4 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5 - per-import/export recovery handling (958, 931, 959)
6 - multiple last-rcvd slots, for serving multiple FSes (949)
7 - connections are again shared between multiple imp/exports (963, 964)
8 - "umount -f" would hang if any requests needed to be sent (393, 978)
9 - avoid pinning large req buffer by copying for queued messages (989)
10 - add "uuid" to "lctl device" command to help upcalls (991)
11 - "open" RPCs with transnos would confuse recovery counters (1037)
12 - do proper endian conversion of all wire messages (288, 340, 891)
13 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
14 - call waiting locks callback from LDLM recovery thread (1127, 1151)
15 - fix ptlrpc_connection leak in target_handle_connect (1174)
16 - fix import refcounting bug in OST and MDS cleanup (1134)
17 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
18 - fix very unlikely obd_types race condition (501)
19 - remove osc_open hack for echo_client (1187)
20 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
21 - a failure in read_inode2 leads to deadlock (1139)
22 - cancel ack-locks as soon as transaction is committed (1072)
23 - fix major leaks and crashes in the bulk I/O path (937, 1057)
24 - make sure to commitrw after any preprw to avoid deadlock (1162)
25 - failing to execute a file in a lustre FS would lock inode (1203)
26 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
27 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
28 - obd_uuid2tgt would walk off the end of the list (1255)
29 - on IA64 the osc would give portals incorrect bulk size (1258)
30 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
31 - fix lock inversion caused by new llite matching code (1282)
32 - limit the number of dirty pages on a client to 10MB (1286)
33 - timed out locks were not being corrected cancelled (1289)
34 - fix O_DIRECT above 4GB on IA-32 (1292)
35 * major user-visible changes
36 - fail out/fail over policy now controlled by the upcall (993)
38 - add OBD_PING to check server availability and failure (954)
39 - lustre messages are now sent in sending host order (288, 340, 891)
40 - add eadatalen to MDS getattr reply (340)
41 - OST read replies may contain second buffer, with per-page status (593)
43 2003-03-11 Phil Schwan <phil@clusterfs.com>
46 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
47 - failed open()s could cause deadlock; fixed (867, 869)
48 - stop cancelling OST locks when files are closed (481)
49 - overlapping XID spaces caused network corruption (851, 853)
50 - fix unsafe fsfilt counter arithmetic; change to atomic_t
51 - setattr_raw added, to do single-RPC, server-side setattrs
52 - lmc/lconf syntax change for OST UUIDs
53 - fix crashy race condition between ptlrpc_free_req and osc_close
54 - don't use request in mdc_enqueue if we hit a timeout (889)
55 - don't set the inode i_size for regular files from the MDS (896)
56 - handle out of order completion AST (842)
57 - don't LBUG if a lock request times out after receiving AST (913)
58 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
59 - if a bad lock AST arrives, send an error instead of dropping entirely
60 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
61 - fix leak in bulk IO when only partially completed (899, 900, 926)
62 - fix O_DIRECT for ia64 (55)
63 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
64 - C-z after timeout could hang a process forever; fixed (977)
66 - client-side I/O cache (678, 924, 929, 941, 970)
68 - READPAGE and SETATTRs which don't take server-side locks get
71 2003-02-11 Phil Schwan <phil@clusterfs.com>
74 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
75 - Fix fsfilt_extN_readpage() to read a full page of directory
76 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
77 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
78 - Fix imbalanced LOV object allocation and out-of-bound access (469)
79 - Most intent operations were removed, in favour of a new RPC mode
80 that does a single RPC to the server and bypasses most of the VFS
81 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
82 - Aggressively cancel local locks on DLM servers
83 - mds_reint_unlink sends EA to the client if it's the last nlink.
84 client uses that EA to unlink OST objects.
