Skip to contentSkip to conversation

Problems accessing USB Mass Storage

DISCUSSION Jonathan Corbet 8 participants 16 messages Sep 17, 2002 – Sep 18, 2002

lore
Share & tools
lore
#0 lore raw

Don't know if this is helpful or not, but, based on my messing around with SmartMedia USB stuff...

SmartMedia cards are weird in that they have a (seemingly) random amount of waste space at the beginning of the card. Your 8MB card, in particular, has nothing of interest in the first 25 sectors. Some cards have a reasonable partition table in the first sector, and some don't. Modern Windows systems (and cameras, of course) seem to be able to access the filesystem on the card without needing to see a partition table.

A little while I posted a Lexar SmartMedia driver patch which hacked around this by substituting a fake partition table when the first sector was read. I'm not sure it's the right solution, though. A better way, perhaps, is a little user-space program which writes the appropriate partition table depending on the card capacity. Note that fdisk doesn't (easily) work for this purpose, since it wants partitions to start on cylinder boundaries.

You might try just using dd to copy your card to disk with an offset of 25 sectors, and see of you can mount the resulting image.

Then again, the interface to some SmartMedia readers is vastly more complicated, as the sddr09 driver shows.

jon

Jonathan Corbet
Executive editor, LWN.net
[email protected]

#1 lore raw

On Tue, 17 Sep 2002, Jonathan Corbet wrote:

| Don't know if this is helpful or not, but, based on my messing around with
| SmartMedia USB stuff...
|
| SmartMedia cards are weird in that they have a (seemingly) random amount of
| waste space at the beginning of the card.  Your 8MB card, in particular,
| has nothing of interest in the first 25 sectors.  Some cards have a
| reasonable partition table in the first sector, and some don't.  Modern
| Windows systems (and cameras, of course) seem to be able to access the
| filesystem on the card without needing to see a partition table.
|
| A little while I posted a Lexar SmartMedia driver patch which hacked around
| this by substituting a fake partition table when the first sector was read.
| I'm not sure it's the right solution, though.  A better way, perhaps, is a
| little user-space program which writes the appropriate partition table
| depending on the card capacity.  Note that fdisk doesn't (easily) work for
| this purpose, since it wants partitions to start on cylinder boundaries.
|
| You might try just using dd to copy your card to disk with an offset of 25
| sectors, and see of you can mount the resulting image.
|
| Then again, the interface to some SmartMedia readers is vastly more
| complicated, as the sddr09 driver shows.

This is a bit like what we (JE, David Brownell, and I) saw at the USB plugfest in 1999. We had a camera device that we couldn't mount as a filesystem, but we could dd it. When we did that and studied the dd-ed file, we could see a FAT filesystem beginning after the first <N> blocks (but more than 25 sectors IIRC -- more like after 50-100 KB, or maybe even more).

Signature · 3 lines
~Randy
"Linux is not a research project. Never was, never will be."
  -- Linus, 2002-09-02
#2 lore raw
Quoted 6 lines · Randy.Dunlap

This is a bit like what we (JE, David Brownell, and I) saw at the USB plugfest in 1999. We had a camera device that we couldn't mount as a filesystem, but we could dd it. When we did that and studied the dd-ed file, we could see a FAT filesystem beginning after the first <N> blocks (but more than 25 sectors IIRC -- more like after 50-100 KB, or maybe even more).

Sorry to sound a bit bewildered, but would be the next best thing for me to do on this?,
I have also been advised by Jonathan Corbet
to use dd to copy your card to disk with an offset of 25

looking through the info and man pages for dd, I can find no mention of offset at all, the next best thing I could find was the command option 'skip'

Sorry to sound abit overwhelmed.

Mark

Signature · 3 lines
---
To steal ideas from one person is plagiarism;
to steal from many is research.
#3 lore raw
Quoted 11 lines
Quoted 6 lines

This is a bit like what we (JE, David Brownell, and I) saw at the USB plugfest in 1999. We had a camera device that we couldn't mount as a filesystem, but we could dd it. When we did that and studied the dd-ed file, we could see a FAT filesystem beginning after the first <N> blocks (but more than 25 sectors IIRC -- more like after 50-100 KB, or maybe even more).

Sorry to sound a bit bewildered, but would be the next best thing for me to do on this?,
I have also been advised by Jonathan Corbet
to use dd to copy your card to disk with an offset of 25

Have you tried reading from a different card? The common opinion seems to be that the actual data starts some way in to the card, and as you get an input/output error when you try to access the card from the begining, it could be that there is a genuine media error, that is not showing up when you use drivers from another OS, that do not attempt to read those first few blocks, because they 'know' where the filesystem begins.

