1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5 - don't overwrite extent policy data in reply if lock was blocked (2901)
6 - drop filter export grants atomically with removal from device (2663)
7 - del obd_self_export from work_list in class_disconnect_exports (2908)
8 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
9 - swab reply message in mdc_close, other PPC fixes (2464)
10 - fix destroying of named logs (2325)
11 - overwrite old logs when running lconf --write_conf (2264)
12 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
13 - fix race in target_handle_connect (2898)
15 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
18 - account for cache space usage on clients to avoid data loss (974)
19 - lfsck support in lustre kernel code (2349)
20 - reduce journal credits needed for BRW writes (2370)
21 - orphan handling to avoid losing space on client/server crashes
22 - ptlrpcd can be blocked, stopping ALL progress (2477)
23 - use lock value blocks to assist in proper KMS, faster stat (1021)
24 - takes i_sem instead of DLM locks internally on obdfilter (2720)
25 - recovery for initial connections (2355)
26 - fixes for mds_cleanup_orphans (1934)
27 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
28 - block all file creations until orphan recovery completes (1901)
29 - client remove rq_connection from request struct (2423)
30 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
31 - recovery timer race (2670)
32 - mdc_close recovey bug (2532)
33 - ptlrpc cleanup bug (2710)
34 - mds timeout on local locks (2588)
35 - namespace lock held during RPCs (2431)
36 - handle interrupted sync write properly (2503)
37 - don't try to handle a message that hasn't been replied to (2699)
38 - client assert failure during cleanup after abort recovery (2701)
39 - leak mdc device after failed mount (2712)
40 - ptlrpc_check_set allows timedout requests to complete (2714)
41 - wait for inflight reqs when ptlrpcd finishes (2710)
42 - make sure unregistered services are removed from the srv_list
43 - reset bulk XID's when resending them (caught by 1138 test)
44 - unregister_bulk after timeout
45 - fix lconf error (2694)
46 - handle write after unfinished setstripe, stripe-only getstripe (2388)
47 - readahead locks pages, leaves pending causing memory pressure (2673)
48 - increase OST request buffers to 4096 on large machines (2729)
49 - fix up permission of existing directories in simple_mkdir (2661)
50 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
51 - don't assign transno to errored transactions (2742)
52 - don't delete objects on OST if given a bogus objid from MDS (2751)
53 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
54 - if rq_no_resend, then timeout request after recovery (2432)
55 - fix MDS llog_logid record size, 64-bit array alignment (2733)
56 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
57 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
58 - MDS recovery shouldn't delete objects that it has given out (2730)
59 - if enqueue arrives after completion, don't clobber LVB (2819)
60 - don't unlock pages twice when trigger_group_io returns error (2814)
61 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
62 - don't write pages to disk if there was an error (1450)
63 - don't ping imports that have recovery disabled (2676)
64 - take buffered bytes into account when balancing socknal conn (2817)
65 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
66 - reconnect unlink llog connection after MDS reconnects to OST (2816)
67 - remove little-endian swabbing of llog records (1987)
68 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
69 - retry reposting request buffers if they fail (1191)
70 - grow extent at grant time to avoid granting a revoked lock (2809)
71 - lock revoke doesn't evict page if covered by a second lock (2765)
72 - disable VM readahead to avoid reading outside lock extents (2805)
74 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
75 - updated kernel patches for hp-2.4.20 kernel (2681)
77 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
80 - fix truncated write corruption (2366)
81 - fix for failed assertion in iopen_connect_dentry (1792,2517)
83 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
84 - ensure len(uuid) < 37 in lmc (1171)
85 - fix ia64 OOPS in llog_test (2255)
86 - zero end of page at obdfilter for partial page writes (2648)
87 - don't leave stale dentries around after renames (bug 2428)
