1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6 Frequency : Clusters with multiple interfaces not on the same subnet
8 Description: Nodes will repeatedly try to reconnect to an interface which it
9 cannot reach and report an error to the log.
10 Details : Extra peer list entries will be created by lconf with some peers
11 unreachable. lconf now validates the peer before adding it.
14 Frequency : Only if a default stripe is set on the filesystem root.
16 Description: Setting a default stripe on the filesystem root prevented the
17 filesystem from being remounted.
18 Details : The client was sending extra request flags in the root getattr
19 request and did not allocate a reply buffer for the dir EA.
22 Frequency : occasional, higher if lots of files are accessed by one client
24 Description: Client trips assertion regarding lsm mismatch/magic
25 Details : While revalidating inodes the VFS looks up inodes with ifind()
26 and in rare cases can find an inode that is being freed.
27 The ll_test_inode() code will free the lsm during ifind()
28 when it finds an existing inode and then the VFS later attaches
29 this free lsm to a new inode.
32 Frequency : occasional
33 Description: While starting a server, the fsfilt_ext3 module could not be
35 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
36 kernels. Previously, lconf would still use the ext3 name
37 when trying to load modules. Now, it will correctly use
38 ext3 on 2.4 and ldiskfs on 2.6.
40 Severity : enhancement
41 Description: The default stripe count has been changed to 1
42 Details : The interpretation of the default stripe count (0, to lfs
43 or lmc) has been changed to mean striping across a single
44 OST, rather than all available. For general usage we have
45 found a stripe count of 1 or 2 works best.
48 Frequency : occasional
50 Description: Creating files with an explicit stripe count may lead to
51 a failed assertion on the MDS
52 Details : If some OSTs are full or unavailable, creating files may
53 trigger a failed assertion on the MDS. Now, Lustre will
54 try to use other servers or return an error to the
58 Frequency : occasional
60 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
61 caused lock timeout during readahead (since 1.4.2).
62 Details : Processes doing readahead might match a lock that hasn't been
63 granted yet if there are overlapping and conflicting lock
64 requests. The readahead process waits on ungranted lock
65 (original lock is CBPENDING), while OST waits for that process
66 to cancel CBPENDING read lock and eventually evicts client.
68 ------------------------------------------------------------------------------
70 2005-06-20 Cluster File Systems, Inc. <info@clusterfs.com>
75 Frequency : rare (extremely heavy IO load with hundreds of clients)
77 Description: Client is evicted, gets IO error writing to file
78 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
79 caused clients to be evicted for AST timeouts. The fixes in
80 bug 5192 mean we no longer need to have such short AST timeouts
81 so ldlm_timeout has been increased.
84 Frequency : occasional during --force or --failover shutdown under load
86 Description: Server oops/LBUG if stopped with --force or --failover under load
87 Details : a collection of import/export refcount and cleanup ordering
88 issues fixed for safer force cleanup
91 Frequency : only filesystems larger than 120 OSTs
93 Description: lfs getstripe would oops on a very large filesystem
94 Details : lov_getconfig used kfree on vmalloc'd memory
97 Frequency : only filesystems exporting via NFS to Solaris 10 clients
99 Description: reading from files that had been truncated to a non-zero size
100 but never opened returned no data
101 Details : ll_file_read() reads zeros from no-object files to EOF
106 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
107 Details : There's little harm in aborting MDS/OSS recovery and letting it
108 try again, so I removed the LASSERT and return an error instead.
110 Severity : enhancement
112 Description: New debugging infrastructure for tracking down data corruption
113 Details : The I/O checksum code was replaced to: (a) control it at runtime,
114 (b) cover more of the client-side code path, and (c) try to narrow
115 down where problems occurred
119 Bugzilla : 3819, 4364, 4397, 6313
120 Description: Racing close and eviction MDS could cause assertion in mds_close
121 Details : It was possible to get multiple mfd references during close and
122 client eviction, leading to one thread referencing a freed mfd.
124 Severity: : enhancement
125 Bugzilla : 3262, 6359
126 Description: Attempts to reconnect to servers are now more aggressive.
127 Details : This builds on the enhanced upcall-less recovery that was added
128 in 1.4.2. When trying to reconnect to servers, clients will
129 now try each server in the failover group every 10 seconds. By
130 default, clients would previously try one server every 25 seconds.
135 Description: After recovery, certain operations trigger a failed
136 assertion on a client.
137 Details : Failing over an mds, using lconf -d --failover, while a
138 client was doing a readdir() call would cause the client to
139 LBUG after recovery completed and the readdir() was resent.
141 Severity : enhancement
143 Description: Default groups are now added by lconf
144 Details : You can now run lconf --group <servicename> without having to
145 manually add groups with lmc.
148 Frequency : occasional
150 Description: Nodes with an elan id of 0 trigger a failed assertion
153 Frequency : always when accessing e.g. tty/console device nodes
155 Description: tty and some other devices nodes cannot be used on lustre
156 Details : file's private_data field is used by device data and lustre
157 values in there got lost. New field was added to struct file to
158 store fs-specific private data.
