linux-ntfs-dev — Development discussions and bug reports.

[Linux-NTFS-Dev] [ linux-ntfs-Bugs-959924 ] the write support of NTFS driver

[SourceForge.net <no...@so...>]

Bugs item #959924, was opened at 2004-05-25 07:46
Message generated for change (Comment added) made by antona
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=959924&group_id=13956

Category: None
Group: None
>Status: Closed
>Resolution: Invalid
Priority: 5
Submitted By: Nobody/Anonymous (nobody)
>Assigned to: Anton Altaparmakov (antona)
Summary: the write support of NTFS driver

Initial Comment:
In the document of linux ntfs, I learned that "The only 
supported write operation is overwriting existing files, 
without changing the file length." So I made some 
experiments to test this characteristic, but encountered 
some problems.

I installed two operating system on my computer, one
is redhat linux, the other is Windows server 2003.

The kernel is linux 2.6.6, and the output of 
"dmesg | grep -i ntfs" are:
    NTFS driver 2.1.8 [Flags: R/W DEBUG].
    NTFS volume version 3.1.

Firstly, I made the following experiment:
(1) Under the Windows 2003, I creat a simple text 
file "1.txt", it only consists of one line of characters,
and the size of the file is 21 bytes;

(2) I reboot the system and switch to the linux, and 
then mount the corresponding ntfs partition:
    mount  /dev/hda6  /mnt/ntfs
this command was completed successfully because when
I typed "mount", I saw:
    ....
    /dev/hda6 on /mnt/ntfs type ntfs (rw)

(3)Then I use the vi editor to open the "1.txt" file and 
try to replace one of its character by another 
character,  which means the content of the file is going 
to change while the size of the file remains the same. 

(4) I tried to save the modification and quit, but failed,
the error message is:
"1.txt" NTFS-fs error (device hda6): ntfs_setattr(): 
Changes in i_size are not supported yet. Sorry.

But I don't think I have changed the file length, so I
am wondering why this experiment failed. Did I do 
something wrong?

Then I remembered that in the document of linux ntfs, 
I read "Note only non-resident files can be written to, 
so you may find that some very small files (<500 bytes 
or so) cannot be written to. "

So I made the second experiment, and created a bigger
file "2.txt", whose length is 12, 324 bytes, and repeated
the above steps, but the result is the same.

Can you tell me why?

Thanks a lot!

Weijun Chen

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

>Comment By: Anton Altaparmakov (antona)
Date: 2004-05-25 09:29

Message:
Logged In: YES 
user_id=42205

Hi,

The problem you are seeing is because you used a text editor
to write to the file.  What these programs do is to truncate
the file to zero size and then write it from scratch and
NTFS doesn't support the truncate to zero size operation so
this fails.  You need to use a disk editor like hexedit or
dd or write your own program so that your file size is
really unaffected.

The second problem is that small files will indeed not be
written out to disk but the write will actually work, it
will just never hit the disk.  The NTFS driver will emit a
warning message telling you this.

Best regards,

Anton

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=959924&group_id=13956

[Linux-NTFS-Dev] [ linux-ntfs-Bugs-958982 ] Doesn't handle ntfs vol created with win2k 'convert'?

[SourceForge.net <no...@so...>]

Bugs item #958982, was opened at 2004-05-23 17:47
Message generated for change (Comment added) made by antona
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

Category: None
Group: None
Status: Open
>Resolution: None
Priority: 5
Submitted By: Andrew Smith (asqm)
Assigned to: Anton Altaparmakov (antona)
Summary: Doesn't handle ntfs vol created with win2k 'convert'?

Initial Comment:
I had 1 ntfs volume and 1 fat32 volume used under win2k
Both worked fine using respective ntfs and vfat
filesystems under fc1

I booted into win2k and 'convert d: /fs:ntfs' (d: was
fat32=/dev/hdb1)
Volume was OK in win2k after convert.
Booted into fc1.
  mount -t ntfs /dev/hdb1 /win2
Succeeds with dmesg = 'NTFS: Warning! NTFS volume
version is Win2k+: Mounting read-only'
However
  ls /win2
Fails with error: 'ls: reading directory /win2: Not a
directory'
and dmegs = 'NTFS: ntfs_getdir_unsorted(): Index block
0x2 is not an index record. Returning -ENOTDIR.'
However, booting back and forward between win2k and fc1
the drive is always OK under win2k
Also, fdisk show the partition as 'id=7' as expected.
And, of course, the original ntfs partition still reads
fine.

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

>Comment By: Anton Altaparmakov (antona)
Date: 2004-05-25 09:26

Message:
Logged In: YES 
user_id=42205

Hi,

Thanks for trying the new driver!

The error message sounds like a corrupt volume rather than a
driver bug so I would be very interested in the results from
running windows chkdsk /f.  If that doesn't report any
errors and you still get the same error message from the new
ntfs driver, please report back so we can investigate this
further.

