On Sunday, May 27, 2012 06:43:43 PM Sean Bruno wrote:> I'm trying to understand the newbus and acpi interactions on this Dell
> R620 that result in the Broadcom adapter board being probed
"backwards"
> or just plain out of order in comparison to the connector layout and the
> linux tg3 driver.
>
> We seem to be detecting PCI0:2:0 before PCI0:1:0. This seems odd to me.
> When I replace the broadcom daughter card with an intel daughter card,
> this does not show up, so I assume either a malfunction of the Dell ACPI
> tables or the bge(4) driver.
Oof, you confused me. You are detecting bus PCI domain 0 bus 1 after PCI
domain 0 bus 2. A dmesg would be more useful here. FreeBSD walks the PCI
tree in a deterministic depth-first order, and we enumerate host bridges in
the order ACPI enumerates them. Looking at the ACPI dump, you have 3 host
bridges, PCI bus 0, and two uncore busses for your CPU sockets. So busses 1
and 2 must be children of bus 0. It would seem that bus 2 comes before bus 1
on PCI bus 0. You can tell this by seeing what the parent pcibX device of
busses 1 and 2 are and looking at the address of that pcibX device on PCI bus
0.
--
John Baldwin