Just an idea, anyway.

John.

#4 lore raw
In reply to Mark C
Quoted 6 lines

I have also been advised by Jonathan Corbet
to use dd to copy your card to disk with an offset of 25

looking through the info and man pages for dd, I can find no mention of offset at all, the next best thing I could find was the command option 'skip'

That's what I meant. Something like:

dd if=/dev/sda of=/somewhere/image skip=25

If that works, please let me know - maybe there *is* a place for my hackish fake partition table patch...

jon

Jonathan Corbet
Executive editor, LWN.net
[email protected]

#5 lore raw
In reply to Randy.Dunlap

Mark C. Wrote:

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

[root@stimpy dev]# dd if=/dev/sda of=/dev/null bs=1k count=1
dd: reading `/dev/sda': Input/output error
0+0 records in
0+0 records out

Quoted 6 lines · Randy.Dunlap

On Tue, 17 Sep 2002, Jonathan Corbet wrote:

|
| You might try just using dd to copy your card to disk with an offset of 25
| sectors, and see of you can mount the resulting image.
Quoted 6 lines

This is a bit like what we (JE, David Brownell, and I) saw at the USB plugfest in 1999. We had a camera device that we couldn't mount as a filesystem, but we could dd it. When we did that and studied the dd-ed file, we could see a FAT filesystem beginning after the first <N> blocks (but more than 25 sectors IIRC -- more like after 50-100 KB, or maybe even more).

See the above form Mark's post. He tried to dd a 1K block and it failed. Granted, The first few blocks may need to be skipped, but right now he cannot even get the raw data out.

If we can get the data, we can use the loop device to mount it.

Any ideas to figure out the why the dd fails?

-Thomas

#6 lore raw

The device may not actually have the beginning few sectors. Use skip= to try to read something from the middle of the media.

Yes, I actually have seen this before. The firmware 'fakes' a partition table on the first attempt to read one, but sector 0 really isn't there.

Matt

Quoted 38 lines · Thomas Dodd

Mark C. Wrote:

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

[root@stimpy dev]# dd if=/dev/sda of=/dev/null bs=1k count=1
dd: reading `/dev/sda': Input/output error
0+0 records in
0+0 records out

Quoted 6 lines · Randy.Dunlap

On Tue, 17 Sep 2002, Jonathan Corbet wrote:

|
| You might try just using dd to copy your card to disk with an offset of 25
| sectors, and see of you can mount the resulting image.
Quoted 6 lines

This is a bit like what we (JE, David Brownell, and I) saw at the USB plugfest in 1999. We had a camera device that we couldn't mount as a filesystem, but we could dd it. When we did that and studied the dd-ed file, we could see a FAT filesystem beginning after the first <N> blocks (but more than 25 sectors IIRC -- more like after 50-100 KB, or maybe even more).

See the above form Mark's post. He tried to dd a 1K block and it failed. Granted, The first few blocks may need to be skipped, but right now he cannot even get the raw data out.

If we can get the data, we can use the loop device to mount it.

Any ideas to figure out the why the dd fails?

-Thomas

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 6 lines
Matthew Dharm                              Home: [email protected] 
Maintainer, Linux USB Mass Storage Driver

How would you like this tie wrapped around your hairy round head?
					-- Greg
User Friendly, 9/2/1998
#7 lore raw
Quoted 2 lines · Matthew Dharm

The device may not actually have the beginning few sectors. Use skip= to try to read something from the middle of the media.

Give that a go Mark.

Try a few values like 25, 50, 75, and 100. with bs=1k and unset (default 512 byte).

Quoted 2 lines

Yes, I actually have seen this before. The firmware 'fakes' a partition table on the first attempt to read one, but sector 0 really isn't there.

Am I glad I have a CF camera:) What kind of
junk is SmartMedia, missing sectors? Ugh!

-Thomas

#8 lore raw
Quoted 5 lines · Thomas Dodd

Give that a go Mark.

Try a few values like 25, 50, 75, and 100. with bs=1k and unset (default 512 byte).

If I'm reading this correctly, I have been trying:

[root@stimpy mark]# dd if=/dev/sda of=tmp/tmp.img skip=50 \
bs=1k dd: reading `/dev/sda': Input/output error
0+0 records in
0+0 records out

Then the output of dmesg:

SCSI device (ioctl) reports ILLEGAL REQUEST. SCSI device sda: 16384 512-byte hdwr sectors (8 MB)
sda: test WP failed, assume Write Enabled
sda: I/O error: dev 08:00, sector 0
I/O error: dev 08:00, sector 0
unable to read partition table
I/O error: dev 08:00, sector 96

I have altered skip from 25 - 100 and received the same errors, except the sectors change in size with relation to altering the skip size.