Best regards,

Anton

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

Comment By: Andrew Smith (asqm)
Date: 2004-05-24 10:12

Message:
Logged In: YES 
user_id=271661

Problem still occurs with kernel 2.6.5-1.358 using the
matching fc2 RPM
dmesg on mount is: 'NTFS volume version 3.0.'
error in ls is: 'ls: reading directory /win2: Input/output
error'
dmesg on ls is: 'NTFS-fs error (device hdb1):
ntfs_readdir(): Actual VCN (0x225a9985efd829ae) of index
buffer is different from expected VCN (0x10). Directory
inode 0x5 is corrupt or driver bug.'

I will run a fs check under win2k to see if it shows up any
errors or removes the error - just to be sure it isn't that
- and then add another comment here

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

Comment By: Anton Altaparmakov (antona)
Date: 2004-05-23 18:03

Message:
Logged In: YES 
user_id=42205

Hi,

From the dmesg outpu it is clear that you are using the old 
NTFS driver which we no longer support/develop.

Could you try using the new driver?  You can either get a 
patch for 2.4 vanilla kernels from our website or use a 2.6 
kernel (2.6 has the new driver).

If you still experience problems with the new driver than please 
let us know / reopen the bug report.

Thanks,

Anton

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

[Linux-NTFS-Dev] [ linux-ntfs-Bugs-959924 ] the write support of NTFS driver

[SourceForge.net <no...@so...>]

Bugs item #959924, was opened at 2004-05-24 23:46
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=959924&group_id=13956

Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: Nobody/Anonymous (nobody)
Assigned to: Nobody/Anonymous (nobody)
Summary: the write support of NTFS driver

Initial Comment:
In the document of linux ntfs, I learned that "The only 
supported write operation is overwriting existing files, 
without changing the file length." So I made some 
experiments to test this characteristic, but encountered 
some problems.

I installed two operating system on my computer, one
is redhat linux, the other is Windows server 2003.

The kernel is linux 2.6.6, and the output of 
"dmesg | grep -i ntfs" are:
    NTFS driver 2.1.8 [Flags: R/W DEBUG].
    NTFS volume version 3.1.

Firstly, I made the following experiment:
(1) Under the Windows 2003, I creat a simple text 
file "1.txt", it only consists of one line of characters,
and the size of the file is 21 bytes;

(2) I reboot the system and switch to the linux, and 
then mount the corresponding ntfs partition:
    mount  /dev/hda6  /mnt/ntfs
this command was completed successfully because when
I typed "mount", I saw:
    ....
    /dev/hda6 on /mnt/ntfs type ntfs (rw)

(3)Then I use the vi editor to open the "1.txt" file and 
try to replace one of its character by another 
character,  which means the content of the file is going 
to change while the size of the file remains the same. 

(4) I tried to save the modification and quit, but failed,
the error message is:
"1.txt" NTFS-fs error (device hda6): ntfs_setattr(): 
Changes in i_size are not supported yet. Sorry.

But I don't think I have changed the file length, so I
am wondering why this experiment failed. Did I do 
something wrong?

Then I remembered that in the document of linux ntfs, 
I read "Note only non-resident files can be written to, 
so you may find that some very small files (<500 bytes 
or so) cannot be written to. "

So I made the second experiment, and created a bigger
file "2.txt", whose length is 12, 324 bytes, and repeated
the above steps, but the result is the same.

Can you tell me why?

Thanks a lot!

Weijun Chen

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=959924&group_id=13956

[Linux-NTFS-Dev] [ linux-ntfs-Bugs-958982 ] Doesn't handle ntfs vol created with win2k 'convert'?

[SourceForge.net <no...@so...>]

Bugs item #958982, was opened at 2004-05-24 02:47
Message generated for change (Comment added) made by asqm
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

Category: None
Group: None
>Status: Open
Resolution: Out of Date
Priority: 5
Submitted By: Andrew Smith (asqm)
Assigned to: Anton Altaparmakov (antona)
Summary: Doesn't handle ntfs vol created with win2k 'convert'?

Initial Comment:
I had 1 ntfs volume and 1 fat32 volume used under win2k
Both worked fine using respective ntfs and vfat
filesystems under fc1

I booted into win2k and 'convert d: /fs:ntfs' (d: was
fat32=/dev/hdb1)
Volume was OK in win2k after convert.
Booted into fc1.
  mount -t ntfs /dev/hdb1 /win2
Succeeds with dmesg = 'NTFS: Warning! NTFS volume
version is Win2k+: Mounting read-only'
However
  ls /win2
Fails with error: 'ls: reading directory /win2: Not a
directory'
and dmegs = 'NTFS: ntfs_getdir_unsorted(): Index block
0x2 is not an index record. Returning -ENOTDIR.'
However, booting back and forward between win2k and fc1
the drive is always OK under win2k
Also, fdisk show the partition as 'id=7' as expected.
And, of course, the original ntfs partition still reads
fine.

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

>Comment By: Andrew Smith (asqm)
Date: 2004-05-24 19:12

Message:
Logged In: YES 
user_id=271661

Problem still occurs with kernel 2.6.5-1.358 using the
matching fc2 RPM
dmesg on mount is: 'NTFS volume version 3.0.'
error in ls is: 'ls: reading directory /win2: Input/output
error'
dmesg on ls is: 'NTFS-fs error (device hdb1):
ntfs_readdir(): Actual VCN (0x225a9985efd829ae) of index
buffer is different from expected VCN (0x10). Directory
inode 0x5 is corrupt or driver bug.'

