Accessing mtd partitions from driver/i2c

Accessing mtd partitions from driver/i2c

[Prabhakar Kushwaha]

Hi All,

We have a requirement of using(read/write) one mtd partition from driver/i2c. 

As per my googling  mtd->_read, mtd->_write APIs has not been used anywhere other than driver/mtd layer.
Do we have any such framework/use case available in Linux which can be referred.

Any pointer/clue will be helpful

--pk

 

Re: Accessing mtd partitions from driver/i2c

[Boris Brezillon]

Hi Prabhakar,

On Wed, 6 Jun 2018 05:02:47 +0000
Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:

> Hi All,
> 
> We have a requirement of using(read/write) one mtd partition from driver/i2c. 
> 
> As per my googling  mtd->_read, mtd->_write APIs has not been used anywhere other than driver/mtd layer.

It's used by at least one filesystem (JFFS2). And you should not look
for mtd->_xxx() users but mtd_xxx() users. You should never access the
mtd hooks directly but the wrappers that are exposed by the MTD API
(mtd_read(), mtd_write(), mtd_erase(), ...).

BTW, what's your use case? If it's about reading a piece of information
from a RO partition, then the work initiated by Alban [1] and exposing
some MTD parts are NVMEM providers should be a better match.

> Do we have any such framework/use case available in Linux which can be referred.
> 
> Any pointer/clue will be helpful

I don't know what you want to do with the MTD device exactly, so it's
hard to tell.

Regards,

Boris

[1]https://patchwork.ozlabs.org/project/linux-mtd/list/?series=35728

RE: Accessing mtd partitions from driver/i2c

[Prabhakar Kushwaha]

Dear Borris,

> -----Original Message-----
> From: Boris Brezillon [mailto:boris.brezillon@bootlin.com]
> Sent: Wednesday, June 6, 2018 12:53 PM
> To: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
> Cc: linux-mtd@lists.infradead.org; marek.vasut@gmail.com; Poonam Aggrwal
> <poonam.aggrwal@nxp.com>; Ashish Kumar <ashish.kumar@nxp.com>; Ying
> Zhang <ying.zhang22455@nxp.com>
> Subject: Re: Accessing mtd partitions from driver/i2c
> 
> Hi Prabhakar,
> 
> On Wed, 6 Jun 2018 05:02:47 +0000
> Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:
> 
> > Hi All,
> >
> > We have a requirement of using(read/write) one mtd partition from driver/i2c.
> >
> > As per my googling  mtd->_read, mtd->_write APIs has not been used
> anywhere other than driver/mtd layer.
> 
> It's used by at least one filesystem (JFFS2). And you should not look for mtd-
> >_xxx() users but mtd_xxx() users. You should never access the mtd hooks
> directly but the wrappers that are exposed by the MTD API (mtd_read(),
> mtd_write(), mtd_erase(), ...).
> 

I can find it's usage in driver/firmware/ folder. Looks like it is using nvram.

> BTW, what's your use case? If it's about reading a piece of information from a
> RO partition, then the work initiated by Alban [1] and exposing some MTD parts
> are NVMEM providers should be a better match.
> 

Currently, Linux has i2c slave framework. It is reading/writing to DDR emulated as EEPROM (drivers/i2c/i2c-slave-eeprom.c).
We want something similar, where DDR will be replaced by MTD partitions for read/write. 

--pk

Re: Accessing mtd partitions from driver/i2c

[Boris Brezillon]

+Alban and Srinivas

On Wed, 6 Jun 2018 13:44:32 +0000
Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:

