Re: [RFC PATCH v1 6/6] mtd: spi-nor: introduce support for displaying deprecation message

[Pratyush Yadav <pratyush@kernel.org>]

On Wed, Apr 17 2024, Michael Walle wrote:

> Hi,
>
> On Wed Apr 17, 2024 at 4:36 PM CEST, Pratyush Yadav wrote:
>> On Fri, Apr 12 2024, Michael Walle wrote:
>>
>> > SPI-NOR will automatically detect the attached flash device most of the
>> > time. We cannot easily find out if boards are using a given flash.
>> > Therefore, introduce a (temporary) flag to display a message on boot if
>>
>> Why temporary? There will always be a need to deprecate one flash or
>> another. Might as well keep the flag around.
>
> Mh, yes I agree. That also means that this flag will not have any
> users (most) of the time (hopefully ;) ).
>
>> Also, this patch/series does not add any users of the deprecated flag.
>> If you have some flashes in mind, it would be good to add them to the
>> patch/series.
>
> This is just an RFC to see if whether you Tudor agree with me :) But
> I was about to add it to the evervision/cypress FRAMs.
>
>> I like the idea in general. Do you think we should also print a rough
>> date for the deprecation as well?
>
> Might make sense, any suggestions?

How about a simple string to flash_info?

/**
 * struct flash_info - SPI NOR flash_info entry.
 * @id:   pointer to struct spi_nor_id or NULL, which means "no ID" (mostly
 *        older chips).
 * @name: (obsolete) the name of the flash. Do not set it for new additions.
 * @size:           the size of the flash in bytes.
 * @deprecation_date: The date after which the support for this flash will be
 *                    removed.
 * [...]
 */
struct flash_info {
	char *name;
	const struct spi_nor_id *id;
	char *deprecation_date;
	[...]
}

And then in everspin.c for example,

	{
		.name = "mr25h128",
		.size = SZ_16K,
		.sector_size = SZ_16K,
		.addr_nbytes = 2,
		.flags = SPI_NOR_NO_ERASE | SPI_NOR_NO_FR,
		.deprecation_date = "2025-01-01",
	}, {

And in spi_nor_get_flash_info() (changed some wording of the message):

	info = jinfo ?: info;

	if (info->deprecation_date)
		pr_warn("Your board or device tree is using a SPI NOR flash (%s) with\n"
			"deprecated driver support. It can be removed in any kernel\n"
			"version after %s. If you feel this shouldn't be the case, please contact\n"
			"us at linux-mtd@lists.infradead.org\n", info->name,
			info->deprecation_date);

	return info;

This would also remove the need for SPI_NOR_DEPRECATED. But it would
make the flash_info 4 or 8 bytes larger.

What do you think?

>
>> > support for a given flash device is scheduled to be removed in the
>> > future.
>> >
>> > Signed-off-by: Michael Walle <mwalle@kernel.org>
>> > ---
>> >  drivers/mtd/spi-nor/core.c | 12 ++++++++++++
>> >  drivers/mtd/spi-nor/core.h |  1 +
>> >  2 files changed, 13 insertions(+)
>> >
>> > diff --git a/drivers/mtd/spi-nor/core.c b/drivers/mtd/spi-nor/core.c
>> > index 58d310427d35..a294eef2e34a 100644
>> > --- a/drivers/mtd/spi-nor/core.c
>> > +++ b/drivers/mtd/spi-nor/core.c
>> > @@ -3312,6 +3312,7 @@ static const struct flash_info *spi_nor_get_flash_info(struct spi_nor *nor,
>> >  						       const char *name)
>> >  {
>> >  	const struct flash_info *jinfo = NULL, *info = NULL;
>> > +	const char *deprecated = NULL;
>> >  
>> >  	if (name)
>> >  		info = spi_nor_match_name(nor, name);
>> > @@ -3326,6 +3327,17 @@ static const struct flash_info *spi_nor_get_flash_info(struct spi_nor *nor,
>> >  			return jinfo;
>> >  	}
>> >  
>> > +	if (info && (info->flags & SPI_NOR_DEPRECATED))
>> > +		deprecated = info->name;
>> > +	else if (jinfo && (jinfo->flags & SPI_NOR_DEPRECATED))
>> > +		deprecated = jinfo->name;
>> > +
>> > +	if (deprecated)
>> > +		pr_warn("Your board or device tree is using a SPI NOR flash (%s) with\n"
>> > +			"deprecated driver support. It will be removed in future kernel\n"
>>
>> Nit: "removed in a future kernel version"
>>
>> > +			"version. If you feel this shouldn't be the case, please contact\n"
>> > +			"us at linux-mtd@lists.infradead.org\n", deprecated);
>> > +
>>
>> Hmm, this isn't so nice. I'd suggest doing something like:
>>
>> 	/*
>>          * If caller has specified name of flash model that can normally be
>>          * ...
>>          */
>> 	info = jinfo ?: info;
>>
>> 	if (info->flags & SPI_NOR_DEPRECATED)
>>         	pr_warn(...);
>
> Actually, I had that, *but* I was thinking we might only check the
> detected flash and not the one specified in the device tree. But
> thinking about that again, I guess it makes sense because:
>  - that's the actually used flash driver
>  - having jinfo != info will print that other warning, thus this
>    case is hopefully very unlikely.
>
>>
>> 	return info;
>>
>> >  	/*
>> >  	 * If caller has specified name of flash model that can normally be
>> >  	 * detected using JEDEC, let's verify it.
>> > diff --git a/drivers/mtd/spi-nor/core.h b/drivers/mtd/spi-nor/core.h
>> > index 8552e31b1b07..0317d8e253f4 100644
>> > --- a/drivers/mtd/spi-nor/core.h
>> > +++ b/drivers/mtd/spi-nor/core.h
>> > @@ -524,6 +524,7 @@ struct flash_info {
>> >  #define SPI_NOR_NO_ERASE		BIT(6)
>> >  #define SPI_NOR_QUAD_PP			BIT(8)
>> >  #define SPI_NOR_RWW			BIT(9)
>> > +#define SPI_NOR_DEPRECATED		BIT(15)
>>
>> If you do agree with my suggestion of making it permanent, would it make
>> more sense to make it BIT(10) instead. Or BIT(9) once you move up the
>> others because we no longer have BIT(7).
>
> Or just BIT(7) and avoid any code churn :)

Yep, that works.

>
> -michael
>
>>
>> >  
>> >  	u8 no_sfdp_flags;
>> >  #define SPI_NOR_SKIP_SFDP		BIT(0)
>

-- 
Regards,
Pratyush Yadav