From mboxrd@z Thu Jan 1 00:00:00 1970 From: Matt Porter Date: Mon, 25 Feb 2013 09:46:38 -0500 Subject: [U-Boot] [PATCH 05/10] am33xx: add ti814x specific register definitions In-Reply-To: References: <1360766644-9901-1-git-send-email-mporter@ti.com> <1360766644-9901-6-git-send-email-mporter@ti.com> <87txpaofbi.fsf@dell.be.48ers.dk> Message-ID: <20130225144638.GJ2244@beef> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On Mon, Feb 18, 2013 at 03:18:42PM +0000, Tom Rini wrote: > On Sun, Feb 17, 2013 at 09:28:33PM +0100, Peter Korsgaard wrote: > > >>>>> "Matt" == Matt Porter writes: > > > > Matt> Support the ti814x specific register definitions within > > Matt> arch-am33xx. > > > > Matt> Signed-off-by: Matt Porter > > Matt> --- > > Matt> arch/arm/cpu/armv7/am33xx/sys_info.c | 3 +++ > > Matt> arch/arm/include/asm/arch-am33xx/cpu.h | 11 +++++---- > > Matt> arch/arm/include/asm/arch-am33xx/hardware.h | 32 +++++++++++++++++++++++++++ > > Matt> arch/arm/include/asm/arch-am33xx/omap.h | 7 ++++++ > > Matt> arch/arm/include/asm/arch-am33xx/spl.h | 5 +++++ > > Matt> 5 files changed, 54 insertions(+), 4 deletions(-) > > > > Matt> diff --git a/arch/arm/include/asm/arch-am33xx/hardware.h b/arch/arm/include/asm/arch-am33xx/hardware.h > > Matt> index 41ab2c0..786c159 100644 > > Matt> --- a/arch/arm/include/asm/arch-am33xx/hardware.h > > Matt> +++ b/arch/arm/include/asm/arch-am33xx/hardware.h > > Matt> @@ -20,9 +20,14 @@ > > Matt> #define __AM33XX_HARDWARE_H > > > > Matt> #include > > Matt> +#include > > > > Quite some of the base addresses are similar, but I wonder if it > > wouldn't be cleaner to simply have a hardware-am33xx.h / > > hardware-ti814x.h instead of all these ifdef / elif? > > Since I suspect the things common from ti814x and am33xx are also common > to ti816x (which has been left as an exercise to whomever needs that > next), I think we can re-structure this into something like that, but > keeping the common parts within hardware.h still. Sounds good. I'll restructure with only the common parts in hardware.h. > > Matt> /* Control Module Base Address */ > > Matt> +#ifdef CONFIG_AM33XX > > Matt> #define CTRL_BASE 0x44E10000 > > Matt> #define CTRL_DEVICE_BASE 0x44E10600 > > Matt> +#elif defined(CONFIG_TI814X) > > Matt> +#define CTRL_BASE 0x48140000 > > Matt> +#endif > > > > No CTRL_DEVICE_BASE on ti814x? > > I think this is a side-effect of Matt not supporting the things attached > to it (USB in the case of am335x). I tried to avoid defining things I'm not yet using. For CTRL_DEVICE_BASE, I would add it once I get to adding cpsw support as we need to read the efused macid value from that area. > > Matt> --- a/arch/arm/include/asm/arch-am33xx/spl.h > > Matt> +++ b/arch/arm/include/asm/arch-am33xx/spl.h > > Matt> @@ -25,8 +25,13 @@ > > > > Matt> #define BOOT_DEVICE_XIP 2 > > Matt> #define BOOT_DEVICE_NAND 5 > > Matt> +#ifdef CONFIG_AM33XX > > Matt> #define BOOT_DEVICE_MMC1 8 > > Matt> #define BOOT_DEVICE_MMC2 9 /* eMMC or daughter card */ > > Matt> +#elif defined(CONFIG_TI814X) > > Matt> +#define BOOT_DEVICE_MMC1 9 > > Matt> +#define BOOT_DEVICE_MMC2 8 /* ROM only supports 2nd instance */ > > > > Argh! Couldn't we just swap the meaning of mmc1/mmc2 or would that be > > too confusing? > > IMHO, that will lead to further confusion down the line. I talked with > Matt about this before and well, it's funky. This is definitely a quirky area wrt TI814x. AFAIK it's the only OMAP-ish part where the ROM only allows boot from a single MMC instance. Further, that single MMC instance is actually the 2nd one as noted in the comment. This means to keep all the existing SPL mmc init logic intact, we need to define as above so that the unimplemented first MMC instance (which does exist on the part) does not get intialized when the ROM drops a bootdevice of "8". -Matt