ethtool failure with tulip 21143

ethtool failure with tulip 21143

[Stefan Jeglinski]

I have a PowerPC 6500 with a Farallon 10/100 PCI card connected to a
Netgear 10/100 switch. I have never gotten this card to work right in
this box. If the number of packets is small, the connection seems to
work, but any increase in the number of packets (my theory) causes
the interface/NIC to choke up. I'm using 2.4.14 (bk).

I downloaded Jeff Garzik's ethtool to work on this, but I'm hitting a
brick wall rather rapidly. As per dmesg:

[root@orion ethtool-1.4]# dmesg
Linux Tulip driver version 0.9.15-pre8 (Oct 11, 2001)
PCI: Enabling device 00:0e.0 (0000 -> 0003)
tulip0:  EEPROM default media type Autosense.
tulip0:  Index #0 - Media 10baseT (#0) described by a 21142 Serial
PHY (2) block.
tulip0:  Index #1 - Media 10baseT-FDX (#4) described by a 21142
Serial PHY (2) block.
tulip0:  Index #2 - Media 100baseTx (#3) described by a 21143 SYM PHY
(4) block.
tulip0:  Index #3 - Media 100baseTx-FDX (#5) described by a 21143 SYM
PHY (4) block.
tulip0:  Index #4 - Media 100baseTx (#3) described by a 21143 reset
method (5) block.
eth0: Digital DS21143 Tulip rev 65 at 0xc81f2000, 00:00:C5:53:3E:3D, IRQ 25.


I can get some basic information with the -i switch:

[root@orion ethtool-1.4]# ethtool-i eth0
driver: tulip
version: 0.9.15-pre8
firmware-version:
bus-info: 00:0e.0


But any attempts to get (much less set) anything else fail in this fashion:

[root@orion ethtool-1.4]# ethtool eth0
Settings for eth0:
No data available

Trying the register dump yields:

[root@orion ethtool-1.4]# ethtool -d eth0
Cannot get register dump: Operation not supported

I've tried this in a completely different box (9500) with the same
card, and get the same result. Does ethtool not work with the 21143
Tulip?


Stefan Jeglinski

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Jeff Garzik]

Stefan Jeglinski wrote:
> I've tried this in a completely different box (9500) with the same
> card, and get the same result. Does ethtool not work with the 21143
> Tulip?

correct, it does not support anything but basic driver info yet

--
Jeff Garzik      | Only so many songs can be sung
Building 1024    | with two lips, two lungs, and one tongue.
MandrakeSoft     |         - nomeansno

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Stefan Jeglinski]

Jeff Garzik wrote:

>correct, it does not support anything but basic driver info yet

oops. Sorry, I didn't know the tool only did this so far. I read all
the dox I could find, but somehow I missed this point?

Benjamin Herrenschmidt wrote:

>Jeff, afaik, this one has the broken bridge I told you about.

How "broken" is it? With the card in the 6500, the tulip driver sort
of works, sometimes. The only error message I get out of it ever is

Internal fault: The skbuff addresses do not match in tulip_rx [plus hex output]

but this message doesn't always seem to correspond with the
driver/interface/NIC failing. And on the 9500, the card seems to work
significantly better than on the 6500 (but we already know the 6500
seems to be a pathological beast in places). For example, I have
never seen the skbuff message on the 9500, and AFAICT, the card just
works.

Actually your comment is an interesting one. When you are saying
"bridge" are you talking about something related to the ethernet
function or the PCI function? Over time on the 6500, I have been
wondering if this card's problems are related to PCI rather than
ethernet.


>Stephan, I'll send you a hack to try out with de4x5 later today.

woohoo! I wait with excitement :-) But I have a Christmas party to go
to so I might not get back to you for a day :-(


Stefan Jeglinski

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Benjamin Herrenschmidt]

>
>Jeff Garzik wrote:
>
>>correct, it does not support anything but basic driver info yet
>
>oops. Sorry, I didn't know the tool only did this so far. I read all
>the dox I could find, but somehow I missed this point?
>
>Benjamin Herrenschmidt wrote:
>
>>Jeff, afaik, this one has the broken bridge I told you about.
>
>How "broken" is it? With the card in the 6500, the tulip driver sort
>of works, sometimes. The only error message I get out of it ever is
>
>Internal fault: The skbuff addresses do not match in tulip_rx [plus hex
>output]

Which match my idea about broken cache coherency on 6400/6500 machines
and possibly other derivatives (5x00 ?).

