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

[Linux-NTFS-Dev] Re: [PATCH] 436 printf length modifier fixes

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

On Wed, 14 Jan 2004, Szakacsits Szabolcs wrote:
> ntfsprogs 1.8.2 added support for FreeBSD. I tested it. The good news, all
> regression tests passed (mostly just resizing related). The bad one is,
> there were a lot of scary and bogus outputs in regards to printed numbers
> (sizes, offsets, run lengths, etc) like relocating data to the bootsector,
> etc, those kind that should never happen.
>
> The reason is, the printf 'L' length modifier is incorrectly used in 436
> places all over the source instead of using 'll'. This is a very
> embarrassing and bad bug. Depending on the libc used and its current
> printf implementation, the numbers can be random thus it can be impossible
> to debug things and if an external tool relies on them then very
> unpleasant things could happen.
>
> Interestingly this never showed up on Linux for me, actually the compiler
> should have even warned :(

On Linux L is accepted to be the same as ll.  At least it is in the
kernel.  This is why I was using L and not ll...  That is what I read one
should use.  Obviously the person who wrote that wasn't aware of the
difference.  I just did a "man 3 printf" and you are quite right. L is for
long double while ll is for long long (unsigned) int which is what we
should be using!

> Bzip2 compressed patch is attached.

Thanks, will apply either later today or tomorrow.  At the moment my
workstation is in a state of limbo as I am migrating from RedHat 9.0 to
SuSE 9.0...

I will then release 1.8.3 or whatever we are at now immediately.

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] [PATCH] 436 printf length modifier fixes

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

Hi,

ntfsprogs 1.8.2 added support for FreeBSD. I tested it. The good news, all
regression tests passed (mostly just resizing related). The bad one is,
there were a lot of scary and bogus outputs in regards to printed numbers
(sizes, offsets, run lengths, etc) like relocating data to the bootsector,
etc, those kind that should never happen.

The reason is, the printf 'L' length modifier is incorrectly used in 436
places all over the source instead of using 'll'. This is a very
embarrassing and bad bug. Depending on the libc used and its current
printf implementation, the numbers can be random thus it can be impossible
to debug things and if an external tool relies on them then very
unpleasant things could happen.

Interestingly this never showed up on Linux for me, actually the compiler
should have even warned :(

Bzip2 compressed patch is attached.

	Szaka

[Linux-NTFS-Dev] Re: [PATCH] broken ntfs_mapping_pairs_build

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

Patch applied.  Thanks.

Sorry for the long delay but wedding and christmas got in the way of
ntfs development...

Best regards,

Anton

On Sun, 2003-12-14 at 15:16, Szakacsits Szabolcs wrote:
> Hi,
> 
> ntfs_mapping_pairs_build in ntfsprogs is broken for sparse and compressed
> files. If one used it for multiply runs with holes in between them then the
> result is minimum lost data.
> 
> Fix attached for 1.8.0 but should apply to 1.9-devel as well. It works fine
> now for all non-resident attributes [decode(mp) = decode(encode(decode(mp)))] 
> on all tested images.
> 
> 	Szaka
-- 
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/

Re: [Linux-NTFS-Dev] Re: libntfs memory leaks?

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

On Tue, 23 Dec 2003, Szakacsits Szabolcs wrote:
> I'm also close to finish ntfsresize with unlimited relocation support
> (Miguel Lastra and Dewey M. Sasser helped a lot to achieve this) and I
> need a very important fix to be applied to libntfs but I didn't get
> response so far for my patches whereever I send them (since the cvs->bk
> change I can't commit).

Do you mean your patch about building mapping pairs?  If yes, then I have
that but it arrived to late for me to do anything with.  I have been out
of action due to wedding and Christmas since 11th December and I am afraid
I don't have  time to look at it until the New Year.  Sorry but there are
more important things in life...  (-:

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/

Re: [Linux-NTFS-Dev] Re: libntfs memory leaks?

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

On Tue, 23 Dec 2003, Szakacsits Szabolcs wrote:
> On Tue, 23 Dec 2003, Jan Kratochvil wrote:
> > I have suspection libntfs is leaking memory.
>
> It's possible. I've found there were ususally 1-3 bugs in the libntfs
> functions I used. This is expectable because, as I was explained, Anton

Yes, it is quite possible.  I don't claim my code doesn't have bugs.  But
at least I write code!  Without me there wouldn't be stable ntfs read
support on Linux today (except for Captive perhaps, that is).  To rephrase
better:  There would be no opensource and freeware linux ntfs support.  I
think it a lot better to write a lot of code than to write only half the
code but to know it is bug free...  And anyway I have often said that
libntfs is a staging area for my kernel development so I don't care about
it that much that I would test my code to death for it...

