diff mbox series

[4/9] mtd: devices: add AT24 eeprom support

Message ID 20240701-b4-v6-10-topic-usbc-tcpci-v1-4-3fd5f4a193cc@pengutronix.de
State New
Headers show
Series AT24 EEPROM MTD Support | expand

Commit Message

Marco Felsch July 1, 2024, 1:53 p.m. UTC
At the moment there are three ways to access EEPROM content from
user-space:

  1st) via the single nvmem device (rw)
  2nd) via the single 'eeprom' device (rw)
  3th) via nvmem-cells (r)

EEPROMs can become quite large nowadays (>=64K). Exposing such devices
as single device isn't always sufficient. There may be partitions which
require different access permissions. Also write access always need to
to verify the offset.

Port the current misc/eeprom/at24.c driver to the MTD framework since
EEPROMs are memory-technology devices and the framework already supports
partitioning. This allow using of-paritions like we do for SPI-NOR
devices already:

&eeprom {
	compatible = "atmel,24c02";
	reg = <0x52>;
	pagesize = <16>;
	#address-cells = <1>;
	#size-cells = <0>;

	partitions {
		compatible = "fixed-partitions";
		#address-cells = <1>;
		#size-cells = <1>;
		label = "board-eeprom";

		otp@0 {
			compatible = "nvmem-cells";
			label = "manufaturing";
			reg = <0x0 0xe0>;

			nvmem-layout {
				compatible = "fixed-layout";
				#address-cells = <1>;
				#size-cells = <1>;

				/* MACs stored in ASCII */
				ethmac1: mac-address@0 {
					reg = <0x0 0xc>;
				};

				ethmac2: mac-address@c {
					reg = <0xc 0xc>;
				};

				test@d8 {
					reg = <0xd8 0x8>;
				};
			};
		j};

		user@e0 {
			reg = <0xe0 0x100>;
		};
	};
};

The porting was straight forward:
 - at24_read/write funcs are adapted to be MTD compatible
 - (re-)adding the write-protection gpio handling since the devices can
   now be accessed via NVMEM and MTD API
 - Replace the NVMEM with MTD registration. The mtdcore is taking care
   of exposing the NVMEM and eeprom device correctly ensure backward
   compatibility.

Signed-off-by: Marco Felsch <m.felsch@pengutronix.de>

Changelog:
v2:
- resync misc/eeprom/at24.c with mtd/at24.c driver
- add backward compatibility
- drop misc/eeprom/at24.c driver
- adapt Kconfig EEPROM_AT24 Kconfig entry
- shift infrastructure support into separate patch
- adapt commit message
---
 MAINTAINERS                                 |   2 +-
 drivers/misc/eeprom/Kconfig                 |  30 +------
 drivers/misc/eeprom/Makefile                |   1 -
 drivers/mtd/devices/Kconfig                 |  31 +++++++
 drivers/mtd/devices/Makefile                |   1 +
 drivers/{misc/eeprom => mtd/devices}/at24.c | 122 +++++++++++++++-------------
 6 files changed, 103 insertions(+), 84 deletions(-)

Comments

Tudor Ambarus July 1, 2024, 4:14 p.m. UTC | #1
On 7/1/24 2:53 PM, Marco Felsch wrote:
> EEPROMs can become quite large nowadays (>=64K). Exposing such devices
> as single device isn't always sufficient. There may be partitions which
> require different access permissions. Also write access always need to
> to verify the offset.
> 
> Port the current misc/eeprom/at24.c driver to the MTD framework since
> EEPROMs are memory-technology devices and the framework already supports

I was under the impression that MTD devices are tightly coupled by erase
blocks. But then we see MTD_NO_ERASE, so what are MTD devices after all?

> partitioning. This allow using of-paritions like we do for SPI-NOR
> devices already:
Pratyush Yadav July 2, 2024, 1:41 p.m. UTC | #2
On Mon, Jul 01 2024, Tudor Ambarus wrote:

> On 7/1/24 2:53 PM, Marco Felsch wrote:
>> EEPROMs can become quite large nowadays (>=64K). Exposing such devices
>> as single device isn't always sufficient. There may be partitions which
>> require different access permissions. Also write access always need to
>> to verify the offset.
>> 
>> Port the current misc/eeprom/at24.c driver to the MTD framework since
>> EEPROMs are memory-technology devices and the framework already supports
>
> I was under the impression that MTD devices are tightly coupled by erase
> blocks. But then we see MTD_NO_ERASE, so what are MTD devices after all?

I was curious as well so I did some digging.

The Kconfig help says:

    Memory Technology Devices are flash, RAM and similar chips, often
    used for solid state file systems on embedded devices [...]

The FAQ on the MTD documentation [0] says:

    Unix traditionally only knew block devices and character devices.
    Character devices were things like keyboards or mice, that you could
    read current data from, but couldn't be seek-ed and didn't have a size.
    Block devices had a fixed size and could be seek-ed. They also happened
    to be organized in blocks of multiple bytes, usually 512.

    Flash doesn't match the description of either block or character
    devices. They behave similar to block device, but have differences. For
    example, block devices don't distinguish between write and erase
    operations. Therefore, a special device type to match flash
    characteristics was created: MTD.

    So MTD is neither a block nor a char device. There are translations to
    use them, as if they were. But those translations are nowhere near the
    original, just like translated Chinese poems.

And in the section below, it lists some properties of an MTD device:

    - Consists of eraseblocks.
    - Eraseblocks are larger (typically 128KiB).
    - Maintains 3 main operations: read from eraseblock, write to
      eraseblock, and erase eraseblock.
    - Bad eraseblocks are not hidden and should be dealt with in
      software.
    - Eraseblocks wear-out and become bad and unusable after about 10^3
      (for MLC NAND) - 10^5 (NOR, SLC NAND) erase cycles.

This does support the assumption you had about MTD devices being tightly
coupled with erase block. It also makes it quite clear that an EEPROM is
not MTD -- since EEPROMs are byte-erasable.

