OS 12.x: probs with NIC 82579/ulimit

English Hardware
1

Hi folks,

i got some probs with this NIC in different versions.

I did a minimal installation 64bit with the patterns
file_server, print_server, devel_C_C++, 32bit and some yast-modules.

Due to the message


rlimit_max: rlimit_max (1024) below minimum Windows limit (16384)

i set


* - nofile 16384

in /etc/security/limits.conf

I setup a minimal workgroup-setup in samba.

Then i copied some files from WINDOWS XP to the share
and i saw some errors occur in the output of ifconfig
on the interface.

I studied the versions of opensuse 11.4, 12.1, 12.2 and found out
that this behavior is only in 12.x.

Removing the entry in /etc/security/limity.conf the errors didn’t appear,
but the speed / throughput was much slower.

Here are the versions:
OS kernel samba intel kernel driver
11.4 2.6.37 3.6.3 1.2.7-k2
12.1 3.1.10 3.6.3 1.4.4-k
12.2 3.4.33 3.6.7 1.9.5-k
(no, i didn’t try 12.3, it seems to unstable for me)

I compiled the latest driver from intel 2.3.2, but it didn’t help really.

Board: Supermicro C7B75
BIOS 1.0a
Intel Ethernet 82579V onboard
kernel driver is e1000e

Searching with google for ‘linux 82579V error’ showed me,
that many distributions have probs with this NIC on different
boards from different vendors.

Is this a hardware related problem or a kernel-driver problem?

Using a RTL8169 NIC i don’t have any errors.

Best regards

2

additional information:

all kernels are kernel-default 64bit.
11.4 has the same Problem.

USB3 is blacklisted via
/etc/modprobe.d/50-blacklist.conf
blacklist xhci_hcd

SAMBA is running with
socket options = SO_KEEPALIVE TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192

Removing the entry

    • nofile 16384
      in /etc/security/limits.conf and setting SAMBA to
      socket options = SO_KEEPALIVE TCP_NODELAY
      i get more speed in iptraf and less errors on ifconfig,
      but still errors.
      and the SAMBA message again:
      rlimit_max: rlimit_max (1024) below minimum Windows limit (16384)

The main goal is oracle express edtion 11 and the second is samba.

Best regards

3

Further additional information:

If i use


ethtool -S eth0 |grep -i error

i get the result:
rx_errors: 92
tx_errors: 0
rx_length_errors: 0
rx_over_errors: 0
rx_crc_errors: 46
rx_frame_errors: 0
rx_missed_errors: 0
tx_aborted_errors: 0
tx_carrier_errors: 0
tx_fifo_errors: 0
tx_heartbeat_errors: 0
tx_window_errors: 0
rx_long_length_errors: 0
rx_short_length_errors: 0
rx_align_errors: 0
rx_csum_offload_errors: 0

I have 4 identical machines with identical instalation, but different values.
Today i used for SAMBA:
socket options = SO_KEEPALIVE TCP_NODELAY
and for /etc/security/limits.conf:

    • nofile 16384
      One machine had only 2 errors (only crc), but a another had
      rx_errors: 200
      rx_crc_errors: 100

I#m not the kind of guy who’s fit up with wireshark, so i have
no reults from that.

Best regards.

4

So i tried to set the MTU-Size to 1500 in /etc/sysconfig/network/ifcfg-eth0
I got:
1 machine with errors 0
1 machine with errors 38
1 machine with errors 124

Errors are less, but greater then 0 is bad.

This sounds like a Intel-NIC problem.

best regards