161 Frequency : when exporting Lustre via NFS
163 Description: NFSD failed occasionally when looking up a path component
164 Details : NFSD is looking up ".." which was broken in ext3 directories
165 that had grown large enough to become hashed.
167 ------------------------------------------------------------------------------
169 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
171 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
172 versions of Lustre. Please update all servers and clients to
173 version 1.4.2 or later at the same time. You must also run
174 "lconf --write-conf {config}.xml" on the MDS while it is stopped
175 to update the configuration logs.
177 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
178 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
179 - swab llog records of type '0' so we get proper header size/idx (5861)
180 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
181 - fix rename of one directory over another leaking an inode (5953)
182 - avoid SetPageDirty on 2.6 (5981)
183 - don't re-add just-being-destroyed locks to the waiting list (5653)
184 - when creating new directories, inherit the parent's custom
185 striping settings if present parent (3048)
186 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
187 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
188 - don't hold client locks on temporary worklist from l_lru (5666)
189 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
190 - automatically evict dead clients (5921)
191 - Update file size properly in create+truncate+fstat case (6196)
192 - Do not unhash mountpoint dentries, do not allow removal of
194 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
195 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
196 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
197 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
198 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
199 - Correctly handle reads of files with no objects (6243)
200 - lctl recover will also mark a device active if deactivate used (5933)
202 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
203 - allow --write-conf on an MDS with different nettype than client (5619)
204 - don't write config llogs to MDS for mounts not from that MDS (5617)
205 - lconf should create multiple TCP connections from a client (5201)
206 - init scripts are now turned off by default; run chkconfig --on
207 lustre and chkconfig --on lustrefs to use them
208 - upcalls are no longer needed for clients to recover to failover
210 - add --abort-recovery option to lconf to abort recovery on device
212 - add support for an arbitrary number of OSTs (3026)
213 - Quota support protocol changes.
214 - forward compatibility changes to wire structs (6007)
215 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
216 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
217 - improved "lustre status" script
218 - initialize blocksize for non-regular files (6062)
219 - added --disable-server and --disable-client configure options (5782)
220 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
221 - Vanilla 2.4.29 support
222 - increase maximum number of obd devices to 520 (6242)
223 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
224 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
226 - limit stripes per file to 160 (the maximum EA size) (6093)
228 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
231 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
232 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
233 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
234 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
235 - small ext3 extents cleanups and fixes (5733)
236 - improved mballoc code, several small races and bugs fixed (5733, 5638)
237 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
238 - avoid needless client->OST connect, fix handle mismatch (5317)
239 - fix DLM error path that led to out-of-sync client, long delays (5779)
240 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
241 - fix several locking issues related to i_size (5492,5624,5654,5672)
242 - don't move pending lock onto export if it is already evicted (5683)
243 - fix kernel oops when creating .foo in unlinked directory (5548)
244 - fix deadlock in obdfilter statistics vs. object create (5811)
245 - use time_{before,after} to avoid timer jiffies wrap (5882)
246 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
247 - Do not leak request if server was not able to process it (5154)
248 - If mds_open unable to find parent dir, make that negative lookup(5154)
249 - don't create new directories with extent-mapping (5909, 5936)
251 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
252 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
253 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
254 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
255 - support for RHEL4 kernel on i686 (5773)
256 - provide error messages when incompatible logs are encountered (5898)
258 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
259 * version 1.4.0.10 (1.4.1 release candidate 1)
261 - don't keep a lock reference when lock is not granted (4238)
262 - unsafe list practices (rarely) led to infinite eviction loop (4908)
263 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
264 - drop import inflight refcount on signal_completed_replay error (5255)
265 - unlock page after async write error during send (3677)
266 - handle missing objects in filter_preprw_read properly (5265)
267 - no transno return for symlink open, don't save no-trasno open (3440)
268 - don't try to complete elan receive that already failed (4012)
269 - free RPC server reply state on error (5406)
270 - clean up thread from ptlrpc_start_thread() on error (5160)
271 - readahead could read extra page into cache that wasn't ejected (5388)
272 - prevent races in class_attach/setup/cleanup/detach (5260)
273 - don't dereference de->d_inode after l_dput of de (5458)
274 - use "int" for stripe value returned from lock_to_stripe (5544)
275 - mballoc allocation and error-checking fixes in 2.6 (5504)
276 - block device patches to fix I/O request sizes in 2.6 (5482)
277 - look up hostnames for IB nals (5602)
278 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
279 - don't start multiple acceptors for the same port (5277)
