Thanks Ngassam. Decoding the MSI address field, the interrupt will occur on the lowest priority APICs with logical ID 3 (usually core 3 in non-large systems) [1].
Can you see if rewriting the address with a simpler one prevents the crashing? Boot without 'pci=nomsi' and run:
$ sudo setpci -s 01:05.0 84.l=fee00000
(note, this won't work on other systems)
If this prevents the crashing, it's possible that I could cook up a MSI-quirk patch to write the simpler address vector only. If it fails, we need to not enable MSI for this board via adding a no-MSI quirk for this PCI device.
Thanks Ngassam. Decoding the MSI address field, the interrupt will occur on the lowest priority APICs with logical ID 3 (usually core 3 in non-large systems) [1].
Can you see if rewriting the address with a simpler one prevents the crashing? Boot without 'pci=nomsi' and run:
$ sudo setpci -s 01:05.0 84.l=fee00000
(note, this won't work on other systems)
If this prevents the crashing, it's possible that I could cook up a MSI-quirk patch to write the simpler address vector only. If it fails, we need to not enable MSI for this board via adding a no-MSI quirk for this PCI device.
Thanks, Daniel
--- [1]
01:05.0 VGA compatible controller: ATI Technologies Inc RC410 [Radeon Xpress 200M]
...
Capabilities: [80] Message Signalled Interrupts: Mask- 64bit- Queue=0/0 Enable+
Address: fee0300c Data: 4181