85 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
86 - recursive symlinks were fixed (439)
87 - fixed NULL deref in DEBUG_REQ
88 - filter_update_lastobjid no longer calls sync, which annoyed extN
89 - fixed multi-client small-writes to a single file problem (445)
90 - fixed mtime updates during file writes (607)
91 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
92 - fixed bug in obd_brw_read/write() (under guise of testing 367)
93 - fixed Linux OST size reporting problem (444, 656)
94 - OST now updates object mtime with writes or setattr (607, 619)
95 - client verifies file size before zeroing page past EOF (445)
96 - OST now writes last allocated objid to disk with allocation (108)
97 - LOV on echo now works (409)
99 - mds_reint_unlink sends a new buffer, with the EA included. this
100 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
101 set if a regular file was being unlinked, and it was the last link
102 - use PtlGet from the target for bulk writes (315)
103 - OST now updates object mtime with writes or setattr (607, 619)
104 - LDLM now has a grant-time callback to revalidate locked items, if
106 - Many MDS operations were reorganized to combat race conditions
108 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
109 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
111 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
114 - Fully reactivate OST imports after reconnection (512, others)
115 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
116 - More graceful error handling for truncating on dead OST (515)
117 - Don't error out unless we're actually accessing dead stripes (474)
118 - Fix garbage sizes when stripes are missing (410)
119 - LRU counters were broken, causing constant lock purge (433, 432)
120 - garbage on read from stripes with failed OSTs (441)
121 - mark OSCs as active before reconnecting during recovery (438)
122 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
123 - lfind did not preserve OST order in output (443)
124 - symlinks cause hung clients, incorrect data (439)
125 - stop dereferencing request after dropping refcount (457)
126 - don't LASSERT(spin_is_locked) on non-SMP (455)
127 - fixes for many rename() bugs
128 - fstat didn't correctly synchronize attributes (399)
129 - server must handle lock cancellation during blocking AST prep (487)
130 - bulk descriptors were free()d too soon (511)
131 - fix paths in lconf, which would load incorrect modules (451, 507)
132 - fix confusing lconf 'host not found' error message (386)
133 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
134 - fix race condition in mdc_blocking_ast() for inode access (526)
135 - fix lov_unpackmd() unpacking wrong number of stripes (537)
136 - fix lov_set_osc_active() marking wrong OSC inactive (440)
137 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
138 - fix multiple writes of stripe MD to MDS (358, maybe 519)
139 - fix lstripe in several ways (kernel side) (527)
140 - fix request leak in ldlm_cli_enqueue (262)
141 - incorrect OSC was marked inactive after OST failure
142 - call mds_fs_cleanup before unmounting filesystem (524)
143 - fix races between taking ns_lock and ldlm_lock_change_resource
144 - fix races updating LOV export open file list
145 - fix lov_enqueue error path, avoid decref-ing bad lock handle
146 - fix recovery NULL deref in ldlm_cli_cancel_unused
147 - fix some DLM races by using new hash table for lock handles (419)
148 - permit the client to specify desired inodes, at replay
149 - duplicate requests when we queue them for replay reintegration
150 - fix last_rcvd offset calculation
151 - sync after each recovered transaction, so we always make progress
152 - never, not always, ERESTART requests without transnos
153 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
155 - skip replay if the MDS says that the client is already connected
156 - don't check for a recovery-enabled export to match lctl's UUID
157 - don't INC_USE_COUNT for phantom exports
158 - don't crash when cleaning up phantom exports (567)
159 - don't double-finish or set replay data for errored mdc_open requests
160 - abort requests when they time out, so we don't get old replies
161 - send/receive replies for AST messages again
162 - if the client says that it doesn't have the lock, cancel it on the
164 - if we timeout during I/O, don't try to cancel an in-use lock; instead
165 mark it as destroyed, it will all work out when decref is called