280 - fix incorrect LASSERT in mds_getattr_name (5635)
281 - export a proc file for general "ping" checking (5628)
282 - fix "lfs check" to not block when the MDS is down (5628)
284 - service request history (4965)
285 - put {ll,lov,osc}_async_page structs in a single slab (4699)
286 - create an "evict_client" /proc entry on OSTs, like the MDS has
287 - fix mount usage message, return errors per mount(8) (5168)
288 - change grep [] to grep "[]" in tests so they work in more UMLs
289 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
290 - remove ancient LOV_MAGIC_V0 EA support (5047)
291 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
292 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
293 - allow readahead allocations to fail when low on memory (5383)
294 - mmap locking landed again, after considerable improvement (2828)
295 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
297 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
300 - send OST transaction number in read/write reply to free req (4966)
301 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
302 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
303 - the watchdog thread now runs as interruptible (5246)
304 - flock/lockf fixes (but it's still disabled, pending 5135)
305 - don't use EXT3 constants in llite code (5094)
306 - memory shortage at startup could cause assertion (5176)
308 - reorganization of lov code
309 - single portals codebase
311 - add extents/mballoc support (5025)
312 - direct I/O reads in the obdfilter (4048)
313 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
315 tbd Cluster File Systems, Inc. <info@clusterfs.com>
318 - send OST transaction number in read/write reply to free req (4966)
319 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
320 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
321 - the watchdog thread now runs as interruptible (5246)
322 - handle missing objects in filter_preprw_read properly (5265)
323 - unsafe list practices (rarely) led to infinite eviction loop (4908)
324 - drop import inflight refcount on signal_completed_replay error (5255)
325 - unlock page after async write error during send (3677)
326 - return original error code on reconstructed replies (3761)
327 - no transno return for symlink open, don't save no-trasno open (3440)
329 - add pid to ldlm debugging output (4922)
330 - bump the watchdog timeouts -- we can't handle 30sec yet
331 - extra debugging for orphan dentry/inode bug (5259)
333 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
336 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
337 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
338 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
339 - fix NULL dereference in /proc/sys/portals/routes (4827)
340 - allow failed mdc_close() operations to be interrupted (4561)
341 - stop precreate on OST before MDS would time out on it (4778)
342 - don't send partial-page writes before EOF from client (4410)
343 - discard client grant for sub-page writes on large-page clients (4520)
344 - don't free dentries not owned by NFS code, check generation (4806)
345 - fix lsm leak if mds_create_objects() fails (4801)
346 - limit debug_daemon file size, always print CERROR messages (4789)
347 - use transno after validating reply (3892)
348 - process timed out requests if import state changes (3754)
349 - update mtime on OST during writes, return in glimpse (4829)
350 - add mkfsoptions to LDAP (4679)
351 - use ->max_readahead method instead of zapping global ra (5039)
352 - don't interrupt __l_wait_event() during strace
354 - add software watchdogs to catch hung threads quickly (4941)
355 - make lustrefs init script start after nfs is mounted
356 - fix CWARN/ERROR duplication (4930)
357 - return async write errors to application if possible (2248)
358 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
359 - print NAL number in %x format (4645)
360 - update barely-supported suse-2.4.21-171 series (4842)
361 - support for sles 9 %post scripts
362 - support for building 2.6 kernel-source packages
363 - support for sles km_* packages
365 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
368 - ignore -ENOENT errors in osc_destroy (3639)
369 - notify osc create thread that OSC is being cleaned up (4600)
370 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
371 - reconstruct ost_handle() like mds_handle() (4657)
372 - create a new thread to do import eviction to avoid deadlock (3969)
373 - let lconf resolve symlinked-to devices (4629)
374 - don't unlink "objects" from directory with default EA (4554)
375 - hold socknal file ref over connect in case target is down (4394)
376 - allow more than 32000 subdirectories in a single directory (3244)
377 - fix blocks count for O_DIRECT writes (3751)
378 - OST returns ENOSPC from object create when no space left (4539)
379 - don't send truncate RPC if file size isn't changing (4410)
380 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
381 - bind to privileged port in socknal and tcpnal (3689)
383 - rate limit CERROR/CWARN console message to avoid overload (4519)
384 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
385 - basic mmap support (3918)
386 - kernel patch series update from b1_4 (4711)
388 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
391 - avoid crash during MDS cleanup with OST shut down (2775)
392 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
393 - don't use bad inodes on the MDS (3744)
394 - dynamic object preallocation to improve recovery speed (4236)
395 - don't hold spinlock over lock dumping or change debug flags (4401)
396 - don't zero obd_dev when it is force cleaned (3651)
397 - print grants to console if they go negative (4431)
398 - "lctl deactivate" will stop automatic recovery attempts (3406)
399 - look for existing locks in ldlm_handle_enqueue() (3764)
400 - don't resolve lock handle twice in recovery avoiding race (4401)
401 - revalidate should check working dir is a directory (4134)
403 - don't always mark "slow" obdfilter messages as errors (4418)
405 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
408 - don't close LustreDB during write_conf until it is done (3860)
409 - fix typo in lconf for_each_profile (3821)
410 - allow dumping logs from multiple threads at one time (3820)