This may be the wrong way of running the command, if so I'm sorry for wasting peoples time on that.

Mark

Signature · 3 lines
---
To steal ideas from one person is plagiarism;
to steal from many is research.
#9 lore raw
Quoted 2 lines · Mark C

[root@stimpy mark]# dd if=/dev/sda of=tmp/tmp.img skip=50 \
bs=1k dd: reading `/dev/sda': Input/output error

That's correct.

Quoted 9 lines

Then the output of dmesg:

SCSI device (ioctl) reports ILLEGAL REQUEST. SCSI device sda: 16384 512-byte hdwr sectors (8 MB)
sda: test WP failed, assume Write Enabled
sda: I/O error: dev 08:00, sector 0
I/O error: dev 08:00, sector 0
unable to read partition table
I/O error: dev 08:00, sector 96

That's what I expected. I didn't help.

Sector 96 is a little odd. 50 1K blocks should be sector 99. (counting starts at zero)

I wonder why 96 is accessed...

This is really starting to look like it's *NOT* really a Mass Storage Class device, just claiming to be. (black list time?)

I noticed the windows driver setup several other interfaces, like audio and VfW (video).

Best sugestion for now is buy 1) a memory card,
2) a card reader that works there are several.

It probably doesn't use the internal memory if a card is loaded, but you'd have to read the manul/test to find out. 8M isn't much room for pics at 1280x1024 any way :)

-Thomas

#10 lore raw
Quoted 8 lines · Thomas Dodd

Sector 96 is a little odd. 50 1K blocks should be sector 99. (counting starts at zero)

I wonder why 96 is accessed...

This is really starting to look like it's *NOT* really a Mass Storage Class device, just claiming to be. (black list time?)

Now how did I know you were going to say that :-)

Quoted 3 lines

I noticed the windows driver setup several other interfaces, like audio and VfW (video).

It can do streaming Audio/Video

Quoted 2 lines

Best sugestion for now is buy 1) a memory card,
2) a card reader that works there are several.

I have a 16mb Stmart Media Card included with the pack.

Can you possibly suggest any card readers that *will* work on Linux => 2.4.18 (ie use a specific type, usb, serial) and do they have to be compatible with specific memory cards?

Sorry to sound vague, but I've never used one before.

Just fishing :-)..I do a google...

And if so any chance of a few quick links on getting the memory card readers to work?
(don't worry too much about that, I just figured if you know any that were maybe bookmarked)

Quoted 4 lines

It probably doesn't use the internal memory if a card is loaded, but you'd have to read the manul/test to find out. 8M isn't much room for pics at 1280x1024 any way :)

True :-)

Signature · 3 lines
---
To steal ideas from one person is plagiarism;
to steal from many is research.
#11 lore raw
In reply to Mark C
Quoted 8 lines · Mark C
Quoted 5 lines · Thomas Dodd

Give that a go Mark.

Try a few values like 25, 50, 75, and 100. with bs=1k and unset (default 512 byte).

Quoted 6 lines

If I'm reading this correctly, I have been trying:

[root@stimpy mark]# dd if=/dev/sda of=tmp/tmp.img skip=50 \
bs=1k dd: reading `/dev/sda': Input/output error
0+0 records in
0+0 records out

Guys,

When dd is told to skip a certain number of input blocks it doesn't seek past them, but reads them and then discards them. Thus if you're not supposed to read sectors 1-100 then this will not work.

Try the following program:

/* seek.c (C) [email protected] */
/*
gcc -Wall -O2 seek.c -o seek
*/

#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <errno.h>
#include <sys/types.h>

#ifndef O_LARGEFILE
#define O_LARGEFILE 0100000
#endif
long long lseek64 (int fd, long long offset, int whence);

int main (int argc,char **argv)
{
long long off;
long long tt;

if(argc < 2)
exit(0); /* don't seek at all */

  if (strncmp (argv[1],"0x",2) )
    sscanf (argv[1],"%Ld",&off);
  else
    sscanf (argv[1],"%Lx",&off);

  if (argc > 3) {
    if (strncmp (argv[3],"0x",2) )
      sscanf (argv[3],"%Ld",&tt);
    else
      sscanf (argv[3],"%Lx",&tt);
    if (argv[2][0] == '+')
      off += tt;
    else
      off -= tt;
  }

errno = 0;
if ((lseek64 (0,off,SEEK_CUR) < 0) &&

      (errno != 0))
    perror ("seek");

exit (0);
}

with the command:

dd if=/dev/sda of=firstpart

(Get the partition table)

	(seek 0x100000;dd of=secondpart) < /dev/sda 

Get everything beyond 1Mb. If this works, then we have to figure out how low we can make the "0x100000" number to get all of the data.