I will run a fs check under win2k to see if it shows up any
errors or removes the error - just to be sure it isn't that
- and then add another comment here

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

Comment By: Anton Altaparmakov (antona)
Date: 2004-05-24 03:03

Message:
Logged In: YES 
user_id=42205

Hi,

From the dmesg outpu it is clear that you are using the old 
NTFS driver which we no longer support/develop.

Could you try using the new driver?  You can either get a 
patch for 2.4 vanilla kernels from our website or use a 2.6 
kernel (2.6 has the new driver).

If you still experience problems with the new driver than please 
let us know / reopen the bug report.

Thanks,

Anton

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

[Linux-NTFS-Dev] [ linux-ntfs-Bugs-958982 ] Doesn't handle ntfs vol created with win2k 'convert'?

[SourceForge.net <no...@so...>]

Bugs item #958982, was opened at 2004-05-23 17:47
Message generated for change (Comment added) made by antona
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

Category: None
Group: None
>Status: Closed
>Resolution: Out of Date
Priority: 5
Submitted By: Andrew Smith (asqm)
>Assigned to: Anton Altaparmakov (antona)
Summary: Doesn't handle ntfs vol created with win2k 'convert'?

Initial Comment:
I had 1 ntfs volume and 1 fat32 volume used under win2k
Both worked fine using respective ntfs and vfat
filesystems under fc1

I booted into win2k and 'convert d: /fs:ntfs' (d: was
fat32=/dev/hdb1)
Volume was OK in win2k after convert.
Booted into fc1.
  mount -t ntfs /dev/hdb1 /win2
Succeeds with dmesg = 'NTFS: Warning! NTFS volume
version is Win2k+: Mounting read-only'
However
  ls /win2
Fails with error: 'ls: reading directory /win2: Not a
directory'
and dmegs = 'NTFS: ntfs_getdir_unsorted(): Index block
0x2 is not an index record. Returning -ENOTDIR.'
However, booting back and forward between win2k and fc1
the drive is always OK under win2k
Also, fdisk show the partition as 'id=7' as expected.
And, of course, the original ntfs partition still reads
fine.

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

>Comment By: Anton Altaparmakov (antona)
Date: 2004-05-23 18:03

Message:
Logged In: YES 
user_id=42205

Hi,

From the dmesg outpu it is clear that you are using the old 
NTFS driver which we no longer support/develop.

Could you try using the new driver?  You can either get a 
patch for 2.4 vanilla kernels from our website or use a 2.6 
kernel (2.6 has the new driver).

If you still experience problems with the new driver than please 
let us know / reopen the bug report.

Thanks,

Anton

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

[Linux-NTFS-Dev] [ linux-ntfs-Bugs-958982 ] Doesn't handle ntfs vol created with win2k 'convert'?

[SourceForge.net <no...@so...>]

Bugs item #958982, was opened at 2004-05-24 02:47
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: Andrew Smith (asqm)
Assigned to: Nobody/Anonymous (nobody)
Summary: Doesn't handle ntfs vol created with win2k 'convert'?

Initial Comment:
I had 1 ntfs volume and 1 fat32 volume used under win2k
Both worked fine using respective ntfs and vfat
filesystems under fc1

I booted into win2k and 'convert d: /fs:ntfs' (d: was
fat32=/dev/hdb1)
Volume was OK in win2k after convert.
Booted into fc1.
  mount -t ntfs /dev/hdb1 /win2
Succeeds with dmesg = 'NTFS: Warning! NTFS volume
version is Win2k+: Mounting read-only'
However
  ls /win2
Fails with error: 'ls: reading directory /win2: Not a
directory'
and dmegs = 'NTFS: ntfs_getdir_unsorted(): Index block
0x2 is not an index record. Returning -ENOTDIR.'
However, booting back and forward between win2k and fc1
the drive is always OK under win2k
Also, fdisk show the partition as 'id=7' as expected.
And, of course, the original ntfs partition still reads
fine.

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=113956&aid=958982&group_id=13956

Re: [Linux-NTFS-Dev] ldmutil code query

[Nasir H. <son...@ga...>]

hi richard,

>> the LDM Documentation specifies that the PRIVHEAD ,512 bytes Private
>> Header) is located immedietly after the (dummy) partition table. So
>> picking up the first 2 sectors of the physical disk should get the
>> partition table and PRIVHEAD.
>
> Hmm, that's not right.  I've just read all my docs, and I don't actually
> mention where to find the PRIVHEAD.

this is what /ldmdoc/overview/database.html says :

"Just after the dummy partition table, there is the first LDM structure.  
The Private Header (PRIVHEAD) is a table of accounting information."

I got my conclusions from 'just after the dummy partition table...'.  
Would'nt that point to the 2nd sector ( if the PT resides on the 1st one) ?