Of course, the existence of MTD_NO_ERASE nullifies a lot of
these points. So it seems the subsystem has evolved. MTD_NO_ERASE was
added by 92cbfdcc3661d ("[MTD] replace MTD_RAM with MTD_GENERIC_TYPE")
in 2006, but this commit only adds the flag. The functionality of "not
requiring an explicit erase" for RAM devices has existed since the start
of the git history at least.

I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
drivers under a single interface. I am not sure what came of it though,
since I can't find any patches that followed up with the proposal.

Overall, I'd say that while originally MTD was written with flash
devices with erase blocks in mind, the subsystem seems to have evolved
with time to include other types of devices.

I don't see anything obviously wrong with adding EEPROMs to the type of
devices in MTD as well. It doesn't seem to be too invasive to the
subsystem (I do see some dubious code when skimming through the patches,
but nothing unfixable). And the EEPROM drivers can get a common
interface. The other option would be to create a separate subsystem for
EEPROMs, but perhaps that would just lead to a bunch of code being
duplicated.

I'd like to hear if somebody thinks otherwise, and sees reasons to _not_
do this.

[0] http://www.linux-mtd.infradead.org/faq/general.html
[1] https://lore.kernel.org/linux-mtd/20130705201118.GM2959@lukather/
Maxime Ripard July 2, 2024, 1:56 p.m. UTC | #3
On Tue, Jul 02, 2024 at 03:41:52PM GMT, Pratyush Yadav wrote:
> On Mon, Jul 01 2024, Tudor Ambarus wrote:
> 
> > On 7/1/24 2:53 PM, Marco Felsch wrote:
> >> EEPROMs can become quite large nowadays (>=64K). Exposing such devices
> >> as single device isn't always sufficient. There may be partitions which
> >> require different access permissions. Also write access always need to
> >> to verify the offset.
> >> 
> >> Port the current misc/eeprom/at24.c driver to the MTD framework since
> >> EEPROMs are memory-technology devices and the framework already supports
> >
> > I was under the impression that MTD devices are tightly coupled by erase
> > blocks. But then we see MTD_NO_ERASE, so what are MTD devices after all?
> 
> I was curious as well so I did some digging.
> 
> The Kconfig help says:
> 
>     Memory Technology Devices are flash, RAM and similar chips, often
>     used for solid state file systems on embedded devices [...]
> 
> The FAQ on the MTD documentation [0] says:
> 
>     Unix traditionally only knew block devices and character devices.
>     Character devices were things like keyboards or mice, that you could
>     read current data from, but couldn't be seek-ed and didn't have a size.
>     Block devices had a fixed size and could be seek-ed. They also happened
>     to be organized in blocks of multiple bytes, usually 512.
> 
>     Flash doesn't match the description of either block or character
>     devices. They behave similar to block device, but have differences. For
>     example, block devices don't distinguish between write and erase
>     operations. Therefore, a special device type to match flash
>     characteristics was created: MTD.
> 
>     So MTD is neither a block nor a char device. There are translations to
>     use them, as if they were. But those translations are nowhere near the
>     original, just like translated Chinese poems.
> 
> And in the section below, it lists some properties of an MTD device:
> 
>     - Consists of eraseblocks.
>     - Eraseblocks are larger (typically 128KiB).
>     - Maintains 3 main operations: read from eraseblock, write to
>       eraseblock, and erase eraseblock.
>     - Bad eraseblocks are not hidden and should be dealt with in
>       software.
>     - Eraseblocks wear-out and become bad and unusable after about 10^3
>       (for MLC NAND) - 10^5 (NOR, SLC NAND) erase cycles.
> 
> This does support the assumption you had about MTD devices being tightly
> coupled with erase block. It also makes it quite clear that an EEPROM is
> not MTD -- since EEPROMs are byte-erasable.
> 
> Of course, the existence of MTD_NO_ERASE nullifies a lot of
> these points. So it seems the subsystem has evolved. MTD_NO_ERASE was
> added by 92cbfdcc3661d ("[MTD] replace MTD_RAM with MTD_GENERIC_TYPE")
> in 2006, but this commit only adds the flag. The functionality of "not
> requiring an explicit erase" for RAM devices has existed since the start
> of the git history at least.
> 
> I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
> EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
> drivers under a single interface. I am not sure what came of it though,
> since I can't find any patches that followed up with the proposal.

That discussion led to drivers/nvmem after I started to work on
some early prototype, and Srinivas took over that work.

Maxime
Pratyush Yadav July 2, 2024, 2:15 p.m. UTC | #4
On Tue, Jul 02 2024, Maxime Ripard wrote:

> On Tue, Jul 02, 2024 at 03:41:52PM GMT, Pratyush Yadav wrote:
>> On Mon, Jul 01 2024, Tudor Ambarus wrote:
>> 
>> > On 7/1/24 2:53 PM, Marco Felsch wrote:
>> >> EEPROMs can become quite large nowadays (>=64K). Exposing such devices
>> >> as single device isn't always sufficient. There may be partitions which
>> >> require different access permissions. Also write access always need to
>> >> to verify the offset.
>> >> 
>> >> Port the current misc/eeprom/at24.c driver to the MTD framework since
>> >> EEPROMs are memory-technology devices and the framework already supports
>> >
>> > I was under the impression that MTD devices are tightly coupled by erase
>> > blocks. But then we see MTD_NO_ERASE, so what are MTD devices after all?
>> 
>> I was curious as well so I did some digging.
>> 
[...]
>> 
>> I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
>> EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
>> drivers under a single interface. I am not sure what came of it though,
>> since I can't find any patches that followed up with the proposal.
>
> That discussion led to drivers/nvmem after I started to work on
> some early prototype, and Srinivas took over that work.

