1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4 - don't dereference NULL peer_ni in ldlm_handle_ast_error (3258)
5 - don't allow unlinking open directory if it isn't empty (2904)
6 - handle partial page writes in filter; fix 512b direct IO (3138)
7 - handle page cache pages in cleanup path for 2.6 (3335)
8 - leave liblustre's partial write handling to filter (3274)
9 - chose better nal ids in liblustre (3292)
10 - initialize liblustre with uid/group membership (2862)
11 - let lconf resolve symlinked-to devices (4629)
13 - drop scimac NAL (unmaintained)
15 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
18 - don't copy lvb into (possibly NULL) reply on error (2983)
19 - don't deref dentry after dput, don't free lvb on error (2922)
20 - use the kms to determine writeback rpc length (2947)
21 - increment oti_logcookies when osc is inactive (2948)
22 - update client's i_blocks count via lvb messages (2543)
23 - handle intent open/close of special files properly (1557)
24 - mount MDS with errors=remount-ro, like obdfilter (2009)
25 - initialize lock handle to avoid ASSERT on error cleanup (3057)
26 - don't use cancelling-locks' kms values (2947)
27 - use highest lock extent for kms, not last one (2925)
28 - don't dereference ERR_PTR() dentry in error handling path (3107)
29 - fix thread race in portals_debug_dumplog() (3122)
30 - create lprocfs device entries at setup instead of at attach (1519)
31 - common AST error handler, don't evict client on completion race (3145)
32 - zero nameidata in detach_mnt in 2.6 (3118)
33 - verify d_inode after revalidate_special is valid in 2.6 (3116)
34 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
35 - initialize RPC timeout timer earlier for 2.6 (3219)
36 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
37 - print nal/nid for unknown nid (3258)
38 - additional checks for oscc recovery before doing precreate (3284)
39 - fix ll_extent_lock() error return code for 64-bit systems (3043)
40 - don't crash in mdc_close for bad permissions on open (3285)
41 - zero i_rdev for non-device files (3147)
42 - clear page->private before handing to FS, better assertion (3119)
43 - fix incorrect decref of invalidated dentry (2350)
44 - don't hold journal transaction open across create RPC (3313)
45 - update atime on MDS at close time (3265)
46 - close LDAP connection when recovering to avoid server load (3315)
47 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
48 - don't leak open file on MDS after open resend (3325)
49 - serialize filter_precreate and filter_destroy_precreated (3329)
50 - loop device shouldn't call sync_dev() for nul device (3092)
51 - clear page cache after eviction (2766)
52 - resynchronize MDS->OST in background (2824)
53 - refuse to mount the same filesystem twice on same mountpoint (3394)
54 - allow llmount to create routes for mounting behind routers (3320)
55 - push lock cancellation to blocking thread for glimpse ASTs (3409)
56 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
57 - fix rare problem with rename on htree directories (3417)
59 - allow default OST striping configuration per directory (1414)
60 - fix compilation for qswnal for 2.6 kernels (3125)
61 - increase maximum number of MDS request buffers for large systems
62 - change liblustreapi to be useful for external progs like lfsck (3098)
63 - increase local configuration timeout for slow disks (3353)
65 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
68 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
69 - don't overwrite extent policy data in reply if lock was blocked (2901)
70 - drop filter export grants atomically with removal from device (2663)
71 - del obd_self_export from work_list in class_disconnect_exports (2908)
72 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
73 - swab reply message in mdc_close, other PPC fixes (2464)
74 - fix destroying of named logs (2325)
75 - overwrite old logs when running lconf --write_conf (2264)
76 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
77 - fix race in target_handle_connect (2898)
78 - mds_reint_create() should take same inode create lock (2926)
79 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
80 - don't close files for self_export to avoid uninitialized obd (2936)
81 - allow MDS with the same name as client node (2939)
82 - hold dentry reference for closed log files for unlink (2325)