>
>> This is an extract from /ldm/ldmutil/main.cpp (ldmutil code) :
>> for(i = 0; i < 7; i++) {
>> Why are the picking up the first 7 sectors instead of picking up only  
>> the
>> first 2 ?
>
> The PRIVHEAD is in the 7th sector (i.e. offset 0xC00).

thanks

nasir

-- 
while(Strarting Windoze)
{
	Illegal OS Error;
	Reboot;
	Boot GNU/Linux;
} ;-)

Re: [Linux-NTFS-Dev] ldmutil code query

[Richard R. <nt...@fl...>]

Hi Nassir,

> the LDM Documentation specifies that the PRIVHEAD ,512 bytes Private  
> Header) is located immedietly after the (dummy) partition table. So  
> picking up the first 2 sectors of the physical disk should get the  
> partition table and PRIVHEAD.

Hmm, that's not right.  I've just read all my docs, and I don't actually
mention where to find the PRIVHEAD.

> This is an extract from /ldm/ldmutil/main.cpp (ldmutil code) :
> for(i = 0; i < 7; i++) {
> Why are the picking up the first 7 sectors instead of picking up only the  
> first 2 ?

The PRIVHEAD is in the 7th sector (i.e. offset 0xC00).

Cheers,
  FlatCap (Rich)

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

On Thu, 20 May 2004 ch...@to... wrote:

> As far as real windows booting is concerned I kinda doubt that that
> process can be hyjacked to boot a linux kernel probably not even some kind
> of custom bootloader. 

So it sounds like a real challange ;)

> If I remember correctly the windows boot process uses the ntldr to
> load other stuff before the windows kernel such as the registry and
> hardware detection code (it's been a while since I read over the
> details commonly avaliable for windows boot process.) If it were any
> less than moderatly difficult to change this process I would think
> microsoft would have used that instead of the load bootsector method
> to load thier own recovery console.

I see. Considering this, your approach makes sense to me now ... 
 
> I guess when I said about avoiding fragmentation I was giving MS too much
> credit. I always hear that the need for defragmentation in *NIX
> filesystems is rare because they try to avoid fragmentation in the first
> place. So I assumed that if MS was building a modern filesystem that they
> would use some of these methods.

True and not. The design has considered but the implementation could be
improved. E.g. concurrent IO's can result very bad fragmentations. The
interesting thing is, in some benchmarks this approach really fly. But in
real world usage, well ...
 
> The recovery console is a piece of crap. It cannot run any external
> commands (although it can run special batch files made of RC internal
> commands) I guess they figured that it would be a security feature.
> But that seems to be based on the assumption that no third party would
> ever figure out NTFS

Booting/running exclusively from an infected disk can't be secure. That's
why I thought it's a CD.
 
	Szaka

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

On Thu, 20 May 2004, Szakacsits Szabolcs wrote:
> On Thu, 20 May 2004 ch...@to... wrote:
> 
> > Take a look at the windows recovery console and how it is installed and
> > booted. 
>
> I don't have windows recovery console, my windows was preinstalled and IBM
> provided only a Windows reinstalling CD set.
> 
> However I'm also not too interested what the recovery console does,
> instead I'm interested what the REAL Windows boot process does.

Hmm, apparently this RC is installed on the disk for some people, it's not
a bootable CD I've thought (e.g. Linux recovery cd's, etc).
 
> > It uses the same method as loading a linux or dos bootsector image.
> > ntldr loads c:\cmdcons\bootsect.dat this is a file that spans several
> > sectors. 
> > Make a copy of bootsect.dat and setup boot.ini to point to
> > the new copy It will not work.... 

As I wrote before, these things also depend on _where_ you move it. To a
higher or lower positions on the disk. In certain cases the 1024 cylinder
limit is still valid _during_ the boot process.

Perhaps you're right, it must be on a fixed location, perhaps you moved it
to a higher position, above 1024 cylinder.

Do you have exact details, facts, URL's?

Thanks,
	Szaka 

Re: [Linux-NTFS-Dev] NTFS defrag question

[<ch...@to...>]

As far as real windows booting is concerned I kinda doubt that that
process can be hyjacked to boot a linux kernel probably not even some kin=
d
of custom bootloader. If I remember correctly the windows boot process
uses the ntldr to load other stuff before the windows kernel such as the
registry and hardware detection code (it's been a while since I read over
the details commonly avaliable for windows boot process.) If it were any
less than moderatly difficult to change this process I would think
microsoft would have used that instead of the load bootsector method to
load thier own recovery console.

I guess when I said about avoiding fragmentation I was giving MS too much
credit. I always hear that the need for defragmentation in *NIX
filesystems is rare because they try to avoid fragmentation in the first
place. So I assumed that if MS was building a modern filesystem that they
would use some of these methods.

yes bootsect.dat file is basicly the recovery console kernel as far as I
know. (it is definalty the first code loaded for RC)

The recovery console is a piece of crap. It cannot run any external
commands (although it can run special batch files made of RC internal
commands)
I guess they figured that it would be a security feature. But that seems
to be based on the assumption that no third party would ever figure out
NTFS

I'll give that faq a good read later.

