[PATCH 8/8] mtd: spi-nor: spansion: Add support for Infineon

[Takahiro Kuwano]

On 8/8/2022 4:34 PM, Tudor.Ambarus at microchip.com wrote:
> On 8/8/22 09:41, Takahiro Kuwano wrote:
>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>>
>> On 8/8/2022 3:08 PM, Tudor.Ambarus at microchip.com wrote:
>>> On 8/8/22 08:42, Takahiro Kuwano wrote:
>>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>>>>
>>>> On 8/8/2022 1:47 PM, Tudor.Ambarus at microchip.com wrote:
>>>>> On 8/6/22 09:34, tkuw584924 at gmail.com wrote:
>>>>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>>>>>>
>>>>>> From: Takahiro Kuwano <Takahiro.Kuwano at infineon.com>
>>>>>
>>>>> Hi!
>>>>>
>>>>>>
>>>>>> s25hl02gt and s25hs02gt
>>>>>>
>>>>>> Add ID, flags, and fixup for s25hl02gt and s25hs02gt.
>>>>>> These parts are
>>>>>>   - Dual-die package parts
>>>>>>   - Not support chip erase
>>>>>>   - 4-byte addressing mode by default
>>>>>
>>>>> CFR2N[7] CFR2V[7] says that: "For the DDP or QDP devices, if ADRBYT = 0
>>>>> only the first 128 Mb of die 1 can be accessed."
>>>>> So there are flashes of the same family that are by default in 3 byte address
>>>>> mode. You added support just for a subset of them and used a generic name,
>>>>> which is not accurate, right?
>>>>>
>>>> We added model #15 (3-byte address mode by default) to address special
>>>> requirement from a customer who needs to use bootrom with 3-byte addressing.
>>>> Anyway, I overlooked model # difference. Thanks for pointing out this.
>>>>
>>>>> Can we instead make an algorithm to determine the current address mode?
>>>>>
>>>> I have just found that we can distinguish model # via BFPT DWORD16.
>>>> If Hardware reset, Software reset, or Power cycle can exit 4-byte address
>>>> mode, that means the device is 3-byte address mode by default.
>>>
>>> I don't think this will help us. It doesn't matter the default mode if you
>>> have a non volatile register that can be updated and changes the default
>>> mode.
>>>
>>> Are there any registers/data that can be read successively in 3 byte addr mode
>>> and then in 4 byte addr mode? We'll then compare what we receive from the flash
>>> with a known value and determine the mode.
>>>
>> As we discussed before [0], if address mode in the controller and device are
> 
> I remember, yes, but without determining the mode, the driver will work only
> with flashes that come with the factory settings. The driver will be unusable
> if someone changes the address mode in a non volatile way, right?
> 
Yes, right.

>> different, the read data will be undetermined.
>>
>> But if we really want...
>> Compare SR1 data read by RDSR1(05h - No Addr) and RDAR(65h - Addr 0).
>> In most cases (without block protection), SR1=00h. The value of 00h would be
>> awkward to determine if this is 'real' output from Flash or not. So, use> WREN(06h) and WRDI(04h) that flips BIT(1) in SR1.
> 
> Would be good to have more fixed/OPT-like bits, or if we could change more bits
> to increase the chances to not hit just some undetermined data.
>>
>> Therefore, something like:
>> 1) RDSR1
>> 2) RDAR with 3-byte addr (000000h)
>> 3) If #1 == #2
>>         4) WREN
>>         5) RDAR with 3-byte addr (000000h)
>>         6) BIT(1) is SR1==1?
>>         ...
>>
>> Or simply WREN -> RDAR -> WRDI -> RDAR then check if only BIT(1) is toggled.
> 
> Both may work, yes, but making the assumption on only one bit is fragile.
> Can we use the Read Any Register Command with 3 and 4 byte address modes and
> compare the values? Are there any registers with fixed values?
>
Register values can vary because it's register:)

So... let's use Data Integrity Check CRC registers. These registers do not
have fixed values but we can calculate expected values by offline. Read
several bytes (>=4) from Flash array with Read(03h) then calculate CRC by
crc32(). Issue Data integrity Check command (5Bh) followed by start and
end address (4-byte for each), wait till ready. Read calculated CRC by
Read Any Register in 3 and 4 byte address (00800095h~0080098h) then compare
the crc32() result and register read result.