1ToDo/Notes: 2 - Find and fix bugs. 3 - The only places in the kernel where a file is resized are 4 ntfs_file_write*() and ntfs_truncate() for both of which i_mutex is 5 held. Just have to be careful in read-/writepage and other helpers 6 not running under i_mutex that we play nice. Also need to be careful 7 with initialized_size extension in ntfs_file_write*() and writepage. 8 UPDATE: The only things that need to be checked are the compressed 9 write and the other attribute resize/write cases like index 10 attributes, etc. For now none of these are implemented so are safe. 11 - Implement filling in of holes in aops.c::ntfs_writepage() and its 12 helpers. 13 - Implement mft.c::sync_mft_mirror_umount(). We currently will just 14 leave the volume dirty on umount if the final iput(vol->mft_ino) 15 causes a write of any mirrored mft records due to the mft mirror 16 inode having been discarded already. Whether this can actually ever 17 happen is unclear however so it is worth waiting until someone hits 18 the problem. 19 202.1.28 - Fix a deadlock. 21 22 - Fix deadlock in fs/ntfs/inode.c::ntfs_put_inode(). Thanks to Sergey 23 Vlasov for the report and detailed analysis of the deadlock. The fix 24 involved getting rid of ntfs_put_inode() altogether and hence NTFS no 25 longer has a ->put_inode super operation. 26 272.1.27 - Various bug fixes and cleanups. 28 29 - Fix two compiler warnings on Alpha. Thanks to Andrew Morton for 30 reporting them. 31 - Fix an (innocent) off-by-one error in the runlist code. 32 - Fix a buggette in an "should be impossible" case handling where we 33 continued the attribute lookup loop instead of aborting it. 34 - Use buffer_migrate_page() for the ->migratepage function of all ntfs 35 address space operations. 36 - Fix comparison of $MFT and $MFTMirr to not bail out when there are 37 unused, invalid mft records which are the same in both $MFT and 38 $MFTMirr. 39 - Add support for sparse files which have a compression unit of 0. 40 - Remove all the make_bad_inode() calls. This should only be called 41 from read inode and new inode code paths. 42 - Limit name length in fs/ntfs/unistr.c::ntfs_nlstoucs() to maximum 43 allowed by NTFS, i.e. 255 Unicode characters, not including the 44 terminating NULL (which is not stored on disk). 45 - Improve comments on file attribute flags in fs/ntfs/layout.h. 46 - Fix a bug in fs/ntfs/inode.c::ntfs_read_locked_index_inode() where we 47 forgot to update a temporary variable so loading index inodes which 48 have an index allocation attribute failed. 49 - Add a missing call to flush_dcache_mft_record_page() in 50 fs/ntfs/inode.c::ntfs_write_inode(). 51 - Handle the recently introduced -ENAMETOOLONG return value from 52 fs/ntfs/unistr.c::ntfs_nlstoucs() in fs/ntfs/namei.c::ntfs_lookup(). 53 - Semaphore to mutex conversion. (Ingo Molnar) 54 552.1.26 - Minor bug fixes and updates. 56 57 - Fix a potential overflow in file.c where a cast to s64 was missing in 58 a left shift of a page index. 59 - The struct inode has had its i_sem semaphore changed to a mutex named 60 i_mutex. 61 - We have struct kmem_cache now so use it instead of the typedef 62 kmem_cache_t. (Pekka Enberg) 63 - Implement support for sector sizes above 512 bytes (up to the maximum 64 supported by NTFS which is 4096 bytes). 65 - Do more detailed reporting of why we cannot mount read-write by 66 special casing the VOLUME_MODIFIED_BY_CHKDSK flag. 67 - Miscellaneous updates to layout.h. 68 - Cope with attribute list attribute having invalid flags. Windows 69 copes with this and even chkdsk does not detect or fix this so we 70 have to cope with it, too. Thanks to Pawel Kot for reporting the 71 problem. 72 732.1.25 - (Almost) fully implement write(2) and truncate(2). 74 75 - Change ntfs_map_runlist_nolock(), ntfs_attr_find_vcn_nolock() and 76 {__,}ntfs_cluster_free() to also take an optional attribute search 77 context as argument. This allows calling these functions with the 78 mft record mapped. Update all callers. 79 - Fix potential deadlock in ntfs_mft_data_extend_allocation_nolock() 80 error handling by passing in the active search context when calling 81 ntfs_cluster_free(). 82 - Change ntfs_cluster_alloc() to take an extra boolean parameter 83 specifying whether the cluster are being allocated to extend an 84 attribute or to fill a hole. 85 - Change ntfs_attr_make_non_resident() to call ntfs_cluster_alloc() 86 with @is_extension set to TRUE and remove the runlist terminator 87 fixup code as this is now done by ntfs_cluster_alloc(). 88 - Change ntfs_attr_make_non_resident to take the attribute value size 89 as an extra parameter. This is needed since we need to know the size 90 before we can map the mft record and our callers always know it. The 91 reason we cannot simply read the size from the vfs inode i_size is 92 that this is not necessarily uptodate. This happens when 93 ntfs_attr_make_non_resident() is called in the ->truncate call path. 94 - Fix ntfs_attr_make_non_resident() to update the vfs inode i_blocks 95 which is zero for a resident attribute but should no longer be zero 96 once the attribute is non-resident as it then has real clusters 97 allocated. 98 - Add fs/ntfs/attrib.[hc]::ntfs_attr_extend_allocation(), a function to 99 extend the allocation of an attributes. Optionally, the data size, 100 but not the initialized size can be extended, too. 101 - Implement fs/ntfs/inode.[hc]::ntfs_truncate(). It only supports 102 uncompressed and unencrypted files and it never creates sparse files 103 at least for the moment (making a file sparse requires us to modify 104 its directory entries and we do not support directory operations at 105 the moment). Also, support for highly fragmented files, i.e. ones 106 whose data attribute is split across multiple extents, is severly 107 limited. When such a case is encountered, EOPNOTSUPP is returned. 108 - Enable ATTR_SIZE attribute changes in ntfs_setattr(). This completes 109 the initial implementation of file truncation. Now both open(2)ing 110 a file with the O_TRUNC flag and the {,f}truncate(2) system calls 111 will resize a file appropriately. The limitations are that only 112 uncompressed and unencrypted files are supported. Also, there is 113 only very limited support for highly fragmented files (the ones whose 114 $DATA attribute is split into multiple attribute extents). 115 - In attrib.c::ntfs_attr_set() call balance_dirty_pages_ratelimited() 116 and cond_resched() in the main loop as we could be dirtying a lot of 117 pages and this ensures we play nice with the VM and the system as a 118 whole. 119 - Implement file operations ->write, ->aio_write, ->writev for regular 120 files. This replaces the old use of generic_file_write(), et al and 121 the address space operations ->prepare_write and ->commit_write. 122 This means that both sparse and non-sparse (unencrypted and 123 uncompressed) files can now be extended using the normal write(2) 124 code path. There are two limitations at present and these are that 125 we never create sparse files and that we only have limited support 126 for highly fragmented files, i.e. ones whose data attribute is split 127 across multiple extents. When such a case is encountered, 128 EOPNOTSUPP is returned. 129 - $EA attributes can be both resident and non-resident. 130 - Use %z for size_t to fix compilation warnings. (Andrew Morton) 131 - Fix compilation warnings with gcc-4.0.2 on SUSE 10.0. 132 - Document extended attribute ($EA) NEED_EA flag. (Based on libntfs 133 patch by Yura Pakhuchiy.) 134 1352.1.24 - Lots of bug fixes and support more clean journal states. 136 137 - Support journals ($LogFile) which have been modified by chkdsk. This 138 means users can boot into Windows after we marked the volume dirty. 139 The Windows boot will run chkdsk and then reboot. The user can then 140 immediately boot into Linux rather than having to do a full Windows 141 boot first before rebooting into Linux and we will recognize such a 142 journal and empty it as it is clean by definition. Note, this only 143 works if chkdsk left the journal in an obviously clean state. 144 - Support journals ($LogFile) with only one restart page as well as 145 journals with two different restart pages. We sanity check both and 146 either use the only sane one or the more recent one of the two in the 147 case that both are valid. 148 - Add fs/ntfs/malloc.h::ntfs_malloc_nofs_nofail() which is analogous to 149 ntfs_malloc_nofs() but it performs allocations with __GFP_NOFAIL and 150 hence cannot fail. 151 - Use ntfs_malloc_nofs_nofail() in the two critical regions in 152 fs/ntfs/runlist.c::ntfs_runlists_merge(). This means we no longer 153 need to panic() if the allocation fails as it now cannot fail. 154 - Fix two nasty runlist merging bugs that had gone unnoticed so far. 155 Thanks to Stefano Picerno for the bug report. 156 - Remove two bogus BUG_ON()s from fs/ntfs/mft.c. 157 - Fix handling of valid but empty mapping pairs array in 158 fs/ntfs/runlist.c::ntfs_mapping_pairs_decompress(). 159 - Report unrepresentable inodes during ntfs_readdir() as KERN_WARNING 160 messages and include the inode number. Thanks to Yura Pakhuchiy for 161 pointing this out. 162 - Change ntfs_rl_truncate_nolock() to throw away the runlist if the new 163 length is zero. 164 - Add runlist.[hc]::ntfs_rl_punch_nolock() which punches a caller 165 specified hole into a runlist. 166 - Fix a bug in fs/ntfs/index.c::ntfs_index_lookup(). When the returned 167 index entry is in the index root, we forgot to set the @ir pointer in 168 the index context. Thanks to Yura Pakhuchiy for finding this bug. 169 - Remove bogus setting of PageError in ntfs_read_compressed_block(). 170 - Add fs/ntfs/attrib.[hc]::ntfs_resident_attr_value_resize(). 171 - Fix a bug in ntfs_map_runlist_nolock() where we forgot to protect 172 access to the allocated size in the ntfs inode with the size lock. 173 - Fix ntfs_attr_vcn_to_lcn_nolock() and ntfs_attr_find_vcn_nolock() to 174 return LCN_ENOENT when there is no runlist and the allocated size is 175 zero. 176 - Fix load_attribute_list() to handle the case of a NULL runlist. 177 - Fix handling of sparse attributes in ntfs_attr_make_non_resident(). 178 - Add BUG() checks to ntfs_attr_make_non_resident() and ntfs_attr_set() 179 to ensure that these functions are never called for compressed or 180 encrypted attributes. 181 - Fix cluster (de)allocators to work when the runlist is NULL and more 182 importantly to take a locked runlist rather than them locking it 183 which leads to lock reversal. 184 - Truncate {a,c,m}time to the ntfs supported time granularity when 185 updating the times in the inode in ntfs_setattr(). 186 - Fixup handling of sparse, compressed, and encrypted attributes in 187 fs/ntfs/inode.c::ntfs_read_locked_{,attr_,index_}inode(), 188 fs/ntfs/aops.c::ntfs_{read,write}page(). 189 - Make ntfs_write_block() not instantiate sparse blocks if they contain 190 only zeroes. 191 - Optimize fs/ntfs/aops.c::ntfs_write_block() by extending the page 192 lock protection over the buffer submission for i/o which allows the 193 removal of the get_bh()/put_bh() pairs for each buffer. 194 - Fix fs/ntfs/aops.c::ntfs_{read,write}_block() to handle the case 195 where a concurrent truncate has truncated the runlist under our feet. 196 - Fix page_has_buffers()/page_buffers() handling in fs/ntfs/aops.c. 197 - In fs/ntfs/aops.c::ntfs_end_buffer_async_read(), use a bit spin lock 198 in the first buffer head instead of a driver global spin lock to 199 improve scalability. 200 - Minor fix to error handling and error message display in 201 fs/ntfs/aops.c::ntfs_prepare_nonresident_write(). 202 - Change the mount options {u,f,d}mask to always parse the number as 203 an octal number to conform to how chmod(1) works, too. Thanks to 204 Giuseppe Bilotta and Horst von Brand for pointing out the errors of 205 my ways. 206 - Fix various bugs in the runlist merging code. (Based on libntfs 207 changes by Richard Russon.) 208 - Fix sparse warnings that have crept in over time. 209 - Change ntfs_cluster_free() to require a write locked runlist on entry 210 since we otherwise get into a lock reversal deadlock if a read locked 211 runlist is passed in. In the process also change it to take an ntfs 212 inode instead of a vfs inode as parameter. 