>but this message doesn't always seem to correspond with the
>driver/interface/NIC failing. And on the 9500, the card seems to work
>significantly better than on the 6500 (but we already know the 6500
>seems to be a pathological beast in places). For example, I have
>never seen the skbuff message on the 9500, and AFAICT, the card just
>works.

Yes, cache coherency in the 9500 is much better, fortunately ;)

>Actually your comment is an interesting one. When you are saying
>"bridge" are you talking about something related to the ethernet
>function or the PCI function? Over time on the 6500, I have been
>wondering if this card's problems are related to PCI rather than
>ethernet.

I suspect the cache coherency isn't properly maintained by the
PCI host bridge, which would break drivers (or cause memory corruptions)
when DMA happens. It's not completely incoherent however, but you'd
rather avoid having cache lines shared.

de4x5 has a tweak to align descriptors so that only one exist per
cache line. This solved the problem for some 6400 users, but
that tweak must be enabled by hacking the driver a bit.

Try defining CACHE_ALIGN to CAL_32LONG, and DESC_ALIGN to u32 dummy[4]
(as in the commented out example).

If this also solve your problem, then we'll need Jeff to add a similar
tweak to tulip, possibly as a config option.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Stefan Jeglinski]

>  >How "broken" is it? With the card in the 6500, the tulip driver sort
>>of works, sometimes. The only error message I get out of it ever is
>>
>>Internal fault: The skbuff addresses do not match in tulip_rx [plus hex
>>output]
>
>Which match my idea about broken cache coherency on 6400/6500 machines
>and possibly other derivatives (5x00 ?).

A long time ago (2.2.13 or so), I could not use another card on the
6500 for very long with the de4x5 driver as well. At that time, Paul
provided a patch for the de4x5 driver which fixed the problem. He
also described the issue as a cache coherency problem.

>de4x5 has a tweak to align descriptors so that only one exist per
>cache line. This solved the problem for some 6400 users, but
>that tweak must be enabled by hacking the driver a bit.

Well, I wonder if that was the fix that Paul put in. Some time ago
(early 2.4) I tried the de4x5 driver with this Farallon 10/100 card
but it worked even more poorly than the tulip driver of that time did.


>Try defining CACHE_ALIGN to CAL_32LONG, and DESC_ALIGN to u32 dummy[4]
>(as in the commented out example).

I will look at the de4x5 code again and try this. It's all vaguely
familiar to me... Thanks much!


Stefan Jeglinski


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Benjamin Herrenschmidt]

>A long time ago (2.2.13 or so), I could not use another card on the
>6500 for very long with the de4x5 driver as well. At that time, Paul
>provided a patch for the de4x5 driver which fixed the problem. He
>also described the issue as a cache coherency problem.

Yes, paul's patch is what I'm talking about.

>Well, I wonder if that was the fix that Paul put in. Some time ago
>(early 2.4) I tried the de4x5 driver with this Farallon 10/100 card
>but it worked even more poorly than the tulip driver of that time did.

Yup.
>
>>Try defining CACHE_ALIGN to CAL_32LONG, and DESC_ALIGN to u32 dummy[4]
>>(as in the commented out example).
>
>I will look at the de4x5 code again and try this. It's all vaguely
>familiar to me... Thanks much!

That should be the same patch paul did.

Ben.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Stefan Jeglinski]

>  >>Try defining CACHE_ALIGN to CAL_32LONG, and DESC_ALIGN to u32 dummy[4]
>>>(as in the commented out example).
>>
>>I will look at the de4x5 code again and try this. It's all vaguely
>  >familiar to me... Thanks much!

Yes, it has painfully all come back to me now. I originally switched
to the tulip driver because the de4x5 driver did not function with
the Farallon 10/100 PCI NIC, and I have now reconfirmed this. There
are no error messages when loading the module, but there is some
problem I can't diagnose. The switch lights indicating "it's alive"
never come on, and the network is not ever reachable. With the tulip
driver I can at least get that.

