TestDisk: Flash Drive="CHS and LBA don't match" [SOLVED]

[Semme]

Sylvander, it's a PILEUP! Cuz, weez'ALL *really* wantcha ta have photos on dat dare flash drive'a yers..

[Sylvander]

proebler:
1. "Does the device still not work?"
Correct.

2. "what is the problem?"
It won't do anything I'd expect a Flash Drive to do.
e.g. Won't mount; can't do anything with it in most programs like GParted.

npierce:
3. "If you still get "Disk /dev/sdb: 1010 MB, 1010826752 bytes . . .""
Yes got that.
Here's the output:

Code: Select all

# fdisk -l /dev/sdb

Disk /dev/sdb: 1010 MB, 1010826752 bytes
196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009dcf9

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    b  W95 FAT32
# 

4. Next tried:

Code: Select all

dd if=/dev/sdb bs=1 skip=446 count=64 2>/dev/null | hexdump -C

And got:

Code: Select all

# dd if=/dev/sdb bs=1 skip=446 count=64 2>/dev/null | hexdump -C
00000000  00 20 21 00 0b c3 09 7a  00 08 00 00 00 10 1e 00  |. !....z........|
00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00000040
# 

amigo:
5.

Code: Select all

# cat /proc/filesystems
nodev	sysfs
nodev	rootfs
nodev	bdev
nodev	proc
nodev	tmpfs
nodev	binfmt_misc
nodev	debugfs
nodev	sockfs
nodev	usbfs
nodev	pipefs
nodev	anon_inodefs
nodev	devpts
	reiserfs
	ext3
	ext2
	ext4
nodev	ramfs
	vfat
	msdos
	iso9660
	ntfs
	udf
	squashfs
nodev	aufs
	fuseblk
nodev	fuse
nodev	fusectl
# 

6.

Code: Select all

# lsmod
Module                  Size  Used by
cpufreq_ondemand        4361  2 
acpi_cpufreq            4301  1 
mperf                    859  1 acpi_cpufreq
i915                  284201  2 
drm_kms_helper         17719  1 i915
drm                   121515  3 i915,drm_kms_helper
video                   9383  1 i915
i2c_algo_bit            3672  1 i915
iptable_mangle           904  0 
iptable_nat             2575  0 
nf_nat                  9805  1 iptable_nat
ipt_REJECT              1517  1 
nf_conntrack_ftp        4065  0 
nf_conntrack_irc        2379  0 
iptable_filter           804  1 
xt_state                 791  4 
nf_conntrack_ipv4       7266  7 iptable_nat,nf_nat
nf_conntrack           38062  6 iptable_nat,nf_nat,nf_conntrack_ftp,nf_conntrack_irc,xt_state,nf_conntrack_ipv4
nf_defrag_ipv4           787  1 nf_conntrack_ipv4
ip_tables               7021  3 iptable_mangle,iptable_nat,iptable_filter
snd_hda_codec_hdmi     18363  1 
snd_hda_codec_realtek   135150  1 
usblp                   7516  0 
pcspkr                  1195  0 
atl1c                  22782  0 
snd_hda_intel          17167  0 
snd_hda_codec          54456  3 snd_hda_codec_hdmi,snd_hda_codec_realtek,snd_hda_intel
snd_hwdep               3708  1 snd_hda_codec
snd_pcm_oss            27363  0 
snd_mixer_oss           9850  1 snd_pcm_oss
snd_pcm                47145  4 snd_hda_codec_hdmi,snd_hda_intel,snd_hda_codec,snd_pcm_oss
snd_seq_dummy            907  0 
i2c_i801                6166  0 
i2c_core               12587  5 i915,drm_kms_helper,drm,i2c_algo_bit,i2c_i801
snd_seq_oss            19155  0 
snd_seq_midi            3248  0 
snd_seq_midi_event      3636  2 snd_seq_oss,snd_seq_midi
snd_rawmidi            11838  1 snd_seq_midi
snd_seq                32204  6 snd_seq_dummy,snd_seq_oss,snd_seq_midi,snd_seq_midi_event
snd_seq_device          3541  5 snd_seq_dummy,snd_seq_oss,snd_seq_midi,snd_rawmidi,snd_seq
snd_timer              11743  2 snd_pcm,snd_seq
shpchp                 17907  0 
snd                    33354  13 snd_hda_codec_hdmi,snd_hda_codec_realtek,snd_hda_intel,snd_hda_codec,snd_hwdep,snd_pcm_oss,snd_mixer_oss,snd_pcm,snd_seq_oss,snd_rawmidi,snd_seq,snd_seq_device,snd_timer
soundcore               3321  1 snd
snd_page_alloc          4697  2 snd_hda_intel,snd_pcm
intel_agp               7984  1 i915
intel_gtt               9153  3 i915,intel_agp
agpgart                17858  3 drm,intel_agp,intel_gtt
parport_pc             18328  0 
parport                20351  1 parport_pc
evdev                   5812  0 
processor              21886  1 acpi_cpufreq
thermal_sys             9730  2 video,processor
hwmon                    877  1 thermal_sys
button                  3275  1 i915
fuse                   47727  0 
aufs                  120424  72 
squashfs               18852  1 
# 

