On Sun, 19 Apr 2015 00:28:39 +0200 Jonas Gorski <jogo@xxxxxxxxxxx> wrote: > Hi, > > On Fri, Apr 17, 2015 at 2:36 PM, Alban Bedel <albeu@xxxxxxx> wrote: > > The define AR71XX_PCI_MEM_SIZE miss one window, there is 7 windows, > > not 6. To make things clearer, and allow simpler code, derive > > AR71XX_PCI_MEM_SIZE from the newly introduced AR71XX_PCI_WIN_COUNT > > and AR71XX_PCI_WIN_SIZE. > > > > The define AR71XX_PCI_WIN7_OFFS also add a typo, fix it. > > I think this will break PCI on ar71xx. > > > > > Signed-off-by: Alban Bedel <albeu@xxxxxxx> > > --- > > arch/mips/include/asm/mach-ath79/ar71xx_regs.h | 6 ++++-- > > 1 file changed, 4 insertions(+), 2 deletions(-) > > > > diff --git a/arch/mips/include/asm/mach-ath79/ar71xx_regs.h > > b/arch/mips/include/asm/mach-ath79/ar71xx_regs.h index > > aa3800c..e2669a8 100644 --- > > a/arch/mips/include/asm/mach-ath79/ar71xx_regs.h +++ > > b/arch/mips/include/asm/mach-ath79/ar71xx_regs.h @@ -41,7 +41,9 @@ > > #define AR71XX_RESET_SIZE 0x100 > > > > #define AR71XX_PCI_MEM_BASE 0x10000000 > > -#define AR71XX_PCI_MEM_SIZE 0x07000000 > > +#define AR71XX_PCI_WIN_COUNT 8 > > +#define AR71XX_PCI_WIN_SIZE 0x01000000 > > +#define AR71XX_PCI_MEM_SIZE (AR71XX_PCI_WIN_COUNT * > > AR71XX_PCI_WIN_SIZE) > > > > #define AR71XX_PCI_WIN0_OFFS 0x10000000 > > #define AR71XX_PCI_WIN1_OFFS 0x11000000 > > @@ -50,7 +52,7 @@ > > #define AR71XX_PCI_WIN4_OFFS 0x14000000 > > #define AR71XX_PCI_WIN5_OFFS 0x15000000 > > #define AR71XX_PCI_WIN6_OFFS 0x16000000 > > -#define AR71XX_PCI_WIN7_OFFS 0x07000000 > > +#define AR71XX_PCI_WIN7_OFFS 0x17000000 > > These values are used in exactly one place, for writing into the PCI > address space offset registers. > The 7th PCI window is a special one for accessing the configuration > space registers, which requires to be set to 0x07000000 for that > purpose. So by changing this value you likely break access to these > values. Sorry, I foolishly assumed it was a typo. > > > > #define AR71XX_PCI_CFG_BASE \ > > (AR71XX_PCI_MEM_BASE + AR71XX_PCI_WIN7_OFFS + 0x10000) > > Also this macro would now be wrong, and calculate a wrong address. I see, I'll drop this patch and rework the following one to match the old code. Alban
