Hi,
I have bought a 2GB MP3 player / flash disk
that erroneously partitions and formats its storage.
The built-in firmware has an off-by-one bug that
creates the partition one cylinder larger that the
disk size allows and then it formats the VFAT fs
according to the buggy partition size. No wonder
when I try to copy large amounts of data to the
flash disk it detects errors and then remounts it
read-only.
I tried to repartition and reformat it three times
with different mformat or mkdosfs options
but as soon as I remove it from the USB port,
the device detects changed disk format and
automatically reformats itself again, so it
stays buggy.
Here's the excerpt from the logs on one occasion
I tried to copy some large stuff onto it:
Oct 3 22:56:31 host-81-17-177-202 kernel: SCSI device sdb: 4095765
512-byte hdwr sectors (2097 MB)
Oct 3 22:56:31 host-81-17-177-202 kernel: sdb: Write Protect is off
Oct 3 22:56:32 host-81-17-177-202 kernel: sdb: assuming drive cache:
write through
Oct 3 22:56:32 host-81-17-177-202 kernel: SCSI device sdb: 4095765
512-byte hdwr sectors (2097 MB)
Oct 3 22:56:32 host-81-17-177-202 kernel: sdb: Write Protect is off
Oct 3 22:56:32 host-81-17-177-202 kernel: sdb: assuming drive cache:
write through
Oct 3 22:56:33 host-81-17-177-202 kernel: sdb: sdb1
Oct 3 22:56:33 host-81-17-177-202 kernel: sdb: p1 exceeds device capacity
Oct 3 22:56:33 host-81-17-177-202 kernel: sd 6:0:0:0: Attached scsi
removable disk sdb
Oct 3 22:56:33 host-81-17-177-202 kernel: sd 6:0:0:0: Attached scsi
generic sg1 type 0
Oct 3 22:56:33 host-81-17-177-202 kernel: sd 5:0:0:0: Attached scsi
removable disk sdc
Oct 3 22:56:33 host-81-17-177-202 kernel: sd 5:0:0:0: Attached scsi
generic sg2 type 0
Oct 3 22:56:33 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095616
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095617
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095618
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095619
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095620
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095621
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095622
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095623
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095616
Oct 3 22:56:34 host-81-17-177-202 kernel: Buffer I/O error on device
sdb1, logical block 4095617
Oct 3 22:56:34 host-81-17-177-202 kernel: SELinux: initialized (dev
sdb1, type vfat), uses genfs_contexts
...
Oct 3 23:26:53 host-81-17-177-202 kernel: FAT: Filesystem panic (dev sdb1)
Oct 3 23:26:53 host-81-17-177-202 kernel: fat_get_cluster: invalid
cluster chain (i_pos 0)
Oct 3 23:26:53 host-81-17-177-202 kernel: File system has been set
read-only
...
Oct 3 23:31:35 host-81-17-177-202 kernel: FAT: Filesystem panic (dev sdb1)
Oct 3 23:31:35 host-81-17-177-202 kernel: fat_get_cluster: invalid
cluster chain (i_pos 0)
Oct 3 23:35:33 host-81-17-177-202 kernel: FAT: Filesystem panic (dev sdb1)
Oct 3 23:35:33 host-81-17-177-202 kernel: fat_bmap_cluster: request
beyond EOF (i_pos 47395744)
Oct 3 23:35:33 host-81-17-177-202 kernel: FAT: Filesystem panic (dev sdb1)
Oct 3 23:35:33 host-81-17-177-202 kernel: fat_bmap_cluster: request
beyond EOF (i_pos 47395744)
Oct 3 23:35:33 host-81-17-177-202 kernel: FAT: Filesystem panic (dev sdb1)
Oct 3 23:35:33 host-81-17-177-202 kernel: fat_bmap_cluster: request
beyond EOF (i_pos 47395744)
Oct 3 23:35:33 host-81-17-177-202 kernel: FAT: Filesystem panic (dev sdb1)
Oct 3 23:35:33 host-81-17-177-202 kernel: fat_bmap_cluster: request
beyond EOF (i_pos 47395744)
...
Unfortunately, the firmware is not upgradeable.
The device in question is a Telstar UFM-102B.
