Older blog entries for AntonA (starting at number 15)

Finished the new NTFS Linux kernel driver (read-only)! Posted it and got some excellent feedback so far. Speedwise it definitely is a significant improvement. I am happy! (-8

Went to Richard M. Stallman's lecture on software patents this afternoon. Brilliant talk. RMS certainly is a very good speaker. And I fully agree with him, too. Software patents are evil full stop!

Had a feature request for mkntfs - to add ability to disable content indexing. Did that and while at it added ability to enable compression, too. After some small cleanups released Linux-NTFS 1.6.0.

Went on holiday to Austria for two weeks and enjoyed some great skiing in St. Anton am Arlberg. Had all kinds of bad weather ranging from two days with sun shine over rain, hail, snow, blizzards, and some real full blown snow storms. Fun, fun. (-:

Managed to sneak in some ntfs tng development into a few of the evenings so the driver is moving along nicely. Now have the options from the previous driver all supported but giving messages about them being deprecated and generally loads of code cleanups and enhancements. Noticed a few subtleties which require quite a bit of a rewrite of my readpage methods. Oh well, better now than after having gone into the kernel... That delays the driver quite a bit more than anticipated, so it won't be being submitted for kernel inclusion yet.

Been very busy with PhD work. First chapter is finished (results wise) and exciting results for the next one appeared which is fanfabulous. (-8

Ntfs tng is now feature complete read-only. It now reads all ntfs volumes from all windows versions (nt4/2k/xp) I threw at it and produces identical output as the old ntfs driver. - Some code cleanup left to do, some small odds and sods to implement /cleanup, bring up to speed with the latest 2.5.x offering and it's ready for the masses. The only thing then remaining is the fact that ntfs tng requires gcc-2.96 or later compiler which some people are not happy with, so I will probably have to play some dirty tricks to fix this temporarily. (Will revert them out once the kernel officially requires gcc 2.96 or later.)

Am leaving for a fortnight of internet deprived skiing holiday so no entries for a while! (-:

More fun with ntfs tng. Many cleanups, like the $ signs in identifiers are all gone now (now should compile on all architectures as long as they use gcc-2.96+), debug output is improved and can be switched on/off via sysctl/proc interface. Cleaned up mft bitmap handling and enabled loading of the complete run list. The mft data is already done which means we now should be coping with any possible mft a user will ever throw at us. (-: As it is all working nicely now I tagged a 0.0.6 release as the next set of changes is going to be enabling of full attribute list / multiple attribute extents support which may destabilize things.

Went to a charity quiz-night. Didn't win but had a lot of fun nevertheless. (-:

Had a play with encrypted ntfs directories and the ntfs 3.x extended system files. Adapted ntfs tng to cope with both and found a buglet in the process. Now all is happy.

Had a look at the compile warnings report for linux-ntfs 1.5.1 for powerpc. It seems powerpc use char == unsigned char which was breaking some signed char comparisons. Fixed that. Now need a powerpc person to test. (-;

More NTFS TNG hacking. Added the new inode allocation super operations, the recent show_options so we have a cool /proc/mounts output, etc. This was followed by a four hour debug session involving repeated smacks of head against wall to find in the end the oops on mount problem was due to a buffer overrun caused by a cut and paste error in the new inode allocator... Fixing that it all worked a treat which prompted me to do a 0.0.5 release before I break it again. (-8

Went to visit friend in Chesham for a games day. Lots of fun. (-: Picked up four Roger Zelazny books I haven't read yet, now I just need the time to read them. (-8

On the NTFS TNG front, finished the second lot of changes and made some cleanups at the same time so the total diff was about the size of the whole driver. (-; Started converting tng to the new ->alloc_inode/- >destroy_inode super block ops and realized that these mean that every ntfs_inode automatically has a vfs_inode which means I can undo a lot of the changes done to divorce the two as doing it the New Way(TM) means we have both anyway. Another set of big changes is appearing on the horizont...

Received debian packages of linux-ntfs 1.5.0 and put them on Sourceforge. With them came a report of compiler warnings and errors from compiling on various architectures. That led me to find and fix two small bugs in ntfsfix and to fixup logfile.h which I had forgotten to cleanup before. Hopefully now it really will compile on all arch, so released linux-ntfs 1.5.1.

Al Viro had a look at NTFS TNG a few days ago and had several suggestions/comments on what should be modified to make it prettier and to make it acceptable for kernel inclusion. Started making the changes. The first change, to rewrite (un)map_mft_record() to use read_cache_page() instead of the deep VM magic was easy, and removed all the VM/VFS exports Al didn't like which made the core kernel patch about 2/3 smaller. The second change, to stop using struct inode * for extent inodes, and just using ntfs_inode * for them is turning into a major rewrite as the driver always accepts struct inode * as argument to functions instead of ntfs_inode *, so about half of the driver needs to be rewritten. Also i_generation and i_ino have to be moved to ntfs_inode *. But it is well worth the effort as it removes iget_no_wait() which was just ugly and with this the core kernel patch becomes only a trivial modification to kernel/ksyms.c to export end_buffer_io_sync().

Long time no entry. Oh, well. An eventful Christmas/New Year vacation which I spent with my sister in France is over. Full with my sister's and her boyfriend's cars catching fire, burning down of their house, subsequent police interrogations, what fun... )-: At least no-one got hurt so we were all lucky in that respect. And as it turned out, accoring to the police investigators, the fire starter was my sister's car - battery leaked and a discharge caused the fire to start. Apparently this can happen to any car but the probability of it occuring is extremely small. Obviously above zero though. I feel like I was part of an action movie for a week...

Back to normal in good old England, got round to releasing linux-ntfs 1.5.0. Several changes had accumulated in cvs so it was about time especially as I found a mkntfs bug relating to the size of the created journal which I fixed.