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)
11 2003-06-15 Phil Schwan <phil@clusterfs.com>
14 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
15 - per-import/export recovery handling (958, 931, 959)
16 - multiple last-rcvd slots, for serving multiple FSes (949)
17 - connections are again shared between multiple imp/exports (963, 964)
18 - "umount -f" would hang if any requests needed to be sent (393, 978)
19 - avoid pinning large req buffer by copying for queued messages (989)
20 - add "uuid" to "lctl device" command to help upcalls (991)
21 - "open" RPCs with transnos would confuse recovery counters (1037)
22 - do proper endian conversion of all wire messages (288, 340, 891)
23 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
24 - call waiting locks callback from LDLM recovery thread (1127, 1151)
25 - fix ptlrpc_connection leak in target_handle_connect (1174)
26 - fix import refcounting bug in OST and MDS cleanup (1134)
27 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
28 - fix very unlikely obd_types race condition (501)
29 - remove osc_open hack for echo_client (1187)
30 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
31 - a failure in read_inode2 leads to deadlock (1139)
32 - cancel ack-locks as soon as transaction is committed (1072)
33 - fix major leaks and crashes in the bulk I/O path (937, 1057)
34 - make sure to commitrw after any preprw to avoid deadlock (1162)
35 - failing to execute a file in a lustre FS would lock inode (1203)
36 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
37 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
38 - obd_uuid2tgt would walk off the end of the list (1255)
39 - on IA64 the osc would give portals incorrect bulk size (1258)
40 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
41 - fix lock inversion caused by new llite matching code (1282)
42 - limit the number of dirty pages on a client to 10MB (1286)
43 - timed out locks were not being corrected cancelled (1289)
44 - fix O_DIRECT above 4GB on IA-32 (1292)
45 * major user-visible changes
46 - fail out/fail over policy now controlled by the upcall (993)
48 - add OBD_PING to check server availability and failure (954)
49 - lustre messages are now sent in sending host order (288, 340, 891)
50 - add eadatalen to MDS getattr reply (340)
51 - OST read replies may contain second buffer, with per-page status (593)
53 2003-03-11 Phil Schwan <phil@clusterfs.com>
56 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
57 - failed open()s could cause deadlock; fixed (867, 869)
58 - stop cancelling OST locks when files are closed (481)
59 - overlapping XID spaces caused network corruption (851, 853)
60 - fix unsafe fsfilt counter arithmetic; change to atomic_t
61 - setattr_raw added, to do single-RPC, server-side setattrs
62 - lmc/lconf syntax change for OST UUIDs
63 - fix crashy race condition between ptlrpc_free_req and osc_close
64 - don't use request in mdc_enqueue if we hit a timeout (889)
65 - don't set the inode i_size for regular files from the MDS (896)
66 - handle out of order completion AST (842)
67 - don't LBUG if a lock request times out after receiving AST (913)
68 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
69 - if a bad lock AST arrives, send an error instead of dropping entirely
70 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
71 - fix leak in bulk IO when only partially completed (899, 900, 926)
72 - fix O_DIRECT for ia64 (55)
73 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
74 - C-z after timeout could hang a process forever; fixed (977)
76 - client-side I/O cache (678, 924, 929, 941, 970)
78 - READPAGE and SETATTRs which don't take server-side locks get
81 2003-02-11 Phil Schwan <phil@clusterfs.com>
84 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
85 - Fix fsfilt_extN_readpage() to read a full page of directory
86 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
87 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
88 - Fix imbalanced LOV object allocation and out-of-bound access (469)
89 - Most intent operations were removed, in favour of a new RPC mode
90 that does a single RPC to the server and bypasses most of the VFS
91 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
92 - Aggressively cancel local locks on DLM servers
93 - mds_reint_unlink sends EA to the client if it's the last nlink.
94 client uses that EA to unlink OST objects.
