Re: [PATCH 2/4] PCI: pciehp: bail out if pci_hp_add_bridge() fails

On Sat, May 04, 2024 at 05:02:34PM +0200, Lukas Wunner wrote:

On Sat, May 04, 2024 at 12:56:30PM +0200, Nam Cao wrote:

...

On Sat, May 04, 2024 at 11:51:54AM +0200, Lukas Wunner wrote:

...

Could you reproduce with pciehp instead of shpchp please?

Same thing for pciehp below. I think the problem is because without pci_stop_and_remove_bus_device(), no one cleans up the device added in pci_scan_slot(). When another device get hot-added, pci_get_slot() wrongly thinks another device is already there, so the hot-plug fails.

pciehp powers down the slot because you're returning a negative errno from pciehp_configure_device(). Please return 0 instead if pci_hp_add_bridge() fails.

Thanks for the suggestion. This is applicable to shpchp as well.

Best regards, Nam