Re: DragonFly kernel not seeing PCI bus

[Bill Hacker <wbh@xxxxxxxxxxxxx>]

dpk wrote:

> On Tue, 19 Apr 2005, Chuck Tuffli wrote:
>
>
> > First steps would be the output from 'pciconf -lv' and the dmesg if the
> > system is booting.
>
>
> I'm guessing this is unusual then; that's good at least.
I've had quite a lot of it with older MB/BIOS - several SATA
controllers - one Intel GigE NIC.

In several cases the scan detects all the PCI bus critters  and
the cards resident, but does not then utilize the card ID to match it to
an appropriate driver.

Not all that unexpected with newish Intel Gig-E cards, but puzzling
when it happens to older storage controllers that have been listed as
supported for several years.

I *suspect* - but do not have time to prove - that:

'missing' the bus or device is due to reduced tolerance for marginal 
timing issues, as it is occuring with the more recent FreeBSD's (4.10, 
4.11) as well as DragonFly, but does not (necessarily) occur with 
FreeBSD 4.8 or 4.9 (in one case 4.7).
 There is no compelling reason to expend scarce resources on older MB 
with out-of-spec bus timing, happy to drop that issue *unless* it is 
also showing up as a problem on newer, common, MB.

that the 'hints' or whatever - that steer a given maker-code, 
device-code to a particular driver, is not as comprehensive as it could 
be w/r rev-levels of common devices.

Unfortunately it will be September-October before I again have time to 
look into these.

- But please do continue to provide the best information you can 
if/as/when this issue bites.

Best,

Bill Hacker