88 - fix timeouts when evicting a client with a single lock held (2642)
89 - set deadline for the initial HELLO message to drain (2634)
90 - print out dotted-quad IP addresses in the socknal (2302)
92 - additional debugging for MDS client eviction problem (2443)
93 - fix mkfsoptions support for osts (2603, 2604)
95 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
98 - add series for the vanilla 2.6.0 kernel
99 - add series for the vanilla 2.4.24 kernel
100 - add series for a cray x86/64 UL kernel drop
101 - fix xattr patches for the vanilla 2.4.19 series
103 - generate true UUIDs in lmc (1171)
104 - have portals stack dumping break in UML (2466)
105 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
106 - allocate with _NOFS in ldlm to avoid deadlock (1933)
107 - wake callback waiting threads on client eviction (2460)
108 - Add --ptldebug and --subsystem to lmc (1719)
109 - update assertion to allow safe interrupt allocation
110 - set rq_no_resend for cancel requests (2432)
111 - recalculate ptlrpcd timeout after resend (2494)
112 - call vfs_rmdir when removing pending directories (2368)
113 - fix renaming a file to itself (2429)
114 - lmc creates a default one-stripe lov (2454)
115 - expand procfs space to handle large clusters (2326)
116 - increase UML stack to avoid overflow
117 - update lconf's list of debug and subsystem masks
118 - fix lfs find --obd (2510)
119 - /proc tunable for disabling filter read caching (2591)
120 - stop rpm packages from altering slapd.conf (2301)
121 - disable nagle in the socknal under 0conf (2578)
122 - choose mds inode size based on stripe count (2572)
123 - fix kernel-source rpm problems (2516)
125 - add --disable-doc to avoid pdf generation (2421)
126 - update documentation, tests, type-os, comments
127 - avoid format warnings on ia64
129 - tiny code cleanups by removing unused fields
131 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
134 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
135 - remove the most blatant lies from BUILDING file (2371)
136 - change default debug level to reasonable production setting
137 - reduce client side cache size to reduce cache flush time
138 - reduce max RPCs in flight to avoid unnecessary file fragmentation
139 - make TCP zerocopy and pinger support enabled by default (2476)
140 - sync writes completed after process exits caused crashes (2319)
141 - maintain correct mount count on the MDS (2356)
142 - backout 1557, because 2316 wasn't really fixed
143 - better file I/O statistics gathering in /proc
144 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
145 - another kernel patch to fix zero-copy TCP function export
146 - don't take duplicate lock when processing re-sent getattr (2420)
147 - lctl uses obd_self_export instead of creating new conn (2353)
148 - MDS/OST recovery case which requires object creation asserted (2425)
149 - move lfs from /usr/sbin to /usr/bin in packages
150 - fix race between mds_client_add and mds_client_free (2417)
151 - use kmalloc instead of slabs in portals (2430)
152 - don't create duplicate records when a failover MDS is present (2442)
153 - remove unnecessary mount age check (2332)
154 - don't remove directory inodes from locks prematurely (2451)
155 - don't break if MDS service name is the same as hostname (2103)
156 - fix races in client write RPC generation when cache full (2482)
158 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
161 - remove now-unused request->rq_obd (278)
162 - if an allocation fails, print out how much memory we've used (1933)
163 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
164 - add the "configurable stack size" patch to most series files (1256)
165 - ability to write large log records, for 100+ OST configs (2306)
166 - fix NULL deref when filter_prep fails (2314)
167 - fix operator precedence error in filter_sync
168 - dynamic allocation of socknal TX descriptors (2315)
169 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
170 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
171 - ensure that CWARN messages in Portals always get to the syslog
172 - __init/__exit are not for prototype decls (ldlm_init/exit)
173 - x86-64 compile warning fixes
174 - fix gateway LMC keyword conflict (2318)
175 - fix MDS lock inversions in getattr/reint paths (1844)
176 - fix a rare lock re-ordering bug, which caused deadlock (2322)
