Understanding jffs2 warning message

Understanding jffs2 warning message

[Jean-Philippe Francois]

Hi,

I am doing some test on a JFFS2 file system.
It is installed on a Macronix Flash 29LV320, with top boot block.
The part are cfi probed, and it works (almost) fine.

I am doing some tar/untar loop [1]. Every six or seven loop, I have a
serie of 8 warning messages :

Node totlen on flash (0xffffffff) != totlen in node ref (0x00001040)
Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
Node totlen on flash (0xffffffff) != totlen in node ref (0x00000768)

I looked at the code, but I don't understand what is the meaning of
these messages.
Do you think it can be related to the Top Boot Block layout of this
flash (ie last 64 k sector is replaced by 8 8k sector) ?
I might be wrong, but I think the sector layout and erase size is taken
into account by CFI probing ?

Regards,

JP Francois

[1] : tar/untar loop

#!/bin/sh
i=0
while [ true ]
do
        tar -xzf /test.tar.gz
        rm /test.tar.gz
        sync
	echo "pass $i"
        md5sum /usr/test/etc/rc3
        tar -czf /test.tar.gz /usr/test/
        rm -r /usr/test/
        sync
        let i=i+1
done

Re: Understanding jffs2 warning message

[Artem B. Bityuckiy]

Jean-Philippe Francois wrote:
> I am doing some test on a JFFS2 file system.
> It is installed on a Macronix Flash 29LV320, with top boot block.
> The part are cfi probed, and it works (almost) fine.
> 
> I am doing some tar/untar loop [1]. Every six or seven loop, I have a
> serie of 8 warning messages :
> 
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001040)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
You don't do unclean reboots? If not, this is wierd. This means that the 
node on flash is somehow got corrupted.


-- 
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: Understanding jffs2 warning message

[Jean-Philippe Francois]

On Tue, 2005-07-19 at 15:06, Artem B. Bityuckiy wrote:
> Jean-Philippe Francois wrote:
> > I am doing some test on a JFFS2 file system.
> > It is installed on a Macronix Flash 29LV320, with top boot block.
> > The part are cfi probed, and it works (almost) fine.
> > 
> > I am doing some tar/untar loop [1]. Every six or seven loop, I have a
> > serie of 8 warning messages :
> > 
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001040)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> You don't do unclean reboots? If not, this is wierd. This means that the 
> node on flash is somehow got corrupted.
Well, things happen, so may be there was not so clean reboot :)
What happens to corrupted nodes ? Are they somehow garbage collected ?
If I use a partition without the top block, these warning messages do
not appear.

JP Francois

Re: Understanding jffs2 warning message

[Jean-Philippe Francois]

On Tue, 2005-07-19 at 14:35, Jean-Philippe Francois wrote:
> Hi,
> 
> I am doing some test on a JFFS2 file system.
> It is installed on a Macronix Flash 29LV320, with top boot block.
> The part are cfi probed, and it works (almost) fine.
> 
> I am doing some tar/untar loop [1]. Every six or seven loop, I have a
> serie of 8 warning messages :
> 
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001040)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> Node totlen on flash (0xffffffff) != totlen in node ref (0x00000768)
> 
> I looked at the code, but I don't understand what is the meaning of
> these messages.
> Do you think it can be related to the Top Boot Block layout of this
> flash (ie last 64 k sector is replaced by 8 8k sector) ?
> I might be wrong, but I think the sector layout and erase size is taken
> into account by CFI probing ?

Looking into the archive, I think I am in the case described by [1].
The driver know about the various erase regions, but doesn't transmit
this info to partition level. What is the status of the patch in [1] ?
It is not part of the mtd code in 2.6.12 kernel.

Regards,

JP Francois

[1]
http://lists.infradead.org/pipermail/linux-mtd/2004-December/011320.html

Re: Understanding jffs2 warning message

[Jean-Philippe Francois]

On Wed, 2005-07-20 at 09:30, Jean-Philippe Francois wrote:
> On Tue, 2005-07-19 at 14:35, Jean-Philippe Francois wrote:
> > Hi,
> > 
> > I am doing some test on a JFFS2 file system.
> > It is installed on a Macronix Flash 29LV320, with top boot block.
> > The part are cfi probed, and it works (almost) fine.
> > 
> > I am doing some tar/untar loop [1]. Every six or seven loop, I have a
> > serie of 8 warning messages :
> > 
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001040)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00001044)
> > Node totlen on flash (0xffffffff) != totlen in node ref (0x00000768)
> > 
> > I looked at the code, but I don't understand what is the meaning of
> > these messages.
> > Do you think it can be related to the Top Boot Block layout of this
> > flash (ie last 64 k sector is replaced by 8 8k sector) ?
> > I might be wrong, but I think the sector layout and erase size is taken
> > into account by CFI probing ?
> 
> Looking into the archive, I think I am in the case described by [1].
> The driver know about the various erase regions, but doesn't transmit
> this info to partition level. What is the status of the patch in [1] ?
> It is not part of the mtd code in 2.6.12 kernel.
> 
> Regards,
> 
> JP Francois
> 
> [1]
> http://lists.infradead.org/pipermail/linux-mtd/2004-December/011320.html
> 
> 
I tested this patch, and it solved my problem, in case anyone is
interested.

JP François