On 8.01.2025 11:44 PM, Stephen Boyd wrote: > Quoting Rob Herring (2025-01-08 06:11:28) >> On Tue, Jan 7, 2025 at 7:28 PM Stephen Boyd <swboyd@xxxxxxxxxxxx> wrote: >>> >>> Extract the simple bus into a self contained driver so that devices are >>> still populated when a node has two (or more) compatibles with the least >>> specific one being the generic "simple-bus". Allow the driver to be a >>> module so that in a fully modular build a driver module for the more >>> specific compatible will be loaded first before trying to match this >>> driver. >>> >>> Cc: Rob Herring <robh@xxxxxxxxxx> >>> Cc: Saravana Kannan <saravanak@xxxxxxxxxx> >>> Cc: <devicetree@xxxxxxxxxxxxxxx> >>> Cc: Arnd Bergmann <arnd@xxxxxxxx> >>> Cc: "Uwe Kleine-König" <u.kleine-koenig@xxxxxxxxxxxx> >>> Cc: Bjorn Andersson <andersson@xxxxxxxxxx> >>> Cc: Konrad Dybcio <konradybcio@xxxxxxxxxx> >>> Cc: <linux-arm-msm@xxxxxxxxxxxxxxx> >>> Signed-off-by: Stephen Boyd <swboyd@xxxxxxxxxxxx> >>> --- [...] > Maybe the best approach is to simply avoid all of this and drop the > "simple-bus" compatible from the soc node? It introduces an annoying > hurdle where you have to enable the new driver that does exactly the > same thing as "simple-bus" does so you continue to have a working > system, but it avoids the headaches of trying to make the fallback to > "simple-bus" work and it would match how new DTs would be written. We > could make the driver 'default ARCH_<SOC_VENDOR>' so that it gets built > for olddefconfig users too. I think it even makes logical sense for the /soc node's compatible to be.. you know.. the model of the SoC we're modeling Konrad