177 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
178 - DLM race condition prevented some lock evictions (2328)
179 - ENOMEM detection and retry on socknal sends (2230)
180 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
181 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
182 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
183 - better error messages when a client is rejected during recovery (1505)
184 - avoid cancelling locks which were never granted, after failure (2330)
185 - fix i_sem/journal inversion in mds_client_add (2333)
186 - fix truncate/getattr lock cycle deadlock (2334)
187 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
188 - fix two rare exit paths which could leak an l_lock() ref (2321)
189 - fencepost error in MDS/OST orphan recovery (2226)
190 - make log record alignment 8 bytes (1988)
191 - lstripe now fails when requested offset > ost_count (2237)
192 - ensure that all kernel series have a complete list.h (1607)
193 - fix crashes in special-file operations (2316)
194 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
196 - allow configurable automake binary, for testing new versions
197 - small update to the lfs documentation
199 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
201 * fix negative export reference count in fsfilt_sync (2312)
203 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
204 * release candidate 0.9.1
206 - orphans are moved into the PENDING directory for possible recovery
207 - replayed opens now open by fid for orphan/rename safety (1042)
208 - last close of an orphan inode generates a transno (683)
209 - chdir() and mount() now pin the directory entry (1020)
210 - avoid CERROR in normal ll_setattr_raw() error case (1500)
211 - discard very old requests without processing them (1502)
212 - remove some common, well-understood CERRORs (1505)
213 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
214 - clear "grant" flags in OST replies until OST grant code lands (1644)
215 - fix read performance by not clobbering i_blksize on client (1598)
216 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
217 - make LRU size a /proc tunable, clears locks when reduced (707)
218 - fix some lprocfs rot that prevented ptlbd from loading (1732)
219 - server locks take references on exports now (1558)
220 - build fixes for 2.4.20-rh trees (1663)
221 - return an error from lov_create if all OSCs are inactive (1751)
222 - fix import levels when a reconnect happens without a timeout (1597)
223 - exit early from mds_open if we get a lookup error (1749)
224 - partial page read at EOF wouldn't wait for disk before sending (1642)
225 - avoid NULL deref in obdfilter when reading page past EOF (1592)
226 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
227 - fix LBUG in ll_it_open_error with rc = -2 (1861)
228 - write/truncate lock inversion (1639)
229 - Don't auto-load obdclass, portals modules during cleanup (1495)
230 - fix timestamps from jumping to "now" (1763)
231 - extra journal assertions (1648)
232 - add an extra multiunlink test (1771)
233 - fix read_record/write_record API (1776)
234 - fix leak of offset_extent, possible incorrect i_size later (1772)
235 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
236 - Debugging for the kqswnal_get_idle_tx problems (1820)
237 - Allow recovery to be attempted multiple times (1536)
238 - Write out MDS last_rcvd file after it is first created (1600)
239 - Fix tx_descriptor leak in failed transmit situations (1827)
240 - ext3 journaling fixes for assertion failure after IO error (1871)
241 - class_export_put() on freed export after completion AST error (1896)
242 - Fix revalidate looping in VFS (1322)
243 - Don't access a freed export during MDS_REINT timeout (1521)
244 - Add open-unlink recovery support on the MDS (1673,1764)
245 - Return an error if no MDS data was read from last_rcvd (1946)
246 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
247 - Don't setup a disk device that doesn't match exported UUID (317)
248 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
249 - avoid committing NULL handle in force close
250 - local.sh is now a one-stripe LOV configuration
251 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
252 - fix ext3 htree duplicate directory entry corruption (1516)
253 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
254 - update to new LOV EA format (2097)
255 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
