1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4 - clean kiobufs before and after use (3485)
5 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
7 - change init script to fail more clearly if not run as root (1528)
8 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
9 - don't do DNS lookups on NIDs too small for IP addresses (3442)
10 - dynamic ptlrpc request buffer allocation (2102)
11 - don't allow unlinking open directory if it isn't empty (2904)
12 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
14 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
16 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
19 - don't copy lvb into (possibly NULL) reply on error (2983)
20 - don't deref dentry after dput, don't free lvb on error (2922)
21 - use the kms to determine writeback rpc length (2947)
22 - increment oti_logcookies when osc is inactive (2948)
23 - update client's i_blocks count via lvb messages (2543)
24 - handle intent open/close of special files properly (1557)
25 - mount MDS with errors=remount-ro, like obdfilter (2009)
26 - initialize lock handle to avoid ASSERT on error cleanup (3057)
27 - don't use cancelling-locks' kms values (2947)
28 - use highest lock extent for kms, not last one (2925)
29 - don't dereference ERR_PTR() dentry in error handling path (3107)
30 - fix thread race in portals_debug_dumplog() (3122)
31 - create lprocfs device entries at setup instead of at attach (1519)
32 - common AST error handler, don't evict client on completion race (3145)
33 - zero nameidata in detach_mnt in 2.6 (3118)
34 - verify d_inode after revalidate_special is valid in 2.6 (3116)
35 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
36 - initialize RPC timeout timer earlier for 2.6 (3219)
37 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
38 - print nal/nid for unknown nid (3258)
39 - additional checks for oscc recovery before doing precreate (3284)
40 - fix ll_extent_lock() error return code for 64-bit systems (3043)
41 - don't crash in mdc_close for bad permissions on open (3285)
42 - zero i_rdev for non-device files (3147)
43 - clear page->private before handing to FS, better assertion (3119)
44 - fix incorrect decref of invalidated dentry (2350)
45 - don't hold journal transaction open across create RPC (3313)
46 - update atime on MDS at close time (3265)
47 - close LDAP connection when recovering to avoid server load (3315)
48 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
49 - don't leak open file on MDS after open resend (3325)
50 - serialize filter_precreate and filter_destroy_precreated (3329)
51 - loop device shouldn't call sync_dev() for nul device (3092)
52 - clear page cache after eviction (2766)
53 - resynchronize MDS->OST in background (2824)
54 - refuse to mount the same filesystem twice on same mountpoint (3394)
55 - allow llmount to create routes for mounting behind routers (3320)
56 - push lock cancellation to blocking thread for glimpse ASTs (3409)
57 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
58 - fix rare problem with rename on htree directories (3417)
60 - allow default OST striping configuration per directory (1414)
61 - fix compilation for qswnal for 2.6 kernels (3125)
62 - increase maximum number of MDS request buffers for large systems
63 - change liblustreapi to be useful for external progs like lfsck (3098)
64 - increase local configuration timeout for slow disks (3353)
65 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
67 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
70 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
71 - don't overwrite extent policy data in reply if lock was blocked (2901)
72 - drop filter export grants atomically with removal from device (2663)
73 - del obd_self_export from work_list in class_disconnect_exports (2908)
74 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
75 - swab reply message in mdc_close, other PPC fixes (2464)
76 - fix destroying of named logs (2325)
77 - overwrite old logs when running lconf --write_conf (2264)
78 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
79 - fix race in target_handle_connect (2898)
80 - mds_reint_create() should take same inode create lock (2926)
81 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
82 - don't close files for self_export to avoid uninitialized obd (2936)
83 - allow MDS with the same name as client node (2939)
84 - hold dentry reference for closed log files for unlink (2325)
85 - reserve space for all logs during transactions (2059)
86 - don't evict page beyond end of stripe extent (2925)
87 - don't oops on a deleted current working directory (2399)
88 - handle hard links to targets without a parent properly (2517)
89 - don't dereference NULL lock when racing during eviction (2867)
90 - don't grow lock extents when lots of conflicting locks (2919)
92 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
95 - account for cache space usage on clients to avoid data loss (974)
96 - lfsck support in lustre kernel code (2349)
97 - reduce journal credits needed for BRW writes (2370)
98 - orphan handling to avoid losing space on client/server crashes
99 - ptlrpcd can be blocked, stopping ALL progress (2477)
100 - use lock value blocks to assist in proper KMS, faster stat (1021)
101 - takes i_sem instead of DLM locks internally on obdfilter (2720)
102 - recovery for initial connections (2355)
103 - fixes for mds_cleanup_orphans (1934)
104 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