411 - don't allow multiple threads in OSC recovery (3812)
412 - fix debug_size parameters (3864)
413 - fix mds_postrecov to initialize import for llog ctxt (3121)
414 - replace config semaphore with spinlock (3306)
415 - be sure to send a reply for a CANCEL rpc with bad export (3863)
416 - don't allow enqueue to complete on a destroyed export (3822)
417 - down write_lock before checking llog header bitmap (3825)
418 - recover from lock replay timeout (3764)
419 - up llog sem before sending rpc (3652)
420 - reduce ns lock hold times when setting kms (3267)
421 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
422 - fix NULL deref and obd_dev leak on setup error (3312)
423 - replace some LBUG about llog ops with error handling (3841)
424 - don't match INVALID dentries from d_lookup and spin (3784)
425 - hold dcache_lock while marking dentries INVALID and hashing (4255)
426 - fix invalid assertion in ptlrpc_set_wait (3880)
428 - add libwrap support for the TCP acceptor (3996)
429 - add /proc/sys/portals/routes for non-root route listing (3994)
430 - allow setting MDS UUID in .xml (2580)
431 - print the stack of a process that LBUGs (4228)
433 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
436 - don't cleanup request in ll_file_open() on failed MDS open (3430)
437 - make sure to unset replay flag from failed open requests (3440)
438 - if default stripe count is 0, use OST count for inode size (3636)
439 - update parent mtime/ctime on client for create/unlink (2611)
440 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
441 - free recovery state on server during a forced cleanup (3571)
442 - unregister_reply for resent reqs (3063)
443 - loop back devices mounting and status check on 2.6 (3563)
444 - fix resource-creation race that can provoke i_size == 0 (3513)
445 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
446 - more debugging for page-accounting assertion (3746)
447 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
448 - avoid deadlock after precreation failure (3758)
449 - fix race and lock order deadlock in orphan handling (3450, 3750)
450 - add validity checks when grabbing inodes from l_ast_data (3599)
452 - add /proc/.../recovery_status to obdfilter (3428)
453 - lightweight CDEBUG infrastructure, debug daemon (3668)
454 - change default OSC RPC parameters to be better on small clusters
455 - turn off OST read cache for files smaller than 32MB
456 - install man pages and include them in rpms (3100)
457 - add new init script for (un)mounting lustre filesystems (2593)
458 - run chkconfig in %post for init scripts (3701)
459 - drop scimac NAL (unmaintained)
461 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
464 - clean kiobufs before and after use (3485)
465 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
466 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
467 - change init script to fail more clearly if not run as root (1528)
468 - allow clients to reconnect during replay (1742)
469 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
470 - don't do DNS lookups on NIDs too small for IP addresses (3442)
471 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
472 - fix cond_resched (3554)
473 - only evict unfinished clients after recovery (3515)
474 - allow bulk resend, prevent data loss (3570)
475 - dynamic ptlrpc request buffer allocation (2102)
476 - don't allow unlinking open directory if it isn't empty (2904)
477 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
478 - remove extraneous obd dereference causing LASSERT failure (3334)
479 - don't use get_cycles() when creating temp. files on the mds (3156)
480 - hold i_sem when setting i_size in ll_extent_lock() (3564)
481 - handle EEXIST for set-stripe, set proper directory name (3336)
483 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
484 - fix ksocknal_fmb_callback() error messages (2918)
486 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
489 - don't copy lvb into (possibly NULL) reply on error (2983)
490 - don't deref dentry after dput, don't free lvb on error (2922)
491 - use the kms to determine writeback rpc length (2947)
492 - increment oti_logcookies when osc is inactive (2948)
493 - update client's i_blocks count via lvb messages (2543)
494 - handle intent open/close of special files properly (1557)
495 - mount MDS with errors=remount-ro, like obdfilter (2009)
496 - initialize lock handle to avoid ASSERT on error cleanup (3057)
497 - don't use cancelling-locks' kms values (2947)
498 - use highest lock extent for kms, not last one (2925)
499 - don't dereference ERR_PTR() dentry in error handling path (3107)
500 - fix thread race in portals_debug_dumplog() (3122)
501 - create lprocfs device entries at setup instead of at attach (1519)
502 - common AST error handler, don't evict client on completion race (3145)
503 - zero nameidata in detach_mnt in 2.6 (3118)
504 - verify d_inode after revalidate_special is valid in 2.6 (3116)
505 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
506 - initialize RPC timeout timer earlier for 2.6 (3219)
507 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
508 - print nal/nid for unknown nid (3258)
509 - additional checks for oscc recovery before doing precreate (3284)
510 - fix ll_extent_lock() error return code for 64-bit systems (3043)
511 - don't crash in mdc_close for bad permissions on open (3285)
512 - zero i_rdev for non-device files (3147)
513 - clear page->private before handing to FS, better assertion (3119)
514 - tune the read pipeline (3236)
515 - fix incorrect decref of invalidated dentry (2350)
516 - provide read-ahead stats and refine rpc in flight stats (3328)
517 - don't hold journal transaction open across create RPC (3313)
518 - update atime on MDS at close time (3265)
519 - close LDAP connection when recovering to avoid server load (3315)
520 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
521 - don't leak open file on MDS after open resend (3325)
522 - serialize filter_precreate and filter_destroy_precreated (3329)
523 - loop device shouldn't call sync_dev() for nul device (3092)
524 - clear page cache after eviction (2766)
525 - resynchronize MDS->OST in background (2824)
526 - refuse to mount the same filesystem twice on same mountpoint (3394)
527 - allow llmount to create routes for mounting behind routers (3320)
528 - push lock cancellation to blocking thread for glimpse ASTs (3409)
