pfSense

For what it's worth, I saw the same behavior with igb interfaces after restoring a 2.3 config with a shaper.

This appears to be caused by r263259 in FreeBSD 11 - Add Tx/Rx multiqueue support to vmx(4).

I believe that ALTQ does not support multiqueues and needs to be complied with the flag VMXNET3_LEGACY_TX flag enabled in order to get it working with ALTQ working again. See Forums posts details relating to ALTQ and IGB when multiqueue was enabled on that driver:

https://forum.pfsense.org/index.php?topic=122367.0

From the revised Makefile in FreeBSD 11 (https://svnweb.freebsd.org/base?view=revision&revision=263259):

# With VMXNET3_LEGACY_TX, the driver will use the non-multiqueue
35    # capable if_start interface.
36    #CFLAGS+= -DVMXNET3_LEGACY_TX

I also see this on Hyper-V virtual NICs (not legacy). Perhaps this is a more generalized issue?

Jim ALTQ does looks its on the a downward path but still pfSense uses it and the traffic shaper is an established key feature.

Looking at FreeBSD problem tickets is confusing, I find some saying igb needs patching, but I also found one where apparently a patch was commited to auto enable legacy queuing when ALTQ is compiled into kernel.

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=208409

Also to confirm igb is broken on pfSense 2.4 and ALTQ

Chris Collins wrote:

Jim ALTQ does looks its on the a downward path but still pfSense uses it and the traffic shaper is an established key feature.

Looking at FreeBSD problem tickets is confusing, I find some saying igb needs patching, but I also found one where apparently a patch was commited to auto enable legacy queuing when ALTQ is compiled into kernel.

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=208409

Also to confirm igb is broken on pfSense 2.4 and ALTQ

ALTQ may be broken, but igb is likely fine.

That "patch" is a bodge, and I am unlikely to being in "legacy tx" support to pfsense.

If there is any issue with igb, I can't replicate it here. On an SG-8860 with igb running 2.4, I can use ALTQ and it has multiple queues on the NICs.

The only type of NICs that fail in this way that I have tried so far are the vmx(4) interfaces from the original report.

Jim Pingle wrote:

If there is any issue with igb, I can't replicate it here. On an SG-8860 with igb running 2.4, I can use ALTQ and it has multiple queues on the NICs.

The only type of NICs that fail in this way that I have tried so far are the vmx(4) interfaces from the original report.

I swapped the VM interfaces from VMXNET3 to EM1000 and no sign of the error anymore. There does appear to be a cosmetic bug in the Traffic Shaper interface which states "This firewall does not have any interfaces assigned that are capable of using ALTQ traffic shaping." Traffic is being assigned to queues and there are no errors in the system logs. Can't seem to see the bug report the OP refers to here (seeing the same with igb interfaces: https://forum.pfsense.org/index.php?topic=123639.0

What page does the error show on, exactly? I can't replicate any problem where that error shows up on a firewall that is actually capable of using the shaper.

Firewall-Traffic Shaper-By Interface. I didn't check the others. If I see traffic in the queues then AltQ should be working right?

Luiz Otavio O Souza wrote:

Should work now: https://github.com/pfsense/FreeBSD-src/commit/91384809a74dc98ef0d2a173718bd79bd77c13bb

Thanks. Will test as soon as the snapshots update (still on 29 Jan at the moment).

Luiz Otavio O Souza wrote:

Should work now: https://github.com/pfsense/FreeBSD-src/commit/91384809a74dc98ef0d2a173718bd79bd77c13bb

Confirmed fixed here. No errors reported using ALTQ and 6 x VMXNET3 interfaces on ESXi 6.5.