Changes since 1.81: =================== SmartFilesystem: - Fixed Enforcer hit, which could crash the machine. - Defragmenter can now defragment very fragmented files a bit faster because it can move multiple fragments. SFSdefragmentGUI: - Updated SFSdefragmentGUI to support new features of the SFS Defragmenter. - Added a DEBUG switch to make SFSdefragmentGUI print out what it is doing. Changes since 1.80: =================== SmartFilesystem: - Reduced CPU usage of Defragmenter considerably. Defragmentation should be a lot faster on 020/030 machines now. - Fixed a bug which was introduced in 1.80. When Defragmentation should be completed, SFS would keep on moving some more data. Changes since 1.78: =================== SmartFilesystem: - Changed Defragmenter to scan for fragments from the end of the disk, instead of from the current point of defragmentation. - Fixed bug which popped a requester saying something like 'couldn't mark 128 blocks from block X because some of them were already marked'. This only happened during defragmenting, and it could have destroyed 128 blocks of your disk. - Added a way to query the version of SFS easily. SFSdefragmentGUI: - Now checks SFS version, and refuses to work with older versions. - Fixed a bug which could crash the machine. SFSformat: - Fixed a bug which could crash the machine. SetCache: - Fixed a bug which could crash the machine. Changes since 1.62: =================== SmartFilesystem: - Optimized a dozen different routines to improve small file writing speed. CPU usage is also lower in those circumstances now. - Copyback cache has been improved. The cache doesn't need to read a line anymore before it can do a write to that line (no read-on-write). The read will be delayed until it is needed, or it won't even be performed at all. Previously SFS always did a read-on-write (similair to the 040 copyback cache). - In 1.58 a bug was introduced which could cause a requester to appear where SFS claims it wants to mark 32 blocks, but some of those were already in use. The requester usually appears when a disk is quite full and quite fragmented. The bug most likely didn't cause any damage (SFScheck reports everything ok after the requester appears), but it was annoying nonetheless. This is now fixed. - The Defragmenter is now finally debugged far enough to allow it to be BETA tested. The speed problems and the problems which could arise when accessing the disk while it was being defragmented should now be eliminated. More on the defragmenter below. - SFS now will try to prevent fragmentation when multiple files are written in small bits to a SFS disk at the same time. Previously worst-case fragmentation could occur, where each of 2 files occupied one block after the other. Now SFS tries to make each part of the file atleast 10 blocks large. - SFS now estimates the size of the Adminstration space it needs and will write file-data after this reserved size to the disk. The reserved area is not fixed, and when the disk gets full it will simply be used normally. The effect of this is that Administration blocks will get grouped at the start of the disk (as long as there is space there). Previously these Administration areas would be scattered all over the disk. - A new packet, ACTION_SFS_SET, has been added. See packets.h for details. It allows you to set some parameters of SFS (until reboot for now). No tool is provided yet to set these parameters. - SFS no longer crashes when used under Kickstart 1.3. It now simply exits without mounting. - A work-around for a (possible) bug in ixemul.library (47.3) has been added. The bug has to do with Soft Links. This bug only shows up under SFS -- FFS is not affected. - Changed directory scanning order slightly. This should reduce problems even further with tools which modify entries in a directory while scanning. - A new version of SFScheck has been added. What's new? ----------- 28 March 1999, changes for version 1.58: - There is an SFSquery tool which displays a bit of information about a SFS drive. Try it. - SFS will now refuse to mount a partition if it is (partially) located after 4 GB on the drive and there was no 64-bit support available (NSD(64), TD64 or SCSI direct). This helps to prevent damage to your data. - Changed the way SFS updates the disk -- it should be slightly faster now. - The read-ahead cache should now be slightly faster on 68040 and 68060 systems by aligning the buffers to 16-byte boundaries. - Made some alterations to the structure of some SFS blocks to reduce CPU usage and to make the defragmenter easier to add in a new version. - The default name of the directory where the deleted files are stored is now '.recycled'. It is now referred to as the Recycled directory in all the documentation. Renaming the Recycled directory is of course still possible. - BUGFIX: Deleting Soft-links now works correctly. - SFS should now work under Kick 2.04. - BUGFIX: Truncating files using SetFileSize() could damage the file truncated under some conditions. - In some cases, calling SetFileSize() twice on an empty file could lead to a damaged file and loss of free space -- fixed. - BUGFIX: Fixed a small problem with disk changes and DMS. - There is no seperate SCSI direct version of SFS anymore. SCSI direct will be used automatically when no NSD 64-bit or TD64 support is found (SCSI direct users: please let me know if this works correctly for you -- use the SFSquery tool to find out if SCSI direct is being used). - BUGFIX: Moving a file from the Recycled directory to the same disk would not lower the file counter which keeps track of how many files are in the Recycled directory. This can lead to files being removed from the Recycled directory too early. - The contents of the Recycled directory are not considered anymore when calculating the free space. - BUGFIX: Overwriting empty directories by a file with the same name is no longer allowed. - SFSformat now has an option to keep the Recycled directory visible if you wish. - Format will now no longer accept names with colons (':') or slashes ('/') in them. - Updated SFScheck and SetCache. The source of SFScheck is now also included in the archive. - Now uses TD_GETGEOMETRY to detect the size of the disk. This means SFS should now work a bit better with DD & HD floppies, MO drives and any other drives which have disks of different sizes. - BUGFIX: Renaming a directory into one of its children now generates an error (Object is in use). - BUGFIX: There was a slight possibility that SFS reported that a valid disk was inserted, but did not yet tell Dos the name of the disk. - Implemented ACTION_FLUSH. - Like FFS, SFS now doesn't allow drives to be inhibited which still have pending changes. Also slightly altered the way SFS handles disk insertion and removal. - BUGFIX: Notification did not correctly check the last character of the notification path which could lead to multiple notifications being sent if there were multiple notifications which only differed by their last character. Note: This version is NOT compatible with the previous releases of SFS. You'll need to reformat existing SFS partitions if you want to use the new features. It is possible to use the new SFS and the older version at the same time, see the guide or see below. Upgrading from 1.1x ------------------- This new version of SFS is not compatible with the SFS versions 1.13 or older. Even though I try to keep SFS compatible to older versions as much as I can, this is not always an option. In this case SFS the changes were to extensive to keep SFS compatible. Most of the changes have to do with the defragmenter (even though it is not available yet) and with reducing CPU usage. This means that you'll have to reformat your SFS partitions to be able to use the latest version. However, it is possible to run two different versions of SFS at the same time -- you simply need to treat this new version of SFS as a completely new filesystem. So just like FFS can't be directly replaced by a SFS filesystem, this new version can't simply replace the old SFS version either. SFS in the RDB -------------- If you've put the old version of SFS in the RDB then you can add the new version there as well. Don't remove the old version until you've converted all of your SFS partitions. Add the new version, but give it a different DosType. Normally you used 0x53465300 (=SFS/0), but give the new version a different DosType. I'd recommend using 0x53465301 (=SFS/1). Now you can choose between the old SFS version (SFS/0) and the new one (SFS/1) for each of your partitions. If you changed one of your old SFS partitions to the new version, SFS will put up a requester during booting telling you that the SFS disk is in a format which is not supported anymore. This is normal. After formatting the disk it should be useable again. You can use SFSformat or the normal Format command with the QUICK option to format the disk. SFS with mountlists ------------------- If you mount your SFS partitions from a mountlist then you can also use two versions of SFS at the same time. Just copy the new version of SFS to your L: directory with a different name (for example 'SmartFilesystem2'). Now in your mountlists you can use the new version of SFS by changing the following lines: Filesystem = l:SmartFilesystem2 DosType = 0x53465301 Also take a look at the Installation section in the guide for more details. Introduction ------------ Smart Filesystem is a new filesystem for your Amiga. A filesystem's main purpose is to store files on a disk in such a way that they can be located and used quickly. How this is done is up to the filesystem. The way information is stored on your disk has a large impact on speed and space efficiency, and so these factors can vary a lot from filesystem to filesystem. Smart Filesystem was created to provide you with an alternative to the Fast Fileystem. Smart Filesystem makes more efficient use of your disk space, has better performance in most areas and will allow for new features never seen before in an Amiga filesystem.