| Message ID | 20110601132206.GF23771@pengutronix.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Jun 01, 2011 at 03:22:06PM +0200, Sascha Hauer wrote: > diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig > index 9adc278..89c3f16 100644 > --- a/arch/arm/Kconfig > +++ b/arch/arm/Kconfig > @@ -193,7 +193,6 @@ config VECTORS_BASE > > config ARM_PATCH_PHYS_VIRT > bool "Patch physical to virtual translations at runtime (EXPERIMENTAL)" > - depends on EXPERIMENTAL Please make this a separate patch - which also removes the (EXPERIMENTAL) suffix as well. Options which are experimental are supposed to have that tag in the text, options which aren't must not have the tag.
On Wednesday 01 June 2011, Sascha Hauer wrote: > > This allows for all armv4 and armv5 based i.MX systems to be compiled > together in one kernel. To accomplish this we need ARM_PATCH_PHYS_VIRT > and AUTO_ZRELADDR which is selected in Kconfig. > As compiling with ARM_PATCH_PHYS_VIRT breaks XIP support and different > PHYS_OFFSETs break uImage support the old way to integrate only one > of these SoCs is kept in place. Great stuff! My main concern is that it's really confusing to have so many symbols with slightly different scopes: * SOC_SELECT_IMX1 HAVE_SOC_IMX1 ARCH_MX1 SOC_IMX1 * SOC_SELECT_IMX21 HAVE_SOC_IMX21 ARCH_MX2 SOC_IMX21 Are you sure that this is actually the minimum set that is required? I assume that you have gone through all the uses of these symbols to check that there is no silent breakage if you enable more than one. The only ones that looks suspicious to me are arch/arm/plat-mxc/include/mach/timex.h and arch/arm/plat-mxc/include/mach/debug-macro.S. The first one is probably harmless, because we don't really use the CLOCK_TICK_RATE any more, the second one I can't tell. Arnd
On Wed, Jun 01, 2011 at 05:24:26PM +0200, Arnd Bergmann wrote: > On Wednesday 01 June 2011, Sascha Hauer wrote: > > > > This allows for all armv4 and armv5 based i.MX systems to be compiled > > together in one kernel. To accomplish this we need ARM_PATCH_PHYS_VIRT > > and AUTO_ZRELADDR which is selected in Kconfig. > > As compiling with ARM_PATCH_PHYS_VIRT breaks XIP support and different > > PHYS_OFFSETs break uImage support the old way to integrate only one > > of these SoCs is kept in place. > > Great stuff! > > My main concern is that it's really confusing to have so many symbols > with slightly different scopes: > > * SOC_SELECT_IMX1 HAVE_SOC_IMX1 ARCH_MX1 SOC_IMX1 > * SOC_SELECT_IMX21 HAVE_SOC_IMX21 ARCH_MX2 SOC_IMX21 The ARCH_MX* macros should be removed, but some driver Kconfig entries still depend on them. > > Are you sure that this is actually the minimum set that is required? I'd love to remove some of these, but I think they are required. SOC_SELECT_IMX1 and HAVE_SOC_IMX1 are only used to offer the right Kconfig magic. > > I assume that you have gone through all the uses of these symbols > to check that there is no silent breakage if you enable more than > one. The only ones that looks suspicious to me are > arch/arm/plat-mxc/include/mach/timex.h See An earlier patch in this series. The ifdefs are removed and CLOCK_TICK_RATE is replaced with a bogus value. CLOCK_TICK_RATE is unused but still required to compile the kernel. > and > arch/arm/plat-mxc/include/mach/debug-macro.S. This can be used only when only one family is compiled in. Doing otherwise should be catched with an error during comppilation. Sascha