213 - Fix the definition of the CHKD ntfs record magic. It had an off by 214 two error causing it to be CHKB instead of CHKD. 215 - Fix a stupid bug in __ntfs_bitmap_set_bits_in_run() which caused the 216 count to become negative and hence we had a wild memset() scribbling 217 all over the system's ram. 218 2192.1.23 - Implement extension of resident files and make writing safe as well as 220 many bug fixes, cleanups, and enhancements... 221 222 - Add printk rate limiting for ntfs_warning() and ntfs_error() when 223 compiled without debug. This avoids a possible denial of service 224 attack. Thanks to Carl-Daniel Hailfinger from SuSE for pointing this 225 out. 226 - Fix compilation warnings on ia64. (Randy Dunlap) 227 - Use i_size_{read,write}() instead of reading i_size by hand and cache 228 the value where apropriate. 229 - Add size_lock to the ntfs_inode structure. This is an rw spinlock 230 and it locks against access to the inode sizes. Note, ->size_lock 231 is also accessed from irq context so you must use the _irqsave and 232 _irqrestore lock and unlock functions, respectively. Protect all 233 accesses to allocated_size, initialized_size, and compressed_size. 234 - Minor optimization to fs/ntfs/super.c::ntfs_statfs() and its helpers. 235 - Implement extension of resident files in the regular file write code 236 paths (fs/ntfs/aops.c::ntfs_{prepare,commit}_write()). At present 237 this only works until the data attribute becomes too big for the mft 238 record after which we abort the write returning -EOPNOTSUPP from 239 ntfs_prepare_write(). 240 - Add disable_sparse mount option together with a per volume sparse 241 enable bit which is set appropriately and a per inode sparse disable 242 bit which is preset on some system file inodes as appropriate. 243 - Enforce that sparse support is disabled on NTFS volumes pre 3.0. 244 - Fix a bug in fs/ntfs/runlist.c::ntfs_mapping_pairs_decompress() in 245 the creation of the unmapped runlist element for the base attribute 246 extent. 247 - Split ntfs_map_runlist() into ntfs_map_runlist() and a non-locking 248 helper ntfs_map_runlist_nolock() which is used by ntfs_map_runlist(). 249 This allows us to map runlist fragments with the runlist lock already 250 held without having to drop and reacquire it around the call. Adapt 251 all callers. 252 - Change ntfs_find_vcn() to ntfs_find_vcn_nolock() which takes a locked 253 runlist. This allows us to find runlist elements with the runlist 254 lock already held without having to drop and reacquire it around the 255 call. Adapt all callers. 256 - Change time to u64 in time.h::ntfs2utc() as it otherwise generates a 257 warning in the do_div() call on sparc32. Thanks to Meelis Roos for 258 the report and analysis of the warning. 259 - Fix a nasty runlist merge bug when merging two holes. 260 - Set the ntfs_inode->allocated_size to the real allocated size in the 261 mft record for resident attributes (fs/ntfs/inode.c). 262 - Small readability cleanup to use "a" instead of "ctx->attr" 263 everywhere (fs/ntfs/inode.c). 264 - Make fs/ntfs/namei.c::ntfs_get_{parent,dentry} static and move the 265 definition of ntfs_export_ops from fs/ntfs/super.c to namei.c. Also, 266 declare ntfs_export_ops in fs/ntfs/ntfs.h. 267 - Correct sparse file handling. The compressed values need to be 268 checked and set in the ntfs inode as done for compressed files and 269 the compressed size needs to be used for vfs inode->i_blocks instead 270 of the allocated size, again, as done for compressed files. 271 - Add AT_EA in addition to AT_DATA to whitelist for being allowed to be 272 non-resident in fs/ntfs/attrib.c::ntfs_attr_can_be_non_resident(). 273 - Add fs/ntfs/attrib.c::ntfs_attr_vcn_to_lcn_nolock() used by the new 274 write code. 275 - Fix bug in fs/ntfs/attrib.c::ntfs_find_vcn_nolock() where after 276 dropping the read lock and taking the write lock we were not checking 277 whether someone else did not already do the work we wanted to do. 278 - Rename fs/ntfs/attrib.c::ntfs_find_vcn_nolock() to 279 ntfs_attr_find_vcn_nolock() and update all callers. 280 - Add fs/ntfs/attrib.[hc]::ntfs_attr_make_non_resident(). 281 - Fix sign of various error return values to be negative in 282 fs/ntfs/lcnalloc.c. 283 - Modify ->readpage and ->writepage (fs/ntfs/aops.c) so they detect and 284 handle the case where an attribute is converted from resident to 285 non-resident by a concurrent file write. 286 - Remove checks for NULL before calling kfree() since kfree() does the 287 checking itself. (Jesper Juhl) 288 - Some utilities modify the boot sector but do not update the checksum. 289 Thus, relax the checking in fs/ntfs/super.c::is_boot_sector_ntfs() to 290 only emit a warning when the checksum is incorrect rather than 291 refusing the mount. Thanks to Bernd Casimir for pointing this 292 problem out. 293 - Update attribute definition handling. 294 - Add NTFS_MAX_CLUSTER_SIZE and NTFS_MAX_PAGES_PER_CLUSTER constants. 295 - Use NTFS_MAX_CLUSTER_SIZE in super.c instead of hard coding 0x10000. 296 - Use MAX_BUF_PER_PAGE instead of variable sized array allocation for 297 better code generation and one less sparse warning in fs/ntfs/aops.c. 298 - Remove spurious void pointer casts from fs/ntfs/. (Pekka Enberg) 299 - Use C99 style structure initialization after memory allocation where 300 possible (fs/ntfs/{attrib.c,index.c,super.c}). Thanks to Al Viro and 301 Pekka Enberg. 302 - Stamp the transaction log ($UsnJrnl), aka user space journal, if it 303 is active on the volume and we are mounting read-write or remounting 304 from read-only to read-write. 305 - Fix a bug in address space operations error recovery code paths where 306 if the runlist was not mapped at all and a mapping error occured we 307 would leave the runlist locked on exit to the function so that the 308 next access to the same file would try to take the lock and deadlock. 309 - Detect the case when Windows has been suspended to disk on the volume 310 to be mounted and if this is the case do not allow (re)mounting 311 read-write. This is done by parsing hiberfil.sys if present. 312 - Fix several occurences of a bug where we would perform 'var & ~const' 313 with a 64-bit variable and a int, i.e. 32-bit, constant. This causes 314 the higher order 32-bits of the 64-bit variable to be zeroed. To fix 315 this cast the 'const' to the same 64-bit type as 'var'. 316 - Change the runlist terminator of the newly allocated cluster(s) to 317 LCN_ENOENT in ntfs_attr_make_non_resident(). Otherwise the runlist 318 code gets confused. 319 - Add an extra parameter @last_vcn to ntfs_get_size_for_mapping_pairs() 320 and ntfs_mapping_pairs_build() to allow the runlist encoding to be 321 partial which is desirable when filling holes in sparse attributes. 322 Update all callers. 323 - Change ntfs_map_runlist_nolock() to only decompress the mapping pairs 324 if the requested vcn is inside it. Otherwise we get into problems 325 when we try to map an out of bounds vcn because we then try to map 326 the already mapped runlist fragment which causes 327 ntfs_mapping_pairs_decompress() to fail and return error. Update 328 ntfs_attr_find_vcn_nolock() accordingly. 329 - Fix a nasty deadlock that appeared in recent kernels. 330 The situation: VFS inode X on a mounted ntfs volume is dirty. For 331 same inode X, the ntfs_inode is dirty and thus corresponding on-disk 332 inode, i.e. mft record, which is in a dirty PAGE_CACHE_PAGE belonging 333 to the table of inodes, i.e. $MFT, inode 0. 334 What happens: 335 Process 1: sys_sync()/umount()/whatever... calls 336 __sync_single_inode() for $MFT -> do_writepages() -> write_page for 337 the dirty page containing the on-disk inode X, the page is now locked 338 -> ntfs_write_mst_block() which clears PageUptodate() on the page to 339 prevent anyone else getting hold of it whilst it does the write out. 340 This is necessary as the on-disk inode needs "fixups" applied before 341 the write to disk which are removed again after the write and 342 PageUptodate is then set again. It then analyses the page looking 343 for dirty on-disk inodes and when it finds one it calls 344 ntfs_may_write_mft_record() to see if it is safe to write this 345 on-disk inode. This then calls ilookup5() to check if the 346 corresponding VFS inode is in icache(). This in turn calls ifind() 347 which waits on the inode lock via wait_on_inode whilst holding the 348 global inode_lock. 349 Process 2: pdflush results in a call to __sync_single_inode for the 350 same VFS inode X on the ntfs volume. This locks the inode (I_LOCK) 351 then calls write-inode -> ntfs_write_inode -> map_mft_record() -> 352 read_cache_page() for the page (in page cache of table of inodes 353 $MFT, inode 0) containing the on-disk inode. This page has 354 PageUptodate() clear because of Process 1 (see above) so 355 read_cache_page() blocks when it tries to take the page lock for the 356 page so it can call ntfs_read_page(). 357 Thus Process 1 is holding the page lock on the page containing the 358 on-disk inode X and it is waiting on the inode X to be unlocked in 359 ifind() so it can write the page out and then unlock the page. 360 And Process 2 is holding the inode lock on inode X and is waiting for 361 the page to be unlocked so it can call ntfs_readpage() or discover 362 that Process 1 set PageUptodate() again and use the page. 363 Thus we have a deadlock due to ifind() waiting on the inode lock. 364 The solution: The fix is to use the newly introduced 365 ilookup5_nowait() which does not wait on the inode's lock and hence 366 avoids the deadlock. This is safe as we do not care about the VFS 367 inode and only use the fact that it is in the VFS inode cache and the 368 fact that the vfs and ntfs inodes are one struct in memory to find 369 the ntfs inode in memory if present. Also, the ntfs inode has its 370 own locking so it does not matter if the vfs inode is locked. 371 - Fix bug in mft record writing where we forgot to set the device in 372 the buffers when mapping them after the VM had discarded them. 373 Thanks to Martin MOKREJ�� for the bug report. 374 3752.1.22 - Many bug and race fixes and error handling improvements. 376 377 - Improve error handling in fs/ntfs/inode.c::ntfs_truncate(). 378 - Change fs/ntfs/inode.c::ntfs_truncate() to return an error code 379 instead of void and provide a helper ntfs_truncate_vfs() for the 380 vfs ->truncate method. 381 - Add a new ntfs inode flag NInoTruncateFailed() and modify 382 fs/ntfs/inode.c::ntfs_truncate() to set and clear it appropriately. 383 - Fix min_size and max_size definitions in ATTR_DEF structure in 384 fs/ntfs/layout.h to be signed. 385 - Add attribute definition handling helpers to fs/ntfs/attrib.[hc]: 386 ntfs_attr_size_bounds_check(), ntfs_attr_can_be_non_resident(), and 387 ntfs_attr_can_be_resident(), which in turn use the new private helper 388 ntfs_attr_find_in_attrdef(). 389 - In fs/ntfs/aops.c::mark_ntfs_record_dirty(), take the 390 mapping->private_lock around the dirtying of the buffer heads 391 analagous to the way it is done in __set_page_dirty_buffers(). 392 - Ensure the mft record size does not exceed the PAGE_CACHE_SIZE at 393 mount time as this cannot work with the current implementation. 394 - Check for location of attribute name and improve error handling in 395 general in fs/ntfs/inode.c::ntfs_read_locked_inode() and friends. 396 - In fs/ntfs/aops.c::ntfs_writepage(), if the page is fully outside 397 i_size, i.e. race with truncate, invalidate the buffers on the page 398 so that they become freeable and hence the page does not leak. 399 - Remove unused function fs/ntfs/runlist.c::ntfs_rl_merge(). (Adrian 400 Bunk) 401 - Fix stupid bug in fs/ntfs/attrib.c::ntfs_attr_find() that resulted in 402 a NULL pointer dereference in the error code path when a corrupt 403 attribute was found. (Thanks to Domen Puncer for the bug report.) 404 - Add MODULE_VERSION() to fs/ntfs/super.c. 405 - Make several functions and variables static. (Adrian Bunk) 406 - Modify fs/ntfs/aops.c::mark_ntfs_record_dirty() so it allocates 407 buffers for the page if they are not present and then marks the 408 buffers belonging to the ntfs record dirty. This causes the buffers 409 to become busy and hence they are safe from removal until the page 410 has been written out. 411 - Fix stupid bug in fs/ntfs/attrib.c::ntfs_external_attr_find() in the 412 error handling code path that resulted in a BUG() due to trying to 413 unmap an extent mft record when the mapping of it had failed and it 414 thus was not mapped. (Thanks to Ken MacFerrin for the bug report.) 