2.4.0 pdev_enable_device() call in setup-bus.c
Hi Ivan,
Can you explain why pci_assign_unassigned_resources() calls pdev_enable_device() for every PCI device instead of having each PCI *driver* call pci_enable_device() as part of driver initialization?
I'm thinking I missed something that a comment in the code should have explained.
After having written the bulk of PCI support for parisc port, I was clearly under the impression the PCI driver was supposed to call pci_enable_device(). IMHO, it's a *bad* idea to enable a device when it's driver might not be present.
thanks,
grant
Code from drivers/pci/setup-bus.c:
void __initpci_assign_unassigned_resources(void)
{
...
pci_for_each_dev(dev) {
pdev_enable_device(dev);
}
}-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/
Quoted 4 lines · Grant Grundler
Can you explain why
pci_assign_unassigned_resources()callspdev_enable_device()for every PCI device instead of having each PCI *driver* callpci_enable_device()as part of driver initialization?
Mainly because there are driverless devices like display adapters, PCI bridges, or PCI devices with legacy drivers (IDE, for example).
OTOH, pdev_enable_device() most likely will be removed, but it's 2.5 material.
Ivan.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/
Quoted 2 lines · Ivan Kokshaysky
Mainly because there are driverless devices like display adapters, PCI bridges, or PCI devices with legacy drivers (IDE, for example).
Ok. It seems that all of those would have to interact with the PCI code someplace. And that's were pci_enable_device() could be called. eg. PCI Bridges could be handled in it's "driver": pci_setup_bridge().
Quoted 2 lines
OTOH,
pdev_enable_device()most likely will be removed, but it's 2.5 material.
Agreed - standardizing the enable path would be good for above devices.
I'd like to see arch hooks added for stuff like default latency and PCI_BRIDGE_CONTROL. My gut feeling is the "root" struct pci_bus allocation and initialization should be done by arch specific code *before* pci_scan_bus() is called. That would allow cfg space defaults to be set to arch specific values on a per bus basis *before* doing the bus walks instead of hacking this in pci_bus_fixup() later.
thanks,
grant
Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
Please read the FAQ at http://www.tux.org/lkml/