On Sat, Mar 01, 2025 at 07:49:23PM +0000, Luís Mendes wrote: > ... I would like to offer my help for testing the patches with your > guidance. > > ... In any way, long story short I've read in some threads about > mvebu that PCIe is broken, and I say that it is not, the secret to > have PCIe working is to use a good u-boot bootloader. Many recent > u-boot versions cause the PCIe not to work on boot and I haven't > figured out why yet. In fact my A388 systems are working with the > amdgpu driver and an AMD Polaris, a RX 550 with 4GB VRAM. I have > been succesfully using PCIe with this old u-boot version: > https://github.com/SolidRun/u-boot-armada38x repo and branch > u-boot-2013.01-15t1-clearfog. Glad to hear that. Making this work with more versions of u-boot would be great, and maybe we can make progress on that eventually. For right now, I want to try to make progress on Pali's patches. I'm pretty much in the dark as far as mvebu and IRQ details, but it might help enlighten me if you could collect these: - complete dmesg log, booted with "pci=nomsi" - complete output of "sudo lspci -vv" - contents of /proc/interrupts - output of "grep -r . /proc/irq/" For now just pick one of your systems, probably the solidrun one since that sounds like one that other folks might have. Bjorn