I remember further contacting the de4x5 maintainer about this, and
his response with apologies was that the driver is no longer being
maintained.

I remember even further contacting Donald Becker about somehow
inserting this cache alignment fix into the tulip driver. No
response. So, any thoughts on how to proceed? I have no idea how to
get the fix into tulip to test...


Stefan Jeglinski

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Kevin B. Hendricks]

Hi Stefan,

Try sending a patch to Jeff Garzik <jgarzik@mandrakesoft.com> the tulip
maintainer that implements this fix.  I have had good luck getting him to
incorporate patches for the PNIC_II driver so if you send him a tested
patch (against the latest 2.4.x kernel), I am sure he will consider it.

Kevin



On December 26, 2001 11:13, Stefan Jeglinski wrote:
> >  >>Try defining CACHE_ALIGN to CAL_32LONG, and DESC_ALIGN to u32
> >  >> dummy[4]
> >>>
> >>>(as in the commented out example).
> >>
> >>I will look at the de4x5 code again and try this. It's all vaguely
> >>
> >  >familiar to me... Thanks much!
>
> Yes, it has painfully all come back to me now. I originally switched
> to the tulip driver because the de4x5 driver did not function with
> the Farallon 10/100 PCI NIC, and I have now reconfirmed this. There
> are no error messages when loading the module, but there is some
> problem I can't diagnose. The switch lights indicating "it's alive"
> never come on, and the network is not ever reachable. With the tulip
> driver I can at least get that.
>
> I remember further contacting the de4x5 maintainer about this, and
> his response with apologies was that the driver is no longer being
> maintained.
>
> I remember even further contacting Donald Becker about somehow
> inserting this cache alignment fix into the tulip driver. No
> response. So, any thoughts on how to proceed? I have no idea how to
> get the fix into tulip to test...
>
>
> Stefan Jeglinski
>

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Benjamin Herrenschmidt]

>I remember even further contacting Donald Becker about somehow
>inserting this cache alignment fix into the tulip driver. No
>response. So, any thoughts on how to proceed? I have no idea how to
>get the fix into tulip to test...
>

I'll try to hack you something in tulip to test.

Ben.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Benjamin Herrenschmidt]

>
>I have a PowerPC 6500 with a Farallon 10/100 PCI card connected to a
>Netgear 10/100 switch. I have never gotten this card to work right in
>this box. If the number of packets is small, the connection seems to
>work, but any increase in the number of packets (my theory) causes
>the interface/NIC to choke up. I'm using 2.4.14 (bk).
>
>I downloaded Jeff Garzik's ethtool to work on this, but I'm hitting a
>brick wall rather rapidly. As per dmesg:

Jeff, afaik, this one has the broken bridge I told you about.


Stephan, I'll send you a hack to try out with de4x5 later today.

Ben.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ethtool failure with tulip 21143

[Tony 'Nicoya' Mantler]

At 11:53 AM -0600 12/16/01, Stefan Jeglinski wrote:
>I have a PowerPC 6500 with a Farallon 10/100 PCI card connected to a
>Netgear 10/100 switch. I have never gotten this card to work right in
>this box. If the number of packets is small, the connection seems to
>work, but any increase in the number of packets (my theory) causes
>the interface/NIC to choke up. I'm using 2.4.14 (bk).
[...]

I had the same problem with the Farallon 10/100 card in my 9600. It seemed
that either the driver or the chip weren't talking FDX properly - the card
worked flawlessly at 100/HDX.

I never figured out how to twiddle the duplex setting in the driver itself,
I usually ended up unplugging and replugging the ethernet cable a few times
until I confused the nway into thinking half-duplex is a good idea.

This was using the 'tulip' driver, iirc. I never had much luck with the
de4x5 driver. I haven't booted linux on my 9600 in a while, though, so
things might have improved since then.


Cheers - Tony 'Nicoya' Mantler :)


--
Tony "Nicoya" Mantler - Renaissance Nerd Extraordinaire - nicoya@apia.dhs.org
Winnipeg, Manitoba, Canada           --           http://nicoya.feline.pp.se/


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/