Semme:
7. Ready to check for corruption.

[npierce]

Sylvander, thanks for the information.

Code: Select all

196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors

. . .

Code: Select all

00000000  00 20 21 00 0b c3 09 7a  00 08 00 00 00 10 1e 00  |. !....z........|

The partition table has fields for both CHS and LBA values. If CHS values are used, they should agree with the LBA values. The CHS values on your partition table do not agree with the LBA values.

If you are unfamiliar with the terms "CHS" and "LBA", you can read the next three paragraphs for a brief explanation. But you don't really need to know what they mean, so may skip ahead if you like.

----------------------------------------------------------------
Partition tables contain the information needed to find the starting sector and the ending sector of each partition. There are two ways of doing this.

The older way was to specify the cylinder, head, and sector (CHS). So a sector was identified by where it physically lived on the disk. That way requires knowing how many sectors were on each track of the cylinder and how many heads there were. As drives got bigger it became impossible to fit the actual cylinder number into the ten-bit field allowed for it, so a bit of poetic license was used to allow bigger drives, like specifying head numbers greater than the actual number of heads. Nowadays these numbers rarely refer to actual cylinders and heads. This is clearly the case for your flash drive, which has no physical cylinders or heads.

The newer way of doing things is to just identify each sector with a number, instead of by where it lives on the drive. This is known as "Logical Block Addressing" (LBA).
----------------------------------------------------------------

Normally, Linux only uses the LBA values, so I am a little surprised by the "CHS and LBA don't match" error message. Perhaps instead of simply ignoring the CHS values, something has noticed that they don't match and assumes that the partition is not really a type b FAT partition.

Two questions:

1. When did you see that error message? (Clicking on the icon? Using pmount? Using the mount command? Using some other utility and/or another non-Puppy operating system? Using gparted? Using fdisk? Something else?)

2. Do you still get that error message, and does it still say "Incorrect number of heads/cylinder 255 (FAT) !=32 (HD)", or have those values changed?


Anyway, some utility seems to have put bad values in the CHS fields. That utility assumed that the drive had 255 heads and 63 sectors per track (not actual physical heads and tracks, of course) -- those numbers are commonly used for large drives, since they are the maximum allowed for CHS, but your drive is not so large, and fdsik indicates that it has 196 heads and 9 sectors per track. So the utility got the values wrong.

Probably the easiest way to fix this is to change your drive type from type b (W95 FAT32) to type c (W95 FAT32 (LBA)), which uses Logical Block Addressing, and so the CHS values should be ignored.

