[Top] [All Lists]

Re: Acer Aspire 1524WLMi and RealTek 8169 - very slow

To: Francois Romieu <romieu@xxxxxxxxxxxxx>
Subject: Re: Acer Aspire 1524WLMi and RealTek 8169 - very slow
From: Richard Dawe <rich@xxxxxxxxxxxxxxxxxxxx>
Date: Tue, 08 Feb 2005 23:00:59 +0000
Cc: netdev@xxxxxxxxxxx, Me <rich@xxxxxxxxxxxxxxxxxxxx>
In-reply-to: <20050205204135.GA13986@xxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <41A24F35.5080106@xxxxxxxxxxxxxxxxxxxx> <20041122213008.GA9618@xxxxxxxxxxxxxxxxxxxxxxxxxx> <41D2844E.5070204@xxxxxxxxxxxxxxxxxxxx> <20041229235203.GA5465@xxxxxxxxxxxxxxxxxxxxxxxxxx> <41F250D1.8000207@xxxxxxxxxxxxxxxxxxxx> <41F26FD1.2060407@xxxxxxxxxxxxxxxxxxxx> <20050122230156.GC24461@xxxxxxxxxxxxxxxxxxxxxxxxxx> <41F3F632.3060800@xxxxxxxxxxxxxxxxxxxx> <20050125214725.GA6093@xxxxxxxxxxxxxxxxxxxxxxxxxx> <4204C981.6050102@xxxxxxxxxxxxxxxxxxxx> <20050205204135.GA13986@xxxxxxxxxxxxxxxxxxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.3) Gecko/20041020

Francois Romieu wrote:
If you can do an extra test, I'd like to know if you can safely bring
the r8169 interface down on your computer once the patch below if applied:

(it should apply correctly on top of you current setup)

I can, with and without that patch.

But I encounter problems bringing the interface back up and down again. If I do the following after a reboot:

  /etc/init.d/network stop
  /etc/init.d/network start
  /etc/init.d/network stop

I find that the "ip link set dev eth0 down" command hangs, consuming most of the CPU time (~99%). Note that I issued the "network" commands with about a second's delay between each command.

This is with a static IP address. If the IP address is assigned via DHCP, then the DHCP client hangs in the "network start" command.

I'm running Fedora Core 3 + all the latest security updates.

One more thing:

With 1.6LK + my PHY patch, I see the message "eth0: PHY reset until link up" every 5 seconds or so, when there is no Ethernet cable plugged in. This is annoying. I think it should only log it once.

Hmmm... One could argue that 1) syslog will cope and 2) the printk level
of the console can be lowered. Some event kamasutra can probably be done
from userspace too so I am a bit reluctant to special case this code
(read as: I'd welcome more requests/complaints).

Is it enough for you if I salt the code with a few netif_msg_xxx() here
and there to control the messages ?

That sounds good.

I'm happy to submit a patch to do this, if you'd like. Note that I'm new to kernel programming, so it might take me a while.

Bye, Rich =]

Richard Dawe [ ]

"You can't evaluate a man by logic alone."
  -- McCoy, "I, Mudd", Star Trek

<Prev in Thread] Current Thread [Next in Thread>