[parisc-linux] Interesting network card ...

[parisc-linux] Interesting network card ...

[Nahkola Mikko]

What's the deal with the various HP-branded tulip-based NICs? Are they 
expected to "just work" or what? I found one that sure doesn't, but then 
again, but my kernel isn't actually new ... (2.4.18-32 packaged)

The card is the HP ANA-6911A/AUI (yes, it's a 10/100BT with coax and 
AUI too), part# A3738-60001, and it _is_ found by the tulip driver, but 
for some reason the driver claims that "no MII transceiver found", and 
does list the 10B2 and AUI connections too. But I can't seem to make it 
form a link with TP or coax, no matter what I do. And the AUI connector 
is blocked by the C240 chassis. Is there a simple way to force the use of 
a certain transceiver? Tried mii-tool already and it didn't help much - 
but then again it doesn't seem to know about the coax and AUI at all...

Then again, I haven't tested this card on HP-UX either yet so it could be 
a hardware problem...


-- 
Mikko Nahkola   <mikko.nahkola@nokia.com>
Tre-IN sysadmin <mnahkola@trein.ntc.nokia.com>

Re: [parisc-linux] Interesting network card ...

[Matthew Wilcox]

On Wed, Jan 15, 2003 at 10:56:20AM +0200, Nahkola Mikko wrote:
> What's the deal with the various HP-branded tulip-based NICs? Are they 
> expected to "just work" or what? I found one that sure doesn't, but then 
> again, but my kernel isn't actually new ... (2.4.18-32 packaged)

we're always interested in hearing about cards that don't
work.. hp-branded or not ;-)

> The card is the HP ANA-6911A/AUI (yes, it's a 10/100BT with coax and 
> AUI too), part# A3738-60001, and it _is_ found by the tulip driver, but 
> for some reason the driver claims that "no MII transceiver found", and 
> does list the 10B2 and AUI connections too. But I can't seem to make it 
> form a link with TP or coax, no matter what I do. And the AUI connector 
> is blocked by the C240 chassis. Is there a simple way to force the use of 
> a certain transceiver? Tried mii-tool already and it didn't help much - 
> but then again it doesn't seem to know about the coax and AUI at all...

how very interesting.  could you send an lspci -vvx for that card?

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

Re: [parisc-linux] Interesting network card ...

[Jochen Friedrich]

Hi,

> > The card is the HP ANA-6911A/AUI (yes, it's a 10/100BT with coax and
> > AUI too), part# A3738-60001, and it _is_ found by the tulip driver, but
> > for some reason the driver claims that "no MII transceiver found", and
> > does list the 10B2 and AUI connections too. But I can't seem to make it
> > form a link with TP or coax, no matter what I do. And the AUI connector
> > is blocked by the C240 chassis. Is there a simple way to force the use of
> > a certain transceiver? Tried mii-tool already and it didn't help much -
> > but then again it doesn't seem to know about the coax and AUI at all...
>
> how very interesting.  could you send an lspci -vvx for that card?

This might be a card with an older DC21x4x chipset. Maybe the de4x5.c
driver works better in this case.

--jochen

Re: [parisc-linux] Interesting network card ...

[Grant Grundler]

On Wed, Jan 15, 2003 at 10:56:20AM +0200, Nahkola Mikko wrote:
> What's the deal with the various HP-branded tulip-based NICs? Are they 
> expected to "just work" or what?

yes.

> I found one that sure doesn't, but then 
> again, but my kernel isn't actually new ... (2.4.18-32 packaged)

...a newer kernel would be more interesting.

> The card is the HP ANA-6911A/AUI (yes, it's a 10/100BT with coax and 
> AUI too), part# A3738-60001, and it _is_ found by the tulip driver, but 
> for some reason the driver claims that "no MII transceiver found",

This does sound like a bug (fixed) in the tulip driver relating to
MII initialization. But I'm surprised to see it manifest on a C240.

