1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4 - remove now-unused request->rq_obd (278)
5 - if an allocation fails, print out how much memory we've used (1933)
6 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
7 - add the "configurable stack size" patch to most series files (1256)
8 - ability to write large log records, for 100+ OST configs (2306)
9 - fix NULL deref when filter_prep fails (2314)
10 - fix operator precedence error in filter_sync
11 - dynamic allocation of socknal TX descriptors (2315)
12 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
13 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
14 - ensure that CWARN messages in Portals always get to the syslog
15 - __init/__exit are not for prototype decls (ldlm_init/exit)
16 - x86-64 compile warning fixes
17 - fix gateway LMC keyword conflict (2318)
18 - fix MDS lock inversions in getattr/reint paths (1844)
19 - fix a rare lock re-ordering bug, which caused deadlock (2322)
20 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
21 - protect MDS inode fsdata with stronger locking (2313)
22 - better error messages when a client is rejected during recovery (1505)
24 - allow configurable automake binary, for testing new versions
25 - small update to the lfs documentation
27 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
29 * fix negative export reference count in fsfilt_sync (2312)
31 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
32 * release candidate 0.9.1
34 - orphans are moved into the PENDING directory for possible recovery
35 - replayed opens now open by fid for orphan/rename safety (1042)
36 - last close of an orphan inode generates a transno (683)
37 - chdir() and mount() now pin the directory entry (1020)
38 - avoid CERROR in normal ll_setattr_raw() error case (1500)
39 - discard very old requests without processing them (1502)
40 - remove some common, well-understood CERRORs (1505)
41 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
42 - clear "grant" flags in OST replies until OST grant code lands (1644)
43 - fix read performance by not clobbering i_blksize on client (1598)
44 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
45 - make LRU size a /proc tunable, clears locks when reduced (707)
46 - fix some lprocfs rot that prevented ptlbd from loading (1732)
47 - server locks take references on exports now (1558)
48 - build fixes for 2.4.20-rh trees (1663)
49 - return an error from lov_create if all OSCs are inactive (1751)
50 - fix import levels when a reconnect happens without a timeout (1597)
51 - exit early from mds_open if we get a lookup error (1749)
52 - partial page read at EOF wouldn't wait for disk before sending (1642)
53 - avoid NULL deref in obdfilter when reading page past EOF (1592)
54 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
55 - fix LBUG in ll_it_open_error with rc = -2 (1861)
56 - write/truncate lock inversion (1639)
57 - Don't auto-load obdclass, portals modules during cleanup (1495)
58 - fix timestamps from jumping to "now" (1763)
59 - extra journal assertions (1648)
60 - add an extra multiunlink test (1771)
61 - fix read_record/write_record API (1776)
62 - fix leak of offset_extent, possible incorrect i_size later (1772)
63 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
64 - Debugging for the kqswnal_get_idle_tx problems (1820)
65 - Allow recovery to be attempted multiple times (1536)
66 - Write out MDS last_rcvd file after it is first created (1600)
67 - Fix tx_descriptor leak in failed transmit situations (1827)
68 - ext3 journaling fixes for assertion failure after IO error (1871)
69 - class_export_put() on freed export after completion AST error (1896)
70 - Fix revalidate looping in VFS (1322)
71 - Don't access a freed export during MDS_REINT timeout (1521)
72 - Add open-unlink recovery support on the MDS (1673,1764)
73 - Return an error if no MDS data was read from last_rcvd (1946)
74 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
75 - Don't setup a disk device that doesn't match exported UUID (317)
76 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
77 - avoid committing NULL handle in force close
78 - local.sh is now a one-stripe LOV configuration
79 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
80 - fix ext3 htree duplicate directory entry corruption (1516)
81 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
82 - update to new LOV EA format (2097)