529 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
530 - fix rare problem with rename on htree directories (3417)
532 - allow default OST striping configuration per directory (1414)
533 - fix compilation for qswnal for 2.6 kernels (3125)
534 - increase maximum number of MDS request buffers for large systems
535 - change liblustreapi to be useful for external progs like lfsck (3098)
536 - increase local configuration timeout for slow disks (3353)
537 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
539 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
542 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
543 - don't overwrite extent policy data in reply if lock was blocked (2901)
544 - drop filter export grants atomically with removal from device (2663)
545 - del obd_self_export from work_list in class_disconnect_exports (2908)
546 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
547 - swab reply message in mdc_close, other PPC fixes (2464)
548 - fix destroying of named logs (2325)
549 - overwrite old logs when running lconf --write_conf (2264)
550 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
551 - fix race in target_handle_connect (2898)
552 - mds_reint_create() should take same inode create lock (2926)
553 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
554 - don't close files for self_export to avoid uninitialized obd (2936)
555 - allow MDS with the same name as client node (2939)
556 - hold dentry reference for closed log files for unlink (2325)
557 - reserve space for all logs during transactions (2059)
558 - don't evict page beyond end of stripe extent (2925)
559 - don't oops on a deleted current working directory (2399)
560 - handle hard links to targets without a parent properly (2517)
561 - don't dereference NULL lock when racing during eviction (2867)
562 - don't grow lock extents when lots of conflicting locks (2919)
564 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
567 - account for cache space usage on clients to avoid data loss (974)
568 - lfsck support in lustre kernel code (2349)
569 - reduce journal credits needed for BRW writes (2370)
570 - orphan handling to avoid losing space on client/server crashes
571 - ptlrpcd can be blocked, stopping ALL progress (2477)
572 - use lock value blocks to assist in proper KMS, faster stat (1021)
573 - takes i_sem instead of DLM locks internally on obdfilter (2720)
574 - recovery for initial connections (2355)
575 - fixes for mds_cleanup_orphans (1934)
576 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
577 - block all file creations until orphan recovery completes (1901)
578 - client remove rq_connection from request struct (2423)
579 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
580 - recovery timer race (2670)
581 - mdc_close recovey bug (2532)
582 - ptlrpc cleanup bug (2710)
583 - mds timeout on local locks (2588)
584 - namespace lock held during RPCs (2431)
585 - handle interrupted sync write properly (2503)
586 - don't try to handle a message that hasn't been replied to (2699)
587 - client assert failure during cleanup after abort recovery (2701)
588 - leak mdc device after failed mount (2712)
589 - ptlrpc_check_set allows timedout requests to complete (2714)
590 - wait for inflight reqs when ptlrpcd finishes (2710)
591 - make sure unregistered services are removed from the srv_list
592 - reset bulk XID's when resending them (caught by 1138 test)
593 - unregister_bulk after timeout
594 - fix lconf error (2694)
595 - handle write after unfinished setstripe, stripe-only getstripe (2388)
596 - readahead locks pages, leaves pending causing memory pressure (2673)
597 - increase OST request buffers to 4096 on large machines (2729)
598 - fix up permission of existing directories in simple_mkdir (2661)
599 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
600 - don't assign transno to errored transactions (2742)
601 - don't delete objects on OST if given a bogus objid from MDS (2751)
602 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
603 - if rq_no_resend, then timeout request after recovery (2432)
604 - fix MDS llog_logid record size, 64-bit array alignment (2733)
605 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
606 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
607 - MDS recovery shouldn't delete objects that it has given out (2730)
608 - if enqueue arrives after completion, don't clobber LVB (2819)
609 - don't unlock pages twice when trigger_group_io returns error (2814)
610 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
611 - don't write pages to disk if there was an error (1450)
612 - don't ping imports that have recovery disabled (2676)
613 - take buffered bytes into account when balancing socknal conn (2817)
614 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
615 - reconnect unlink llog connection after MDS reconnects to OST (2816)
616 - remove little-endian swabbing of llog records (1987)
617 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
618 - retry reposting request buffers if they fail (1191)
619 - grow extent at grant time to avoid granting a revoked lock (2809)
620 - lock revoke doesn't evict page if covered by a second lock (2765)
621 - disable VM readahead to avoid reading outside lock extents (2805)
623 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
624 - updated kernel patches for hp-2.4.20 kernel (2681)
626 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
629 - fix truncated write corruption (2366)
630 - fix for failed assertion in iopen_connect_dentry (1792,2517)
632 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
633 - ensure len(uuid) < 37 in lmc (1171)
634 - fix ia64 OOPS in llog_test (2255)
635 - zero end of page at obdfilter for partial page writes (2648)
636 - don't leave stale dentries around after renames (bug 2428)
637 - fix timeouts when evicting a client with a single lock held (2642)
638 - set deadline for the initial HELLO message to drain (2634)
639 - print out dotted-quad IP addresses in the socknal (2302)
641 - additional debugging for MDS client eviction problem (2443)
642 - fix mkfsoptions support for osts (2603, 2604)
644 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
647 - add series for the vanilla 2.6.0 kernel
648 - add series for the vanilla 2.4.24 kernel
649 - add series for a cray x86/64 UL kernel drop
