On Sat, Sep 15, 2012 at 05:42:41PM +0000, Arnd Bergmann wrote:

On Friday 14 September 2012, Fabio Estevam wrote:

quoted

On Fri, Sep 14, 2012 at 6:34 PM, Arnd Bergmann [off-list ref] wrote:

quoted

ARM is moving to stricter checks on readl/write functions,
so we need to use the correct types everywhere.

This found a bug in mach-armadillo5x0.c, where we attempt mmio
on the MXC_CCM_RCSR address that is currently defined to 0xc
and consequently causes an illegal address access.

...

quoted

        /* set NAND page size to 2k if not configured via boot mode pins */
-       __raw_writel(__raw_readl(MXC_CCM_RCSR) | (1 << 30), MXC_CCM_RCSR);
+       /* FIXME __raw_writel(__raw_readl(MXC_CCM_RCSR) | (1 << 30), MXC_CCM_RCSR); */

Good catch, the correct access would be:
__raw_readl(mx3_ccm_base + MXC_CCM_RCSR) ...

I will fix this after your series reaches linux-next.

I think I'd prefer it if you can fix it on the current kernel. This seems to
be a serious bug that we actually want to fix in 3.6 or maybe backport to
the stable series (I don't know when it was introduced). I'll just drop
this hunk from my patch then.

This was introduced between 3.4 and 3,5 with:

commit eb92044eb3d59d29c9812e85e3a4bf41f6f38e3a
Author: Sascha Hauer [off-list ref]
Date:   Tue Apr 3 12:42:27 2012 +0200

    ARM i.MX3: Make ccm base address a variable
    
    Instead of having a cpu_is_* in each ccm register access it
    is more efficient to make it a variable.
    
    Signed-off-by: Sascha Hauer [off-list ref]


-- 
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 |