83 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
85 2003-06-15 Phil Schwan <phil@clusterfs.com>
88 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
89 - per-import/export recovery handling (958, 931, 959)
90 - multiple last-rcvd slots, for serving multiple FSes (949)
91 - connections are again shared between multiple imp/exports (963, 964)
92 - "umount -f" would hang if any requests needed to be sent (393, 978)
93 - avoid pinning large req buffer by copying for queued messages (989)
94 - add "uuid" to "lctl device" command to help upcalls (991)
95 - "open" RPCs with transnos would confuse recovery counters (1037)
96 - do proper endian conversion of all wire messages (288, 340, 891)
97 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
98 - call waiting locks callback from LDLM recovery thread (1127, 1151)
99 - fix ptlrpc_connection leak in target_handle_connect (1174)
100 - fix import refcounting bug in OST and MDS cleanup (1134)
101 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
102 - fix very unlikely obd_types race condition (501)
103 - remove osc_open hack for echo_client (1187)
104 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
105 - a failure in read_inode2 leads to deadlock (1139)
106 - cancel ack-locks as soon as transaction is committed (1072)
107 - fix major leaks and crashes in the bulk I/O path (937, 1057)
108 - make sure to commitrw after any preprw to avoid deadlock (1162)
109 - failing to execute a file in a lustre FS would lock inode (1203)
110 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
111 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
112 - obd_uuid2tgt would walk off the end of the list (1255)
113 - on IA64 the osc would give portals incorrect bulk size (1258)
114 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
115 - fix lock inversion caused by new llite matching code (1282)
116 - limit the number of dirty pages on a client to 10MB (1286)
117 - timed out locks were not being corrected cancelled (1289)
118 - fix O_DIRECT above 4GB on IA-32 (1292)
119 * major user-visible changes
120 - fail out/fail over policy now controlled by the upcall (993)
122 - add OBD_PING to check server availability and failure (954)
123 - lustre messages are now sent in sending host order (288, 340, 891)
124 - add eadatalen to MDS getattr reply (340)
125 - OST read replies may contain second buffer, with per-page status (593)
127 2003-03-11 Phil Schwan <phil@clusterfs.com>
130 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
131 - failed open()s could cause deadlock; fixed (867, 869)
132 - stop cancelling OST locks when files are closed (481)
133 - overlapping XID spaces caused network corruption (851, 853)
134 - fix unsafe fsfilt counter arithmetic; change to atomic_t
135 - setattr_raw added, to do single-RPC, server-side setattrs
136 - lmc/lconf syntax change for OST UUIDs
137 - fix crashy race condition between ptlrpc_free_req and osc_close
138 - don't use request in mdc_enqueue if we hit a timeout (889)
139 - don't set the inode i_size for regular files from the MDS (896)
140 - handle out of order completion AST (842)
141 - don't LBUG if a lock request times out after receiving AST (913)
142 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
143 - if a bad lock AST arrives, send an error instead of dropping entirely
144 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
145 - fix leak in bulk IO when only partially completed (899, 900, 926)
146 - fix O_DIRECT for ia64 (55)
147 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
148 - C-z after timeout could hang a process forever; fixed (977)
150 - client-side I/O cache (678, 924, 929, 941, 970)
152 - READPAGE and SETATTRs which don't take server-side locks get
155 2003-02-11 Phil Schwan <phil@clusterfs.com>
158 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
159 - Fix fsfilt_extN_readpage() to read a full page of directory
160 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
161 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
162 - Fix imbalanced LOV object allocation and out-of-bound access (469)
163 - Most intent operations were removed, in favour of a new RPC mode
164 that does a single RPC to the server and bypasses most of the VFS
165 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
166 - Aggressively cancel local locks on DLM servers
167 - mds_reint_unlink sends EA to the client if it's the last nlink.
168 client uses that EA to unlink OST objects.
