4 - orphans are moved into the PENDING directory for possible recovery
5 - replayed opens now open by fid for orphan/rename safety (1042)
6 - last close of an orphan inode generates a transno (683)
7 - chdir() and mount() now pin the directory entry (1020)
8 - avoid CERROR in normal ll_setattr_raw() error case (1500)
9 - discard very old requests without processing them (1502)
10 - remove some common, well-understood CERRORs (1505)
11 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
12 - clear "grant" flags in OST replies until OST grant code lands (1644)
13 - fix read performance by not clobbering i_blksize on client (1598)
14 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
15 - make LRU size a /proc tunable, clears locks when reduced (707)
16 - fix some lprocfs rot that prevented ptlbd from loading (1732)
17 - server locks take references on exports now (1558)
18 - build fixes for 2.4.20-rh trees (1663)
19 - return an error from lov_create if all OSCs are inactive (1751)
20 - fix import levels when a reconnect happens without a timeout (1597)
21 - exit early from mds_open if we get a lookup error (1749)
22 - partial page read at EOF wouldn't wait for disk before sending (1642)
23 - avoid NULL deref in obdfilter when reading page past EOF (1592)
24 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
25 - fix LBUG in ll_it_open_error with rc = -2 (1861)
26 - write/truncate lock inversion (1639)
27 - Don't auto-load obdclass, portals modules during cleanup (1495)
28 - fix timestamps from jumping to "now" (1763)
29 - extra journal assertions (1648)
30 - add an extra multiunlink test (1771)
31 - fix read_record/write_record API (1776)
32 - fix leak of offset_extent, possible incorrect i_size later (1772)
33 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
34 - Debugging for the kqswnal_get_idle_tx problems (1820)
35 - Allow recovery to be attempted multiple times (1536)
36 - Write out MDS last_rcvd file after it is first created (1600)
37 - Fix tx_descriptor leak in failed transmit situations (1827)
38 - ext3 journaling fixes for assertion failure after IO error (1871)
39 - class_export_put() on freed export after completion AST error (1896)
40 - Fix revalidate looping in VFS (1322)
41 - Don't access a freed export during MDS_REINT timeout (1521)
42 - Add open-unlink recovery support on the MDS (1673,1764)
43 - Return an error if no MDS data was read from last_rcvd (1946)
44 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
45 - Don't setup a disk device that doesn't match exported UUID (317)
46 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
47 - avoid committing NULL handle in force close
48 - local.sh is now a one-stripe LOV configuration
49 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
50 - fix ext3 htree duplicate directory entry corruption (1516)
51 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
52 - update to new LOV EA format (2097)
53 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
55 2003-06-15 Phil Schwan <phil@clusterfs.com>
58 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
59 - per-import/export recovery handling (958, 931, 959)
60 - multiple last-rcvd slots, for serving multiple FSes (949)
61 - connections are again shared between multiple imp/exports (963, 964)
62 - "umount -f" would hang if any requests needed to be sent (393, 978)
63 - avoid pinning large req buffer by copying for queued messages (989)
64 - add "uuid" to "lctl device" command to help upcalls (991)
65 - "open" RPCs with transnos would confuse recovery counters (1037)
66 - do proper endian conversion of all wire messages (288, 340, 891)
67 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
68 - call waiting locks callback from LDLM recovery thread (1127, 1151)
69 - fix ptlrpc_connection leak in target_handle_connect (1174)
70 - fix import refcounting bug in OST and MDS cleanup (1134)
71 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
72 - fix very unlikely obd_types race condition (501)
73 - remove osc_open hack for echo_client (1187)
74 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
75 - a failure in read_inode2 leads to deadlock (1139)
76 - cancel ack-locks as soon as transaction is committed (1072)
77 - fix major leaks and crashes in the bulk I/O path (937, 1057)
78 - make sure to commitrw after any preprw to avoid deadlock (1162)
79 - failing to execute a file in a lustre FS would lock inode (1203)
80 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
81 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
82 - obd_uuid2tgt would walk off the end of the list (1255)
83 - on IA64 the osc would give portals incorrect bulk size (1258)
84 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
85 - fix lock inversion caused by new llite matching code (1282)
86 - limit the number of dirty pages on a client to 10MB (1286)
87 - timed out locks were not being corrected cancelled (1289)
88 - fix O_DIRECT above 4GB on IA-32 (1292)
89 * major user-visible changes
90 - fail out/fail over policy now controlled by the upcall (993)
92 - add OBD_PING to check server availability and failure (954)
93 - lustre messages are now sent in sending host order (288, 340, 891)
