[PATCH 1/2] mtd: spi-nor: introduce SNOR_ID3()

[Tudor.Ambarus at microchip.com]

On 7/19/22 10:57, Michael Walle wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
> Am 2022-07-19 09:33, schrieb Tudor.Ambarus at microchip.com:
>> On 7/19/22 10:07, Michael Walle wrote:
>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know
>>> the content is safe
>>>
>>> Am 2022-07-19 07:57, schrieb Tudor.Ambarus at microchip.com:
>>>> On 5/10/22 17:02, Michael Walle wrote:
>>>>> EXTERNAL EMAIL: Do not click links or open attachments unless you
>>>>> know
>>>>> the content is safe
>>>>>
>>>>> Up until now, flashes were defined by specifying the JEDEC ID, the
>>>>> sector size and the number of sectors. This can be read by parsing
>>>>> the
>>>>> SFDP, we don't have to specify it. Thus provide a new macro
>>>>> SNOR_ID3()
>>>>> which just takes the JEDEC ID and implicitly set .parse_sfdp = true.
>>>>> All
>>>>> new flashes which have SFDP should use this macro.
>>>>
>>>> I like the idea, but you need to refine it a bit.
>>>> Your assumptions are true only for flashes that are compliant with
>>>> SFDP
>>>> revB or
>>>> later because params->page_size is initialized by querying BFPT DWORD
>>>> 11. I think it would be good to specify this in the comment section.
>>>
>>> Sure.
>>>
>>>> Also, I think you introduce
>>>> a bug in spi_nor_select_erase() when
>>>> CONFIG_MTD_SPI_NOR_USE_4K_SECTORS
>>>> is not
>>>> selected. wanted_size will be zero, will you select an invalid erase
>>>> type?
>>>
>>> You mean to squeeze [1] into this one? If so, sure.
>>>
>>> -michael
>>>
>>> [1]
>>> https://lore.kernel.org/linux-mtd/20220716000643.3541839-1-quic_jaehyoo@quicinc.com/
>>
>> No, these are orthogonal. If you keep wanted_size to zero, then
>> spi_nor_select_uniform_erase() will return NULL, doesn't it?
> 
> No, have a look at the second condition
> 
> if (!erase && tested_erase->size)
>   erase = ..
> 
> So it will return the first non-empty slot. Thus it will
> only return NULL if all the slots are empty (given the
> fix is included).
> 
> Actually, I'd have expected that to mask out an erase
> type, you clear the corresponding bit in uniform_erase_type,
> in which case the for loop in spi_nor_select_uniform_erase()
> would have just worked. But apparently there are two differnt
> mechanism here to mark an entry as unused, either the size
> is zero or the bit is not set. But that is a topic for another
> patch.


Right, I remember I leaned towards using just the erase mask to mask
out an erase, but I have to reassess this. Here's a patch that is
related and I left behind:
https://patchwork.ozlabs.org/project/linux-mtd/patch/20211119081412.29732-1-alexander.sverdlin@nokia.com/

> 

something else that looks wrong:
drivers/mtd/spi-nor/swp.c:              return nor->info->sector_size <<
drivers/mtd/spi-nor/swp.c:              return nor->info->sector_size;

How do we progress on this? I like the SNOR_ID3 idea, but I think it
should have a different form. Do you want to spend more time on this
or do you think I should invest more time on this?


> -michael
> 
>> Maybe to adapt the code to something like
>> diff --git a/drivers/mtd/spi-nor/core.c b/drivers/mtd/spi-nor/core.c
>> index 21cefe2864ba..dd6cd852d1ef 100644
>> --- a/drivers/mtd/spi-nor/core.c
>> +++ b/drivers/mtd/spi-nor/core.c
>> @@ -2148,7 +2148,7 @@ static int spi_nor_select_erase(struct spi_nor
>> *nor)
>>         struct spi_nor_erase_map *map = &nor->params->erase_map;
>>         const struct spi_nor_erase_type *erase = NULL;
>>         struct mtd_info *mtd = &nor->mtd;
>> -       u32 wanted_size = nor->info->sector_size;
>> +       u32 wanted_size = nor->params->sector_size;
>>
>> and fill nor->params->sector_size even when no SFDP
>>
>> Also, params->size = (u64)info->sector_size * info->n_sectors; from
>> spi_nor_init_default_params() becomes superfluous. I would check
>> the fields that I don't initialize in flash_info with SNOR_ID3
>> and check how I can mitigate their absence throughout the code.