A3740A PCI config space

[Steve Jones_4]

I'm trying to figure out why an A3740A FC PCI adapter is not recognised by a HP9000 server running HP-UX 11.0. This is a supported config, and I have all the requisite drivers in the kernel, but 11.0 refuses to play ball.

ODE MAPPER sees the card and reports it as a 'fibre channel adaptor'. If I boot HP-UX 11i, I get a boot time warning on the console which states that the 'A3740A is not supported by HP-UX 11i ....". This leads me to believe that the card is functional.

If anyone has a working A3740A configuration, would you be so kind as to post the output from either ODE MAPPER, or the STM information relating to this adapter? I'm particularly interested in the PCI config space info (Vendor ID, Device ID, etc.).

Hoping you can help ...

Steve

[Steve Jones_4]

in addition to mapper/stm, fcmsutil will give me most of what I want;

fcmsutil /dev/fcmsN

thanks

Steve

[Florian Heigl (new acc)]

hi steve,

we don't have this card in any system here, but I'm a bit confused about where You get the problem - do I get it right that both stm and fcmsutil see the card? this would mean that the driver is correctly bound to the device and would - to my eyes - only leave a patch problem or a configuration issue. (If I remember correctly this adapter wouldn't work in a fabric, for example)

If You want, please add some information.

[melvyn burnard]

so can you post your 11.00 system file?
What Server is it installed in?

[Steve Jones_4]

Hi Florian

Neither fcmsutil nor stm can see the card in my server; that's my problem. I was hoping to compare ODE MAPPER output from a working system with that which I see on my server. ODE MAPPER on MY server does report the card, but no driver (fcT1) binds to it when 11.0 boots.

As it happens, I just found a production server on a client site which has a similar config to mine (A3740A in an L-class box running 11.0 [mine's an A-class]). The L-class setup is working and allows me to confirm, by comparison, that I have all the requisite fcT1 and fcp drivers configured in my system. It also allows me to see that neither 'fcmsutil' nor 'stm' produces any PCI ID info when run against the A3740A. So, I think only ODE MAPPER output would be of any use. I'll try it on this L-class when I next do some maint on it.


Melvyn

The server is an A180c. I don't have the system file immediately to hand, but I can post it later.

Thanks for your input guys.

Steve

[Florian Heigl (new acc)]

I've got an A180 here at and will get two A3470 for it in a week or two.
I was kind of a stupid idea as I'm running 11.11 on it, but if noone can provide You with the details until then I at least found *some* use for the adapter.

One more thing - if You do an ioscan -fn do the cards show up (as UNCLAIMED I suppose)?
If You don't see them there, something else is wrong, maybe downlevel firmware.

PF_CSTD3932 should be current, see openpa.net on that. (I haven't upgraded either, so maybe I can reproduce the problem. *g*)

[Florian Heigl (new acc)]

I just pray noone will mention here!

[rick jones]

well, given that the manual says nothing about installing on an A Class of any stripe... it does not bode well for it being officially supported in an A180. Of course the manual might be wrong... got to the manual with a serach for the product number on www.hp.com

as for getting the vendor ids et al, might be documented on agilent's website - IIRC it is an old Agilent Tachyon (not tachlite) chip. i suspect the subvendor ids are HP-specific.

if all else fails you could place the card in a system running linux - I think that lspci would give the id's

[melvyn burnard]

Well sorry to say this, but Yes, it is unsupported in the A180 server.
It IS supported in the A400/A500 servers.

[Steve Jones_4]

All

According to http://docs.hp.com/en/J6374-90014/ch02s04.html A180/A3740A with 11.0 _is_ a supported config (albeit without FC boot support, which doesn't bother me).

According to http://docs.hp.com/en/FC_SM_20050107/FC_SM_20050107.htm A180/A3740A with 11.0 _is_ a supported config (again, boot support is spcifically excluded).

Which doc says A180 isn't supported? And if it isn't supported, why do the above HP docs say it is?

In any case, I'm not looking for 'official' support; I'm just curious as to why this stuff isn't working.

Florian

ioscan says:
unknown -1 8/0/1/0 UNCLAIMED UNKNOWN PCI(107e0004) -- 1

Now, 107e0004 is the PCI VID/DID for the card (this is confirmed by ODE MAPPER). 107e is the VID for Interphase, who manufactures a Tachyon adapter (model PowerSAN 5226) which has a more than passing resemblance to the A3740A.

What I wanted was the VID/DID (and maybe the SVID/SDID) for a working A3740A (working, as in recognised by HP-UX) so that I could see if _my_ card matches.