105 - block all file creations until orphan recovery completes (1901)
106 - client remove rq_connection from request struct (2423)
107 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
108 - recovery timer race (2670)
109 - mdc_close recovey bug (2532)
110 - ptlrpc cleanup bug (2710)
111 - mds timeout on local locks (2588)
112 - namespace lock held during RPCs (2431)
113 - handle interrupted sync write properly (2503)
114 - don't try to handle a message that hasn't been replied to (2699)
115 - client assert failure during cleanup after abort recovery (2701)
116 - leak mdc device after failed mount (2712)
117 - ptlrpc_check_set allows timedout requests to complete (2714)
118 - wait for inflight reqs when ptlrpcd finishes (2710)
119 - make sure unregistered services are removed from the srv_list
120 - reset bulk XID's when resending them (caught by 1138 test)
121 - unregister_bulk after timeout
122 - fix lconf error (2694)
123 - handle write after unfinished setstripe, stripe-only getstripe (2388)
124 - readahead locks pages, leaves pending causing memory pressure (2673)
125 - increase OST request buffers to 4096 on large machines (2729)
126 - fix up permission of existing directories in simple_mkdir (2661)
127 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
128 - don't assign transno to errored transactions (2742)
129 - don't delete objects on OST if given a bogus objid from MDS (2751)
130 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
131 - if rq_no_resend, then timeout request after recovery (2432)
132 - fix MDS llog_logid record size, 64-bit array alignment (2733)
133 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
134 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
135 - MDS recovery shouldn't delete objects that it has given out (2730)
136 - if enqueue arrives after completion, don't clobber LVB (2819)
137 - don't unlock pages twice when trigger_group_io returns error (2814)
138 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
139 - don't write pages to disk if there was an error (1450)
140 - don't ping imports that have recovery disabled (2676)
141 - take buffered bytes into account when balancing socknal conn (2817)
142 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
143 - reconnect unlink llog connection after MDS reconnects to OST (2816)
144 - remove little-endian swabbing of llog records (1987)
145 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
146 - retry reposting request buffers if they fail (1191)
147 - grow extent at grant time to avoid granting a revoked lock (2809)
148 - lock revoke doesn't evict page if covered by a second lock (2765)
149 - disable VM readahead to avoid reading outside lock extents (2805)
151 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
152 - updated kernel patches for hp-2.4.20 kernel (2681)
154 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
157 - fix truncated write corruption (2366)
158 - fix for failed assertion in iopen_connect_dentry (1792,2517)
160 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
161 - ensure len(uuid) < 37 in lmc (1171)
162 - fix ia64 OOPS in llog_test (2255)
163 - zero end of page at obdfilter for partial page writes (2648)
164 - don't leave stale dentries around after renames (bug 2428)
165 - fix timeouts when evicting a client with a single lock held (2642)
166 - set deadline for the initial HELLO message to drain (2634)
167 - print out dotted-quad IP addresses in the socknal (2302)
169 - additional debugging for MDS client eviction problem (2443)
170 - fix mkfsoptions support for osts (2603, 2604)
172 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
175 - add series for the vanilla 2.6.0 kernel
176 - add series for the vanilla 2.4.24 kernel
177 - add series for a cray x86/64 UL kernel drop
178 - fix xattr patches for the vanilla 2.4.19 series
180 - generate true UUIDs in lmc (1171)
181 - have portals stack dumping break in UML (2466)
182 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
183 - allocate with _NOFS in ldlm to avoid deadlock (1933)
184 - wake callback waiting threads on client eviction (2460)
185 - Add --ptldebug and --subsystem to lmc (1719)
186 - update assertion to allow safe interrupt allocation
187 - set rq_no_resend for cancel requests (2432)
188 - recalculate ptlrpcd timeout after resend (2494)
189 - call vfs_rmdir when removing pending directories (2368)
190 - fix renaming a file to itself (2429)
191 - lmc creates a default one-stripe lov (2454)
192 - expand procfs space to handle large clusters (2326)
193 - increase UML stack to avoid overflow
194 - update lconf's list of debug and subsystem masks
195 - fix lfs find --obd (2510)
196 - /proc tunable for disabling filter read caching (2591)
197 - stop rpm packages from altering slapd.conf (2301)
198 - disable nagle in the socknal under 0conf (2578)
199 - choose mds inode size based on stripe count (2572)
200 - fix kernel-source rpm problems (2516)
202 - add --disable-doc to avoid pdf generation (2421)
203 - update documentation, tests, type-os, comments
204 - avoid format warnings on ia64
206 - tiny code cleanups by removing unused fields
208 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
211 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
212 - remove the most blatant lies from BUILDING file (2371)
213 - change default debug level to reasonable production setting
214 - reduce client side cache size to reduce cache flush time
215 - reduce max RPCs in flight to avoid unnecessary file fragmentation
