In article <linux.kernel.Pine.LNX.3.95.990324233439.5546A-100000@catbert.desm.plconline.com>,
Dave Weis <djweis@plconline.com> wrote:
>
>> Out of curiosity, why do you refer to the Tulip clones as
>> `abominations'? I'm running a few of them in high-performance
>
>The spontaneous reboots and/or lack of functionality in via chipset based
>machines made me call them much worse than that.
Now I'm really curious, because the machines I'm putting them
are pretty close to exclusively VIA-based, and the only downtime
the any of them have had (in the past 90 days, which is when I
got the 8-pack and started installing the chips) have been when
nfsd jammed up talking to a AIX machine and I had to reboot the
file server to get socket 2049 out of CLOSE_WAIT.
We are talking about the same tulips and VIA chipsets, I hope?
The Netgear cards I've got have one nearly tulip sized chip
on them titled:
NETGEAR
NGMC1698
9843
AN6258.1
which is described in hinv (on the server, which I'll describe
exclusively, because it's been running longest and has the most
to lose if the card turns out to be icky) as
eth0: Lite-On 82c168 PNIC
eth0: MII transceiver found
And the VIA chipset I've got on the motherboards (FIC PA2013)
is Vendor id 1106, Device id 586, which dmesg reports as
ide: VT82C586 VXPRO+ (Apollo) on PCI bus 0 function 57
ide0: BM-DMA at 0x6400-0x6407
ide1: BM-DMA at 0x6408-0x640f
hda: IBM-DTTA-351350, 8063MB w/464kB Cache, CHS=1644/255/63, DMA
hdc: IBM-DTTA-351350, 8063MB w/464kB Cache, CHS=16383/16/63, DMA
Of course, I'm not running 2.2.x on this machine; I'm running a much
hacked 2.0.28, with the startup logo patch, the Webshield "this is Unix,
so yes it's okay to have network modules autoprobe" patch, one of the
VIA chipset DMA patches, my new.new.memory patch, a patch so that the
IDE driver doesn't screw up disk geometry by trying to outguess the
IBM drives, and a large crop of additional and backported 2.0.3x and
2.2.x drivers (including the netgear-tweaked tulip.c)
The hardware on this machine is pretty simple:
400mb core
AMD K6/333
the netgear card
FIC PA-2013 mb (1mb cache)
BusLogic 958, driving a Sony SDT9000 and Exabyte 8505
2 IBM 13.5gb IDE drives
some throwaway Intel 740 AGP video card (the machine runs
in text mode, using SVGATextMode to get a satisfying 20x9
character cell.)
And the network topology is:
[this box] ---100mb/full dup--- [netgear switch] -- firewall --- dmz
|||||| ||||||
Wad of Wad of NT machines,
Unix talking SAMBA.
machines
(SGI, HP, IBM, DEC)
talking nfs to
this box, because
/home is mounted
from here.
With this box providing
home directories
mail
the public ftp site
the public web site
dns for the internal network
So, as you can see, I'm rather surprised to hear that this particular
combination of hardware is BAD, given that it's not yet bitten me in
the ass.
____
david parsons \bi/ And since Netgear is shipping Linux drivers with their
\/ hardware, guess who's getting all my business for
ethernet cards? It would be distressing if all this
suddenly failed.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/