I know that you intended to set the type to c earlier, but both "fdisk -l" and the raw partition table show that it is still b.

1. Run fdisk /dev/sdb

2. Press p then Enter to and look to make sure it is working on your flash drive ("Disk /dev/sdb: 1010 MB, 1010826752 bytes . . ."). (If not, press q then Enter to quit, and tell us what it said.)

3. Press t then Enter to prepare to change the type.

4. Press 1 then Enter to choose partition 1.

5. Press c then Enter to set type c.

6. Press p then Enter and look to ensure that the Id column says c and the System column says "W95 FAT32 (LBA)" (If not, press q then Enter to quit, and tell us what it said.)

7. Press w then Enter to write to the partition table.

[amigo]

What exact command did you use to format the drive? (mkdosfs...)

[Sylvander]

npierce:
1. "When did you see that error message?"
Oh dear, I can only try to remember and guess.
Might have been in Testdisk, or Gparted, or Pmount->fdisk, or Falconfour's UBCD->XP->CheckDisk.
Hey, I notice I reported in the title, that it was TestDisk that gave that report.

2. "Do you still get that error message?"
I'd need to attempt to retrace my steps and see if I encounter that message again.
Is that really necessary?

3. Followed your steps [with tiny deviations to truly put your intent into effect], as follows:

Code: Select all

Command (m for help): m
Command action
   a   toggle a bootable flag
   b   edit bsd disklabel
   c   toggle the dos compatibility flag
   d   delete a partition
   l   list known partition types
   m   print this menu
   n   add a new partition
   o   create a new empty DOS partition table
   p   print the partition table
   q   quit without saving changes
   s   create a new empty Sun disklabel
   t   change a partition's system id
   u   change display/entry units
   v   verify the partition table
   w   write table to disk and exit
   x   extra functionality (experts only)

Command (m for help): p

Disk /dev/sdb: 1010 MB, 1010826752 bytes
196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009dcf9

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    b  W95 FAT32

Command (m for help): t
Selected partition 1
Hex code (type L to list codes): c
Changed system type of partition 1 to c (W95 FAT32 (LBA))

Command (m for help): p

Disk /dev/sdb: 1010 MB, 1010826752 bytes
196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009dcf9

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    c  W95 FAT32 (LBA)

Command (m for help): 

Then entered w to write, and the console closed.
Should it do that?
Would that write only take effect after a reboot WITH SAVE of the session changes?
I notice that when I re-run "System->Pdisk->fdisk" on sdb1, and enter the command p, the Id=b and System=w95 FAT32!
In full:

Code: Select all

Command (m for help): p

Disk /dev/sdb: 1010 MB, 1010826752 bytes
196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009dcf9

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    b  W95 FAT32

4. "The CHS values on your partition table do not agree with the LBA values.
Might we attempt to change so as to correct the CHS values?
Can you give me instructions to follow?

amigo:
5. "What exact command did you use to format the drive? (mkdosfs...)"
Originally, I used Gparted to reformat [the existing partition] from EXT3 to FAT32.
Then, through the course of this thread I've followed various instructions...
[And reported back on those]
Rather parrot fashion...
Since most are beyond my understanding.
Did any of those include yet another reformat?
GParted always reports a failure to reformat.
Also reports failure to delete the partition.

[npierce]

Hey, I notice I reported in the title, that it was TestDisk that gave that report.

Oh, ya. Sorry about that -- time I got some new glasses!

I'd need to attempt to retrace my steps and see if I encounter that message again.
Is that really necessary?

No. At least it is not a priority right now. We have something bigger to chase.

Then entered w to write, and the console closed.
Should it do that?

Oh dear, no. It should print some messages and exit back to the command prompt, like it did for you on Tuesday.

Code: Select all

Command (m for help): w
The partition table has been altered!

Calling ioctl() to re-read partition table.
Syncing disks.
# 

But it shouldn't close the terminal window.