169 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
170 - recursive symlinks were fixed (439)
171 - fixed NULL deref in DEBUG_REQ
172 - filter_update_lastobjid no longer calls sync, which annoyed extN
173 - fixed multi-client small-writes to a single file problem (445)
174 - fixed mtime updates during file writes (607)
175 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
176 - fixed bug in obd_brw_read/write() (under guise of testing 367)
177 - fixed Linux OST size reporting problem (444, 656)
178 - OST now updates object mtime with writes or setattr (607, 619)
179 - client verifies file size before zeroing page past EOF (445)
180 - OST now writes last allocated objid to disk with allocation (108)
181 - LOV on echo now works (409)
183 - mds_reint_unlink sends a new buffer, with the EA included. this
184 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
185 set if a regular file was being unlinked, and it was the last link
186 - use PtlGet from the target for bulk writes (315)
187 - OST now updates object mtime with writes or setattr (607, 619)
188 - LDLM now has a grant-time callback to revalidate locked items, if
190 - Many MDS operations were reorganized to combat race conditions
192 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
193 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
195 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
198 - Fully reactivate OST imports after reconnection (512, others)
199 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
200 - More graceful error handling for truncating on dead OST (515)
201 - Don't error out unless we're actually accessing dead stripes (474)
202 - Fix garbage sizes when stripes are missing (410)
203 - LRU counters were broken, causing constant lock purge (433, 432)
204 - garbage on read from stripes with failed OSTs (441)
205 - mark OSCs as active before reconnecting during recovery (438)
206 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
207 - lfind did not preserve OST order in output (443)
208 - symlinks cause hung clients, incorrect data (439)
209 - stop dereferencing request after dropping refcount (457)
210 - don't LASSERT(spin_is_locked) on non-SMP (455)
211 - fixes for many rename() bugs
212 - fstat didn't correctly synchronize attributes (399)
213 - server must handle lock cancellation during blocking AST prep (487)
214 - bulk descriptors were free()d too soon (511)
215 - fix paths in lconf, which would load incorrect modules (451, 507)
216 - fix confusing lconf 'host not found' error message (386)
217 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
218 - fix race condition in mdc_blocking_ast() for inode access (526)
219 - fix lov_unpackmd() unpacking wrong number of stripes (537)
220 - fix lov_set_osc_active() marking wrong OSC inactive (440)
221 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
222 - fix multiple writes of stripe MD to MDS (358, maybe 519)
223 - fix lstripe in several ways (kernel side) (527)
224 - fix request leak in ldlm_cli_enqueue (262)
225 - incorrect OSC was marked inactive after OST failure
226 - call mds_fs_cleanup before unmounting filesystem (524)
227 - fix races between taking ns_lock and ldlm_lock_change_resource
228 - fix races updating LOV export open file list
229 - fix lov_enqueue error path, avoid decref-ing bad lock handle
230 - fix recovery NULL deref in ldlm_cli_cancel_unused
231 - fix some DLM races by using new hash table for lock handles (419)
232 - permit the client to specify desired inodes, at replay
233 - duplicate requests when we queue them for replay reintegration
234 - fix last_rcvd offset calculation
235 - sync after each recovered transaction, so we always make progress
236 - never, not always, ERESTART requests without transnos
237 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
239 - skip replay if the MDS says that the client is already connected
240 - don't check for a recovery-enabled export to match lctl's UUID
241 - don't INC_USE_COUNT for phantom exports
242 - don't crash when cleaning up phantom exports (567)
243 - don't double-finish or set replay data for errored mdc_open requests
244 - abort requests when they time out, so we don't get old replies
245 - send/receive replies for AST messages again
246 - if the client says that it doesn't have the lock, cancel it on the
248 - if we timeout during I/O, don't try to cancel an in-use lock; instead
249 mark it as destroyed, it will all work out when decref is called
