1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4 - fixes from lustre 1.2.8
5 - print NAL number in %x format (4645)
6 - the watchdog thread now runs as interruptible (5246)
7 - drop import inflight refcount on signal_completed_replay error (5255)
9 - add pid to ldlm debugging output (4922)
10 - add --disable-server and --disable-client configure options (5786)
12 2004-10-08 Cluster File Systems, Inc. <info@clusterfs.com>
15 - properly handle portals process identifiers in messages (4165)
16 - finish default directory EA handling (3048)
17 - fixes from lustre 1.2.7
18 - removed PTL_MD_KIOV usage under CRAY_PORTALS (4420)
19 - allow EADDRNOTAVAIL as retry for connect in liblustre tcpnal (4822)
21 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
24 - many liblustre fixes
25 - fixes from lustre 1.2.6
27 - update to new libsysio-head-0806
28 - reorganization of lov code
30 2004-08-30 Cluster File Systems, Inc. <info@clusterfs.com>
33 - add locking for mmapped files (2828)
34 - lmc/lconf changes to support multiple interfaces (3376)
35 - fixes from lustre 1.2.5
37 2004-08-14 Cluster File Systems, Inc. <info@clusterfs.com>
40 - don't dereference NULL peer_ni in ldlm_handle_ast_error (3258)
41 - don't allow unlinking open directory if it isn't empty (2904)
42 - handle partial page writes in filter; fix 512b direct IO (3138)
43 - handle page cache pages in cleanup path for 2.6 (3335)
44 - leave liblustre's partial write handling to filter (3274)
45 - chose better nal ids in liblustre (3292)
46 - initialize liblustre with uid/group membership (2862)
47 - let lconf resolve symlinked-to devices (4629)
48 - balance journal closure when 2.6 filter write fails (3401)
49 - add second rpc_lock and last_rcvd info for close reqs (3462)
50 - don't hold llog sem during network request (3652)
51 - update server last transno after client disconnects (2525)
52 - replace config semaphore with spinlock (3306)
53 - ext3 exents and multi-block allocation (3024)
54 - service time statistics in /proc
55 - minor fixes to liblustre build (3317)
56 - client recovery without upcall (3262)
57 - use transno after validating reply (3892)
58 - use different name for 2nd ptlrpcd thread (3887)
59 - get a client lock in ll_inode_revalidate_it (3597)
60 - direct IO reads on OST (4048)
61 - process timed out requests if import state changes (3754)
62 - ignore -ENOENT errors in osc_destroy (3639)
63 - fixes from lustre 1.2.0-1.2.4
65 - use "CATALOGS" for the llog catalogs, not "CATLIST" (old) (b=2841)
66 - added kernel patch for /dev/sd I/O stats (4385)
68 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
71 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
72 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
73 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
74 - fix NULL dereference in /proc/sys/portals/routes (4827)
75 - allow failed mdc_close() operations to be interrupted (4561)
76 - stop precreate on OST before MDS would time out on it (4778)
77 - don't free dentries not owned by NFS code, check generation (4806)
78 - fix lsm leak if mds_create_objects() fails (4801)
79 - limit debug_daemon file size, always print CERROR messages (4789)
80 - use transno after validating reply (3892)
81 - process timed out requests if import state changes (3754)
82 - update mtime on OST during writes, return in glimpse (4829)
83 - add mkfsoptions to LDAP (4679)
84 - use ->max_readahead method instead of zapping global ra (5039)
85 - don't interrupt __l_wait_event() during strace
87 - add software watchdogs to catch hung threads quickly (4941)
88 - make lustrefs init script start after nfs is mounted
89 - fix CWARN/ERROR duplication (4930)
90 - return async write errors to application if possible (2248)
91 - update barely-supported suse-2.4.21-171 series (4842)
92 - support for sles 9 %post scripts
93 - support for building 2.6 kernel-source packages
94 - support for sles km_* packages
96 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
99 - ignore -ENOENT errors in osc_destroy (3639)
100 - notify osc create thread that OSC is being cleaned up (4600)
101 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
102 - reconstruct ost_handle() like mds_handle() (4657)
103 - create a new thread to do import eviction to avoid deadlock (3969)
104 - let lconf resolve symlinked-to devices (4629)
105 - don't unlink "objects" from directory with default EA (4554)
106 - hold socknal file ref over connect in case target is down (4394)
107 - allow more than 32000 subdirectories in a single directory (3244)
108 - OST returns ENOSPC from object create when no space left (4539)
109 - don't send truncate RPC if file size isn't changing (4410)
110 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
111 - bind to privileged port in socknal and tcpnal (3689)
113 - rate limit CERROR/CWARN console message to avoid overload (4519)
114 - basic mmap support (3918)
115 - kernel patch series update from b1_4 (4711)
117 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
120 - avoid crash during MDS cleanup with OST shut down (2775)
121 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
122 - don't use bad inodes on the MDS (3744)
123 - dynamic object preallocation to improve recovery speed (4236)
124 - don't hold spinlock over lock dumping or change debug flags (4401)
125 - don't zero obd_dev when it is force cleaned (3651)
