REGRESSION in 6.0-rc7 caused by patch "ARM/dma-mapping: use dma-direct unconditionally"

[Russell King (Oracle)]

On Fri, Oct 21, 2022 at 06:30:50PM +0200, Christoph Hellwig wrote:
> On Thu, Oct 20, 2022 at 07:22:07PM +0100, Russell King (Oracle) wrote:
> > Is it possible that this would also cause data corruption reading from
> > a SATA card on armada-38x platforms?
> 
> armada-38x eems to be a mvebu platform, so on the account yes.
> I'm not sure what SATA host driver is used there, but that must also be
> doing weird thing with the DMA API for the issue to hit.

None of the devices on armada-38x mark themselves as DMA-coherent, but
I believe they are all DMA-coherent - certainly in
arch/arm/mach-mvebu/coherency.c, all platform devices will be marked 
as DMA-coherent via the notifier, which basically means all devices
declared in DT need to be DMA-coherent.

Since the crypto engine, SD, USB and so on and so forth all use DMA,
it's highly likely that all those were broken as well and this isn't
limited to just mvneta as originally reported. Basically the entire
platform got broken, which really isn't good.

At least in this case, I have a test platform I try new kernels out on
before putting it on my connectivity-critical hardware - such as the
machine that my public connectivity comes through!

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!