Recently I added two PCIe 4.0 NVMe m.2 disks to a system (AMD Epyc 7402P on a ROMED8-2T board). Since then I receive warnings telling me that there were correctable AER errors on the physical layer.
Because the mainboard has only one m.2 slot of sufficient length (22110) I used a carrier card which according to the manufacturer supports PCIe 4.0. Today I found forum post stating that there are quite some mainboards which were originally designed for PCIe 3.0 and that even if they support PCIe 4.0 now they often have problems with carrier boards. Just to check I set PCIe link speed to Gen3 and it seems the errors are gone now.
Can somebody tell me if the mainboard mentioned is such a case (i.e. originally designed for PCIe 3.0)?
Checking the logs I found
acpi PNP0A08:00: _OSC: platform does not support [AER LTR],
but a few seconds later logs are full of
kernel: pci 0000:03:00.0: AER: aer_status: 0x00000001, aer_mask: 0x00000000
kernel: pci 0000:03:00.0: [ 0] RxErr (First)
kernel: pci 0000:03:00.0: AER: aer_layer=Physical Layer, aer_agent=Receiver ID
Why do I receive those errors if the platform is apparently unable to even support AER?
A final addition: I know that U.2 disks are better than M.2 disks but we can't use those due to thermal limitations of this system.
