On Tue, Mar 10, 2026 at 07:07:39PM +0530, Manivannan Sadhasivam wrote: > > > > I tested your patch series + your suggested change above, and after a: > > > > ## On EP side: > > # echo 0 > /sys/kernel/config/pci_ep/controllers/a40000000.pcie-ep/start && > > \ > > sleep 0.1 && echo 1 > > > /sys/kernel/config/pci_ep/controllers/a40000000.pcie-ep/start > > > > Instead of: > > > > # FAILED: 1 / 16 tests passed. > > > > I now get: > > # FAILED: 7 / 16 tests passed. > > > > Test cases 1-7 now passes (the test cases related to BARs), > > all other test cases still fail:
(snip) > I guess what is going wrong here is that you might be rebooting the device > *after* restoring the config space in 'slot_reset()' callback. So all the > previous config space contents would be lost and the device would've started > afresh. I'm confused. # echo 0 > /sys/kernel/config/pci_ep/controllers/a40000000.pcie-ep/start && \ sleep 0.1 && echo 1 > /sys/kernel/config/pci_ep/controllers/a40000000.pcie-ep/start Will not reboot the device. > > I don't think we can restore the config space after rebooting the EP device, > because host has no idea about it. It can only restore the content after > recovering the device through platform specific means. Am I missing something, or why are we talking about rebooting the device? Kind regards, Niklas