257 2003-06-15 Phil Schwan <phil@clusterfs.com>
260 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
261 - per-import/export recovery handling (958, 931, 959)
262 - multiple last-rcvd slots, for serving multiple FSes (949)
263 - connections are again shared between multiple imp/exports (963, 964)
264 - "umount -f" would hang if any requests needed to be sent (393, 978)
265 - avoid pinning large req buffer by copying for queued messages (989)
266 - add "uuid" to "lctl device" command to help upcalls (991)
267 - "open" RPCs with transnos would confuse recovery counters (1037)
268 - do proper endian conversion of all wire messages (288, 340, 891)
269 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
270 - call waiting locks callback from LDLM recovery thread (1127, 1151)
271 - fix ptlrpc_connection leak in target_handle_connect (1174)
272 - fix import refcounting bug in OST and MDS cleanup (1134)
273 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
274 - fix very unlikely obd_types race condition (501)
275 - remove osc_open hack for echo_client (1187)
276 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
277 - a failure in read_inode2 leads to deadlock (1139)
278 - cancel ack-locks as soon as transaction is committed (1072)
279 - fix major leaks and crashes in the bulk I/O path (937, 1057)
280 - make sure to commitrw after any preprw to avoid deadlock (1162)
281 - failing to execute a file in a lustre FS would lock inode (1203)
282 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
283 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
284 - obd_uuid2tgt would walk off the end of the list (1255)
285 - on IA64 the osc would give portals incorrect bulk size (1258)
286 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
287 - fix lock inversion caused by new llite matching code (1282)
288 - limit the number of dirty pages on a client to 10MB (1286)
289 - timed out locks were not being corrected cancelled (1289)
290 - fix O_DIRECT above 4GB on IA-32 (1292)
291 * major user-visible changes
292 - fail out/fail over policy now controlled by the upcall (993)
294 - add OBD_PING to check server availability and failure (954)
295 - lustre messages are now sent in sending host order (288, 340, 891)
296 - add eadatalen to MDS getattr reply (340)
297 - OST read replies may contain second buffer, with per-page status (593)
299 2003-03-11 Phil Schwan <phil@clusterfs.com>
302 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
303 - failed open()s could cause deadlock; fixed (867, 869)
304 - stop cancelling OST locks when files are closed (481)
305 - overlapping XID spaces caused network corruption (851, 853)
306 - fix unsafe fsfilt counter arithmetic; change to atomic_t
307 - setattr_raw added, to do single-RPC, server-side setattrs
308 - lmc/lconf syntax change for OST UUIDs
309 - fix crashy race condition between ptlrpc_free_req and osc_close
310 - don't use request in mdc_enqueue if we hit a timeout (889)
311 - don't set the inode i_size for regular files from the MDS (896)
312 - handle out of order completion AST (842)
313 - don't LBUG if a lock request times out after receiving AST (913)
314 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
315 - if a bad lock AST arrives, send an error instead of dropping entirely
316 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
317 - fix leak in bulk IO when only partially completed (899, 900, 926)
318 - fix O_DIRECT for ia64 (55)
319 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
320 - C-z after timeout could hang a process forever; fixed (977)
322 - client-side I/O cache (678, 924, 929, 941, 970)
324 - READPAGE and SETATTRs which don't take server-side locks get
327 2003-02-11 Phil Schwan <phil@clusterfs.com>
330 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
331 - Fix fsfilt_extN_readpage() to read a full page of directory
332 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
333 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
334 - Fix imbalanced LOV object allocation and out-of-bound access (469)
335 - Most intent operations were removed, in favour of a new RPC mode
336 that does a single RPC to the server and bypasses most of the VFS
337 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
338 - Aggressively cancel local locks on DLM servers
339 - mds_reint_unlink sends EA to the client if it's the last nlink.
340 client uses that EA to unlink OST objects.