126 - "lctl deactivate" will stop automatic recovery attempts (3406)
127 - look for existing replayed locks to avoid duplicates (3764)
128 - don't resolve lock handle twice in recovery avoiding race (4401)
129 - revalidate should check working dir is a directory (4134)
131 - don't always mark "slow" obdfilter messages as errors (4418)
133 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
136 - don't close LustreDB during write_conf until it is done (3860)
137 - fix typo in lconf for_each_profile (3821)
138 - allow dumping logs from multiple threads at one time (3820)
139 - don't allow multiple threads in OSC recovery (3812)
140 - fix debug_size parameters (3864)
141 - fix mds_postrecov to initialize import for llog ctxt (3121)
142 - replace config semaphore with spinlock (3306)
143 - be sure to send a reply for a CANCEL rpc with bad export (3863)
144 - don't allow enqueue to complete on a destroyed export (3822)
145 - down write_lock before checking llog header bitmap (3825)
146 - recover from lock replay timeout (3764)
147 - up llog sem before sending rpc (3652)
148 - reduce ns lock hold times when setting kms (3267)
149 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
150 - fix NULL deref and obd_dev leak on setup error (3312)
151 - replace some LBUG about llog ops with error handling (3841)
152 - don't match INVALID dentries from d_lookup and spin (3784)
153 - hold dcache_lock while marking dentries INVALID and hashing (4255)
154 - fix invalid assertion in ptlrpc_set_wait (3880)
156 - add libwrap support for the TCP acceptor (3996)
157 - add /proc/sys/portals/routes for non-root route listing (3994)
158 - allow setting MDS UUID in .xml (2580)
159 - print the stack of a process that LBUGs (4228)
161 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
164 - don't cleanup request in ll_file_open() on failed MDS open (3430)
165 - make sure to unset replay flag from failed open requests (3440)
166 - if default stripe count is 0, use OST count for inode size (3636)
167 - update parent mtime/ctime on client for create/unlink (2611)
168 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
169 - free recovery state on server during a forced cleanup (3571)
170 - unregister_reply for resent reqs (3063)
171 - loop back devices mounting and status check on 2.6 (3563)
172 - fix resource-creation race that can provoke i_size == 0 (3513)
173 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
174 - more debugging for page-accounting assertion (3746)
175 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
176 - avoid deadlock after precreation failure (3758)
177 - fix race and lock order deadlock in orphan handling (3450, 3750)
178 - add validity checks when grabbing inodes from l_ast_data (3599)
180 - drop scimac NAL (unmaintained)
182 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
185 - don't copy lvb into (possibly NULL) reply on error (2983)
186 - don't deref dentry after dput, don't free lvb on error (2922)
187 - use the kms to determine writeback rpc length (2947)
188 - increment oti_logcookies when osc is inactive (2948)
189 - update client's i_blocks count via lvb messages (2543)
190 - handle intent open/close of special files properly (1557)
191 - mount MDS with errors=remount-ro, like obdfilter (2009)
192 - initialize lock handle to avoid ASSERT on error cleanup (3057)
193 - don't use cancelling-locks' kms values (2947)
194 - use highest lock extent for kms, not last one (2925)
195 - don't dereference ERR_PTR() dentry in error handling path (3107)
196 - fix thread race in portals_debug_dumplog() (3122)
197 - create lprocfs device entries at setup instead of at attach (1519)
198 - common AST error handler, don't evict client on completion race (3145)
199 - zero nameidata in detach_mnt in 2.6 (3118)
200 - verify d_inode after revalidate_special is valid in 2.6 (3116)
201 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
202 - initialize RPC timeout timer earlier for 2.6 (3219)
203 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
204 - print nal/nid for unknown nid (3258)
205 - additional checks for oscc recovery before doing precreate (3284)
206 - fix ll_extent_lock() error return code for 64-bit systems (3043)
207 - don't crash in mdc_close for bad permissions on open (3285)
208 - zero i_rdev for non-device files (3147)
209 - clear page->private before handing to FS, better assertion (3119)
210 - fix incorrect decref of invalidated dentry (2350)
211 - don't hold journal transaction open across create RPC (3313)
212 - update atime on MDS at close time (3265)
213 - close LDAP connection when recovering to avoid server load (3315)
214 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
215 - don't leak open file on MDS after open resend (3325)
216 - serialize filter_precreate and filter_destroy_precreated (3329)
217 - loop device shouldn't call sync_dev() for nul device (3092)
218 - clear page cache after eviction (2766)
219 - resynchronize MDS->OST in background (2824)
220 - refuse to mount the same filesystem twice on same mountpoint (3394)
221 - allow llmount to create routes for mounting behind routers (3320)
222 - push lock cancellation to blocking thread for glimpse ASTs (3409)
223 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
224 - fix rare problem with rename on htree directories (3417)
226 - allow default OST striping configuration per directory (1414)