Is there a way to tell the VFAT driver to exclude
the last N sectors from the allocation strategy?
Best regards,
Zolt?n B?sz?rm?nyi
Am Montag 16 Oktober 2006 12:06 schrieb Zoltan Boszormenyi:
> Is there a way to tell the VFAT driver to exclude
> the last N sectors from the allocation strategy?
Can't you mark that clusters as bad with a diskeditor?
--
(?= =?)
//\ Prakash Punnoor /\\
V_/ \_V
Hi,
On Mon, 16 Oct 2006, Zoltan Boszormenyi wrote:
> I have bought a 2GB MP3 player / flash disk
> that erroneously partitions and formats its storage.
> The built-in firmware has an off-by-one bug that
> creates the partition one cylinder larger that the
> disk size allows and then it formats the VFAT fs
> according to the buggy partition size. No wonder
> when I try to copy large amounts of data to the
> flash disk it detects errors and then remounts it
> read-only.
>
> I tried to repartition and reformat it three times
> with different mformat or mkdosfs options
> but as soon as I remove it from the USB port,
> the device detects changed disk format and
> automatically reformats itself again, so it
> stays buggy.
[snip]
> Unfortunately, the firmware is not upgradeable.
> The device in question is a Telstar UFM-102B.
>
> Is there a way to tell the VFAT driver to exclude
> the last N sectors from the allocation strategy?
Maybe try to setup device mapper linear mapping backed by portion of that
partition that is ok (== one cylinder smaller) instead of messing with
filesystem drivers. And then create filesystem on top of that new device.
Grzegorz Kulewski
Prakash Punnoor ?rta:
> Am Montag 16 Oktober 2006 12:06 schrieb Zoltan Boszormenyi:
>
>> Is there a way to tell the VFAT driver to exclude
>> the last N sectors from the allocation strategy?
>>
>
> Can't you mark that clusters as bad with a diskeditor?
>
Can you suggest one that works on Linux?
Or which bits should I change if I use LDE?
(lde.sourceforge.net)
Zoltan Boszormenyi schrieb:
> Prakash Punnoor ?rta:
>> Am Montag 16 Oktober 2006 12:06 schrieb Zoltan Boszormenyi:
>>
>>> Is there a way to tell the VFAT driver to exclude
>>> the last N sectors from the allocation strategy?
>>>
>>
>> Can't you mark that clusters as bad with a diskeditor?
>>
>
> Can you suggest one that works on Linux?
> Or which bits should I change if I use LDE?
> (lde.sourceforge.net)
Try mbadblocks, part of the Mtools package (http://mtools.linux.lu/).
If it doesn't help, and you are brave, you may want to play with
mdoctorfat, which comes with Mtools, too, but is hidden for some reason. :->
Ren?
Rene Scharfe ?rta:
> Zoltan Boszormenyi schrieb:
>
>> Prakash Punnoor ?rta:
>>
>>> Am Montag 16 Oktober 2006 12:06 schrieb Zoltan Boszormenyi:
>>>
>>>
>>>> Is there a way to tell the VFAT driver to exclude
>>>> the last N sectors from the allocation strategy?
>>>>
>>>>
>>> Can't you mark that clusters as bad with a diskeditor?
>>>
>>>
>> Can you suggest one that works on Linux?
>> Or which bits should I change if I use LDE?
>> (lde.sourceforge.net)
>>
>
> Try mbadblocks, part of the Mtools package (http://mtools.linux.lu/).
> If it doesn't help, and you are brave, you may want to play with
> mdoctorfat, which comes with Mtools, too, but is hidden for some reason. :->
>
> Ren?
>
Thanks. I have set up my /root/.mtoolsrc and:
# mbadblocks x:
plain_io: Input/output error
Bad cluster 63985 found
plain_io: Input/output error
Bad cluster 63986 found
plain_io: Input/output error
Bad cluster 63987 found
plain_io: Input/output error
Bad cluster 63988 found
plain_io: Input/output error
Bad cluster 63989 found
After unplugging the device didn't reformat itself.
Filling the disk seems to be working. It gives me
"disk full" without any new "Buffer I/O error on device sdc1"
or FAT panic messages in the kernel logs.
Thanks and best regards,
Zolt?n B?sz?rm?nyi