Re: [PATCH] ahci: force 32-bit DMA for JMicron JMB582/JMB585

[Niklas Cassel <cassel@kernel.org>]

Hello Sven,

On Sat, Sep 05, 2026 at 09:04:57PM +0000, snoep wrote:
> So I was unable to reproduce the corruption on this Raspberry Pi 5
> with either the normal 64-bit DMA configuration or the Raspberry Pi
> 32-bit PCIe DMA overlay.
> 
> Again, since this JMB585 has no IOMMU group, I do not consider this a
> test of the Intel/ARM IOMMU enforcing-mode case.

Thank you for your testing!

Considering that Rasberry Pi seems to hand out 64-bit PCI addresses by
default, which is why they even have the (from JMB585's point of view)
completely meaningless pcie-32bit-dma-pi5 DT overlay.

Considering that you could run the reproducer for several iterations
without any issues, I am confident that JMB585 does indeed handle
64-bit DMA addresses perfectly fine, and we will drop the JMB585 quirk
as soon as there is fix for the AMD IOMMU issue.


> 
> However, I am still trying to determine what caused the repeated EXT4
> metadata corruption in the original RAID1 setup, and I would appreciate
> guidance on what would be most useful to test next.
> 
> Would repeating a sustained checksumming workload with the original
> two-SSD md RAID1 setup be useful, or is there another workload, tracing
> option, kernel configuration or test that would better help distinguish
> between the JMB585/PCIe path, md RAID1, EXT4, or another component?

I have no idea why you saw corruption on ext4 using 6.12.75+rpt-rpi-2712
downstream kernel.

Considering that they seem okay with carrying dirty workarounds for issues
for several months, that they don't even bother reporting to upstream, I
would strongly recommend running an upstream kernel instead.


I did see some reports from people that the JMB585 could overheat:
https://github.com/geerlingguy/raspberry-pi-pcie-devices/issues/354
https://forums.truenas.com/t/jmb585-m-2-to-sata-card-doesnt-see-ssd/8231

If you don't have a heatsink, perhaps it could be something to try.


Kind regards,
Niklas