94 - add eadatalen to MDS getattr reply (340)
95 - OST read replies may contain second buffer, with per-page status (593)
97 2003-03-11 Phil Schwan <phil@clusterfs.com>
100 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
101 - failed open()s could cause deadlock; fixed (867, 869)
102 - stop cancelling OST locks when files are closed (481)
103 - overlapping XID spaces caused network corruption (851, 853)
104 - fix unsafe fsfilt counter arithmetic; change to atomic_t
105 - setattr_raw added, to do single-RPC, server-side setattrs
106 - lmc/lconf syntax change for OST UUIDs
107 - fix crashy race condition between ptlrpc_free_req and osc_close
108 - don't use request in mdc_enqueue if we hit a timeout (889)
109 - don't set the inode i_size for regular files from the MDS (896)
110 - handle out of order completion AST (842)
111 - don't LBUG if a lock request times out after receiving AST (913)
112 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
113 - if a bad lock AST arrives, send an error instead of dropping entirely
114 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
115 - fix leak in bulk IO when only partially completed (899, 900, 926)
116 - fix O_DIRECT for ia64 (55)
117 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
118 - C-z after timeout could hang a process forever; fixed (977)
120 - client-side I/O cache (678, 924, 929, 941, 970)
122 - READPAGE and SETATTRs which don't take server-side locks get
125 2003-02-11 Phil Schwan <phil@clusterfs.com>
128 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
129 - Fix fsfilt_extN_readpage() to read a full page of directory
130 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
131 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
132 - Fix imbalanced LOV object allocation and out-of-bound access (469)
133 - Most intent operations were removed, in favour of a new RPC mode
134 that does a single RPC to the server and bypasses most of the VFS
135 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
136 - Aggressively cancel local locks on DLM servers
137 - mds_reint_unlink sends EA to the client if it's the last nlink.
138 client uses that EA to unlink OST objects.
139 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
140 - recursive symlinks were fixed (439)
141 - fixed NULL deref in DEBUG_REQ
142 - filter_update_lastobjid no longer calls sync, which annoyed extN
143 - fixed multi-client small-writes to a single file problem (445)
144 - fixed mtime updates during file writes (607)
145 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
146 - fixed bug in obd_brw_read/write() (under guise of testing 367)
147 - fixed Linux OST size reporting problem (444, 656)
148 - OST now updates object mtime with writes or setattr (607, 619)
149 - client verifies file size before zeroing page past EOF (445)
150 - OST now writes last allocated objid to disk with allocation (108)
151 - LOV on echo now works (409)
153 - mds_reint_unlink sends a new buffer, with the EA included. this
154 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
155 set if a regular file was being unlinked, and it was the last link
156 - use PtlGet from the target for bulk writes (315)
157 - OST now updates object mtime with writes or setattr (607, 619)
158 - LDLM now has a grant-time callback to revalidate locked items, if
160 - Many MDS operations were reorganized to combat race conditions
162 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
163 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
165 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
168 - Fully reactivate OST imports after reconnection (512, others)
169 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
170 - More graceful error handling for truncating on dead OST (515)
171 - Don't error out unless we're actually accessing dead stripes (474)
172 - Fix garbage sizes when stripes are missing (410)
173 - LRU counters were broken, causing constant lock purge (433, 432)
174 - garbage on read from stripes with failed OSTs (441)
175 - mark OSCs as active before reconnecting during recovery (438)
176 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
177 - lfind did not preserve OST order in output (443)
178 - symlinks cause hung clients, incorrect data (439)
179 - stop dereferencing request after dropping refcount (457)
180 - don't LASSERT(spin_is_locked) on non-SMP (455)
181 - fixes for many rename() bugs
182 - fstat didn't correctly synchronize attributes (399)
183 - server must handle lock cancellation during blocking AST prep (487)
184 - bulk descriptors were free()d too soon (511)
185 - fix paths in lconf, which would load incorrect modules (451, 507)
186 - fix confusing lconf 'host not found' error message (386)
187 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
188 - fix race condition in mdc_blocking_ast() for inode access (526)
189 - fix lov_unpackmd() unpacking wrong number of stripes (537)
190 - fix lov_set_osc_active() marking wrong OSC inactive (440)
191 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
192 - fix multiple writes of stripe MD to MDS (358, maybe 519)
193 - fix lstripe in several ways (kernel side) (527)