415 - Drop the runlist lock after the vcn has been read in 416 fs/ntfs/lcnalloc.c::__ntfs_cluster_free(). 417 - Rewrite handling of multi sector transfer errors. We now do not set 418 PageError() when such errors are detected in the async i/o handler 419 fs/ntfs/aops.c::ntfs_end_buffer_async_read(). All users of mst 420 protected attributes now check the magic of each ntfs record as they 421 use it and act appropriately. This has the effect of making errors 422 granular per ntfs record rather than per page which solves the case 423 where we cannot access any of the ntfs records in a page when a 424 single one of them had an mst error. (Thanks to Ken MacFerrin for 425 the bug report.) 426 - Fix error handling in fs/ntfs/quota.c::ntfs_mark_quotas_out_of_date() 427 where we failed to release i_mutex on the $Quota/$Q attribute inode. 428 - Fix bug in handling of bad inodes in fs/ntfs/namei.c::ntfs_lookup(). 429 - Add mapping of unmapped buffers to all remaining code paths, i.e. 430 fs/ntfs/aops.c::ntfs_write_mst_block(), mft.c::ntfs_sync_mft_mirror(), 431 and write_mft_record_nolock(). From now on we require that the 432 complete runlist for the mft mirror is always mapped into memory. 433 - Add creation of buffers to fs/ntfs/mft.c::ntfs_sync_mft_mirror(). 434 - Improve error handling in fs/ntfs/aops.c::ntfs_{read,write}_block(). 435 - Cleanup fs/ntfs/aops.c::ntfs_{read,write}page() since we know that a 436 resident attribute will be smaller than a page which makes the code 437 simpler. Also make the code more tolerant to concurrent ->truncate. 438 4392.1.21 - Fix some races and bugs, rewrite mft write code, add mft allocator. 440 441 - Implement extent mft record deallocation 442 fs/ntfs/mft.c::ntfs_extent_mft_record_free(). 443 - Splitt runlist related functions off from attrib.[hc] to runlist.[hc]. 444 - Add vol->mft_data_pos and initialize it at mount time. 445 - Rename init_runlist() to ntfs_init_runlist(), ntfs_vcn_to_lcn() to 446 ntfs_rl_vcn_to_lcn(), decompress_mapping_pairs() to 447 ntfs_mapping_pairs_decompress(), ntfs_merge_runlists() to 448 ntfs_runlists_merge() and adapt all callers. 449 - Add fs/ntfs/runlist.[hc]::ntfs_get_nr_significant_bytes(), 450 ntfs_get_size_for_mapping_pairs(), ntfs_write_significant_bytes(), 451 and ntfs_mapping_pairs_build(), adapted from libntfs. 452 - Make fs/ntfs/lcnalloc.c::ntfs_cluster_free_from_rl_nolock() not 453 static and add a declaration for it to lcnalloc.h. 454 - Add fs/ntfs/lcnalloc.h::ntfs_cluster_free_from_rl() which is a static 455 inline wrapper for ntfs_cluster_free_from_rl_nolock() which takes the 456 cluster bitmap lock for the duration of the call. 457 - Add fs/ntfs/attrib.[hc]::ntfs_attr_record_resize(). 458 - Implement the equivalent of memset() for an ntfs attribute in 459 fs/ntfs/attrib.[hc]::ntfs_attr_set() and switch 460 fs/ntfs/logfile.c::ntfs_empty_logfile() to using it. 461 - Remove unnecessary casts from LCN_* constants. 462 - Implement fs/ntfs/runlist.c::ntfs_rl_truncate_nolock(). 463 - Add MFT_RECORD_OLD as a copy of MFT_RECORD in fs/ntfs/layout.h and 464 change MFT_RECORD to contain the NTFS 3.1+ specific fields. 465 - Add a helper function fs/ntfs/aops.c::mark_ntfs_record_dirty() which 466 marks all buffers belonging to an ntfs record dirty, followed by 467 marking the page the ntfs record is in dirty and also marking the vfs 468 inode containing the ntfs record dirty (I_DIRTY_PAGES). 469 - Switch fs/ntfs/index.h::ntfs_index_entry_mark_dirty() to using the 470 new helper fs/ntfs/aops.c::mark_ntfs_record_dirty() and remove the no 471 longer needed fs/ntfs/index.[hc]::__ntfs_index_entry_mark_dirty(). 472 - Move ntfs_{un,}map_page() from ntfs.h to aops.h and fix resulting 473 include errors. 474 - Move the typedefs for runlist_element and runlist from types.h to 475 runlist.h and fix resulting include errors. 476 - Remove unused {__,}format_mft_record() from fs/ntfs/mft.c. 477 - Modify fs/ntfs/mft.c::__mark_mft_record_dirty() to use the helper 478 mark_ntfs_record_dirty() which also changes the behaviour in that we 479 now set the buffers belonging to the mft record dirty as well as the 480 page itself. 481 - Update fs/ntfs/mft.c::write_mft_record_nolock() and sync_mft_mirror() 482 to cope with the fact that there now are dirty buffers in mft pages. 483 - Update fs/ntfs/inode.c::ntfs_write_inode() to also use the helper 484 mark_ntfs_record_dirty() and thus to set the buffers belonging to the 485 mft record dirty as well as the page itself. 486 - Fix compiler warnings on x86-64 in fs/ntfs/dir.c. (Randy Dunlap, 487 slightly modified by me) 488 - Add fs/ntfs/mft.c::try_map_mft_record() which fails with -EALREADY if 489 the mft record is already locked and otherwise behaves the same way 490 as fs/ntfs/mft.c::map_mft_record(). 491 - Modify fs/ntfs/mft.c::write_mft_record_nolock() so that it only 492 writes the mft record if the buffers belonging to it are dirty. 493 Otherwise we assume that it was written out by other means already. 494 - Attempting to write outside initialized size is _not_ a bug so remove 495 the bug check from fs/ntfs/aops.c::ntfs_write_mst_block(). It is in 496 fact required to write outside initialized size when preparing to 497 extend the initialized size. 498 - Map the page instead of using page_address() before writing to it in 499 fs/ntfs/aops.c::ntfs_mft_writepage(). 500 - Provide exclusion between opening an inode / mapping an mft record 501 and accessing the mft record in fs/ntfs/mft.c::ntfs_mft_writepage() 502 by setting the page not uptodate throughout ntfs_mft_writepage(). 503 - Clear the page uptodate flag in fs/ntfs/aops.c::ntfs_write_mst_block() 504 to ensure noone can see the page whilst the mst fixups are applied. 505 - Add the helper fs/ntfs/mft.c::ntfs_may_write_mft_record() which 506 checks if an mft record may be written out safely obtaining any 507 necessary locks in the process. This is used by 508 fs/ntfs/aops.c::ntfs_write_mst_block(). 509 - Modify fs/ntfs/aops.c::ntfs_write_mst_block() to also work for 510 writing mft records and improve its error handling in the process. 511 Now if any of the records in the page fail to be written out, all 512 other records will be written out instead of aborting completely. 513 - Remove ntfs_mft_aops and update all users to use ntfs_mst_aops. 514 - Modify fs/ntfs/inode.c::ntfs_read_locked_inode() to set the 515 ntfs_mst_aops for all inodes which are NInoMstProtected() and 516 ntfs_aops for all other inodes. 517 - Rename fs/ntfs/mft.c::sync_mft_mirror{,_umount}() to 518 ntfs_sync_mft_mirror{,_umount}() and change their parameters so they 519 no longer require an ntfs inode to be present. Update all callers. 520 - Cleanup the error handling in fs/ntfs/mft.c::ntfs_sync_mft_mirror(). 521 - Clear the page uptodate flag in fs/ntfs/mft.c::ntfs_sync_mft_mirror() 522 to ensure noone can see the page whilst the mst fixups are applied. 523 - Remove the no longer needed fs/ntfs/mft.c::ntfs_mft_writepage() and 524 fs/ntfs/mft.c::try_map_mft_record(). 525 - Fix callers of fs/ntfs/aops.c::mark_ntfs_record_dirty() to call it 526 with the ntfs inode which contains the page rather than the ntfs 527 inode the mft record of which is in the page. 528 - Fix race condition in fs/ntfs/inode.c::ntfs_put_inode() by moving the 529 index inode bitmap inode release code from there to 530 fs/ntfs/inode.c::ntfs_clear_big_inode(). (Thanks to Christoph 531 Hellwig for spotting this.) 532 - Fix race condition in fs/ntfs/inode.c::ntfs_put_inode() by taking the 533 inode semaphore around the code that sets ni->itype.index.bmp_ino to 534 NULL and reorganize the code to optimize it a bit. (Thanks to 535 Christoph Hellwig for spotting this.) 536 - Modify fs/ntfs/aops.c::mark_ntfs_record_dirty() to no longer take the 537 ntfs inode as a parameter as this is confusing and misleading and the 538 needed ntfs inode is available via NTFS_I(page->mapping->host). 539 Adapt all callers to this change. 540 - Modify fs/ntfs/mft.c::write_mft_record_nolock() and 541 fs/ntfs/aops.c::ntfs_write_mst_block() to only check the dirty state 542 of the first buffer in a record and to take this as the ntfs record 543 dirty state. We cannot look at the dirty state for subsequent 544 buffers because we might be racing with 545 fs/ntfs/aops.c::mark_ntfs_record_dirty(). 546 - Move the static inline ntfs_init_big_inode() from fs/ntfs/inode.c to 547 inode.h and make fs/ntfs/inode.c::__ntfs_init_inode() non-static and 548 add a declaration for it to inode.h. Fix some compilation issues 549 that resulted due to #includes and header file interdependencies. 550 - Simplify setup of i_mode in fs/ntfs/inode.c::ntfs_read_locked_inode(). 551 - Add helpers fs/ntfs/layout.h::MK_MREF() and MK_LE_MREF(). 552 - Modify fs/ntfs/mft.c::map_extent_mft_record() to only verify the mft 553 record sequence number if it is specified (i.e. not zero). 554 - Add fs/ntfs/mft.[hc]::ntfs_mft_record_alloc() and various helper 555 functions used by it. 556 - Update Documentation/filesystems/ntfs.txt with instructions on how to 557 use the Device-Mapper driver with NTFS ftdisk/LDM raid. This removes 558 the linear raid problem with the Software RAID / MD driver when one 559 or more of the devices has an odd number of sectors. 560 5612.1.20 - Fix two stupid bugs introduced in 2.1.18 release. 562 563 - Fix stupid bug in fs/ntfs/attrib.c::ntfs_attr_reinit_search_ctx() 564 where we did not clear ctx->al_entry but it was still set due to 565 changes in ntfs_attr_lookup() and ntfs_external_attr_find() in 566 particular. 567 - Fix another stupid bug in fs/ntfs/attrib.c::ntfs_external_attr_find() 568 where we forgot to unmap the extent mft record when we had finished 569 enumerating an attribute which caused a bug check to trigger when the 570 VFS calls ->clear_inode. 571 5722.1.19 - Many cleanups, improvements, and a minor bug fix. 573 574 - Update ->setattr (fs/ntfs/inode.c::ntfs_setattr()) to refuse to 575 change the uid, gid, and mode of an inode as we do not support NTFS 576 ACLs yet. 577 - Remove BKL use from ntfs_setattr() syncing up with the rest of the 578 kernel. 579 - Get rid of the ugly transparent union in fs/ntfs/dir.c::ntfs_readdir() 580 and ntfs_filldir() as per suggestion from Al Viro. 581 - Change '\0' and L'\0' to simply 0 as per advice from Linus Torvalds. 582 - Update ->truncate (fs/ntfs/inode.c::ntfs_truncate()) to check if the 583 inode size has changed and to only output an error if so. 584 - Rename fs/ntfs/attrib.h::attribute_value_length() to ntfs_attr_size(). 585 - Add le{16,32,64} as well as sle{16,32,64} data types to 586 fs/ntfs/types.h. 587 - Change ntfschar to be le16 instead of u16 in fs/ntfs/types.h. 588 - Add le versions of VCN, LCN, and LSN called leVCN, leLCN, and leLSN, 589 respectively, to fs/ntfs/types.h. 590 - Update endianness conversion macros in fs/ntfs/endian.h to use the 591 new types as appropriate. 592 - Do proper type casting when using sle64_to_cpup() in fs/ntfs/dir.c 593 and index.c. 594 - Add leMFT_REF data type to fs/ntfs/layout.h. 595 - Update all NTFS header files with the new little endian data types. 596 Affected files are fs/ntfs/layout.h, logfile.h, and time.h. 597 - Do proper type casting when using ntfs_is_*_recordp() in 598 fs/ntfs/logfile.c, mft.c, and super.c. 599 - Fix all the sparse bitwise warnings. Had to change all the typedef 600 enums storing little endian values to simple enums plus a typedef for 601 the datatype to make sparse happy. 602 - Fix a bug found by the new sparse bitwise warnings where the default 603 upcase table was defined as a pointer to wchar_t rather than ntfschar 604 in fs/ntfs/ntfs.h and super.c. 605 - Change {const_,}cpu_to_le{16,32}(0) to just 0 as suggested by Al Viro. 606 6072.1.18 - Fix scheduling latencies at mount time as well as an endianness bug. 608 609 - Remove vol->nr_mft_records as it was pretty meaningless and optimize 610 the calculation of total/free inodes as used by statfs(). 