Hypothesis: The partition table specifies that the data starts on sector 200, and they didn't implement sectors 1-199..... Cheap basterds.

(My memory stick is just over 128 * 10^6 bytes, and not even close to 128 * 2^20 bytes....)

Roger.

Signature · 5 lines
** [email protected] ** http://www.BitWizard.nl/ ** +31-15-2600998 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
* The Worlds Ecosystem is a stable system. Stable systems may experience *
* excursions from the stable situation. We are currenly in such an       * 
* excursion: The stable situation does not include humans. ***************
#12 lore raw
Quoted 5 lines · Rogier Wolff

On Tue, Sep 17, 2002 at 10:13:13PM +0100, Mark C wrote:

When dd is told to skip a certain number of input blocks it doesn't seek past them, but reads them and then discards them. Thus if you're not supposed to read sectors 1-100 then this will not work.

Fair enough. I, and the others though it did a seek.

Quoted 1 line

Try the following program:

<snip>

Quoted 6 lines

with the command:

dd if=/dev/sda of=firstpart

(Get the partition table)

Quoted 7 lines
	(seek 0x100000;dd of=secondpart) < /dev/sda 

Get everything beyond 1Mb. If this works, then we have to figure out how low we can make the "0x100000" number to get all of the data.

Hypothesis: The partition table specifies that the data starts on sector 200, and they didn't implement sectors 1-199.....

Where did the sector 200 come from? Something in the dmesg output from before? (I don't really grok SCSI or USB at that level :( )

Quoted 1 line

Cheap basterds.

Agree:)

-Thomas

#13 lore raw

On Tue, 17 Sep 2002, Thomas Dodd wrote:

| Rogier Wolff wrote:
| > On Tue, Sep 17, 2002 at 10:13:13PM +0100, Mark C wrote:
|
| > When dd is told to skip a certain number of input blocks it doesn't
| > seek past them, but reads them and then discards them. Thus if you're
| > not supposed to read sectors 1-100 then this will not work.
|
| Fair enough. I, and the others though it did a seek.
|
| > Try the following program:
| <snip>
| > with the command:
| >
| > 	dd if=/dev/sda of=firstpart
| >
| > (Get the partition table)
| >
| > 	(seek 0x100000;dd of=secondpart) < /dev/sda
| >
| > Get everything beyond 1Mb. If this works, then we have to figure out
| > how low we can make the "0x100000" number to get all of the data.
| >
| > Hypothesis: The partition table specifies that the data starts
| > on sector 200, and they didn't implement sectors 1-199.....
|
| Where did the sector 200 come from?
| Something in the dmesg output from before?
| (I don't really grok SCSI or USB at that level :( )

I think that's part of the hypothesis, but if we can read the first sector, it should be trivial to decode the partition table, if it's a typical DOS/Windows/PC-type partition table.

If someone can read the first sector, I'll be glad to decode it; just send it.

Signature · 3 lines
~Randy
"Linux is not a research project. Never was, never will be."
  -- Linus, 2002-09-02
#15 lore raw
In reply to Jonathan Corbet
Quoted 9 lines · Jonathan Corbet

SmartMedia cards are weird in that they have a (seemingly) random amount of waste space at the beginning of the card. Your 8MB card, in particular, has nothing of interest in the first 25 sectors. Some cards have a reasonable partition table in the first sector, and some don't. Modern Windows systems (and cameras, of course) seem to be able to access the filesystem on the card without needing to see a partition table.

A little while I posted a Lexar SmartMedia driver patch which hacked around this by substituting a fake partition table when the first sector was read.

The structure of SmartMedia cards is well-known and precisely documented. See, e.g., http://www.win.tue.nl/~aeb/linux/smartmedia/SmartMedia_Format.pdf
What varies is how readers present this to the outside world.

The card starts with the CIS, followed by a remapping table that tells where on the card a given user sector lives. There are also sector checksums and information about bad sectors.

Some readers have this as built-in knowledge, and present to card to the outside world as an ordinary block device.

Some readers give direct access to the bits on the card, and then understanding this remapping business is a job for the driver.

If you see non-understood space at the beginning of the card, then one might suspect that your reader is of the latter kind. The partition table will be on the card, but not on sector zero.

Andries