650 - fix xattr patches for the vanilla 2.4.19 series
652 - generate true UUIDs in lmc (1171)
653 - have portals stack dumping break in UML (2466)
654 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
655 - allocate with _NOFS in ldlm to avoid deadlock (1933)
656 - wake callback waiting threads on client eviction (2460)
657 - Add --ptldebug and --subsystem to lmc (1719)
658 - update assertion to allow safe interrupt allocation
659 - set rq_no_resend for cancel requests (2432)
660 - recalculate ptlrpcd timeout after resend (2494)
661 - call vfs_rmdir when removing pending directories (2368)
662 - fix renaming a file to itself (2429)
663 - lmc creates a default one-stripe lov (2454)
664 - expand procfs space to handle large clusters (2326)
665 - increase UML stack to avoid overflow
666 - update lconf's list of debug and subsystem masks
667 - fix lfs find --obd (2510)
668 - /proc tunable for disabling filter read caching (2591)
669 - stop rpm packages from altering slapd.conf (2301)
670 - disable nagle in the socknal under 0conf (2578)
671 - choose mds inode size based on stripe count (2572)
672 - fix kernel-source rpm problems (2516)
674 - add --disable-doc to avoid pdf generation (2421)
675 - update documentation, tests, type-os, comments
676 - avoid format warnings on ia64
678 - tiny code cleanups by removing unused fields
680 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
683 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
684 - remove the most blatant lies from BUILDING file (2371)
685 - change default debug level to reasonable production setting
686 - reduce client side cache size to reduce cache flush time
687 - reduce max RPCs in flight to avoid unnecessary file fragmentation
688 - make TCP zerocopy and pinger support enabled by default (2476)
689 - sync writes completed after process exits caused crashes (2319)
690 - maintain correct mount count on the MDS (2356)
691 - backout 1557, because 2316 wasn't really fixed
692 - better file I/O statistics gathering in /proc
693 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
694 - another kernel patch to fix zero-copy TCP function export
695 - don't take duplicate lock when processing re-sent getattr (2420)
696 - lctl uses obd_self_export instead of creating new conn (2353)
697 - MDS/OST recovery case which requires object creation asserted (2425)
698 - move lfs from /usr/sbin to /usr/bin in packages
699 - fix race between mds_client_add and mds_client_free (2417)
700 - use kmalloc instead of slabs in portals (2430)
701 - don't create duplicate records when a failover MDS is present (2442)
702 - remove unnecessary mount age check (2332)
703 - don't remove directory inodes from locks prematurely (2451)
704 - don't break if MDS service name is the same as hostname (2103)
705 - fix races in client write RPC generation when cache full (2482)
707 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
710 - remove now-unused request->rq_obd (278)
711 - if an allocation fails, print out how much memory we've used (1933)
712 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
713 - add the "configurable stack size" patch to most series files (1256)
714 - ability to write large log records, for 100+ OST configs (2306)
715 - fix NULL deref when filter_prep fails (2314)
716 - fix operator precedence error in filter_sync
717 - dynamic allocation of socknal TX descriptors (2315)
718 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
719 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
720 - ensure that CWARN messages in Portals always get to the syslog
721 - __init/__exit are not for prototype decls (ldlm_init/exit)
722 - x86-64 compile warning fixes
723 - fix gateway LMC keyword conflict (2318)
724 - fix MDS lock inversions in getattr/reint paths (1844)
725 - fix a rare lock re-ordering bug, which caused deadlock (2322)
726 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
727 - DLM race condition prevented some lock evictions (2328)
728 - ENOMEM detection and retry on socknal sends (2230)
729 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
730 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
731 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
732 - better error messages when a client is rejected during recovery (1505)
733 - avoid cancelling locks which were never granted, after failure (2330)
734 - fix i_sem/journal inversion in mds_client_add (2333)
735 - fix truncate/getattr lock cycle deadlock (2334)
736 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
737 - fix two rare exit paths which could leak an l_lock() ref (2321)
738 - fencepost error in MDS/OST orphan recovery (2226)
739 - make log record alignment 8 bytes (1988)
740 - lstripe now fails when requested offset > ost_count (2237)
741 - ensure that all kernel series have a complete list.h (1607)
742 - fix crashes in special-file operations (2316)
743 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
745 - allow configurable automake binary, for testing new versions
746 - small update to the lfs documentation
748 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
750 * fix negative export reference count in fsfilt_sync (2312)
752 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
753 * release candidate 0.9.1
755 - orphans are moved into the PENDING directory for possible recovery
756 - replayed opens now open by fid for orphan/rename safety (1042)
757 - last close of an orphan inode generates a transno (683)
758 - chdir() and mount() now pin the directory entry (1020)
759 - avoid CERROR in normal ll_setattr_raw() error case (1500)
760 - discard very old requests without processing them (1502)
761 - remove some common, well-understood CERRORs (1505)
762 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
763 - clear "grant" flags in OST replies until OST grant code lands (1644)
764 - fix read performance by not clobbering i_blksize on client (1598)
765 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
766 - make LRU size a /proc tunable, clears locks when reduced (707)
767 - fix some lprocfs rot that prevented ptlbd from loading (1732)
768 - server locks take references on exports now (1558)
769 - build fixes for 2.4.20-rh trees (1663)
770 - return an error from lov_create if all OSCs are inactive (1751)
771 - fix import levels when a reconnect happens without a timeout (1597)