83 - reserve space for all logs during transactions (2059)
84 - don't evict page beyond end of stripe extent (2925)
85 - don't oops on a deleted current working directory (2399)
86 - handle hard links to targets without a parent properly (2517)
87 - don't dereference NULL lock when racing during eviction (2867)
88 - don't grow lock extents when lots of conflicting locks (2919)
90 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
93 - account for cache space usage on clients to avoid data loss (974)
94 - lfsck support in lustre kernel code (2349)
95 - reduce journal credits needed for BRW writes (2370)
96 - orphan handling to avoid losing space on client/server crashes
97 - ptlrpcd can be blocked, stopping ALL progress (2477)
98 - use lock value blocks to assist in proper KMS, faster stat (1021)
99 - takes i_sem instead of DLM locks internally on obdfilter (2720)
100 - recovery for initial connections (2355)
101 - fixes for mds_cleanup_orphans (1934)
102 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
103 - block all file creations until orphan recovery completes (1901)
104 - client remove rq_connection from request struct (2423)
105 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
106 - recovery timer race (2670)
107 - mdc_close recovey bug (2532)
108 - ptlrpc cleanup bug (2710)
109 - mds timeout on local locks (2588)
110 - namespace lock held during RPCs (2431)
111 - handle interrupted sync write properly (2503)
112 - don't try to handle a message that hasn't been replied to (2699)
113 - client assert failure during cleanup after abort recovery (2701)
114 - leak mdc device after failed mount (2712)
115 - ptlrpc_check_set allows timedout requests to complete (2714)
116 - wait for inflight reqs when ptlrpcd finishes (2710)
117 - make sure unregistered services are removed from the srv_list
118 - reset bulk XID's when resending them (caught by 1138 test)
119 - unregister_bulk after timeout
120 - fix lconf error (2694)
121 - handle write after unfinished setstripe, stripe-only getstripe (2388)
122 - readahead locks pages, leaves pending causing memory pressure (2673)
123 - increase OST request buffers to 4096 on large machines (2729)
124 - fix up permission of existing directories in simple_mkdir (2661)
125 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
126 - don't assign transno to errored transactions (2742)
127 - don't delete objects on OST if given a bogus objid from MDS (2751)
128 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
129 - if rq_no_resend, then timeout request after recovery (2432)
130 - fix MDS llog_logid record size, 64-bit array alignment (2733)
131 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
132 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
133 - MDS recovery shouldn't delete objects that it has given out (2730)
134 - if enqueue arrives after completion, don't clobber LVB (2819)
135 - don't unlock pages twice when trigger_group_io returns error (2814)
136 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
137 - don't write pages to disk if there was an error (1450)
138 - don't ping imports that have recovery disabled (2676)
139 - take buffered bytes into account when balancing socknal conn (2817)
140 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
141 - reconnect unlink llog connection after MDS reconnects to OST (2816)
142 - remove little-endian swabbing of llog records (1987)
143 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
144 - retry reposting request buffers if they fail (1191)
145 - grow extent at grant time to avoid granting a revoked lock (2809)
146 - lock revoke doesn't evict page if covered by a second lock (2765)
147 - disable VM readahead to avoid reading outside lock extents (2805)
149 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
150 - updated kernel patches for hp-2.4.20 kernel (2681)
152 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
155 - fix truncated write corruption (2366)
156 - fix for failed assertion in iopen_connect_dentry (1792,2517)
158 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
159 - ensure len(uuid) < 37 in lmc (1171)
160 - fix ia64 OOPS in llog_test (2255)
161 - zero end of page at obdfilter for partial page writes (2648)
162 - don't leave stale dentries around after renames (bug 2428)
163 - fix timeouts when evicting a client with a single lock held (2642)
