $25K worth of gear all with 4 ixgbe dual 10 gig card in them with the intention of using pfsense as our primary firewall going forward, this way was way more economical than going the Cisco/Juniper/Whathaveyou route given our need for 10 gig interfaces. All was well and good until we tried to get things working when the gear arrived last week. We were immediately hit by bug 3250 with 2.1-REL. Hearing that the drivers were more stable in 2.0.x we tried installing 2.0.3 only to find out that there was not adequate support for our RAID controllers and thus could not install. With previous 2.1.1 snapshots we got further but still suffered from the mbuf leak that eventually caused the systems to panic and crash. Now that the decision has been made to revert to the 2.1-REL versions of the drivers (which I agree is the right call to have made), we are basically stuck needing 2.2 (for which I can find no builds publicly available) for these boxes which were slated to go into production within a week. If you were in my shoes, had hardware which was already purchased and cannot be returned and a deadline that really can't move…what would you do?

Supermule

Run it en ESXi on your new hardware. Then it works fine.

dostrow

Definitely something to try, though somehow running my edge devices under a hypervisor makes me a little queasy, not to mention setting things up with BGP and CARP over virtual interfaces all running at 10 gig adds some odd complexity but if it is my only option so be it.

stephenw10

Do you have a support subscription? If so I'd go straight to ESF who will have a much better handle on what can/can't be achieved by using whichever driver version might be available.
Do they run OK in 2.0.3? You could maybe reconfigure the disk controllers into some mode that is supported or there maybe backported drivers for 8.1.

Steve

athurdent

Maybe running pfSense from USB flash is an option?

qseb

Hello

is it still possible to insert manually new igb drivers to boot.conf for next 2.1.1 snaphots?
Currently, I have a 2.1.1 from january with new igb, how to extract modules?

stephenw10

The drivers are compiled into the kernel so you can't extract them as modules. However if you did get hold of the modules there should be no problem loading them via loader.conf.local.

Steve

demco

Hi,

Anyone know if the snapshot dated 20140217-1607 contains the newer driver?

Tried the image on a Rangeley box and boot log seem to indicate no network driver.

pci0: <network, ethernet="">at device 20.0 (no driver attached)
pci0: <network, ethernet="">at device 20.1 (no driver attached)
pci0: <network, ethernet="">at device 20.2 (no driver attached)
pci0: <network, ethernet="">at device 20.3 (no driver attached)

Just want to confirm that it is the driver and not something else, before trying to build the new driver.

full boot log attached.

Thanks.

bootlog.txt</network,></network,></network,></network,>

qseb

@stephenw10:

The drivers are compiled into the kernel so you can't extract them as modules. However if you did get hold of the modules there should be no problem loading them via loader.conf.local.

Steve

Yes I saw kldstat not returning modules. How to get the igb module file?

stephenw10

Compile it yourself.
Ask ESF nicely.
Search the forum, I think a few others have compiled newer modules but I also think that the version included in earlier 2.1.1 snapshots had patches.

They were removed from 2.1.1 for good reason though. I guess if it works at all for you that's better than than nothing.

Steve

eri--

The drivers from HEAD have been enabled again in the snapshots after fixing the cause of panics.

Please try them again in later snapshots.

stephenw10

Woot! ;D
Great news. Updating…
They are in current snapshots I presume? (08-Mar-2014) Or wait for new builds?

Steve

maverick_slo

So NEW drivers are again included?
Also em drivers?

Thanks!

maverick_slo

Huh only March 8 snapshots are available…
Something went wrong?

stephenw10

The build server maybe not running for some other reason. The new drivers may be in the 8/3/14 build. Unfortunately we I …uh...can't check the (cough) tools. ;)
Logs show it built images successfully yesterday but hasn't uploaded them.
I'll check my test box.

Looks 8/3/14 is still using the 7.3.2 em driver so wait for a new snapshot.

[2.1.1-PRERELEASE][root@pfsense.localdomain]/root(1): uname -a
FreeBSD pfsense.localdomain 8.3-RELEASE-p14 FreeBSD 8.3-RELEASE-p14 #0: Sat Mar  8 12:47:31 EST 2014     root@snapshots-8_3-i386.builders.pfsense.org:/usr/obj.pfSense/usr/pfSensesrc/src/sys/pfSense_wrap.8.i386  i386
[2.1.1-PRERELEASE][root@pfsense.localdomain]/root(2): sysctl dev.em
dev.em.0.%desc: Intel(R) PRO/1000 Network Connection 7.3.2
dev.em.0.%driver: em

Steve

jimp

Builder VMs are offline at the moment, shuffling some things around again completely unrelated to anything else. Once things settle we'll fire them back up and get new snapshots out.

stephenw10 - you can get access to the tools repo if you follow the instructions on the list, mail in a copy of your ssh public key to the address given on the list in the thread.

stephenw10

Yes, I realise I could but thanks. I haven't done so because this is the first time I've had reason to look at the tools repo. I have (currently) no need to modify or compile the code only occasionally reviewing what's been changed.

Steve

jasonlitka

@ermal:

The drivers from HEAD have been enabled again in the snapshots after fixing the cause of panics.

Please try them again in later snapshots.

SWEET!

I'll update to a newer snapshot next week. Right now I'm in the middle of trying to migrate about 55TB of data from two failed storage arrays…

stephenw10

em drivers back to current. :)

[2.1.1-PRERELEASE][root@pfsense.localdomain]/root(3): sysctl dev.em.0
dev.em.0.%desc: Intel(R) PRO/1000 Network Connection 7.3.8

Steve

jasonlitka

@Jason:

SWEET!

I'll update to a newer snapshot next week. Right now I'm in the middle of trying to migrate about 55TB of data from two failed storage arrays…

So the good news is that it's not crashing any more.

The bad news is that I still seem to be hitting a pretty hard wall at ~2.1Gbit/s across 10Gb ix interfaces (tried both SFP+ Direct Attach and Intel SR optics with a 1M OM3 patch). Anyone have any thoughts on that? It's 2.1Gbit/s of TOTAL throughput, not in each direction.