Access entire device with mtdblock0

Access entire device with mtdblock0

[Adam Yergovich]

Hello MTD,

There was a grand feature in the 2.4 flavor of MTD that i used wherein 
mtdblock0 was assigned the entirety of the device.  This made updating 
the flash image very easy.

I've since started messing around with 2.6 flavors, and while i have a 
working map that gives the partitions i can no longer access the entire 
device.  I've tried a few things without success to get it to work.

Any recommendations on a simple way to create a partition using the 
entirety of flash?

-Adam

Re: Access entire device with mtdblock0

[Jamie Lokier]

Adam Yergovich wrote:
> There was a grand feature in the 2.4 flavor of MTD that i used wherein 
> mtdblock0 was assigned the entirety of the device.  This made updating 
> the flash image very easy.
> 
> I've since started messing around with 2.6 flavors, and while i have a 
> working map that gives the partitions i can no longer access the entire 
> device.  I've tried a few things without success to get it to work.
> 
> Any recommendations on a simple way to create a partition using the 
> entirety of flash?

Hey, I use that feature with 2.4 too, and I'm looking at porting a
device to 2.6.  I'll be interested in the answer.

-- Jamie

Re: Access entire device with mtdblock0

[Mike Frysinger]

On Fri, Jan 9, 2009 at 19:58, Adam Yergovich wrote:
> There was a grand feature in the 2.4 flavor of MTD that i used wherein
> mtdblock0 was assigned the entirety of the device.  This made updating
> the flash image very easy.
>
> I've since started messing around with 2.6 flavors, and while i have a
> working map that gives the partitions i can no longer access the entire
> device.  I've tried a few things without success to get it to work.
>
> Any recommendations on a simple way to create a partition using the
> entirety of flash?

looks like no one responded.  i'm pretty sure the answer is to use
add_mtd_device() even in face of add_mtd_partitions().
-mike

Re: Access entire device with mtdblock0

[Peter Korsgaard]

>>>>> "Mike" == Mike Frysinger <vapier.adi@gmail.com> writes:

Hi,

 >> Any recommendations on a simple way to create a partition using the
 >> entirety of flash?

 Mike> looks like no one responded.  i'm pretty sure the answer is to use
 Mike> add_mtd_device() even in face of add_mtd_partitions().

Why not just add a partition spanning the entire device in addition to
the other partitions?

-- 
Bye, Peter Korsgaard

Re: Access entire device with mtdblock0

[Mike Frysinger]

On Wed, Jan 14, 2009 at 04:00, Peter Korsgaard wrote:
> "Mike" == Mike Frysinger <writes:
>>> Any recommendations on a simple way to create a partition using the
>>> entirety of flash?
>>
>> looks like no one responded.  i'm pretty sure the answer is to use
>> add_mtd_device() even in face of add_mtd_partitions().
>
> Why not just add a partition spanning the entire device in addition to
> the other partitions?

i was not aware partition tables allowed for overlaps.  if that's
true, then the answer is obvious and easy to implement in boards ...
-mike

Re: Access entire device with mtdblock0

[Peter Korsgaard]

>>>>> "Mike" == Mike Frysinger <vapier.adi@gmail.com> writes:

Hi,

 >> Why not just add a partition spanning the entire device in
 >> addition to the other partitions?

 Mike> i was not aware partition tables allowed for overlaps.  if
 Mike> that's true, then the answer is obvious and easy to implement
 Mike> in boards ...

I'm pretty sure it does, and I don't see any checks for it in
mtdpart.c:add_one_partition()

-- 
Bye, Peter Korsgaard

Re: Access entire device with mtdblock0

[Ricard Wanderlof]

On Wed, 14 Jan 2009, Mike Frysinger wrote:

> On Wed, Jan 14, 2009 at 04:00, Peter Korsgaard wrote:
>> "Mike" == Mike Frysinger <writes:
>>>> Any recommendations on a simple way to create a partition using the
>>>> entirety of flash?
>>>
>> Why not just add a partition spanning the entire device in addition to
>> the other partitions?
>
> i was not aware partition tables allowed for overlaps.  if that's
> true, then the answer is obvious and easy to implement in boards ...