164 - set deadline for the initial HELLO message to drain (2634)
165 - print out dotted-quad IP addresses in the socknal (2302)
167 - additional debugging for MDS client eviction problem (2443)
168 - fix mkfsoptions support for osts (2603, 2604)
170 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
173 - add series for the vanilla 2.6.0 kernel
174 - add series for the vanilla 2.4.24 kernel
175 - add series for a cray x86/64 UL kernel drop
176 - fix xattr patches for the vanilla 2.4.19 series
178 - generate true UUIDs in lmc (1171)
179 - have portals stack dumping break in UML (2466)
180 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
181 - allocate with _NOFS in ldlm to avoid deadlock (1933)
182 - wake callback waiting threads on client eviction (2460)
183 - Add --ptldebug and --subsystem to lmc (1719)
184 - update assertion to allow safe interrupt allocation
185 - set rq_no_resend for cancel requests (2432)
186 - recalculate ptlrpcd timeout after resend (2494)
187 - call vfs_rmdir when removing pending directories (2368)
188 - fix renaming a file to itself (2429)
189 - lmc creates a default one-stripe lov (2454)
190 - expand procfs space to handle large clusters (2326)
191 - increase UML stack to avoid overflow
192 - update lconf's list of debug and subsystem masks
193 - fix lfs find --obd (2510)
194 - /proc tunable for disabling filter read caching (2591)
195 - stop rpm packages from altering slapd.conf (2301)
196 - disable nagle in the socknal under 0conf (2578)
197 - choose mds inode size based on stripe count (2572)
198 - fix kernel-source rpm problems (2516)
200 - add --disable-doc to avoid pdf generation (2421)
201 - update documentation, tests, type-os, comments
202 - avoid format warnings on ia64
204 - tiny code cleanups by removing unused fields
206 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
209 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
210 - remove the most blatant lies from BUILDING file (2371)
211 - change default debug level to reasonable production setting
212 - reduce client side cache size to reduce cache flush time
213 - reduce max RPCs in flight to avoid unnecessary file fragmentation
214 - make TCP zerocopy and pinger support enabled by default (2476)
215 - sync writes completed after process exits caused crashes (2319)
216 - maintain correct mount count on the MDS (2356)
217 - backout 1557, because 2316 wasn't really fixed
218 - better file I/O statistics gathering in /proc
219 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
220 - another kernel patch to fix zero-copy TCP function export
221 - don't take duplicate lock when processing re-sent getattr (2420)
222 - lctl uses obd_self_export instead of creating new conn (2353)
223 - MDS/OST recovery case which requires object creation asserted (2425)
224 - move lfs from /usr/sbin to /usr/bin in packages
225 - fix race between mds_client_add and mds_client_free (2417)
226 - use kmalloc instead of slabs in portals (2430)
227 - don't create duplicate records when a failover MDS is present (2442)
228 - remove unnecessary mount age check (2332)
229 - don't remove directory inodes from locks prematurely (2451)
230 - don't break if MDS service name is the same as hostname (2103)
231 - fix races in client write RPC generation when cache full (2482)
233 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
236 - remove now-unused request->rq_obd (278)
237 - if an allocation fails, print out how much memory we've used (1933)
238 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
239 - add the "configurable stack size" patch to most series files (1256)
240 - ability to write large log records, for 100+ OST configs (2306)
241 - fix NULL deref when filter_prep fails (2314)
242 - fix operator precedence error in filter_sync
243 - dynamic allocation of socknal TX descriptors (2315)
244 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
245 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
246 - ensure that CWARN messages in Portals always get to the syslog
247 - __init/__exit are not for prototype decls (ldlm_init/exit)
248 - x86-64 compile warning fixes
249 - fix gateway LMC keyword conflict (2318)
250 - fix MDS lock inversions in getattr/reint paths (1844)
251 - fix a rare lock re-ordering bug, which caused deadlock (2322)
252 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
253 - DLM race condition prevented some lock evictions (2328)