One should be aware that HP sold 4 different flavors of single port
100BT cards based on tulip chipsets. IIRC, they were:
o PCI "server" 
o GSC (card-mode Dino)
o PCI "workstation"
o PCI V-class only

(and an NIO 100BT also that I don't know anything about).

And each had a different HP-UX driver to go with it (btlan3-6).
Eventually all 4 drivers were merged into one but that's still fairly
recent history.

grant

Re: [parisc-linux] Interesting network card ...

[Nahkola Mikko]

On Wed, Jan 15, 2003 at 07:11:11PM -0700, ext Grant Grundler wrote:
> On Wed, Jan 15, 2003 at 10:56:20AM +0200, Nahkola Mikko wrote:

> > What's the deal with the various HP-branded tulip-based NICs? Are they 
> > expected to "just work" or what?

> yes.

Well, at least then there's hope in trying to get it to work ...

> > I found one that sure doesn't, but then 
> > again, but my kernel isn't actually new ... (2.4.18-32 packaged)

> ...a newer kernel would be more interesting.

Recommendations? For some reason I haven't managed to get a working 
kernel by building from source ... I do still have a 2.4.19-pa24 that I 
did get to boot even if it wasn't all that good. I mean, at least the .18 
has all the basic functionality that I tend to use.

Could be because I still have only the Debian-packaged tools. I haven't 
taken the time to run after the toolchain development ... didn't even get 
a newer modutils package yet.

Any ready-built package that I should try, or sould I just pull a newer 
(which version?) toolchain and compile from sources (which version?) - I 
haven't really paid enough attention to that stuff lately...

> > The card is the HP ANA-6911A/AUI (yes, it's a 10/100BT with coax and 
> > AUI too), part# A3738-60001, and it _is_ found by the tulip driver, but 
> > for some reason the driver claims that "no MII transceiver found",

> This does sound like a bug (fixed) in the tulip driver relating to
> MII initialization. But I'm surprised to see it manifest on a C240.

What do you mean, surprised to see it manifest on a C240? How so?

> One should be aware that HP sold 4 different flavors of single port
> 100BT cards based on tulip chipsets. IIRC, they were:
> o PCI "server" 
> o GSC (card-mode Dino)
> o PCI "workstation"
> o PCI V-class only
> (and an NIO 100BT also that I don't know anything about).

This might be the PCI "server" card. Not sure, partsurfer only knows an 
ANA-6911 without the A and AUI, and not much about that either... and I 
only see the 4-port cards usually. 

BTW, the NIO 100BT is an interesting beast. It (or the driver) is the 
only reason why one of our HP-UX boxes only gets about half an year max 
uptime usually ... and has other "interesting" "features". 

> And each had a different HP-UX driver to go with it (btlan3-6).
> Eventually all 4 drivers were merged into one but that's still fairly
> recent history.

Well, recent enough to predate all the entries that partsurfer showed 
about the ANA-6911. 

And it was still better than the deal with the EISA drivers - I mean, 
duh, I did get (the devices, not drivers) lan0 and btlan0 as different 
adapters once ... 

Anyway, here's some from lspci and dmesg. The adapter in question is the 
first one - it usurped eth0 and still didn't go anywhere - and as usual,
things got interesting, as the box is a NIS and NFS client... 
Notably it is a higher revision than the builtin adapter, and therefore I 
wouldn't expect to need the de4x5 driver.

--------------------
# lspci -vvx
01:01.0 Ethernet controller: Digital Equipment Corporation DECchip 21142/43 (rev 41)
        Subsystem: Hewlett-Packard Company: Unknown device 1067
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV+ VGASnoop- ParErr+ Stepping- SERR+ FastB2B-
        Status: Cap- 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Latency: 255 (5000ns min, 10000ns max), cache line size 08
        Interrupt: pin A routed to IRQ 129
        Region 0: I/O ports at 1fd00 [size=128]
        Region 1: Memory at f2ffc000 (32-bit, non-prefetchable) [size=1K]
        Expansion ROM at f2f80000 [disabled] [size=256K]
00: 11 10 19 00 57 01 80 02 41 00 00 02 08 ff 00 00
10: 01 fd 00 00 00 c0 ff f2 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 3c 10 67 10
30: 00 00 f8 f2 00 00 00 00 00 00 00 00 01 01 14 28

01:13.0 SCSI storage controller: LSI Logic / Symbios Logic (formerly NCR) 53c875 (rev 04)
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV+ VGASnoop- ParErr+ Stepping- SERR+ FastB2B-
        Status: Cap- 66Mhz- UDF- FastB2B- ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Latency: 255 (4250ns min, 16000ns max), cache line size 08
        Interrupt: pin A routed to IRQ 131
        Region 0: I/O ports at 1fe00 [size=256]
        Region 1: Memory at f2ffd000 (32-bit, non-prefetchable) [size=256]
        Region 2: Memory at f2ffe000 (32-bit, non-prefetchable) [size=4K]
00: 00 10 0f 00 57 01 00 02 04 00 00 01 08 ff 00 00
10: 01 fe 00 00 00 d0 ff f2 00 e0 ff f2 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 00 00 00 00 00 00 00 00 03 01 11 40

01:14.0 Ethernet controller: Digital Equipment Corporation DECchip 21142/43 (rev 30)
        Subsystem: Hewlett-Packard Company: Unknown device 104f
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV+ VGASnoop- ParErr+ Stepping- SERR+ FastB2B-
        Status: Cap- 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Latency: 255 (5000ns min, 10000ns max), cache line size 08
        Interrupt: pin A routed to IRQ 128
        Region 0: I/O ports at 1ff00 [size=128]
        Region 1: Memory at f2fff000 (32-bit, non-prefetchable) [size=128]
        Expansion ROM at f2f80000 [disabled] [size=256K]
00: 11 10 19 00 57 01 80 02 30 00 00 02 08 ff 00 00
10: 01 ff 00 00 00 f0 ff f2 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 3c 10 4f 10
30: 00 00 f8 f2 00 00 00 00 00 00 00 00 00 01 14 28

# dmesg |grep -i tulip
Linux Tulip driver version 0.9.15-pre9 (Nov 6, 2001)
tulip0:  EEPROM default media type Autosense.
tulip0:  Index #0 - Media MII (#11) described by a 21142 MII PHY (3) block.
tulip0:  Index #1 - Media 10base2 (#1) described by a 21142 Serial PHY (2) block.
tulip0:  Index #2 - Media AUI (#2) described by a 21142 Serial PHY (2) block.
tulip0: ***WARNING***: No MII transceiver found!
eth0: Digital DS21143 Tulip rev 65 at 0x1fd00, 00:10:83:F7:84:33, IRQ 129.
tulip1: no phy info, aborting mtable build
tulip1:  MII transceiver #1 config 1000 status 782d advertising 0061.
eth1: Digital DS21143 Tulip rev 48 at 0x1ff00, 00:60:B0:D0:1B:9F, IRQ 128.

# uname -mrspv
Linux 2.4.18-32 #1 Sat Mar 30 18:26:22 MST 2002 parisc unknown
------------------


-- 
Mikko Nahkola   <mikko.nahkola@nokia.com>
Tre-IN sysadmin <mnahkola@trein.ntc.nokia.com>

Re: [parisc-linux] Interesting network card ...

[Grant Grundler]

On Thu, Jan 16, 2003 at 11:30:28AM +0200, Nahkola Mikko wrote:
> > ...a newer kernel would be more interesting.
> 
> Recommendations?
...
> Any ready-built package that I should try,
...

Ideally, our CVS kernels: 32-2.4.20-pa21.tar.gz
autobuilt 32-bit kernels are on ftp.parisc-linux.org/kernels/32/ 
I can't say if they work "just fine" on C240 or not though
since I don't test that.

