On Thu, Jul 02, 2020 at 12:15:22PM +0100, Mark Brown wrote: > On Wed, Jul 01, 2020 at 08:32:50PM -0300, Jason Gunthorpe wrote: > > On Wed, Jul 01, 2020 at 10:50:49AM +0100, Mark Brown wrote: > > > > Another part of this is that there's not a clean cut over between MMIO > > > and not using any hardware resources at all - for example a device might > > > be connected over I2C but use resources to distribute interrupts to > > > subdevices. > > > How does the subdevice do anything if it only received an interrupt? > > Via some bus that isn't memory mapped like I2C or SPI. > > > That sounds rather more like virtual bus's use case.. > > These are very much physical devices often with distinct IPs in distinct > address ranges and so on, it's just that those addresses happen not to > be on buses it is sensible to memory map. But platform bus is all about memmory mapping, so how does the subdevice learn the address range and properly share the underlying transport? Jason