341 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
342 - recursive symlinks were fixed (439)
343 - fixed NULL deref in DEBUG_REQ
344 - filter_update_lastobjid no longer calls sync, which annoyed extN
345 - fixed multi-client small-writes to a single file problem (445)
346 - fixed mtime updates during file writes (607)
347 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
348 - fixed bug in obd_brw_read/write() (under guise of testing 367)
349 - fixed Linux OST size reporting problem (444, 656)
350 - OST now updates object mtime with writes or setattr (607, 619)
351 - client verifies file size before zeroing page past EOF (445)
352 - OST now writes last allocated objid to disk with allocation (108)
353 - LOV on echo now works (409)
355 - mds_reint_unlink sends a new buffer, with the EA included. this
356 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
357 set if a regular file was being unlinked, and it was the last link
358 - use PtlGet from the target for bulk writes (315)
359 - OST now updates object mtime with writes or setattr (607, 619)
360 - LDLM now has a grant-time callback to revalidate locked items, if
362 - Many MDS operations were reorganized to combat race conditions
364 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
365 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
367 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
370 - Fully reactivate OST imports after reconnection (512, others)
371 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
372 - More graceful error handling for truncating on dead OST (515)
373 - Don't error out unless we're actually accessing dead stripes (474)
374 - Fix garbage sizes when stripes are missing (410)
375 - LRU counters were broken, causing constant lock purge (433, 432)
376 - garbage on read from stripes with failed OSTs (441)
377 - mark OSCs as active before reconnecting during recovery (438)
378 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
379 - lfind did not preserve OST order in output (443)
380 - symlinks cause hung clients, incorrect data (439)
381 - stop dereferencing request after dropping refcount (457)
382 - don't LASSERT(spin_is_locked) on non-SMP (455)
383 - fixes for many rename() bugs
384 - fstat didn't correctly synchronize attributes (399)
385 - server must handle lock cancellation during blocking AST prep (487)
386 - bulk descriptors were free()d too soon (511)
387 - fix paths in lconf, which would load incorrect modules (451, 507)
388 - fix confusing lconf 'host not found' error message (386)
389 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
390 - fix race condition in mdc_blocking_ast() for inode access (526)
391 - fix lov_unpackmd() unpacking wrong number of stripes (537)
392 - fix lov_set_osc_active() marking wrong OSC inactive (440)
393 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
394 - fix multiple writes of stripe MD to MDS (358, maybe 519)
395 - fix lstripe in several ways (kernel side) (527)
396 - fix request leak in ldlm_cli_enqueue (262)
397 - incorrect OSC was marked inactive after OST failure
398 - call mds_fs_cleanup before unmounting filesystem (524)
399 - fix races between taking ns_lock and ldlm_lock_change_resource
400 - fix races updating LOV export open file list
401 - fix lov_enqueue error path, avoid decref-ing bad lock handle
402 - fix recovery NULL deref in ldlm_cli_cancel_unused
403 - fix some DLM races by using new hash table for lock handles (419)
404 - permit the client to specify desired inodes, at replay
405 - duplicate requests when we queue them for replay reintegration
406 - fix last_rcvd offset calculation
407 - sync after each recovered transaction, so we always make progress
408 - never, not always, ERESTART requests without transnos
409 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
411 - skip replay if the MDS says that the client is already connected
412 - don't check for a recovery-enabled export to match lctl's UUID
413 - don't INC_USE_COUNT for phantom exports
414 - don't crash when cleaning up phantom exports (567)
415 - don't double-finish or set replay data for errored mdc_open requests
416 - abort requests when they time out, so we don't get old replies
417 - send/receive replies for AST messages again
418 - if the client says that it doesn't have the lock, cancel it on the
420 - if we timeout during I/O, don't try to cancel an in-use lock; instead
421 mark it as destroyed, it will all work out when decref is called
