Re: X86_64 + VIA + 4g problems

[Andi Kleen <ak@suse.de>]

Dieter Stüken <stueken@conterra.de> writes:
[can you please not always drop me from cc with each reply?]

Dieter Stüken <stueken@conterra.de> writes:

> OK, here are my last results for today:
>
> using "iommu=allowed" did not work. System freezes during initialization
> of the PDC20318, which is on the external PCI bus.
>
> But swiotlb=force works well!

This means your PCI bridge doesn't support addresses >4GB.

> The pci-gart.c patch seems to disable dma. 

Only DMA for addresses >4GB.

> Is this the DMA my PCI devices
> perform them self? As I learned, they may perform DMA even above 4g if all
> works well. Thus I may be happy without any IOMMU. As I saw my system working
> even without this patch, I will turn back to the original 2.6.15-rc7 an continue
> running this torture test during this night.

The patch should perform slightly better than swiotlb=force
because it will only force bounce buffering for addresses >4GB.

If your torture test involves more than 64MB of IO in flight
you might also need to increase the bounce buffer area
with swiotlb=128M or somesuch.

-Andi