On Thu, Sep 27, 2012 at 07:06:20PM +0100, Jon Medhurst (Tixy) wrote:
> From: Jon Medhurst <tixy@xxxxxxxxxx>
>
> The A15xA7 models simulate a Cache Coherent Interconnect (CCI) and this
> needs to be initialised correctly for Linux to boot.
>
> To perform this initiation we add the new function configure_from_fdt()
> which will look in the fdt for devices to initialise. In this first case
> we look for the CCI node and if found then setup this device.
>
> Signed-off-by: Jon Medhurst <tixy@xxxxxxxxxx>
> ---
> semi_loader.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 50 insertions(+)
>
> diff --git a/semi_loader.c b/semi_loader.c
> index c9750be..ca70633 100644
> --- a/semi_loader.c
> +++ b/semi_loader.c
> @@ -242,6 +242,54 @@ libfdt_error:
> fatal("libfdt: ", fdt_strerror(e), ", while updating device tree\n");
> }
>
> +/* For accessing 32-bit device ports */
> +#define io32(p) (*(volatile uint32_t *)(p))
> +
> +static void init_cci(unsigned cci)
> +{
> + info("Initialising CCI\n");
> +
> + /*
> + * Ideally, the CCI device tree binding would include suitable
> + * information so we can correctly configure the CCI, but for
> + * now we'll just hard-code settings for the present A15xA7
> + * models.
> + */
> +
> + /* Turn on CCI snoops and DVM messages */
> + io32(cci+0x4000) = 0x3; /* A15 cluster */
> + io32(cci+0x5000) = 0x3; /* A7 cluster */
Ultimately, CCI slave port assignments will be deducible from the DT
topology, but we don't have this yet, so I guess this is OK for now.
For the intergrated switcher work we can have a simple local patch for
this until/unless this (or the switcher) becomes more generic.
> +
> + /* Wait while change pending bit of status register is set */
> + while(io32(cci+0xc) & 0x1)
> + {}
> +}
> +
> +static void configure_from_fdt(struct loader_info *info)
> +{
> + void *fdt = (void *)info->fdt_start;
> + uint32_t const *p;
> + int addrcells, sizecells;
> + int offset, len;
> +
> + if(!fdt)
> + return;
> +
> + _fdt_address_and_size_cells(fdt, &addrcells, &sizecells);
> +
> + /* See if there is a CCI device to initialise */
> + offset = fdt_node_offset_by_compatible(fdt, 0, "arm,cci");
I'm guessing this binding isn't official yet, but it'll do for now...
> + if (offset >= 0) {
> + p = fdt_getprop(fdt, offset, "reg", &len);
> + if(len != (addrcells + sizecells) * 4)
> + info("Failed parsing device-tree node for CCI\n");
> + else
> + init_cci(fdt32_to_cpu(p[addrcells - 1]));
I think this is worth a comment. I presume you're trying to get the low
32 bits of the address here (and assuming that the high 32 bits are zero).
Cheers
---Dave
> + }
> +
> + return;
> +}
> +
> static int is_space(char c)
> {
> return c == ' ';
> @@ -598,4 +646,6 @@ args_done:
> atag_append(&atagp, ATAG_NONE, 0, 0);
>
> update_fdt(&phys, info);
> +
> + configure_from_fdt(info);
> }
> --
> 1.7.10.4
>
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