> > This does sound like a bug (fixed) in the tulip driver relating to
> > MII initialization. But I'm surprised to see it manifest on a C240.
> 
> What do you mean, surprised to see it manifest on a C240? How so?

The original problem seemed to be a CPU timing problem where the PHY reset
was not recovering faster than the CPU could pound down the next accesses
to the MII. I've only seen this on problem systems with 550Mhz or
faster CPUs (both parisc and ia64).

The fix is to follow the 802.x spec which describes the MII register
set and how to properly do a MII reset (involves some polling).
Details should be in the parisc-linux mail archive or lkml.

> BTW, the NIO 100BT is an interesting beast. It (or the driver) is the 
> only reason why one of our HP-UX boxes only gets about half an year max 
> uptime usually ... and has other "interesting" "features". 

If you have extras of those, I know someone who wants one.

> Notably it is a higher revision than the builtin adapter, and therefore I 
> wouldn't expect to need the de4x5 driver.

I don't either. I thought we had the tulip cards working just fine.

cheers,
grant

Re: [parisc-linux] Interesting network card ...

[Nahkola Mikko]

On Thu, Jan 16, 2003 at 11:54:27PM -0700, ext Grant Grundler wrote:
> On Thu, Jan 16, 2003 at 11:30:28AM +0200, Nahkola Mikko wrote:

> > > ...a newer kernel would be more interesting.
> > Recommendations? Any ready-built package that I should try,

> Ideally, our CVS kernels: 32-2.4.20-pa21.tar.gz
> autobuilt 32-bit kernels are on ftp.parisc-linux.org/kernels/32/ 
> I can't say if they work "just fine" on C240 or not though
> since I don't test that.

Installed 32-2.4.20-pa20 as the -21 only had the Config, no vmlinux or 
System.map there. 

> > > This does sound like a bug (fixed) in the tulip driver relating to
> > > MII initialization. But I'm surprised to see it manifest on a C240.
> > What do you mean, surprised to see it manifest on a C240? How so?

> The original problem seemed to be a CPU timing problem where the PHY reset
> was not recovering faster than the CPU could pound down the next accesses

> The fix is to follow the 802.x spec which describes the MII register
> set and how to properly do a MII reset (involves some polling).
> Details should be in the parisc-linux mail archive or lkml.

Well, at any rate, it doesn't complain about tulip0/eth0 anymore ... but 
it still doesn't pass any traffic or obey mii-tool. 
"10 Mbit, half duplex, autonegotiation complete, remote fault, link ok" 
... even after "mii-tool -F 100baseTx-HD eth0".

Looks better now anyway ;-/

> > BTW, the NIO 100BT is an interesting beast. It (or the driver) is the 
> > only reason why one of our HP-UX boxes only gets about half an year max 
> > uptime usually ... and has other "interesting" "features". 

> If you have extras of those, I know someone who wants one.

No extras, sorry. They're in real use. (don't tell me that's a bad choice 
- well, if you know of something else to get a H to do 100BT ...)

> > Notably it is a higher revision than the builtin adapter, and therefore I 
> > wouldn't expect to need the de4x5 driver.
> I don't either. I thought we had the tulip cards working just fine.

Well, the builtin adapter isn't doing "just fine" either, but that may be 
firmware-related - when set to 100-HD with mii-tool it goes to 100-FD or 
some such that the hub (yes) doesn't like at all. I suppose I should get 
around to updating the box's firmware. 
 
Oh well, it autonegotiates to 10HD and that works. The other end is a HP 
J3294A (ProCurve 10/100 hub) that obivously can't do anything FD ... and 
yes, I did use other stuff to test the 6911 too. 



And additionally, there's some (non-network-related?) stuff in dmesg that 
I don't remember seeing before, "HP SDC: Transaction add failed: 
transaction already queued?" Is that the HIL stub or something? Should I 
pursue that too?


-- 
Mikko Nahkola   <mikko.nahkola@nokia.com>
Tre-IN sysadmin <mnahkola@trein.ntc.nokia.com>