mtd_info->write

mtd_info->write

[Abraham vd Merwe]

[-- Attachment #1: Type: text/plain, Size: 726 bytes --]

Hi!

Can you safely assume in the MTD write function that the blocks on which the
data is have been erased or do you actually have to do some error checking
and return with -EINVAL if you discover that it's not been erased?

-- 

Regards
 Abraham

Just because he's dead is no reason to lay off work.

__________________________________________________________
 Abraham vd Merwe - 2d3D, Inc.

 Device Driver Development, Outsourcing, Embedded Systems

  Cell: +27 82 565 4451         Snailmail:
   Tel: +27 21 761 7549            Block C, Antree Park
   Fax: +27 21 761 7648            Doncaster Road
 Email: abraham@2d3d.co.za         Kenilworth, 7700
  Http: http://www.2d3d.com        South Africa


[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

RE: mtd_info->write

[Frederic Giasson]

I am almost sure that you have to validate the erase state of the sector you
want to write to.  For example, I use cfi_amdstd_write function, and
absolutely no validation is made in that fonction for this.

Frederic Giasson





-----Original Message-----
From: Abraham vd Merwe [mailto:abraham@2d3d.co.za]
Sent: Wednesday, June 06, 2001 9:28 AM
To: Frederic Giasson
Cc: MTD for Linux
Subject: Re: mtd_info->write


Hi Frederic!

> If the data is all ones ("1"), it has been erased.   If not, it has not.

Yes, I know, but I want to know whether I have to check if that's the case
or not.

> Can you safely assume in the MTD write function that the blocks on which
the
> data is have been erased or do you actually have to do some error checking
> and return with -EINVAL if you discover that it's not been erased?

-- 

Regards
 Abraham

It is your concern when your neighbor's wall is on fire.
		-- Quintus Horatius Flaccus (Horace)

__________________________________________________________
 Abraham vd Merwe - 2d3D, Inc.

 Device Driver Development, Outsourcing, Embedded Systems

  Cell: +27 82 565 4451         Snailmail:
   Tel: +27 21 761 7549            Block C, Antree Park
   Fax: +27 21 761 7648            Doncaster Road
 Email: abraham@2d3d.co.za         Kenilworth, 7700
  Http: http://www.2d3d.com        South Africa

Re: mtd_info->write

[Abraham vd Merwe]

[-- Attachment #1: Type: text/plain, Size: 950 bytes --]

Hi Frederic!

> If the data is all ones ("1"), it has been erased.   If not, it has not.

Yes, I know, but I want to know whether I have to check if that's the case
or not.

> Can you safely assume in the MTD write function that the blocks on which the
> data is have been erased or do you actually have to do some error checking
> and return with -EINVAL if you discover that it's not been erased?

-- 

Regards
 Abraham

It is your concern when your neighbor's wall is on fire.
		-- Quintus Horatius Flaccus (Horace)

__________________________________________________________
 Abraham vd Merwe - 2d3D, Inc.

 Device Driver Development, Outsourcing, Embedded Systems

  Cell: +27 82 565 4451         Snailmail:
   Tel: +27 21 761 7549            Block C, Antree Park
   Fax: +27 21 761 7648            Doncaster Road
 Email: abraham@2d3d.co.za         Kenilworth, 7700
  Http: http://www.2d3d.com        South Africa


[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

RE: mtd_info->write

[Frederic Giasson]

In my opinion, the low level driver does not have to handle validation, it
just have to provide an interface to primitive function.  The high level
module ( the user interface part of the driver ) has to do the job.  I think
that's the way of MTD people too. 

Frederic Giasson





-----Original Message-----
From: Abraham vd Merwe [mailto:abraham@2d3d.co.za]
Sent: Wednesday, June 06, 2001 9:35 AM
To: Frederic Giasson
Cc: MTD for Linux
Subject: Re: mtd_info->write


Hi Frederic!

> I am almost sure that you have to validate the erase state of the sector
you
> want to write to.  For example, I use cfi_amdstd_write function, and
> absolutely no validation is made in that fonction for this.

You mean the user process/module has to, not the chip driver? I'm writing a
chip driver and I want to know whether the chip driver is responsible for
the checking.

I noticed the other chip drivers (like you say here) doesn't so I'm assuming
I don't have to but I just want to make sure.

> > If the data is all ones ("1"), it has been erased.   If not, it has not.
> 
> Yes, I know, but I want to know whether I have to check if that's the case
> or not.
> 
> > Can you safely assume in the MTD write function that the blocks on which
> the
> > data is have been erased or do you actually have to do some error
checking
> > and return with -EINVAL if you discover that it's not been erased?

-- 

Regards
 Abraham

Chemistry professors never die, they just fail to react.

__________________________________________________________
 Abraham vd Merwe - 2d3D, Inc.

 Device Driver Development, Outsourcing, Embedded Systems

  Cell: +27 82 565 4451         Snailmail:
   Tel: +27 21 761 7549            Block C, Antree Park
   Fax: +27 21 761 7648            Doncaster Road
 Email: abraham@2d3d.co.za         Kenilworth, 7700
  Http: http://www.2d3d.com        South Africa

RE: mtd_info->write

[Frederic Giasson]

That's right, but a read back validation has to show the error.