194 - fix request leak in ldlm_cli_enqueue (262)
195 - incorrect OSC was marked inactive after OST failure
196 - call mds_fs_cleanup before unmounting filesystem (524)
197 - fix races between taking ns_lock and ldlm_lock_change_resource
198 - fix races updating LOV export open file list
199 - fix lov_enqueue error path, avoid decref-ing bad lock handle
200 - fix recovery NULL deref in ldlm_cli_cancel_unused
201 - fix some DLM races by using new hash table for lock handles (419)
202 - permit the client to specify desired inodes, at replay
203 - duplicate requests when we queue them for replay reintegration
204 - fix last_rcvd offset calculation
205 - sync after each recovered transaction, so we always make progress
206 - never, not always, ERESTART requests without transnos
207 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
209 - skip replay if the MDS says that the client is already connected
210 - don't check for a recovery-enabled export to match lctl's UUID
211 - don't INC_USE_COUNT for phantom exports
212 - don't crash when cleaning up phantom exports (567)
213 - don't double-finish or set replay data for errored mdc_open requests
214 - abort requests when they time out, so we don't get old replies
215 - send/receive replies for AST messages again
216 - if the client says that it doesn't have the lock, cancel it on the
218 - if we timeout during I/O, don't try to cancel an in-use lock; instead
219 mark it as destroyed, it will all work out when decref is called
220 - fix module use counts (22, 581)
222 - ASTs now expect a reply (server cancels lock on error reply)
224 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
227 - fix many simultaneous client startup (392)
228 - fix dentry->d_it clobbering
229 - credentials weren't being shipped for readdir/getattr operations
230 - remove invalid assertions triggered during some concurrent MD
232 - proper Lustre versions added (336, 389)
233 - fix memory leak for create error case (398)
234 - fix LOV locking bug that would get cli/srv out of sync
235 - fix echo client over LOV (409)
236 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
237 - fix double-O_EXCL intent crash (424)
238 - avoid sending multiple lock CANCELs (352)
240 - MDS can do multi-client recovery (modulo bugs in new code)
242 - many updates, edits, cleanups
244 2002-11-18 Phil Schwan <phil@clusterfs.com>
247 - fix null d_it dereference (346)
248 - fix full OST/dbench hang (333)
249 - fix permission problem with file removal (286)
250 - fix removal of OSCs from LOV when they fail
251 - fix NULL deref during bulk timeout (214)
252 - fix problems related to multiple filesystems on one MDS (241)
253 - fixed serious subtle metadata locking bugs
254 - free locks on clients when inodes are removed due to memory
256 - fix inode pointer in lock data (285)
257 - partial support for multiple MDS on a single host (241)
258 - data locks weren't cancelled at clear_inode time (290, 311)
259 - intent locks could lead to unbounded lock growth (205)
260 - added a maximum lock count, an LRU list, and a flusher
261 - fix multiple rename (365)
262 - properly abstracted the echo client
263 - OSC locked 1 byte too many; fixed
264 - rewrote brw callback code:
265 - fixed recovery bugs related to LOVs (306)
266 - fixed too-many-pages-in-one-write crash (191)
267 - fixed (again) crash in sync_io_timeout (214)
268 - probably fixed callback-related race (385)
270 - Add capability to MDS protocol
271 - LDLM cancellations and callbacks on different portals
273 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
276 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
277 - defer open object destruction to close time (601981, 138)
278 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
279 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
280 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
281 - don't call obd_setattr in ll_file_release for destroyed objects
282 * protocol change to lustre_msg: move |version| and add |flags|
283 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
284 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
285 * added replay of create, unlink, link and rename operations during
286 MDS failover; recovery should be much more robust now
287 * remove failed OSCs from LOVs (only lov_create uses this so far)
288 * the lustre-HOWTO was brought (more) up to date (582544)
290 2002-10-23 Phil Schwan <phil@clusterfs.com>
293 - in-use dentries weren't being reused properly (617851)
294 - prevent multiple LDLM setup (599178)
295 - fix LOV size calculations for truncate (617853)
296 - fix client handling of MDS intent errors (POSIX)
297 - fix permission bug in lovstripe.c test (624321)
298 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
299 - truncate past end of file could corrupt data
300 - proper cleanup after timeouts, crashes, etc (592524, 550815)
301 - a race in recovery could return ETIMEDOUT to apps (623947)
302 - building outside the source directory was fixed
303 * the lustre-HOWTO was brought (more) up to date (582544)
304 * major progress was made on recovery functionality
306 2002-10-10 Phil Schwan <phil@clusterfs.com>
309 - recovery deadlock fix