Thanks for the input.


Steve

[melvyn burnard]

Hmm, an internal document I found lisy=ts it as unsupported.
Ok, so again, what does your /stand/system file look like for this kernel and card, as it seems a driver is missing?

[rick jones]

Um, are you using a _real_ A3740A, or some other card that happens to have a Tachyon chip on it?

HP-UX I/O drivers check for _HP_ IDs and will only attach to _HP_ cards. They will not attach to cards with other IDs (subvendor IDs I believe).

So, if you have a pseudo-random Tachyon-based card other than an A3740A, it would be consistent with a) the driver being present in the kernel and b) the card remaining unclaimed.

[Florian Heigl (new acc)]

rick: it seems he has bought an A3740A that thinks it is something different, i.e. got mixed with the interphase line in the fab. personality disorder might also be a reason :)

that's why I just love 3rd party adapters like JNI. I can put them in a windows pc, into my HP L1000, into my RS/6000 7026, into a Sun x86, into a Sun Ultra, ...

steve:
I'll post back on this in two weeks or so when I got the cards.

[Steve Jones_14]

Rick

The label on the bracket says A3740A.

Florian

Thanks for taking the trouble with this.

Here's what MAPPER says on my system:

Type HW SW Revisions
Path Component Name ID Mod Mod Hdwr Firm
---------- ----------------------------------- ----- ----- ----- ----- -----
8 Phantom PseudoBC GSC+ Port 7H 504H 0H 0 0
8/0 Host PCI Bridge DH 680H AH 3 0
8/0/2:0 Fibre Channel Controller 107EH 0004H 107EH 0004H 0CH
8/0/20:0 Ethernet Controller 1011H 0019H 103CH 104FH 02H
8/4 Upper Bus Converter 7H 500H CH 0 0
8/4/0 HP-PB Mux 5H 10H DH 0 0

The A3740A (at 8/0/2) has the Interphase strings in the VID(107Eh)/DID fields AND in the SVID/SDID. The core Ethernet adapter (at 8/0/20) has a DEC VID(1011h)/DID, but a HP SVID(103Ch)/SDID. I suspect that my A3740A should hava a HP SVID/SDID.

The following drivers are in my system file:

:
fc
fcT1
fcT1_cntl
fcT1_fcp
fcms
fcp
fcp_cdio
fcparray
fcpdev
fcpmux
:

Thanks for the continued interest.

Steve

[rick jones]

Didn't you say yours was an A180?

Something does not compute - I did not think there was any HP-PB in an A180, yet your mapper output is listing an HP-PB mux? (I must confess that the last time I was hands-on with an A180 was quite a long time ago...)

Anyhow, if the card does not have the HP SVID it won't be claimed. I do not know if there was ever an issue with improperly manufatured cards - if there was then if you got the card from HP you can try contacting some official support folks - unless my assertion that all cards are supposed to have HP SVIDS is mistaken - I do know that ages ago, the EISA TokenRing card was "pure" Madge.

[Steve Jones_4]

It is ans A180. The HP-PB mux is provided by an access-port card (early service processor)taken from an old D-class box. It plugs into the first HSC slot (and has some additional pins which dictate that it must use the first HSC slot). Don't know why it announces itself as HP-PB - the D-class didn't have PB slots either... Anyway, that card works ok.

I suspected that the non-HP SVID string was at the root of this. Looks like I was correct.

I'd still like to see MAPPER output from a working A3740A, if anyone can oblige.

Thanks for all the input.

Steve

[Florian Heigl (new acc)]

Hey Steve - I think I'll have to give up on this one - I just put on of the A3740 into my A180 (it's an A180C now, btw. :) - and now I've got steady power and heartbeat and a (slowly) blinking LAN led, which the user guide translates to a IO-bound error. :/
(Also, the box doesn't come up and gives no console output...)

Florian

[Florian Heigl (new acc)]

Card removed, System boots fine.

I'll do one more try by checking my PDC firmware levels.

[Florian Heigl (new acc)]

one more and for now last time:

PDC update solved the boot issue, now the system is coming up with the adapter inside, but it remains unclaimed and prints a nice boot message that the adapter is outdated, as I run 11.11

OTOH - I booted to ODE and started MAPPER and I'm at a loss how to get the pci diag information You'd need.

I'll try to gather some info from our techs, but for now the card is out of the system again (and ~120kg of others are on top of it)

If I find out what to do, or You can provide me with a short walkthrough, I'll give it another try. Heck - can You use a linux lspci -vv for what You need? It would appear to me.