(Note: if you run fdisk from the menu, it will close the window. To see the messages it prints before it exits, it needs to be run from a terminal window (e.g., urxvt).)

It seems to me I remember that w would sometimes not exit gracefully. Since it worked for you on Tuesday, perhaps its failure today was a fluke. Probably not, but it might be worth trying to change the type one more time. I know this is probably getting a bit tedious, but if you can get it to work, it might be easier than trying other things.

I'm going to dig up a flash drive and see if I have similar problems trying to change its partition table.

Would that write only take effect after a reboot WITH SAVE of the session changes?
I notice that when I re-run "System->Pdisk->fdisk" on sdb1, and enter the command p, the Id=b and System=w95 FAT32!

The write, if it had worked, would write to the flash drive. So it should be readable immediately -- no save to the save file is necessary.

4. "The CHS values on your partition table do not agree with the LBA values.
Might we attempt to change so as to correct the CHS values?
Can you give me instructions to follow?

That is a possibility, although it is a little more complicated. For instance, there should be a backup partition table somewhere which also would need to be changed, so I'll have to go see where that lives.

I'll dig up a flash drive and do some experiments. In the meantime, you might try using fdisk once more to change the partition type, if you've not yet run out of patience with fdisk. If it works it could save us some time.

[Semme]

NPierce- my 1g SanDisk Cruzer. Using fdisk alone, 5.2.8 has no problem with the mismatch.

Code: Select all

Disk /dev/sdb: 1027 MB, 1027416576 bytes
32 heads, 62 sectors/track, 1011 cylinders
Units = cylinders of 1984 * 512 = 1015808 bytes
Sector size (logical/physical): 512 bytes / 512 bytes

Device     Boot      Start         End      Blocks   Id  System
/dev/sdb1   *           2         530      524288   83  Linux
Partition 1 does not end on cylinder boundary.
/dev/sdb2             530        1012      478024+  83  Linux
Partition 2 does not end on cylinder boundary.

Code: Select all

00000000  80 01 03 01 83 11 a2 11  00 08 00 00 00 00 10 00  |................|
00000010  00 11 a3 11 83 0d eb f3  00 08 10 00 91 96 0e 00  |................|
00000020  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00000040

These from my 2g Cruzer >> Id=c..

Code: Select all

Disk /dev/sdb: 2055 MB, 2055021056 bytes
255 heads, 63 sectors/track, 249 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes

Device  Boot     Start    End     Blocks      Id  System
/dev/sdb1        1        250     2006825     c   W95 FAT32 (LBA)
Partition 1 has different physical/logical endings:
     phys=(248, 254, 63) logical=(249, 214, 46)

Code: Select all

00000000  00 01 01 00 0c fe 3f f8  3f 00 00 00 52 3e 3d 00  |......?.?...R>=.|
00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00000040

[Sylvander]

npierce:
1. "it needs to be run from a terminal window"
Did that, and used the write command [w], and the console window didn't close.
Instead, it gave the following [also see the end of the results of the p command prior to it]:

Code: Select all

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    c  W95 FAT32 (LBA)

Command (m for help): w
The partition table has been altered!

Calling ioctl() to re-read partition table.

WARNING: If you have created or modified any DOS 6.x
partitions, please see the fdisk manual page for additional
information.
Syncing disks.
# 

Will now close the console window [NOT using q command prior to close], re-open it, enter the p command.
Oh dear, here's what I see:

Code: Select all

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    b  W95 FAT32

2. "it might be worth trying to change the type one more time. I know this is probably getting a bit tedious, but if you can get it to work, it might be easier than trying other things"
I find this interesting, not tedious at all [not yet anyway].
This is certainly good practise in using commands in a terminal.

[npierce]

Semme,

Thanks for the report. That's good to know. Yes, since your partitions are Linux partitions, and since Linux always uses LBA, I would expect that the CHS fields are ignored.

