Hi there: A revised SiS645 SMBus driver is now here: http://home.pacbell.net/mh0ffman/sensors/i2c-sis645.c It uses PCI hotplug; the SMBus device shows up in lspci now. I couldn't see any reason to remove the device on module unload though. I am still interested in your comments and suggestions. Regards, -- Mark M. Hoffman mhoffman at lightlink.com On Sat, Aug 24, 2002 at 04:20:13PM -0400, Mark Studebaker wrote: > With respect to the PCI "junk", > the "right" way to do it is to use the kernel's PCI hotplug facility. > Two examples are the m7101 and p4b_smbus modules in prog/hotlug in our > package. > I think it's the only acceptable way. > We have many people using these modules, they work well. > > These are "fixup" modules that enable a PCI device. They must > be run before the corresponding i2c driver. > > We've always treated boards that failed to enable a PCI device > as the exception - and haven't put the hotplug code into any i2c driver > itself. You could certainly do that for yours, it would have to > be wrapped with #ifdef CONFIG_HOTPLUG. You're most of the way > there already. > > How many people run kernels with CONFIG_HOTPLUG, I have no idea - > but you can always put out a printk telling them to > enable it if they want to get their SMBus device. > > I also checked in your ID. > > mds > > > > > "Mark M. Hoffman" wrote: > > > > Hello sensors list: > > > > I'm also ready for an ID for SiS645 SMbus. You can put my name down for > > 635, 645, and 735; although this first attempt only supports 645. > > > > According to SiS, it should also work for the others. But this is the *only* > > information I got from them - still no datasheets. Thanks to Alexander for > > finding the 630 datasheet though... it's close enough to put this over the > > top. > > > > It tests OK on my Asus P4S333. You can find the driver here... > > http://home.pacbell.net/mh0ffman/sensors/i2c-sis645.c > > > > I'll follow this up w/ docs, sensors-detect patch, etc. as soon as I can. > > > > I'm interested in any comments or suggestions; especially would someone > > with more PCI experience care to comment on the junk starting at line 168? > > > > Thanks and regards, > > > > -- > > Mark M. Hoffman > > mhoffman at lightlink.com