227 - fix compilation for qswnal for 2.6 kernels (3125)
228 - increase maximum number of MDS request buffers for large systems
229 - change liblustreapi to be useful for external progs like lfsck (3098)
230 - increase local configuration timeout for slow disks (3353)
232 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
235 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
236 - don't overwrite extent policy data in reply if lock was blocked (2901)
237 - drop filter export grants atomically with removal from device (2663)
238 - del obd_self_export from work_list in class_disconnect_exports (2908)
239 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
240 - swab reply message in mdc_close, other PPC fixes (2464)
241 - fix destroying of named logs (2325)
242 - overwrite old logs when running lconf --write_conf (2264)
243 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
244 - fix race in target_handle_connect (2898)
245 - mds_reint_create() should take same inode create lock (2926)
246 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
247 - don't close files for self_export to avoid uninitialized obd (2936)
248 - allow MDS with the same name as client node (2939)
249 - hold dentry reference for closed log files for unlink (2325)
250 - reserve space for all logs during transactions (2059)
251 - don't evict page beyond end of stripe extent (2925)
252 - don't oops on a deleted current working directory (2399)
253 - handle hard links to targets without a parent properly (2517)
254 - don't dereference NULL lock when racing during eviction (2867)
255 - don't grow lock extents when lots of conflicting locks (2919)
257 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
260 - account for cache space usage on clients to avoid data loss (974)
261 - lfsck support in lustre kernel code (2349)
262 - reduce journal credits needed for BRW writes (2370)
263 - orphan handling to avoid losing space on client/server crashes
264 - ptlrpcd can be blocked, stopping ALL progress (2477)
265 - use lock value blocks to assist in proper KMS, faster stat (1021)
266 - takes i_sem instead of DLM locks internally on obdfilter (2720)
267 - recovery for initial connections (2355)
268 - fixes for mds_cleanup_orphans (1934)
269 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
270 - block all file creations until orphan recovery completes (1901)
271 - client remove rq_connection from request struct (2423)
272 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
273 - recovery timer race (2670)
274 - mdc_close recovey bug (2532)
275 - ptlrpc cleanup bug (2710)
276 - mds timeout on local locks (2588)
277 - namespace lock held during RPCs (2431)
278 - handle interrupted sync write properly (2503)
279 - don't try to handle a message that hasn't been replied to (2699)
280 - client assert failure during cleanup after abort recovery (2701)
281 - leak mdc device after failed mount (2712)
282 - ptlrpc_check_set allows timedout requests to complete (2714)
283 - wait for inflight reqs when ptlrpcd finishes (2710)
284 - make sure unregistered services are removed from the srv_list
285 - reset bulk XID's when resending them (caught by 1138 test)
286 - unregister_bulk after timeout
287 - fix lconf error (2694)
288 - handle write after unfinished setstripe, stripe-only getstripe (2388)
289 - readahead locks pages, leaves pending causing memory pressure (2673)
290 - increase OST request buffers to 4096 on large machines (2729)
291 - fix up permission of existing directories in simple_mkdir (2661)
292 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
293 - don't assign transno to errored transactions (2742)
294 - don't delete objects on OST if given a bogus objid from MDS (2751)
295 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
296 - if rq_no_resend, then timeout request after recovery (2432)
297 - fix MDS llog_logid record size, 64-bit array alignment (2733)
298 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
299 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
300 - MDS recovery shouldn't delete objects that it has given out (2730)
301 - if enqueue arrives after completion, don't clobber LVB (2819)
302 - don't unlock pages twice when trigger_group_io returns error (2814)
303 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
304 - don't write pages to disk if there was an error (1450)
305 - don't ping imports that have recovery disabled (2676)
306 - take buffered bytes into account when balancing socknal conn (2817)
307 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
308 - reconnect unlink llog connection after MDS reconnects to OST (2816)
309 - remove little-endian swabbing of llog records (1987)
310 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
311 - retry reposting request buffers if they fail (1191)
312 - grow extent at grant time to avoid granting a revoked lock (2809)
313 - lock revoke doesn't evict page if covered by a second lock (2765)
314 - disable VM readahead to avoid reading outside lock extents (2805)
316 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
317 - updated kernel patches for hp-2.4.20 kernel (2681)
319 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
322 - fix truncated write corruption (2366)
323 - fix for failed assertion in iopen_connect_dentry (1792,2517)
325 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