In fact, I suspect that mount may also ignore the CHS fields even though it is mounting a type b (not-LBA) FAT partition. It is likely that the TestDisk program is simply more critical, and reported that inconsistency, since it might make the drive unusable on an old MS-DOS system.

If so, and if mount really doesn't care about the CHS fields, then the problem is elsewhere. But with few other clues, it was worth investigating the TestDisk error message. And it would be nice if simply (or not so simply, as this case seems to be) changing to type c made mount happy.

I'm hoping that mount or the kernel may provide a more specific error message in the dmesg log. Read on . . .

----------------------------------------------------------------

Sylvander,

Oh dear, here's what I see:

Code: Select all

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    b  W95 FAT32

Oh dear.

Do you know if this is one of those flash drives that have some kind of software or hardware (switch) write protection? If so, perhaps that has been confused somehow.

I'm hoping mount or the kernel may tell us more.

Running these commands, in this order, might provide enlightening output:

Code: Select all

guess_fstype /dev/sdb1
mkdir -p /mnt/sdb1
mount -t vfat /dev/sdb1 /mnt/sdb1
dmesg | tail -9

I know you already provided the output of mount. But I'm asking you to run it again because I would like to see if it writes anything to the dmesg log.

I've been playing with an old 8 GB Toshiba TransMemory flash drive. Currently, it has a type c partition at /dev/sdb1 and a type 83 (Linux) partition at /dev/sdb2. I have repeatedly been able to change partition 1 from type c to type b, remove the drive, plug it in again, verify that the type really changed with "fdisk -l /dev/sdb", and change it back again. So it can be done -- at least with my flash drive, and using the version of fdisk from util-linux-ng 2.18, which came with Racy 5.2.2.

By the way, if you still have that copy of testdisk.log that you mentioned in your first post, I'd be glad to take a look at it if you gzip it and add it as an attachment -- I'm assuming it may be too long to just insert as text. You probably know how to gzip a file, but if not, this should do it. (This assumes that it is in your /root/ directory, if not adjust accordingly.):

Code: Select all

gzip /root/testdisk.log

That's not a big priority, and it might not have anything useful in it, but if you have the time to attach it, it just might provide a clue.

I find this interesting, not tedious at all [not yet anyway].

Oh good.

[npierce]

Sylvander,

I forgot to ask: Assuming that there is an LED on your flash drive, does it blink just after you use the write command in fdisk? Mine does.

[Semme]

While we're in Q & A.. This drive come with Store 'n' Go or V-Safe software (sitting inside the opening 2048 sectors)?

[amigo]

So, try again to manually format the drive. Make sure it is unmounted and from the terminal run:
mkdosfs -F 32 /dev/sdb1

[Sylvander]

npierce:
1. "Do you know if this is one of those flash drives that have some kind of software or hardware (switch) write protection?"
I'm certain theres none.
Neither software on the drive, nor any hardware switch/protection; just a plain drive.

2. "Running these commands, in this order, might provide enlightening output:"
I'll try those now and report back.
This is what is given:

Code: Select all

# guess_fstype /dev/sdb1 
unknown
# mkdir -p /mnt/sdb1 
# mount -t vfat /dev/sdb1 /mnt/sdb1 
mount: wrong fs type, bad option, bad superblock on /dev/sdb1,
       missing codepage or helper program, or other error
       In some cases useful info is found in syslog - try
       dmesg | tail  or so

# dmesg | tail -9
[ 5420.484714] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[ 5420.489795] sd 4:0:0:0: [sdb] Asking for cache data failed
[ 5420.489799] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[ 5420.600103]  sdb: sdb1
[ 5420.603798] sd 4:0:0:0: [sdb] Asking for cache data failed
[ 5420.603801] sd 4:0:0:0: [sdb] Assuming drive cache: write through
[ 5420.603804] sd 4:0:0:0: [sdb] Attached SCSI removable disk
[ 5516.321762] FAT-fs (sdb1): bogus number of reserved sectors
[ 5516.321766] FAT-fs (sdb1): Can't find a valid FAT filesystem
# 

3. "I have repeatedly been able to change partition 1 from type c to type b, remove the drive, plug it in again, verify that the type really changed with "fdisk -l /dev/sdb", and change it back again."
I haven't been unplugging and re-plugging the Flash Drive.
Might this be why all my efforts come to naught?

