On Tue, Nov 29, 2022 at 12:35:43PM -0600, Bjorn Helgaas wrote: > On Sun, Nov 27, 2022 at 04:10:05AM +0300, Serge Semin wrote: > > On Wed, Nov 23, 2022 at 01:44:36PM -0600, Bjorn Helgaas wrote: > > > On Sun, Nov 13, 2022 at 10:12:58PM +0300, Serge Semin wrote: > > > > Thanks for these new generic interfaces in the DWC core! And thanks > > > for the changes in this patch to take advantage of them in the > > > pcie-designware drivers. > > > > > > Do you plan similar changes to other drivers to take advantage of > > > these DWC-generic data and interfaces? If you add generic things to > > > the DWC core but only take advantage of them in your driver, I don't > > > think they are really usefully generic. > > > > Could you be more specific what generic things you are referring to? I > > am asking because the only part of the changes which is used in my > > low-level driver only is introduced in another patch of this series. > > I asked because three of your patches mention "generic" things, but I > didn't see any changes to drivers except pcie-designware: > > PCI: dwc: Introduce generic platform clocks and reset This patch introduces a method to request a generic platform clocks and resets by their names. As I already said these names are defined by the DT-bindings, which are platform-specific. That's why the most of the currently available drivers can't be converted to using it. Instead the new drivers are supposed to be encouraged to use the generic names (in accordance with the generic DW PCIe DT-schema) and the resources request interface (based on the generic DT-bindings) if it suits their design. Anyway I honestly tried to come up with an even more generic interface, which could be used by all the low-level drivers. But due to too much variations of the resource names and their sometimes too complex utilization in the drivers any solution looked too complex. After all of thoughts I decided to keep things simpler. > PCI: dwc: Introduce generic resources getter This patch defines a generic resource getter for the DW PCIe host and end-point drivers. That's why it's called generic. > PCI: dwc: Introduce generic controller capabilities interface This patch introduces an interface to set the device-specific capabilities. Since these capabilities can be marked as available by both the core driver (at least two of them already defined within this patchset) and low-level platform drivers the interface is called as generic. > > I hoped that we would be able to use these to remove some code from > existing drivers, but if they only improve maintainability of future > drivers, that's useful, too. Removing some code is possible for instance from the pcie-visconti.c driver by using the new generic clocks and resets request interface. I've scheduled to create a small patchset which would do that after the rest of my patches pass the review process. -Serge(y) > > Bjorn