95 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
96 - recursive symlinks were fixed (439)
97 - fixed NULL deref in DEBUG_REQ
98 - filter_update_lastobjid no longer calls sync, which annoyed extN
99 - fixed multi-client small-writes to a single file problem (445)
100 - fixed mtime updates during file writes (607)
101 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
102 - fixed bug in obd_brw_read/write() (under guise of testing 367)
103 - fixed Linux OST size reporting problem (444, 656)
104 - OST now updates object mtime with writes or setattr (607, 619)
105 - client verifies file size before zeroing page past EOF (445)
106 - OST now writes last allocated objid to disk with allocation (108)
107 - LOV on echo now works (409)
109 - mds_reint_unlink sends a new buffer, with the EA included. this
110 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
111 set if a regular file was being unlinked, and it was the last link
112 - use PtlGet from the target for bulk writes (315)
113 - OST now updates object mtime with writes or setattr (607, 619)
114 - LDLM now has a grant-time callback to revalidate locked items, if
116 - Many MDS operations were reorganized to combat race conditions
118 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
119 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
121 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
124 - Fully reactivate OST imports after reconnection (512, others)
125 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
126 - More graceful error handling for truncating on dead OST (515)
127 - Don't error out unless we're actually accessing dead stripes (474)
128 - Fix garbage sizes when stripes are missing (410)
129 - LRU counters were broken, causing constant lock purge (433, 432)
130 - garbage on read from stripes with failed OSTs (441)
131 - mark OSCs as active before reconnecting during recovery (438)
132 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
133 - lfind did not preserve OST order in output (443)
134 - symlinks cause hung clients, incorrect data (439)
135 - stop dereferencing request after dropping refcount (457)
136 - don't LASSERT(spin_is_locked) on non-SMP (455)
137 - fixes for many rename() bugs
138 - fstat didn't correctly synchronize attributes (399)
139 - server must handle lock cancellation during blocking AST prep (487)
140 - bulk descriptors were free()d too soon (511)
141 - fix paths in lconf, which would load incorrect modules (451, 507)
142 - fix confusing lconf 'host not found' error message (386)
143 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
144 - fix race condition in mdc_blocking_ast() for inode access (526)
145 - fix lov_unpackmd() unpacking wrong number of stripes (537)
146 - fix lov_set_osc_active() marking wrong OSC inactive (440)
147 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
148 - fix multiple writes of stripe MD to MDS (358, maybe 519)
149 - fix lstripe in several ways (kernel side) (527)
150 - fix request leak in ldlm_cli_enqueue (262)
151 - incorrect OSC was marked inactive after OST failure
152 - call mds_fs_cleanup before unmounting filesystem (524)
153 - fix races between taking ns_lock and ldlm_lock_change_resource
154 - fix races updating LOV export open file list
155 - fix lov_enqueue error path, avoid decref-ing bad lock handle
156 - fix recovery NULL deref in ldlm_cli_cancel_unused
157 - fix some DLM races by using new hash table for lock handles (419)
158 - permit the client to specify desired inodes, at replay
159 - duplicate requests when we queue them for replay reintegration
160 - fix last_rcvd offset calculation
161 - sync after each recovered transaction, so we always make progress
162 - never, not always, ERESTART requests without transnos
163 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
165 - skip replay if the MDS says that the client is already connected
166 - don't check for a recovery-enabled export to match lctl's UUID
167 - don't INC_USE_COUNT for phantom exports
168 - don't crash when cleaning up phantom exports (567)
169 - don't double-finish or set replay data for errored mdc_open requests
170 - abort requests when they time out, so we don't get old replies
171 - send/receive replies for AST messages again
172 - if the client says that it doesn't have the lock, cancel it on the
174 - if we timeout during I/O, don't try to cancel an in-use lock; instead
175 mark it as destroyed, it will all work out when decref is called