772 - exit early from mds_open if we get a lookup error (1749)
773 - partial page read at EOF wouldn't wait for disk before sending (1642)
774 - avoid NULL deref in obdfilter when reading page past EOF (1592)
775 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
776 - fix LBUG in ll_it_open_error with rc = -2 (1861)
777 - write/truncate lock inversion (1639)
778 - Don't auto-load obdclass, portals modules during cleanup (1495)
779 - fix timestamps from jumping to "now" (1763)
780 - extra journal assertions (1648)
781 - add an extra multiunlink test (1771)
782 - fix read_record/write_record API (1776)
783 - fix leak of offset_extent, possible incorrect i_size later (1772)
784 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
785 - Debugging for the kqswnal_get_idle_tx problems (1820)
786 - Allow recovery to be attempted multiple times (1536)
787 - Write out MDS last_rcvd file after it is first created (1600)
788 - Fix tx_descriptor leak in failed transmit situations (1827)
789 - ext3 journaling fixes for assertion failure after IO error (1871)
790 - class_export_put() on freed export after completion AST error (1896)
791 - Fix revalidate looping in VFS (1322)
792 - Don't access a freed export during MDS_REINT timeout (1521)
793 - Add open-unlink recovery support on the MDS (1673,1764)
794 - Return an error if no MDS data was read from last_rcvd (1946)
795 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
796 - Don't setup a disk device that doesn't match exported UUID (317)
797 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
798 - avoid committing NULL handle in force close
799 - local.sh is now a one-stripe LOV configuration
800 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
801 - fix ext3 htree duplicate directory entry corruption (1516)
802 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
803 - update to new LOV EA format (2097)
804 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
806 2003-06-15 Phil Schwan <phil@clusterfs.com>
809 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
810 - per-import/export recovery handling (958, 931, 959)
811 - multiple last-rcvd slots, for serving multiple FSes (949)
812 - connections are again shared between multiple imp/exports (963, 964)
813 - "umount -f" would hang if any requests needed to be sent (393, 978)
814 - avoid pinning large req buffer by copying for queued messages (989)
815 - add "uuid" to "lctl device" command to help upcalls (991)
816 - "open" RPCs with transnos would confuse recovery counters (1037)
817 - do proper endian conversion of all wire messages (288, 340, 891)
818 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
819 - call waiting locks callback from LDLM recovery thread (1127, 1151)
820 - fix ptlrpc_connection leak in target_handle_connect (1174)
821 - fix import refcounting bug in OST and MDS cleanup (1134)
822 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
823 - fix very unlikely obd_types race condition (501)
824 - remove osc_open hack for echo_client (1187)
825 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
826 - a failure in read_inode2 leads to deadlock (1139)
827 - cancel ack-locks as soon as transaction is committed (1072)
828 - fix major leaks and crashes in the bulk I/O path (937, 1057)
829 - make sure to commitrw after any preprw to avoid deadlock (1162)
830 - failing to execute a file in a lustre FS would lock inode (1203)
831 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
832 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
833 - obd_uuid2tgt would walk off the end of the list (1255)
834 - on IA64 the osc would give portals incorrect bulk size (1258)
835 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
836 - fix lock inversion caused by new llite matching code (1282)
837 - limit the number of dirty pages on a client to 10MB (1286)
838 - timed out locks were not being corrected cancelled (1289)
839 - fix O_DIRECT above 4GB on IA-32 (1292)
840 * major user-visible changes
841 - fail out/fail over policy now controlled by the upcall (993)
843 - add OBD_PING to check server availability and failure (954)
844 - lustre messages are now sent in sending host order (288, 340, 891)
845 - add eadatalen to MDS getattr reply (340)
846 - OST read replies may contain second buffer, with per-page status (593)
848 2003-03-11 Phil Schwan <phil@clusterfs.com>
851 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
852 - failed open()s could cause deadlock; fixed (867, 869)
853 - stop cancelling OST locks when files are closed (481)
854 - overlapping XID spaces caused network corruption (851, 853)
855 - fix unsafe fsfilt counter arithmetic; change to atomic_t
856 - setattr_raw added, to do single-RPC, server-side setattrs
857 - lmc/lconf syntax change for OST UUIDs
858 - fix crashy race condition between ptlrpc_free_req and osc_close
859 - don't use request in mdc_enqueue if we hit a timeout (889)
860 - don't set the inode i_size for regular files from the MDS (896)
861 - handle out of order completion AST (842)
862 - don't LBUG if a lock request times out after receiving AST (913)
863 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
864 - if a bad lock AST arrives, send an error instead of dropping entirely
865 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
866 - fix leak in bulk IO when only partially completed (899, 900, 926)
867 - fix O_DIRECT for ia64 (55)
868 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
869 - C-z after timeout could hang a process forever; fixed (977)
871 - client-side I/O cache (678, 924, 929, 941, 970)
873 - READPAGE and SETATTRs which don't take server-side locks get
876 2003-02-11 Phil Schwan <phil@clusterfs.com>
879 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
880 - Fix fsfilt_extN_readpage() to read a full page of directory
881 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
882 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
883 - Fix imbalanced LOV object allocation and out-of-bound access (469)
884 - Most intent operations were removed, in favour of a new RPC mode
885 that does a single RPC to the server and bypasses most of the VFS
886 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
887 - Aggressively cancel local locks on DLM servers
888 - mds_reint_unlink sends EA to the client if it's the last nlink.