422 - fix module use counts (22, 581)
424 - ASTs now expect a reply (server cancels lock on error reply)
426 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
429 - fix many simultaneous client startup (392)
430 - fix dentry->d_it clobbering
431 - credentials weren't being shipped for readdir/getattr operations
432 - remove invalid assertions triggered during some concurrent MD
434 - proper Lustre versions added (336, 389)
435 - fix memory leak for create error case (398)
436 - fix LOV locking bug that would get cli/srv out of sync
437 - fix echo client over LOV (409)
438 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
439 - fix double-O_EXCL intent crash (424)
440 - avoid sending multiple lock CANCELs (352)
442 - MDS can do multi-client recovery (modulo bugs in new code)
444 - many updates, edits, cleanups
446 2002-11-18 Phil Schwan <phil@clusterfs.com>
449 - fix null d_it dereference (346)
450 - fix full OST/dbench hang (333)
451 - fix permission problem with file removal (286)
452 - fix removal of OSCs from LOV when they fail
453 - fix NULL deref during bulk timeout (214)
454 - fix problems related to multiple filesystems on one MDS (241)
455 - fixed serious subtle metadata locking bugs
456 - free locks on clients when inodes are removed due to memory
458 - fix inode pointer in lock data (285)
459 - partial support for multiple MDS on a single host (241)
460 - data locks weren't cancelled at clear_inode time (290, 311)
461 - intent locks could lead to unbounded lock growth (205)
462 - added a maximum lock count, an LRU list, and a flusher
463 - fix multiple rename (365)
464 - properly abstracted the echo client
465 - OSC locked 1 byte too many; fixed
466 - rewrote brw callback code:
467 - fixed recovery bugs related to LOVs (306)
468 - fixed too-many-pages-in-one-write crash (191)
469 - fixed (again) crash in sync_io_timeout (214)
470 - probably fixed callback-related race (385)
472 - Add capability to MDS protocol
473 - LDLM cancellations and callbacks on different portals
475 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
478 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
479 - defer open object destruction to close time (601981, 138)
480 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
481 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
482 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
483 - don't call obd_setattr in ll_file_release for destroyed objects
484 * protocol change to lustre_msg: move |version| and add |flags|
485 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
486 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
487 * added replay of create, unlink, link and rename operations during
488 MDS failover; recovery should be much more robust now
489 * remove failed OSCs from LOVs (only lov_create uses this so far)
490 * the lustre-HOWTO was brought (more) up to date (582544)
492 2002-10-23 Phil Schwan <phil@clusterfs.com>
495 - in-use dentries weren't being reused properly (617851)
496 - prevent multiple LDLM setup (599178)
497 - fix LOV size calculations for truncate (617853)
498 - fix client handling of MDS intent errors (POSIX)
499 - fix permission bug in lovstripe.c test (624321)
500 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
501 - truncate past end of file could corrupt data
502 - proper cleanup after timeouts, crashes, etc (592524, 550815)
503 - a race in recovery could return ETIMEDOUT to apps (623947)
504 - building outside the source directory was fixed
505 * the lustre-HOWTO was brought (more) up to date (582544)
506 * major progress was made on recovery functionality
508 2002-10-10 Phil Schwan <phil@clusterfs.com>
511 - recovery deadlock fix
512 - rm -rf causes LBUG fix (617817)
513 - file open by multiple tasks fix (618962)
514 - directory permissions bugs (602707 and 620007)
515 - journal_stop fixed with locking (611313)
516 - O_APPEND failures resolved (618273, perhaps 614459)
517 - lconf PATH fix (619770)
518 - IA64 build fix (621450)
519 - RPC buffer sizes scale with amount of memory
521 2002-10-01 Phil Schwan <phil@clusterfs.com>
524 - locks would be cancelled without throwing away data pages,
525 resulting in inconsistent data (605627)
526 - inode attributes were not always being refreshed (605627, 612449)
527 - lconf now continues to cleanup after lctl reports an error
528 - MDS now enforces user permissions (602707)
529 - lprocfs cleanup fixed, but not yet enabled (614157)