176 - fix module use counts (22, 581)
178 - ASTs now expect a reply (server cancels lock on error reply)
180 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
183 - fix many simultaneous client startup (392)
184 - fix dentry->d_it clobbering
185 - credentials weren't being shipped for readdir/getattr operations
186 - remove invalid assertions triggered during some concurrent MD
188 - proper Lustre versions added (336, 389)
189 - fix memory leak for create error case (398)
190 - fix LOV locking bug that would get cli/srv out of sync
191 - fix echo client over LOV (409)
192 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
193 - fix double-O_EXCL intent crash (424)
194 - avoid sending multiple lock CANCELs (352)
196 - MDS can do multi-client recovery (modulo bugs in new code)
198 - many updates, edits, cleanups
200 2002-11-18 Phil Schwan <phil@clusterfs.com>
203 - fix null d_it dereference (346)
204 - fix full OST/dbench hang (333)
205 - fix permission problem with file removal (286)
206 - fix removal of OSCs from LOV when they fail
207 - fix NULL deref during bulk timeout (214)
208 - fix problems related to multiple filesystems on one MDS (241)
209 - fixed serious subtle metadata locking bugs
210 - free locks on clients when inodes are removed due to memory
212 - fix inode pointer in lock data (285)
213 - partial support for multiple MDS on a single host (241)
214 - data locks weren't cancelled at clear_inode time (290, 311)
215 - intent locks could lead to unbounded lock growth (205)
216 - added a maximum lock count, an LRU list, and a flusher
217 - fix multiple rename (365)
218 - properly abstracted the echo client
219 - OSC locked 1 byte too many; fixed
220 - rewrote brw callback code:
221 - fixed recovery bugs related to LOVs (306)
222 - fixed too-many-pages-in-one-write crash (191)
223 - fixed (again) crash in sync_io_timeout (214)
224 - probably fixed callback-related race (385)
226 - Add capability to MDS protocol
227 - LDLM cancellations and callbacks on different portals
229 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
232 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
233 - defer open object destruction to close time (601981, 138)
234 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
235 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
236 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
237 - don't call obd_setattr in ll_file_release for destroyed objects
238 * protocol change to lustre_msg: move |version| and add |flags|
239 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
240 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
241 * added replay of create, unlink, link and rename operations during
242 MDS failover; recovery should be much more robust now
243 * remove failed OSCs from LOVs (only lov_create uses this so far)
244 * the lustre-HOWTO was brought (more) up to date (582544)
246 2002-10-23 Phil Schwan <phil@clusterfs.com>
249 - in-use dentries weren't being reused properly (617851)
250 - prevent multiple LDLM setup (599178)
251 - fix LOV size calculations for truncate (617853)
252 - fix client handling of MDS intent errors (POSIX)
253 - fix permission bug in lovstripe.c test (624321)
254 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
255 - truncate past end of file could corrupt data
256 - proper cleanup after timeouts, crashes, etc (592524, 550815)
257 - a race in recovery could return ETIMEDOUT to apps (623947)
258 - building outside the source directory was fixed
259 * the lustre-HOWTO was brought (more) up to date (582544)
260 * major progress was made on recovery functionality
262 2002-10-10 Phil Schwan <phil@clusterfs.com>
265 - recovery deadlock fix
266 - rm -rf causes LBUG fix (617817)
267 - file open by multiple tasks fix (618962)
268 - directory permissions bugs (602707 and 620007)
269 - journal_stop fixed with locking (611313)
270 - O_APPEND failures resolved (618273, perhaps 614459)
271 - lconf PATH fix (619770)
272 - IA64 build fix (621450)
273 - RPC buffer sizes scale with amount of memory
275 2002-10-01 Phil Schwan <phil@clusterfs.com>
278 - locks would be cancelled without throwing away data pages,
279 resulting in inconsistent data (605627)
280 - inode attributes were not always being refreshed (605627, 612449)
281 - lconf now continues to cleanup after lctl reports an error
282 - MDS now enforces user permissions (602707)
283 - lprocfs cleanup fixed, but not yet enabled (614157)
