Re: [PATCH 1/5] x86_64: get mp_bus_to_node as early v3

["Yinghai Lu" <yhlu.kernel@gmail.com>]

On 8/8/07, Muli Ben-Yehuda <muli@il.ibm.com> wrote:
>
> Can you explain why this patch (and your other patches in this area)
> are needed? is this a performance issue? the patches seem complex, is
> there a good argument for that complexity?
>
1. AMD K8 system with two HT chain: when acpi=off or there is not _PXM
in DSDT, still make numa_node of device get correct value. ...
/sys/devices/pci0000:80/*/numa_node will be correct instead of 0 for
all. ---- so far i didn't find one two chain AMD k8 based system have
_PXM in DSDT...
2. could use dev_to_node instead of pcibus_to_node.
3. complex? I don't think so. It just try to get bus_to_node mapping
from pci conf space and put into one 256 byte array early, and then
use it some link pxm_to_node.
4. make it support quad core opteron by change to scan four regs in NB
func1 intead of func0.

YH
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/