Linksys PNIC II glitches
Robin Whittle
rw@firstpr.com.au
Fri Aug 13 11:34:37 1999
Short version:
1 - The Linksys LNE100TX NICs I recently purchased with the
*excellent* super-cost-effective EZXS55W 5 port 10/100 Meg switch
are more trouble that they are worth for me, due to difficulties
with drivers under Linux, and a few other intermittent problems
which have been reported, and observed.
2 - Positive appraisal and defanning of the EZXS55W.
- - - -
Dear Donald and Linksys people,
(Also to the Tulip driver development list.)
Thanks, Donald, for your Ethernet drivers for Linux!
I am writing to report that your currently valid driver version (v0.91
4/14/99) at
http://cesdis.gsfc.nasa.gov/linux/drivers/tulip.c
does not work with my particular Linksys NICs, but that the version
which comes on a floppy from Linksys does.
"v0.90f 12/17/98 Originally written by
becker@cesdis.gsfc.nasa.gov
Driver modified for Linksys LNE100TX Fast Ethernet
PCI Adapter on 01/13/99"
Here is a description of the NIC.
PCB says:
Part No.: LNE100TX
Version 2.0
P/N:6804057404 REV:A2
(Date code on foil side of PCB: "9816".)
LSI, 128 pins:
LINKSYS
LNE100TX
LC82C115
C9914
T4023702
37BDX
This card has a "WOL" Wake On LAN connector and a 6 pin jumper to
configure it.
These NICs works fine with Windows 98.
They are not recognised at all by the Red Hat 6.0 installation
procedure. (This is a pest even if the card is subsequently made to
work, since I find it easier to install via LAN and set up all the
networking details during installation, rather than later.)
Once RH6.0 is installed, attempts to use the normal tulip.o driver
fail, with an error message pointing out that the card ID of:
11ad c115
is not recognised.
Therefore, I looked up the Ethernet HowTo, which took me to the
Linksys page:
http://www.linksys.com/support/solution/nos/linux_lne100tx.htm
which directed me to use the latest available driver at:
http://cesdis.gsfc.nasa.gov/linux/drivers/tulip.c
(v0.91 4/14/99) which I compiled and installed. The driver was happy
with the NIC, but it would not in fact send or receive any packets to
the LAN (actually to a magnificent Linksys 100 Mbps 5 port *switch*).
(BTW, the Linksys page says to "depmod -a", which I did, but this is
not mentioned on the development page.)
I tried every other possible configuration thing, and still it would
not work.
When I compiled the driver from the floppy, the worked fine. (Althoug
I have not extensively tested it.)
Clearly the "v0.91 4/14/99" driver is supposed to work with this chip,
since:
1 - The source mentions its PCI ID and has special provisions for it.
2 - The Linksys site indicates that this is the driver code to use.
3 - The driver recognises the chip and installs itself happily, with
the message:
"Lite-On PNIC-II rev 37 at . . . . IRQ 9"
4 - TCP-DUMP reports activity of packets sent to this interface.
However, it simply doesn't work! (I have not tried the "testing"
version of the driver, currently "v0.91g".)
Here are some suggestions for Linksys to improve matters:
1 - Correct the page:
http://www.linksys.com/support/solution/nos/linux_lne100tx.htm
which states that for RH5.2 and 6.0 the standard drivers will
work. (Below that, there is a qualification, for if they don't!)
Also, it states that the current driver from the Tulip team
should work, which it doesn't - for these particular cards.
2 - Make the Floppy 2 /linux/ directory more helpful than simply the
tulip.c file and the GPL copying file. Point the user to the
web support page, and the Tulip development site. Give explicit
compilation and installation instructions in a readme.txt file.
Other concerns
--------------
I note in the Errata of:
http://cesdis.gsfc.nasa.gov/linux/drivers/tulip.html
"Due inaccurate documentation, autonegotiation cannot be enabled on
the PNIC chip."
This sounds like a good reason to use another kind of NIC!
This NIC card in my Windows machine occasionally (say every ten or
twenty minutes - or more often with lots of activity?) appears to die
for a second or so. The three LEDs for this card on the 5 port switch
go off together, and then come back on together.
According to the mailing list archives, a number of incorrect
behaviours have been observed with some PNIC based cards. I am not
sure if these are exactly the same cards as I have, but I would prefer
to avoid using cards which occasionally send invalid data to the LAN.
Whether these problems are caused by faulty driver design or not, I
cannot tell.
The point is, life is too way short to be spending time debugging
other people's hardware and software when all you want is a handfull
of bulletproof NICs! I can buy tried-and-tested Intel NICs (S/H
82557s for USD$33) and not have to worry about any of this! I only
got these Linksys NICs as part of the switch package, and I don't plan
to buy any more, or use them except perhaps on Windows machines . . .
but even then, I want real-time audio through the LAN, so I can't have
NICs going to sleep for a moment.
The Tulip Development Page has details of joining the mailing list.
Here are some URLs of recent postings in the archive pointing to
anomalous behaviour of PNIC cards (not necessarily the same type as
mine):
Problem with Linksys etherfast 10/100
http://www.tux.org/hypermail/linux-tulip-bug/1999-Jul/0014.html
Linksys PNIC card hangs up under load (v. 0.91g)
http://www.tux.org/hypermail/linux-tulip-bug/1999-Jul/0012.html
Odd bug when using tulip v0.91g on a LinkSys EtherFast 10/100 card
http://www.tux.org/hypermail/linux-tulip-bug/1999-Aug/0003.html
" . . . Whenever I send data from this card, it randomly (as far as
I can tell) cuts 2 characters within the file and moves them
forward a character before replacing them. . . . "
(Seriously low-level user debugging!)
- - - -
I have put the tulip.c from the floppy, this email, and any other
related files at:
http://gair.firstpr.com.au/temp/tulip/
- - - -
All hail the Linksys EZXS55W 5 port 10/100 Ethernet Switch!
-----------------------------------------------------------
Finally, lest someone think I have something against Linksys, let me
say that I think their new 5 port 10/100 Ethernet EZXS55W *Switch* is
a beauty! The cost here, ex-tax, with two NICs and two cables as a kit
was AUD$330, which is USD$215. If you look around, such as:
http://www.1stcomputersource.com/cgi-win/detail.exe?443980
the switch itself can be purchased for USD$124.
It is small, and has a fan - which is a source of noise and a dust and
long-term maintenance issue. (The fan is a 40mm sleeve bearing
Sunon.)
I modified my EZXS55W for fanless operation by removing the rear
panel, replaceing the top panel with a metal grid, adding a large
heatsink to the heatsink on the main LSI (I found one which
interdigitated nicely, used silicone grease, and cut some sheet
aluminium to slot into the side-slots of the case and work as a spring
to hold it in place) and adding a heatsink to the second largest LSI
(Allayer AL102).
As far as I know, this device is a faultless full-duplex 5 port 10/100
meg switch. Apart from perhaps its use of a fan, I think it is
beautifully made - and its cost is extremely attractive!
- Robin
===============================================================
Robin Whittle rw@firstpr.com.au http://www.firstpr.com.au
Heidelberg Heights, Melbourne, Australia
First Principles Research and expression: Consulting and
technical writing. Music. Internet music
marketing. Telecommunications. Consumer
advocacy in telecommunications, especially
privacy. M-F relationships. Kinetic sculpture.
Real World Electronics and software for music including:
Interfaces Devil Fish mods for the TB-303, Akai sampler
memory and Csound synthesis software.
===============================================================