ethtool (latest version) only shows rx_queue_0_packets/bytes queue counter even though ixgbe is loaded with default parameters and allocated 16 receive queues. All queues are getting used nevertheless as can be seen from /proc/interrupts. This is on RHEL 5.8 stock kernel 2.6.18-308 but the same is on stock 5.11 2.6.18-398.el5.
I have tried several ixgbe versions and found out that last ixgbe version that ethtool properly displays counters for all allocated queues is 3.16.1 so perhaps this is expected in later versions - would be good to know the reason if that indeed is the case.
Interesting also is that RX queue 0 is shared with TX queue now meaning that RSS=1 mean single queue for RX and TX and RSS=2 mean TX/RX shares queue 0 and RX queue 1 and so on. Obivously, 2.6.18 does not support multiple TX queues.
I tried to look for an answer in all mailing lists but didn't find anything.
Attaching an output from dmesg, ethtool, /proc/interruprs and modinfo.
We also need the system details. Please provide ethtool -i results.
And do include the network card model and module used.
Sure here it is. Card is Intel(R) 10GbE 2P X520 Adapter.
[root@prodsdw80 ~]# ethtool -i eth0
firmware-version: 0x800007f5, 14.5.8
[root@prodsdw80 ~]# lspci | grep 825
42:00.0 Ethernet controller: Intel Corporation 82599EB 10-Gigabit SFP+ Network Connection (rev 01)
42:00.1 Ethernet controller: Intel Corporation 82599EB 10-Gigabit SFP+ Network Connection (rev 01)
Good day. I apologize for the delay.
With regards to the issue, we don't control their distributions, we suggest to contact Red Hat* Support directly.
Again, thank you for contacting Intel.
I am not sure I understand your reply. The problem is not with RedHat or any other distribution.
Problem I am having is with driver from intel.com ixgbe v4.3.13 which is distribution indifferent. It is not part of RedHat distribution. For performance reasons, I am not using driver that come with RH but the one from Intel which does have a problem I have explained in a post.