> If you don't get answers (linux-ntfs-dev became moderated due to spams and
> sometimes it takes 3+ days to approve a message) I'm afraid you have to
> look at it yourself.

There are no messages except spam that I can see waiting for approval...
And I just got married so I am not really checking email (haven't done for
about a week and won't again until next week probably).

> need a very important fix to be applied to libntfs but I didn't get
> response so far for my patches whereever I send them (since the cvs->bk
> change I can't commit).

I haven't received any email from you.  I don't think I have touched my
spam folder recently so when I get back I will search it for emails from
you.  Why don't you send your patches to linux-ntfs-dev?  Those should
always manage to get to me and if not I should be able to see them in the
archives...  And there is the patch submission thing in the tracker on
sourceforge if you prefer that.  I can assure you I have _never_ ignored a
patch that I have seen!

> Merry Christmas :)

Merry Christmas and a Happy New Year to you all!

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] Re: libntfs memory leaks?

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

On Tue, 23 Dec 2003, Jan Kratochvil wrote:

> I have suspection libntfs is leaking memory. 

It's possible. I've found there were ususally 1-3 bugs in the libntfs
functions I used. This is expectable because, as I was explained, Anton
doesn't verify/test the code he writes. Please see more here for example,

   http://marc.theaimsgroup.com/?l=linux-ntfs-dev&m=104626284218891&w=2

I revise functions/patches when I use them moreover use electric fence,
valgrind and automatic testing to catch new errors.

> I did not yet investigate it much 

Sorry but I don't have time either, in one hour I'm more or less away from
computers for a week.

If you don't get answers (linux-ntfs-dev became moderated due to spams and
sometimes it takes 3+ days to approve a message) I'm afraid you have to
look at it yourself. 

I'm also close to finish ntfsresize with unlimited relocation support
(Miguel Lastra and Dewey M. Sasser helped a lot to achieve this) and I
need a very important fix to be applied to libntfs but I didn't get
response so far for my patches whereever I send them (since the cvs->bk
change I can't commit).

The other solution you might have is by looking for the driver this way:

 - read /boot.ini
 - get the system directory from boot.ini (e.g. /winnt)
 - get the drivers from system_directory/system32/drivers (I don't
   know if other places are also possible)

Merry Christmas :)

	Szaka

[Linux-NTFS-Dev] libntfs memory leaks?

[Jan K. <rcp...@ja...>]

Hi,

I have suspection libntfs is leaking memory. During NTFS disks scanning by
captive-install-acquire of Captive NTFS through GnomeVFS and ntfsprogs-gnomevfs
its leaking a lot, short scanning produces http://dmalloc.com output:

	top 10 allocations:
	 total-size  count in-use-size  count  source
	   75055104  18324           0      0  runlist.c:759
	   74174464  18109    74174464  18109  dir.c:238
	   29713408  29017      602112    588  mft.c:214
	    3069184  95912           0      0  attrib.c:1907
	    1468928  22952       52160    815  attrib.c:335
	    1160720  29018       23520    588  inode.c:44

Out of
	http://www.jankratochvil.net/project/captive/dist/captive-1.1.4.tar.gz

it can be produced by
	(rm -f /tmp/dlog;eval `dmalloc -b runtime -l /tmp/dlog`;./captive-install-acquire --text -v --scan-path=file:///dev/hdc1#libntfs: )

or better
	(rm -f /tmp/dlog;eval `dmalloc -b runtime -l /tmp/dlog`;./captive-install-acquire --text -v )
and specify custom scanning path 'file:///dev/hdc1#libntfs:' by hand to be able
to abort it soon by ENTER on the keyboard ('--scan-path' is abortable just by
CTRL-C which prevents output of the dmalloc '/tmp/dlog' report).


I did not yet investigate it much - are you aware of this leakage?
I do not ntfs_umount() it but I hope I correctly ntfs_attr_close() and
ntfs_inode_close() everything.



Regards,
Lace

[Linux-NTFS-Dev] [PATCH] broken ntfs_mapping_pairs_build

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

Hi,

ntfs_mapping_pairs_build in ntfsprogs is broken for sparse and compressed
files. If one used it for multiply runs with holes in between them then the
result is minimum lost data.

Fix attached for 1.8.0 but should apply to 1.9-devel as well. It works fine
now for all non-resident attributes [decode(mp) = decode(encode(decode(mp)))] 
on all tested images.

	Szaka

Re: [Linux-NTFS-Dev] resizing ntfs

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

On Mon, 8 Dec 2003, Raphael Clifford wrote:

> Thank you very much for all the top tips. I am partiticularly excited 
> about the new features that are in the BETA versions of ntfsresize.

In the weekend I made some progress to handle your case as well but the
needed additional feature not ready yet to be included even in the BETA.
Lots of things to do but not much time for them ...

> Unfortunately, I couldn't bring myself to try any of the more exciting
> options you described as they all seemed a little destructive (as I
> don't have a full backup of Windows 2000).  

OK, no problem :) I _think_ the Windows Recovery Console trick should have
succeeded. 

But the Captive NTFS way could have resulted loss data. I've never used
this new way of accessing NTFS (I'm not even sure it's legal) but things
looked promising ... until this weekend when several people using Windows
2000 and Captive NTFS reported data loss. Jan released a fix, we will see
how it works. If Captive NTFS were reliable then it could be useful for
additionally testing Linux-NTFS code (no need to reboot, start/buy VMWare,
etc).

Cheers,
	Szaka

Re: [Linux-NTFS-Dev] resizing ntfs

[Raphael C. <ra...@cl...>]

Hi,

Thank you very much for all the top tips. I am partiticularly excited 
about the new features that are in the BETA versions of ntfsresize.  
Unfortunately, I couldn't bring myself to try any of the more exciting 
options you described as they all seemed a  little destructive (as I 
don't have a full backup of Windows 2000).  Partition magic 8.x saved 
the day.  It really is magic... it removed 1GB from my ntfs partition 
and simply stuck it on one of my ext3 partitions without batting an eye 
lid :)

Cheers and thanks again,
Raphael

Szakacsits Szabolcs wrote:

>>> find -inum 2694
>>>./WINNT/system32/config/software.LOG
>>>
>>>a) Can I simply delete/copy+delete that file. It looks like it might be 
>>>important :)
>>>      
>>>
>>A possible strategy,
>>    
>>
>
>A less exciting but simpler one,
> 
>  
>
>>	1) backup everything, you can do an exact _copy_ (sector-to-sector) 
>>           of your NTFS by either partimage or ntfsclone. Both is on the
>>           SystemResceCD. 
>>    
>>
>
>	2) Start the Windows Recovery Console and 
>
>	        copy software.log software.log.orig
>        	erase software.log
>	        rename software.log.orig software.log
>
>  
>
>>	4) check if Windows still works, if not, recover from backup, we
>>           will try something else.
>>    
>>
>
>Cheers,
>
>	Szaka
>
>
>  
>

Re: [Linux-NTFS-Dev] resizing ntfs

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

> >  find -inum 2694
> > ./WINNT/system32/config/software.LOG
> > 
> > a) Can I simply delete/copy+delete that file. It looks like it might be 
> > important :)
> 
> A possible strategy,

A less exciting but simpler one,
 
> 	1) backup everything, you can do an exact _copy_ (sector-to-sector) 
>            of your NTFS by either partimage or ntfsclone. Both is on the
>            SystemResceCD. 

	2) Start the Windows Recovery Console and 

	        copy software.log software.log.orig
        	erase software.log
	        rename software.log.orig software.log

> 	4) check if Windows still works, if not, recover from backup, we
>            will try something else.

Cheers,

	Szaka

Re: [Linux-NTFS-Dev] ntfsundelete - output ?!

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

Hi Pavel,

> how is working the '-o' or '--output' ?

Er...

> -rw------- 1 root    512 2003-10-10 09:35 ~DFC5A.tmp  <----- !!!!!!!!

Damn, you're right.  I'm not sure how I overlooked that.  It works for
the mft dump, but nobody is interested in that.

> tested on gentoo ntfsprogs 1.7.1 and 1.8.0

Thanks.  I've just fixed the source.  I'll upload a copy to the
snapshots dir as soon as I get some free time.

Thanks for the bug report.

Cheers,
  FlatCap (Rich)
  nt...@fl...

WWW: http://linux-ntfs.sourceforge.net
IRC: #ntfs on irc.freenode.net

Re: [Linux-NTFS-Dev] resizing ntfs

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

On Wed, 3 Dec 2003, Raphael Clifford wrote:

> The answer is
> 
>  find -inum 2694
> ./WINNT/system32/config/software.LOG
> 
> a) Can I simply delete/copy+delete that file. It looks like it might be 
> important :)

Yes, a search on google says it's part of the registry so you won't be
able to delete it even if you tried.

A possible strategy,

	1) backup everything, you can do an exact _copy_ (sector-to-sector) 
           of your NTFS by either partimage or ntfsclone. Both is on the
           SystemResceCD. 

	2) try to install Captive NTFS. It's a free read-write NTFS
           driver that copies your Windows ntfs.sys to the Linux 
	   then uses it to access NTFS.:

 	     http://www.jankratochvil.net/project/captive/

	   Check the law and licensing conditions of your Windows in
           your country if you can use it:
	
	     http://www.jankratochvil.net/project/captive/doc/Architecture.html.pl#law

	3) now you should be able to copy + delete + rename the file.

	4) check if Windows still works, if not, recover from backup, we
           will try something else.

Well, looks like a quite exciting project ;) If you're not willing to do
it, I'll suggest something else. Thanks,

	Szaka

Re: [Linux-NTFS-Dev] resizing ntfs

[Raphael C. <ra...@cl...>]

Hi,

Thanks very much for the great reply.

>This is a very fragmented file. Handling this case is not yet supported
>unfortunately. But maybe you can overcome this. You can check what's this
>file if you mount your NTFS partition:
>
>	mkdir windows
>	mount -t ntfs <device> windows
>	cd windows 
>	find -inum 2694
>  
>

The answer is

 find -inum 2694
./WINNT/system32/config/software.LOG

>If this file is pagefile.sys (great chance) then you could temporarily
>turn off Windows' Virtual Memory under 
>
>	Start->Settings->Control Panel->System->Advanced->Performance
>
>and try again the ntfsresize process.
>
>If the file is something else then on Windwos you may either just delete
>it if you don't need (but be careful you don't delete something useful) or
>
>	copy file file_new
>	erase file
>	rename file_new file
>
>Using just 'rename' is not enough. 
>
>  
>
a) Can I simply delete/copy+delete that file. It looks like it might be 
important :)
b) Will it just be instantly recreated when I do that (before I can exit 
windows 2000)?

Cheers,
Raphael

Re: [Linux-NTFS-Dev] resizing ntfs

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

Hi,

First of all, sorry for the late reply for your first email. The list
became moderated due to the huge amount of spam lately and your email was
approved only 3 days later, I didn't get it earlier. [Anton, as I noticed
SourceForge started to use spam filtering. Is there still a lot of spam?
If not, it could be switched back to unmoderated, also less job for you]

> As I am running it from the sysrescuecd I have copied out the important 
> bits by hand and abbreviated some bits.

OK. Do you have floppy or is sysrescuecd capable to automaticallly setup
the network for you? If yes, you could redirect ntfsresize output to a
file, e.g by

    ntfsresize -i device > ntfsresize.out 
or 
    ntfsresize -i device | tee ntfsresize.out  # you see also the output

then save ntfsresize.out to a floppy or to somewhere else on the network.

> volume/device size 4294MB
> Space in use 2385MB (55.5%)
>
>             Last used at            By Inode
> $MFT                  33                   0

Ok, the problem is not here.

> $MFTMirr            4220                   1

The development (BETA) version of ntfsresize can cope with this.

> Compressed          2898                9577  

This one also but it's disabled even in the BETA since it's untested. Thus
in the best case you could resize at 2899 MB with the BETA version (unless
I enable support for this as well) if we find solution for the below
problem.

> Multi-Record        4287                2694

This is a very fragmented file. Handling this case is not yet supported
unfortunately. But maybe you can overcome this. You can check what's this
file if you mount your NTFS partition:

	mkdir windows
	mount -t ntfs <device> windows
	cd windows 
	find -inum 2694

If this file is pagefile.sys (great chance) then you could temporarily
turn off Windows' Virtual Memory under 

	Start->Settings->Control Panel->System->Advanced->Performance

and try again the ntfsresize process.

If the file is something else then on Windwos you may either just delete
it if you don't need (but be careful you don't delete something useful) or

	copy file file_new
	erase file
	rename file_new file

Using just 'rename' is not enough. 

> Ordinary            4288               14944

The ntfsresize BETA can cope with this also.
 
> P.S. I have tried defragging over and over again in Win 2000 as someone 
> suggested. This is the result after the 5th defrag.

Windows defragmenter can't defragment files in use (e.g. system files,
NTFS internal files). Some commercial defragmenter can achieve this at
boot time. ntfsresize BETA can also overcome most of these problems
(supporting all cases still needs some time). If it make sense and you're
willing to try it out, please let me know.

> P.P.S. In Win 2000 it is clear from the defrag picture that some system 
> files are kept at the end of the ntfs partition and they are causing the 
> problem.

Usually it's the Windows pagefile.sys and that can be turned off. Cheers,

	Szaka

Re: [Linux-NTFS-Dev] resizing ntfs

[Raphael C. <ra...@cl...>]

Szakacsits Szabolcs wrote:

>On Sun, 30 Nov 2003, Raphael Clifford wrote:
>
>  
>
>>I have a dual boot system with windows 2000/linux and wanted to grab 
>>some of the spare space on the ntfs partition for linux.  I made a 
>>systemrescuecd (0.2.8) but it couldn't resize the ntfs partition as it 
>>    
>>
>
>systemrescuecd has ntfsresize 1.8.0. Could you do send me the output of
>ntfsresize -i device? Then I can tell what you might do.
>
>Cheers,
>	Szak
>
Hi,

As I am running it from the sysrescuecd I have copied out the important 
bits by hand and abbreviated some bits.

NTFS Volume version 3.0

volume/device size 4294MB
Space in use 2385MB (55.5%)

                            Last used at            By Inode
$MFT                  33                              0
$MFTMirr           4220                        1
Compressed       2898                        9577  
Multi-Record      4287                        2694
Ordinary             4288                        14944

Could resize to 4288MB (freeing 6MB)


Cheers,
Raphael
P.S. I have tried defragging over and over again in Win 2000 as someone 
suggested. This is the result after the 5th defrag. 
P.P.S. In Win 2000 it is clear from the defrag picture that some system 
files are kept at the end of the ntfs partition and they are causing the 
problem.

Re: [Linux-NTFS-Dev] resizing ntfs

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

On Sun, 30 Nov 2003, Raphael Clifford wrote:

> I have a dual boot system with windows 2000/linux and wanted to grab 
> some of the spare space on the ntfs partition for linux.  I made a 
> systemrescuecd (0.2.8) but it couldn't resize the ntfs partition as it 

systemrescuecd has ntfsresize 1.8.0. Could you do send me the output of
ntfsresize -i device? Then I can tell what you might do.

Cheers,
	Szaka

[Linux-NTFS-Dev] NTFS driver 2.1.5a released for Linux kernel 2.4.23

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

The latest NTFS driver (2.1.5a) is now available for Linux kernel
2.4.23.

This updated release for Linux kernel 2.4.23 also includes a fix for a
problem encountered by at least two Windows XP users when accessing
certain fragmented files.

As always you can download a patch for the kernel from our website:

	http://linux-ntfs.sf.net/downloads.html

Or you can pull the changes from our bitkeeper repository:

	http://linux-ntfs.bkbits.net/ntfs-2.4

Note that the latest NTFS version (2.1.5) is also present in the current 2.6.0-test kernel and has been since 2.6.0-test9.

Happy NTFSing!  (-:

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] ntfsundelete - output ?!

[Pavel R. <Pav...@tr...>]

Hi,

how is working the '-o' or '--output' ?
the man page say, that it's the name of output file for undelete (or save=
)

but:

(root)~# ntfsundelete -u 49026 --output out /dev/hdb1     =20
Inode    Flags  %age  Date            Size  Filename
---------------------------------------------------------------
49026    FR..     0%  2003-10-10       512  ~DFC5A.tmp

Undeleted '~DFC5A.tmp' successfully.

(root)~# ls -l
celkem 77
drwxr-xr-x 2 root     48 2003-12-01 17:56 a
-rw------- 1 root    512 2003-10-10 09:35 ~DFC5A.tmp  <----- !!!!!!!!
drwxr-xr-x 2 root    176 2003-12-02 08:15 files
-rw-r--r-- 1 root  47109 2003-12-01 13:08 index.txt
-rw-r--r-- 1 root  18274 2003-12-01 17:46 inodes.txt
-rwxr-xr-x 1 root    304 2003-12-02 08:17 undelete.sh


tested on gentoo ntfsprogs 1.7.1 and 1.8.0
kernel 2.6.0-test11

Pavel
--=20
Pavel Riha
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
V=B9echny nepomiluje=B9, ale m=E1=B9 se sna=BEit.
		-- Napoleon

Re: [Linux-NTFS-Dev] unable to install - reboot and select kernel

[James B. <jba...@ms...>]

Thanks that did it.  I had the wrong RPM.  It works GREAT NOW!!!!  I am =
about to try and have lilo boot the NTFS drive next just for kicks and =
grins..

James
  ----- Original Message -----=20
  From: Richard Russon<mailto:nt...@fl...>=20
  To: James Baker<mailto:jba...@ms...>=20
  Cc: sf<mailto:Lin...@li...>=20
  Sent: Sunday, November 30, 2003 2:05 AM
  Subject: Re: [Linux-NTFS-Dev] unable to install - reboot and select =
kernel


  Hi James,

  > [root@jbakerlinux ~]# rpm -ihv kernel-ntfs-2.4.20-20.8.i386.rpm

  I tried to make the rpms foolproof, looks like I failed :-)
  You've installed the wrong rpm.  Uninstall it

    rpm -e kernel-ntfs

  and read the instructions.  They tell you which rpm you need

    =
http://linux-ntfs.sourceforge.net/rpm/instructions.html<http://linux-ntfs=
.sourceforge.net/rpm/instructions.html>

  Cheers,
    FlatCap (Rich)
    nt...@fl...<mailto:nt...@fl...>

  WWW: =
http://linux-ntfs.sourceforge.net<http://linux-ntfs.sourceforge.net/>
  IRC: #ntfs on irc.freenode.net

[Linux-NTFS-Dev] resizing ntfs

[Raphael C. <ra...@cl...>]

I have a dual boot system with windows 2000/linux and wanted to grab 
some of the spare space on the ntfs partition for linux.  I made a 
systemrescuecd (0.2.8) but it couldn't resize the ntfs partition as it 
was fragmented. I rebooted into 2000 and defragmented it (noticing that 
it gave a list of files it hadn't actually defragmented at the end). 
After rebooting using the systemrescuecd I got the same error about the 
ntfs filesystem being fragmented. I then defragmented it again, it 
appeared to actually do some more but still had a few files it said it 
had not defragmented. After rebooting, systemrescuecd  gave me a very 
similar (but not identically phrased) error to before and still wouldn't 
resize the partition.

Essentially I have two questions and a request

a) What am I to do?  Is there a different more aggressive way of 
defragmenting in 2000 that I have missed?
b) Could there be some docs on how to defragment so that ntfsresize will 
work?  I very rarely use Windows and am quite lost.
c) If ntfsresize were able to do this for you it would be fantasic!

Kind regards,
Raphael

Re: [Linux-NTFS-Dev] unable to install - reboot and select kernel

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

Hi James,

> [root@jbakerlinux ~]# rpm -ihv kernel-ntfs-2.4.20-20.8.i386.rpm

I tried to make the rpms foolproof, looks like I failed :-)
You've installed the wrong rpm.  Uninstall it

  rpm -e kernel-ntfs

and read the instructions.  They tell you which rpm you need

  http://linux-ntfs.sourceforge.net/rpm/instructions.html

Cheers,
  FlatCap (Rich)
  nt...@fl...

WWW: http://linux-ntfs.sourceforge.net
IRC: #ntfs on irc.freenode.net

[Linux-NTFS-Dev] unable to install - reboot and select kernel

[James B. <jba...@ms...>]

I try and install the rpm for NTFS support and I get the following error.  I 
am not sure how I do what is expected.  Can anyone help.  This is 
kernel-ntfs-2.4.20-20.8.i386.rpm my kernel version as listed at the boot up 
screen is 2.4.20-20.8
NTFS version is XP

Thanks
J

[root@jbakerlinux ~]# rpm -ihv kernel-ntfs-2.4.20-20.8.i386.rpm
Preparing...                ########################################### 
[100%]
_______________________________________________________________________________

ERROR: This RPM cannot be installed yet.

This RPM requires kernel-2.4.20-20.8.
You already have it installed, but it is not running.
Please reboot and select that kernel, then retry the install.
_______________________________________________________________________________

error: %pre(kernel-ntfs-2.4.20-20.8) scriptlet failed, exit status 1
error:   install: %pre scriptlet failed (2), skipping 
kernel-ntfs-2.4.20-20.8
[root@jbakerlinux ~]#

_________________________________________________________________
Share holiday photos without swamping your Inbox.  Get MSN Extra Storage 
now!  http://join.msn.com/?PAGE=features/es

[Linux-NTFS-Dev] Some open questions in your ntfs docs are documented by MS!

[<JB...@Da...>]

Some open questions in your ntfs docs are documented by MS!
(or at least obvious from knowing their style)

Specifically:

Why do some Metadata files on NTFS 3.0+ still have Security Descriptors?
On NTFS 3.0+, $Volume, $AttrDef, dot and $Boot have Security Descriptors. Is
this to save time at boot up? Perhaps to reduce the number of files it has
to parse? Or is this the same as the previous question?

My Guess: To make sure older Microsoft NTFS drivers do not allow insecure
access to these critical files.
Remember: No security descriptor = all access for everyone = UNIX mode 777


Attribute Header
When is "Initialized" not the same size as "Real"?

This is documented in connection with a new XP IOCTL to change it directly.
The short summary:

The bytes from 0 to real -1 are the current file data

The bytes from Initialized to allocated -1 have not been initialized
securely and
may contain bytes that used to be in entirely different files that were not
readable
by those who can read the current file.  C2 level security (and basic
security thinking)
require that such bytes not be revealed to applications that simply use
documented
file I/O calls (such as seek() past end then write bytes there) do not
suddenly see those
bytes as part of an area in their file which they have allowed to contain
old garbage in
that way.  So if an application tries to set Real > Initialized without
writing to every
byte in between, NTFS.SYS makes sure the old contents of those clusters is
overwritten with zeroes or other nonsecret garbage 

The bytes from real to Initialized - 1 have already been cleared by NTFS.SYS
or
contains bytes that used to be part of this file itself.  If the use tries
to increase
the value of Real, NTFS.SYS makes no attempt to clear the contents of these
bytes.



Security Descriptors
How are ACEs inherited?

The full format, layout and inheritance of Security Descriptors, including
ACEs,
Sids etc. is a generic NT mechanism not at all specific to NTFS.  NT even
exposes
user mode functions to dream up your own kind of data, including your own
way
of storing SDs within that data, and then call the NT functions that
manipulate
ACLs, ACEs etc. 

Officially, the entire SD format is opaque and you should access it only
through
the (complete) set of manipulation functions available in ADVAPI32.DLL.  In
practice,
because SDs can be legitimately stored on disk, in databases, in your own
dreamed
up kind of data etc., the format is carved in stone.

The actual data format (more complete than your description) is described by
comments
in winnt.h in the Microsoft Platform SDK headers.

The inheritance rules are in the SDK.  There are two VERSIONS of the
inheritance rules:

The classic, pre-Windows 2000 rules are used when NTFS is accessed from NT
4.0 or older,
no matter what the NTFS version.

The new, Windows 2000 and later rules are used when NTFS is accessed from
Windows 2000,
no matter what the NTFS version.  The New rules simply add the various
"autoinherit" bits in
the SD, ACL and ACE headers, these bits were 0 in NT 4.0 created SDs and 0
means NT 4.0
semantics, while each 1 bit enables a new feature.  If a Windows 2000
created SD happens
to have all the autoinherit bits turned off by the user (not uncommon) it is
binary identical to
an NT 4.0 originated SD.  If NT 4 accesses an SD with some of these bits
set, it will ignore
those bits, and the specific semantics are designed so the read-only
semantics are still
correct, and the write semantics are ok if the write clears the bits.

There is a 3 part article series on MSDN named "the guts of security" which
is a great intro
to the NT 4.0 rules.

Re: [Linux-NTFS-Dev] [Fwd: [ linux-ntfs-Bugs-849517 ] ntfs doc : directory]

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

Hi Anton,

> Summary: ntfs doc : directory

OK, I'll add it to the list.

Cheers,
  FlatCap (Rich)
  nt...@fl...