611 - Fix scheduling latencies in ntfs_fill_super() by dropping the BKL 612 because the code itself is using the ntfs_lock semaphore which 613 provides safe locking. (Ingo Molnar) 614 - Fix a potential bug in fs/ntfs/mft.c::map_extent_mft_record() that 615 could occur in the future for when we start closing/freeing extent 616 inodes if we don't set base_ni->ext.extent_ntfs_inos to NULL after 617 we free it. 618 - Rename {find,lookup}_attr() to ntfs_attr_{find,lookup}() as well as 619 find_external_attr() to ntfs_external_attr_find() to cleanup the 620 namespace a bit and to be more consistent with libntfs. 621 - Rename {{re,}init,get,put}_attr_search_ctx() to 622 ntfs_attr_{{re,}init,get,put}_search_ctx() as well as the type 623 attr_search_context to ntfs_attr_search_ctx. 624 - Force use of ntfs_attr_find() in ntfs_attr_lookup() when searching 625 for the attribute list attribute itself. 626 - Fix endianness bug in ntfs_external_attr_find(). 627 - Change ntfs_{external_,}attr_find() to return 0 on success, -ENOENT 628 if the attribute is not found, and -EIO on real error. In the case 629 of -ENOENT, the search context is updated to describe the attribute 630 before which the attribute being searched for would need to be 631 inserted if such an action were to be desired and in the case of 632 ntfs_external_attr_find() the search context is also updated to 633 indicate the attribute list entry before which the attribute list 634 entry of the attribute being searched for would need to be inserted 635 if such an action were to be desired. Also make ntfs_find_attr() 636 static and remove its prototype from attrib.h as it is not used 637 anywhere other than attrib.c. Update ntfs_attr_lookup() and all 638 callers of ntfs_{external,}attr_{find,lookup}() for the new return 639 values. 640 - Minor cleanup of fs/ntfs/inode.c::ntfs_init_locked_inode(). 641 6422.1.17 - Fix bugs in mount time error code paths and other updates. 643 644 - Implement bitmap modification code (fs/ntfs/bitmap.[hc]). This 645 includes functions to set/clear a single bit or a run of bits. 646 - Add fs/ntfs/attrib.[hc]::ntfs_find_vcn() which returns the locked 647 runlist element containing a particular vcn. It also takes care of 648 mapping any needed runlist fragments. 649 - Implement cluster (de-)allocation code (fs/ntfs/lcnalloc.[hc]). 650 - Load attribute definition table from $AttrDef at mount time. 651 - Fix bugs in mount time error code paths involving (de)allocation of 652 the default and volume upcase tables. 653 - Remove ntfs_nr_mounts as it is no longer used. 654 6552.1.16 - Implement access time updates, file sync, async io, and read/writev. 656 657 - Add support for readv/writev and aio_read/aio_write (fs/ntfs/file.c). 658 This is done by setting the appropriate file operations pointers to 659 the generic helper functions provided by mm/filemap.c. 660 - Implement fsync, fdatasync, and msync both for files (fs/ntfs/file.c) 661 and directories (fs/ntfs/dir.c). 662 - Add support for {a,m,c}time updates to inode.c::ntfs_write_inode(). 663 Note, except for the root directory and any other system files opened 664 by the user, the system files will not have their access times 665 updated as they are only accessed at the inode level an hence the 666 file level functions which cause the times to be updated are never 667 invoked. 668 6692.1.15 - Invalidate quotas when (re)mounting read-write. 670 671 - Add new element itype.index.collation_rule to the ntfs inode 672 structure and set it appropriately in ntfs_read_locked_inode(). 673 - Implement a new inode type "index" to allow efficient access to the 674 indices found in various system files and adapt inode handling 675 accordingly (fs/ntfs/inode.[hc]). An index inode is essentially an 676 attribute inode (NInoAttr() is true) with an attribute type of 677 AT_INDEX_ALLOCATION. As such, it is no longer allowed to call 678 ntfs_attr_iget() with an attribute type of AT_INDEX_ALLOCATION as 679 there would be no way to distinguish between normal attribute inodes 680 and index inodes. The function to obtain an index inode is 681 ntfs_index_iget() and it uses the helper function 682 ntfs_read_locked_index_inode(). Note, we do not overload 683 ntfs_attr_iget() as indices consist of multiple attributes so using 684 ntfs_attr_iget() to obtain an index inode would be confusing. 685 - Ensure that there is no overflow when doing page->index << 686 PAGE_CACHE_SHIFT by casting page->index to s64 in fs/ntfs/aops.c. 687 - Use atomic kmap instead of kmap() in fs/ntfs/aops.c::ntfs_read_page() 688 and ntfs_read_block(). 689 - Use case sensitive attribute lookups instead of case insensitive ones. 690 - Lock all page cache pages belonging to mst protected attributes while 691 accessing them to ensure we never see corrupt data while the page is 692 under writeout. 693 - Add framework for generic ntfs collation (fs/ntfs/collation.[hc]). 694 We have ntfs_is_collation_rule_supported() to check if the collation 695 rule you want to use is supported and ntfs_collation() which actually 696 collates two data items. We currently only support COLLATION_BINARY 697 and COLLATION_NTOFS_ULONG but support for other collation rules will 698 be added as the need arises. 699 - Add a new type, ntfs_index_context, to allow retrieval of an index 700 entry using the corresponding index key. To get an index context, 701 use ntfs_index_ctx_get() and to release it, use ntfs_index_ctx_put(). 702 This also adds a new slab cache for the index contexts. To lookup a 703 key in an index inode, use ntfs_index_lookup(). After modifying an 704 index entry, call ntfs_index_entry_flush_dcache_page() followed by 705 ntfs_index_entry_mark_dirty() to ensure the changes are written out 706 to disk. For details see fs/ntfs/index.[hc]. Note, at present, if 707 an index entry is in the index allocation attribute rather than the 708 index root attribute it will not be written out (you will get a 709 warning message about discarded changes instead). 710 - Load the quota file ($Quota) and check if quota tracking is enabled 711 and if so, mark the quotas out of date. This causes windows to 712 rescan the volume on boot and update all quota entries. 713 - Add a set_page_dirty address space operation for ntfs_m[fs]t_aops. 714 It is simply set to __set_page_dirty_nobuffers() to make sure that 715 running set_page_dirty() on a page containing mft/ntfs records will 716 not affect the dirty state of the page buffers. 717 - Add fs/ntfs/index.c::__ntfs_index_entry_mark_dirty() which sets all 718 buffers that are inside the ntfs record in the page dirty after which 719 it sets the page dirty. This allows ->writepage to only write the 720 dirty index records rather than having to write all the records in 721 the page. Modify fs/ntfs/index.h::ntfs_index_entry_mark_dirty() to 722 use this rather than __set_page_dirty_nobuffers(). 723 - Implement fs/ntfs/aops.c::ntfs_write_mst_block() which enables the 724 writing of page cache pages belonging to mst protected attributes 725 like the index allocation attribute in directory indices and other 726 indices like $Quota/$Q, etc. This means that the quota is now marked 727 out of date on all volumes rather than only on ones where the quota 728 defaults entry is in the index root attribute of the $Quota/$Q index. 729 7302.1.14 - Fix an NFSd caused deadlock reported by several users. 731 732 - Modify fs/ntfs/ntfs_readdir() to copy the index root attribute value 733 to a buffer so that we can put the search context and unmap the mft 734 record before calling the filldir() callback. We need to do this 735 because of NFSd which calls ->lookup() from its filldir callback() 736 and this causes NTFS to deadlock as ntfs_lookup() maps the mft record 737 of the directory and since ntfs_readdir() has got it mapped already 738 ntfs_lookup() deadlocks. 739 7402.1.13 - Enable overwriting of resident files and housekeeping of system files. 741 742 - Implement writing of mft records (fs/ntfs/mft.[hc]), which includes 743 keeping the mft mirror in sync with the mft when mirrored mft records 744 are written. The functions are write_mft_record{,_nolock}(). The 745 implementation is quite rudimentary for now with lots of things not 746 implemented yet but I am not sure any of them can actually occur so 747 I will wait for people to hit each one and only then implement it. 748 - Commit open system inodes at umount time. This should make it 749 virtually impossible for sync_mft_mirror_umount() to ever be needed. 750 - Implement ->write_inode (fs/ntfs/inode.c::ntfs_write_inode()) for the 751 ntfs super operations. This gives us inode writing via the VFS inode 752 dirty code paths. Note: Access time updates are not implemented yet. 753 - Implement fs/ntfs/mft.[hc]::{,__}mark_mft_record_dirty() and make 754 fs/ntfs/aops.c::ntfs_writepage() and ntfs_commit_write() use it, thus 755 finally enabling resident file overwrite! (-8 This also includes a 756 placeholder for ->writepage (ntfs_mft_writepage()), which for now 757 just redirties the page and returns. Also, at umount time, we for 758 now throw away all mft data page cache pages after the last call to 759 ntfs_commit_inode() in the hope that all inodes will have been 760 written out by then and hence no dirty (meta)data will be lost. We 761 also check for this case and emit an error message telling the user 762 to run chkdsk. 763 - Use set_page_writeback() and end_page_writeback() in the resident 764 attribute code path of fs/ntfs/aops.c::ntfs_writepage() otherwise 765 the radix-tree tag PAGECACHE_TAG_DIRTY remains set even though the 766 page is clean. 767 - Implement ntfs_mft_writepage() so it now checks if any of the mft 768 records in the page are dirty and if so redirties the page and 769 returns. Otherwise it just returns (after doing set_page_writeback(), 770 unlock_page(), end_page_writeback() or the radix-tree tag 771 PAGECACHE_TAG_DIRTY remains set even though the page is clean), thus 772 alowing the VM to do with the page as it pleases. Also, at umount 773 time, now only throw away dirty mft (meta)data pages if dirty inodes 774 are present and ask the user to email us if they see this happening. 775 - Add functions ntfs_{clear,set}_volume_flags(), to modify the volume 776 information flags (fs/ntfs/super.c). 777 - Mark the volume dirty when (re)mounting read-write and mark it clean 778 when unmounting or remounting read-only. If any volume errors are 779 found, the volume is left marked dirty to force chkdsk to run. 780 - Add code to set the NT4 compatibility flag when (re)mounting 781 read-write for newer NTFS versions but leave it commented out for now 782 since we do not make any modifications that are NTFS 1.2 specific yet 783 and since setting this flag breaks Captive-NTFS which is not nice. 784 This code must be enabled once we start writing NTFS 1.2 specific 785 changes otherwise Windows NTFS driver might crash / cause corruption. 786 7872.1.12 - Fix the second fix to the decompression engine and some cleanups. 788 789 - Add a new address space operations struct, ntfs_mst_aops, for mst 790 protected attributes. This is because the default ntfs_aops do not 791 make sense with mst protected data and were they to write anything to 792 such an attribute they would cause data corruption so we provide 793 ntfs_mst_aops which does not have any write related operations set. 794 - Cleanup dirty ntfs inode handling (fs/ntfs/inode.[hc]) which also 795 includes an adapted ntfs_commit_inode() and an implementation of 796 ntfs_write_inode() which for now just cleans dirty inodes without 797 writing them (it does emit a warning that this is happening). 798 - Undo the second decompression engine fix (see 2.1.9 release ChangeLog 799 entry) as it was only fixing a theoretical bug but at the same time 800 it badly broke the handling of sparse and uncompressed compression 801 blocks. 802 8032.1.11 - Driver internal cleanups. 804 805 - Only build logfile.o if building the driver with read-write support. 806 - Really final white space cleanups. 807 - Use generic_ffs() instead of ffs() in logfile.c which allows the 808 log_page_size variable to be optimized by gcc into a constant. 809 - Rename uchar_t to ntfschar everywhere as uchar_t is unsigned 1-byte 810 char as defined by POSIX and as found on some systems. 