284 - fixed infinite server hang, should a client not respond to an AST
285 - avoid going into recovery if user calls readlink() with a buffer
286 that's too small (613941)
287 - AST RPCs no longer require replies (614867) -- this may be changed
288 - don't crash server if client sends an IOV that's too big (611336)
289 - fixed lock conversion deadlock (611892)
290 - fixed the following of symlinks (614622)
291 * recovery: the server can remove locks from a client that dies, other
292 clients can make progress
293 * more extN patch fixes
294 * compile-time configurable ptlrpc buffer allocations
296 - collaborative read cache document
297 - Lustre Lite Performance CDR document-in-progress
299 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
302 - fix typo in patch-2.4.18
304 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
307 - clear ptlrpc request each time in handle_incoming_request()
308 - unlink of files now destroys the object on the OST
310 2002-09-19 Peter Braam <braam@clusterfs.com>
312 * add hard link support
313 * change obdfile creation method
314 * kernel patch changed
316 2002-09-19 Peter Braam <braam@clusterfs.com>
319 - stack overflow bug in extN fixed
321 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
323 * documentation updates
324 - add man pages for config tools
325 - update tests/README to describe testing with new config tools
326 - finish metadata API descriptions
327 * bug fixes and cleanups
328 - statfs workaround for 16TB limit
329 - LOV stripe allocation improved, can stripe on subset of OSTs
330 - LOV file size/IO offset was wrong for files > 4GB in size
331 - object EA data was being dropped, caused files to be unreadable
332 - memory overflow with non-LOV OST caused memory corruption
333 - fixed regression tests to work with new config tools, obdfilter
334 - fixed bug when directory size became larger than 1 block
335 - fixed bug (for single client case) when PWD was deleted
336 - invalidate local directory pages when doing intent-based ops
337 - avoid LDLM oops when lock callback contained bad data
339 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
341 * documentation updates
342 * bug fixes and cleanups
343 - configuration tools
346 - 64-bit compile warnings
347 - 64-bit internal statfs data
349 * test_brw on persistent OST devices
353 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
355 * documentation updates
356 * bug fixes and cleanups
357 * configuration tools
359 2002-08-30 Peter J. Braam <braam@clusterfs.com>
362 * many small fixes to 0_5_4
363 * io/network handling
364 * thinkos in MDS operations
366 2002-08-24 Peter J. Braam <braam@clusterfs.com>
369 * crucial basic fixes to 0.5.3
370 * IOR, Iozone work over Elan
373 2002-08-07 Phil Schwan <phil@clusterfs.com>
374 * version 0_5_3, our first alpha
375 * we use the new Portals iovs
376 * documentation updates
377 * bug fixes and cleanups
378 * small changes in the DLM wire protocol
380 2002-07-25 Peter J. Braam <braam@clusterfs.com>
381 * version 0_5_1 with some initial stability,
382 * locking on MD and file I/O.
383 * documentation updates
384 * several bug fixes since 0.5.0
385 * small changes in wire protocol
387 2002-07-18 Phil Schwan <phil@clusterfs.com>
389 * delivered as Lustre Light Alpha
390 * fixed a crash after handling invalid MDS requests
391 * fixed directory pages for architectures with non-4k pages sizes
393 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
394 * release version v0_4_4
395 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
397 * Fixes a number of interruption problems with OST operations.
398 * Update documentation for portals header changes
399 * Move all wire protocol structs/defines to lustre_idl.h
400 * Fixes symlink length bug.
401 * Add tcpdump to repository.
403 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
404 * release version v0_4_3
405 * Fixes statfs for inodes on extN.
406 * Fixes bug in runtests which would delete /etc/hosts.
407 * Use 64-bit object IDs wherever possible (not into VFS though)
408 Remove ost_get_info, which is unused by lustre, and out of date.
410 2002-07-03 Peter Braam <braam@clusterfs.com>
411 * release version v0_4_2 Fixes a lookup error (type not passed)
412 * move forward to head of Portals
413 * move forward to latest Lustre kernel
415 2002-06-25 Peter Braam <braam@clusterfs.com>
416 * release version v0_4_1. Hopefully stable on single node use.