310 - rm -rf causes LBUG fix (617817)
311 - file open by multiple tasks fix (618962)
312 - directory permissions bugs (602707 and 620007)
313 - journal_stop fixed with locking (611313)
314 - O_APPEND failures resolved (618273, perhaps 614459)
315 - lconf PATH fix (619770)
316 - IA64 build fix (621450)
317 - RPC buffer sizes scale with amount of memory
319 2002-10-01 Phil Schwan <phil@clusterfs.com>
322 - locks would be cancelled without throwing away data pages,
323 resulting in inconsistent data (605627)
324 - inode attributes were not always being refreshed (605627, 612449)
325 - lconf now continues to cleanup after lctl reports an error
326 - MDS now enforces user permissions (602707)
327 - lprocfs cleanup fixed, but not yet enabled (614157)
328 - fixed infinite server hang, should a client not respond to an AST
329 - avoid going into recovery if user calls readlink() with a buffer
330 that's too small (613941)
331 - AST RPCs no longer require replies (614867) -- this may be changed
332 - don't crash server if client sends an IOV that's too big (611336)
333 - fixed lock conversion deadlock (611892)
334 - fixed the following of symlinks (614622)
335 * recovery: the server can remove locks from a client that dies, other
336 clients can make progress
337 * more extN patch fixes
338 * compile-time configurable ptlrpc buffer allocations
340 - collaborative read cache document
341 - Lustre Lite Performance CDR document-in-progress
343 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
346 - fix typo in patch-2.4.18
348 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
351 - clear ptlrpc request each time in handle_incoming_request()
352 - unlink of files now destroys the object on the OST
354 2002-09-19 Peter Braam <braam@clusterfs.com>
356 * add hard link support
357 * change obdfile creation method
358 * kernel patch changed
360 2002-09-19 Peter Braam <braam@clusterfs.com>
363 - stack overflow bug in extN fixed
365 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
367 * documentation updates
368 - add man pages for config tools
369 - update tests/README to describe testing with new config tools
370 - finish metadata API descriptions
371 * bug fixes and cleanups
372 - statfs workaround for 16TB limit
373 - LOV stripe allocation improved, can stripe on subset of OSTs
374 - LOV file size/IO offset was wrong for files > 4GB in size
375 - object EA data was being dropped, caused files to be unreadable
376 - memory overflow with non-LOV OST caused memory corruption
377 - fixed regression tests to work with new config tools, obdfilter
378 - fixed bug when directory size became larger than 1 block
379 - fixed bug (for single client case) when PWD was deleted
380 - invalidate local directory pages when doing intent-based ops
381 - avoid LDLM oops when lock callback contained bad data
383 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
385 * documentation updates
386 * bug fixes and cleanups
387 - configuration tools
390 - 64-bit compile warnings
391 - 64-bit internal statfs data
393 * test_brw on persistent OST devices
397 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
399 * documentation updates
400 * bug fixes and cleanups
401 * configuration tools
403 2002-08-30 Peter J. Braam <braam@clusterfs.com>
406 * many small fixes to 0_5_4
407 * io/network handling
408 * thinkos in MDS operations
410 2002-08-24 Peter J. Braam <braam@clusterfs.com>
413 * crucial basic fixes to 0.5.3
414 * IOR, Iozone work over Elan
417 2002-08-07 Phil Schwan <phil@clusterfs.com>
418 * version 0_5_3, our first alpha
419 * we use the new Portals iovs
420 * documentation updates
421 * bug fixes and cleanups
422 * small changes in the DLM wire protocol
424 2002-07-25 Peter J. Braam <braam@clusterfs.com>
425 * version 0_5_1 with some initial stability,
426 * locking on MD and file I/O.
427 * documentation updates
428 * several bug fixes since 0.5.0
429 * small changes in wire protocol
431 2002-07-18 Phil Schwan <phil@clusterfs.com>
433 * delivered as Lustre Light Alpha
434 * fixed a crash after handling invalid MDS requests
435 * fixed directory pages for architectures with non-4k pages sizes
437 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
438 * release version v0_4_4
439 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
441 * Fixes a number of interruption problems with OST operations.
442 * Update documentation for portals header changes
443 * Move all wire protocol structs/defines to lustre_idl.h
444 * Fixes symlink length bug.
445 * Add tcpdump to repository.
447 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
448 * release version v0_4_3
449 * Fixes statfs for inodes on extN.
450 * Fixes bug in runtests which would delete /etc/hosts.
451 * Use 64-bit object IDs wherever possible (not into VFS though)
452 Remove ost_get_info, which is unused by lustre, and out of date.
454 2002-07-03 Peter Braam <braam@clusterfs.com>
455 * release version v0_4_2 Fixes a lookup error (type not passed)
456 * move forward to head of Portals
457 * move forward to latest Lustre kernel
459 2002-06-25 Peter Braam <braam@clusterfs.com>
460 * release version v0_4_1. Hopefully stable on single node use.