So would you say it is better for EEPROM drivers to use nvmem instead of
moving under MTD?
Maxime Ripard July 2, 2024, 2:34 p.m. UTC | #5
On Tue, Jul 02, 2024 at 04:15:20PM GMT, Pratyush Yadav wrote:
> On Tue, Jul 02 2024, Maxime Ripard wrote:
> 
> > On Tue, Jul 02, 2024 at 03:41:52PM GMT, Pratyush Yadav wrote:
> >> On Mon, Jul 01 2024, Tudor Ambarus wrote:
> >> 
> >> > On 7/1/24 2:53 PM, Marco Felsch wrote:
> >> >> EEPROMs can become quite large nowadays (>=64K). Exposing such devices
> >> >> as single device isn't always sufficient. There may be partitions which
> >> >> require different access permissions. Also write access always need to
> >> >> to verify the offset.
> >> >> 
> >> >> Port the current misc/eeprom/at24.c driver to the MTD framework since
> >> >> EEPROMs are memory-technology devices and the framework already supports
> >> >
> >> > I was under the impression that MTD devices are tightly coupled by erase
> >> > blocks. But then we see MTD_NO_ERASE, so what are MTD devices after all?
> >> 
> >> I was curious as well so I did some digging.
> >> 
> [...]
> >> 
> >> I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
> >> EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
> >> drivers under a single interface. I am not sure what came of it though,
> >> since I can't find any patches that followed up with the proposal.
> >
> > That discussion led to drivers/nvmem after I started to work on
> > some early prototype, and Srinivas took over that work.
> 
> So would you say it is better for EEPROM drivers to use nvmem instead of
> moving under MTD?

I thought so at the time, but that was more than 10y ago, and I have
followed neither nvmem nor MTD since so I don't really have an opinion
there.

It looks like drivers/misc/eeprom/at24.c has support for nvmem though,
and MTD can be used as an nvmem provider too, so it's not clear to me
why we would want to create yet another variant.

But again, you shouldn't really ask me in the first place :)

I'm sure Miquel, Srinivas, and surely others, are much more relevant to
answer that question.

Maxime
Miquel Raynal July 8, 2024, 6:44 a.m. UTC | #6
Hi,

> > >> >> Port the current misc/eeprom/at24.c driver to the MTD framework since
> > >> >> EEPROMs are memory-technology devices and the framework already supports  
> > >> >
> > >> > I was under the impression that MTD devices are tightly coupled by erase
> > >> > blocks. But then we see MTD_NO_ERASE, so what are MTD devices after all?  
> > >> 
> > >> I was curious as well so I did some digging.
> > >>   
> > [...]  
> > >> 
> > >> I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
> > >> EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
> > >> drivers under a single interface. I am not sure what came of it though,
> > >> since I can't find any patches that followed up with the proposal.  
> > >
> > > That discussion led to drivers/nvmem after I started to work on
> > > some early prototype, and Srinivas took over that work.  
> > 
> > So would you say it is better for EEPROM drivers to use nvmem instead of
> > moving under MTD?  
> 
> I thought so at the time, but that was more than 10y ago, and I have
> followed neither nvmem nor MTD since so I don't really have an opinion
> there.
> 
> It looks like drivers/misc/eeprom/at24.c has support for nvmem though,
> and MTD can be used as an nvmem provider too, so it's not clear to me
> why we would want to create yet another variant.
> 
> But again, you shouldn't really ask me in the first place :)
> 
> I'm sure Miquel, Srinivas, and surely others, are much more relevant to
> answer that question.

More relevant, I doubt, but just a feeling: EEPROMs have their own
subsystem now, NVMEM, which, as Maxime said, was initially written for
that very specific case. EEPROMs don't have the complexity of MTD
devices, and thus pulling the whole MTD subsystem just for getting
partitions seems counter intuitive to me. You can definitely "split"
EEPROM devices with NVMEM as well anyway.

Overall I think the idea of getting rid of these misc/ drivers is goes
into the right direction, but registering directly into NVMEM makes
more sense IMO.

Thanks,
Miquèl
Marco Felsch July 9, 2024, 9:22 a.m. UTC | #7
Hi Miquel,

On 24-07-08, Miquel Raynal wrote:
> Hi,
> 
> > > >> >> Port the current misc/eeprom/at24.c driver to the MTD framework since
> > > >> >> EEPROMs are memory-technology devices and the framework already supports  
> > > >> >
> > > >> > I was under the impression that MTD devices are tightly coupled by erase
> > > >> > blocks. But then we see MTD_NO_ERASE, so what are MTD devices after all?  
> > > >> 
> > > >> I was curious as well so I did some digging.
> > > >>   
> > > [...]  
> > > >> 
> > > >> I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
> > > >> EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
> > > >> drivers under a single interface. I am not sure what came of it though,
> > > >> since I can't find any patches that followed up with the proposal.  
> > > >
> > > > That discussion led to drivers/nvmem after I started to work on
> > > > some early prototype, and Srinivas took over that work.  
> > > 
> > > So would you say it is better for EEPROM drivers to use nvmem instead of
> > > moving under MTD?  
> > 
> > I thought so at the time, but that was more than 10y ago, and I have
> > followed neither nvmem nor MTD since so I don't really have an opinion
> > there.
> > 
> > It looks like drivers/misc/eeprom/at24.c has support for nvmem though,
> > and MTD can be used as an nvmem provider too, so it's not clear to me
> > why we would want to create yet another variant.
> > 
> > But again, you shouldn't really ask me in the first place :)
> > 
> > I'm sure Miquel, Srinivas, and surely others, are much more relevant to
> > answer that question.
> 
> More relevant, I doubt, but just a feeling: EEPROMs have their own
> subsystem now, NVMEM, which, as Maxime said, was initially written for
> that very specific case. EEPROMs don't have the complexity of MTD
> devices, and thus pulling the whole MTD subsystem just for getting
> partitions seems counter intuitive to me. You can definitely "split"
> EEPROM devices with NVMEM as well anyway.

I asked for feedback on my RFC [1] and all I got was to merge both
drivers into one and make the driver backward compatible, which I did by
this commit.

> Overall I think the idea of getting rid of these misc/ drivers is goes
> into the right direction, but registering directly into NVMEM makes
> more sense IMO.