811 8122.1.10 - Force read-only (re)mounting of volumes with unsupported volume flags. 813 814 - Finish off the white space cleanups (remove trailing spaces, etc). 815 - Clean up ntfs_fill_super() and ntfs_read_inode_mount() by removing 816 the kludges around the first iget(). Instead of (re)setting ->s_op 817 we have the $MFT inode set up by explicit new_inode() / set ->i_ino / 818 insert_inode_hash() / call ntfs_read_inode_mount() directly. This 819 kills the need for second super_operations and allows to return error 820 from ntfs_read_inode_mount() without resorting to ugly "poisoning" 821 tricks. (Al Viro) 822 - Force read-only (re)mounting if any of the following bits are set in 823 the volume information flags: 824 VOLUME_IS_DIRTY, VOLUME_RESIZE_LOG_FILE, 825 VOLUME_UPGRADE_ON_MOUNT, VOLUME_DELETE_USN_UNDERWAY, 826 VOLUME_REPAIR_OBJECT_ID, VOLUME_MODIFIED_BY_CHKDSK 827 To make this easier we define VOLUME_MUST_MOUNT_RO_MASK with all the 828 above bits set so the test is made easy. 829 8302.1.9 - Fix two bugs in decompression engine. 831 832 - Fix a bug where we would not always detect that we have reached the 833 end of a compression block because we were ending at minus one byte 834 which is effectively the same as being at the end. The fix is to 835 check whether the uncompressed buffer has been fully filled and if so 836 we assume we have reached the end of the compression block. A big 837 thank you to Marcin Gibu��a for the bug report, the assistance in 838 tracking down the bug and testing the fix. 839 - Fix a possible bug where when a compressed read is truncated to the 840 end of the file, the offset inside the last page was not truncated. 841 8422.1.8 - Handle $MFT mirror and $LogFile, improve time handling, and cleanups. 843 844 - Use get_bh() instead of manual atomic_inc() in fs/ntfs/compress.c. 845 - Modify fs/ntfs/time.c::ntfs2utc(), get_current_ntfs_time(), and 846 utc2ntfs() to work with struct timespec instead of time_t on the 847 Linux UTC time side thus preserving the full precision of the NTFS 848 time and only loosing up to 99 nano-seconds in the Linux UTC time. 849 - Move fs/ntfs/time.c to fs/ntfs/time.h and make the time functions 850 static inline. 851 - Remove unused ntfs_dirty_inode(). 852 - Cleanup super operations declaration in fs/ntfs/super.c. 853 - Wrap flush_dcache_mft_record_page() in #ifdef NTFS_RW. 854 - Add NInoTestSetFoo() and NInoTestClearFoo() macro magic to 855 fs/ntfs/inode.h and use it to declare NInoTest{Set,Clear}Dirty. 856 - Move typedefs for ntfs_attr and test_t from fs/ntfs/inode.c to 857 fs/ntfs/inode.h so they can be used elsewhere. 858 - Determine the mft mirror size as the number of mirrored mft records 859 and store it in ntfs_volume->mftmirr_size (fs/ntfs/super.c). 860 - Load the mft mirror at mount time and compare the mft records stored 861 in it to the ones in the mft. Force a read-only mount if the two do 862 not match (fs/ntfs/super.c). 863 - Fix type casting related warnings on 64-bit architectures. Thanks 864 to Meelis Roos for reporting them. 865 - Move %L to %ll as %L is floating point and %ll is integer which is 866 what we want. 867 - Read the journal ($LogFile) and determine if the volume has been 868 shutdown cleanly and force a read-only mount if not (fs/ntfs/super.c 869 and fs/ntfs/logfile.c). This is a little bit of a crude check in 870 that we only look at the restart areas and not at the actual log 871 records so that there will be a very small number of cases where we 872 think that a volume is dirty when in fact it is clean. This should 873 only affect volumes that have not been shutdown cleanly and did not 874 have any pending, non-check-pointed i/o. 875 - If the $LogFile indicates a clean shutdown and a read-write (re)mount 876 is requested, empty $LogFile by overwriting it with 0xff bytes to 877 ensure that Windows cannot cause data corruption by replaying a stale 878 journal after Linux has written to the volume. 879 8802.1.7 - Enable NFS exporting of mounted NTFS volumes. 881 882 - Set i_generation in the VFS inode from the seq_no of the NTFS inode. 883 - Make ntfs_lookup() NFS export safe, i.e. use d_splice_alias(), etc. 884 - Implement ->get_dentry() in fs/ntfs/namei.c::ntfs_get_dentry() as the 885 default doesn't allow inode number 0 which is a valid inode on NTFS 886 and even if it did allow that it uses iget() instead of ntfs_iget() 887 which makes it useless for us. 888 - Implement ->get_parent() in fs/ntfs/namei.c::ntfs_get_parent() as the 889 default just returns -EACCES which is not very useful. 890 - Define export operations (->s_export_op) for NTFS (ntfs_export_ops) 891 and set them up in the super block at mount time (super.c) this 892 allows mounted NTFS volumes to be exported via NFS. 893 - Add missing return -EOPNOTSUPP; in 894 fs/ntfs/aops.c::ntfs_commit_nonresident_write(). 895 - Enforce no atime and no dir atime updates at mount/remount time as 896 they are not implemented yet anyway. 897 - Move a few assignments in fs/ntfs/attrib.c::load_attribute_list() to 898 after a NULL check. Thanks to Dave Jones for pointing this out. 899 9002.1.6 - Fix minor bug in handling of compressed directories. 901 902 - Fix bug in handling of compressed directories. A compressed 903 directory is not really compressed so when we set the ->i_blocks 904 field of a compressed directory inode we were setting it from the 905 non-existing field ni->itype.compressed.size which gave random 906 results... For directories we now always use ni->allocated_size. 907 9082.1.5 - Fix minor bug in attribute list attribute handling. 909 910 - Fix bug in attribute list handling. Actually it is not as much a bug 911 as too much protection in that we were not allowing attribute lists 912 which waste space on disk while Windows XP clearly allows it and in 913 fact creates such attribute lists so our driver was failing. 914 - Update NTFS documentation ready for 2.6 kernel release. 915 9162.1.4 - Reduce compiler requirements. 917 918 - Remove all uses of unnamed structs and unions in the driver to make 919 old and newer gcc versions happy. Makes it a bit uglier IMO but at 920 least people will stop hassling me about it. 921 9222.1.3 - Important bug fixes in corner cases. 923 924 - super.c::parse_ntfs_boot_sector(): Correct the check for 64-bit 925 clusters. (Philipp Thomas) 926 - attrib.c::load_attribute_list(): Fix bug when initialized_size is a 927 multiple of the block_size but not the cluster size. (Szabolcs 928 Szakacsits) 929 9302.1.2 - Important bug fixes aleviating the hangs in statfs. 931 932 - Fix buggy free cluster and free inode determination logic. 933 9342.1.1 - Minor updates. 935 936 - Add handling for initialized_size != data_size in compressed files. 937 - Reduce function local stack usage from 0x3d4 bytes to just noise in 938 fs/ntfs/upcase.c. (Randy Dunlap) 939 - Remove compiler warnings for newer gcc. 940 - Pages are no longer kmapped by mm/filemap.c::generic_file_write() 941 around calls to ->{prepare,commit}_write. Adapt NTFS appropriately 942 in fs/ntfs/aops.c::ntfs_prepare_nonresident_write() by using 943 kmap_atomic(KM_USER0). 944 9452.1.0 - First steps towards write support: implement file overwrite. 946 947 - Add configuration option for developmental write support with an 948 appropriately scary configuration help text. 949 - Initial implementation of fs/ntfs/aops.c::ntfs_writepage() and its 950 helper fs/ntfs/aops.c::ntfs_write_block(). This enables mmap(2) based 951 overwriting of existing files on ntfs. Note: Resident files are 952 only written into memory, and not written out to disk at present, so 953 avoid writing to files smaller than about 1kiB. 954 - Initial implementation of fs/ntfs/aops.c::ntfs_prepare_write(), its 955 helper fs/ntfs/aops.c::ntfs_prepare_nonresident_write() and their 956 counterparts, fs/ntfs/aops.c::ntfs_commit_write(), and 957 fs/ntfs/aops.c::ntfs_commit_nonresident_write(), respectively. Also, 958 add generic_file_write() to the ntfs file operations (fs/ntfs/file.c). 959 This enables write(2) based overwriting of existing files on ntfs. 960 Note: As with mmap(2) based overwriting, resident files are only 961 written into memory, and not written out to disk at present, so avoid 962 writing to files smaller than about 1kiB. 963 - Implement ->truncate (fs/ntfs/inode.c::ntfs_truncate()) and 964 ->setattr() (fs/ntfs/inode.c::ntfs_setattr()) inode operations for 965 files with the purpose of intercepting and aborting all i_size 966 changes which we do not support yet. ntfs_truncate() actually only 967 emits a warning message but AFAICS our interception of i_size changes 968 elsewhere means ntfs_truncate() never gets called for i_size changes. 969 It is only called from generic_file_write() when we fail in 970 ntfs_prepare_{,nonresident_}write() in order to discard any 971 instantiated buffers beyond i_size. Thus i_size is not actually 972 changed so our warning message is enough. Unfortunately it is not 973 possible to easily determine if i_size is being changed or not hence 974 we just emit an appropriately worded error message. 975 9762.0.25 - Small bug fixes and cleanups. 977 978 - Unlock the page in an out of memory error code path in 979 fs/ntfs/aops.c::ntfs_read_block(). 980 - If fs/ntfs/aops.c::ntfs_read_page() is called on an uptodate page, 981 just unlock the page and return. (This can happen due to ->writepage 982 clearing PageUptodate() during write out of MstProtected() 983 attributes. 984 - Remove leaked write code again. 985 9862.0.24 - Cleanups. 987 988 - Treat BUG_ON() as ASSERT() not VERIFY(), i.e. do not use side effects 989 inside BUG_ON(). (Adam J. Richter) 990 - Split logical OR expressions inside BUG_ON() into individual BUG_ON() 991 calls for improved debugging. (Adam J. Richter) 992 - Add errors flag to the ntfs volume state, accessed via 993 NVol{,Set,Clear}Errors(vol). 994 - Do not allow read-write remounts of read-only volumes with errors. 995 - Clarify comment for ntfs file operation sendfile which was added by 996 Christoph Hellwig a while ago (just using generic_file_sendfile()) 997 to say that ntfs ->sendfile is only used for the case where the 998 source data is on the ntfs partition and the destination is 999 somewhere else, i.e. nothing we need to concern ourselves with. 1000 - Add generic_file_write() as our ntfs file write operation. 1001 10022.0.23 - Major bug fixes (races, deadlocks, non-i386 architectures). 1003 1004 - Massive internal locking changes to mft record locking. Fixes lock 1005 recursion and replaces the mrec_lock read/write semaphore with a 1006 mutex. Also removes the now superfluous mft_count. This fixes several 1007 race conditions and deadlocks, especially in the future write code. 1008 - Fix ntfs over loopback for compressed files by adding an 1009 optimization barrier. (gcc was screwing up otherwise ?) 1010 - Miscellaneous cleanups all over the code and a fix or two in error 1011 handling code paths. 1012 Thanks go to Christoph Hellwig for pointing out the following two: 1013 - Remove now unused function fs/ntfs/malloc.h::vmalloc_nofs(). 1014 - Fix ntfs_free() for ia64 and parisc by checking for VMALLOC_END, too. 1015 10162.0.22 - Cleanups, mainly to ntfs_readdir(), and use C99 initializers. 1017 1018 - Change fs/ntfs/dir.c::ntfs_reddir() to only read/write ->f_pos once 1019 at entry/exit respectively. 1020 - Use C99 initializers for structures. 1021 - Remove unused variable blocks from fs/ntfs/aops.c::ntfs_read_block(). 1022 10232.0.21 - Check for, and refuse to work with too large files/directories/volumes. 1024 1025 - Limit volume size at mount time to 2TiB on architectures where 1026 unsigned long is 32-bits (fs/ntfs/super.c::parse_ntfs_boot_sector()). 1027 This is the most we can do without overflowing the 32-bit limit of 1028 the block device size imposed on us by sb_bread() and sb_getblk() 1029 for the time being. 1030 - Limit file/directory size at open() time to 16TiB on architectures 1031 where unsigned long is 32-bits (fs/ntfs/file.c::ntfs_file_open() and 1032 fs/ntfs/dir.c::ntfs_dir_open()). This is the most we can do without 1033 overflowing the page cache page index. 1034 10352.0.20 - Support non-resident directory index bitmaps, fix page leak in readdir. 