216 - make TCP zerocopy and pinger support enabled by default (2476)
217 - sync writes completed after process exits caused crashes (2319)
218 - maintain correct mount count on the MDS (2356)
219 - backout 1557, because 2316 wasn't really fixed
220 - better file I/O statistics gathering in /proc
221 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
222 - another kernel patch to fix zero-copy TCP function export
223 - don't take duplicate lock when processing re-sent getattr (2420)
224 - lctl uses obd_self_export instead of creating new conn (2353)
225 - MDS/OST recovery case which requires object creation asserted (2425)
226 - move lfs from /usr/sbin to /usr/bin in packages
227 - fix race between mds_client_add and mds_client_free (2417)
228 - use kmalloc instead of slabs in portals (2430)
229 - don't create duplicate records when a failover MDS is present (2442)
230 - remove unnecessary mount age check (2332)
231 - don't remove directory inodes from locks prematurely (2451)
232 - don't break if MDS service name is the same as hostname (2103)
233 - fix races in client write RPC generation when cache full (2482)
235 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
238 - remove now-unused request->rq_obd (278)
239 - if an allocation fails, print out how much memory we've used (1933)
240 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
241 - add the "configurable stack size" patch to most series files (1256)
242 - ability to write large log records, for 100+ OST configs (2306)
243 - fix NULL deref when filter_prep fails (2314)
244 - fix operator precedence error in filter_sync
245 - dynamic allocation of socknal TX descriptors (2315)
246 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
247 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
248 - ensure that CWARN messages in Portals always get to the syslog
249 - __init/__exit are not for prototype decls (ldlm_init/exit)
250 - x86-64 compile warning fixes
251 - fix gateway LMC keyword conflict (2318)
252 - fix MDS lock inversions in getattr/reint paths (1844)
253 - fix a rare lock re-ordering bug, which caused deadlock (2322)
254 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
255 - DLM race condition prevented some lock evictions (2328)
256 - ENOMEM detection and retry on socknal sends (2230)
257 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
258 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
259 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
260 - better error messages when a client is rejected during recovery (1505)
261 - avoid cancelling locks which were never granted, after failure (2330)
262 - fix i_sem/journal inversion in mds_client_add (2333)
263 - fix truncate/getattr lock cycle deadlock (2334)
264 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
265 - fix two rare exit paths which could leak an l_lock() ref (2321)
266 - fencepost error in MDS/OST orphan recovery (2226)
267 - make log record alignment 8 bytes (1988)
268 - lstripe now fails when requested offset > ost_count (2237)
269 - ensure that all kernel series have a complete list.h (1607)
270 - fix crashes in special-file operations (2316)
271 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
273 - allow configurable automake binary, for testing new versions
274 - small update to the lfs documentation
276 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
278 * fix negative export reference count in fsfilt_sync (2312)
280 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
281 * release candidate 0.9.1
283 - orphans are moved into the PENDING directory for possible recovery
284 - replayed opens now open by fid for orphan/rename safety (1042)
285 - last close of an orphan inode generates a transno (683)
286 - chdir() and mount() now pin the directory entry (1020)
287 - avoid CERROR in normal ll_setattr_raw() error case (1500)
288 - discard very old requests without processing them (1502)
289 - remove some common, well-understood CERRORs (1505)
290 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
291 - clear "grant" flags in OST replies until OST grant code lands (1644)
292 - fix read performance by not clobbering i_blksize on client (1598)
293 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
294 - make LRU size a /proc tunable, clears locks when reduced (707)
295 - fix some lprocfs rot that prevented ptlbd from loading (1732)
296 - server locks take references on exports now (1558)
297 - build fixes for 2.4.20-rh trees (1663)
298 - return an error from lov_create if all OSCs are inactive (1751)
299 - fix import levels when a reconnect happens without a timeout (1597)
300 - exit early from mds_open if we get a lookup error (1749)
301 - partial page read at EOF wouldn't wait for disk before sending (1642)
302 - avoid NULL deref in obdfilter when reading page past EOF (1592)
303 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
304 - fix LBUG in ll_it_open_error with rc = -2 (1861)
305 - write/truncate lock inversion (1639)
306 - Don't auto-load obdclass, portals modules during cleanup (1495)
307 - fix timestamps from jumping to "now" (1763)
308 - extra journal assertions (1648)
309 - add an extra multiunlink test (1771)
310 - fix read_record/write_record API (1776)
311 - fix leak of offset_extent, possible incorrect i_size later (1772)
312 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
313 - Debugging for the kqswnal_get_idle_tx problems (1820)
314 - Allow recovery to be attempted multiple times (1536)
