Wrong flash type in m25p80 driver

[Carl-Daniel Hailfinger]

On 17.05.2010 06:54, Mike Frysinger wrote:
> On Sun, May 16, 2010 at 15:55, Shachar Shemesh wrote:
>   
>> Mike Frysinger wrote:
>>     
>>> so your answer is "there is no problem with calling it NOR flash"
>>>
>>> if it looks like a duck, quacks like a duck, and walks like a duck,
>>> then why waste time calling it a DataDuck ?  for all intents and
>>> purposes, all the kernel code/utilities that work with the flashes can
>>> treat these SPI flashes as NOR flashes and everything works fine.
>>>       
>> Actually, that is not the case. Not even remotely.
>>
>> NOR flashes (and the m25p80 driver broadcasts it this way) have the ability
>> to change individual bytes (and, some flashes, individual bits). The SPI
>> flashes need to erase an entire block. NOR flashes are accessible via the
>> data bus, these require complicated SPI transactions. They neither quack nor
>> look anywhere near the same.
>>     
>
> you havent shown how the current behavior results in non-working
> flashes.  you probably can claim that things dont work as optimally as
> possible according to the hardware, but as you noted, people work on
> what interests/annoys them rather than what other people report.
>   

ST M25P80 can change individual bytes.


>> You might want to claim that they are NAND flashes, but, again, those are
>> somewhat different. Dataflashes are the closest these come to, as far as I
>> can tell.
>>     
>
> it doesnt make sense to lump SPI and NAND together because NAND allows
> for bad blocks.  both NOR and SPI are guaranteed to be good throughout
> or the device is dead.  so for all practical uses thus far, SPI and
> NOR operate the same way.
>   

Yes.


>> I am not so sure about the "properties which other SPI flashes do not". Not,
>> at least, according to my research (which does not beat experience, but does
>> beat obscure abstract statements).
>>     
>
> iirc, dataflashes allow programming on individual bytes just like NOR
> flashes.  most serial flashes (like the ST micro that started the SPI
> flash driver in the first place) have to be programmed on a page
> basis.  dataflashes also have more flexible erase structures (4kb,
> 32kb, or 64kb) while ST micro can only be erased at 64kb sectors.
>   

The ST M25P80 has a minimum write size of 1 bit (datasheet is a bit
unclear, could also be 1 byte) and a maximum write size of 256 bytes.


> my point is just that calling all SPI flashes "dataflashes" is
> inherently wrong since only Atmel makes dataflashes.  it's also a bit
> funny since you started off the thread with the complaint that calling
> SPI flashes NOR flashes was wrong because SPI flashes arent NOR
> flashes.  if you're going to declare a new category for SPI flashes,
> then it should be something that applies to all the current SPI
> flashes (just look in the m25p80.c driver to see all the ones
> supported).
>   

DataFlash is special because some of the chips have a user configurable
page size: 512 or 528 Bytes per page.

There's always the option of looking at how flashrom
<http://www.flashrom.org/> handles those chips. flashrom an
OS-independent userspace tool specialized on chips which are used for
BIOS/firmware, but it handles some other flash chips as well.


Regards,
Carl-Daniel

-- 
http://www.hailfinger.org/