1036 1037 - Move the directory index bitmap to use an attribute inode instead of 1038 having special fields for it inside the ntfs inode structure. This 1039 means that the index bitmaps now use the page cache for i/o, too, 1040 and also as a side effect we get support for non-resident index 1041 bitmaps for free. 1042 - Simplify/cleanup error handling in fs/ntfs/dir.c::ntfs_readdir() and 1043 fix a page leak that manifested itself in some cases. 1044 - Add fs/ntfs/inode.c::ntfs_put_inode(), which we need to release the 1045 index bitmap inode on the final iput(). 1046 10472.0.19 - Fix race condition, improvements, and optimizations in i/o interface. 1048 1049 - Apply block optimization added to fs/ntfs/aops.c::ntfs_read_block() 1050 to fs/ntfs/compress.c::ntfs_file_read_compressed_block() as well. 1051 - Drop the "file" from ntfs_file_read_compressed_block(). 1052 - Rename fs/ntfs/aops.c::ntfs_enb_buffer_read_async() to 1053 ntfs_end_buffer_async_read() (more like the fs/buffer.c counterpart). 1054 - Update ntfs_end_buffer_async_read() with the improved logic from 1055 its updated counterpart fs/buffer.c::end_buffer_async_read(). Apply 1056 further logic improvements to better determine when we set PageError. 1057 - Update submission of buffers in fs/ntfs/aops.c::ntfs_read_block() to 1058 check for the buffers being uptodate first in line with the updated 1059 fs/buffer.c::block_read_full_page(). This plugs a small race 1060 condition. 1061 10622.0.18 - Fix race condition in reading of compressed files. 1063 1064 - There was a narrow window between checking a buffer head for being 1065 uptodate and locking it in ntfs_file_read_compressed_block(). We now 1066 lock the buffer and then check whether it is uptodate or not. 1067 10682.0.17 - Cleanups and optimizations - shrinking the ToDo list. 1069 1070 - Modify fs/ntfs/inode.c::ntfs_read_locked_inode() to return an error 1071 code and update callers, i.e. ntfs_iget(), to pass that error code 1072 up instead of just using -EIO. 1073 - Modifications to super.c to ensure that both mount and remount 1074 cannot set any write related options when the driver is compiled 1075 read-only. 1076 - Optimize block resolution in fs/ntfs/aops.c::ntfs_read_block() to 1077 cache the current runlist element. This should improve performance 1078 when reading very large and/or very fragmented data. 1079 10802.0.16 - Convert access to $MFT/$BITMAP to attribute inode API. 1081 1082 - Fix a stupid bug introduced in 2.0.15 where we were unmapping the 1083 wrong inode in fs/ntfs/inode.c::ntfs_attr_iget(). 1084 - Fix debugging check in fs/ntfs/aops.c::ntfs_read_block(). 1085 - Convert $MFT/$BITMAP access to attribute inode API and remove all 1086 remnants of the ugly mftbmp address space and operations hack. This 1087 means we finally have only one readpage function as well as only one 1088 async io completion handler. Yey! The mft bitmap is now just an 1089 attribute inode and is accessed from vol->mftbmp_ino just as if it 1090 were a normal file. Fake inodes rule. (-: 1091 10922.0.15 - Fake inodes based attribute i/o via the pagecache, fixes and cleanups. 1093 1094 - Fix silly bug in fs/ntfs/super.c::parse_options() which was causing 1095 remounts to fail when the partition had an entry in /etc/fstab and 1096 the entry specified the nls= option. 1097 - Apply same macro magic used in fs/ntfs/inode.h to fs/ntfs/volume.h to 1098 expand all the helper functions NVolFoo(), NVolSetFoo(), and 1099 NVolClearFoo(). 1100 - Move copyright statement from driver initialisation message to 1101 module description (fs/super.c). This makes the initialisation 1102 message fit on one line and fits in better with rest of kernel. 1103 - Update fs/ntfs/attrib.c::map_run_list() to work on both real and 1104 attribute inodes, and both for files and directories. 1105 - Implement fake attribute inodes allowing all attribute i/o to go via 1106 the page cache and to use all the normal vfs/mm functionality: 1107 - Add ntfs_attr_iget() and its helper ntfs_read_locked_attr_inode() 1108 to fs/ntfs/inode.c. 1109 - Add needed cleanup code to ntfs_clear_big_inode(). 1110 - Merge address space operations for files and directories (aops.c), 1111 now just have ntfs_aops: 1112 - Rename: 1113 end_buffer_read_attr_async() -> ntfs_end_buffer_read_async(), 1114 ntfs_attr_read_block() -> ntfs_read_block(), 1115 ntfs_file_read_page() -> ntfs_readpage(). 1116 - Rewrite fs/ntfs/aops.c::ntfs_readpage() to work on both real and 1117 attribute inodes, and both for files and directories. 1118 - Remove obsolete fs/ntfs/aops.c::ntfs_mst_readpage(). 1119 11202.0.14 - Run list merging code cleanup, minor locking changes, typo fixes. 1121 1122 - Change fs/ntfs/super.c::ntfs_statfs() to not rely on BKL by moving 1123 the locking out of super.c::get_nr_free_mft_records() and taking and 1124 dropping the mftbmp_lock rw_semaphore in ntfs_statfs() itself. 1125 - Bring attribute runlist merging code (fs/ntfs/attrib.c) in sync with 1126 current userspace ntfs library code. This means that if a merge 1127 fails the original runlists are always left unmodified instead of 1128 being silently corrupted. 1129 - Misc typo fixes. 1130 11312.0.13 - Use iget5_locked() in preparation for fake inodes and small cleanups. 1132 1133 - Remove nr_mft_bits and the now superfluous union with nr_mft_records 1134 from ntfs_volume structure. 1135 - Remove nr_lcn_bits and the now superfluous union with nr_clusters 1136 from ntfs_volume structure. 1137 - Use iget5_locked() and friends instead of conventional iget(). Wrap 1138 the call in fs/ntfs/inode.c::ntfs_iget() and update callers of iget() 1139 to use ntfs_iget(). Leave only one iget() call at mount time so we 1140 don't need an ntfs_iget_mount(). 1141 - Change fs/ntfs/inode.c::ntfs_new_extent_inode() to take mft_no as an 1142 additional argument. 1143 11442.0.12 - Initial cleanup of address space operations following 2.0.11 changes. 1145 1146 - Merge fs/ntfs/aops.c::end_buffer_read_mst_async() and 1147 fs/ntfs/aops.c::end_buffer_read_file_async() into one function 1148 fs/ntfs/aops.c::end_buffer_read_attr_async() using NInoMstProtected() 1149 to determine whether to apply mst fixups or not. 1150 - Above change allows merging fs/ntfs/aops.c::ntfs_file_read_block() 1151 and fs/ntfs/aops.c::ntfs_mst_readpage() into one function 1152 fs/ntfs/aops.c::ntfs_attr_read_block(). Also, create a tiny wrapper 1153 fs/ntfs/aops.c::ntfs_mst_readpage() to transform the parameters from 1154 the VFS readpage function prototype to the ntfs_attr_read_block() 1155 function prototype. 1156 11572.0.11 - Initial preparations for fake inode based attribute i/o. 1158 1159 - Move definition of ntfs_inode_state_bits to fs/ntfs/inode.h and 1160 do some macro magic (adapted from include/linux/buffer_head.h) to 1161 expand all the helper functions NInoFoo(), NInoSetFoo(), and 1162 NInoClearFoo(). 1163 - Add new flag to ntfs_inode_state_bits: NI_Sparse. 1164 - Add new fields to ntfs_inode structure to allow use of fake inodes 1165 for attribute i/o: type, name, name_len. Also add new state bits: 1166 NI_Attr, which, if set, indicates the inode is a fake inode, and 1167 NI_MstProtected, which, if set, indicates the attribute uses multi 1168 sector transfer protection, i.e. fixups need to be applied after 1169 reads and before/after writes. 1170 - Rename fs/ntfs/inode.c::ntfs_{new,clear,destroy}_inode() to 1171 ntfs_{new,clear,destroy}_extent_inode() and update callers. 1172 - Use ntfs_clear_extent_inode() in fs/ntfs/inode.c::__ntfs_clear_inode() 1173 instead of ntfs_destroy_extent_inode(). 1174 - Cleanup memory deallocations in {__,}ntfs_clear_{,big_}inode(). 1175 - Make all operations on ntfs inode state bits use the NIno* functions. 1176 - Set up the new ntfs inode fields and state bits in 1177 fs/ntfs/inode.c::ntfs_read_inode() and add appropriate cleanup of 1178 allocated memory to __ntfs_clear_inode(). 1179 - Cleanup ntfs_inode structure a bit for better ordering of elements 1180 w.r.t. their size to allow better packing of the structure in memory. 1181 11822.0.10 - There can only be 2^32 - 1 inodes on an NTFS volume. 1183 1184 - Add check at mount time to verify that the number of inodes on the 1185 volume does not exceed 2^32 - 1, which is the maximum allowed for 1186 NTFS according to Microsoft. 1187 - Change mft_no member of ntfs_inode structure to be unsigned long. 1188 Update all users. This makes ntfs_inode->mft_no just a copy of struct 1189 inode->i_ino. But we can't just always use struct inode->i_ino and 1190 remove mft_no because extent inodes do not have an attached struct 1191 inode. 1192 11932.0.9 - Decompression engine now uses a single buffer and other cleanups. 1194 1195 - Change decompression engine to use a single buffer protected by a 1196 spin lock instead of per-CPU buffers. (Rusty Russell) 1197 - Do not update cb_pos when handling a partial final page during 1198 decompression of a sparse compression block, as the value is later 1199 reset without being read/used. (Rusty Russell) 1200 - Switch to using the new KM_BIO_SRC_IRQ for atomic kmap()s. (Andrew 1201 Morton) 1202 - Change buffer size in ntfs_readdir()/ntfs_filldir() to use 1203 NLS_MAX_CHARSET_SIZE which makes the buffers almost 1kiB each but 1204 it also makes everything safer so it is a good thing. 1205 - Miscellaneous minor cleanups to comments. 1206 12072.0.8 - Major updates for handling of case sensitivity and dcache aliasing. 1208 1209 Big thanks go to Al Viro and other inhabitants of #kernel for investing 1210 their time to discuss the case sensitivity and dcache aliasing issues. 1211 1212 - Remove unused source file fs/ntfs/attraops.c. 1213 - Remove show_inodes mount option(s), thus dropping support for 1214 displaying of short file names. 1215 - Remove deprecated mount option posix. 1216 - Restore show_sys_files mount option. 1217 - Add new mount option case_sensitive, to determine if the driver 1218 treats file names as case sensitive or not. If case sensitive, create 1219 file names in the POSIX namespace. Otherwise create file names in the 1220 LONG/WIN32 namespace. Note, files remain accessible via their short 1221 file name, if it exists. 1222 - Remove really dumb logic bug in boot sector recovery code. 1223 - Fix dcache aliasing issues wrt short/long file names via changes 1224 to fs/ntfs/dir.c::ntfs_lookup_inode_by_name() and 1225 fs/ntfs/namei.c::ntfs_lookup(): 1226 - Add additional argument to ntfs_lookup_inode_by_name() in which we 1227 return information about the matching file name if the case is not 1228 matching or the match is a short file name. See comments above the 1229 function definition for details. 1230 - Change ntfs_lookup() to only create dcache entries for the correctly 1231 cased file name and only for the WIN32 namespace counterpart of DOS 1232 namespace file names. This ensures we have only one dentry per 1233 directory and also removes all dcache aliasing issues between short 1234 and long file names once we add write support. See comments above 1235 function for details. 1236 - Fix potential 1 byte overflow in fs/ntfs/unistr.c::ntfs_ucstonls(). 1237 12382.0.7 - Minor cleanups and updates for changes in core kernel code. 1239 1240 - Remove much of the NULL struct element initializers. 1241 - Various updates to make compatible with recent kernels. 1242 - Remove defines of MAX_BUF_PER_PAGE and include linux/buffer_head.h 1243 in fs/ntfs/ntfs.h instead. 1244 - Remove no longer needed KERNEL_VERSION checks. We are now in the 1245 kernel proper so they are no longer needed. 1246 12472.0.6 - Major bugfix to make compatible with other kernel changes. 1248 1249 - Initialize the mftbmp address space properly now that there are more 1250 fields in the struct address_space. This was leading to hangs and 1251 oopses on umount since 2.5.12 because of changes to other parts of 1252 the kernel. We probably want a kernel generic init_address_space() 1253 function... 1254 - Drop BKL from ntfs_readdir() after consultation with Al Viro. The 1255 only caller of ->readdir() is vfs_readdir() which holds i_mutex 1256 during the call, and i_mutex is sufficient protection against changes 1257 in the directory inode (including ->i_size). 1258 - Use generic_file_llseek() for directories (as opposed to 1259 default_llseek()) as this downs i_mutex instead of the BKL which is 1260 what we now need for exclusion against ->f_pos changes considering we 1261 no longer take the BKL in ntfs_readdir(). 