FAT filesystem errors on Ipod Video 5.5 80GB

[swimmer]

Hi all,

recently I get filesystem errors when I try to update Rockbox via rbutilqt

Code: [Select]

May 28 18:20:58 swimmer FAT: Filesystem panic (dev sdd2)
May 28 18:20:58 swimmer fat_free_clusters: deleting FAT entry beyond EOF
May 28 18:20:58 swimmer File system has been set read-only

This does not  occur when I change anything by hand as in editing cfg-files or unzipping new themes ...

The Rockbox Utility
Version 1.2.1 (SVN $ Revision: 20807 $), built May 28 2009 18:01:59

[gevaerts]

This has nothing at all to do with rbutilqt. You have a corrupted filesystem. Try fsck.vfat

[swimmer]

May I doubt that I have a corrupted filesystem? It occurs *only* if I use rbutilqt and *never* when I do something manually ... or does rbutilqt modify things differently?

[bluebrother]

May I doubt that I have a corrupted filesystem?

You may, but it doesn't change the facts. Rockbox Utility is simply a program that accesses the file system using the OS. If the filesystem is broken this doesn't make a difference -- the OS has to handle the filesystem. The only disk access that is done without using the filesystem is when installing the bootloader. As the bootloader is on a different partition than the data this won't have any effect on the data partition and thus its filesystem.

Why don't you simply go and check the filesystem first? By the way, the log you pasted is from the kernel. Which definitely tells you that the kernel FAT driver has a problem with the filesystem.

[swimmer]

I see - this is a one-way thing here ;-)

I'll obey and do the filesystem checks ok?

Edit: and these are the results ...

Code: [Select]

swimmer ~ # fsck.vfat /dev/sdd
dosfsck 3.0.2, 28 Feb 2009, FAT32, LFN
Too many clusters (4884357) for FAT16 filesystem.
swimmer ~ # fdisk -l /dev/sdd
Note: sector size is 2048 (not 512)

Disk /dev/sdd: 80.0 GB, 80026361856 bytes
84 heads, 62 sectors/track, 7502 cylinders
Units = cylinders of 5208 * 2048 = 10665984 bytes
Disk identifier: 0x20202020

   Device Boot      Start         End      Blocks   Id  System
/dev/sdd1               1          10       96264    0  Empty
Partition 1 does not end on cylinder boundary.
/dev/sdd2              10        7503    78054352    b  W95 FAT32Does it make more sense now?

[gevaerts]

You checked /dev/sdd, while your filesystem is on /dev/sdd2

[swimmer]

Gnaaa - my copy+paste was not complete :-(

Code: [Select]

swimmer ~ # fsck.vfat /dev/sdd2
dosfsck 3.0.2, 28 Feb 2009, FAT32, LFN
open /dev/sdd2:No such file or directory
swimmer ~ # fsck.vfat /dev/sdd
dosfsck 3.0.2, 28 Feb 2009, FAT32, LFN
Too many clusters (4884357) for FAT16 filesystem.
swimmer ~ # fdisk -l /dev/sdd
Note: sector size is 2048 (not 512)

Disk /dev/sdd: 80.0 GB, 80026361856 bytes
84 heads, 62 sectors/track, 7502 cylinders
Units = cylinders of 5208 * 2048 = 10665984 bytes
Disk identifier: 0x20202020

   Device Boot      Start         End      Blocks   Id  System
/dev/sdd1               1          10       96264    0  Empty
Partition 1 does not end on cylinder boundary.
/dev/sdd2              10        7503    78054352    b  W95 FAT32
But thanks to your intervention I was put on the right track - there is a udev-rule that moves /dev/sdd2 to /dev/ipod and doing a 'fsck.fvat -a /dev/ipod' revealed some unused clusters which were repaired ...

Many thanks gevaerts!! 

Edit: I still don't get it why I could perform all those actions manually (editing cfg-files & installing themes) while rbutilqt bailed out immediately?!? Anyway - it all seems to work now again and I'm happy to thank you guys again ... especially gevaerts!!!   

[bluebrother]

I still don't get it why I could perform all those actions manually (editing cfg-files & installing themes) while rbutilqt bailed out immediately?!?

That's rather simple: FAT uses simply linked lists and tables holding the directory entries. If you now modify files that have broken file information the system will get into trouble: if you modify files that do not contain broken information (or seem to not contain broken information -- if for example two files pointing to the same data this can cause corruption as well) everything at least seems to work fine. Unless you check the directory structures you won't notice errors like two files pointing to the same data, so the OS won't notice it immediately.

Judging from the syslog you posted initially rbutil tried to delete a file which was broken. If you unzip a new build / theme / etc. to the player you obviously don't delete files, thus you won't notice it if a file can't get deleted properly.

[swimmer]

Well - I tried to install a theme with rbutilqt which did not succeed so I tried it manually with unzip which succeeded and this puzzled me ;-)

Thanks for your explanation and the time you invested to help me and let's hope that it's fixed now!