chris

>
> On Thu, 20 May 2004 ch...@to... wrote:
>
>> Take a look at the windows recovery console and how it is installed an=
d
>> booted.
>
> I don't have windows recovery console, my windows was preinstalled and =
IBM
> provided only a Windows reinstalling CD set.
>
> However I'm also not too interested what the recovery console does,
> instead I'm interested what the REAL Windows boot process does.
>
>> It uses the same method as loading a linux or dos bootsector image.
>> ntldr loads c:\cmdcons\bootsect.dat this is a file that spans several
>> sectors.
>
> I don't have any such file on my XP. A quick seach on google suggests i=
t's
> a recovery console related thing only.
>
>> Make a copy of bootsect.dat and setup boot.ini to point to
>> the new copy It will not work....
>
> Sure but this has nothing to do with real booting.
>
>> I did do alot of googleing on this issue. All I found was that microso=
ft
>> defragmenters respect the system and readonly attributes and do not mo=
ve
>> those files. Some third party defrag tools use other methods such as
>> filename lists.
>
> No. Please read carefully again what I wrote previously in my other
> emails.
>
>> I kinda wonder if the automatic defrag "rumor" is just a way of
>> "explaining"  that modern filesystem avoid fragmentation.
>
> They don't avoid.
>
>> Isn't defraging NTFS kinda silly? The registry and dll's are shure to =
be
>> FU**ed long before any significant framentation takes place. So by the
>> time a defrag would be needed a clean reinstall would be long overdue.
>
> Based on what I've seen/met in the last two years (very badly fragmente=
d
> volumes and probably due to these lots of corruptions, lost clusters [t=
he
> worst was 40% of the disk space!]), probably it's not a bad idea to def=
rag
> daily, weekly. But this really depends on the usage ...
>
>> Also I did a defrag once and something on the middle of the disk staye=
d
>> put (probably MFT copy)
>
> Depends on the defragmenter, if the file is in use by Windwos, etc. See
> that defrag FAQ I've referred earlier.
>
> 	Szaka
>
>

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

On Thu, 20 May 2004 ch...@to... wrote:

> Take a look at the windows recovery console and how it is installed and
> booted. 

I don't have windows recovery console, my windows was preinstalled and IBM
provided only a Windows reinstalling CD set.

However I'm also not too interested what the recovery console does,
instead I'm interested what the REAL Windows boot process does.

> It uses the same method as loading a linux or dos bootsector image.
> ntldr loads c:\cmdcons\bootsect.dat this is a file that spans several
> sectors. 

I don't have any such file on my XP. A quick seach on google suggests it's
a recovery console related thing only.

> Make a copy of bootsect.dat and setup boot.ini to point to
> the new copy It will not work.... 

Sure but this has nothing to do with real booting.

> I did do alot of googleing on this issue. All I found was that microsoft
> defragmenters respect the system and readonly attributes and do not move
> those files. Some third party defrag tools use other methods such as
> filename lists.

No. Please read carefully again what I wrote previously in my other
emails.
 
> I kinda wonder if the automatic defrag "rumor" is just a way of
> "explaining"  that modern filesystem avoid fragmentation.

They don't avoid. 
 
> Isn't defraging NTFS kinda silly? The registry and dll's are shure to be
> FU**ed long before any significant framentation takes place. So by the
> time a defrag would be needed a clean reinstall would be long overdue.

Based on what I've seen/met in the last two years (very badly fragmented
volumes and probably due to these lots of corruptions, lost clusters [the
worst was 40% of the disk space!]), probably it's not a bad idea to defrag
daily, weekly. But this really depends on the usage ...

> Also I did a defrag once and something on the middle of the disk stayed
> put (probably MFT copy)

Depends on the defragmenter, if the file is in use by Windwos, etc. See
that defrag FAQ I've referred earlier. 

	Szaka

Re: [Linux-NTFS-Dev] NTFS defrag question

[<ch...@to...>]

Take a look at the windows recovery console and how it is installed and
booted. It uses the same method as loading a linux or dos bootsector
image.
ntldr loads c:\cmdcons\bootsect.dat this is a file that spans several
sectors. Make a copy of bootsect.dat and setup boot.ini to point to the
new copy It will not work.... only the first sector's worth is loaded int=
o
memory. Microsoft stores a map in this sector when installing and this ma=
p
is used to load the remainder.

I did do alot of googleing on this issue. All I found was that microsoft
defragmenters respect the system and readonly attributes and do not move
those files. Some third party defrag tools use other methods such as
filename lists.

I kinda wonder if the automatic defrag "rumor" is just a way of
"explaining"  that modern filesystem avoid fragmentation.

Isn't defraging NTFS kinda silly? The registry and dll's are shure to be
FU**ed long before any significant framentation takes place. So by the
time a defrag would be needed a clean reinstall would be long overdue.

Also I did a defrag once and something on the middle of the disk stayed
put (probably MFT copy)

chris

>
> On Thu, 20 May 2004, Digital Infra, Inc. wrote:
>
>> Thanks. but I want to know that how to fix the location of some files,
>> even you do defrag for whole HDD. for example, in DOS, Defrag of
>> MS-DOS 6 does not change the location of MSDOS.SYS or IO.SYS.  I think
>> it does not change any file which has System attribute. I think defrag
>> util of Windows2000/XP can change the location of these kind of files.
>> How to avoid this?
>
> FAT is a joke, NTFS is modern ;) Sorry but I really think itt doesn't m=
ake
> sense what you want, moreover if it were possible then my and many othe=
r
> ntfs resizer would be broken. But nobody reported any such problems in =
the
> almost last two years.
>
>> Urgh! even compressing automatically??! if NT compresses virtual
>> partition, there is no hope to run Linux from NTFS.
>
> Linux runs fine from NTFS :-) You want to make it run on/with Windows.
> I don't think it would be impossible. There are many softwares that can
> run several OS'es at the same time from the same filesystem.
>
>> is there any measure to avoid this?
>
> AFAIR maybe it can be turned off, you should google for it.
>
> But that's the wrong way solving this problem.
>
> 	Szaka
>
>

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

On Thu, 20 May 2004, Tobias Svensson wrote:
> just a small note according the stage2 file and ntfs support..  Samuel 
> Leo fixed a bug in that old stage2 so this stage2 file is better to use
> http://hem.passagen.se/svto/upload/

Cool!

