On Wed, Sep 27, 2023 at 12:55:42PM -0500, Bjorn Helgaas wrote: > On Wed, Sep 27, 2023 at 05:46:02PM +0200, Manivannan Sadhasivam wrote: > > Instead of hardcoding the link speed in MBps, let's make use of the > > existing PCIE_SPEED2MBS_ENC() macro that does the encoding of the > > link speed for us. Also, let's Wrap it with QCOM_PCIE_LINK_SPEED_TO_BW() > > macro to do the conversion to ICC speed. > > In subject, /Make use of/Use/ would make better use of the subject > line. Lots of "uses" there :) > > Above, s/let's//, and also s/make use of/use/. > Ok, I will reword accordingly. > > +#define QCOM_PCIE_LINK_SPEED_TO_BW(speed) \ > > + Mbps_to_icc(PCIE_SPEED2MBS_ENC(pcie_link_speed[speed])) > > It's a shame to have to duplicate this macro in pcie-qcom.c and in > pcie-qcom-ep.c, especially since there's nothing qcom-specific in it. > > Would a macro like this fit in interconnect.h? > Unfortunately, no. This macro is neither interconnect specific nor PCI, but a kind of both used by the driver. So it makes sense to keep it in the driver itself for now. If we happen to create a common header for host and ep drivers in the future, it can be moved to that. > > - ret = icc_set_bw(pcie_ep->icc_mem, 0, MBps_to_icc(PCIE_GEN1_BW_MBPS)); > > + ret = icc_set_bw(pcie_ep->icc_mem, 0, QCOM_PCIE_LINK_SPEED_TO_BW(1)); > > "1" is not very informative here. Maybe PCIE_SPEED_2_5GT? (I didn't > completely verify that this is equivalent.) > No. PCIE_SPEED_2_5GT is defined as 0x14 in pci.h. And I do not want to add a macro for just "1" here. - Mani > Bjorn -- மணிவண்ணன் சதாசிவம்