> Dear Borris,
> 
> > -----Original Message-----
> > From: Boris Brezillon [mailto:boris.brezillon@bootlin.com]
> > Sent: Wednesday, June 6, 2018 12:53 PM
> > To: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
> > Cc: linux-mtd@lists.infradead.org; marek.vasut@gmail.com; Poonam
> > Aggrwal <poonam.aggrwal@nxp.com>; Ashish Kumar
> > <ashish.kumar@nxp.com>; Ying Zhang <ying.zhang22455@nxp.com>
> > Subject: Re: Accessing mtd partitions from driver/i2c
> > 
> > Hi Prabhakar,
> > 
> > On Wed, 6 Jun 2018 05:02:47 +0000
> > Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:
> >   
> > > Hi All,
> > >
> > > We have a requirement of using(read/write) one mtd partition from
> > > driver/i2c.
> > >
> > > As per my googling  mtd->_read, mtd->_write APIs has not been
> > > used  
> > anywhere other than driver/mtd layer.
> > 
> > It's used by at least one filesystem (JFFS2). And you should not
> > look for mtd-  
> > >_xxx() users but mtd_xxx() users. You should never access the mtd
> > >hooks  
> > directly but the wrappers that are exposed by the MTD API
> > (mtd_read(), mtd_write(), mtd_erase(), ...).
> >   
> 
> I can find it's usage in driver/firmware/ folder. Looks like it is
> using nvram.
> 
> > BTW, what's your use case? If it's about reading a piece of
> > information from a RO partition, then the work initiated by Alban
> > [1] and exposing some MTD parts are NVMEM providers should be a
> > better match. 
> 
> Currently, Linux has i2c slave framework. It is reading/writing to
> DDR emulated as EEPROM (drivers/i2c/i2c-slave-eeprom.c). We want
> something similar, where DDR will be replaced by MTD partitions for
> read/write. 

Then NVMEM is the solution you're looking for. Alban, Srinivas, can you
agree on something regarding the DT bindings so that we can finally get
this feature merged?

Thanks,

Boris

RE: Accessing mtd partitions from driver/i2c

[Prabhakar Kushwaha]

Hi Boris,

> -----Original Message-----
> From: Boris Brezillon [mailto:boris.brezillon@bootlin.com]
> Sent: Wednesday, June 6, 2018 7:26 PM
> To: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
> Cc: linux-mtd@lists.infradead.org; marek.vasut@gmail.com; Poonam Aggrwal
> <poonam.aggrwal@nxp.com>; Ashish Kumar <ashish.kumar@nxp.com>; Ying
> Zhang <ying.zhang22455@nxp.com>; Alban Bedel <albeu@free.fr>; Srinivas
> Kandagatla <srinivas.kandagatla@linaro.org>
> Subject: Re: Accessing mtd partitions from driver/i2c
> 
> +Alban and Srinivas
> 
> On Wed, 6 Jun 2018 13:44:32 +0000
> Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:
> 
> > Dear Borris,
> >
> > > -----Original Message-----
> > > From: Boris Brezillon [mailto:boris.brezillon@bootlin.com]
> > > Sent: Wednesday, June 6, 2018 12:53 PM
> > > To: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
> > > Cc: linux-mtd@lists.infradead.org; marek.vasut@gmail.com; Poonam
> > > Aggrwal <poonam.aggrwal@nxp.com>; Ashish Kumar
> > > <ashish.kumar@nxp.com>; Ying Zhang <ying.zhang22455@nxp.com>
> > > Subject: Re: Accessing mtd partitions from driver/i2c
> > >
> > > Hi Prabhakar,
> > >
> > > On Wed, 6 Jun 2018 05:02:47 +0000
> > > Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:
> > >
> > > > Hi All,
> > > >
> > > > We have a requirement of using(read/write) one mtd partition from
> > > > driver/i2c.
> > > >
> > > > As per my googling  mtd->_read, mtd->_write APIs has not been used
> > > anywhere other than driver/mtd layer.
> > >
> > > It's used by at least one filesystem (JFFS2). And you should not
> > > look for mtd-
> > > >_xxx() users but mtd_xxx() users. You should never access the mtd
> > > >hooks
> > > directly but the wrappers that are exposed by the MTD API
> > > (mtd_read(), mtd_write(), mtd_erase(), ...).
> > >
> >
> > I can find it's usage in driver/firmware/ folder. Looks like it is
> > using nvram.
> >
> > > BTW, what's your use case? If it's about reading a piece of
> > > information from a RO partition, then the work initiated by Alban
> > > [1] and exposing some MTD parts are NVMEM providers should be a
> > > better match.
> >
> > Currently, Linux has i2c slave framework. It is reading/writing to DDR
> > emulated as EEPROM (drivers/i2c/i2c-slave-eeprom.c). We want something
> > similar, where DDR will be replaced by MTD partitions for read/write.
> 
> Then NVMEM is the solution you're looking for. Alban, Srinivas, can you agree
> on something regarding the DT bindings so that we can finally get this feature
> merged?
> 

Patch set[1] shared by you, only supports read. 
Can this framework also support write?

