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 - make TCP zerocopy and pinger support enabled by default (2476)
10 - sync writes completed after process exits caused crashes (2319)
11 - maintain correct mount count on the MDS (2356)
12 - backout 1557, because 2316 wasn't really fixed
13 - better file I/O statistics gathering in /proc
14 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
15 - another kernel patch to fix zero-copy TCP function export
16 - don't take duplicate lock when processing re-sent getattr (2420)
17 - lctl uses obd_self_export instead of creating new conn (2353)
18 - MDS/OST recovery case which requires object creation asserted (2425)
19 - move lfs from /usr/sbin to /usr/bin in packages
20 - fix race between mds_client_add and mds_client_free (2417)
21 - use kmalloc instead of slabs in portals (2430)
22 - don't create duplicate records when a failover MDS is present (2442)
23 - remove unnecessary mount age check (2332)
24 - don't remove directory inodes from locks prematurely (2451)
26 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
29 - remove now-unused request->rq_obd (278)
30 - if an allocation fails, print out how much memory we've used (1933)
31 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
32 - add the "configurable stack size" patch to most series files (1256)
33 - ability to write large log records, for 100+ OST configs (2306)
34 - fix NULL deref when filter_prep fails (2314)
35 - fix operator precedence error in filter_sync
36 - dynamic allocation of socknal TX descriptors (2315)
37 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
38 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
39 - ensure that CWARN messages in Portals always get to the syslog
40 - __init/__exit are not for prototype decls (ldlm_init/exit)
41 - x86-64 compile warning fixes
42 - fix gateway LMC keyword conflict (2318)
43 - fix MDS lock inversions in getattr/reint paths (1844)
44 - fix a rare lock re-ordering bug, which caused deadlock (2322)
45 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
46 - protect MDS inode fsdata with stronger locking (2313)
47 - better error messages when a client is rejected during recovery (1505)
48 - avoid cancelling locks which were never granted, after failure (2330)
49 - fix i_sem/journal inversion in mds_client_add (2333)
50 - fix truncate/getattr lock cycle deadlock (2334)
52 - allow configurable automake binary, for testing new versions
53 - small update to the lfs documentation
55 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
57 * fix negative export reference count in fsfilt_sync (2312)
59 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
60 * release candidate 0.9.1
62 - orphans are moved into the PENDING directory for possible recovery
63 - replayed opens now open by fid for orphan/rename safety (1042)
64 - last close of an orphan inode generates a transno (683)
65 - chdir() and mount() now pin the directory entry (1020)
66 - avoid CERROR in normal ll_setattr_raw() error case (1500)
67 - discard very old requests without processing them (1502)
68 - remove some common, well-understood CERRORs (1505)
69 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
70 - clear "grant" flags in OST replies until OST grant code lands (1644)
71 - fix read performance by not clobbering i_blksize on client (1598)
72 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
73 - make LRU size a /proc tunable, clears locks when reduced (707)
74 - fix some lprocfs rot that prevented ptlbd from loading (1732)
75 - server locks take references on exports now (1558)
76 - build fixes for 2.4.20-rh trees (1663)
77 - return an error from lov_create if all OSCs are inactive (1751)
78 - fix import levels when a reconnect happens without a timeout (1597)
79 - exit early from mds_open if we get a lookup error (1749)
80 - partial page read at EOF wouldn't wait for disk before sending (1642)
81 - avoid NULL deref in obdfilter when reading page past EOF (1592)
82 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
83 - fix LBUG in ll_it_open_error with rc = -2 (1861)
84 - write/truncate lock inversion (1639)
85 - Don't auto-load obdclass, portals modules during cleanup (1495)
86 - fix timestamps from jumping to "now" (1763)
87 - extra journal assertions (1648)
88 - add an extra multiunlink test (1771)
89 - fix read_record/write_record API (1776)
90 - fix leak of offset_extent, possible incorrect i_size later (1772)
91 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
92 - Debugging for the kqswnal_get_idle_tx problems (1820)
93 - Allow recovery to be attempted multiple times (1536)
94 - Write out MDS last_rcvd file after it is first created (1600)
95 - Fix tx_descriptor leak in failed transmit situations (1827)
96 - ext3 journaling fixes for assertion failure after IO error (1871)
97 - class_export_put() on freed export after completion AST error (1896)
98 - Fix revalidate looping in VFS (1322)
99 - Don't access a freed export during MDS_REINT timeout (1521)
100 - Add open-unlink recovery support on the MDS (1673,1764)