326 - ensure len(uuid) < 37 in lmc (1171)
327 - fix ia64 OOPS in llog_test (2255)
328 - zero end of page at obdfilter for partial page writes (2648)
329 - don't leave stale dentries around after renames (bug 2428)
330 - fix timeouts when evicting a client with a single lock held (2642)
331 - set deadline for the initial HELLO message to drain (2634)
332 - print out dotted-quad IP addresses in the socknal (2302)
334 - additional debugging for MDS client eviction problem (2443)
335 - fix mkfsoptions support for osts (2603, 2604)
337 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
340 - add series for the vanilla 2.6.0 kernel
341 - add series for the vanilla 2.4.24 kernel
342 - add series for a cray x86/64 UL kernel drop
343 - fix xattr patches for the vanilla 2.4.19 series
345 - generate true UUIDs in lmc (1171)
346 - have portals stack dumping break in UML (2466)
347 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
348 - allocate with _NOFS in ldlm to avoid deadlock (1933)
349 - wake callback waiting threads on client eviction (2460)
350 - Add --ptldebug and --subsystem to lmc (1719)
351 - update assertion to allow safe interrupt allocation
352 - set rq_no_resend for cancel requests (2432)
353 - recalculate ptlrpcd timeout after resend (2494)
354 - call vfs_rmdir when removing pending directories (2368)
355 - fix renaming a file to itself (2429)
356 - lmc creates a default one-stripe lov (2454)
357 - expand procfs space to handle large clusters (2326)
358 - increase UML stack to avoid overflow
359 - update lconf's list of debug and subsystem masks
360 - fix lfs find --obd (2510)
361 - /proc tunable for disabling filter read caching (2591)
362 - stop rpm packages from altering slapd.conf (2301)
363 - disable nagle in the socknal under 0conf (2578)
364 - choose mds inode size based on stripe count (2572)
365 - fix kernel-source rpm problems (2516)
367 - add --disable-doc to avoid pdf generation (2421)
368 - update documentation, tests, type-os, comments
369 - avoid format warnings on ia64
371 - tiny code cleanups by removing unused fields
373 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
376 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
377 - remove the most blatant lies from BUILDING file (2371)
378 - change default debug level to reasonable production setting
379 - reduce client side cache size to reduce cache flush time
380 - reduce max RPCs in flight to avoid unnecessary file fragmentation
381 - make TCP zerocopy and pinger support enabled by default (2476)
382 - sync writes completed after process exits caused crashes (2319)
383 - maintain correct mount count on the MDS (2356)
384 - backout 1557, because 2316 wasn't really fixed
385 - better file I/O statistics gathering in /proc
386 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
387 - another kernel patch to fix zero-copy TCP function export
388 - don't take duplicate lock when processing re-sent getattr (2420)
389 - lctl uses obd_self_export instead of creating new conn (2353)
390 - MDS/OST recovery case which requires object creation asserted (2425)
391 - move lfs from /usr/sbin to /usr/bin in packages
392 - fix race between mds_client_add and mds_client_free (2417)
393 - use kmalloc instead of slabs in portals (2430)
394 - don't create duplicate records when a failover MDS is present (2442)
395 - remove unnecessary mount age check (2332)
396 - don't remove directory inodes from locks prematurely (2451)
397 - don't break if MDS service name is the same as hostname (2103)
398 - fix races in client write RPC generation when cache full (2482)
400 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
403 - remove now-unused request->rq_obd (278)
404 - if an allocation fails, print out how much memory we've used (1933)
405 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
406 - add the "configurable stack size" patch to most series files (1256)
407 - ability to write large log records, for 100+ OST configs (2306)
408 - fix NULL deref when filter_prep fails (2314)
409 - fix operator precedence error in filter_sync
410 - dynamic allocation of socknal TX descriptors (2315)
411 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
412 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
413 - ensure that CWARN messages in Portals always get to the syslog
414 - __init/__exit are not for prototype decls (ldlm_init/exit)
415 - x86-64 compile warning fixes
416 - fix gateway LMC keyword conflict (2318)
417 - fix MDS lock inversions in getattr/reint paths (1844)
418 - fix a rare lock re-ordering bug, which caused deadlock (2322)
419 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
420 - DLM race condition prevented some lock evictions (2328)
421 - ENOMEM detection and retry on socknal sends (2230)
422 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
423 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
424 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
425 - better error messages when a client is rejected during recovery (1505)
426 - avoid cancelling locks which were never granted, after failure (2330)
427 - fix i_sem/journal inversion in mds_client_add (2333)
428 - fix truncate/getattr lock cycle deadlock (2334)
429 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
430 - fix two rare exit paths which could leak an l_lock() ref (2321)