530 - fixed infinite server hang, should a client not respond to an AST
531 - avoid going into recovery if user calls readlink() with a buffer
532 that's too small (613941)
533 - AST RPCs no longer require replies (614867) -- this may be changed
534 - don't crash server if client sends an IOV that's too big (611336)
535 - fixed lock conversion deadlock (611892)
536 - fixed the following of symlinks (614622)
537 * recovery: the server can remove locks from a client that dies, other
538 clients can make progress
539 * more extN patch fixes
540 * compile-time configurable ptlrpc buffer allocations
542 - collaborative read cache document
543 - Lustre Lite Performance CDR document-in-progress
545 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
548 - fix typo in patch-2.4.18
550 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
553 - clear ptlrpc request each time in handle_incoming_request()
554 - unlink of files now destroys the object on the OST
556 2002-09-19 Peter Braam <braam@clusterfs.com>
558 * add hard link support
559 * change obdfile creation method
560 * kernel patch changed
562 2002-09-19 Peter Braam <braam@clusterfs.com>
565 - stack overflow bug in extN fixed
567 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
569 * documentation updates
570 - add man pages for config tools
571 - update tests/README to describe testing with new config tools
572 - finish metadata API descriptions
573 * bug fixes and cleanups
574 - statfs workaround for 16TB limit
575 - LOV stripe allocation improved, can stripe on subset of OSTs
576 - LOV file size/IO offset was wrong for files > 4GB in size
577 - object EA data was being dropped, caused files to be unreadable
578 - memory overflow with non-LOV OST caused memory corruption
579 - fixed regression tests to work with new config tools, obdfilter
580 - fixed bug when directory size became larger than 1 block
581 - fixed bug (for single client case) when PWD was deleted
582 - invalidate local directory pages when doing intent-based ops
583 - avoid LDLM oops when lock callback contained bad data
585 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
587 * documentation updates
588 * bug fixes and cleanups
589 - configuration tools
592 - 64-bit compile warnings
593 - 64-bit internal statfs data
595 * test_brw on persistent OST devices
599 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
601 * documentation updates
602 * bug fixes and cleanups
603 * configuration tools
605 2002-08-30 Peter J. Braam <braam@clusterfs.com>
608 * many small fixes to 0_5_4
609 * io/network handling
610 * thinkos in MDS operations
612 2002-08-24 Peter J. Braam <braam@clusterfs.com>
615 * crucial basic fixes to 0.5.3
616 * IOR, Iozone work over Elan
619 2002-08-07 Phil Schwan <phil@clusterfs.com>
620 * version 0_5_3, our first alpha
621 * we use the new Portals iovs
622 * documentation updates
623 * bug fixes and cleanups
624 * small changes in the DLM wire protocol
626 2002-07-25 Peter J. Braam <braam@clusterfs.com>
627 * version 0_5_1 with some initial stability,
628 * locking on MD and file I/O.
629 * documentation updates
630 * several bug fixes since 0.5.0
631 * small changes in wire protocol
633 2002-07-18 Phil Schwan <phil@clusterfs.com>
635 * delivered as Lustre Light Alpha
636 * fixed a crash after handling invalid MDS requests
637 * fixed directory pages for architectures with non-4k pages sizes
639 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
640 * release version v0_4_4
641 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
643 * Fixes a number of interruption problems with OST operations.
644 * Update documentation for portals header changes
645 * Move all wire protocol structs/defines to lustre_idl.h
646 * Fixes symlink length bug.
647 * Add tcpdump to repository.
649 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
650 * release version v0_4_3
651 * Fixes statfs for inodes on extN.
652 * Fixes bug in runtests which would delete /etc/hosts.
653 * Use 64-bit object IDs wherever possible (not into VFS though)
654 Remove ost_get_info, which is unused by lustre, and out of date.
656 2002-07-03 Peter Braam <braam@clusterfs.com>
657 * release version v0_4_2 Fixes a lookup error (type not passed)
658 * move forward to head of Portals
659 * move forward to latest Lustre kernel
661 2002-06-25 Peter Braam <braam@clusterfs.com>
662 * release version v0_4_1. Hopefully stable on single node use.