* Ky?sti M?lkki <kmalkki at cc.hut.fi> [2003-01-21 12:39:44 +0200]: > On Mon, 20 Jan 2003, Mark M. Hoffman wrote: > > > * Ky?sti M?lkki <kmalkki at cc.hut.fi> [2003-01-20 11:06:56 +0200]: > > > Functionally it doesn't matter, but you should avoid all that > > > pci_find_device in sis645_setup(), a call to pci_module_init has done > > > the work for you. You can take i2c-amdXX as a sample that got accepted > > > in Linus's 2.5 tree. > > > > OK, I understand. Some of that I can't avoid though. The BIOS tends > > to hide the device I really want; so I have to flip a magic bit then > > search for it. > > It looks to me, that the SMBus device has its own PCI_ID, > and test for southbridge is only required if SMBus is disabled. > > I don't get the point of testing the north bridge type, > if the SMBus is somewhere else. E.g. although I have an Asus P4S333 which has 645/961, the south bridge claims to be 503 anyway. I assume this is some kind of compatability mode. Anyway, I can't be sure it's really a 961 unless I check against the list of compatible north bridges. > I don't know if the PCI subsystem cares, but I get the impression > that i2c-sis645 driver is not registered to control the SMBus > device, but the device listed in sis645_ids. Yes what this driver really wants is to register against the SMBus (sub) device only... and that all the detection and magic enabling stuff belongs elsewhere; perhaps in a separate module. MDS first suggested separate modules; I should have taken his advice. i2c-sis645-pci-fixup.c? Regards, -- Mark M. Hoffman mhoffman at lightlink.com
