[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Hidden download performance problem with Fedora 6-10



On Mon, 22 Dec 2008 13:42:50 -0800
Chuck <ctleis gmail com> wrote:

> Kevin wrote:
> 
> > I wonder if you are running into
> > "82573(V/L/E) TX Unit Hang Messages"
> >
> > see: http://downloadmirror.intel.com/9180/eng/README.txt
> > look for "82573(V/L/E) TX Unit Hang Messages"
> >
> > perhaps the debian kernel doesn't hit the power management issue
> > with that?
> >
> > Can you see if the test listed there shows you are affected?
> >
> 
> I just ran the F10 live CD on 3 other machines.  2 desktops and one 
> wireless laptop.  Fedora worked fine on them all.  One desktop
> had a 3com adapter, the other had a Linksys NIC. One was a slower
> machine ( 1GHz cpu, 90KBps xfer speeds ), the other fast ( 2.8GHz cpu,
> 250KBps xfer speeds )
>
> It is now just this machine with the Intel and National Semi NICs.

ok. Thats good...isolates things some. ;) 
 
> So I powered it down and switched the cables to the 3 enet ports; so
> now the Nat Semi is the one going to the internet.
> I Booted with F10 live CD and tried my wget experiment again.  Same
> old problems: Data error, interrupted downloads.  This box with
> Fedora just sucks.

Interesting. 

> BTW, The Intel connection which is my usual internet connection does 
> have the "suspect byte" set in the "bad" state as per the test above.
> but since the same thing happens with the Nat Semi NIC, this should
> not be the cause.

ok. The intel card isn't being used at all in that case?
Did you remove it? 

> However, and I keep repeating this, OpenSUSE 11, Debian Lenny, Ubuntu
> 8.04, PCLinuxOS 2007. Windows XP, Windows 2000 and ArchLinux all seem
> to handle things just fine. Only flavors of Fedora have a problem
> with this PC.

Right. This points to some strange kernel bug or other issue with the
Fedora kernel. Is there anything in dmesg when it's having the issues? 

> Is there anyone I can get in contact with to analyze this further?
> Run some patched kernel? module?

I would file a bug against the kernel and provide all the info you can
from this thread. It might also be interesting to do a tcpdump of the
failure and look at it against one from one of the non affected oses,
but if thats needed the kernel folks should ask for it in the bug. 

> Thanks
> 
> charlie

kevin

Attachment: signature.asc
Description: PGP signature


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]