Are these improvements and fixes contributed back to GRUB (grubinstall) 
or these are forks? What's the status?

	Szaka

Re: [Linux-NTFS-Dev] NTFS defrag question

[Tobias S. <tob...@ho...>]

Anton Altaparmakov wrote:

>On Wed, 2004-05-19 at 10:30, Digital Infra, Inc. wrote:
>  
>
>>Hello NTFS developers.
>>
>>we are working on "loop back install" of Linux.
>>it means, you install Linux on NTFS partition as c:/linux/vmlinux or such
>>and boot it from NTLDR.
>>after booting, you can use c:/linux/diskimage.ext2 , which is very big file
>>on NTFS as an ext2 virtual partition.
>>In other words, you dont have to make a new partiton for dual booting.
>>
>>we succeeded to boot it already, but one problem.
>>we mapped sector geometories directly to the boot loader (hacked grub or lilo),
>>then if there is defrag, of course it does not boot.
>>
>>I have two questions:
>>
>>   1. Is there any way to avoid defrag for certain file?
>>      for example, is that impossible to give "no defrag" attribute
>>      for some system files?
>>    
>>
>
>Not as far as I know.
>
>  
>
>>   2. Some people say that NTFS does defrag automatically and implicitly.
>>      Even you dont do any defrag, but NTFS sometimes does defrag automatically.
>>      I have not checked this issue, but is this "rumor" real?
>>      and if it really happens, how to avoid it?
>>      NTFS defrags even the files are set to be System and Hidden attribute?
>>    
>>
>
>You would have to try, I do not know.
>
>But a solution to your problem is perhaps to use grub with NTFS support
>so you do not need to hardcode the sector geometries.  See:
>
>http://www.geocities.com/lode_leroy/grubinstall/
>
>Best regards,
>
>	Anton
>  
>
just a small note according the stage2 file and ntfs support..  Samuel 
Leo fixed a bug in that old stage2 so this stage2 file is better to use
http://hem.passagen.se/svto/upload/

/Tobias

Re: [Linux-NTFS-Dev] NTFS defrag question

[Anton A. <ai...@ca...>]

On Wed, 2004-05-19 at 10:30, Digital Infra, Inc. wrote:
> Hello NTFS developers.
> 
> we are working on "loop back install" of Linux.
> it means, you install Linux on NTFS partition as c:/linux/vmlinux or such
> and boot it from NTLDR.
> after booting, you can use c:/linux/diskimage.ext2 , which is very big file
> on NTFS as an ext2 virtual partition.
> In other words, you dont have to make a new partiton for dual booting.
> 
> we succeeded to boot it already, but one problem.
> we mapped sector geometories directly to the boot loader (hacked grub or lilo),
> then if there is defrag, of course it does not boot.
> 
> I have two questions:
> 
>    1. Is there any way to avoid defrag for certain file?
>       for example, is that impossible to give "no defrag" attribute
>       for some system files?

Not as far as I know.

>    2. Some people say that NTFS does defrag automatically and implicitly.
>       Even you dont do any defrag, but NTFS sometimes does defrag automatically.
>       I have not checked this issue, but is this "rumor" real?
>       and if it really happens, how to avoid it?
>       NTFS defrags even the files are set to be System and Hidden attribute?

You would have to try, I do not know.

But a solution to your problem is perhaps to use grub with NTFS support
so you do not need to hardcode the sector geometries.  See:

http://www.geocities.com/lode_leroy/grubinstall/

Best regards,

	Anton
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer / IRC: #ntfs on irc.freenode.net
WWW: http://linux-ntfs.sf.net/ &
http://www-stu.christs.cam.ac.uk/~aia21/

[Linux-NTFS-Dev] ldmutil code query

[Nasir H. <son...@ga...>]

hi all,

the LDM Documentation specifies that the PRIVHEAD ,512 bytes Private  
Header) is located immedietly after the (dummy) partition table. So  
picking up the first 2 sectors of the physical disk should get the  
partition table and PRIVHEAD. This is an extract from  
/ldm/ldmutil/main.cpp (ldmutil code) :

// First 7 sectors contains partiontable & privhead #1

for(i = 0; i < 7; i++) {
       dev.Read(sect);
       odev.Write(sect);
}

Why are the picking up the first 7 sectors instead of picking up only the  
first 2 ?

thanks,

-- 
while(Strarting Windoze)
{
	Illegal OS Error;
	Reboot;
	Boot GNU/Linux;
} ;-)

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

On Wed, 19 May 2004, Szakacsits Szabolcs wrote:
> On Thu, 20 May 2004, Digital Infra, Inc. wrote:
> 
> > Thanks. but I want to know that how to fix the location of some files,
[...]
> if it were possible then my and many other ntfs resizer would be
> broken. But nobody reported any such problems in the almost last two
> years.

