Re: flash driver interface improvement

[Mark Salter <msalter at redhat dot com>]

>>>>> Robin Farine writes:

> On Fri, 2002-08-23 at 13:46, Mark Salter wrote:
>> 
>> I don't have a comment about this specific proposal, but I have
>> been thinking recently about the current limitations of the
>> flash drivers. In addition to the above, the current flash driver
>> layer does not support multiple flash regions and it lacks support
>> for NAND flash. I am planning on adding these things in the not
>> too distant future. If anyone else has other ideas about missing
>> feature, I would love to hear them.

> A thought I had a long time ago when my board crashed hard (power cycle
> required) during flash operations ...

> Replacing direct access to memory mapped flash with something like 'int
> flash_read(void* dest, size_t length, off_t offset)' might encapsulate
> hardware details such as:

>         * strange things like paged flashes
>         * some chips do not allow any access during sector erase / write
>           operations, others allow access to sectors non currently
>           involved in a write / erase operation; the driver could handle
>           this internally and return an error code like EBUSY when
>           approriate.

Yes. I think a read interface is required for NAND (paged flashes?)support.
It is probably advisable to use that interface for general access on other
types of flash as you note. 

--Mark