Sure does. We've used it for quite a while, using /dev/mtd0 to refer to 
the whole flash, with /dev/mtd1,2,3... referring to the individual 
partitions.

/Ricard
--
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

Re: Access entire device with mtdblock0

[Wolfgang Wegner]

On Wed, Jan 14, 2009 at 05:58:43AM -0500, Mike Frysinger wrote:
> On Wed, Jan 14, 2009 at 04:00, Peter Korsgaard wrote:
> > "Mike" == Mike Frysinger <writes:
> >>> Any recommendations on a simple way to create a partition using the
> >>> entirety of flash?
> >>
> >> looks like no one responded.  i'm pretty sure the answer is to use
> >> add_mtd_device() even in face of add_mtd_partitions().
> >
> > Why not just add a partition spanning the entire device in addition to
> > the other partitions?
> 
> i was not aware partition tables allowed for overlaps.  if that's
> true, then the answer is obvious and easy to implement in boards ...

They are allowed, but I have not been able yet to do a correct query
for erase_size in case of a partition covering a whole device having
multiple different-sized erase regions.

Regards,
Wolfgang

Re: Access entire device with mtdblock0

[Adam Yergovich]

Wolfgang Wegner wrote:
> On Wed, Jan 14, 2009 at 05:58:43AM -0500, Mike Frysinger wrote:
>> On Wed, Jan 14, 2009 at 04:00, Peter Korsgaard wrote:
>>> "Mike" == Mike Frysinger <writes:
>>>>> Any recommendations on a simple way to create a partition using the
>>>>> entirety of flash?
>>>> looks like no one responded.  i'm pretty sure the answer is to use
>>>> add_mtd_device() even in face of add_mtd_partitions().
>>> Why not just add a partition spanning the entire device in addition to
>>> the other partitions?
>> i was not aware partition tables allowed for overlaps.  if that's
>> true, then the answer is obvious and easy to implement in boards ...
> 
> They are allowed, but I have not been able yet to do a correct query
> for erase_size in case of a partition covering a whole device having
> multiple different-sized erase regions.

Thanks for the advice all.  I will give these a go in the next day or two.

-Adam

Re: Access entire device with mtdblock0

[Adam Yergovich]

Wolfgang Wegner wrote:
> On Wed, Jan 14, 2009 at 05:58:43AM -0500, Mike Frysinger wrote:
>> On Wed, Jan 14, 2009 at 04:00, Peter Korsgaard wrote:
>>> "Mike" == Mike Frysinger <writes:
>>>>> Any recommendations on a simple way to create a partition using the
>>>>> entirety of flash?
>>>> looks like no one responded.  i'm pretty sure the answer is to use
>>>> add_mtd_device() even in face of add_mtd_partitions().
>>> Why not just add a partition spanning the entire device in addition to
>>> the other partitions?
>> i was not aware partition tables allowed for overlaps.  if that's
>> true, then the answer is obvious and easy to implement in boards ...
> 
> They are allowed, but I have not been able yet to do a correct query
> for erase_size in case of a partition covering a whole device having
> multiple different-sized erase regions.
> 
> Regards,
> Wolfgang

Hello All,

Sorry for the delay in trying this out, I have been exceptionally busy.

The short story is: i got it all to work.

The long story is:

In reviewing the code i found that add_mtd_device was already being 
called after the partitioning.  As it wasn't giving mtdblock0 as the 
entire region it didn't seem to be the solution.

I played around some more, and found disabling the mtd partitioning 
would give me one partition including all of flash.  If you'll note in 
the code (which has been present in physmap.c for some time i believe) 
it was getting to the add_mtd_device and adding it if and only if the 
redboot section was not called.

I moved the add_mtd_device() call to before the redboot partitioning and 
it all worked brilliantly.

I would like to suggest that this is perhaps a better construct overall 
as it adds a nice feature.

Hopefully this is at very least useful to someone who looks at this in 
the future.  I thank you all for your help in this.

