On Sat, May 24, 2025 at 02:40:23PM +0200, Lukas Wunner wrote: > On Fri, May 23, 2025 at 09:00:27PM +0530, Manivannan Sadhasivam wrote: > > I thought that it *might* be possible to reset individual ports, > > so that's why I passed the root port 'pci_dev' to the callback > > in a hope that the controller drivers could use it to identify > > the root port they are resetting. > > Makes sense. > > > You are right. We should check if the parent bus of the bridge > > is a root bus or not. > > Okay, that's simple enough: pci_is_root_bus(dev->bus) > > > Yes, pretty much so. I could rename it to reset_root_port(), > > since I still believe that multi root port setups may be able > > to reset them separately. > > Conceivably, a PCIe host controller might also possess RCiEPs > in addition to Root Ports. Those are allowed to be FLR-capable, > but could also be reset through a platform-specific means. > > PCIe r6.3 page 121 defines the term "Root Complex Component", > which encompasses Root Ports but also RCiEPs. So if you want to > be super generic, you could use that term in lieu of Root Port, > though it consumes more characters. > We don't have any PCI controller driver incorporating RCiEPs afaik. So I'll stick to 'reset_root_port' for now. - Mani -- மணிவண்ணன் சதாசிவம்