101 - Return an error if no MDS data was read from last_rcvd (1946)
102 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
103 - Don't setup a disk device that doesn't match exported UUID (317)
104 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
105 - avoid committing NULL handle in force close
106 - local.sh is now a one-stripe LOV configuration
107 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
108 - fix ext3 htree duplicate directory entry corruption (1516)
109 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
110 - update to new LOV EA format (2097)
111 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
113 2003-06-15 Phil Schwan <phil@clusterfs.com>
116 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
117 - per-import/export recovery handling (958, 931, 959)
118 - multiple last-rcvd slots, for serving multiple FSes (949)
119 - connections are again shared between multiple imp/exports (963, 964)
120 - "umount -f" would hang if any requests needed to be sent (393, 978)
121 - avoid pinning large req buffer by copying for queued messages (989)
122 - add "uuid" to "lctl device" command to help upcalls (991)
123 - "open" RPCs with transnos would confuse recovery counters (1037)
124 - do proper endian conversion of all wire messages (288, 340, 891)
125 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
126 - call waiting locks callback from LDLM recovery thread (1127, 1151)
127 - fix ptlrpc_connection leak in target_handle_connect (1174)
128 - fix import refcounting bug in OST and MDS cleanup (1134)
129 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
130 - fix very unlikely obd_types race condition (501)
131 - remove osc_open hack for echo_client (1187)
132 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
133 - a failure in read_inode2 leads to deadlock (1139)
134 - cancel ack-locks as soon as transaction is committed (1072)
135 - fix major leaks and crashes in the bulk I/O path (937, 1057)
136 - make sure to commitrw after any preprw to avoid deadlock (1162)
137 - failing to execute a file in a lustre FS would lock inode (1203)
138 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
139 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
140 - obd_uuid2tgt would walk off the end of the list (1255)
141 - on IA64 the osc would give portals incorrect bulk size (1258)
142 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
143 - fix lock inversion caused by new llite matching code (1282)
144 - limit the number of dirty pages on a client to 10MB (1286)
145 - timed out locks were not being corrected cancelled (1289)
146 - fix O_DIRECT above 4GB on IA-32 (1292)
147 * major user-visible changes
148 - fail out/fail over policy now controlled by the upcall (993)
150 - add OBD_PING to check server availability and failure (954)
151 - lustre messages are now sent in sending host order (288, 340, 891)
152 - add eadatalen to MDS getattr reply (340)
153 - OST read replies may contain second buffer, with per-page status (593)
155 2003-03-11 Phil Schwan <phil@clusterfs.com>
158 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
159 - failed open()s could cause deadlock; fixed (867, 869)
160 - stop cancelling OST locks when files are closed (481)
161 - overlapping XID spaces caused network corruption (851, 853)
162 - fix unsafe fsfilt counter arithmetic; change to atomic_t
163 - setattr_raw added, to do single-RPC, server-side setattrs
164 - lmc/lconf syntax change for OST UUIDs
165 - fix crashy race condition between ptlrpc_free_req and osc_close
166 - don't use request in mdc_enqueue if we hit a timeout (889)
167 - don't set the inode i_size for regular files from the MDS (896)
168 - handle out of order completion AST (842)
169 - don't LBUG if a lock request times out after receiving AST (913)
170 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
171 - if a bad lock AST arrives, send an error instead of dropping entirely
172 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
173 - fix leak in bulk IO when only partially completed (899, 900, 926)
174 - fix O_DIRECT for ia64 (55)
175 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
176 - C-z after timeout could hang a process forever; fixed (977)
178 - client-side I/O cache (678, 924, 929, 941, 970)
180 - READPAGE and SETATTRs which don't take server-side locks get
183 2003-02-11 Phil Schwan <phil@clusterfs.com>
186 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
187 - Fix fsfilt_extN_readpage() to read a full page of directory
188 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
189 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
190 - Fix imbalanced LOV object allocation and out-of-bound access (469)
191 - Most intent operations were removed, in favour of a new RPC mode
192 that does a single RPC to the server and bypasses most of the VFS
193 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
194 - Aggressively cancel local locks on DLM servers
195 - mds_reint_unlink sends EA to the client if it's the last nlink.