Sorry, this wasn't true: ntfsresize can move any files (except $boot of
course) to lower locations only since November.

However if you insist to the "fix the location" approach, maybe you find
some tips below and probably on other defragmenters, etc support pages to
minimize the risk of moving around some files
http://smallbiz.executive.com/support/diskeeperfaq.asp?RId=228&CId=1&SId=1

I think I also misunderstood you previously. In the loopback install mode
Linux would run alone, not alongside with Windows however started through
NTLDR what has its own restrictions.

	Szaka

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

On Thu, 20 May 2004, Digital Infra, Inc. wrote:

> Thanks. but I want to know that how to fix the location of some files,
> even you do defrag for whole HDD. for example, in DOS, Defrag of
> MS-DOS 6 does not change the location of MSDOS.SYS or IO.SYS.  I think
> it does not change any file which has System attribute. I think defrag
> util of Windows2000/XP can change the location of these kind of files.
> How to avoid this?

FAT is a joke, NTFS is modern ;) Sorry but I really think itt doesn't make
sense what you want, moreover if it were possible then my and many other
ntfs resizer would be broken. But nobody reported any such problems in the
almost last two years.
 
> Urgh! even compressing automatically??! if NT compresses virtual
> partition, there is no hope to run Linux from NTFS. 

Linux runs fine from NTFS :-) You want to make it run on/with Windows. 
I don't think it would be impossible. There are many softwares that can 
run several OS'es at the same time from the same filesystem.

> is there any measure to avoid this?

AFAIR maybe it can be turned off, you should google for it.

But that's the wrong way solving this problem.

	Szaka

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

On Thu, 20 May 2004, Digital Infra, Inc. wrote:
>   MBR -> NTLDR -> 1st Grub -> 2nd Grub on NTFS -> vmlinux on NTFS.
> as you may know, 1st Grub does not understand NTFS, so we directly map
> the sector location of 2nd Grub.

I don't know grub internals but I can't see why you couldn't add support.
All the needed code exists and gpl. Win boot the same way, it has a mini
ntfs driver.

> and this is the reason we really want to avoid defrag.

Technically it would be very inferior allowing what you want. And I bet MS
is well aware of this. 

Think about for example when one must shrink a volume. Where would you
move the fixed blocks outside of the new volume size? Because you MUST
move it somewhere. The only solution would be if you looked for who and
how using them and modify them also. LILO worked like that and it caused
neverending problems. No wonder users happily moved to GRUB.

	Szaka

Re: [Linux-NTFS-Dev] NTFS defrag question

[Digital I. Inc. <ok...@di...>]

Hello Altaparmakov

>
>
>But a solution to your problem is perhaps to use grub with NTFS support
>so you do not need to hardcode the sector geometries.  See:
>
>http://www.geocities.com/lode_leroy/grubinstall/
>

Yes, I actually use this.
but this is 2nd stage of Grub ....
As far as I know, NTLDR loads only 512 bytes MBR type boot code.
(one exeception is windows kernel. NTLDR can load any size of windows kernel.)
1st Grub fits this size. but not 2nd Grub.
then how to load 2nd of Grub?
what we did is like this:
  MBR -> NTLDR -> 1st Grub -> 2nd Grub on NTFS -> vmlinux on NTFS.
as you may know, 1st Grub does not understand NTFS, so we directly map
the sector location of 2nd Grub.
and this is the reason we really want to avoid defrag.


              --- Okajima.

Re: [Linux-NTFS-Dev] NTFS defrag question

[Digital I. Inc. <ok...@di...>]

Hello Szabolcs

>
>> we are working on "loop back install" of Linux. it means, you install
>> Linux on NTFS partition as c:/linux/vmlinux or such and boot it from
>> NTLDR. after booting, you can use c:/linux/diskimage.ext2 , which is
>> very big file on NTFS as an ext2 virtual partition. In other words,
>> you dont have to make a new partiton for dual booting.
>
>In short, you want to do the same as TopologiLinux but running on Windows.
>

Yes, and we are cooperate with Topologi team.
our distribution, called "conoppix", which KNOPPIX+coLinux has three mode -
   - coLinux mode     you can run it as Win32 app with autorun.inf.
                      inserting CD and wait a while.
                      just it makes KDE appeared on Windows.
   - loop back mode   same as topologi one.
                      you can install conoppix on a very big file of NTFS.
                      and no need to create a new partition.
   - CD Boot mode     same as KNOPPIX.

These three mode makes Linux very newbie friendly.

>>    1. Is there any way to avoid defrag for certain file?
>>       for example, is that impossible to give "no defrag" attribute
>>       for some system files?
>
>You can't move $Boot but otherwise anything to anywhere, considering that
>some systems files are recommended to be below 1024 cylinders otherwise
>Win won't boot, in cases.
>

Thanks. but I want to know that how to fix the location of some files,
even you do defrag for whole HDD.
for example, in DOS, Defrag of MS-DOS 6 does not change the location of
MSDOS.SYS or IO.SYS.  I think it does not change any file which has System attribute.
I think defrag util of Windows2000/XP can change the location of these kind of files.
How to avoid this?