So you propose to have two places for the partition handling (one for
MTD and one for NVMEM) instead of one and moving the code into NVMEM
directly? That doesn't sound right to me either. Also I don't get the
point why EEPROMs can't be handled by the MTD layer? The layer already
supports devices of type MTD_RAM which are very simple and don't require
an erase-op at least I don't see one.

[1] https://lore.kernel.org/all/20231127164623.1008176-1-m.felsch@pengutronix.de

Regards,
  Marco

> 
> Thanks,
> Miquèl
>
Miquel Raynal July 9, 2024, 9:43 a.m. UTC | #8
Hi Marco,

> > > > >> I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
> > > > >> EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
> > > > >> drivers under a single interface. I am not sure what came of it though,
> > > > >> since I can't find any patches that followed up with the proposal.    
> > > > >
> > > > > That discussion led to drivers/nvmem after I started to work on
> > > > > some early prototype, and Srinivas took over that work.    
> > > > 
> > > > So would you say it is better for EEPROM drivers to use nvmem instead of
> > > > moving under MTD?    
> > > 
> > > I thought so at the time, but that was more than 10y ago, and I have
> > > followed neither nvmem nor MTD since so I don't really have an opinion
> > > there.
> > > 
> > > It looks like drivers/misc/eeprom/at24.c has support for nvmem though,
> > > and MTD can be used as an nvmem provider too, so it's not clear to me
> > > why we would want to create yet another variant.
> > > 
> > > But again, you shouldn't really ask me in the first place :)
> > > 
> > > I'm sure Miquel, Srinivas, and surely others, are much more relevant to
> > > answer that question.  
> > 
> > More relevant, I doubt, but just a feeling: EEPROMs have their own
> > subsystem now, NVMEM, which, as Maxime said, was initially written for
> > that very specific case. EEPROMs don't have the complexity of MTD
> > devices, and thus pulling the whole MTD subsystem just for getting
> > partitions seems counter intuitive to me. You can definitely "split"
> > EEPROM devices with NVMEM as well anyway.  
> 
> I asked for feedback on my RFC [1] and all I got was to merge both
> drivers into one and make the driver backward compatible, which I did by
> this commit.

I'm sorry for not bringing this earlier.

> > Overall I think the idea of getting rid of these misc/ drivers is goes
> > into the right direction, but registering directly into NVMEM makes
> > more sense IMO.  
> 
> So you propose to have two places for the partition handling (one for
> MTD and one for NVMEM) instead of one and moving the code into NVMEM
> directly?

Why two places for the partitions handling? Just one, in NVMEM. Also
usually EEPROMs don't require very advanced partitioning schemes,
unlike flashes (which are the most common MTD devices today).

> That doesn't sound right to me either. Also I don't get the
> point why EEPROMs can't be handled by the MTD layer?

They can, but should they? Just compile the two layers and observe
the size difference. MTD is complex and old, carries a lot of history,
and the user interface is also not straightforward because you need to
handle pages, blocks, erases, bitflips, ECC stats, OOB bytes and
positions, two OTP areas... None of that exists in the EEPROM world. So
why would you want to register into MTD and pull a huge subsystem while
there is a much more recent, simpler and way lighter subsystem fitting
much better your device?

> The layer already
> supports devices of type MTD_RAM which are very simple and don't require
> an erase-op at least I don't see one.

MTD_RAM has been there forever, probably for "bad" reasons. BTW there
has been an attempt at removing it which was reverted in _2006_ and then
felt into the cracks:
21c8db9eff95 ("[MTD] Restore MTD_ROM and MTD_RAM types")

Thanks,
Miquèl
Marco Felsch July 9, 2024, 10:38 a.m. UTC | #9
On 24-07-09, Miquel Raynal wrote:
> Hi Marco,
> 
> > > > > >> I also found a thread from 2013 by Maxime Ripard (+Cc) suggesting adding
> > > > > >> EEPROMs to MTD [1]. The main purpose would have been unifying the EEPROM
> > > > > >> drivers under a single interface. I am not sure what came of it though,
> > > > > >> since I can't find any patches that followed up with the proposal.    
> > > > > >
> > > > > > That discussion led to drivers/nvmem after I started to work on
> > > > > > some early prototype, and Srinivas took over that work.    
> > > > > 
> > > > > So would you say it is better for EEPROM drivers to use nvmem instead of
> > > > > moving under MTD?    
> > > > 
> > > > I thought so at the time, but that was more than 10y ago, and I have
> > > > followed neither nvmem nor MTD since so I don't really have an opinion
> > > > there.
> > > > 
> > > > It looks like drivers/misc/eeprom/at24.c has support for nvmem though,
> > > > and MTD can be used as an nvmem provider too, so it's not clear to me
> > > > why we would want to create yet another variant.
> > > > 
> > > > But again, you shouldn't really ask me in the first place :)
> > > > 
> > > > I'm sure Miquel, Srinivas, and surely others, are much more relevant to
> > > > answer that question.  
> > > 
> > > More relevant, I doubt, but just a feeling: EEPROMs have their own
> > > subsystem now, NVMEM, which, as Maxime said, was initially written for
> > > that very specific case. EEPROMs don't have the complexity of MTD
> > > devices, and thus pulling the whole MTD subsystem just for getting
> > > partitions seems counter intuitive to me. You can definitely "split"
> > > EEPROM devices with NVMEM as well anyway.  
> > 
> > I asked for feedback on my RFC [1] and all I got was to merge both
> > drivers into one and make the driver backward compatible, which I did by
> > this commit.
> 
> I'm sorry for not bringing this earlier.

The purpose of the RFC was exactly to figure out the way to go therefore
I'm a bit surprised now :/

> > > Overall I think the idea of getting rid of these misc/ drivers is goes
> > > into the right direction, but registering directly into NVMEM makes
> > > more sense IMO.  
> > 
> > So you propose to have two places for the partition handling (one for
> > MTD and one for NVMEM) instead of one and moving the code into NVMEM
> > directly?
> 
> Why two places for the partitions handling? Just one, in NVMEM. Also

Without checking the details I think that converting the MTD
partitioning code into NVMEM partitioning code is a bigger task. As you
said below there are many legacy code paths you need to consider so they
still work afterwards as well.