431 - fencepost error in MDS/OST orphan recovery (2226)
432 - make log record alignment 8 bytes (1988)
433 - lstripe now fails when requested offset > ost_count (2237)
434 - ensure that all kernel series have a complete list.h (1607)
435 - fix crashes in special-file operations (2316)
436 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
438 - allow configurable automake binary, for testing new versions
439 - small update to the lfs documentation
441 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
443 * fix negative export reference count in fsfilt_sync (2312)
445 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
446 * release candidate 0.9.1
448 - orphans are moved into the PENDING directory for possible recovery
449 - replayed opens now open by fid for orphan/rename safety (1042)
450 - last close of an orphan inode generates a transno (683)
451 - chdir() and mount() now pin the directory entry (1020)
452 - avoid CERROR in normal ll_setattr_raw() error case (1500)
453 - discard very old requests without processing them (1502)
454 - remove some common, well-understood CERRORs (1505)
455 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
456 - clear "grant" flags in OST replies until OST grant code lands (1644)
457 - fix read performance by not clobbering i_blksize on client (1598)
458 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
459 - make LRU size a /proc tunable, clears locks when reduced (707)
460 - fix some lprocfs rot that prevented ptlbd from loading (1732)
461 - server locks take references on exports now (1558)
462 - build fixes for 2.4.20-rh trees (1663)
463 - return an error from lov_create if all OSCs are inactive (1751)
464 - fix import levels when a reconnect happens without a timeout (1597)
465 - exit early from mds_open if we get a lookup error (1749)
466 - partial page read at EOF wouldn't wait for disk before sending (1642)
467 - avoid NULL deref in obdfilter when reading page past EOF (1592)
468 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
469 - fix LBUG in ll_it_open_error with rc = -2 (1861)
470 - write/truncate lock inversion (1639)
471 - Don't auto-load obdclass, portals modules during cleanup (1495)
472 - fix timestamps from jumping to "now" (1763)
473 - extra journal assertions (1648)
474 - add an extra multiunlink test (1771)
475 - fix read_record/write_record API (1776)
476 - fix leak of offset_extent, possible incorrect i_size later (1772)
477 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
478 - Debugging for the kqswnal_get_idle_tx problems (1820)
479 - Allow recovery to be attempted multiple times (1536)
480 - Write out MDS last_rcvd file after it is first created (1600)
481 - Fix tx_descriptor leak in failed transmit situations (1827)
482 - ext3 journaling fixes for assertion failure after IO error (1871)
483 - class_export_put() on freed export after completion AST error (1896)
484 - Fix revalidate looping in VFS (1322)
485 - Don't access a freed export during MDS_REINT timeout (1521)
486 - Add open-unlink recovery support on the MDS (1673,1764)
487 - Return an error if no MDS data was read from last_rcvd (1946)
488 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
489 - Don't setup a disk device that doesn't match exported UUID (317)
490 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
491 - avoid committing NULL handle in force close
492 - local.sh is now a one-stripe LOV configuration
493 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
494 - fix ext3 htree duplicate directory entry corruption (1516)
495 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
496 - update to new LOV EA format (2097)
497 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
499 2003-06-15 Phil Schwan <phil@clusterfs.com>
502 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
503 - per-import/export recovery handling (958, 931, 959)
504 - multiple last-rcvd slots, for serving multiple FSes (949)
505 - connections are again shared between multiple imp/exports (963, 964)
506 - "umount -f" would hang if any requests needed to be sent (393, 978)
507 - avoid pinning large req buffer by copying for queued messages (989)
508 - add "uuid" to "lctl device" command to help upcalls (991)
509 - "open" RPCs with transnos would confuse recovery counters (1037)
510 - do proper endian conversion of all wire messages (288, 340, 891)
511 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
512 - call waiting locks callback from LDLM recovery thread (1127, 1151)
513 - fix ptlrpc_connection leak in target_handle_connect (1174)
514 - fix import refcounting bug in OST and MDS cleanup (1134)
515 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
516 - fix very unlikely obd_types race condition (501)
517 - remove osc_open hack for echo_client (1187)
518 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
519 - a failure in read_inode2 leads to deadlock (1139)
520 - cancel ack-locks as soon as transaction is committed (1072)
521 - fix major leaks and crashes in the bulk I/O path (937, 1057)
522 - make sure to commitrw after any preprw to avoid deadlock (1162)
523 - failing to execute a file in a lustre FS would lock inode (1203)
524 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
525 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
526 - obd_uuid2tgt would walk off the end of the list (1255)