166 - fix module use counts (22, 581)
168 - ASTs now expect a reply (server cancels lock on error reply)
170 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
173 - fix many simultaneous client startup (392)
174 - fix dentry->d_it clobbering
175 - credentials weren't being shipped for readdir/getattr operations
176 - remove invalid assertions triggered during some concurrent MD
178 - proper Lustre versions added (336, 389)
179 - fix memory leak for create error case (398)
180 - fix LOV locking bug that would get cli/srv out of sync
181 - fix echo client over LOV (409)
182 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
183 - fix double-O_EXCL intent crash (424)
184 - avoid sending multiple lock CANCELs (352)
186 - MDS can do multi-client recovery (modulo bugs in new code)
188 - many updates, edits, cleanups
190 2002-11-18 Phil Schwan <phil@clusterfs.com>
193 - fix null d_it dereference (346)
194 - fix full OST/dbench hang (333)
195 - fix permission problem with file removal (286)
196 - fix removal of OSCs from LOV when they fail
197 - fix NULL deref during bulk timeout (214)
198 - fix problems related to multiple filesystems on one MDS (241)
199 - fixed serious subtle metadata locking bugs
200 - free locks on clients when inodes are removed due to memory
202 - fix inode pointer in lock data (285)
203 - partial support for multiple MDS on a single host (241)
204 - data locks weren't cancelled at clear_inode time (290, 311)
205 - intent locks could lead to unbounded lock growth (205)
206 - added a maximum lock count, an LRU list, and a flusher
207 - fix multiple rename (365)
208 - properly abstracted the echo client
209 - OSC locked 1 byte too many; fixed
210 - rewrote brw callback code:
211 - fixed recovery bugs related to LOVs (306)
212 - fixed too-many-pages-in-one-write crash (191)
213 - fixed (again) crash in sync_io_timeout (214)
214 - probably fixed callback-related race (385)
216 - Add capability to MDS protocol
217 - LDLM cancellations and callbacks on different portals
219 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
222 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
223 - defer open object destruction to close time (601981, 138)
224 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
225 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
226 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
227 - don't call obd_setattr in ll_file_release for destroyed objects
228 * protocol change to lustre_msg: move |version| and add |flags|
229 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
230 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
231 * added replay of create, unlink, link and rename operations during
232 MDS failover; recovery should be much more robust now
233 * remove failed OSCs from LOVs (only lov_create uses this so far)
234 * the lustre-HOWTO was brought (more) up to date (582544)
236 2002-10-23 Phil Schwan <phil@clusterfs.com>
239 - in-use dentries weren't being reused properly (617851)
240 - prevent multiple LDLM setup (599178)
241 - fix LOV size calculations for truncate (617853)
242 - fix client handling of MDS intent errors (POSIX)
243 - fix permission bug in lovstripe.c test (624321)
244 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
245 - truncate past end of file could corrupt data
246 - proper cleanup after timeouts, crashes, etc (592524, 550815)
247 - a race in recovery could return ETIMEDOUT to apps (623947)
248 - building outside the source directory was fixed
249 * the lustre-HOWTO was brought (more) up to date (582544)
250 * major progress was made on recovery functionality
252 2002-10-10 Phil Schwan <phil@clusterfs.com>
255 - recovery deadlock fix
256 - rm -rf causes LBUG fix (617817)
257 - file open by multiple tasks fix (618962)
258 - directory permissions bugs (602707 and 620007)
259 - journal_stop fixed with locking (611313)
260 - O_APPEND failures resolved (618273, perhaps 614459)
261 - lconf PATH fix (619770)
262 - IA64 build fix (621450)
263 - RPC buffer sizes scale with amount of memory
265 2002-10-01 Phil Schwan <phil@clusterfs.com>
268 - locks would be cancelled without throwing away data pages,
269 resulting in inconsistent data (605627)
270 - inode attributes were not always being refreshed (605627, 612449)
271 - lconf now continues to cleanup after lctl reports an error
272 - MDS now enforces user permissions (602707)
273 - lprocfs cleanup fixed, but not yet enabled (614157)