315 - Write out MDS last_rcvd file after it is first created (1600)
316 - Fix tx_descriptor leak in failed transmit situations (1827)
317 - ext3 journaling fixes for assertion failure after IO error (1871)
318 - class_export_put() on freed export after completion AST error (1896)
319 - Fix revalidate looping in VFS (1322)
320 - Don't access a freed export during MDS_REINT timeout (1521)
321 - Add open-unlink recovery support on the MDS (1673,1764)
322 - Return an error if no MDS data was read from last_rcvd (1946)
323 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
324 - Don't setup a disk device that doesn't match exported UUID (317)
325 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
326 - avoid committing NULL handle in force close
327 - local.sh is now a one-stripe LOV configuration
328 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
329 - fix ext3 htree duplicate directory entry corruption (1516)
330 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
331 - update to new LOV EA format (2097)
332 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
334 2003-06-15 Phil Schwan <phil@clusterfs.com>
337 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
338 - per-import/export recovery handling (958, 931, 959)
339 - multiple last-rcvd slots, for serving multiple FSes (949)
340 - connections are again shared between multiple imp/exports (963, 964)
341 - "umount -f" would hang if any requests needed to be sent (393, 978)
342 - avoid pinning large req buffer by copying for queued messages (989)
343 - add "uuid" to "lctl device" command to help upcalls (991)
344 - "open" RPCs with transnos would confuse recovery counters (1037)
345 - do proper endian conversion of all wire messages (288, 340, 891)
346 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
347 - call waiting locks callback from LDLM recovery thread (1127, 1151)
348 - fix ptlrpc_connection leak in target_handle_connect (1174)
349 - fix import refcounting bug in OST and MDS cleanup (1134)
350 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
351 - fix very unlikely obd_types race condition (501)
352 - remove osc_open hack for echo_client (1187)
353 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
354 - a failure in read_inode2 leads to deadlock (1139)
355 - cancel ack-locks as soon as transaction is committed (1072)
356 - fix major leaks and crashes in the bulk I/O path (937, 1057)
357 - make sure to commitrw after any preprw to avoid deadlock (1162)
358 - failing to execute a file in a lustre FS would lock inode (1203)
359 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
360 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
361 - obd_uuid2tgt would walk off the end of the list (1255)
362 - on IA64 the osc would give portals incorrect bulk size (1258)
363 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
364 - fix lock inversion caused by new llite matching code (1282)
365 - limit the number of dirty pages on a client to 10MB (1286)
366 - timed out locks were not being corrected cancelled (1289)
367 - fix O_DIRECT above 4GB on IA-32 (1292)
368 * major user-visible changes
369 - fail out/fail over policy now controlled by the upcall (993)
371 - add OBD_PING to check server availability and failure (954)
372 - lustre messages are now sent in sending host order (288, 340, 891)
373 - add eadatalen to MDS getattr reply (340)
374 - OST read replies may contain second buffer, with per-page status (593)
376 2003-03-11 Phil Schwan <phil@clusterfs.com>
379 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
380 - failed open()s could cause deadlock; fixed (867, 869)
381 - stop cancelling OST locks when files are closed (481)
382 - overlapping XID spaces caused network corruption (851, 853)
383 - fix unsafe fsfilt counter arithmetic; change to atomic_t
384 - setattr_raw added, to do single-RPC, server-side setattrs
385 - lmc/lconf syntax change for OST UUIDs
386 - fix crashy race condition between ptlrpc_free_req and osc_close
387 - don't use request in mdc_enqueue if we hit a timeout (889)
388 - don't set the inode i_size for regular files from the MDS (896)
389 - handle out of order completion AST (842)
390 - don't LBUG if a lock request times out after receiving AST (913)
391 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
392 - if a bad lock AST arrives, send an error instead of dropping entirely
393 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
394 - fix leak in bulk IO when only partially completed (899, 900, 926)
395 - fix O_DIRECT for ia64 (55)
396 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
397 - C-z after timeout could hang a process forever; fixed (977)
399 - client-side I/O cache (678, 924, 929, 941, 970)
401 - READPAGE and SETATTRs which don't take server-side locks get
404 2003-02-11 Phil Schwan <phil@clusterfs.com>
407 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
408 - Fix fsfilt_extN_readpage() to read a full page of directory
409 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
410 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
411 - Fix imbalanced LOV object allocation and out-of-bound access (469)
412 - Most intent operations were removed, in favour of a new RPC mode
413 that does a single RPC to the server and bypasses most of the VFS
414 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
415 - Aggressively cancel local locks on DLM servers
416 - mds_reint_unlink sends EA to the client if it's the last nlink.
