Question about ubiformat

Question about ubiformat

[Daniel Bowen]

Let's say that I've prepared an image using mkfs.ubifs and ubinize.  Is
there any advantage, disadvantage or difference to doing ubiformat twice -
once without specifying an image file, then running it a second time with an
image file?  In other words, is there a difference between these two
sequences:

ubiformat /dev/mtd4
ubiformat /dev/mtd4 -f /mnt/source/ubi.img

and

ubiformat /dev/mtd4 -f /mnt/source/ubi.img

Thanks!
-Daniel

Re: Question about ubiformat

[Brian Norris]

Hi Daniel,

On Thu, Aug 07, 2014 at 06:06:27PM -0600, Daniel Bowen wrote:
> Let's say that I've prepared an image using mkfs.ubifs and ubinize.  Is
> there any advantage, disadvantage or difference to doing ubiformat twice -
> once without specifying an image file, then running it a second time with an
> image file?  In other words, is there a difference between these two
> sequences:
> 
> ubiformat /dev/mtd4
> ubiformat /dev/mtd4 -f /mnt/source/ubi.img
> 
> and
> 
> ubiformat /dev/mtd4 -f /mnt/source/ubi.img

ubiformat always saves erase counters, so running it twice should not be
extra-destructive. The primary effect is that you will put the whole
device through an extra erase cycle unnecessarily (slightly reducing the
flash's lifetime), because ubiformat.c's format() function loops through
all eraseblocks and erases them.

But I think either sequence should have the same functional result.

Brian

RE: Question about ubiformat

[Daniel Bowen]

Great, thanks for the response!  So, if the mtd partition were say 500 erase
blocks large, and the image was 300 erase blocks large, would "ubiformat
/dev/mtd4 -f /mnt/source/ubi.img" 
1. only touch the first 300 blocks of the partition, and leave the other 200
unmodified?  Or 
2. would it write the image into the first 300 blocks, and erase the
remaining 200 blocks?

Thanks!
-Daniel

-----Original Message-----
From: Brian Norris [mailto:computersforpeace@gmail.com] 
Sent: Friday, August 08, 2014 12:20 AM
To: Daniel Bowen
Cc: linux-mtd@lists.infradead.org
Subject: Re: Question about ubiformat

Hi Daniel,

On Thu, Aug 07, 2014 at 06:06:27PM -0600, Daniel Bowen wrote:
> Let's say that I've prepared an image using mkfs.ubifs and ubinize.  Is
> there any advantage, disadvantage or difference to doing ubiformat twice -
> once without specifying an image file, then running it a second time with
an
> image file?  In other words, is there a difference between these two
> sequences:
> 
> ubiformat /dev/mtd4
> ubiformat /dev/mtd4 -f /mnt/source/ubi.img
> 
> and
> 
> ubiformat /dev/mtd4 -f /mnt/source/ubi.img

ubiformat always saves erase counters, so running it twice should not be
extra-destructive. The primary effect is that you will put the whole
device through an extra erase cycle unnecessarily (slightly reducing the
flash's lifetime), because ubiformat.c's format() function loops through
all eraseblocks and erases them.

But I think either sequence should have the same functional result.

Brian

Re: Question about ubiformat

[Brian Norris]

On Fri, Aug 08, 2014 at 09:10:44AM -0600, Daniel Bowen wrote:
> Great, thanks for the response!  So, if the mtd partition were say 500 erase
> blocks large, and the image was 300 erase blocks large, would "ubiformat
> /dev/mtd4 -f /mnt/source/ubi.img" 
> 1. only touch the first 300 blocks of the partition, and leave the other 200
> unmodified?  Or 
> 2. would it write the image into the first 300 blocks, and erase the
> remaining 200 blocks?

It erases every block, writes the image to the first 300, and programs
EC headers to the last 200.

The source for ubiformat is actually pretty simple and easy to read.
Check it out yourself!

  http://git.infradead.org/mtd-utils.git

Brian

Re: Question about ubiformat

[Artem Bityutskiy]

On Fri, 2014-08-08 at 09:27 -0700, Brian Norris wrote:
> It erases every block, writes the image to the first 300, and programs
> EC headers to the last 200.

Yes, confirmed.

-- 
Best Regards,
Artem Bityutskiy

RE: Question about ubiformat

[Daniel Bowen]

> > It erases every block, writes the image to the first 300, and programs
> > EC headers to the last 200.
>
> Yes, confirmed.

Perfect - thanks so much to both of you!

-Daniel