From mboxrd@z Thu Jan  1 00:00:00 1970
From: itooo@itooo.com (Greg)
Date: Wed, 22 May 2013 13:46:53 +0200
Subject: [PATCH 0/9] Switch internal registers address to 0xF1 on Armada
 370/XP
In-Reply-To: <519C9779.1040104@gmail.com>
References: <1369132414-18959-1-git-send-email-thomas.petazzoni@free-electrons.com>
 <20130521193803.GA20882@1wt.eu> <519C9779.1040104@gmail.com>
Message-ID: <519CB02D.3090304@itooo.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

Le 22/05/2013 12:01, Sebastian Hesselbarth a ?crit :
> On 05/21/2013 09:38 PM, Willy Tarreau wrote:
>> On Tue, May 21, 2013 at 12:33:25PM +0200, Thomas Petazzoni wrote:
>> (...)
>>> As it was explained above, we unfortunately can't read the
>>> current base address of the internal register window, so we need a
>>> different mechanism to know if the bootloader has done the remapping
>>> at 0xF1000000 (new generation bootloader) or has left the internal
>>> registers at 0xD0000000 (old generation bootloader).
>>
>> Just out of curiosity, what happens if you touch the register at the
>> wrong address ? I mean, if you blindly write to the 0xD0xxxxxx address
>> that you want the registers to be mapped at 0xF1xxxxxx. Will the chip
>> hang, will it silently ignore the sequence ? Because maybe you don't
>> need to detect using CP15 whether the remapping was done, you could
>> simply perform it unconditionally.
>
> Willy,
>
> I tried the above on Dove which is possibly affected by the workaround
> as it is also ARMv7. In Dove you can access unmapped addresses without
> any problems, but as Thomas already stated Armada 370/XP just hang.
> Must be some default AHB slave in Dove, which Armada 370/XP is missing.

According to functionnal specs, when accessing a unmapped address, you 
will trigger a "coherency fabric" error named AccessErr (register 
0x00020258).
I guess this will trigger an interrupt which will hang the CPU if the 
vector is not handled.

Greg