254 - ENOMEM detection and retry on socknal sends (2230)
255 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
256 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
257 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
258 - better error messages when a client is rejected during recovery (1505)
259 - avoid cancelling locks which were never granted, after failure (2330)
260 - fix i_sem/journal inversion in mds_client_add (2333)
261 - fix truncate/getattr lock cycle deadlock (2334)
262 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
263 - fix two rare exit paths which could leak an l_lock() ref (2321)
264 - fencepost error in MDS/OST orphan recovery (2226)
265 - make log record alignment 8 bytes (1988)
266 - lstripe now fails when requested offset > ost_count (2237)
267 - ensure that all kernel series have a complete list.h (1607)
268 - fix crashes in special-file operations (2316)
269 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
271 - allow configurable automake binary, for testing new versions
272 - small update to the lfs documentation
274 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
276 * fix negative export reference count in fsfilt_sync (2312)
278 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
279 * release candidate 0.9.1
281 - orphans are moved into the PENDING directory for possible recovery
282 - replayed opens now open by fid for orphan/rename safety (1042)
283 - last close of an orphan inode generates a transno (683)
284 - chdir() and mount() now pin the directory entry (1020)
285 - avoid CERROR in normal ll_setattr_raw() error case (1500)
286 - discard very old requests without processing them (1502)
287 - remove some common, well-understood CERRORs (1505)
288 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
289 - clear "grant" flags in OST replies until OST grant code lands (1644)
290 - fix read performance by not clobbering i_blksize on client (1598)
291 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
292 - make LRU size a /proc tunable, clears locks when reduced (707)
293 - fix some lprocfs rot that prevented ptlbd from loading (1732)
294 - server locks take references on exports now (1558)
295 - build fixes for 2.4.20-rh trees (1663)
296 - return an error from lov_create if all OSCs are inactive (1751)
297 - fix import levels when a reconnect happens without a timeout (1597)
298 - exit early from mds_open if we get a lookup error (1749)
299 - partial page read at EOF wouldn't wait for disk before sending (1642)
300 - avoid NULL deref in obdfilter when reading page past EOF (1592)
301 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
302 - fix LBUG in ll_it_open_error with rc = -2 (1861)
303 - write/truncate lock inversion (1639)
304 - Don't auto-load obdclass, portals modules during cleanup (1495)
305 - fix timestamps from jumping to "now" (1763)
306 - extra journal assertions (1648)
307 - add an extra multiunlink test (1771)
308 - fix read_record/write_record API (1776)
309 - fix leak of offset_extent, possible incorrect i_size later (1772)
310 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
311 - Debugging for the kqswnal_get_idle_tx problems (1820)
312 - Allow recovery to be attempted multiple times (1536)
313 - Write out MDS last_rcvd file after it is first created (1600)
314 - Fix tx_descriptor leak in failed transmit situations (1827)
315 - ext3 journaling fixes for assertion failure after IO error (1871)
316 - class_export_put() on freed export after completion AST error (1896)
317 - Fix revalidate looping in VFS (1322)
318 - Don't access a freed export during MDS_REINT timeout (1521)
319 - Add open-unlink recovery support on the MDS (1673,1764)
320 - Return an error if no MDS data was read from last_rcvd (1946)
321 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
322 - Don't setup a disk device that doesn't match exported UUID (317)
323 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
324 - avoid committing NULL handle in force close
325 - local.sh is now a one-stripe LOV configuration
326 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
327 - fix ext3 htree duplicate directory entry corruption (1516)
328 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
329 - update to new LOV EA format (2097)
330 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
332 2003-06-15 Phil Schwan <phil@clusterfs.com>