527 - on IA64 the osc would give portals incorrect bulk size (1258)
528 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
529 - fix lock inversion caused by new llite matching code (1282)
530 - limit the number of dirty pages on a client to 10MB (1286)
531 - timed out locks were not being corrected cancelled (1289)
532 - fix O_DIRECT above 4GB on IA-32 (1292)
533 * major user-visible changes
534 - fail out/fail over policy now controlled by the upcall (993)
536 - add OBD_PING to check server availability and failure (954)
537 - lustre messages are now sent in sending host order (288, 340, 891)
538 - add eadatalen to MDS getattr reply (340)
539 - OST read replies may contain second buffer, with per-page status (593)
541 2003-03-11 Phil Schwan <phil@clusterfs.com>
544 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
545 - failed open()s could cause deadlock; fixed (867, 869)
546 - stop cancelling OST locks when files are closed (481)
547 - overlapping XID spaces caused network corruption (851, 853)
548 - fix unsafe fsfilt counter arithmetic; change to atomic_t
549 - setattr_raw added, to do single-RPC, server-side setattrs
550 - lmc/lconf syntax change for OST UUIDs
551 - fix crashy race condition between ptlrpc_free_req and osc_close
552 - don't use request in mdc_enqueue if we hit a timeout (889)
553 - don't set the inode i_size for regular files from the MDS (896)
554 - handle out of order completion AST (842)
555 - don't LBUG if a lock request times out after receiving AST (913)
556 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
557 - if a bad lock AST arrives, send an error instead of dropping entirely
558 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
559 - fix leak in bulk IO when only partially completed (899, 900, 926)
560 - fix O_DIRECT for ia64 (55)
561 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
562 - C-z after timeout could hang a process forever; fixed (977)
564 - client-side I/O cache (678, 924, 929, 941, 970)
566 - READPAGE and SETATTRs which don't take server-side locks get
569 2003-02-11 Phil Schwan <phil@clusterfs.com>
572 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
573 - Fix fsfilt_extN_readpage() to read a full page of directory
574 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
575 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
576 - Fix imbalanced LOV object allocation and out-of-bound access (469)
577 - Most intent operations were removed, in favour of a new RPC mode
578 that does a single RPC to the server and bypasses most of the VFS
579 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
580 - Aggressively cancel local locks on DLM servers
581 - mds_reint_unlink sends EA to the client if it's the last nlink.
582 client uses that EA to unlink OST objects.
583 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
584 - recursive symlinks were fixed (439)
585 - fixed NULL deref in DEBUG_REQ
586 - filter_update_lastobjid no longer calls sync, which annoyed extN
587 - fixed multi-client small-writes to a single file problem (445)
588 - fixed mtime updates during file writes (607)
589 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
590 - fixed bug in obd_brw_read/write() (under guise of testing 367)
591 - fixed Linux OST size reporting problem (444, 656)
592 - OST now updates object mtime with writes or setattr (607, 619)
593 - client verifies file size before zeroing page past EOF (445)
594 - OST now writes last allocated objid to disk with allocation (108)
595 - LOV on echo now works (409)
597 - mds_reint_unlink sends a new buffer, with the EA included. this
598 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
599 set if a regular file was being unlinked, and it was the last link
600 - use PtlGet from the target for bulk writes (315)
601 - OST now updates object mtime with writes or setattr (607, 619)
602 - LDLM now has a grant-time callback to revalidate locked items, if
604 - Many MDS operations were reorganized to combat race conditions
606 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
607 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
609 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
612 - Fully reactivate OST imports after reconnection (512, others)
613 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
614 - More graceful error handling for truncating on dead OST (515)
615 - Don't error out unless we're actually accessing dead stripes (474)
616 - Fix garbage sizes when stripes are missing (410)
617 - LRU counters were broken, causing constant lock purge (433, 432)
618 - garbage on read from stripes with failed OSTs (441)
619 - mark OSCs as active before reconnecting during recovery (438)
620 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
621 - lfind did not preserve OST order in output (443)
622 - symlinks cause hung clients, incorrect data (439)
623 - stop dereferencing request after dropping refcount (457)
624 - don't LASSERT(spin_is_locked) on non-SMP (455)
625 - fixes for many rename() bugs
626 - fstat didn't correctly synchronize attributes (399)
627 - server must handle lock cancellation during blocking AST prep (487)
628 - bulk descriptors were free()d too soon (511)
629 - fix paths in lconf, which would load incorrect modules (451, 507)
630 - fix confusing lconf 'host not found' error message (386)
631 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
