Re: [Qemu-devel] [PATCH] block: mask NOR flash buffered write length

[Roy Franz]

On Fri, Oct 18, 2013 at 7:01 AM, Peter Maydell <address@hidden> wrote:
> On 18 October 2013 14:54, Roy Franz <address@hidden> wrote:
>> On Fri, Oct 18, 2013 at 6:36 AM, Peter Maydell <address@hidden> wrote:
>>> Probably instead of a single "width" property we should have two,
>>> similar to the device tree binding's pair:
>>>  - bank-width : Width (in bytes) of the bank.  Equal to the
>>>    device width times the number of interleaved chips.
>>>  - device-width : (optional) Width of a single mtd chip.  If
>>>    omitted, assumed to be equal to 'bank-width'.
>
>>> However I'm not very familiar with how flash hardware works...
>>     You are correct - we really do want to mask based on the device
>> width, as that is what the
>> actual flash chips will see.  Lacking the device width I used the
>> writeblock size.  Thinking about this more,
>> this will not work for 8 bit devices used together, as the mask size
>> will be greater than 8 bits and the writeblock size
>> will be mis-interpreted like it is now.
>> I'll work on adding a device-size property to the pflash*
>> implementations.  It looks like this will affect about 20 platforms.
>> For the platforms that I am not familiar with I plan just set
>> bank-width==device-width as that should result in the unchanged
>> behavior.
>
> Yes, you should make the default for the device-width property
> be to be the same as the bank-width, since that's what we
> currently implement; then we can just change the platforms
> where we know that's wrong.
>
> NB: probably best to leave the existing 'width' property with
> the name it has, rather than renaming it to 'bank-width'.
>
> thanks
> -- PMM

Thanks Peter.  I'm not familiar with the "properties" and how they are
used.  I think that
the device width is likely only of interest internally, so I won't add
a device-width property.

Roy