Issue with PASID on some platforms

[barabanshek]

Hi,

I'm trying to run the micro, and on some machines it works, on some the DSA halts with a weird PASID-relate error in dmesg. All the machines I'm using are identical SPR servers (but slightly different CPU versions) with identical Linux Kernels and BIOS configurations.

[ 232.023841] idxd 0000:6a:01.0: idxd halted, need FLR. [ 232.023907] DMAR: DRHD: handling fault status reg 3 [ 232.023974] DMAR: [DMA Read PASID 0x2] Request device [6a:01.0] fault addr 0x7fe0c9943000 [fault reason 0x3a] SM: Non-zero reserved field set in Root Entry

I'm using the following sequence of actions to reproduce it:

sudo ./scripts/setup_dsa.sh -d dsa0 -w 1 -m d -e 4
sudo ./src/dsa_perf_micros -n128 -s4k -j -c -f -i1000 -k5 -w0 -zF,F -o3.

Have you seen this error on any machine before? If so, what can it be related to?

I'm using Linux Kernel 5.15, patched with enqcmd and using the same kernel on all the machines.

[barabanshek]

A follow-up on this. I think this might be related to this issue with PASID directory consistency , but after I moved to Kernel 6.5.0 (which has this patch), the issue still persists.