4. "if you still have that copy of testdisk.log that you mentioned in your first post, I'd be glad to take a look at it if you gzip it and add it as an attachment"
I have a copy of the file saved to a different folder on the 24th Feb...
Then I appended additional info on the 27th.
I've gzip'd the later version OK.
Not sure if a .gz file will attach, but will try it.
Do I need to change the extension to txt to attach?
Apparently not.
It attached OK.
I was then able to download, unzip, and read OK.

5. "Assuming that there is an LED on your flash drive, does it blink just after you use the write command in fdisk?"
No, it doesn't.
Not a good sign, eh?
It blinks when I 1st plug it in, as the Puppy detects it [and then the Puppy displays its icon on the desktop].

6. "This drive come with Store 'n' Go or V-Safe software (sitting inside the opening 2048 bytes)?"
I believe it does NOT.
Will keep on the lookout for that.
I usually do keep a watch for stuff like that, and haven't seen any sign of it.

amigo:
7. "try again to manually format the drive. Make sure it is unmounted and from the terminal run:
mkdosfs -F 32 /dev/sdb1"
Done.
Here's what I get:

Code: Select all

# mkdosfs -F 32 /dev/sdb1
mkdosfs 3.0.11 (24 Dec 2010)
# 

[Semme]

Ya think? Dwnlds here! Is your FalconFour system enough?

[Sylvander]

Not sure what you mean Semme.

My wife had this Verbatim Flash Drive on her keychain given her by someone at her workplace, for use in her employment, but she never used it, so gave it to me.

I'm likely to have written zeros to the partition, then re-partitioned and re-formatted.
Was probably used as a bootable Puppy drive...
And then for file storage.
Eventually I either deleted all of the partition contents or wrote zeros or random patterns to the partition.
It always worked well.

Never was any sign of anything on the drive.
If there had been, I'd have eliminated it.
I don't like such stuff.

Semme:
You suggested in a PM [why don't you do such stuff here?], that I have another [and more careful] attempt at a fix using TestDisk, and that seems like a good idea to me.
Anyone think it's not a good idea, or other things should be tried first?

[Semme]

*For the sake of get'n you on your way, I'm gonna overlook the privacy breach.

==

fdisk the drive again. Delete the sole partition, then write the op. Remove, wait a few seconds, reinsert.

Pup probably won't see it now, but it's there along with your other partitions.

Code: Select all

fdisk -l

This should be empty:

Code: Select all

Device Boot      Start         End      Blocks   Id  System

And the total readout after disk /dev/sdb:?

Whether empty or not, run this:

Code: Select all

fsck /dev/sdb

The response?

==

No quoted reply- simple answers.

We can all follow..

[Sylvander]

Semme:
1. "fdisk the drive again. Delete the sole partition, then write the op."
Treat me like a beginner.
I need explicit instructions/commands on how to do this, otherwise I'm likely to get it wrong.
What's an op?

[Semme]

Unless switches are included, this type of opening never changes.

Code: Select all

fdisk /dev/sdb

For giggles, hit: m

Code: Select all

Command action
   a   toggle a bootable flag
   b   edit bsd disklabel
   c   toggle the dos compatibility flag
   d   delete a partition
   l   list known partition types
   m   print this menu
   n   add a new partition
   o   create a new empty DOS partition table
   p   print the partition table
   q   quit without saving changes
   s   create a new empty Sun disklabel
   t   change a partition's system id
   u   change display/entry units
   v   verify the partition table
   w   write table to disk and exit
   x   extra functionality (experts only)

ALWAYS hit "m" to start- so you're aware of your options.