196 client uses that EA to unlink OST objects.
197 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
198 - recursive symlinks were fixed (439)
199 - fixed NULL deref in DEBUG_REQ
200 - filter_update_lastobjid no longer calls sync, which annoyed extN
201 - fixed multi-client small-writes to a single file problem (445)
202 - fixed mtime updates during file writes (607)
203 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
204 - fixed bug in obd_brw_read/write() (under guise of testing 367)
205 - fixed Linux OST size reporting problem (444, 656)
206 - OST now updates object mtime with writes or setattr (607, 619)
207 - client verifies file size before zeroing page past EOF (445)
208 - OST now writes last allocated objid to disk with allocation (108)
209 - LOV on echo now works (409)
211 - mds_reint_unlink sends a new buffer, with the EA included. this
212 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
213 set if a regular file was being unlinked, and it was the last link
214 - use PtlGet from the target for bulk writes (315)
215 - OST now updates object mtime with writes or setattr (607, 619)
216 - LDLM now has a grant-time callback to revalidate locked items, if
218 - Many MDS operations were reorganized to combat race conditions
220 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
221 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
223 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
226 - Fully reactivate OST imports after reconnection (512, others)
227 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
228 - More graceful error handling for truncating on dead OST (515)
229 - Don't error out unless we're actually accessing dead stripes (474)
230 - Fix garbage sizes when stripes are missing (410)
231 - LRU counters were broken, causing constant lock purge (433, 432)
232 - garbage on read from stripes with failed OSTs (441)
233 - mark OSCs as active before reconnecting during recovery (438)
234 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
235 - lfind did not preserve OST order in output (443)
236 - symlinks cause hung clients, incorrect data (439)
237 - stop dereferencing request after dropping refcount (457)
238 - don't LASSERT(spin_is_locked) on non-SMP (455)
239 - fixes for many rename() bugs
240 - fstat didn't correctly synchronize attributes (399)
241 - server must handle lock cancellation during blocking AST prep (487)
242 - bulk descriptors were free()d too soon (511)
243 - fix paths in lconf, which would load incorrect modules (451, 507)
244 - fix confusing lconf 'host not found' error message (386)
245 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
246 - fix race condition in mdc_blocking_ast() for inode access (526)
247 - fix lov_unpackmd() unpacking wrong number of stripes (537)
248 - fix lov_set_osc_active() marking wrong OSC inactive (440)
249 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
250 - fix multiple writes of stripe MD to MDS (358, maybe 519)
251 - fix lstripe in several ways (kernel side) (527)
252 - fix request leak in ldlm_cli_enqueue (262)
253 - incorrect OSC was marked inactive after OST failure
254 - call mds_fs_cleanup before unmounting filesystem (524)
255 - fix races between taking ns_lock and ldlm_lock_change_resource
256 - fix races updating LOV export open file list
257 - fix lov_enqueue error path, avoid decref-ing bad lock handle
258 - fix recovery NULL deref in ldlm_cli_cancel_unused
259 - fix some DLM races by using new hash table for lock handles (419)
260 - permit the client to specify desired inodes, at replay
261 - duplicate requests when we queue them for replay reintegration
262 - fix last_rcvd offset calculation
263 - sync after each recovered transaction, so we always make progress
264 - never, not always, ERESTART requests without transnos
265 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
267 - skip replay if the MDS says that the client is already connected
268 - don't check for a recovery-enabled export to match lctl's UUID
269 - don't INC_USE_COUNT for phantom exports
270 - don't crash when cleaning up phantom exports (567)
271 - don't double-finish or set replay data for errored mdc_open requests
272 - abort requests when they time out, so we don't get old replies
273 - send/receive replies for AST messages again
274 - if the client says that it doesn't have the lock, cancel it on the
276 - if we timeout during I/O, don't try to cancel an in-use lock; instead