1262 12632.0.5 - Major bugfix. Buffer overflow in extent inode handling. 1264 1265 - No need to set old blocksize in super.c::ntfs_fill_super() as the 1266 VFS does so via invocation of deactivate_super() calling 1267 fs->fill_super() calling block_kill_super() which does it. 1268 - BKL moved from VFS into dir.c::ntfs_readdir(). (Linus Torvalds) 1269 -> Do we really need it? I don't think so as we have exclusion on 1270 the directory ntfs_inode rw_semaphore mrec_lock. We mmight have to 1271 move the ->f_pos accesses under the mrec_lock though. Check this... 1272 - Fix really, really, really stupid buffer overflow in extent inode 1273 handling in mft.c::map_extent_mft_record(). 1274 12752.0.4 - Cleanups and updates for kernel 2.5.11. 1276 1277 - Add documentation on how to use the MD driver to be able to use NTFS 1278 stripe and volume sets in Linux and generally cleanup documentation 1279 a bit. 1280 Remove all uses of kdev_t in favour of struct block_device *: 1281 - Change compress.c::ntfs_file_read_compressed_block() to use 1282 sb_getblk() instead of getblk(). 1283 - Change super.c::ntfs_fill_super() to use bdev_hardsect_size() instead 1284 of get_hardsect_size(). 1285 - No need to get old blocksize in super.c::ntfs_fill_super() as 1286 fs/super.c::get_sb_bdev() already does this. 1287 - Set bh->b_bdev instead of bh->b_dev throughout aops.c. 1288 12892.0.3 - Small bug fixes, cleanups, and performance improvements. 1290 1291 - Remove some dead code from mft.c. 1292 - Optimize readpage and read_block functions throughout aops.c so that 1293 only initialized blocks are read. Non-initialized ones have their 1294 buffer head mapped, zeroed, and set up to date, without scheduling 1295 any i/o. Thanks to Al Viro for advice on how to avoid the device i/o. 1296 Thanks go to Andrew Morton for spotting the below: 1297 - Fix buglet in allocate_compression_buffers() error code path. 1298 - Call flush_dcache_page() after modifying page cache page contents in 1299 ntfs_file_readpage(). 1300 - Check for existence of page buffers throughout aops.c before calling 1301 create_empty_buffers(). This happens when an I/O error occurs and the 1302 read is retried. (It also happens once writing is implemented so that 1303 needed doing anyway but I had left it for later...) 1304 - Don't BUG_ON() uptodate and/or mapped buffers throughout aops.c in 1305 readpage and read_block functions. Reasoning same as above (i.e. I/O 1306 error retries and future write code paths.) 1307 13082.0.2 - Minor updates and cleanups. 1309 1310 - Cleanup: rename mst.c::__post_read_mst_fixup to post_write_mst_fixup 1311 and cleanup the code a bit, removing the unused size parameter. 1312 - Change default fmask to 0177 and update documentation. 1313 - Change attrib.c::get_attr_search_ctx() to return the search context 1314 directly instead of taking the address of a pointer. A return value 1315 of NULL means the allocation failed. Updated all callers 1316 appropriately. 1317 - Update to 2.5.9 kernel (preserving backwards compatibility) by 1318 replacing all occurences of page->buffers with page_buffers(page). 1319 - Fix minor bugs in runlist merging, also minor cleanup. 1320 - Updates to bootsector layout and mft mirror contents descriptions. 1321 - Small bug fix in error detection in unistr.c and some cleanups. 1322 - Grow name buffer allocations in unistr.c in aligned mutlipled of 64 1323 bytes. 1324 13252.0.1 - Minor updates. 1326 1327 - Make default umask correspond to documentation. 1328 - Improve documentation. 1329 - Set default mode to include execute bit. The {u,f,d}mask can be used 1330 to take it away if desired. This allows binaries to be executed from 1331 a mounted ntfs partition. 1332 13332.0.0 - New version number. Remove TNG from the name. Now in the kernel. 1334 1335 - Add kill_super, just keeping up with the vfs changes in the kernel. 1336 - Repeat some changes from tng-0.0.8 that somehow got lost on the way 1337 from the CVS import into BitKeeper. 1338 - Begin to implement proper handling of allocated_size vs 1339 initialized_size vs data_size (i.e. i_size). Done are 1340 mft.c::ntfs_mft_readpage(), aops.c::end_buffer_read_index_async(), 1341 and attrib.c::load_attribute_list(). 1342 - Lock the runlist in attrib.c::load_attribute_list() while using it. 1343 - Fix memory leak in ntfs_file_read_compressed_block() and generally 1344 clean up compress.c a little, removing some uncommented/unused debug 1345 code. 1346 - Tidy up dir.c a little bit. 1347 - Don't bother getting the runlist in inode.c::ntfs_read_inode(). 1348 - Merge mft.c::ntfs_mft_readpage() and aops.c::ntfs_index_readpage() 1349 creating aops.c::ntfs_mst_readpage(), improving the handling of 1350 holes and overflow in the process and implementing the correct 1351 equivalent of ntfs_file_get_block() in ntfs_mst_readpage() itself. 1352 I am aiming for correctness at the moment. Modularisation can come 1353 later. 1354 - Rename aops.c::end_buffer_read_index_async() to 1355 end_buffer_read_mst_async() and optimize the overflow checking and 1356 handling. 1357 - Use the host of the mftbmp address space mapping to hold the ntfs 1358 volume. This is needed so the async i/o completion handler can 1359 retrieve a pointer to the volume. Hopefully this will not cause 1360 problems elsewhere in the kernel... Otherwise will need to use a 1361 fake inode. 1362 - Complete implementation of proper handling of allocated_size vs 1363 initialized_size vs data_size (i.e. i_size) in whole driver. 1364 Basically aops.c is now completely rewritten. 1365 - Change NTFS driver name to just NTFS and set version number to 2.0.0 1366 to make a clear distinction from the old driver which is still on 1367 version 1.1.22. 1368 1369tng-0.0.8 - 08/03/2002 - Now using BitKeeper, http://linux-ntfs.bkbits.net/ 1370 1371 - Replace bdevname(sb->s_dev) with sb->s_id. 1372 - Remove now superfluous new-line characters in all callers of 1373 ntfs_debug(). 1374 - Apply kludge in ntfs_read_inode(), setting i_nlink to 1 for 1375 directories. Without this the "find" utility gets very upset which is 1376 fair enough as Linux/Unix do not support directory hard links. 1377 - Further runlist merging work. (Richard Russon) 1378 - Backwards compatibility for gcc-2.95. (Richard Russon) 1379 - Update to kernel 2.5.5-pre1 and rediff the now tiny patch. 1380 - Convert to new filesystem declaration using ->ntfs_get_sb() and 1381 replacing ntfs_read_super() with ntfs_fill_super(). 1382 - Set s_maxbytes to MAX_LFS_FILESIZE to avoid page cache page index 1383 overflow on 32-bit architectures. 1384 - Cleanup upcase loading code to use ntfs_(un)map_page(). 1385 - Disable/reenable preemtion in critical sections of compession engine. 1386 - Replace device size determination in ntfs_fill_super() with 1387 sb->s_bdev->bd_inode->i_size (in bytes) and remove now superfluous 1388 function super.c::get_nr_blocks(). 1389 - Implement a mount time option (show_inodes) allowing choice of which 1390 types of inode names readdir() returns and modify ntfs_filldir() 1391 accordingly. There are several parameters to show_inodes: 1392 system: system files 1393 win32: long file names (including POSIX file names) [DEFAULT] 1394 long: same as win32 1395 dos: short file names only (excluding POSIX file names) 1396 short: same as dos 1397 posix: same as both win32 and dos 1398 all: all file names 1399 Note that the options are additive, i.e. specifying: 1400 -o show_inodes=system,show_inodes=win32,show_inodes=dos 1401 is the same as specifying: 1402 -o show_inodes=all 1403 Note that the "posix" and "all" options will show all directory 1404 names, BUT the link count on each directory inode entry is set to 1, 1405 due to Linux not supporting directory hard links. This may well 1406 confuse some userspace applications, since the directory names will 1407 have the same inode numbers. Thus it is NOT advisable to use the 1408 "posix" or "all" options. We provide them only for completeness sake. 1409 - Add copies of allocated_size, initialized_size, and compressed_size to 1410 the ntfs inode structure and set them up in 1411 inode.c::ntfs_read_inode(). These reflect the unnamed data attribute 1412 for files and the index allocation attribute for directories. 1413 - Add copies of allocated_size and initialized_size to ntfs inode for 1414 $BITMAP attribute of large directories and set them up in 1415 inode.c::ntfs_read_inode(). 1416 - Add copies of allocated_size and initialized_size to ntfs volume for 1417 $BITMAP attribute of $MFT and set them up in 1418 super.c::load_system_files(). 1419 - Parse deprecated ntfs driver options (iocharset, show_sys_files, 1420 posix, and utf8) and tell user what the new options to use are. Note 1421 we still do support them but they will be removed with kernel 2.7.x. 1422 - Change all occurences of integer long long printf formatting to hex 1423 as printk() will not support long long integer format if/when the 1424 div64 patch goes into the kernel. 1425 - Make slab caches have stable names and change the names to what they 1426 were intended to be. These changes are required/made possible by the 1427 new slab cache name handling which removes the length limitation by 1428 requiring the caller of kmem_cache_create() to supply a stable name 1429 which is then referenced but not copied. 1430 - Rename run_list structure to run_list_element and create a new 1431 run_list structure containing a pointer to a run_list_element 1432 structure and a read/write semaphore. Adapt all users of runlists 1433 to new scheme and take and release the lock as needed. This fixes a 1434 nasty race as the run_list changes even when inodes are locked for 1435 reading and even when the inode isn't locked at all, so we really 1436 needed the serialization. We use a semaphore rather than a spinlock 1437 as memory allocations can sleep and doing everything GFP_ATOMIC 1438 would be silly. 1439 - Cleanup read_inode() removing all code checking for lowest_vcn != 0. 1440 This can never happen due to the nature of lookup_attr() and how we 1441 support attribute lists. If it did happen it would imply the inode 1442 being corrupt. 1443 - Check for lowest_vcn != 0 in ntfs_read_inode() and mark the inode as 1444 bad if found. 1445 - Update to 2.5.6-pre2 changes in struct address_space. 1446 - Use parent_ino() when accessing d_parent inode number in dir.c. 1447 - Import Sourceforge CVS repository into BitKeeper repository: 1448 http://linux-ntfs.bkbits.net/ntfs-tng-2.5 1449 - Update fs/Makefile, fs/Config.help, fs/Config.in, and 1450 Documentation/filesystems/ntfs.txt for NTFS TNG. 1451 - Create kernel configuration option controlling whether debugging 1452 is enabled or not. 1453 - Add the required export of end_buffer_io_sync() from the patches 1454 directory to the kernel code. 1455 - Update inode.c::ntfs_show_options() with show_inodes mount option. 1456 - Update errors mount option. 1457 1458tng-0.0.7 - 13/02/2002 - The driver is now feature complete for read-only! 1459 1460 - Cleanup mft.c and it's debug/error output in particular. Fix a minor 1461 bug in mapping of extent inodes. Update all the comments to fit all 1462 the recent code changes. 1463 - Modify vcn_to_lcn() to cope with entirely unmapped runlists. 1464 - Cleanups in compress.c, mostly comments and folding help. 1465 - Implement attrib.c::map_run_list() as a generic helper. 1466 - Make compress.c::ntfs_file_read_compressed_block() use map_run_list() 1467 thus making code shorter and enabling attribute list support. 1468 - Cleanup incorrect use of [su]64 with %L printf format specifier in 1469 all source files. Type casts to [unsigned] long long added to correct 1470 the mismatches (important for architectures which have long long not 1471 being 64 bits). 1472 - Merge async io completion handlers for directory indexes and $MFT 1473 data into one by setting the index_block_size{_bits} of the ntfs 1474 inode for $MFT to the mft_record_size{_bits} of the ntfs_volume. 1475 - Cleanup aops.c, update comments. 1476 - Make ntfs_file_get_block() use map_run_list() so all files now 1477 support attribute lists. 