After each (op)eration, hit: p

This allows you to view the layout and track your work.

Now we wanna (d)elete a partition, right? Right.

Since there seems to be only one partition, it'll probably submit a: 1

Go ahead: d

Whadda we wanna see? That's right- the (p)artition table again.

It's: p

If anything remains, again hit: d

Once clear, (w)rite the (op)erations to disk: w

Easy when you've spent some time with it.

[npierce]

Code: Select all

# dmesg | tail -9
. . .
[ 5516.321762] FAT-fs (sdb1): bogus number of reserved sectors
[ 5516.321766] FAT-fs (sdb1): Can't find a valid FAT filesystem 

OK, so this is what mount is choking on: "bogus number of reserved sectors". If it is looking at the correct sector (by using the LBA fields, not the CHS fields) then this would indicate a formatting error, since the number of reserved sectors would be written to the first sector (the "boot sector") of the partition by the utility that created the FAT filesystem (mkdosfs).

I see that, last Sunday, TestDisk reported the number of reserved sectors (and all of the other values it found in the boot sector) as zero, which is certainly wrong. This was after you used gparted to re-format the partition and it failed, so its not too surprising that the boot sector was incorrect.

Based on reports from Monsie who "had problems with gparted in the past when it came to formatting a disk with Windows file systems," and from proebler's experience "where gparted could no longer be used to reformat a flash drive," this could certainly be attributed to a failing of gparted. Or, of course, it could be attributed to a failing of the flash drive itself.

I noticed that when you first tried to format it with gparted, you had success, and after you started having problems, gparted then failed to format successfully, as it had done previously. That sounds ominous.

But you formatted it again on Tuesday, using mkfs.vfat (which I think is usually a symlink to mkdosfs), and then again on Thursday, using mkdosfs, so good values should have been written to it. Unfortunately, the error message that was logged when you tried to mount the partition doesn't tell us the actual value that it considered bogus.

And my guess is that there are probably other bogus numbers, the "reserved sectors" value was just the first one encountered, so the mount operation quit when it saw that.

If you haven't yet tired of trying new things, you could try this command so we can see what's currently in the first half of your boot sector:

Code: Select all

dd if=/dev/sdb1 bs=256 skip=4096 count=1 2>/dev/null | hexdump -C

[CORRECTION, Mar-04: Sorry, the above command is wrong. The correct command is:

Code: Select all

dd if=/dev/sdb bs=256 skip=4096 count=1 2>/dev/null | hexdump -C

The skip offset I gave was from the start of the drive, not the partition, so I should have said "if=/dev/sda".]

Then to see what is in the first half of the sector that the CHS start field points to, in case mkdosfs believed the CHS numbers and put the boot sector there:

Code: Select all

dd if=/dev/sdb1 bs=256 skip=640 count=1 2>/dev/null | hexdump -C

[CORRECTION, Mar-04: Sorry, the above command is also wrong. The correct command is:

Code: Select all

dd if=/dev/sdb bs=256 skip=640 count=1 2>/dev/null | hexdump -C

The skip offset I gave was from the start of the drive, not the partition, so I should have said "if=/dev/sdb".]

Note that this time if=/dev/sdb1 since we are looking at the partition, unlike the other day when we were looking at the drive so it was if=/dev/sdb.

[CORRECTION, Mar-04: So the above paragraph is also wrong. Normally we would use /dev/sdb1 to read from the partition. But since your partition table has problems I wanted the command to ignore it, and use the desired offsets from the start of the drive, so I should have said "if=/dev/sdb". I thought I'd tested the command, but somehow I managed to confuse things. Sorry about that. (See my latest post: http://www.murga-linux.com/puppy/viewto ... 480#689480).]

The bs=256 lets us deal in half-sectors, and the LBA start field indicates that the first sector of the partition is at sector 2048 on the drive, so skip=4096 (half-sectors).