335 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
336 - per-import/export recovery handling (958, 931, 959)
337 - multiple last-rcvd slots, for serving multiple FSes (949)
338 - connections are again shared between multiple imp/exports (963, 964)
339 - "umount -f" would hang if any requests needed to be sent (393, 978)
340 - avoid pinning large req buffer by copying for queued messages (989)
341 - add "uuid" to "lctl device" command to help upcalls (991)
342 - "open" RPCs with transnos would confuse recovery counters (1037)
343 - do proper endian conversion of all wire messages (288, 340, 891)
344 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
345 - call waiting locks callback from LDLM recovery thread (1127, 1151)
346 - fix ptlrpc_connection leak in target_handle_connect (1174)
347 - fix import refcounting bug in OST and MDS cleanup (1134)
348 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
349 - fix very unlikely obd_types race condition (501)
350 - remove osc_open hack for echo_client (1187)
351 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
352 - a failure in read_inode2 leads to deadlock (1139)
353 - cancel ack-locks as soon as transaction is committed (1072)
354 - fix major leaks and crashes in the bulk I/O path (937, 1057)
355 - make sure to commitrw after any preprw to avoid deadlock (1162)
356 - failing to execute a file in a lustre FS would lock inode (1203)
357 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
358 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
359 - obd_uuid2tgt would walk off the end of the list (1255)
360 - on IA64 the osc would give portals incorrect bulk size (1258)
361 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
362 - fix lock inversion caused by new llite matching code (1282)
363 - limit the number of dirty pages on a client to 10MB (1286)
364 - timed out locks were not being corrected cancelled (1289)
365 - fix O_DIRECT above 4GB on IA-32 (1292)
366 * major user-visible changes
367 - fail out/fail over policy now controlled by the upcall (993)
369 - add OBD_PING to check server availability and failure (954)
370 - lustre messages are now sent in sending host order (288, 340, 891)
371 - add eadatalen to MDS getattr reply (340)
372 - OST read replies may contain second buffer, with per-page status (593)
374 2003-03-11 Phil Schwan <phil@clusterfs.com>
377 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
378 - failed open()s could cause deadlock; fixed (867, 869)
379 - stop cancelling OST locks when files are closed (481)
380 - overlapping XID spaces caused network corruption (851, 853)
381 - fix unsafe fsfilt counter arithmetic; change to atomic_t
382 - setattr_raw added, to do single-RPC, server-side setattrs
383 - lmc/lconf syntax change for OST UUIDs
384 - fix crashy race condition between ptlrpc_free_req and osc_close
385 - don't use request in mdc_enqueue if we hit a timeout (889)
386 - don't set the inode i_size for regular files from the MDS (896)
387 - handle out of order completion AST (842)
388 - don't LBUG if a lock request times out after receiving AST (913)
389 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
390 - if a bad lock AST arrives, send an error instead of dropping entirely
391 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
392 - fix leak in bulk IO when only partially completed (899, 900, 926)
393 - fix O_DIRECT for ia64 (55)
394 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
395 - C-z after timeout could hang a process forever; fixed (977)
397 - client-side I/O cache (678, 924, 929, 941, 970)
399 - READPAGE and SETATTRs which don't take server-side locks get
402 2003-02-11 Phil Schwan <phil@clusterfs.com>
405 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
406 - Fix fsfilt_extN_readpage() to read a full page of directory
407 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
408 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
409 - Fix imbalanced LOV object allocation and out-of-bound access (469)
410 - Most intent operations were removed, in favour of a new RPC mode
411 that does a single RPC to the server and bypasses most of the VFS
412 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
413 - Aggressively cancel local locks on DLM servers
414 - mds_reint_unlink sends EA to the client if it's the last nlink.
415 client uses that EA to unlink OST objects.