-Adam

RE: Access entire device with mtdblock0

[hartleys]

On Friday, January 30, 2009 5:08 PM, Adam Yergovich wrote:
> Wolfgang Wegner wrote:
>> On Wed, Jan 14, 2009 at 05:58:43AM -0500, Mike Frysinger wrote:
>>> On Wed, Jan 14, 2009 at 04:00, Peter Korsgaard wrote:
>>>> "Mike" == Mike Frysinger <writes:
>>>>>> Any recommendations on a simple way to create a partition using 
>>>>>> the entirety of flash?
>>>>> looks like no one responded.  i'm pretty sure the answer is to use
>>>>> add_mtd_device() even in face of add_mtd_partitions().
>>>> Why not just add a partition spanning the entire device in addition

>>>> to the other partitions?
>>> i was not aware partition tables allowed for overlaps.  if that's 
>>> true, then the answer is obvious and easy to implement in boards ...
>> 
>> They are allowed, but I have not been able yet to do a correct query 
>> for erase_size in case of a partition covering a whole device having 
>> multiple different-sized erase regions.
>> 
>> Regards,
>> Wolfgang
>
> Hello All,
> 
> Sorry for the delay in trying this out, I have been exceptionally
busy.
> 
> The short story is: i got it all to work.
> 
> The long story is:
> 
> In reviewing the code i found that add_mtd_device was already being
> called after the partitioning.  As it wasn't giving mtdblock0 as the
> entire region it didn't seem to be the solution.
> 
> I played around some more, and found disabling the mtd partitioning
> would give me one partition including all of flash.  If you'll note
> in the code (which has been present in physmap.c for some time I
> believe) it was getting to the add_mtd_device and adding it if and
> only if the redboot section was not called.
> 
> I moved the add_mtd_device() call to before the redboot partitioning
> and it all worked brilliantly.
> 
> I would like to suggest that this is perhaps a better construct
> overall as it adds a nice feature.
> 
> Hopefully this is at very least useful to someone who looks at this
> in the future.  I thank you all for your help in this.

It appears this was originally thought about for mtd partitions.

In mdtpart.c, add_mtd_partitions() has a comment above it:

/*
 * This function, given a master MTD object and a partition table,
creates
 * and registers slave MTD objects which are bound to the master
according to
 * the partition definitions.
 * (Q: should we register the master MTD object as well?)
 */

Maybe it was just never revisited?

Hartley

Re: Access entire device with mtdblock0

[Adam Yergovich]

hartleys wrote:
> It appears this was originally thought about for mtd partitions.
> 
> In mdtpart.c, add_mtd_partitions() has a comment above it:
> 
> /*
>  * This function, given a master MTD object and a partition table,
> creates
>  * and registers slave MTD objects which are bound to the master
> according to
>  * the partition definitions.
>  * (Q: should we register the master MTD object as well?)
>  */
> 
> Maybe it was just never revisited?
> 

Could well be, some great sleuthing there.  I think the bigger question 
is "Does calling add_mtd_device() after the partitioning has been done 
work?"

I believe the answer to this is no, though some confirmation would be a 
good idea.

If that assumption is correct, I think it should be changed to either:

1) Call it before partitioning so it does add the device.

-or-

2) Change the function such that it works any time and thus the physmap 
call should work as originally written.

I'm not familiar enough with the intended function to know though.  The 
best i can say is that it seems to behave in this way, and it appears to 
me to have been intended to behave differently.

-Adam

Re: Access entire device with mtdblock0

[Peter Korsgaard]

>>>>> "Adam" == Adam Yergovich <ayergo@jkmicro.com> writes:

Hi,

 Adam> If that assumption is correct, I think it should be changed to either:

 Adam> 1) Call it before partitioning so it does add the device.

 Adam> -or-

 Adam> 2) Change the function such that it works any time and thus the physmap 
 Adam> call should work as originally written.

It would be bad to change this now to do it automatically as it would
screw up the mtd numbering.

Why not just create an extra partition with the entire size if you
need it?

-- 
Bye, Peter Korsgaard