>>    2. Some people say that NTFS does defrag automatically and implicitly.
>>       Even you dont do any defrag, but NTFS sometimes does defrag automatically.
>>       I have not checked this issue, but is this "rumor" real?
>
>Me neither but why not? It make sense, Apple does this and basically
>everybody who cares about performance. MS improved a lot on NTFS lately.
>
>>       NTFS defrags even the files are set to be System and Hidden attribute?
>
>Sure, sometimes it even compresses them for better performance or if disk
>is close to full, etc.
>
>However I'm not a Windows guru so please don't take anything granted.
>

Urgh! even compressing automatically??!
if NT compresses virtual partition, there is no hope to run Linux from NTFS.
is there any measure to avoid this?

                          --- Okajima.

Re: [Linux-NTFS-Dev] NTFS defrag question

[Szakacsits S. <sz...@si...>]

Hello!

On Wed, 19 May 2004, Digital Infra, Inc. wrote:

> we are working on "loop back install" of Linux. it means, you install
> Linux on NTFS partition as c:/linux/vmlinux or such and boot it from
> NTLDR. after booting, you can use c:/linux/diskimage.ext2 , which is
> very big file on NTFS as an ext2 virtual partition. In other words,
> you dont have to make a new partiton for dual booting.

In short, you want to do the same as TopologiLinux but running on Windows.

>    1. Is there any way to avoid defrag for certain file?
>       for example, is that impossible to give "no defrag" attribute
>       for some system files?

You can't move $Boot but otherwise anything to anywhere, considering that
some systems files are recommended to be below 1024 cylinders otherwise
Win won't boot, in cases.

>    2. Some people say that NTFS does defrag automatically and implicitly.
>       Even you dont do any defrag, but NTFS sometimes does defrag automatically.
>       I have not checked this issue, but is this "rumor" real?

Me neither but why not? It make sense, Apple does this and basically
everybody who cares about performance. MS improved a lot on NTFS lately.

>       NTFS defrags even the files are set to be System and Hidden attribute?

Sure, sometimes it even compresses them for better performance or if disk
is close to full, etc.

However I'm not a Windows guru so please don't take anything granted.

	Szaka

[Linux-NTFS-Dev] NTFS defrag question

[Digital I. Inc. <ok...@di...>]

Hello NTFS developers.

we are working on "loop back install" of Linux.
it means, you install Linux on NTFS partition as c:/linux/vmlinux or such
and boot it from NTLDR.
after booting, you can use c:/linux/diskimage.ext2 , which is very big file
on NTFS as an ext2 virtual partition.
In other words, you dont have to make a new partiton for dual booting.

we succeeded to boot it already, but one problem.
we mapped sector geometories directly to the boot loader (hacked grub or lilo),
then if there is defrag, of course it does not boot.

I have two questions:

   1. Is there any way to avoid defrag for certain file?
      for example, is that impossible to give "no defrag" attribute
      for some system files?
   2. Some people say that NTFS does defrag automatically and implicitly.
      Even you dont do any defrag, but NTFS sometimes does defrag automatically.
      I have not checked this issue, but is this "rumor" real?
      and if it really happens, how to avoid it?
      NTFS defrags even the files are set to be System and Hidden attribute?


              Okajima, Jun. Digital Infra, Inc. Tokyo, Japan.
              coLinux development team   http://www.colinux.org/

[Linux-NTFS-Dev] Re: Time management

[Anton A. <ai...@ca...>]

Hi all,

On Wed, 2004-05-19 at 01:17, Richard Russon wrote:
> Gentlemen,
> 
> Recent emails have suggested (not for the first time) that we:
> 
> > stop offering help to people
> > do not need to support NTFS users
> 
> This is the major time killer.  We all know it, and yet we all feel
> obliged to help everyone.  We're just too nice.
> 
> > do we care about people who need help?
> 
> Of course we do and we'll feel bad for ignoring them.
> But by ignoring them we're committing ourselves to something bigger.
> Something that will help so many more.
> 
> > no one is paying us for it
> 
> We're volunteers; we can do what we like; only we can finish write
> support.  We just need to focus.
> 
> > we could charge for answering support questions
> 
> Not a bad suggestion, but obviously (for free), we still need to
> 
> > answer bug reports
> 
> The website has helped countless thousands of people, but it takes a lot
> of effort to keep it up-to-date.  So, for now, there will be no new
> additions - no wiki, no forums.
> 
> Unless any of you have any strong objections, I will:
> 
>   make the forums read-only
>   remove all email addresses
>   add a banner
>     explaining our absence
>     asking for volunteers
>     mailing list is dev + bugs only

Do it!  Perhaps leave the user list existing.  Doesn't mean we have to
answer it but at least it is easy to do so if we wanted...  (-;

Best regards,

	Anton
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer / IRC: #ntfs on irc.freenode.net
WWW: http://linux-ntfs.sf.net/ &
http://www-stu.christs.cam.ac.uk/~aia21/