Likewise, the incorrect CHS start field indicates that the first sector of the partition is at sector 320, so skip=640 (half-sectors).

I haven't been unplugging and re-plugging the Flash Drive.
Might this be why all my efforts come to naught?

.
No. Running "fdisk -l /dev/sdb" immediately after writing the partition table to the drive would show any changes you made if the write had worked. I only removed and reinserted my drive to be sure that the changes I saw were really getting written to the drive, and not just sitting in a buffer in RAM somewhere.


Thanks for attaching the TestDisk.log file. As well as the "CHS and LBA don't match" problem (for both "heads/cylinder" and "sect/track"), it showed (as mentioned above) that the boot sector of the partition wasn't created correctly.

5. "Assuming that there is an LED on your flash drive, does it blink just after you use the write command in fdisk?"
No, it doesn't.
Not a good sign, eh?

No, it's not.

The fact that it has not been possible to write to the partition table is certainly troublesome. I am curious to see if you have any better luck with writing to it when you try deleting the partition from the table.

Do you remember if the LED blinked when running mkdosfs?

It will be interesting to see what is in the partition's boot sector, to see if there is any sign of it being formatted properly. If not, perhaps it is no longer possible to write to the partition either.

[Edited 2013-Mar-04 to add above corrections in red.]

[Sylvander]

npierce:
1.

Code: Select all

dd if=/dev/sdb1 bs=256 skip=4096 count=1 2>/dev/null | hexdump -C

Gave:

Code: Select all

# dd if=/dev/sdb1 bs=256 skip=4096 count=1 2>/dev/null | hexdump -C
00000000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00000100
# 

And....
2.

Code: Select all

dd if=/dev/sdb1 bs=256 skip=640 count=1 2>/dev/null | hexdump -C

Gave:

Code: Select all

# dd if=/dev/sdb1 bs=256 skip=640 count=1 2>/dev/null | hexdump -C
00000000  28 80 00 00 28 80 01 00  28 80 02 00 28 80 03 00  |(...(...(...(...|
00000010  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00000100
# 

3. "Do you remember if the LED blinked when running mkdosfs?"
No, didn't look, but will look for an opportunity to do so ASAP.
Just tried it, and...
I thought I saw the faintest blink, but may have imagined it.
There was certainly no obvious blinking.

Semme:
4. "fdisk the drive again"
Here's what I get:

Code: Select all

# fdisk /dev/sdb

Command (m for help): m
Command action
   a   toggle a bootable flag
   b   edit bsd disklabel
   c   toggle the dos compatibility flag
   d   delete a partition
   l   list known partition types
   m   print this menu
   n   add a new partition
   o   create a new empty DOS partition table
   p   print the partition table
   q   quit without saving changes
   s   create a new empty Sun disklabel
   t   change a partition's system id
   u   change display/entry units
   v   verify the partition table
   w   write table to disk and exit
   x   extra functionality (experts only)

Command (m for help): p

Disk /dev/sdb: 1010 MB, 1010826752 bytes
196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009dcf9

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    b  W95 FAT32

Command (m for help): d
Selected partition 1

Command (m for help): p

Disk /dev/sdb: 1010 MB, 1010826752 bytes
196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009dcf9

   Device Boot      Start         End      Blocks   Id  System

Command (m for help): w
The partition table has been altered!

Calling ioctl() to re-read partition table.
Syncing disks.
# fdisk /dev/sdb

Command (m for help): p

Disk /dev/sdb: 1010 MB, 1010826752 bytes
196 heads, 9 sectors/track, 1119 cylinders, total 1974271 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009dcf9

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1            2048     1972223      985088    b  W95 FAT32

Command (m for help): 

I read that as a FAILURE to delete.
Am I correct?
p.s. The Flash Drive was plugged in, its icon [sdb1] was displayed on the desktop, and no attempt had been made to mount the partition.