1308 replies to this topic

[betacentauri]

How big is the generated file /tmp/vmlinux.gz ?
Please rename the file to kernel_cfe_auto.bin and put it into backup folder and flash that backup. Does it now works?

Can you log the flash/boot process with null modem cable?

[scottyboyz]

3.88mb

[scottyboyz]

betacentauri : ill try this now and let you know

[Pedro_Newbie]

The size of the kernel looks too small, it should be, as I'm not mistaken, 4194304 bytes (4096KB, 4MB) and yours is, according to the log you placed (http://openpli.org/f...attach_id=58273 ), 4063232 bytes (3968 KB).

 

So why this is I haven't a clue?

 

 

Is there really no solo user who can confirm or deny if the back-ups work?

[betacentauri]

Yes, the file should have 4194304 bytes. The difference is 131072 bytes and that's exactly the size of 1 eraseblock. So I still think that there is a bad block, but perhaps it's not marked as bad block.
How big is the kernel shipped by OpenPli?

[Pedro_Newbie]

The size of the kernel is at the moment 3,530,628.

 

Is there an easy way to check if there is a bad block and if so mark it as bad?

[Pedro_Newbie]

@scottyboyz

can you execute this command and report the exact number of bytes of the file and the output of the command (if it differs from the previous one)

/usr/sbin/nanddump --bb=dumpbad /dev/mtd1 > /tmp/vmlinux.gz

 

It should make no difference because no bad blocks were reported, but you'll never know

[betacentauri]

The size of the kernel is at the moment 3,530,628.

 

Is there an easy way to check if there is a bad block and if so mark it as bad?

 

I heard that writing to that block should mark it as bad.

 

But it's still strange. If the kernel is only 3,530,628 big and only 1 eraseblock is missing in the dump file and the kerneldump is not usable, then it seems that the 1 missing eraseblock contains kernel data. And the bootloader can read this data(as the box normally boots) and nanddump not. How can that be??

Edited by betacentauri, 22 April 2014 - 19:48.

[Pedro_Newbie]

Maybe Milo, Pieterg or Littlesat have an idea what could be causing this and more important how to solve this?

[Lost in Space]

Well, maybe you should stop using a nanddump binary from Stoneage (version 1.29) and switch to the latest version of mtd-utils (dflash uses 1.50)

 

And a recent nanddum would offer skipbad:

 

nanddump --bb=skipbad --omitoob --quiet  /dev/mtdX > kernel_cfe_auto.bin

 

If you use a kernel in a raw mtd device, then the bootloader is clever enough to skip blocks marked bad and recent nanddump does this also nicely. You are omitting it and then your extract runs short in exactly this length.

 

At least nobody so far complained when doing backups in BA with the above command.

 

And nand_check binary in dFlash for checking for bad blocks is non-toxic :-)

Edited by gutemine, 22 April 2014 - 22:10.

[betacentauri]

I'm quite sure, that OpenPli uses a newer version, but I cannot verify it, because sourceforge is down.

I had a similar strange case some months ago: In this case the kernel partition had 1 bad eraseblock. The kernel was quite big, so that it didn't fit in the kernel partition with the 1 bad eraseblock. But the box was booting without problems and the via nanddump generated kernel dump didn't work. If somebody is interested and can understand German, here is the thread:
http://www.opena.tv/...ei-ImageBackup-(Komplett-Backup)&p=22847&viewfull=1#post22847

Link don't work properly. The interesting part starts at post 36

Edited by betacentauri, 22 April 2014 - 22:49.

[scottyboyz]

Just been reading the above quotes with some differences of opinion.  So what command should i run or do?  How do i update mtd utils to 1.50?  cheers

[Pedro_Newbie]

The version in Openpli is 1.50, at least on my Solo2.
The parameters as mentioned by you are already the default values as I read the helptext

root@vusolo2:~# /usr/sbin/nanddump --help                                      
Usage: nanddump [OPTIONS] MTD-device                                           
Dumps the contents of a nand mtd partition.                                    
                                                                               
           --help               Display this help and exit                     
           --version            Output version information and exit            
           --bb=METHOD          Choose bad block handling method (see below).  
-a         --forcebinary        Force printing of binary data to tty           
-c         --canonicalprint     Print canonical Hex+ASCII dump                 
-f file    --file=file          Dump to file                                   
-l length  --length=length      Length                                         
-n         --noecc              Read without error correction                  
           --omitoob            Omit OOB data (default)                        
-o         --oob                Dump OOB data                                  
-p         --prettyprint        Print nice (hexdump)                           
-q         --quiet              Don't display progress and status messages     
-s addr    --startaddress=addr  Start address                                  
                                                                               
--bb=METHOD, where METHOD can be `padbad', `dumpbad', or `skipbad':            
    padbad:  dump flash data, substituting 0xFF for any bad blocks             
    dumpbad: dump flash data, including any bad blocks                         
    skipbad: dump good data, completely skipping any bad blocks (default)     

root@vusolo2:~# /usr/sbin/nanddump --version        
nanddump 1.5.0                                      
                                                    
nanddump comes with NO WARRANTY                     
to the extent permitted by law.                     
                                                    
You may redistribute copies of nanddump             
under the terms of the GNU General Public Licence.  
See the file `COPYING' for more information.        

As you can see in my plugin I'm no coder/programmer whatsoever, so for a good laugh on your side take a look at the code ,

But I'm trying to make something which can benefit some people.

But when skipbad is the default why is it still running short and is there no report of finding bad blocks?

 

[Pedro_Newbie]

I'm quite sure, that OpenPli uses a newer version, but I cannot verify it, because sourceforge is down.

I had a similar strange case some months ago: In this case the kernel partition had 1 bad eraseblock. The kernel was quite big, so that it didn't fit in the kernel partition with the 1 bad eraseblock. But the box was booting without problems and the via nanddump generated kernel dump didn't work. If somebody is interested and can understand German, here is the thread:
http://www.opena.tv/...ei-ImageBackup-(Komplett-Backup)&p=22847&viewfull=1#post22847

Link don't work properly. The interesting part starts at post 36.

 

Will read this tonight after work.

 

 

 

@scottyboyz

Please execute the command in post 267

[betacentauri]

@scottyboys: Please post afterwards the created vmlinux.gz
And do you use OpenPli or another image? If not OpenPli which one do you use?

[scottyboyz]

pedro : the file is now 4mb after i copy and pasted the command on post 267.  Its the same output from the console as my previous post.  Shall i upload the file?

[scottyboyz]

Im using latest open pli 4.0 image

[scottyboyz]

When i do this command though

 

nanddump --bb=skipbad --omitoob --quiet  /dev/mtd1 > /tmp/kernel_cfe_auto.bin i get a file size of 3.88mb ?????

 

 

Should i upload both files so you can look at them?

Edited by scottyboyz, 23 April 2014 - 21:26.

[scottyboyz]

 vmlinux.gz   4MB   3 downloads

 vmlinux2.gz   3.88MB   4 downloads

 

Here you go

[Pedro_Newbie]

Did you try to restore your image with replacing the kernel_cfe_auto.bin with the biggest vmlinux.gz.

 

So rename your vmlinux.gz (the big one) to kernel_cfe_auto.bin and copy this to your corrupt back-up and try to restore it.

 

I don't see any means to help you, fact is probably that there is a badblock and that's interfering with the restore.

 

Anybody an educated guess how to go further?