277 mark it as destroyed, it will all work out when decref is called
278 - fix module use counts (22, 581)
280 - ASTs now expect a reply (server cancels lock on error reply)
282 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
285 - fix many simultaneous client startup (392)
286 - fix dentry->d_it clobbering
287 - credentials weren't being shipped for readdir/getattr operations
288 - remove invalid assertions triggered during some concurrent MD
290 - proper Lustre versions added (336, 389)
291 - fix memory leak for create error case (398)
292 - fix LOV locking bug that would get cli/srv out of sync
293 - fix echo client over LOV (409)
294 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
295 - fix double-O_EXCL intent crash (424)
296 - avoid sending multiple lock CANCELs (352)
298 - MDS can do multi-client recovery (modulo bugs in new code)
300 - many updates, edits, cleanups
302 2002-11-18 Phil Schwan <phil@clusterfs.com>
305 - fix null d_it dereference (346)
306 - fix full OST/dbench hang (333)
307 - fix permission problem with file removal (286)
308 - fix removal of OSCs from LOV when they fail
309 - fix NULL deref during bulk timeout (214)
310 - fix problems related to multiple filesystems on one MDS (241)
311 - fixed serious subtle metadata locking bugs
312 - free locks on clients when inodes are removed due to memory
314 - fix inode pointer in lock data (285)
315 - partial support for multiple MDS on a single host (241)
316 - data locks weren't cancelled at clear_inode time (290, 311)
317 - intent locks could lead to unbounded lock growth (205)
318 - added a maximum lock count, an LRU list, and a flusher
319 - fix multiple rename (365)
320 - properly abstracted the echo client
321 - OSC locked 1 byte too many; fixed
322 - rewrote brw callback code:
323 - fixed recovery bugs related to LOVs (306)
324 - fixed too-many-pages-in-one-write crash (191)
325 - fixed (again) crash in sync_io_timeout (214)
326 - probably fixed callback-related race (385)
328 - Add capability to MDS protocol
329 - LDLM cancellations and callbacks on different portals
331 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
334 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
335 - defer open object destruction to close time (601981, 138)
336 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
337 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
338 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
339 - don't call obd_setattr in ll_file_release for destroyed objects
340 * protocol change to lustre_msg: move |version| and add |flags|
341 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
342 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
343 * added replay of create, unlink, link and rename operations during
344 MDS failover; recovery should be much more robust now
345 * remove failed OSCs from LOVs (only lov_create uses this so far)
346 * the lustre-HOWTO was brought (more) up to date (582544)
348 2002-10-23 Phil Schwan <phil@clusterfs.com>
351 - in-use dentries weren't being reused properly (617851)
352 - prevent multiple LDLM setup (599178)
353 - fix LOV size calculations for truncate (617853)
354 - fix client handling of MDS intent errors (POSIX)
355 - fix permission bug in lovstripe.c test (624321)
356 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
357 - truncate past end of file could corrupt data
358 - proper cleanup after timeouts, crashes, etc (592524, 550815)
359 - a race in recovery could return ETIMEDOUT to apps (623947)
360 - building outside the source directory was fixed
361 * the lustre-HOWTO was brought (more) up to date (582544)
362 * major progress was made on recovery functionality
364 2002-10-10 Phil Schwan <phil@clusterfs.com>
367 - recovery deadlock fix
368 - rm -rf causes LBUG fix (617817)
369 - file open by multiple tasks fix (618962)
370 - directory permissions bugs (602707 and 620007)
371 - journal_stop fixed with locking (611313)
372 - O_APPEND failures resolved (618273, perhaps 614459)
373 - lconf PATH fix (619770)
374 - IA64 build fix (621450)
375 - RPC buffer sizes scale with amount of memory
377 2002-10-01 Phil Schwan <phil@clusterfs.com>
380 - locks would be cancelled without throwing away data pages,
381 resulting in inconsistent data (605627)
382 - inode attributes were not always being refreshed (605627, 612449)
383 - lconf now continues to cleanup after lctl reports an error
384 - MDS now enforces user permissions (602707)