> usually EEPROMs don't require very advanced partitioning schemes,
> unlike flashes (which are the most common MTD devices today).

As said in my cover letter EEPROMs can become quite large and MTD
supports partitioning storage devices which is very handy for large
EEPROMs as well.

> > That doesn't sound right to me either. Also I don't get the
> > point why EEPROMs can't be handled by the MTD layer?
> 
> They can, but should they? Just compile the two layers and observe
> the size difference. MTD is complex and old, carries a lot of history,
> and the user interface is also not straightforward because you need to
> handle pages, blocks, erases, bitflips, ECC stats, OOB bytes and
> positions, two OTP areas... None of that exists in the EEPROM world. So
> why would you want to register into MTD and pull a huge subsystem while
> there is a much more recent, simpler and way lighter subsystem fitting
> much better your device?

Didn't checked the size but the honest, MTD provides many Kconfig
switches to trim the size down. As of now the mtd.o is made of up to 5
(6 if chipreg.o is counted -> should be an opt) files which is of course
larger than the pure misc/eeprom/at24.c driver but not that large. 

Regards,
  Marco

> > The layer already
> > supports devices of type MTD_RAM which are very simple and don't require
> > an erase-op at least I don't see one.
> 
> MTD_RAM has been there forever, probably for "bad" reasons. BTW there
> has been an attempt at removing it which was reverted in _2006_ and then
> felt into the cracks:
> 21c8db9eff95 ("[MTD] Restore MTD_ROM and MTD_RAM types")
> 
> Thanks,
> Miquèl
>
Miquel Raynal July 17, 2024, 8:19 a.m. UTC | #10
Hi Marco,

> > > > Overall I think the idea of getting rid of these misc/ drivers is goes
> > > > into the right direction, but registering directly into NVMEM makes
> > > > more sense IMO.    
> > > 
> > > So you propose to have two places for the partition handling (one for
> > > MTD and one for NVMEM) instead of one and moving the code into NVMEM
> > > directly?  
> > 
> > Why two places for the partitions handling? Just one, in NVMEM. Also  
> 
> Without checking the details I think that converting the MTD
> partitioning code into NVMEM partitioning code is a bigger task. As you
> said below there are many legacy code paths you need to consider so they
> still work afterwards as well.
> 
> > usually EEPROMs don't require very advanced partitioning schemes,
> > unlike flashes (which are the most common MTD devices today).  
> 
> As said in my cover letter EEPROMs can become quite large and MTD
> supports partitioning storage devices which is very handy for large
> EEPROMs as well.

Did you had a look at nvmem-layouts ? In particular the fixed-layout.

Is there anything you would like to achieve already that is not
possible with nvmem but is with mtd?

Thanks,
Miquèl
Marco Felsch July 18, 2024, 9:17 a.m. UTC | #11
Hi Miquel,

On 24-07-17, Miquel Raynal wrote:
> Hi Marco,
> 
> > > > > Overall I think the idea of getting rid of these misc/ drivers is goes
> > > > > into the right direction, but registering directly into NVMEM makes
> > > > > more sense IMO.    
> > > > 
> > > > So you propose to have two places for the partition handling (one for
> > > > MTD and one for NVMEM) instead of one and moving the code into NVMEM
> > > > directly?  
> > > 
> > > Why two places for the partitions handling? Just one, in NVMEM. Also  
> > 
> > Without checking the details I think that converting the MTD
> > partitioning code into NVMEM partitioning code is a bigger task. As you
> > said below there are many legacy code paths you need to consider so they
> > still work afterwards as well.
> > 
> > > usually EEPROMs don't require very advanced partitioning schemes,
> > > unlike flashes (which are the most common MTD devices today).  
> > 
> > As said in my cover letter EEPROMs can become quite large and MTD
> > supports partitioning storage devices which is very handy for large
> > EEPROMs as well.
> 
> Did you had a look at nvmem-layouts ? In particular the fixed-layout.

Yes I had a look at nvmem-layouts and we use them within a
mtd-partition. Using them instead of a mtd-partition is not sufficient
since they:
 1) don't support user-space write (I send a patch for it but it doesn't
    seem to be accepted soon).
 2) If write would be supported the user-space need to write the
    complete cell e.g. no partial writes.

> Is there anything you would like to achieve already that is not
> possible with nvmem but is with mtd?

Please see above.

Regards,
  Marco
Miquel Raynal Aug. 23, 2024, 3:37 p.m. UTC | #12
Hi Marco,

m.felsch@pengutronix.de wrote on Thu, 18 Jul 2024 11:17:53 +0200:

> Hi Miquel,
> 
> On 24-07-17, Miquel Raynal wrote:
> > Hi Marco,
> >   
> > > > > > Overall I think the idea of getting rid of these misc/ drivers is goes
> > > > > > into the right direction, but registering directly into NVMEM makes
> > > > > > more sense IMO.      
> > > > > 
> > > > > So you propose to have two places for the partition handling (one for
> > > > > MTD and one for NVMEM) instead of one and moving the code into NVMEM
> > > > > directly?    
> > > > 
> > > > Why two places for the partitions handling? Just one, in NVMEM. Also    
> > > 
> > > Without checking the details I think that converting the MTD
> > > partitioning code into NVMEM partitioning code is a bigger task. As you
> > > said below there are many legacy code paths you need to consider so they
> > > still work afterwards as well.
> > >   
> > > > usually EEPROMs don't require very advanced partitioning schemes,
> > > > unlike flashes (which are the most common MTD devices today).    
> > > 
> > > As said in my cover letter EEPROMs can become quite large and MTD
> > > supports partitioning storage devices which is very handy for large
> > > EEPROMs as well.  
> > 
> > Did you had a look at nvmem-layouts ? In particular the fixed-layout.  
> 
> Yes I had a look at nvmem-layouts and we use them within a
> mtd-partition. Using them instead of a mtd-partition is not sufficient
> since they:
>  1) don't support user-space write (I send a patch for it but it doesn't
>     seem to be accepted soon).

Yes, this needed improvements maybe but was not refused either.

>  2) If write would be supported the user-space need to write the
>     complete cell e.g. no partial writes.

Maybe that can also be brought to nvmem. Again, nvmem was introduced
for handling EEPROMs in the first place.

Anyway, if other people in Cc: want to share some thoughts, they are
also welcomed, I don't want to block this series for bad reasons. I'm
also adding Michael Walle in Cc: who might have an opinion on that.

Link: https://lore.kernel.org/linux-mtd/20240701-b4-v6-10-topic-usbc-tcpci-v1-4-3fd5f4a193cc@pengutronix.de/

Thanks,
Miquèl
diff mbox series

Patch

diff --git a/MAINTAINERS b/MAINTAINERS
index d6c90161c7bf..a3a56f34d754 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3355,7 +3355,7 @@  L:	linux-i2c@vger.kernel.org
 S:	Maintained
 T:	git git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux.git
 F:	Documentation/devicetree/bindings/eeprom/at24.yaml
-F:	drivers/misc/eeprom/at24.c
+F:	drivers/mtd/devices/at24.c
 
 ATA OVER ETHERNET (AOE) DRIVER
 M:	"Justin Sanders" <justin@coraid.com>
diff --git a/drivers/misc/eeprom/Kconfig b/drivers/misc/eeprom/Kconfig
index 4e61ac18cc96..3a9aaec2f2c7 100644
--- a/drivers/misc/eeprom/Kconfig
+++ b/drivers/misc/eeprom/Kconfig
@@ -3,34 +3,10 @@  menu "EEPROM support"
 
 config EEPROM_AT24
 	tristate "I2C EEPROMs / RAMs / ROMs from most vendors"
-	depends on I2C && SYSFS
-	select NVMEM
-	select NVMEM_SYSFS
-	select REGMAP
-	select REGMAP_I2C
+	select MTD_EEPROM_AT24
 	help
-	  Enable this driver to get read/write support to most I2C EEPROMs
-	  and compatible devices like FRAMs, SRAMs, ROMs etc. After you
-	  configure the driver to know about each chip on your target
-	  board.  Use these generic chip names, instead of vendor-specific
-	  ones like at24c64, 24lc02 or fm24c04:
-
-	     24c00, 24c01, 24c02, spd (readonly 24c02), 24c04, 24c08,
-	     24c16, 24c32, 24c64, 24c128, 24c256, 24c512, 24c1024, 24c2048
-
-	  Unless you like data loss puzzles, always be sure that any chip
-	  you configure as a 24c32 (32 kbit) or larger is NOT really a
-	  24c16 (16 kbit) or smaller, and vice versa. Marking the chip
-	  as read-only won't help recover from this. Also, if your chip
-	  has any software write-protect mechanism you may want to review the
-	  code to make sure this driver won't turn it on by accident.
-
-	  If you use this with an SMBus adapter instead of an I2C adapter,
-	  full functionality is not available.  Only smaller devices are
-	  supported (24c16 and below, max 4 kByte).
-
-	  This driver can also be built as a module.  If so, the module
-	  will be called at24.
+	  Deprecated config option to get read/write support to most I2C
+	  EEPROMs. Please use MTD_EEPROM_AT24 instead.
 
 config EEPROM_AT25
 	tristate "SPI EEPROMs (FRAMs) from most vendors"
diff --git a/drivers/misc/eeprom/Makefile b/drivers/misc/eeprom/Makefile
index 65794e526d5d..2ace4f09d0f0 100644
--- a/drivers/misc/eeprom/Makefile
+++ b/drivers/misc/eeprom/Makefile
@@ -1,5 +1,4 @@ 
 # SPDX-License-Identifier: GPL-2.0
-obj-$(CONFIG_EEPROM_AT24)	+= at24.o
 obj-$(CONFIG_EEPROM_AT25)	+= at25.o
 obj-$(CONFIG_EEPROM_MAX6875)	+= max6875.o
 obj-$(CONFIG_EEPROM_93CX6)	+= eeprom_93cx6.o
diff --git a/drivers/mtd/devices/Kconfig b/drivers/mtd/devices/Kconfig
index ff2f9e55ef28..8c6720ea84e7 100644
--- a/drivers/mtd/devices/Kconfig
+++ b/drivers/mtd/devices/Kconfig
@@ -3,6 +3,37 @@  menu "Self-contained MTD device drivers"
 	depends on MTD!=n
 	depends on HAS_IOMEM
 
+config MTD_EEPROM_AT24
+	tristate "MTD I2C EEPROMs / RAMs / ROMs from most vendors"
+	depends on I2C && SYSFS
+	select NVMEM
+	select NVMEM_SYSFS
+	select REGMAP
+	select REGMAP_I2C
+	help
+	  Enable this driver to get read/write support to most I2C EEPROMs
+	  and compatible devices like FRAMs, SRAMs, ROMs etc. After you
+	  configure the driver to know about each chip on your target
+	  board.  Use these generic chip names, instead of vendor-specific
+	  ones like at24c64, 24lc02 or fm24c04:
+
+	     24c00, 24c01, 24c02, spd (readonly 24c02), 24c04, 24c08,
+	     24c16, 24c32, 24c64, 24c128, 24c256, 24c512, 24c1024, 24c2048
+
+	  Unless you like data loss puzzles, always be sure that any chip
+	  you configure as a 24c32 (32 kbit) or larger is NOT really a
+	  24c16 (16 kbit) or smaller, and vice versa. Marking the chip
+	  as read-only won't help recover from this. Also, if your chip
+	  has any software write-protect mechanism you may want to review the
+	  code to make sure this driver won't turn it on by accident.
+
+	  If you use this with an SMBus adapter instead of an I2C adapter,
+	  full functionality is not available.  Only smaller devices are
+	  supported (24c16 and below, max 4 kByte).
+
+	  This driver can also be built as a module.  If so, the module
+	  will be called at24.
+
 config MTD_PMC551
 	tristate "Ramix PMC551 PCI Mezzanine RAM card support"
 	depends on PCI
diff --git a/drivers/mtd/devices/Makefile b/drivers/mtd/devices/Makefile
index d11eb2b8b6f8..00ef5d87a957 100644
--- a/drivers/mtd/devices/Makefile
+++ b/drivers/mtd/devices/Makefile
@@ -4,6 +4,7 @@ 
 #
 
 obj-$(CONFIG_MTD_DOCG3)		+= docg3.o
+obj-$(CONFIG_MTD_EEPROM_AT24)	+= at24.o
 obj-$(CONFIG_MTD_SLRAM)		+= slram.o
 obj-$(CONFIG_MTD_PHRAM)		+= phram.o
 obj-$(CONFIG_MTD_PMC551)	+= pmc551.o
diff --git a/drivers/misc/eeprom/at24.c b/drivers/mtd/devices/at24.c
similarity index 92%
rename from drivers/misc/eeprom/at24.c
rename to drivers/mtd/devices/at24.c
index 4bd4f32bcdab..713f40a1d95c 100644
--- a/drivers/misc/eeprom/at24.c
+++ b/drivers/mtd/devices/at24.c
@@ -4,6 +4,7 @@ 
  *
  * Copyright (C) 2005-2007 David Brownell
  * Copyright (C) 2008 Wolfram Sang, Pengutronix
+ * Copyright (C) 2024 Pengutronix, Marco Felsch <kernel@pengutronix.de>
  */
 
 #include <linux/acpi.h>
@@ -16,8 +17,8 @@ 
 #include <linux/kernel.h>
 #include <linux/mod_devicetable.h>
 #include <linux/module.h>
+#include <linux/mtd/mtd.h>
 #include <linux/mutex.h>
-#include <linux/nvmem-provider.h>
 #include <linux/of.h>
 #include <linux/of_device.h>
 #include <linux/pm_runtime.h>
@@ -84,8 +85,9 @@  struct at24_data {
 	u16 page_size;
 	u8 flags;
 
-	struct nvmem_device *nvmem;
+	struct mtd_info mtd;
 	struct regulator *vcc_reg;
+	struct gpio_desc *wp_gpio;
 	void (*read_post)(unsigned int off, char *buf, size_t count);
 
 	/*
@@ -96,6 +98,11 @@  struct at24_data {
 	struct regmap *client_regmaps[] __counted_by(num_addresses);
 };
 
+static struct at24_data *mtd_to_at24(struct mtd_info *mtd)
+{
+	return container_of(mtd, struct at24_data, mtd);
+}
+
 /*
  * This parameter is to help this driver avoid blocking other drivers out
  * of I2C for potentially troublesome amounts of time. With a 100 kHz I2C
@@ -424,20 +431,19 @@  static ssize_t at24_regmap_write(struct at24_data *at24, const char *buf,
 	return -ETIMEDOUT;
 }
 
-static int at24_read(void *priv, unsigned int off, void *val, size_t count)
+static int at24_read(struct mtd_info *mtd, loff_t from, size_t len,
+		     size_t *retlen, u_char *buf)
 {
-	struct at24_data *at24;
+	struct at24_data *at24 = mtd_to_at24(mtd);
 	struct device *dev;
-	char *buf = val;
 	int i, ret;
 
-	at24 = priv;
 	dev = at24_base_client_dev(at24);
 
-	if (unlikely(!count))
-		return count;
+	if (unlikely(!len))
+		return len;
 
-	if (off + count > at24->byte_len)
+	if (from + len > at24->byte_len)
 		return -EINVAL;
 
 	ret = pm_runtime_resume_and_get(dev);
@@ -449,13 +455,14 @@  static int at24_read(void *priv, unsigned int off, void *val, size_t count)
 	 */
 	mutex_lock(&at24->lock);
 
-	for (i = 0; count; i += ret, count -= ret) {
-		ret = at24_regmap_read(at24, buf + i, off + i, count);
+	for (i = 0; len; i += ret, len -= ret) {
+		ret = at24_regmap_read(at24, buf + i, from + i, len);
 		if (ret < 0) {
 			mutex_unlock(&at24->lock);
 			pm_runtime_put(dev);
 			return ret;
 		}
+		*retlen += ret;
 	}
 
 	mutex_unlock(&at24->lock);
@@ -463,25 +470,24 @@  static int at24_read(void *priv, unsigned int off, void *val, size_t count)
 	pm_runtime_put(dev);
 
 	if (unlikely(at24->read_post))
-		at24->read_post(off, buf, i);
+		at24->read_post(from, buf, i);
 
 	return 0;
 }
 
-static int at24_write(void *priv, unsigned int off, void *val, size_t count)
+static int at24_write(struct mtd_info *mtd, loff_t to, size_t len,
+		      size_t *retlen, const u_char *buf)
 {
-	struct at24_data *at24;
+	struct at24_data *at24 = mtd_to_at24(mtd);
 	struct device *dev;
-	char *buf = val;
 	int ret;
 
-	at24 = priv;
 	dev = at24_base_client_dev(at24);
 
-	if (unlikely(!count))
+	if (unlikely(!len))
 		return -EINVAL;
 
-	if (off + count > at24->byte_len)
+	if (to + len > at24->byte_len)
 		return -EINVAL;
 
 	ret = pm_runtime_resume_and_get(dev);
@@ -493,18 +499,23 @@  static int at24_write(void *priv, unsigned int off, void *val, size_t count)
 	 */
 	mutex_lock(&at24->lock);
 
-	while (count) {
-		ret = at24_regmap_write(at24, buf, off, count);
+	gpiod_set_value_cansleep(at24->wp_gpio, 0);
+
+	while (len) {
+		ret = at24_regmap_write(at24, buf, to, len);
 		if (ret < 0) {
 			mutex_unlock(&at24->lock);
 			pm_runtime_put(dev);
 			return ret;
 		}
 		buf += ret;
-		off += ret;
-		count -= ret;
+		to += ret;
+		len -= ret;
+		*retlen += ret;
 	}
 