632 - fix race condition in mdc_blocking_ast() for inode access (526)
633 - fix lov_unpackmd() unpacking wrong number of stripes (537)
634 - fix lov_set_osc_active() marking wrong OSC inactive (440)
635 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
636 - fix multiple writes of stripe MD to MDS (358, maybe 519)
637 - fix lstripe in several ways (kernel side) (527)
638 - fix request leak in ldlm_cli_enqueue (262)
639 - incorrect OSC was marked inactive after OST failure
640 - call mds_fs_cleanup before unmounting filesystem (524)
641 - fix races between taking ns_lock and ldlm_lock_change_resource
642 - fix races updating LOV export open file list
643 - fix lov_enqueue error path, avoid decref-ing bad lock handle
644 - fix recovery NULL deref in ldlm_cli_cancel_unused
645 - fix some DLM races by using new hash table for lock handles (419)
646 - permit the client to specify desired inodes, at replay
647 - duplicate requests when we queue them for replay reintegration
648 - fix last_rcvd offset calculation
649 - sync after each recovered transaction, so we always make progress
650 - never, not always, ERESTART requests without transnos
651 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
653 - skip replay if the MDS says that the client is already connected
654 - don't check for a recovery-enabled export to match lctl's UUID
655 - don't INC_USE_COUNT for phantom exports
656 - don't crash when cleaning up phantom exports (567)
657 - don't double-finish or set replay data for errored mdc_open requests
658 - abort requests when they time out, so we don't get old replies
659 - send/receive replies for AST messages again
660 - if the client says that it doesn't have the lock, cancel it on the
662 - if we timeout during I/O, don't try to cancel an in-use lock; instead
663 mark it as destroyed, it will all work out when decref is called
664 - fix module use counts (22, 581)
666 - ASTs now expect a reply (server cancels lock on error reply)
668 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
671 - fix many simultaneous client startup (392)
672 - fix dentry->d_it clobbering
673 - credentials weren't being shipped for readdir/getattr operations
674 - remove invalid assertions triggered during some concurrent MD
676 - proper Lustre versions added (336, 389)
677 - fix memory leak for create error case (398)
678 - fix LOV locking bug that would get cli/srv out of sync
679 - fix echo client over LOV (409)
680 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
681 - fix double-O_EXCL intent crash (424)
682 - avoid sending multiple lock CANCELs (352)
684 - MDS can do multi-client recovery (modulo bugs in new code)
686 - many updates, edits, cleanups
688 2002-11-18 Phil Schwan <phil@clusterfs.com>
691 - fix null d_it dereference (346)
692 - fix full OST/dbench hang (333)
693 - fix permission problem with file removal (286)
694 - fix removal of OSCs from LOV when they fail
695 - fix NULL deref during bulk timeout (214)
696 - fix problems related to multiple filesystems on one MDS (241)
697 - fixed serious subtle metadata locking bugs
698 - free locks on clients when inodes are removed due to memory
700 - fix inode pointer in lock data (285)
701 - partial support for multiple MDS on a single host (241)
702 - data locks weren't cancelled at clear_inode time (290, 311)
703 - intent locks could lead to unbounded lock growth (205)
704 - added a maximum lock count, an LRU list, and a flusher
705 - fix multiple rename (365)
706 - properly abstracted the echo client
707 - OSC locked 1 byte too many; fixed
708 - rewrote brw callback code:
709 - fixed recovery bugs related to LOVs (306)
710 - fixed too-many-pages-in-one-write crash (191)
711 - fixed (again) crash in sync_io_timeout (214)
712 - probably fixed callback-related race (385)
714 - Add capability to MDS protocol
715 - LDLM cancellations and callbacks on different portals
717 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
720 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
721 - defer open object destruction to close time (601981, 138)
722 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
723 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
724 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
725 - don't call obd_setattr in ll_file_release for destroyed objects
726 * protocol change to lustre_msg: move |version| and add |flags|
727 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
728 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
729 * added replay of create, unlink, link and rename operations during
730 MDS failover; recovery should be much more robust now
731 * remove failed OSCs from LOVs (only lov_create uses this so far)
732 * the lustre-HOWTO was brought (more) up to date (582544)
734 2002-10-23 Phil Schwan <phil@clusterfs.com>
737 - in-use dentries weren't being reused properly (617851)
738 - prevent multiple LDLM setup (599178)
739 - fix LOV size calculations for truncate (617853)
740 - fix client handling of MDS intent errors (POSIX)
741 - fix permission bug in lovstripe.c test (624321)
742 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
743 - truncate past end of file could corrupt data
744 - proper cleanup after timeouts, crashes, etc (592524, 550815)
745 - a race in recovery could return ETIMEDOUT to apps (623947)
746 - building outside the source directory was fixed