416 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
417 - recursive symlinks were fixed (439)
418 - fixed NULL deref in DEBUG_REQ
419 - filter_update_lastobjid no longer calls sync, which annoyed extN
420 - fixed multi-client small-writes to a single file problem (445)
421 - fixed mtime updates during file writes (607)
422 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
423 - fixed bug in obd_brw_read/write() (under guise of testing 367)
424 - fixed Linux OST size reporting problem (444, 656)
425 - OST now updates object mtime with writes or setattr (607, 619)
426 - client verifies file size before zeroing page past EOF (445)
427 - OST now writes last allocated objid to disk with allocation (108)
428 - LOV on echo now works (409)
430 - mds_reint_unlink sends a new buffer, with the EA included. this
431 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
432 set if a regular file was being unlinked, and it was the last link
433 - use PtlGet from the target for bulk writes (315)
434 - OST now updates object mtime with writes or setattr (607, 619)
435 - LDLM now has a grant-time callback to revalidate locked items, if
437 - Many MDS operations were reorganized to combat race conditions
439 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
440 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
442 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
445 - Fully reactivate OST imports after reconnection (512, others)
446 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
447 - More graceful error handling for truncating on dead OST (515)
448 - Don't error out unless we're actually accessing dead stripes (474)
449 - Fix garbage sizes when stripes are missing (410)
450 - LRU counters were broken, causing constant lock purge (433, 432)
451 - garbage on read from stripes with failed OSTs (441)
452 - mark OSCs as active before reconnecting during recovery (438)
453 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
454 - lfind did not preserve OST order in output (443)
455 - symlinks cause hung clients, incorrect data (439)
456 - stop dereferencing request after dropping refcount (457)
457 - don't LASSERT(spin_is_locked) on non-SMP (455)
458 - fixes for many rename() bugs
459 - fstat didn't correctly synchronize attributes (399)
460 - server must handle lock cancellation during blocking AST prep (487)
461 - bulk descriptors were free()d too soon (511)
462 - fix paths in lconf, which would load incorrect modules (451, 507)
463 - fix confusing lconf 'host not found' error message (386)
464 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
465 - fix race condition in mdc_blocking_ast() for inode access (526)
466 - fix lov_unpackmd() unpacking wrong number of stripes (537)
467 - fix lov_set_osc_active() marking wrong OSC inactive (440)
468 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
469 - fix multiple writes of stripe MD to MDS (358, maybe 519)
470 - fix lstripe in several ways (kernel side) (527)
471 - fix request leak in ldlm_cli_enqueue (262)
472 - incorrect OSC was marked inactive after OST failure
473 - call mds_fs_cleanup before unmounting filesystem (524)
474 - fix races between taking ns_lock and ldlm_lock_change_resource
475 - fix races updating LOV export open file list
476 - fix lov_enqueue error path, avoid decref-ing bad lock handle
477 - fix recovery NULL deref in ldlm_cli_cancel_unused
478 - fix some DLM races by using new hash table for lock handles (419)
479 - permit the client to specify desired inodes, at replay
480 - duplicate requests when we queue them for replay reintegration
481 - fix last_rcvd offset calculation
482 - sync after each recovered transaction, so we always make progress
483 - never, not always, ERESTART requests without transnos
484 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
486 - skip replay if the MDS says that the client is already connected
487 - don't check for a recovery-enabled export to match lctl's UUID
488 - don't INC_USE_COUNT for phantom exports
489 - don't crash when cleaning up phantom exports (567)
490 - don't double-finish or set replay data for errored mdc_open requests
491 - abort requests when they time out, so we don't get old replies
492 - send/receive replies for AST messages again
493 - if the client says that it doesn't have the lock, cancel it on the
495 - if we timeout during I/O, don't try to cancel an in-use lock; instead
496 mark it as destroyed, it will all work out when decref is called