On Wednesday 01 June 2011 18:47:31 Sascha Hauer wrote: > > > > Are you sure that this is actually the minimum set that is required? > > I'd love to remove some of these, but I think they are required. > SOC_SELECT_IMX1 and HAVE_SOC_IMX1 are only used to offer the right > Kconfig magic. > > > > > I assume that you have gone through all the uses of these symbols > > to check that there is no silent breakage if you enable more than > > one. The only ones that looks suspicious to me are > > arch/arm/plat-mxc/include/mach/timex.h > > See An earlier patch in this series. The ifdefs are removed and > CLOCK_TICK_RATE is replaced with a bogus value. CLOCK_TICK_RATE > is unused but still required to compile the kernel. > > > and > > arch/arm/plat-mxc/include/mach/debug-macro.S. > > This can be used only when only one family is compiled in. Doing > otherwise should be catched with an error during comppilation. > Ok, thanks for the explanations. Arnd
diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig index 9adc278..89c3f16 100644 --- a/arch/arm/Kconfig +++ b/arch/arm/Kconfig @@ -193,7 +193,6 @@ config VECTORS_BASE config ARM_PATCH_PHYS_VIRT bool "Patch physical to virtual translations at runtime (EXPERIMENTAL)" - depends on EXPERIMENTAL depends on !XIP_KERNEL && MMU depends on !ARCH_REALVIEW || !SPARSEMEM help diff --git a/arch/arm/mach-imx/Kconfig b/arch/arm/mach-imx/Kconfig index 59c97a3..501cf9b 100644 --- a/arch/arm/mach-imx/Kconfig +++ b/arch/arm/mach-imx/Kconfig @@ -11,8 +11,33 @@ config ARCH_MX31 config ARCH_MX35 bool +config ARCH_MX1 + bool + +config ARCH_MX2 + bool + +config ARCH_MX25 + bool + +config MACH_MX27 + bool + +config HAVE_SOC_IMX1 + bool + +config HAVE_SOC_IMX21 + bool + +config HAVE_SOC_IMX25 + bool + +config HAVE_SOC_IMX27 + bool + config SOC_IMX1 bool + select ARCH_MX1 select CPU_ARM920T select IMX_HAVE_DMA_V1 select IMX_HAVE_IOMUX_V1 @@ -20,6 +45,7 @@ config SOC_IMX1 config SOC_IMX21 bool + select ARCH_MX2 select CPU_ARM926T select ARCH_MXC_AUDMUX_V1 select IMX_HAVE_DMA_V1 @@ -28,6 +54,7 @@ config SOC_IMX21 config SOC_IMX25 bool + select ARCH_MX25 select CPU_ARM926T select ARCH_MXC_AUDMUX_V2 select ARCH_MXC_IOMUX_V3 @@ -35,6 +62,8 @@ config SOC_IMX25 config SOC_IMX27 bool + select MACH_MX27 + select ARCH_MX2 select CPU_ARM926T select ARCH_MXC_AUDMUX_V1 select IMX_HAVE_DMA_V1 @@ -59,7 +88,7 @@ config SOC_IMX35 select MXC_AVIC -if ARCH_MX1 +if HAVE_SOC_IMX1 comment "MX1 platforms:" config MACH_MXLADS @@ -89,27 +118,7 @@ config MACH_APF9328 endif -if ARCH_MX2 - -choice - prompt "CPUs:" - default MACH_MX21 - -config MACH_MX21 - bool "i.MX21 support" - help - This enables support for Freescale's MX2 based i.MX21 processor. - -config MACH_MX27 - bool "i.MX27 support" - help - This enables support for Freescale's MX2 based i.MX27 processor. - -endchoice - -endif - -if MACH_MX21 +if HAVE_SOC_IMX21 comment "MX21 platforms:" @@ -126,7 +135,7 @@ config MACH_MX21ADS endif -if ARCH_MX25 +if HAVE_SOC_IMX25 comment "MX25 platforms:" @@ -175,8 +184,7 @@ endchoice endif -if MACH_MX27 - +if HAVE_SOC_IMX27 comment "MX27 platforms:" config MACH_MX27ADS diff --git a/arch/arm/mach-imx/Makefile b/arch/arm/mach-imx/Makefile index e9eb36d..db6394a 100644 --- a/arch/arm/mach-imx/Makefile +++ b/arch/arm/mach-imx/Makefile @@ -1,12 +1,12 @@ obj-$(CONFIG_IMX_HAVE_DMA_V1) += dma-v1.o -obj-$(CONFIG_ARCH_MX1) += clock-imx1.o mm-imx1.o -obj-$(CONFIG_MACH_MX21) += clock-imx21.o mm-imx21.o +obj-$(CONFIG_SOC_IMX1) += clock-imx1.o mm-imx1.o +obj-$(CONFIG_SOC_IMX21) += clock-imx21.o mm-imx21.o -obj-$(CONFIG_ARCH_MX25) += clock-imx25.o mm-imx25.o ehci-imx25.o +obj-$(CONFIG_SOC_IMX25) += clock-imx25.o mm-imx25.o ehci-imx25.o -obj-$(CONFIG_MACH_MX27) += cpu-imx27.o pm-imx27.o -obj-$(CONFIG_MACH_MX27) += clock-imx27.o mm-imx27.o ehci-imx27.o +obj-$(CONFIG_SOC_IMX27) += cpu-imx27.o pm-imx27.o +obj-$(CONFIG_SOC_IMX27) += clock-imx27.o mm-imx27.o ehci-imx27.o obj-$(CONFIG_SOC_IMX31) += mm-imx31.o cpu-imx31.o clock-imx31.o iomux-imx31.o ehci-imx31.o obj-$(CONFIG_SOC_IMX35) += mm-imx35.o cpu-imx35.o clock-imx35.o ehci-imx35.o diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig index a5353fc..44ebde3 100644 --- a/arch/arm/plat-mxc/Kconfig +++ b/arch/arm/plat-mxc/Kconfig @@ -14,21 +14,41 @@ choice prompt "Freescale CPU family:" default ARCH_MX3 -config ARCH_MX1 - bool "MX1-based" +config SOC_SELECT_IMX1 + bool "i.MX1 based" + select HAVE_SOC_IMX1 help This enables support for systems based on the Freescale i.MX1 family -config ARCH_MX2 - bool "MX2-based" +config SOC_SELECT_IMX21 + bool "i.MX21 based" + select HAVE_SOC_IMX21 help - This enables support for systems based on the Freescale i.MX2 family + This enables support for systems based on the Freescale i.MX21 family -config ARCH_MX25 - bool "MX25-based" +config SOC_SELECT_IMX25 + bool "i.MX25 based" + select HAVE_SOC_IMX25 help This enables support for systems based on the Freescale i.MX25 family +config SOC_SELECT_IMX27 + bool "i.MX27 based" + select HAVE_SOC_IMX27 + help + This enables support for systems based on the Freescale i.MX27 family + +config ARCH_IMX_V4_V5 + select ARM_PATCH_PHYS_VIRT + select AUTO_ZRELADDR + select HAVE_SOC_IMX1 + select HAVE_SOC_IMX21 + select HAVE_SOC_IMX25 + select HAVE_SOC_IMX27 + bool "i.MX1, i.MX21, i.MX25 and i.MX27 based" + help + This enables support for all armv4 and armv5 based i.MX systems + config ARCH_MX3 bool "MX3-based" help diff --git a/arch/arm/plat-mxc/devices/platform-imx-dma.c b/arch/arm/plat-mxc/devices/platform-imx-dma.c index 3538b85..735b81d 100644 --- a/arch/arm/plat-mxc/devices/platform-imx-dma.c +++ b/arch/arm/plat-mxc/devices/platform-imx-dma.c @@ -76,7 +76,7 @@ static struct platform_device __init __maybe_unused *imx_add_imx_dma(void) return imx_add_platform_device("imx-dma", -1, NULL, 0, NULL, 0); } -#ifdef CONFIG_ARCH_MX25 +#ifdef CONFIG_SOC_IMX25 static struct sdma_script_start_addrs addr_imx25_to1 = { .ap_2_ap_addr = 729, .uart_2_mcu_addr = 904,
This allows for all armv4 and armv5 based i.MX systems to be compiled together in one kernel. To accomplish this we need ARM_PATCH_PHYS_VIRT and AUTO_ZRELADDR which is selected in Kconfig. As compiling with ARM_PATCH_PHYS_VIRT breaks XIP support and different PHYS_OFFSETs break uImage support the old way to integrate only one of these SoCs is kept in place. Also, As the ARM_PATCH_PHYS_VIRT now has a user this removes the dependency to EXPERIMENTAL. Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de> --- arch/arm/Kconfig | 1 - arch/arm/mach-imx/Kconfig | 58 +++++++++++++++----------- arch/arm/mach-imx/Makefile | 10 ++-- arch/arm/plat-mxc/Kconfig | 34 ++++++++++++--- arch/arm/plat-mxc/devices/platform-imx-dma.c | 2 +- 5 files changed, 66 insertions(+), 39 deletions(-)