On Wed, Sep 26, 2012 at 9:45 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:

On Wed, Sep 26, 2012 at 12:59:51AM +0200, vj wrote:
> ---
> arch/arm/cpu/cpu.c | 1 +
> arch/arm/lib/barebox.lds.S | 10 +
> arch/arm/mach-omap/Kconfig | 21 +
> arch/arm/mach-omap/Makefile | 1 +
> arch/arm/mach-omap/include/mach/omap4_rom_usb.h | 146 +++++
> arch/arm/mach-omap/omap4_generic.c | 2 +
> arch/arm/mach-omap/omap4_rom_usb.c | 189 ++++++
> arch/arm/mach-omap/xload.c | 26 +
> scripts/.gitignore | 1 +
> scripts/Makefile | 3 +
> scripts/usbboot.c | 797 ++++++++++++++++++++++++
> 11 files changed, 1197 insertions(+)
> create mode 100644 arch/arm/mach-omap/include/mach/omap4_rom_usb.h
> create mode 100644 arch/arm/mach-omap/omap4_rom_usb.c
> create mode 100644 scripts/usbboot.c
>
> diff --git a/arch/arm/cpu/cpu.c b/arch/arm/cpu/cpu.c
> index 71ef8c0..05343de 100644
> --- a/arch/arm/cpu/cpu.c
> +++ b/arch/arm/cpu/cpu.c
> @@ -89,6 +89,7 @@ void arch_shutdown(void)
> : "r0", "r1", "r2", "r3", "r6", "r10", "r12", "lr", "cc", "memory"
> );
> #endif
> + __asm__ __volatile__ ("cpsid i\n");
> }
>
> #ifdef CONFIG_THUMB2_BAREBOX
> diff --git a/arch/arm/lib/barebox.lds.S b/arch/arm/lib/barebox.lds.S
> index a69013f..ff7b63d 100644
> --- a/arch/arm/lib/barebox.lds.S
> +++ b/arch/arm/lib/barebox.lds.S
> @@ -97,6 +97,16 @@ SECTIONS
> __bss_start = .;
> .bss : { *(.bss*) }
> __bss_stop = .;
> +#ifdef CONFIG_SHARE_USB_HANDLE
> + /*
> + * Reserve space for the USB handle
> + */
> + . = CONFIG_USB_HANDLE_HANDOVER;
> + . = ALIGN(4);
> + __usb_handle = .;
> + /* . += sizeof(struct usb); */
> + . += 84;
> +#endif

If I understand correctly the first stage loader puts some data here
where a second stage barebox (configured with the same value for
CONFIG_USB_HANDLE_HANDOVER) will pick it up.

This seems to be a bit fragile. Fortunately Jan has posted a series
that gives the control of the reset vector to the boards, which means
that both stages do not have to agree on an addres, but instead a
pointer can be passed from the first to the second stage.

> +int usb_open(void)

So this uses the OMAP4 ROM to do USB communication, right? The functions
should have a omap_rom_* prefix then, otherwise it's not clear for a
reader which context these functions have.

The calling of the ROM code btw explains why the setting of VBAR breaks
USB boot. The ROM probably uses the vectors.

> +
> + n = rom_get_per_driver(&pusb->io, boot->device_type);
> + if (n)
> + return n;
> +
> + pusb->dread.xfer_mode = boot->xfer_mode;
> + pusb->dread.options = boot->options;
> + pusb->dread.device_type = boot->device_type;
> +
> + pusb->dwrite.xfer_mode = boot->xfer_mode;
> + pusb->dwrite.options = boot->options;
> + pusb->dwrite.device_type = boot->device_type;
> +#else
> +#ifdef CONFIG_MMU
> +#error USB communications not working under MMU

You can add the following to CONFIG_OMAP4_USB_BOOT instead:

depends on !MMU

So that an invalid config is not allowed in the first place.

> diff --git a/scripts/usbboot.c b/scripts/usbboot.c

This should also be named omap4-usbboot.c, or maybe, if the works on
other OMAPs, just omap-usboot.c

Sascha

--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |