1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5 - remove the most blatant lies from BUILDING file (2371)
6 - change default debug level to reasonable production setting
7 - reduce client side cache size to reduce cache flush time
8 - reduce max RPCs in flight to avoid unnecessary file fragmentation
9 - sync writes completed after process exits caused crashes (2319)
10 - maintain correct mount count on the MDS (2356)
11 - backout 1557, because 2316 wasn't really fixed
12 - better file I/O statistics gathering in /proc
13 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
14 - another kernel patch to fix zero-copy TCP function export
15 - don't take duplicate lock when processing re-sent getattr (2420)
16 - lctl uses obd_self_export instead of creating new conn (2353)
17 - MDS/OST recovery case which requires object creation asserted (2425)
18 - move lfs from /usr/sbin to /usr/bin in packages
19 - fix race between mds_client_add and mds_client_free (2417)
20 - use kmalloc instead of slabs in portals (2430)
21 - don't create duplicate records when a failover MDS is present (2442)
22 - remove unnecessary mount age check (2332)
23 - don't remove directory inodes from locks prematurely (2451)
25 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
28 - remove now-unused request->rq_obd (278)
29 - if an allocation fails, print out how much memory we've used (1933)
30 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
31 - add the "configurable stack size" patch to most series files (1256)
32 - ability to write large log records, for 100+ OST configs (2306)
33 - fix NULL deref when filter_prep fails (2314)
34 - fix operator precedence error in filter_sync
35 - dynamic allocation of socknal TX descriptors (2315)
36 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
37 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
38 - ensure that CWARN messages in Portals always get to the syslog
39 - __init/__exit are not for prototype decls (ldlm_init/exit)
40 - x86-64 compile warning fixes
41 - fix gateway LMC keyword conflict (2318)
42 - fix MDS lock inversions in getattr/reint paths (1844)
43 - fix a rare lock re-ordering bug, which caused deadlock (2322)
44 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
45 - protect MDS inode fsdata with stronger locking (2313)
46 - better error messages when a client is rejected during recovery (1505)
47 - avoid cancelling locks which were never granted, after failure (2330)
48 - fix i_sem/journal inversion in mds_client_add (2333)
49 - fix truncate/getattr lock cycle deadlock (2334)
51 - allow configurable automake binary, for testing new versions
52 - small update to the lfs documentation
54 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
56 * fix negative export reference count in fsfilt_sync (2312)
58 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
59 * release candidate 0.9.1
61 - orphans are moved into the PENDING directory for possible recovery
62 - replayed opens now open by fid for orphan/rename safety (1042)
63 - last close of an orphan inode generates a transno (683)
64 - chdir() and mount() now pin the directory entry (1020)
65 - avoid CERROR in normal ll_setattr_raw() error case (1500)
66 - discard very old requests without processing them (1502)
67 - remove some common, well-understood CERRORs (1505)
68 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
69 - clear "grant" flags in OST replies until OST grant code lands (1644)
70 - fix read performance by not clobbering i_blksize on client (1598)
71 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
72 - make LRU size a /proc tunable, clears locks when reduced (707)
73 - fix some lprocfs rot that prevented ptlbd from loading (1732)
74 - server locks take references on exports now (1558)
75 - build fixes for 2.4.20-rh trees (1663)
76 - return an error from lov_create if all OSCs are inactive (1751)
77 - fix import levels when a reconnect happens without a timeout (1597)
78 - exit early from mds_open if we get a lookup error (1749)
79 - partial page read at EOF wouldn't wait for disk before sending (1642)
80 - avoid NULL deref in obdfilter when reading page past EOF (1592)
81 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
82 - fix LBUG in ll_it_open_error with rc = -2 (1861)
83 - write/truncate lock inversion (1639)
84 - Don't auto-load obdclass, portals modules during cleanup (1495)
85 - fix timestamps from jumping to "now" (1763)
86 - extra journal assertions (1648)
87 - add an extra multiunlink test (1771)
88 - fix read_record/write_record API (1776)
89 - fix leak of offset_extent, possible incorrect i_size later (1772)
90 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
91 - Debugging for the kqswnal_get_idle_tx problems (1820)
92 - Allow recovery to be attempted multiple times (1536)
93 - Write out MDS last_rcvd file after it is first created (1600)
94 - Fix tx_descriptor leak in failed transmit situations (1827)
95 - ext3 journaling fixes for assertion failure after IO error (1871)
96 - class_export_put() on freed export after completion AST error (1896)
97 - Fix revalidate looping in VFS (1322)
98 - Don't access a freed export during MDS_REINT timeout (1521)
99 - Add open-unlink recovery support on the MDS (1673,1764)
100 - Return an error if no MDS data was read from last_rcvd (1946)
101 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
102 - Don't setup a disk device that doesn't match exported UUID (317)
103 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
104 - avoid committing NULL handle in force close
105 - local.sh is now a one-stripe LOV configuration
106 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
107 - fix ext3 htree duplicate directory entry corruption (1516)
108 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
109 - update to new LOV EA format (2097)
110 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
112 2003-06-15 Phil Schwan <phil@clusterfs.com>
115 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
116 - per-import/export recovery handling (958, 931, 959)
117 - multiple last-rcvd slots, for serving multiple FSes (949)
118 - connections are again shared between multiple imp/exports (963, 964)
119 - "umount -f" would hang if any requests needed to be sent (393, 978)
120 - avoid pinning large req buffer by copying for queued messages (989)
121 - add "uuid" to "lctl device" command to help upcalls (991)
122 - "open" RPCs with transnos would confuse recovery counters (1037)
123 - do proper endian conversion of all wire messages (288, 340, 891)
124 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
125 - call waiting locks callback from LDLM recovery thread (1127, 1151)
126 - fix ptlrpc_connection leak in target_handle_connect (1174)
127 - fix import refcounting bug in OST and MDS cleanup (1134)
128 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
129 - fix very unlikely obd_types race condition (501)
130 - remove osc_open hack for echo_client (1187)
131 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
132 - a failure in read_inode2 leads to deadlock (1139)
133 - cancel ack-locks as soon as transaction is committed (1072)
134 - fix major leaks and crashes in the bulk I/O path (937, 1057)
135 - make sure to commitrw after any preprw to avoid deadlock (1162)
136 - failing to execute a file in a lustre FS would lock inode (1203)
137 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
138 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
139 - obd_uuid2tgt would walk off the end of the list (1255)
140 - on IA64 the osc would give portals incorrect bulk size (1258)
141 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
142 - fix lock inversion caused by new llite matching code (1282)
143 - limit the number of dirty pages on a client to 10MB (1286)
144 - timed out locks were not being corrected cancelled (1289)
145 - fix O_DIRECT above 4GB on IA-32 (1292)
146 * major user-visible changes
147 - fail out/fail over policy now controlled by the upcall (993)
149 - add OBD_PING to check server availability and failure (954)
150 - lustre messages are now sent in sending host order (288, 340, 891)
151 - add eadatalen to MDS getattr reply (340)
152 - OST read replies may contain second buffer, with per-page status (593)
154 2003-03-11 Phil Schwan <phil@clusterfs.com>
157 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
158 - failed open()s could cause deadlock; fixed (867, 869)
159 - stop cancelling OST locks when files are closed (481)
160 - overlapping XID spaces caused network corruption (851, 853)
161 - fix unsafe fsfilt counter arithmetic; change to atomic_t
162 - setattr_raw added, to do single-RPC, server-side setattrs
163 - lmc/lconf syntax change for OST UUIDs
164 - fix crashy race condition between ptlrpc_free_req and osc_close
165 - don't use request in mdc_enqueue if we hit a timeout (889)
166 - don't set the inode i_size for regular files from the MDS (896)
167 - handle out of order completion AST (842)
168 - don't LBUG if a lock request times out after receiving AST (913)
169 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
170 - if a bad lock AST arrives, send an error instead of dropping entirely
171 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
172 - fix leak in bulk IO when only partially completed (899, 900, 926)
173 - fix O_DIRECT for ia64 (55)
174 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
175 - C-z after timeout could hang a process forever; fixed (977)
177 - client-side I/O cache (678, 924, 929, 941, 970)
179 - READPAGE and SETATTRs which don't take server-side locks get
182 2003-02-11 Phil Schwan <phil@clusterfs.com>
185 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
186 - Fix fsfilt_extN_readpage() to read a full page of directory
187 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
188 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
189 - Fix imbalanced LOV object allocation and out-of-bound access (469)
190 - Most intent operations were removed, in favour of a new RPC mode
191 that does a single RPC to the server and bypasses most of the VFS
192 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
193 - Aggressively cancel local locks on DLM servers
194 - mds_reint_unlink sends EA to the client if it's the last nlink.
195 client uses that EA to unlink OST objects.
196 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
197 - recursive symlinks were fixed (439)
198 - fixed NULL deref in DEBUG_REQ
199 - filter_update_lastobjid no longer calls sync, which annoyed extN
200 - fixed multi-client small-writes to a single file problem (445)
201 - fixed mtime updates during file writes (607)
202 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
203 - fixed bug in obd_brw_read/write() (under guise of testing 367)
204 - fixed Linux OST size reporting problem (444, 656)
205 - OST now updates object mtime with writes or setattr (607, 619)
206 - client verifies file size before zeroing page past EOF (445)
207 - OST now writes last allocated objid to disk with allocation (108)
208 - LOV on echo now works (409)
210 - mds_reint_unlink sends a new buffer, with the EA included. this
211 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
212 set if a regular file was being unlinked, and it was the last link
213 - use PtlGet from the target for bulk writes (315)
214 - OST now updates object mtime with writes or setattr (607, 619)
215 - LDLM now has a grant-time callback to revalidate locked items, if
217 - Many MDS operations were reorganized to combat race conditions
219 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
220 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
222 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
225 - Fully reactivate OST imports after reconnection (512, others)
226 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
227 - More graceful error handling for truncating on dead OST (515)
228 - Don't error out unless we're actually accessing dead stripes (474)
229 - Fix garbage sizes when stripes are missing (410)
230 - LRU counters were broken, causing constant lock purge (433, 432)
231 - garbage on read from stripes with failed OSTs (441)
232 - mark OSCs as active before reconnecting during recovery (438)
233 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
234 - lfind did not preserve OST order in output (443)
235 - symlinks cause hung clients, incorrect data (439)
236 - stop dereferencing request after dropping refcount (457)
237 - don't LASSERT(spin_is_locked) on non-SMP (455)
238 - fixes for many rename() bugs
239 - fstat didn't correctly synchronize attributes (399)
240 - server must handle lock cancellation during blocking AST prep (487)
241 - bulk descriptors were free()d too soon (511)
242 - fix paths in lconf, which would load incorrect modules (451, 507)
243 - fix confusing lconf 'host not found' error message (386)
244 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
245 - fix race condition in mdc_blocking_ast() for inode access (526)
246 - fix lov_unpackmd() unpacking wrong number of stripes (537)
247 - fix lov_set_osc_active() marking wrong OSC inactive (440)
248 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
249 - fix multiple writes of stripe MD to MDS (358, maybe 519)
250 - fix lstripe in several ways (kernel side) (527)
251 - fix request leak in ldlm_cli_enqueue (262)
252 - incorrect OSC was marked inactive after OST failure
253 - call mds_fs_cleanup before unmounting filesystem (524)
254 - fix races between taking ns_lock and ldlm_lock_change_resource
255 - fix races updating LOV export open file list
256 - fix lov_enqueue error path, avoid decref-ing bad lock handle
257 - fix recovery NULL deref in ldlm_cli_cancel_unused
258 - fix some DLM races by using new hash table for lock handles (419)
259 - permit the client to specify desired inodes, at replay
260 - duplicate requests when we queue them for replay reintegration
261 - fix last_rcvd offset calculation
262 - sync after each recovered transaction, so we always make progress
263 - never, not always, ERESTART requests without transnos
264 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
266 - skip replay if the MDS says that the client is already connected
267 - don't check for a recovery-enabled export to match lctl's UUID
268 - don't INC_USE_COUNT for phantom exports
269 - don't crash when cleaning up phantom exports (567)
270 - don't double-finish or set replay data for errored mdc_open requests
271 - abort requests when they time out, so we don't get old replies
272 - send/receive replies for AST messages again
273 - if the client says that it doesn't have the lock, cancel it on the
275 - if we timeout during I/O, don't try to cancel an in-use lock; instead
276 mark it as destroyed, it will all work out when decref is called
277 - fix module use counts (22, 581)
279 - ASTs now expect a reply (server cancels lock on error reply)
281 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
284 - fix many simultaneous client startup (392)
285 - fix dentry->d_it clobbering
286 - credentials weren't being shipped for readdir/getattr operations
287 - remove invalid assertions triggered during some concurrent MD
289 - proper Lustre versions added (336, 389)
290 - fix memory leak for create error case (398)
291 - fix LOV locking bug that would get cli/srv out of sync
292 - fix echo client over LOV (409)
293 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
294 - fix double-O_EXCL intent crash (424)
295 - avoid sending multiple lock CANCELs (352)
297 - MDS can do multi-client recovery (modulo bugs in new code)
299 - many updates, edits, cleanups
301 2002-11-18 Phil Schwan <phil@clusterfs.com>
304 - fix null d_it dereference (346)
305 - fix full OST/dbench hang (333)
306 - fix permission problem with file removal (286)
307 - fix removal of OSCs from LOV when they fail
308 - fix NULL deref during bulk timeout (214)
309 - fix problems related to multiple filesystems on one MDS (241)
310 - fixed serious subtle metadata locking bugs
311 - free locks on clients when inodes are removed due to memory
313 - fix inode pointer in lock data (285)
314 - partial support for multiple MDS on a single host (241)
315 - data locks weren't cancelled at clear_inode time (290, 311)
316 - intent locks could lead to unbounded lock growth (205)
317 - added a maximum lock count, an LRU list, and a flusher
318 - fix multiple rename (365)
319 - properly abstracted the echo client
320 - OSC locked 1 byte too many; fixed
321 - rewrote brw callback code:
322 - fixed recovery bugs related to LOVs (306)
323 - fixed too-many-pages-in-one-write crash (191)
324 - fixed (again) crash in sync_io_timeout (214)
325 - probably fixed callback-related race (385)
327 - Add capability to MDS protocol
328 - LDLM cancellations and callbacks on different portals
330 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
333 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
334 - defer open object destruction to close time (601981, 138)
335 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
336 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
337 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
338 - don't call obd_setattr in ll_file_release for destroyed objects
339 * protocol change to lustre_msg: move |version| and add |flags|
340 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
341 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
342 * added replay of create, unlink, link and rename operations during
343 MDS failover; recovery should be much more robust now
344 * remove failed OSCs from LOVs (only lov_create uses this so far)
345 * the lustre-HOWTO was brought (more) up to date (582544)
347 2002-10-23 Phil Schwan <phil@clusterfs.com>
350 - in-use dentries weren't being reused properly (617851)
351 - prevent multiple LDLM setup (599178)
352 - fix LOV size calculations for truncate (617853)
353 - fix client handling of MDS intent errors (POSIX)
354 - fix permission bug in lovstripe.c test (624321)
355 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
356 - truncate past end of file could corrupt data
357 - proper cleanup after timeouts, crashes, etc (592524, 550815)
358 - a race in recovery could return ETIMEDOUT to apps (623947)
359 - building outside the source directory was fixed
360 * the lustre-HOWTO was brought (more) up to date (582544)
361 * major progress was made on recovery functionality
363 2002-10-10 Phil Schwan <phil@clusterfs.com>
366 - recovery deadlock fix
367 - rm -rf causes LBUG fix (617817)
368 - file open by multiple tasks fix (618962)
369 - directory permissions bugs (602707 and 620007)
370 - journal_stop fixed with locking (611313)
371 - O_APPEND failures resolved (618273, perhaps 614459)
372 - lconf PATH fix (619770)
373 - IA64 build fix (621450)
374 - RPC buffer sizes scale with amount of memory
376 2002-10-01 Phil Schwan <phil@clusterfs.com>
379 - locks would be cancelled without throwing away data pages,
380 resulting in inconsistent data (605627)
381 - inode attributes were not always being refreshed (605627, 612449)
382 - lconf now continues to cleanup after lctl reports an error
383 - MDS now enforces user permissions (602707)
384 - lprocfs cleanup fixed, but not yet enabled (614157)
385 - fixed infinite server hang, should a client not respond to an AST
386 - avoid going into recovery if user calls readlink() with a buffer
387 that's too small (613941)
388 - AST RPCs no longer require replies (614867) -- this may be changed
389 - don't crash server if client sends an IOV that's too big (611336)
390 - fixed lock conversion deadlock (611892)
391 - fixed the following of symlinks (614622)
392 * recovery: the server can remove locks from a client that dies, other
393 clients can make progress
394 * more extN patch fixes
395 * compile-time configurable ptlrpc buffer allocations
397 - collaborative read cache document
398 - Lustre Lite Performance CDR document-in-progress
400 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
403 - fix typo in patch-2.4.18
405 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
408 - clear ptlrpc request each time in handle_incoming_request()
409 - unlink of files now destroys the object on the OST
411 2002-09-19 Peter Braam <braam@clusterfs.com>
413 * add hard link support
414 * change obdfile creation method
415 * kernel patch changed
417 2002-09-19 Peter Braam <braam@clusterfs.com>
420 - stack overflow bug in extN fixed
422 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
424 * documentation updates
425 - add man pages for config tools
426 - update tests/README to describe testing with new config tools
427 - finish metadata API descriptions
428 * bug fixes and cleanups
429 - statfs workaround for 16TB limit
430 - LOV stripe allocation improved, can stripe on subset of OSTs
431 - LOV file size/IO offset was wrong for files > 4GB in size
432 - object EA data was being dropped, caused files to be unreadable
433 - memory overflow with non-LOV OST caused memory corruption
434 - fixed regression tests to work with new config tools, obdfilter
435 - fixed bug when directory size became larger than 1 block
436 - fixed bug (for single client case) when PWD was deleted
437 - invalidate local directory pages when doing intent-based ops
438 - avoid LDLM oops when lock callback contained bad data
440 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
442 * documentation updates
443 * bug fixes and cleanups
444 - configuration tools
447 - 64-bit compile warnings
448 - 64-bit internal statfs data
450 * test_brw on persistent OST devices
454 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
456 * documentation updates
457 * bug fixes and cleanups
458 * configuration tools
460 2002-08-30 Peter J. Braam <braam@clusterfs.com>
463 * many small fixes to 0_5_4
464 * io/network handling
465 * thinkos in MDS operations
467 2002-08-24 Peter J. Braam <braam@clusterfs.com>
470 * crucial basic fixes to 0.5.3
471 * IOR, Iozone work over Elan
474 2002-08-07 Phil Schwan <phil@clusterfs.com>
475 * version 0_5_3, our first alpha
476 * we use the new Portals iovs
477 * documentation updates
478 * bug fixes and cleanups
479 * small changes in the DLM wire protocol
481 2002-07-25 Peter J. Braam <braam@clusterfs.com>
482 * version 0_5_1 with some initial stability,
483 * locking on MD and file I/O.
484 * documentation updates
485 * several bug fixes since 0.5.0
486 * small changes in wire protocol
488 2002-07-18 Phil Schwan <phil@clusterfs.com>
490 * delivered as Lustre Light Alpha
491 * fixed a crash after handling invalid MDS requests
492 * fixed directory pages for architectures with non-4k pages sizes
494 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
495 * release version v0_4_4
496 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
498 * Fixes a number of interruption problems with OST operations.
499 * Update documentation for portals header changes
500 * Move all wire protocol structs/defines to lustre_idl.h
501 * Fixes symlink length bug.
502 * Add tcpdump to repository.
504 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
505 * release version v0_4_3
506 * Fixes statfs for inodes on extN.
507 * Fixes bug in runtests which would delete /etc/hosts.
508 * Use 64-bit object IDs wherever possible (not into VFS though)
509 Remove ost_get_info, which is unused by lustre, and out of date.
511 2002-07-03 Peter Braam <braam@clusterfs.com>
512 * release version v0_4_2 Fixes a lookup error (type not passed)
513 * move forward to head of Portals
514 * move forward to latest Lustre kernel
516 2002-06-25 Peter Braam <braam@clusterfs.com>
517 * release version v0_4_1. Hopefully stable on single node use.