+	gpiod_set_value_cansleep(at24->wp_gpio, 1);
+
 	mutex_unlock(&at24->lock);
 
 	pm_runtime_put(dev);
@@ -562,6 +573,8 @@  static void at24_probe_temp_sensor(struct i2c_client *client)
 {
 	struct at24_data *at24 = i2c_get_clientdata(client);
 	struct i2c_board_info info = { .type = "jc42" };
+	struct mtd_info *mtd = &at24->mtd;
+	size_t len;
 	int ret;
 	u8 val;
 
@@ -569,12 +582,12 @@  static void at24_probe_temp_sensor(struct i2c_client *client)
 	 * Byte 2 has value 11 for DDR3, earlier versions don't
 	 * support the thermal sensor present flag
 	 */
-	ret = at24_read(at24, 2, &val, 1);
+	ret = at24_read(mtd, 2, 1, &len, &val);
 	if (ret || val != 11)
 		return;
 
 	/* Byte 32, bit 7 is set if temp sensor is present */
-	ret = at24_read(at24, 32, &val, 1);
+	ret = at24_read(mtd, 32, 1, &len, &val);
 	if (ret || !(val & BIT(7)))
 		return;
 
@@ -586,17 +599,19 @@  static void at24_probe_temp_sensor(struct i2c_client *client)
 static int at24_probe(struct i2c_client *client)
 {
 	struct regmap_config regmap_config = { };
-	struct nvmem_config nvmem_config = { };
 	u32 byte_len, page_size, flags, addrw;
 	const struct at24_chip_data *cdata;
 	struct device *dev = &client->dev;
 	bool i2c_fn_i2c, i2c_fn_block;
 	unsigned int i, num_addresses;
 	struct at24_data *at24;
+	struct device_node *np;
+	struct mtd_info *mtd;
 	bool full_power;
 	struct regmap *regmap;
 	bool writable;
 	u8 test_byte;
+	size_t len;
 	int err;
 
 	i2c_fn_i2c = i2c_check_functionality(client->adapter, I2C_FUNC_I2C);
@@ -703,6 +718,10 @@  static int at24_probe(struct i2c_client *client)
 					page_size, at24_io_limit);
 		if (!i2c_fn_i2c && at24->write_max > I2C_SMBUS_BLOCK_MAX)
 			at24->write_max = I2C_SMBUS_BLOCK_MAX;
+
+		at24->wp_gpio = gpiod_get_optional(dev, "wp", GPIOD_OUT_HIGH);
+		if (IS_ERR(at24->wp_gpio))
+			return PTR_ERR(at24->wp_gpio);
 	}
 
 	/* use dummy devices for multiple-address chips */
@@ -712,37 +731,31 @@  static int at24_probe(struct i2c_client *client)
 			return err;
 	}
 
-	/*
-	 * We initialize nvmem_config.id to NVMEM_DEVID_AUTO even if the
-	 * label property is set as some platform can have multiple eeproms
-	 * with same label and we can not register each of those with same
-	 * label. Failing to register those eeproms trigger cascade failure
-	 * on such platform.
-	 */
-	nvmem_config.id = NVMEM_DEVID_AUTO;
-
+	mtd = &at24->mtd;
 	if (device_property_present(dev, "label")) {
 		err = device_property_read_string(dev, "label",
-						  &nvmem_config.name);
+						  &mtd->name);
 		if (err)
 			return err;
 	} else {
-		nvmem_config.name = dev_name(dev);
+		mtd->name = dev_name(dev);
 	}
-
-	nvmem_config.type = NVMEM_TYPE_EEPROM;
-	nvmem_config.dev = dev;
-	nvmem_config.read_only = !writable;
-	nvmem_config.root_only = !(flags & AT24_FLAG_IRUGO);
-	nvmem_config.owner = THIS_MODULE;
-	nvmem_config.compat = true;
-	nvmem_config.base_dev = dev;
-	nvmem_config.reg_read = at24_read;
-	nvmem_config.reg_write = at24_write;
-	nvmem_config.priv = at24;
-	nvmem_config.stride = 1;
-	nvmem_config.word_size = 1;
-	nvmem_config.size = byte_len;
+	mtd->dev.parent = dev;
+	mtd->type = MTD_EEPROM;
+	mtd->flags = MTD_CAP_EEPROM;
+	if (!writable)
+		mtd->flags = MTD_CAP_ROM;
+
+	mtd->writesize = 1;
+	mtd->writebufsize = 1;
+	mtd->size = byte_len;
+	mtd->_read = at24_read;
+	mtd->_write = at24_write;
+
+	/* Fixed partitions are only supported on OF plaforms */
+	np = to_of_node(dev_fwnode(dev));
+	if (np)
+		mtd_set_of_node(mtd, np);
 
 	i2c_set_clientdata(client, at24);
 
@@ -764,7 +777,7 @@  static int at24_probe(struct i2c_client *client)
 	 * it's powered off right now).
 	 */
 	if (full_power) {
-		err = at24_read(at24, 0, &test_byte, 1);
+		err = at24_read(mtd, 0, 1, &len, &test_byte);
 		if (err) {
 			pm_runtime_disable(dev);
 			if (!pm_runtime_status_suspended(dev))
@@ -773,13 +786,12 @@  static int at24_probe(struct i2c_client *client)
 		}
 	}
 
-	at24->nvmem = devm_nvmem_register(dev, &nvmem_config);
-	if (IS_ERR(at24->nvmem)) {
+	err = mtd_device_register(mtd, NULL, 0);
+	if (err) {
 		pm_runtime_disable(dev);
 		if (!pm_runtime_status_suspended(dev))
 			regulator_disable(at24->vcc_reg);
-		return dev_err_probe(dev, PTR_ERR(at24->nvmem),
-				     "failed to register nvmem\n");
+		return dev_err_probe(dev, err, "failed to register mtd\n");
 	}
 
 	/* If this a SPD EEPROM, probe for DDR3 thermal sensor */