On 08/16/2017 12:39 AM, Bjorn Helgaas wrote: > [+cc Kishon, Pratyush] > > On Fri, Jul 14, 2017 at 02:07:33PM +0200, Niklas Cassel wrote: >> Since the introduction of the dw_pcie_readX_dbi/dw_pcie_writeX_dbi macros, >> most dw_pcie_read(pci->dbi_base, ..)/dw_pcie_write(pci->dbi_base, ..) calls >> have been converted to dw_pcie_readX_dbi/dw_pcie_writeX_dbi calls. >> Convert the remaining calls. >> >> Niklas Cassel (2): >> PCI: dwc: dra7xx: utilize dw_pcie_readX_dbi/dw_pcie_writeX_dbi macros >> PCI: dwc: spear13xx: utilize dw_pcie_readX_dbi/dw_pcie_writeX_dbi >> macros >> >> drivers/pci/dwc/pci-dra7xx.c | 14 ++++++-------- >> drivers/pci/dwc/pcie-spear13xx.c | 22 ++++++++++------------ >> 2 files changed, 16 insertions(+), 20 deletions(-) > > These patches are short, but obscure, e.g., > > - dw_pcie_read(pci->dbi_base + exp_cap_off + PCI_EXP_LNKCAP, 4, &val); > + val = dw_pcie_readl_dbi(pci, exp_cap_off + PCI_EXP_LNKCAP); > > The original call path was: > > dw_pcie_read(addr, ...) > readl(addr) > > and we're replacing it with the much more complicated: > > dw_pcie_readl_dbi > __dw_pcie_read_dbi > if (pci->ops->read_dbi) > return pci->ops->read_dbi(pci, base, reg, size) > dw_pcie_read(base + reg, size, &val) > readl > > This is not functionally equivalent (the new path uses ops->read_dbi > if it's set), so the changelog should say something about why we need > this change. > > Neither dra7xx nor spear13xx sets .read_dbi, so my guess is this > doesn't actually fix anything that's broken and this is just for > consistency. Hello Bjorn, It is just for consistency. Sorry for my misleading cover letter. It should have said: "For consistency, convert all dw_pcie_read(pci->dbi_base, ..)/dw_pcie_write(pci->dbi_base, ..) calls to use the dw_pcie_readX_dbi/dw_pcie_writeX_dbi macros". It is true that dw_pcie_readX_dbi/dw_pcie_writeX_dbi will use ops->read_dbi if it is set. However, if ops->read_dbi is set, you will most likely want to use that function for reading the dbi. Anyway, neither spear13xx nor dra7xx has ops->read_dbi set, so these patches shouldn't change their current behavior. Do you want me to resend these patches or would you prefer to just drop them? Best regards, Niklas > > That's fine, but I'm confused because I looked for some of the > previous conversions you mentioned but couldn't find any. I'm lost in > a maze of twisty little passages. > > Bjorn >