497 - fix module use counts (22, 581)
499 - ASTs now expect a reply (server cancels lock on error reply)
501 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
504 - fix many simultaneous client startup (392)
505 - fix dentry->d_it clobbering
506 - credentials weren't being shipped for readdir/getattr operations
507 - remove invalid assertions triggered during some concurrent MD
509 - proper Lustre versions added (336, 389)
510 - fix memory leak for create error case (398)
511 - fix LOV locking bug that would get cli/srv out of sync
512 - fix echo client over LOV (409)
513 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
514 - fix double-O_EXCL intent crash (424)
515 - avoid sending multiple lock CANCELs (352)
517 - MDS can do multi-client recovery (modulo bugs in new code)
519 - many updates, edits, cleanups
521 2002-11-18 Phil Schwan <phil@clusterfs.com>
524 - fix null d_it dereference (346)
525 - fix full OST/dbench hang (333)
526 - fix permission problem with file removal (286)
527 - fix removal of OSCs from LOV when they fail
528 - fix NULL deref during bulk timeout (214)
529 - fix problems related to multiple filesystems on one MDS (241)
530 - fixed serious subtle metadata locking bugs
531 - free locks on clients when inodes are removed due to memory
533 - fix inode pointer in lock data (285)
534 - partial support for multiple MDS on a single host (241)
535 - data locks weren't cancelled at clear_inode time (290, 311)
536 - intent locks could lead to unbounded lock growth (205)
537 - added a maximum lock count, an LRU list, and a flusher
538 - fix multiple rename (365)
539 - properly abstracted the echo client
540 - OSC locked 1 byte too many; fixed
541 - rewrote brw callback code:
542 - fixed recovery bugs related to LOVs (306)
543 - fixed too-many-pages-in-one-write crash (191)
544 - fixed (again) crash in sync_io_timeout (214)
545 - probably fixed callback-related race (385)
547 - Add capability to MDS protocol
548 - LDLM cancellations and callbacks on different portals
550 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
553 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
554 - defer open object destruction to close time (601981, 138)
555 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
556 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
557 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
558 - don't call obd_setattr in ll_file_release for destroyed objects
559 * protocol change to lustre_msg: move |version| and add |flags|
560 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
561 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
562 * added replay of create, unlink, link and rename operations during
563 MDS failover; recovery should be much more robust now
564 * remove failed OSCs from LOVs (only lov_create uses this so far)
565 * the lustre-HOWTO was brought (more) up to date (582544)
567 2002-10-23 Phil Schwan <phil@clusterfs.com>
570 - in-use dentries weren't being reused properly (617851)
571 - prevent multiple LDLM setup (599178)
572 - fix LOV size calculations for truncate (617853)
573 - fix client handling of MDS intent errors (POSIX)
574 - fix permission bug in lovstripe.c test (624321)
575 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
576 - truncate past end of file could corrupt data
577 - proper cleanup after timeouts, crashes, etc (592524, 550815)
578 - a race in recovery could return ETIMEDOUT to apps (623947)
579 - building outside the source directory was fixed
580 * the lustre-HOWTO was brought (more) up to date (582544)
581 * major progress was made on recovery functionality
583 2002-10-10 Phil Schwan <phil@clusterfs.com>
586 - recovery deadlock fix
587 - rm -rf causes LBUG fix (617817)
588 - file open by multiple tasks fix (618962)
589 - directory permissions bugs (602707 and 620007)
590 - journal_stop fixed with locking (611313)
591 - O_APPEND failures resolved (618273, perhaps 614459)
592 - lconf PATH fix (619770)
593 - IA64 build fix (621450)
594 - RPC buffer sizes scale with amount of memory
596 2002-10-01 Phil Schwan <phil@clusterfs.com>
599 - locks would be cancelled without throwing away data pages,
600 resulting in inconsistent data (605627)
601 - inode attributes were not always being refreshed (605627, 612449)
602 - lconf now continues to cleanup after lctl reports an error
603 - MDS now enforces user permissions (602707)
604 - lprocfs cleanup fixed, but not yet enabled (614157)