1478 - Make ntfs_dir_readpage() almost verbatim copy of 1479 block_read_full_page() by using ntfs_file_get_block() with only real 1480 difference being the use of our own async io completion handler 1481 rather than the default one, thus reducing the amount of code and 1482 automatically enabling attribute list support for directory indices. 1483 - Fix bug in load_attribute_list() - forgot to call brelse in error 1484 code path. 1485 - Change parameters to find_attr() and lookup_attr(). We no longer 1486 pass in the upcase table and its length. These can be gotten from 1487 ctx->ntfs_ino->vol->upcase{_len}. Update all callers. 1488 - Cleanups in attrib.c. 1489 - Implement merging of runlists, attrib.c::merge_run_lists() and its 1490 helpers. (Richard Russon) 1491 - Attribute lists part 2, attribute extents and multi part runlists: 1492 enable proper support for LCN_RL_NOT_MAPPED and automatic mapping of 1493 further runlist parts via attrib.c::map_run_list(). 1494 - Tiny endianness bug fix in decompress_mapping_pairs(). 1495 1496tng-0.0.6 - Encrypted directories, bug fixes, cleanups, debugging enhancements. 1497 1498 - Enable encrypted directories. (Their index root is marked encrypted 1499 to indicate that new files in that directory should be created 1500 encrypted.) 1501 - Fix bug in NInoBmpNonResident() macro. (Cut and paste error.) 1502 - Enable $Extend system directory. Most (if not all) extended system 1503 files do not have unnamed data attributes so ntfs_read_inode() had to 1504 special case them but that is ok, as the special casing recovery 1505 happens inside an error code path so there is zero slow down in the 1506 normal fast path. The special casing is done by introducing a new 1507 function inode.c::ntfs_is_extended_system_file() which checks if any 1508 of the hard links in the inode point to $Extend as being their parent 1509 directory and if they do we assume this is an extended system file. 1510 - Create a sysctl/proc interface to allow {dis,en}abling of debug output 1511 when compiled with -DDEBUG. Default is debug messages to be disabled. 1512 To enable them, one writes a non-zero value to /proc/sys/fs/ntfs-debug 1513 (if /proc is enabled) or uses sysctl(2) to effect the same (if sysctl 1514 interface is enabled). Inspired by old ntfs driver. 1515 - Add debug_msgs insmod/kernel boot parameter to set whether debug 1516 messages are {dis,en}abled. This is useful to enable debug messages 1517 during ntfs initialization and is the only way to activate debugging 1518 when the sysctl interface is not enabled. 1519 - Cleanup debug output in various places. 1520 - Remove all dollar signs ($) from the source (except comments) to 1521 enable compilation on architectures whose gcc compiler does not 1522 support dollar signs in the names of variables/constants. Attribute 1523 types now start with AT_ instead of $ and $I30 is now just I30. 1524 - Cleanup ntfs_lookup() and add consistency check of sequence numbers. 1525 - Load complete runlist for $MFT/$BITMAP during mount and cleanup 1526 access functions. This means we now cope with $MFT/$BITMAP being 1527 spread accross several mft records. 1528 - Disable modification of mft_zone_multiplier on remount. We can always 1529 reenable this later on if we really want to, but we will need to make 1530 sure we readjust the mft_zone size / layout accordingly. 1531 1532tng-0.0.5 - Modernize for 2.5.x and further in line-ing with Al Viro's comments. 1533 1534 - Use sb_set_blocksize() instead of set_blocksize() and verify the 1535 return value. 1536 - Use sb_bread() instead of bread() throughout. 1537 - Add index_vcn_size{_bits} to ntfs_inode structure to store the size 1538 of a directory index block vcn. Apply resulting simplifications in 1539 dir.c everywhere. 1540 - Fix a small bug somewhere (but forgot what it was). 1541 - Change ntfs_{debug,error,warning} to enable gcc to do type checking 1542 on the printf-format parameter list and fix bugs reported by gcc 1543 as a result. (Richard Russon) 1544 - Move inode allocation strategy to Al's new stuff but maintain the 1545 divorce of ntfs_inode from struct inode. To achieve this we have two 1546 separate slab caches, one for big ntfs inodes containing a struct 1547 inode and pure ntfs inodes and at the same time fix some faulty 1548 error code paths in ntfs_read_inode(). 1549 - Show mount options in proc (inode.c::ntfs_show_options()). 1550 1551tng-0.0.4 - Big changes, getting in line with Al Viro's comments. 1552 1553 - Modified (un)map_mft_record functions to be common for read and write 1554 case. To specify which is which, added extra parameter at front of 1555 parameter list. Pass either READ or WRITE to this, each has the 1556 obvious meaning. 1557 - General cleanups to allow for easier folding in vi. 1558 - attrib.c::decompress_mapping_pairs() now accepts the old runlist 1559 argument, and invokes attrib.c::merge_run_lists() to merge the old 1560 and the new runlists. 1561 - Removed attrib.c::find_first_attr(). 1562 - Implemented loading of attribute list and complete runlist for $MFT. 1563 This means we now cope with $MFT being spread across several mft 1564 records. 1565 - Adapt to 2.5.2-pre9 and the changed create_empty_buffers() syntax. 1566 - Adapt major/minor/kdev_t/[bk]devname stuff to new 2.5.x kernels. 1567 - Make ntfs_volume be allocated via kmalloc() instead of using a slab 1568 cache. There are too little ntfs_volume structures at any one time 1569 to justify a private slab cache. 1570 - Fix bogus kmap() use in async io completion. Now use kmap_atomic(). 1571 Use KM_BIO_IRQ on advice from IRC/kernel... 1572 - Use ntfs_map_page() in map_mft_record() and create ->readpage method 1573 for reading $MFT (ntfs_mft_readpage). In the process create dedicated 1574 address space operations (ntfs_mft_aops) for $MFT inode mapping. Also 1575 removed the now superfluous exports from the kernel core patch. 1576 - Fix a bug where kfree() was used insted of ntfs_free(). 1577 - Change map_mft_record() to take ntfs_inode as argument instead of 1578 vfs inode. Dito for unmap_mft_record(). Adapt all callers. 1579 - Add pointer to ntfs_volume to ntfs_inode. 1580 - Add mft record number and sequence number to ntfs_inode. Stop using 1581 i_ino and i_generation for in-driver purposes. 1582 - Implement attrib.c::merge_run_lists(). (Richard Russon) 1583 - Remove use of proper inodes by extent inodes. Move i_ino and 1584 i_generation to ntfs_inode to do this. Apply simplifications that 1585 result and remove iget_no_wait(), etc. 1586 - Pass ntfs_inode everywhere in the driver (used to be struct inode). 1587 - Add reference counting in ntfs_inode for the ntfs inode itself and 1588 for the mapped mft record. 1589 - Extend mft record mapping so we can (un)map extent mft records (new 1590 functions (un)map_extent_mft_record), and so mappings are reference 1591 counted and don't have to happen twice if already mapped - just ref 1592 count increases. 1593 - Add -o iocharset as alias to -o nls for backwards compatibility. 1594 - The latest core patch is now tiny. In fact just a single additional 1595 export is necessary over the base kernel. 1596 1597tng-0.0.3 - Cleanups, enhancements, bug fixes. 1598 1599 - Work on attrib.c::decompress_mapping_pairs() to detect base extents 1600 and setup the runlist appropriately using knowledge provided by the 1601 sizes in the base attribute record. 1602 - Balance the get_/put_attr_search_ctx() calls so we don't leak memory 1603 any more. 1604 - Introduce ntfs_malloc_nofs() and ntfs_free() to allocate/free a single 1605 page or use vmalloc depending on the amount of memory requested. 1606 - Cleanup error output. The __FUNCTION__ "(): " is now added 1607 automatically. Introduced a new header file debug.h to support this 1608 and also moved ntfs_debug() function into it. 1609 - Make reading of compressed files more intelligent and especially get 1610 rid of the vmalloc_nofs() from readpage(). This now uses per CPU 1611 buffers (allocated at first mount with cluster size <= 4kiB and 1612 deallocated on last umount with cluster size <= 4kiB), and 1613 asynchronous io for the compressed data using a list of buffer heads. 1614 Er, we use synchronous io as async io only works on whole pages 1615 covered by buffers and not on individual buffer heads... 1616 - Bug fix for reading compressed files with sparse compression blocks. 1617 1618tng-0.0.2 - Now handles larger/fragmented/compressed volumes/files/dirs. 1619 1620 - Fixed handling of directories when cluster size exceeds index block 1621 size. 1622 - Hide DOS only name space directory entries from readdir() but allow 1623 them in lookup(). This should fix the problem that Linux doesn't 1624 support directory hard links, while still allowing access to entries 1625 via their short file name. This also has the benefit of mimicking 1626 what Windows users are used to, so it is the ideal solution. 1627 - Implemented sync_page everywhere so no more hangs in D state when 1628 waiting for a page. 1629 - Stop using bforget() in favour of brelse(). 1630 - Stop locking buffers unnecessarily. 1631 - Implemented compressed files (inode->mapping contains uncompressed 1632 data, raw compressed data is currently bread() into a vmalloc()ed 1633 memory buffer). 1634 - Enable compressed directories. (Their index root is marked compressed 1635 to indicate that new files in that directory should be created 1636 compressed.) 1637 - Use vsnprintf rather than vsprintf in the ntfs_error and ntfs_warning 1638 functions. (Thanks to Will Dyson for pointing this out.) 1639 - Moved the ntfs_inode and ntfs_volume (the former ntfs_inode_info and 1640 ntfs_sb_info) out of the common inode and super_block structures and 1641 started using the generic_ip and generic_sbp pointers instead. This 1642 makes ntfs entirely private with respect to the kernel tree. 1643 - Detect compiler version and abort with error message if gcc less than 1644 2.96 is used. 1645 - Fix bug in name comparison function in unistr.c. 1646 - Implement attribute lists part 1, the infrastructure: search contexts 1647 and operations, find_external_attr(), lookup_attr()) and make the 1648 code use the infrastructure. 1649 - Fix stupid buffer overflow bug that became apparent on larger run 1650 list containing attributes. 1651 - Fix bugs in readdir() that became apparent on larger directories. 1652 1653 The driver is now really useful and survives the test 1654 find . -type f -exec md5sum "{}" \; 1655 without any error messages on a over 1GiB sized partition with >16k 1656 files on it, including compressed files and directories and many files 1657 and directories with attribute lists. 1658 1659tng-0.0.1 - The first useful version. 1660 1661 - Added ntfs_lookup(). 1662 - Added default upcase generation and handling. 1663 - Added compile options to be shown on module init. 1664 - Many bug fixes that were "hidden" before. 1665 - Update to latest kernel. 1666 - Added ntfs_readdir(). 1667 - Added file operations for mmap(), read(), open() and llseek(). We just 1668 use the generic ones. The whole point of going through implementing 1669 readpage() methods and where possible get_block() call backs is that 1670 this allows us to make use of the generic high level methods provided 1671 by the kernel. 1672 1673 The driver is now actually useful! Yey. (-: It undoubtedly has got bugs 1674 though and it doesn't implement accesssing compressed files yet. Also, 1675 accessing files with attribute list attributes is not implemented yet 1676 either. But for small or simple filesystems it should work and allow 1677 you to list directories, use stat on directory entries and the file 1678 system, open, read, mmap and llseek around in files. A big mile stone 1679 has been reached! 1680 1681tng-0.0.0 - Initial version tag. 1682 1683 Initial driver implementation. The driver can mount and umount simple 1684 NTFS filesystems (i.e. ones without attribute lists in the system 1685 files). If the mount fails there might be problems in the error handling 1686 code paths, so be warned. Otherwise it seems to be loading the system 1687 files nicely and the mft record read mapping/unmapping seems to be 1688 working nicely, too. Proof of inode metadata in the page cache and non- 1689 resident file unnamed stream data in the page cache concepts is thus 1690 complete. 1691