--pk

[1] https://patchwork.ozlabs.org/project/linux-mtd/list/?series=35728

Re: Accessing mtd partitions from driver/i2c

[Boris Brezillon]

On Thu, 7 Jun 2018 11:50:51 +0000
Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:

> Hi Boris,
> 
> > -----Original Message-----
> > From: Boris Brezillon [mailto:boris.brezillon@bootlin.com]
> > Sent: Wednesday, June 6, 2018 7:26 PM
> > To: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
> > Cc: linux-mtd@lists.infradead.org; marek.vasut@gmail.com; Poonam Aggrwal
> > <poonam.aggrwal@nxp.com>; Ashish Kumar <ashish.kumar@nxp.com>; Ying
> > Zhang <ying.zhang22455@nxp.com>; Alban Bedel <albeu@free.fr>; Srinivas
> > Kandagatla <srinivas.kandagatla@linaro.org>
> > Subject: Re: Accessing mtd partitions from driver/i2c
> > 
> > +Alban and Srinivas
> > 
> > On Wed, 6 Jun 2018 13:44:32 +0000
> > Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:
> >   
> > > Dear Borris,
> > >  
> > > > -----Original Message-----
> > > > From: Boris Brezillon [mailto:boris.brezillon@bootlin.com]
> > > > Sent: Wednesday, June 6, 2018 12:53 PM
> > > > To: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
> > > > Cc: linux-mtd@lists.infradead.org; marek.vasut@gmail.com; Poonam
> > > > Aggrwal <poonam.aggrwal@nxp.com>; Ashish Kumar
> > > > <ashish.kumar@nxp.com>; Ying Zhang <ying.zhang22455@nxp.com>
> > > > Subject: Re: Accessing mtd partitions from driver/i2c
> > > >
> > > > Hi Prabhakar,
> > > >
> > > > On Wed, 6 Jun 2018 05:02:47 +0000
> > > > Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:
> > > >  
> > > > > Hi All,
> > > > >
> > > > > We have a requirement of using(read/write) one mtd partition from
> > > > > driver/i2c.
> > > > >
> > > > > As per my googling  mtd->_read, mtd->_write APIs has not been used  
> > > > anywhere other than driver/mtd layer.
> > > >
> > > > It's used by at least one filesystem (JFFS2). And you should not
> > > > look for mtd-  
> > > > >_xxx() users but mtd_xxx() users. You should never access the mtd
> > > > >hooks  
> > > > directly but the wrappers that are exposed by the MTD API
> > > > (mtd_read(), mtd_write(), mtd_erase(), ...).
> > > >  
> > >
> > > I can find it's usage in driver/firmware/ folder. Looks like it is
> > > using nvram.
> > >  
> > > > BTW, what's your use case? If it's about reading a piece of
> > > > information from a RO partition, then the work initiated by Alban
> > > > [1] and exposing some MTD parts are NVMEM providers should be a
> > > > better match.  
> > >
> > > Currently, Linux has i2c slave framework. It is reading/writing to DDR
> > > emulated as EEPROM (drivers/i2c/i2c-slave-eeprom.c). We want something
> > > similar, where DDR will be replaced by MTD partitions for read/write.  
> > 
> > Then NVMEM is the solution you're looking for. Alban, Srinivas, can you agree
> > on something regarding the DT bindings so that we can finally get this feature
> > merged?
> >   
> 
> Patch set[1] shared by you, only supports read. 
> Can this framework also support write?

The NVMEM framework already supports write, you just need to add write
support to the nvmem -> MTD implementation.