Hi Thomas,

Below commit is supposed to fix this issue. It's currently in
x86/vt-d branch of the iommu tree, and will be part of v5.6-rc1.

commit f5a68bb0752e0cf77c06f53f72258e7beb41381b
Author: Barret Rhoden <brho@google.com>
Date:   Wed Jan 15 11:03:56 2020 +0800

     iommu/vt-d: Mark firmware tainted if RMRR fails sanity check

     RMRR entries describe memory regions that are DMA targets for devices
     outside the kernel's control.

     RMRR entries that fail the sanity check are pointing to regions of
     memory that the firmware did not tell the kernel are reserved or
     otherwise should not be used.

     Instead of aborting DMAR processing, this commit marks the firmware
     as tainted. These RMRRs will still be identity mapped, otherwise,
     some devices, e.x. graphic devices, will not work during boot.

     Signed-off-by: Barret Rhoden <brho@google.com>
     Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
     Fixes: f036c7fa0ab60 ("iommu/vt-d: Check VT-d RMRR region in BIOS
is reported as reserved")
     Signed-off-by: Joerg Roedel <jroedel@suse.de>

Best regards,
baolu

On 2020/2/3 5:06, Thomas G wrote:
> Dear all,
>
> Vt-d seems to be broken since I upgraded to kernel 5.5, there is a new
> dmesg that was not there in 5.4:
> "DMAR: [Firmware Bug]: No firmware reserved region can cover this RMRR
> [0x0000000041d92000-0x0000000041d94fff], contact BIOS vendor for fixes"
>
> System motherboard is Asrock X299 Taichi XE, bios 1.70, CPU is Core i9 7920
> Vt-d and gpu passthrough was working ok since kernel 4.20 when I
> purchased it
>
> I attached the dmesg I get with kernel 5.5.1, and the dmesg from 5.4.17.
>
> Also, I use the following script that displays the IOMMU groups, I found
> the script there
> https://wiki.archlinux.org/index.php/PCI_passthrough_via_OVMF
>
> #!/bin/bash
> shopt -s nullglob
> for d in /sys/kernel/iommu_groups/*/devices/*; do
>      n=${d#*/iommu_groups/*}; n=${n%%/*}
>      printf 'IOMMU Group %s ' "$n"
>      lspci -nns "${d##*/}"
> done;
>
> The output of this script with kernel 5.4.17 is attached to this email.
> The same script with kernel 5.5.1 produces an empty output
>
> Hope it helps narrowing down the problem, please ask additional info if
> needed. Sorry if this is not submitted as it should be (then please
> advise), first time I report -what apprears to be- a bug.
>
> Best,
> Thomas