Intel 82574L hang issue

hahacc · 04-09-2012, 05:49 PM

Hi guys,
There's now a problem with Intel 82574L Gigabit Network Connection NIC on CentOS 6.2 X86_64 server. I've installed kmod-e1000e-1.9.5-1.el6.elrepo.x86_64 driver and append the following to /etc/grub.conf:

Code:

pcie_aspm=off e1000e.IntMode=1,1 acpi=off noapic

But still, the NIC hang now and then, and after NIC hang, I can see that TX bytes from ifconfig shows a great value:

Code:

eth0      Link encap:Ethernet  HWaddr 00:1E:67:03:36:A5
          inet addr:202.142.26.115  Bcast:202.142.26.255  Mask:255.255.255.0
          inet6 addr: fe80::21e:67ff:fe03:36a5/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:13658 errors:0 dropped:0 overruns:0 frame:0
          TX packets:1735158 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:1053021 (1.0 MiB)  TX bytes:2444504404 (2.2 GiB)
          Interrupt:10 Memory:b1100000-b1120000

This seems quite impossible since this server is not expecting so much traffic in such a short time period.

Can any one help on this please? I'm really suffering from this for a long time. Thanks very much!

hahacc · 04-13-2012, 01:14 AM

Just for guys who may arrive here after searching:
1.there's kernel bug in intel 82574L e1000e driver on centos 6(MSI/MSI-X interrupts issue), we can resolve this by install kmod-e1000e package from ELrepo.org and later add pcie_aspm=off e1000e.IntMode=1,1 e1000e.InterruptThrottleRate=10000,10000 acpi=off to kernel parameters. You can read more info Intel e1000e driver bug on 82574L Ethernet controller causing network blipping.
2.For the high Tx traffic, this was caused by port 53 dns flooding attack. I've resolved this by writing some iptable rules. More info here: port 53 dns flooding attack