250 - fix module use counts (22, 581)
252 - ASTs now expect a reply (server cancels lock on error reply)
254 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
257 - fix many simultaneous client startup (392)
258 - fix dentry->d_it clobbering
259 - credentials weren't being shipped for readdir/getattr operations
260 - remove invalid assertions triggered during some concurrent MD
262 - proper Lustre versions added (336, 389)
263 - fix memory leak for create error case (398)
264 - fix LOV locking bug that would get cli/srv out of sync
265 - fix echo client over LOV (409)
266 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
267 - fix double-O_EXCL intent crash (424)
268 - avoid sending multiple lock CANCELs (352)
270 - MDS can do multi-client recovery (modulo bugs in new code)
272 - many updates, edits, cleanups
274 2002-11-18 Phil Schwan <phil@clusterfs.com>
277 - fix null d_it dereference (346)
278 - fix full OST/dbench hang (333)
279 - fix permission problem with file removal (286)
280 - fix removal of OSCs from LOV when they fail
281 - fix NULL deref during bulk timeout (214)
282 - fix problems related to multiple filesystems on one MDS (241)
283 - fixed serious subtle metadata locking bugs
284 - free locks on clients when inodes are removed due to memory
286 - fix inode pointer in lock data (285)
287 - partial support for multiple MDS on a single host (241)
288 - data locks weren't cancelled at clear_inode time (290, 311)
289 - intent locks could lead to unbounded lock growth (205)
290 - added a maximum lock count, an LRU list, and a flusher
291 - fix multiple rename (365)
292 - properly abstracted the echo client
293 - OSC locked 1 byte too many; fixed
294 - rewrote brw callback code:
295 - fixed recovery bugs related to LOVs (306)
296 - fixed too-many-pages-in-one-write crash (191)
297 - fixed (again) crash in sync_io_timeout (214)
298 - probably fixed callback-related race (385)
300 - Add capability to MDS protocol
301 - LDLM cancellations and callbacks on different portals
303 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
306 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
307 - defer open object destruction to close time (601981, 138)
308 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
309 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
310 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
311 - don't call obd_setattr in ll_file_release for destroyed objects
312 * protocol change to lustre_msg: move |version| and add |flags|
313 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
314 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
315 * added replay of create, unlink, link and rename operations during
316 MDS failover; recovery should be much more robust now
317 * remove failed OSCs from LOVs (only lov_create uses this so far)
318 * the lustre-HOWTO was brought (more) up to date (582544)
320 2002-10-23 Phil Schwan <phil@clusterfs.com>
323 - in-use dentries weren't being reused properly (617851)
324 - prevent multiple LDLM setup (599178)
325 - fix LOV size calculations for truncate (617853)
326 - fix client handling of MDS intent errors (POSIX)
327 - fix permission bug in lovstripe.c test (624321)
328 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
329 - truncate past end of file could corrupt data
330 - proper cleanup after timeouts, crashes, etc (592524, 550815)
331 - a race in recovery could return ETIMEDOUT to apps (623947)
332 - building outside the source directory was fixed
333 * the lustre-HOWTO was brought (more) up to date (582544)
334 * major progress was made on recovery functionality
336 2002-10-10 Phil Schwan <phil@clusterfs.com>
339 - recovery deadlock fix
340 - rm -rf causes LBUG fix (617817)
341 - file open by multiple tasks fix (618962)
342 - directory permissions bugs (602707 and 620007)
343 - journal_stop fixed with locking (611313)
344 - O_APPEND failures resolved (618273, perhaps 614459)
345 - lconf PATH fix (619770)
346 - IA64 build fix (621450)
347 - RPC buffer sizes scale with amount of memory
349 2002-10-01 Phil Schwan <phil@clusterfs.com>
352 - locks would be cancelled without throwing away data pages,
353 resulting in inconsistent data (605627)
354 - inode attributes were not always being refreshed (605627, 612449)
355 - lconf now continues to cleanup after lctl reports an error
356 - MDS now enforces user permissions (602707)
357 - lprocfs cleanup fixed, but not yet enabled (614157)
358 - fixed infinite server hang, should a client not respond to an AST
359 - avoid going into recovery if user calls readlink() with a buffer
360 that's too small (613941)
361 - AST RPCs no longer require replies (614867) -- this may be changed
362 - don't crash server if client sends an IOV that's too big (611336)
363 - fixed lock conversion deadlock (611892)
364 - fixed the following of symlinks (614622)
365 * recovery: the server can remove locks from a client that dies, other
366 clients can make progress
367 * more extN patch fixes
368 * compile-time configurable ptlrpc buffer allocations
370 - collaborative read cache document
371 - Lustre Lite Performance CDR document-in-progress
373 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
376 - fix typo in patch-2.4.18
378 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
381 - clear ptlrpc request each time in handle_incoming_request()
382 - unlink of files now destroys the object on the OST
384 2002-09-19 Peter Braam <braam@clusterfs.com>
386 * add hard link support
387 * change obdfile creation method
388 * kernel patch changed
390 2002-09-19 Peter Braam <braam@clusterfs.com>
393 - stack overflow bug in extN fixed
395 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
397 * documentation updates
398 - add man pages for config tools
399 - update tests/README to describe testing with new config tools
400 - finish metadata API descriptions
401 * bug fixes and cleanups
402 - statfs workaround for 16TB limit
403 - LOV stripe allocation improved, can stripe on subset of OSTs
404 - LOV file size/IO offset was wrong for files > 4GB in size
405 - object EA data was being dropped, caused files to be unreadable
406 - memory overflow with non-LOV OST caused memory corruption
407 - fixed regression tests to work with new config tools, obdfilter
408 - fixed bug when directory size became larger than 1 block
409 - fixed bug (for single client case) when PWD was deleted
410 - invalidate local directory pages when doing intent-based ops
411 - avoid LDLM oops when lock callback contained bad data
413 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
415 * documentation updates
416 * bug fixes and cleanups
417 - configuration tools
420 - 64-bit compile warnings
421 - 64-bit internal statfs data
423 * test_brw on persistent OST devices
427 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
429 * documentation updates
430 * bug fixes and cleanups
431 * configuration tools
433 2002-08-30 Peter J. Braam <braam@clusterfs.com>
436 * many small fixes to 0_5_4
437 * io/network handling
438 * thinkos in MDS operations
440 2002-08-24 Peter J. Braam <braam@clusterfs.com>
443 * crucial basic fixes to 0.5.3
444 * IOR, Iozone work over Elan
447 2002-08-07 Phil Schwan <phil@clusterfs.com>
448 * version 0_5_3, our first alpha
449 * we use the new Portals iovs
450 * documentation updates
451 * bug fixes and cleanups
452 * small changes in the DLM wire protocol
454 2002-07-25 Peter J. Braam <braam@clusterfs.com>
455 * version 0_5_1 with some initial stability,
456 * locking on MD and file I/O.
457 * documentation updates
458 * several bug fixes since 0.5.0
459 * small changes in wire protocol
461 2002-07-18 Phil Schwan <phil@clusterfs.com>
463 * delivered as Lustre Light Alpha
464 * fixed a crash after handling invalid MDS requests
465 * fixed directory pages for architectures with non-4k pages sizes
467 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
468 * release version v0_4_4
469 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
471 * Fixes a number of interruption problems with OST operations.
472 * Update documentation for portals header changes
473 * Move all wire protocol structs/defines to lustre_idl.h
474 * Fixes symlink length bug.
475 * Add tcpdump to repository.
477 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
478 * release version v0_4_3
479 * Fixes statfs for inodes on extN.
480 * Fixes bug in runtests which would delete /etc/hosts.
481 * Use 64-bit object IDs wherever possible (not into VFS though)
482 Remove ost_get_info, which is unused by lustre, and out of date.
484 2002-07-03 Peter Braam <braam@clusterfs.com>
485 * release version v0_4_2 Fixes a lookup error (type not passed)
486 * move forward to head of Portals
487 * move forward to latest Lustre kernel
489 2002-06-25 Peter Braam <braam@clusterfs.com>
490 * release version v0_4_1. Hopefully stable on single node use.