417 client uses that EA to unlink OST objects.
418 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
419 - recursive symlinks were fixed (439)
420 - fixed NULL deref in DEBUG_REQ
421 - filter_update_lastobjid no longer calls sync, which annoyed extN
422 - fixed multi-client small-writes to a single file problem (445)
423 - fixed mtime updates during file writes (607)
424 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
425 - fixed bug in obd_brw_read/write() (under guise of testing 367)
426 - fixed Linux OST size reporting problem (444, 656)
427 - OST now updates object mtime with writes or setattr (607, 619)
428 - client verifies file size before zeroing page past EOF (445)
429 - OST now writes last allocated objid to disk with allocation (108)
430 - LOV on echo now works (409)
432 - mds_reint_unlink sends a new buffer, with the EA included. this
433 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
434 set if a regular file was being unlinked, and it was the last link
435 - use PtlGet from the target for bulk writes (315)
436 - OST now updates object mtime with writes or setattr (607, 619)
437 - LDLM now has a grant-time callback to revalidate locked items, if
439 - Many MDS operations were reorganized to combat race conditions
441 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
442 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
444 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
447 - Fully reactivate OST imports after reconnection (512, others)
448 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
449 - More graceful error handling for truncating on dead OST (515)
450 - Don't error out unless we're actually accessing dead stripes (474)
451 - Fix garbage sizes when stripes are missing (410)
452 - LRU counters were broken, causing constant lock purge (433, 432)
453 - garbage on read from stripes with failed OSTs (441)
454 - mark OSCs as active before reconnecting during recovery (438)
455 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
456 - lfind did not preserve OST order in output (443)
457 - symlinks cause hung clients, incorrect data (439)
458 - stop dereferencing request after dropping refcount (457)
459 - don't LASSERT(spin_is_locked) on non-SMP (455)
460 - fixes for many rename() bugs
461 - fstat didn't correctly synchronize attributes (399)
462 - server must handle lock cancellation during blocking AST prep (487)
463 - bulk descriptors were free()d too soon (511)
464 - fix paths in lconf, which would load incorrect modules (451, 507)
465 - fix confusing lconf 'host not found' error message (386)
466 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
467 - fix race condition in mdc_blocking_ast() for inode access (526)
468 - fix lov_unpackmd() unpacking wrong number of stripes (537)
469 - fix lov_set_osc_active() marking wrong OSC inactive (440)
470 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
471 - fix multiple writes of stripe MD to MDS (358, maybe 519)
472 - fix lstripe in several ways (kernel side) (527)
473 - fix request leak in ldlm_cli_enqueue (262)
474 - incorrect OSC was marked inactive after OST failure
475 - call mds_fs_cleanup before unmounting filesystem (524)
476 - fix races between taking ns_lock and ldlm_lock_change_resource
477 - fix races updating LOV export open file list
478 - fix lov_enqueue error path, avoid decref-ing bad lock handle
479 - fix recovery NULL deref in ldlm_cli_cancel_unused
480 - fix some DLM races by using new hash table for lock handles (419)
481 - permit the client to specify desired inodes, at replay
482 - duplicate requests when we queue them for replay reintegration
483 - fix last_rcvd offset calculation
484 - sync after each recovered transaction, so we always make progress
485 - never, not always, ERESTART requests without transnos
486 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
488 - skip replay if the MDS says that the client is already connected
489 - don't check for a recovery-enabled export to match lctl's UUID
490 - don't INC_USE_COUNT for phantom exports
491 - don't crash when cleaning up phantom exports (567)
492 - don't double-finish or set replay data for errored mdc_open requests
493 - abort requests when they time out, so we don't get old replies
494 - send/receive replies for AST messages again
495 - if the client says that it doesn't have the lock, cancel it on the
497 - if we timeout during I/O, don't try to cancel an in-use lock; instead