Frédéric Giasson




-----Original Message-----
From: David Woodhouse [mailto:dwmw2@infradead.org]
Sent: Wednesday, June 06, 2001 9:36 AM
To: Abraham vd Merwe
Cc: Frederic Giasson; MTD for Linux
Subject: Re: mtd_info->write 




abraham@2d3d.co.za said:
> > If the data is all ones ("1"), it has been erased.   If not, it has
> > not.

> Yes, I know, but I want to know whether I have to check if that's the
> case or not. 

The chip driver is not required to check that the data are all ones. You 
are permitted to write to the same location of a chip repeatedly. If you 
write a 1 where previously there was a zero, obviously the result will be a 
zero. But that is not an error.

--
dwmw2

Re: mtd_info->write

[Abraham vd Merwe]

[-- Attachment #1: Type: text/plain, Size: 1442 bytes --]

Hi Frederic!

> I am almost sure that you have to validate the erase state of the sector you
> want to write to.  For example, I use cfi_amdstd_write function, and
> absolutely no validation is made in that fonction for this.

You mean the user process/module has to, not the chip driver? I'm writing a
chip driver and I want to know whether the chip driver is responsible for
the checking.

I noticed the other chip drivers (like you say here) doesn't so I'm assuming
I don't have to but I just want to make sure.

> > If the data is all ones ("1"), it has been erased.   If not, it has not.
> 
> Yes, I know, but I want to know whether I have to check if that's the case
> or not.
> 
> > Can you safely assume in the MTD write function that the blocks on which
> the
> > data is have been erased or do you actually have to do some error checking
> > and return with -EINVAL if you discover that it's not been erased?

-- 

Regards
 Abraham

Chemistry professors never die, they just fail to react.

__________________________________________________________
 Abraham vd Merwe - 2d3D, Inc.

 Device Driver Development, Outsourcing, Embedded Systems

  Cell: +27 82 565 4451         Snailmail:
   Tel: +27 21 761 7549            Block C, Antree Park
   Fax: +27 21 761 7648            Doncaster Road
 Email: abraham@2d3d.co.za         Kenilworth, 7700
  Http: http://www.2d3d.com        South Africa


[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: mtd_info->write

[David Woodhouse]

abraham@2d3d.co.za said:
> > If the data is all ones ("1"), it has been erased.   If not, it has
> > not.

> Yes, I know, but I want to know whether I have to check if that's the
> case or not. 

The chip driver is not required to check that the data are all ones. You 
are permitted to write to the same location of a chip repeatedly. If you 
write a 1 where previously there was a zero, obviously the result will be a 
zero. But that is not an error.

--
dwmw2

RE: mtd_info->write

[Frederic Giasson]

There is validation, in do_write_oneword ( cfi_cmdset_0002.c ), there is
pooling to see if data has changed.

Frédéric Giasson
 




-----Original Message-----
From: David Woodhouse [mailto:dwmw2@infradead.org]
Sent: Wednesday, June 06, 2001 9:41 AM
To: Frederic Giasson
Cc: Abraham vd Merwe; MTD for Linux
Subject: Re: mtd_info->write 



fgiasson@mediatrix.com said:
> That's right, but a read back validation has to show the error.

There is no such read back validation in any of the MTD code, is there?
If you write to a region which already contains data, then read back and 
compare with the data you wrote, you're likely to see differences.

Film at 11. :)

--
dwmw2

Re: mtd_info->write

[David Woodhouse]

fgiasson@mediatrix.com said:
> That's right, but a read back validation has to show the error.

There is no such read back validation in any of the MTD code, is there?
If you write to a region which already contains data, then read back and 
compare with the data you wrote, you're likely to see differences.

Film at 11. :)

--
dwmw2

Re: mtd_info->write

[kira brown]

On Wed, 6 Jun 2001, David Woodhouse wrote:

> If you write to a region which already contains data, then read back and
> compare with the data you wrote, you're likely to see differences.

And if you don't see differences, then what's the problem?

kira.

Re: mtd_info->write

[David Woodhouse]

fgiasson@mediatrix.com said:
> There is validation, in do_write_oneword ( cfi_cmdset_0002.c ), there
> is pooling to see if data has changed. 

Then it's probably b0rken. It should do the appropriate mask before
comparing.

--
dwmw2

Re: mtd_info->write

[David Woodhouse]

kira@linuxgrrls.org said:
> > If you write to a region which already contains data, then read back
> > and compare with the data you wrote, you're likely to see differences.

> And if you don't see differences, then what's the problem? 

There is no problem. Either way. I may want to write 
~JFFS2_NODETYPE_ACCURATE to a word on the flash chip. I don't care about 
any bits apart from the _one_ bit I asked it to zero.


--
dwmw2

Re: mtd_info->write

[Nicolas Pitre]

On Wed, 6 Jun 2001, Abraham vd Merwe wrote:

> Hi Frederic!
>
> > If the data is all ones ("1"), it has been erased.   If not, it has not.
>
> Yes, I know, but I want to know whether I have to check if that's the case
> or not.

The chip should return an error code if it couldn't erase it correctly.
When it returns "done" then you just have to trust it, or signal the error
condition appropriately.


Nicolas