netdev - RE: e1000

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <F169D4F5E1F1974DBFAFABF47F60C10A17673ABC@orsmsx507.amr.corp.intel.com>
Date:	Wed, 17 Dec 2008 12:14:53 -0800
From:	"Brandeburg, Jesse" <jesse.brandeburg@...el.com>
To:	"nik-lists@...na.net" <nik-lists@...na.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>
CC:	"e1000-devel@...ts.sourceforge.net" 
	<e1000-devel@...ts.sourceforge.net>
Subject: RE: e1000 

First, I included the maintainer's list, it is more appropriate since
you're running an out of tree driver.

nik-lists@...na.net wrote:
> Hi all,
> 
> I am experiencing a big amount of tx_deferred_ok errors with
> e1000-8.0.6-NAPI. The error rate varies from 1 to 50 per second, this
> interface is loaded no more that 15 kpps and 100 Mbit/s. What is the
> typical reason for this type of errors?

Are you at 100Mb half duplex?  tx_deferred_ok happens when the
transmitter in the card is waiting for the wire to become available.  I
don't necessarily expect it at 100Mb, but I would if you were in half
duplex mode, or if flow control was enabled and you were getting a lot
of rx_flow_control_off packets

> I am getting about the same amount of tx_restart_queue errors on the
> same interface. My suggestion is that I need to increase the tx ring
> size. Am I right?

increasing the tx ring size will help some.  If your cpu/application is
pushing more data than can be completed out the transmit portion of the
hardware (for multiple possible different reasons, including link speed,
collisions, flow control, bus bandwidth, etc) then the hardware's
transmit ring will fill.

You could also try increasing the interrupt throttle rate (modprobe
e1000 InterruptThrottleRate=20000,20000)

since you're linked at 100 Mb?

> Thank you very much in advance.

if you don't solve your problem with the above info, please include
output of /proc/interrupts, lspci -vvv, ethtool -S ethX output, and a
full dmesg from boot in the next mail.

Download attachment "smime.p7s" of type "application/x-pkcs7-signature" (6703 bytes)