[PATCH 1/2] mtd: spi-nor: make n_sectors in flash_info 32 bit wide

[Bastian Stender]

Hi Cyrille,

On 01/16/2017 06:34 PM, Cyrille Pitchen wrote:
> Le 16/01/2017 à 16:12, Bastian Stender a écrit :
>> Since MRAM chips (like the Everspin mr25h40) are not sector
>> organized they should be defined as n_sectors * 1 byte sectors. To
>> be able to store the higher number of sectors n_sectors should be
>> an unsigned 32 bit integer just like sector_size.
>>
>
> Could you comment a little bit more on what this series improves or
> fixes?

Sure. I want to create multiple writeable partitions on a mr25h40 MRAM 
chip. This does not currently work because drivers/mtd/mtdpart.c makes 
sure that partitions start on erase block boundaries and end on erase 
blocks. That is obviously not the case for multiple partitions with the 
current definition:

{ "mr25h40",  CAT25_INFO(1, 512 * 1024, 256, 3, SPI_NOR_NO_ERASE | 
SPI_NOR_NO_FR) },

So I get

mtd: partition "mtd2" doesn't start on an erase block boundary -- force 
read-only

Adding a check for SPI_NOR_NO_ERASE does not seem to be the way to go:

http://lists.infradead.org/pipermail/linux-mtd/2017-January/071386.html

Setting the erasesize to 1 neither:

http://lists.infradead.org/pipermail/linux-mtd/2014-March/052410.html

I could use SECT_4K, but I'd like to use the advantage of MRAM without 
limitations.

> Swapping the info->sector_size and info->n_sectors doesn't change the
> value of mtd->size: mtd->size = info->sector_size * info->n_sectors;

To be clear: mtd->size is perfectly fine, no need to change it.

> info->n_sectors is used nowhere else whereas info->sector_size is
> only used once more to initialized mtd->erasesize but due to the
> SPI_NOR_NO_ERASE info->flags, mtd->erasesize should be never used.

mtd->erasesize gets used even with SPI_NOR_NO_ERASE set in the check 
mentioned above in mtd_mod_by_eb().

> However changing the type of n_sectors from u16 to unsigned int
> would increase the size of the struct flash_info hence the size of
> the spi_nor_ids[] array. So I don't know whether it's worth it.
>
> However maybe swapping the values would better respect the meaning
> of sector "size" and "number".

MRAM is not page, sector, or block organized memory, so I do not know 
about this one.

Regards,
Bastian Stender

-- 
Pengutronix e.K.
Industrial Linux Solutions
http://www.pengutronix.de/
Peiner Str. 6-8, 31137 Hildesheim, Germany
Amtsgericht Hildesheim, HRA 2686