747 * the lustre-HOWTO was brought (more) up to date (582544)
748 * major progress was made on recovery functionality
750 2002-10-10 Phil Schwan <phil@clusterfs.com>
753 - recovery deadlock fix
754 - rm -rf causes LBUG fix (617817)
755 - file open by multiple tasks fix (618962)
756 - directory permissions bugs (602707 and 620007)
757 - journal_stop fixed with locking (611313)
758 - O_APPEND failures resolved (618273, perhaps 614459)
759 - lconf PATH fix (619770)
760 - IA64 build fix (621450)
761 - RPC buffer sizes scale with amount of memory
763 2002-10-01 Phil Schwan <phil@clusterfs.com>
766 - locks would be cancelled without throwing away data pages,
767 resulting in inconsistent data (605627)
768 - inode attributes were not always being refreshed (605627, 612449)
769 - lconf now continues to cleanup after lctl reports an error
770 - MDS now enforces user permissions (602707)
771 - lprocfs cleanup fixed, but not yet enabled (614157)
772 - fixed infinite server hang, should a client not respond to an AST
773 - avoid going into recovery if user calls readlink() with a buffer
774 that's too small (613941)
775 - AST RPCs no longer require replies (614867) -- this may be changed
776 - don't crash server if client sends an IOV that's too big (611336)
777 - fixed lock conversion deadlock (611892)
778 - fixed the following of symlinks (614622)
779 * recovery: the server can remove locks from a client that dies, other
780 clients can make progress
781 * more extN patch fixes
782 * compile-time configurable ptlrpc buffer allocations
784 - collaborative read cache document
785 - Lustre Lite Performance CDR document-in-progress
787 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
790 - fix typo in patch-2.4.18
792 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
795 - clear ptlrpc request each time in handle_incoming_request()
796 - unlink of files now destroys the object on the OST
798 2002-09-19 Peter Braam <braam@clusterfs.com>
800 * add hard link support
801 * change obdfile creation method
802 * kernel patch changed
804 2002-09-19 Peter Braam <braam@clusterfs.com>
807 - stack overflow bug in extN fixed
809 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
811 * documentation updates
812 - add man pages for config tools
813 - update tests/README to describe testing with new config tools
814 - finish metadata API descriptions
815 * bug fixes and cleanups
816 - statfs workaround for 16TB limit
817 - LOV stripe allocation improved, can stripe on subset of OSTs
818 - LOV file size/IO offset was wrong for files > 4GB in size
819 - object EA data was being dropped, caused files to be unreadable
820 - memory overflow with non-LOV OST caused memory corruption
821 - fixed regression tests to work with new config tools, obdfilter
822 - fixed bug when directory size became larger than 1 block
823 - fixed bug (for single client case) when PWD was deleted
824 - invalidate local directory pages when doing intent-based ops
825 - avoid LDLM oops when lock callback contained bad data
827 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
829 * documentation updates
830 * bug fixes and cleanups
831 - configuration tools
834 - 64-bit compile warnings
835 - 64-bit internal statfs data
837 * test_brw on persistent OST devices
841 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
843 * documentation updates
844 * bug fixes and cleanups
845 * configuration tools
847 2002-08-30 Peter J. Braam <braam@clusterfs.com>
850 * many small fixes to 0_5_4
851 * io/network handling
852 * thinkos in MDS operations
854 2002-08-24 Peter J. Braam <braam@clusterfs.com>
857 * crucial basic fixes to 0.5.3
858 * IOR, Iozone work over Elan
861 2002-08-07 Phil Schwan <phil@clusterfs.com>
862 * version 0_5_3, our first alpha
863 * we use the new Portals iovs
864 * documentation updates
865 * bug fixes and cleanups
866 * small changes in the DLM wire protocol
868 2002-07-25 Peter J. Braam <braam@clusterfs.com>
869 * version 0_5_1 with some initial stability,
870 * locking on MD and file I/O.
871 * documentation updates
872 * several bug fixes since 0.5.0
873 * small changes in wire protocol
875 2002-07-18 Phil Schwan <phil@clusterfs.com>
877 * delivered as Lustre Light Alpha
878 * fixed a crash after handling invalid MDS requests
879 * fixed directory pages for architectures with non-4k pages sizes
881 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
882 * release version v0_4_4
883 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
885 * Fixes a number of interruption problems with OST operations.
886 * Update documentation for portals header changes
887 * Move all wire protocol structs/defines to lustre_idl.h
888 * Fixes symlink length bug.
889 * Add tcpdump to repository.
891 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
892 * release version v0_4_3
893 * Fixes statfs for inodes on extN.
894 * Fixes bug in runtests which would delete /etc/hosts.
895 * Use 64-bit object IDs wherever possible (not into VFS though)
896 Remove ost_get_info, which is unused by lustre, and out of date.
898 2002-07-03 Peter Braam <braam@clusterfs.com>
899 * release version v0_4_2 Fixes a lookup error (type not passed)
900 * move forward to head of Portals
901 * move forward to latest Lustre kernel
903 2002-06-25 Peter Braam <braam@clusterfs.com>
904 * release version v0_4_1. Hopefully stable on single node use.