2006-01-13 22:46:07

by Joshua Kwan

[permalink] [raw]
Subject: [?] PCI BIOS masks some IDs to prevent OS detection?

Hi,

I'd like to tap some of the Linux-PCI gurus about something weird I've
been helping a friend with...

He recently installed a PCI RAID card, and ever since, his Ethernet
card stopped working. Further investigation revealed that his
Realtek 8139 (10ec:8139) card had become 10ec:0139, and his 3Com Cyclone
card had become 10b7:1055 from 10b7:9055.

Did the PCI bus decide to mask those PCI IDs to prevent some sort of
resource conflict that would ensue from loading an appropriate driver
for these devices?

--
Joshua Kwan


2006-01-14 00:27:42

by Alan

[permalink] [raw]
Subject: Re: [?] PCI BIOS masks some IDs to prevent OS detection?

On Gwe, 2006-01-13 at 14:45 -0800, Joshua Kwan wrote:
> He recently installed a PCI RAID card, and ever since, his Ethernet
> card stopped working. Further investigation revealed that his
> Realtek 8139 (10ec:8139) card had become 10ec:0139, and his 3Com Cyclone
> card had become 10b7:1055 from 10b7:9055.
>
> Did the PCI bus decide to mask those PCI IDs to prevent some sort of
> resource conflict that would ensue from loading an appropriate driver
> for these devices?

You just need to plug the cards in properly.

Alan

2006-01-14 19:12:58

by Onur Küçük

[permalink] [raw]
Subject: Re: [?] PCI BIOS masks some IDs to prevent OS detection?


> He recently installed a PCI RAID card, and ever since, his Ethernet
> card stopped working. Further investigation revealed that his
> Realtek 8139 (10ec:8139) card had become 10ec:0139, and his 3Com
> Cyclone card had become 10b7:1055 from 10b7:9055.
>
> Did the PCI bus decide to mask those PCI IDs to prevent some sort of
> resource conflict that would ensue from loading an appropriate driver
> for these devices?

Probably just 1 pin somewhere does not connect well to the board, if
it did not burn or got scratched. I once had a soundblaster turn into a
graphics card :) Make sure you plug in clean slots (clean the dust) and
plug the cards well.

--
Onur Küçük Knowledge speaks,
<onur.--.-.delipenguen.net> but wisdom listens

2006-01-16 15:53:55

by Lennart Sorensen

[permalink] [raw]
Subject: Re: [?] PCI BIOS masks some IDs to prevent OS detection?

On Fri, Jan 13, 2006 at 02:45:29PM -0800, Joshua Kwan wrote:
> I'd like to tap some of the Linux-PCI gurus about something weird I've
> been helping a friend with...
>
> He recently installed a PCI RAID card, and ever since, his Ethernet
> card stopped working. Further investigation revealed that his
> Realtek 8139 (10ec:8139) card had become 10ec:0139, and his 3Com Cyclone
> card had become 10b7:1055 from 10b7:9055.
>
> Did the PCI bus decide to mask those PCI IDs to prevent some sort of
> resource conflict that would ensue from loading an appropriate driver
> for these devices?

Maybe the raid card has a pin shorted to ground, so all reads of that
bit read as 0. That would explain why all the cards lost the same bit.
It appears to be the highest bit on the bus that is stuck low.

I see this fairly frequently on our own mainboards due to a problem with
the soldering of a surface mount pci bridge chip on the bottom of
the board.

So check if the problem goes away if the raid card is removed, and you
could check the raid card in another machine too. Either it isn't
seated properly, or it is broken, or the pci slot is defective perhaps.

Len Sorensen

2006-01-16 16:15:19

by Kyle Moffett

[permalink] [raw]
Subject: Re: [?] PCI BIOS masks some IDs to prevent OS detection?

On Jan 16, 2006, at 10:53, Lennart Sorensen wrote:
> Maybe the raid card has a pin shorted to ground, so all reads of
> that bit read as 0. That would explain why all the cards lost the
> same bit. It appears to be the highest bit on the bus that is stuck
> low.
>
> I see this fairly frequently on our own mainboards due to a problem
> with the soldering of a surface mount pci bridge chip on the bottom
> of the board.

I've also seen this problem with a faulty case where a small metal
support shorted a couple pins on the PCI riser. It gave odd PCI ids,
corrupted data, and crashed within a minute of booting. After
removing the metal bracket, it's been stable for almost a year.

Cheers,
Kyle Moffett

--
I have yet to see any problem, however complicated, which, when you
looked at it in the right way, did not become still more complicated.
-- Poul Anderson