498 mark it as destroyed, it will all work out when decref is called
499 - fix module use counts (22, 581)
501 - ASTs now expect a reply (server cancels lock on error reply)
503 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
506 - fix many simultaneous client startup (392)
507 - fix dentry->d_it clobbering
508 - credentials weren't being shipped for readdir/getattr operations
509 - remove invalid assertions triggered during some concurrent MD
511 - proper Lustre versions added (336, 389)
512 - fix memory leak for create error case (398)
513 - fix LOV locking bug that would get cli/srv out of sync
514 - fix echo client over LOV (409)
515 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
516 - fix double-O_EXCL intent crash (424)
517 - avoid sending multiple lock CANCELs (352)
519 - MDS can do multi-client recovery (modulo bugs in new code)
521 - many updates, edits, cleanups
523 2002-11-18 Phil Schwan <phil@clusterfs.com>
526 - fix null d_it dereference (346)
527 - fix full OST/dbench hang (333)
528 - fix permission problem with file removal (286)
529 - fix removal of OSCs from LOV when they fail
530 - fix NULL deref during bulk timeout (214)
531 - fix problems related to multiple filesystems on one MDS (241)
532 - fixed serious subtle metadata locking bugs
533 - free locks on clients when inodes are removed due to memory
535 - fix inode pointer in lock data (285)
536 - partial support for multiple MDS on a single host (241)
537 - data locks weren't cancelled at clear_inode time (290, 311)
538 - intent locks could lead to unbounded lock growth (205)
539 - added a maximum lock count, an LRU list, and a flusher
540 - fix multiple rename (365)
541 - properly abstracted the echo client
542 - OSC locked 1 byte too many; fixed
543 - rewrote brw callback code:
544 - fixed recovery bugs related to LOVs (306)
545 - fixed too-many-pages-in-one-write crash (191)
546 - fixed (again) crash in sync_io_timeout (214)
547 - probably fixed callback-related race (385)
549 - Add capability to MDS protocol
550 - LDLM cancellations and callbacks on different portals
552 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
555 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
556 - defer open object destruction to close time (601981, 138)
557 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
558 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
559 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
560 - don't call obd_setattr in ll_file_release for destroyed objects
561 * protocol change to lustre_msg: move |version| and add |flags|
562 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
563 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
564 * added replay of create, unlink, link and rename operations during
565 MDS failover; recovery should be much more robust now
566 * remove failed OSCs from LOVs (only lov_create uses this so far)
567 * the lustre-HOWTO was brought (more) up to date (582544)
569 2002-10-23 Phil Schwan <phil@clusterfs.com>
572 - in-use dentries weren't being reused properly (617851)
573 - prevent multiple LDLM setup (599178)
574 - fix LOV size calculations for truncate (617853)
575 - fix client handling of MDS intent errors (POSIX)
576 - fix permission bug in lovstripe.c test (624321)
577 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
578 - truncate past end of file could corrupt data
579 - proper cleanup after timeouts, crashes, etc (592524, 550815)
580 - a race in recovery could return ETIMEDOUT to apps (623947)
581 - building outside the source directory was fixed
582 * the lustre-HOWTO was brought (more) up to date (582544)
583 * major progress was made on recovery functionality
585 2002-10-10 Phil Schwan <phil@clusterfs.com>
588 - recovery deadlock fix
589 - rm -rf causes LBUG fix (617817)
590 - file open by multiple tasks fix (618962)
591 - directory permissions bugs (602707 and 620007)
592 - journal_stop fixed with locking (611313)
593 - O_APPEND failures resolved (618273, perhaps 614459)
594 - lconf PATH fix (619770)
595 - IA64 build fix (621450)
596 - RPC buffer sizes scale with amount of memory
598 2002-10-01 Phil Schwan <phil@clusterfs.com>
601 - locks would be cancelled without throwing away data pages,
602 resulting in inconsistent data (605627)
603 - inode attributes were not always being refreshed (605627, 612449)
604 - lconf now continues to cleanup after lctl reports an error
605 - MDS now enforces user permissions (602707)
