Messages in this thread | | | Date | Wed, 13 Aug 2003 16:29:21 -0400 | From | Jeff Garzik <> | Subject | Re: C99 Initialisers |
| |
Matthew Wilcox wrote: > On Wed, Aug 13, 2003 at 03:44:44PM -0400, Jeff Garzik wrote: > >>enums are easy putting direct references would be annoying, but I also >>argue it's potentially broken and wrong to store and export that >>information publicly anyway. The use of enums instead of pointers is >>practically required because there is a many-to-one relationship of ids >>to board information structs. > > > The hard part is that it's actually many-to-many. The same card can have > multiple drivers. one driver can support many cards.
pci_device_tables are (and must be) at per-driver granularity. Sure the same card can have multiple drivers, but that doesn't really matter in this context, simply because I/we cannot break that per-driver granularity. Any solution must maintain per-driver granularity.
> Let me give you a true story that your solution needs to address. > I recently got myself a Compaq Evo with an eepro100 onboard. So I took > my Debian 3.0 CD and tried to install on it. Failed because the eepro > on the board had PCI IDs that were more recent than the driver. > > So I took the driver module, put it on a floppy, hand-edited the binary > to replace one of the PCI IDs with the ones that came back from lspci. > Stuck the floppy back in the Evo, loaded the hacked module and finished > the install. Then compiled a new kernel ;-) > > I haven't seen anything to address this in a nicer way yet.
Well, the step I wish to take - moving the ids from location X to location Y, would have no effect on that scenario, positive or negative.
Jeff
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |