original development tree for Linux kernel GTP module; now long in mainline.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

343 lines
9.1 KiB

  1. (Note: I consider version numbers as cheap. That means
  2. that I do not like numbers like 0.1 and the like for
  3. things that can be used since quite some time. But
  4. then, 3.1 doesn't mean 'perfectly stable', too.)
  5. Known bugs:
  6. -----------
  7. - Doesn't work on the alpha. The only 64/32-bit
  8. problem that I'm aware of (pointer/int conversion
  9. in readdir()) gives compiler warnings but is
  10. apparently not causing the failure, as directory
  11. reads basically work (but all files are of size 0).
  12. Alas, I've got no alpha to debug. :-(
  13. - The partition checker (drivers/block/genhd.c)
  14. doesn't work with devices which have 256 byte
  15. blocks (some very old SCSI drives).
  16. - The feature to automatically make the fs clean
  17. might leave a trashed file system with the
  18. bitmap flag set valid.
  19. - When a file is truncated to a size that is not
  20. a multiple of the blocksize, the rest of the
  21. last allocated block is not cleared. Well,
  22. this fs never claimed to be Posix conformant.
  23. Please direct bug reports to: zippel@linux-m68k.org
  24. Version 3.20
  25. ------------
  26. - kill kernel lock
  27. - fix for a possible bitmap corruption
  28. Version 3.19
  29. ------------
  30. - sizeof changes from Kernel Janitor Project
  31. - several bug fixes found with fsx
  32. Version 3.18
  33. ------------
  34. - change to global min macro + warning fixes
  35. - add module tags
  36. Version 3.17
  37. ------------
  38. - locking fixes
  39. - wrong sign in __affs_hash_dentry
  40. - remove unnecessary check in affs_new_inode
  41. - enable international mode for dircache fs
  42. Version 3.16
  43. ------------
  44. - use mark_buffer_dirty_inode instead of mark_buffer_dirty.
  45. - introduce affs_lock_{link|dir|ext}.
  46. Version 3.15
  47. ------------
  48. - disable link to directories until we can properly support them.
  49. - locking fixes for link creation/removal.
  50. Version 3.14
  51. ------------
  52. - correctly cut off long file names for compares
  53. - correctly initialize s_last_bmap
  54. Version 3.13
  55. ------------
  56. Major cleanup for 2.4 [Roman Zippel]
  57. - new extended block handling
  58. - new bitmap allocation functions
  59. - locking should be safe for the future
  60. - cleanup of some interfaces
  61. Version 3.12
  62. ------------
  63. more 2.4 fixes: [Roman Zippel]
  64. - s_lock changes
  65. - increased getblock mess
  66. - clear meta blocks
  67. Version 3.11
  68. ------------
  69. - Converted to use 2.3.x page cache [Dave Jones]
  70. - Corruption in truncate() bugfix [Ken Tyler <kent@werple.net.au>]
  71. Version 3.10
  72. ------------
  73. - Changed partition checker to allow devices
  74. with physical blocks != 512 bytes.
  75. - The partition checker now also ignores the
  76. word at 0xd0 that Windows likes to write to.
  77. Version 3.9
  78. -----------
  79. - Moved cleanup from release_file() to put_inode().
  80. This makes the first one obsolete.
  81. - truncate() zeroes the unused remainder of a
  82. partially used last block when a file is truncated.
  83. It also marks the inode dirty now (which is not
  84. really necessary as notify_change() will do
  85. it anyway).
  86. - Added a few comments, fixed some typos (and
  87. introduced some new ones), made the debug messages
  88. more consistent. Changed a bad example in the
  89. doc file (affs.txt).
  90. - Sets the NOEXEC flag in read_super() for old file
  91. systems, since you can't run programs on them.
  92. Version 3.8
  93. -----------
  94. Bill Hawes kindly reviewed the affs and sent me the
  95. patches he did. They're marked (BH). Thanks, Bill!
  96. - Cleanup of error handling in read_super().
  97. Didn't release all resources in case of an
  98. error. (BH)
  99. - put_inode() releases the ext cache only if it's
  100. no longer needed. (BH)
  101. - One set of dentry callbacks is enough. (BH)
  102. - Cleanup of error handling in namei.c. (BH)
  103. - Cleanup of error handling in file.c. (BH)
  104. - The original blocksize of the device is
  105. restored when the fs is unmounted. (BH)
  106. - getblock() did not invalidate the key cache
  107. when it allocated a new block.
  108. - Removed some unnecessary locks as Bill
  109. suggested.
  110. - Simplified match_name(), changed all hashing
  111. and case insensitive name comparisons to use
  112. uppercase. This makes the tolower() routines
  113. obsolete.
  114. - Added mount option 'mufs' to force muFS
  115. uid/gid interpretation.
  116. - File mode changes were not updated on disk.
  117. This was fixed before, but somehow got lost.
  118. Version 3.7
  119. -----------
  120. - Added dentry callbacks to allow the dcache to
  121. operate case insensitive and length ignorant
  122. like the affs itself.
  123. - getblock() didn't update the lastblock field in the
  124. inode if the fs was not an OFS. This bug only shows
  125. up if a file was enlarged via truncate() and there
  126. was not enough space.
  127. - Remove some more superfluous code left over from
  128. the old link days ...
  129. - Fixed some oversights which were in patch 2.1.78.
  130. - Fixed a few typos.
  131. Version 3.6
  132. -----------
  133. - dentry changes. (Thanks to Jes Sorensen for his help.)
  134. - Fixed bug in balloc(): Superblock was not set dirty after
  135. the bitmap was changed, so the bitmap wasn't sync'd.
  136. - Fixed nasty bug in find_new_zone(): If the current
  137. zone number was zero, the loop didn't terminate,
  138. causing a solid lock-up.
  139. - Removed support for old-style directory reads.
  140. - Fixed bug in add_entry(): When doing a sorted insert,
  141. the pointer to the next entry in the hash chain wasn't
  142. correctly byte-swapped. Since most of the users of the
  143. affs use it on a 68k, they didn't notice. But why did
  144. I not find this during my tests?
  145. - Fixed some oversights (version wasn't updated on some
  146. directory changes).
  147. - Handling of hard links rewritten. To the VFS
  148. they appear now as normal Unix links. They are
  149. now resolved only once in lookup(). The backside
  150. is that unlink(), rename() and rmdir() have to
  151. be smart about them, but the result is worth the
  152. effort. This also led to some code cleanup.
  153. - Changed name type to unsigned char; the test for
  154. invalid filenames didn't work correctly.
  155. (Thanks to Michael Krause for pointing at this.)
  156. - Changed mapping of executable flag.
  157. - Changed all network byte-order macros to the
  158. recommended ones.
  159. - Added a remount function, so attempts to remount
  160. a dircache filesystem or one with errors read/write
  161. can be trapped. Previously, ro remounts didn't
  162. flush the super block, and rw remounts didn't
  163. create allocation zones ...
  164. - Call shrink_dcache_parent() in rmdir().
  165. (Thanks to Bill Hawes.)
  166. - Permission checks in unlink().
  167. - Allow mounting of volumes with superfluous
  168. bitmap pointers read only, also allows them
  169. to be remounted read/write.
  170. - Owner/Group defaults now to the fs user (i.e.
  171. the one that mounted it) instead of root. This
  172. obsoletes the mount options uid and gid.
  173. - Argument to volume option could overflow the
  174. name buffer. It is now silently truncated to
  175. 30 characters. (Damn it! This kind of bug
  176. is too embarrassing.)
  177. - Split inode.c into 2 files, the superblock
  178. routines desperately wanted their own file.
  179. - truncate() didn't allocate an extension block
  180. cache. If a file was extended by means of
  181. truncate(), this led to an Oops.
  182. - fsuser is now checked last.
  183. - rename() will not ignore changes in filename
  184. casing any more (though mv(1) still won't allow
  185. you to do "mv oldname OldName").
  186. Version 3.5
  187. -----------
  188. - Extension block caches are now allocated on
  189. demand instead of when a file is opened, as
  190. files can be read and written without opening
  191. them (e. g. the loopback device does this).
  192. - Removed an unused function.
  193. Version 3.4
  194. -----------
  195. - Hash chains are now sorted by block numbers.
  196. (Thanks to Kars de Jong for finding this.)
  197. - Removed all unnecessary external symbols.
  198. Version 3.3
  199. -----------
  200. - Tried to make all types 'correct' and consistent.
  201. - Errors and warnings are now reported via a
  202. function. They are all prefixed by a severity
  203. and have the same appearance:
  204. "AFFS: <function>: <error message>"
  205. (There's one exception to this, as in that function
  206. is no pointer to the super block available.)
  207. - The filesystem is remounted read-only after an
  208. error.
  209. - The names of newly created filesystem objects are
  210. now checked for validity.
  211. - Minor cleanups in comments.
  212. - Added this Changes file. At last!
  213. Version 3.2
  214. -----------
  215. - Extension block cache: Reading/writing of huge files
  216. (several MB) is much faster (of course the added
  217. overhead slows down opening, but this is hardly
  218. noticeable).
  219. - The same get_block()-routine can now be used for
  220. both OFS and FFS.
  221. - The super block is now searched in the block that
  222. was calculated and in the one following. This
  223. should remedy the round-off error introduced by
  224. the 1-k blocks that Linux uses.
  225. - Minor changes to adhere to the new VFS interface.
  226. - The number of used blocks is now also calculated
  227. if the filesystem is mounted read-only.
  228. - Prefixed some constants with AFFS_ to avoid name
  229. clashes.
  230. - Removed 'EXPERIMENTAL' status.
  231. Version 3.1
  232. -----------
  233. - Fixed a nasty bug which didn't allow read-only
  234. mounts.
  235. - Allow dir-cache filesystems to be mounted
  236. read only.
  237. - OFS support.
  238. - Several other changes I just cannot remember
  239. any more.
  240. Version 3.0
  241. -----------
  242. - Almost complete rewrite for the new VFS
  243. interface in Linux 1.3.
  244. - Write support.
  245. - Support for hard and symbolic links.
  246. - Lots of things I remember even less ...
  247. Version 2.0
  248. -----------
  249. - Fixed a few things to get it compiled.
  250. - Automatic root block calculation.
  251. - Partition checker for genhd.c
  252. ========================================
  253. Let's just call Ray Burr's original affs
  254. 'Version 1.0'.