605 - fixed infinite server hang, should a client not respond to an AST
606 - avoid going into recovery if user calls readlink() with a buffer
607 that's too small (613941)
608 - AST RPCs no longer require replies (614867) -- this may be changed
609 - don't crash server if client sends an IOV that's too big (611336)
610 - fixed lock conversion deadlock (611892)
611 - fixed the following of symlinks (614622)
612 * recovery: the server can remove locks from a client that dies, other
613 clients can make progress
614 * more extN patch fixes
615 * compile-time configurable ptlrpc buffer allocations
617 - collaborative read cache document
618 - Lustre Lite Performance CDR document-in-progress
620 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
623 - fix typo in patch-2.4.18
625 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
628 - clear ptlrpc request each time in handle_incoming_request()
629 - unlink of files now destroys the object on the OST
631 2002-09-19 Peter Braam <braam@clusterfs.com>
633 * add hard link support
634 * change obdfile creation method
635 * kernel patch changed
637 2002-09-19 Peter Braam <braam@clusterfs.com>
640 - stack overflow bug in extN fixed
642 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
644 * documentation updates
645 - add man pages for config tools
646 - update tests/README to describe testing with new config tools
647 - finish metadata API descriptions
648 * bug fixes and cleanups
649 - statfs workaround for 16TB limit
650 - LOV stripe allocation improved, can stripe on subset of OSTs
651 - LOV file size/IO offset was wrong for files > 4GB in size
652 - object EA data was being dropped, caused files to be unreadable
653 - memory overflow with non-LOV OST caused memory corruption
654 - fixed regression tests to work with new config tools, obdfilter
655 - fixed bug when directory size became larger than 1 block
656 - fixed bug (for single client case) when PWD was deleted
657 - invalidate local directory pages when doing intent-based ops
658 - avoid LDLM oops when lock callback contained bad data
660 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
662 * documentation updates
663 * bug fixes and cleanups
664 - configuration tools
667 - 64-bit compile warnings
668 - 64-bit internal statfs data
670 * test_brw on persistent OST devices
674 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
676 * documentation updates
677 * bug fixes and cleanups
678 * configuration tools
680 2002-08-30 Peter J. Braam <braam@clusterfs.com>
683 * many small fixes to 0_5_4
684 * io/network handling
685 * thinkos in MDS operations
687 2002-08-24 Peter J. Braam <braam@clusterfs.com>
690 * crucial basic fixes to 0.5.3
691 * IOR, Iozone work over Elan
694 2002-08-07 Phil Schwan <phil@clusterfs.com>
695 * version 0_5_3, our first alpha
696 * we use the new Portals iovs
697 * documentation updates
698 * bug fixes and cleanups
699 * small changes in the DLM wire protocol
701 2002-07-25 Peter J. Braam <braam@clusterfs.com>
702 * version 0_5_1 with some initial stability,
703 * locking on MD and file I/O.
704 * documentation updates
705 * several bug fixes since 0.5.0
706 * small changes in wire protocol
708 2002-07-18 Phil Schwan <phil@clusterfs.com>
710 * delivered as Lustre Light Alpha
711 * fixed a crash after handling invalid MDS requests
712 * fixed directory pages for architectures with non-4k pages sizes
714 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
715 * release version v0_4_4
716 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
718 * Fixes a number of interruption problems with OST operations.
719 * Update documentation for portals header changes
720 * Move all wire protocol structs/defines to lustre_idl.h
721 * Fixes symlink length bug.
722 * Add tcpdump to repository.
724 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
725 * release version v0_4_3
726 * Fixes statfs for inodes on extN.
727 * Fixes bug in runtests which would delete /etc/hosts.
728 * Use 64-bit object IDs wherever possible (not into VFS though)
729 Remove ost_get_info, which is unused by lustre, and out of date.
731 2002-07-03 Peter Braam <braam@clusterfs.com>
732 * release version v0_4_2 Fixes a lookup error (type not passed)
733 * move forward to head of Portals
734 * move forward to latest Lustre kernel
736 2002-06-25 Peter Braam <braam@clusterfs.com>
737 * release version v0_4_1. Hopefully stable on single node use.