274 - fixed infinite server hang, should a client not respond to an AST
275 - avoid going into recovery if user calls readlink() with a buffer
276 that's too small (613941)
277 - AST RPCs no longer require replies (614867) -- this may be changed
278 - don't crash server if client sends an IOV that's too big (611336)
279 - fixed lock conversion deadlock (611892)
280 - fixed the following of symlinks (614622)
281 * recovery: the server can remove locks from a client that dies, other
282 clients can make progress
283 * more extN patch fixes
284 * compile-time configurable ptlrpc buffer allocations
286 - collaborative read cache document
287 - Lustre Lite Performance CDR document-in-progress
289 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
292 - fix typo in patch-2.4.18
294 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
297 - clear ptlrpc request each time in handle_incoming_request()
298 - unlink of files now destroys the object on the OST
300 2002-09-19 Peter Braam <braam@clusterfs.com>
302 * add hard link support
303 * change obdfile creation method
304 * kernel patch changed
306 2002-09-19 Peter Braam <braam@clusterfs.com>
309 - stack overflow bug in extN fixed
311 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
313 * documentation updates
314 - add man pages for config tools
315 - update tests/README to describe testing with new config tools
316 - finish metadata API descriptions
317 * bug fixes and cleanups
318 - statfs workaround for 16TB limit
319 - LOV stripe allocation improved, can stripe on subset of OSTs
320 - LOV file size/IO offset was wrong for files > 4GB in size
321 - object EA data was being dropped, caused files to be unreadable
322 - memory overflow with non-LOV OST caused memory corruption
323 - fixed regression tests to work with new config tools, obdfilter
324 - fixed bug when directory size became larger than 1 block
325 - fixed bug (for single client case) when PWD was deleted
326 - invalidate local directory pages when doing intent-based ops
327 - avoid LDLM oops when lock callback contained bad data
329 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
331 * documentation updates
332 * bug fixes and cleanups
333 - configuration tools
336 - 64-bit compile warnings
337 - 64-bit internal statfs data
339 * test_brw on persistent OST devices
343 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
345 * documentation updates
346 * bug fixes and cleanups
347 * configuration tools
349 2002-08-30 Peter J. Braam <braam@clusterfs.com>
352 * many small fixes to 0_5_4
353 * io/network handling
354 * thinkos in MDS operations
356 2002-08-24 Peter J. Braam <braam@clusterfs.com>
359 * crucial basic fixes to 0.5.3
360 * IOR, Iozone work over Elan
363 2002-08-07 Phil Schwan <phil@clusterfs.com>
364 * version 0_5_3, our first alpha
365 * we use the new Portals iovs
366 * documentation updates
367 * bug fixes and cleanups
368 * small changes in the DLM wire protocol
370 2002-07-25 Peter J. Braam <braam@clusterfs.com>
371 * version 0_5_1 with some initial stability,
372 * locking on MD and file I/O.
373 * documentation updates
374 * several bug fixes since 0.5.0
375 * small changes in wire protocol
377 2002-07-18 Phil Schwan <phil@clusterfs.com>
379 * delivered as Lustre Light Alpha
380 * fixed a crash after handling invalid MDS requests
381 * fixed directory pages for architectures with non-4k pages sizes
383 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
384 * release version v0_4_4
385 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
387 * Fixes a number of interruption problems with OST operations.
388 * Update documentation for portals header changes
389 * Move all wire protocol structs/defines to lustre_idl.h
390 * Fixes symlink length bug.
391 * Add tcpdump to repository.
393 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
394 * release version v0_4_3
395 * Fixes statfs for inodes on extN.
396 * Fixes bug in runtests which would delete /etc/hosts.
397 * Use 64-bit object IDs wherever possible (not into VFS though)
398 Remove ost_get_info, which is unused by lustre, and out of date.
400 2002-07-03 Peter Braam <braam@clusterfs.com>
401 * release version v0_4_2 Fixes a lookup error (type not passed)
402 * move forward to head of Portals
403 * move forward to latest Lustre kernel
405 2002-06-25 Peter Braam <braam@clusterfs.com>
406 * release version v0_4_1. Hopefully stable on single node use.