606 - lprocfs cleanup fixed, but not yet enabled (614157)
607 - fixed infinite server hang, should a client not respond to an AST
608 - avoid going into recovery if user calls readlink() with a buffer
609 that's too small (613941)
610 - AST RPCs no longer require replies (614867) -- this may be changed
611 - don't crash server if client sends an IOV that's too big (611336)
612 - fixed lock conversion deadlock (611892)
613 - fixed the following of symlinks (614622)
614 * recovery: the server can remove locks from a client that dies, other
615 clients can make progress
616 * more extN patch fixes
617 * compile-time configurable ptlrpc buffer allocations
619 - collaborative read cache document
620 - Lustre Lite Performance CDR document-in-progress
622 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
625 - fix typo in patch-2.4.18
627 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
630 - clear ptlrpc request each time in handle_incoming_request()
631 - unlink of files now destroys the object on the OST
633 2002-09-19 Peter Braam <braam@clusterfs.com>
635 * add hard link support
636 * change obdfile creation method
637 * kernel patch changed
639 2002-09-19 Peter Braam <braam@clusterfs.com>
642 - stack overflow bug in extN fixed
644 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
646 * documentation updates
647 - add man pages for config tools
648 - update tests/README to describe testing with new config tools
649 - finish metadata API descriptions
650 * bug fixes and cleanups
651 - statfs workaround for 16TB limit
652 - LOV stripe allocation improved, can stripe on subset of OSTs
653 - LOV file size/IO offset was wrong for files > 4GB in size
654 - object EA data was being dropped, caused files to be unreadable
655 - memory overflow with non-LOV OST caused memory corruption
656 - fixed regression tests to work with new config tools, obdfilter
657 - fixed bug when directory size became larger than 1 block
658 - fixed bug (for single client case) when PWD was deleted
659 - invalidate local directory pages when doing intent-based ops
660 - avoid LDLM oops when lock callback contained bad data
662 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
664 * documentation updates
665 * bug fixes and cleanups
666 - configuration tools
669 - 64-bit compile warnings
670 - 64-bit internal statfs data
672 * test_brw on persistent OST devices
676 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
678 * documentation updates
679 * bug fixes and cleanups
680 * configuration tools
682 2002-08-30 Peter J. Braam <braam@clusterfs.com>
685 * many small fixes to 0_5_4
686 * io/network handling
687 * thinkos in MDS operations
689 2002-08-24 Peter J. Braam <braam@clusterfs.com>
692 * crucial basic fixes to 0.5.3
693 * IOR, Iozone work over Elan
696 2002-08-07 Phil Schwan <phil@clusterfs.com>
697 * version 0_5_3, our first alpha
698 * we use the new Portals iovs
699 * documentation updates
700 * bug fixes and cleanups
701 * small changes in the DLM wire protocol
703 2002-07-25 Peter J. Braam <braam@clusterfs.com>
704 * version 0_5_1 with some initial stability,
705 * locking on MD and file I/O.
706 * documentation updates
707 * several bug fixes since 0.5.0
708 * small changes in wire protocol
710 2002-07-18 Phil Schwan <phil@clusterfs.com>
712 * delivered as Lustre Light Alpha
713 * fixed a crash after handling invalid MDS requests
714 * fixed directory pages for architectures with non-4k pages sizes
716 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
717 * release version v0_4_4
718 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
720 * Fixes a number of interruption problems with OST operations.
721 * Update documentation for portals header changes
722 * Move all wire protocol structs/defines to lustre_idl.h
723 * Fixes symlink length bug.
724 * Add tcpdump to repository.
726 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
727 * release version v0_4_3
728 * Fixes statfs for inodes on extN.
729 * Fixes bug in runtests which would delete /etc/hosts.
730 * Use 64-bit object IDs wherever possible (not into VFS though)
731 Remove ost_get_info, which is unused by lustre, and out of date.
733 2002-07-03 Peter Braam <braam@clusterfs.com>
734 * release version v0_4_2 Fixes a lookup error (type not passed)
735 * move forward to head of Portals
736 * move forward to latest Lustre kernel
738 2002-06-25 Peter Braam <braam@clusterfs.com>
739 * release version v0_4_1. Hopefully stable on single node use.