On Fri, Aug 14, 2020 at 02:53:20PM -0400, Mohammed Billoo wrote: > Andrew, > > Thanks for the feedback. That makes sense. > > My use case was very specific that required these changes and I thought I'd > push them upstream for others. Basically, we have an FPGA over PCI that has > some IP, vendor-specific bridging to AHB, and then grlib to go to i2c (among > other controllers). When we load the pci driver we create and add a mfd device > that has the specific BAR as the resource, and the offset within the BAR for > the i2c controller. Looking back, all of this could be added to the device > tree, so what we've done so far may be an anti-pattern. > > Would you still want this patch? I would prefer device tree. Andrew