385 - lprocfs cleanup fixed, but not yet enabled (614157)
386 - fixed infinite server hang, should a client not respond to an AST
387 - avoid going into recovery if user calls readlink() with a buffer
388 that's too small (613941)
389 - AST RPCs no longer require replies (614867) -- this may be changed
390 - don't crash server if client sends an IOV that's too big (611336)
391 - fixed lock conversion deadlock (611892)
392 - fixed the following of symlinks (614622)
393 * recovery: the server can remove locks from a client that dies, other
394 clients can make progress
395 * more extN patch fixes
396 * compile-time configurable ptlrpc buffer allocations
398 - collaborative read cache document
399 - Lustre Lite Performance CDR document-in-progress
401 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
404 - fix typo in patch-2.4.18
406 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
409 - clear ptlrpc request each time in handle_incoming_request()
410 - unlink of files now destroys the object on the OST
412 2002-09-19 Peter Braam <braam@clusterfs.com>
414 * add hard link support
415 * change obdfile creation method
416 * kernel patch changed
418 2002-09-19 Peter Braam <braam@clusterfs.com>
421 - stack overflow bug in extN fixed
423 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
425 * documentation updates
426 - add man pages for config tools
427 - update tests/README to describe testing with new config tools
428 - finish metadata API descriptions
429 * bug fixes and cleanups
430 - statfs workaround for 16TB limit
431 - LOV stripe allocation improved, can stripe on subset of OSTs
432 - LOV file size/IO offset was wrong for files > 4GB in size
433 - object EA data was being dropped, caused files to be unreadable
434 - memory overflow with non-LOV OST caused memory corruption
435 - fixed regression tests to work with new config tools, obdfilter
436 - fixed bug when directory size became larger than 1 block
437 - fixed bug (for single client case) when PWD was deleted
438 - invalidate local directory pages when doing intent-based ops
439 - avoid LDLM oops when lock callback contained bad data
441 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
443 * documentation updates
444 * bug fixes and cleanups
445 - configuration tools
448 - 64-bit compile warnings
449 - 64-bit internal statfs data
451 * test_brw on persistent OST devices
455 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
457 * documentation updates
458 * bug fixes and cleanups
459 * configuration tools
461 2002-08-30 Peter J. Braam <braam@clusterfs.com>
464 * many small fixes to 0_5_4
465 * io/network handling
466 * thinkos in MDS operations
468 2002-08-24 Peter J. Braam <braam@clusterfs.com>
471 * crucial basic fixes to 0.5.3
472 * IOR, Iozone work over Elan
475 2002-08-07 Phil Schwan <phil@clusterfs.com>
476 * version 0_5_3, our first alpha
477 * we use the new Portals iovs
478 * documentation updates
479 * bug fixes and cleanups
480 * small changes in the DLM wire protocol
482 2002-07-25 Peter J. Braam <braam@clusterfs.com>
483 * version 0_5_1 with some initial stability,
484 * locking on MD and file I/O.
485 * documentation updates
486 * several bug fixes since 0.5.0
487 * small changes in wire protocol
489 2002-07-18 Phil Schwan <phil@clusterfs.com>
491 * delivered as Lustre Light Alpha
492 * fixed a crash after handling invalid MDS requests
493 * fixed directory pages for architectures with non-4k pages sizes
495 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
496 * release version v0_4_4
497 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
499 * Fixes a number of interruption problems with OST operations.
500 * Update documentation for portals header changes
501 * Move all wire protocol structs/defines to lustre_idl.h
502 * Fixes symlink length bug.
503 * Add tcpdump to repository.
505 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
506 * release version v0_4_3
507 * Fixes statfs for inodes on extN.
508 * Fixes bug in runtests which would delete /etc/hosts.
509 * Use 64-bit object IDs wherever possible (not into VFS though)
510 Remove ost_get_info, which is unused by lustre, and out of date.
512 2002-07-03 Peter Braam <braam@clusterfs.com>
513 * release version v0_4_2 Fixes a lookup error (type not passed)
514 * move forward to head of Portals
515 * move forward to latest Lustre kernel
517 2002-06-25 Peter Braam <braam@clusterfs.com>
518 * release version v0_4_1. Hopefully stable on single node use.