889 client uses that EA to unlink OST objects.
890 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
891 - recursive symlinks were fixed (439)
892 - fixed NULL deref in DEBUG_REQ
893 - filter_update_lastobjid no longer calls sync, which annoyed extN
894 - fixed multi-client small-writes to a single file problem (445)
895 - fixed mtime updates during file writes (607)
896 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
897 - fixed bug in obd_brw_read/write() (under guise of testing 367)
898 - fixed Linux OST size reporting problem (444, 656)
899 - OST now updates object mtime with writes or setattr (607, 619)
900 - client verifies file size before zeroing page past EOF (445)
901 - OST now writes last allocated objid to disk with allocation (108)
902 - LOV on echo now works (409)
904 - mds_reint_unlink sends a new buffer, with the EA included. this
905 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
906 set if a regular file was being unlinked, and it was the last link
907 - use PtlGet from the target for bulk writes (315)
908 - OST now updates object mtime with writes or setattr (607, 619)
909 - LDLM now has a grant-time callback to revalidate locked items, if
911 - Many MDS operations were reorganized to combat race conditions
913 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
914 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
916 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
919 - Fully reactivate OST imports after reconnection (512, others)
920 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
921 - More graceful error handling for truncating on dead OST (515)
922 - Don't error out unless we're actually accessing dead stripes (474)
923 - Fix garbage sizes when stripes are missing (410)
924 - LRU counters were broken, causing constant lock purge (433, 432)
925 - garbage on read from stripes with failed OSTs (441)
926 - mark OSCs as active before reconnecting during recovery (438)
927 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
928 - lfind did not preserve OST order in output (443)
929 - symlinks cause hung clients, incorrect data (439)
930 - stop dereferencing request after dropping refcount (457)
931 - don't LASSERT(spin_is_locked) on non-SMP (455)
932 - fixes for many rename() bugs
933 - fstat didn't correctly synchronize attributes (399)
934 - server must handle lock cancellation during blocking AST prep (487)
935 - bulk descriptors were free()d too soon (511)
936 - fix paths in lconf, which would load incorrect modules (451, 507)
937 - fix confusing lconf 'host not found' error message (386)
938 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
939 - fix race condition in mdc_blocking_ast() for inode access (526)
940 - fix lov_unpackmd() unpacking wrong number of stripes (537)
941 - fix lov_set_osc_active() marking wrong OSC inactive (440)
942 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
943 - fix multiple writes of stripe MD to MDS (358, maybe 519)
944 - fix lstripe in several ways (kernel side) (527)
945 - fix request leak in ldlm_cli_enqueue (262)
946 - incorrect OSC was marked inactive after OST failure
947 - call mds_fs_cleanup before unmounting filesystem (524)
948 - fix races between taking ns_lock and ldlm_lock_change_resource
949 - fix races updating LOV export open file list
950 - fix lov_enqueue error path, avoid decref-ing bad lock handle
951 - fix recovery NULL deref in ldlm_cli_cancel_unused
952 - fix some DLM races by using new hash table for lock handles (419)
953 - permit the client to specify desired inodes, at replay
954 - duplicate requests when we queue them for replay reintegration
955 - fix last_rcvd offset calculation
956 - sync after each recovered transaction, so we always make progress
957 - never, not always, ERESTART requests without transnos
958 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
960 - skip replay if the MDS says that the client is already connected
961 - don't check for a recovery-enabled export to match lctl's UUID
962 - don't INC_USE_COUNT for phantom exports
963 - don't crash when cleaning up phantom exports (567)
964 - don't double-finish or set replay data for errored mdc_open requests
965 - abort requests when they time out, so we don't get old replies
966 - send/receive replies for AST messages again
967 - if the client says that it doesn't have the lock, cancel it on the
969 - if we timeout during I/O, don't try to cancel an in-use lock; instead
970 mark it as destroyed, it will all work out when decref is called
971 - fix module use counts (22, 581)
973 - ASTs now expect a reply (server cancels lock on error reply)
975 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
978 - fix many simultaneous client startup (392)
979 - fix dentry->d_it clobbering
980 - credentials weren't being shipped for readdir/getattr operations
981 - remove invalid assertions triggered during some concurrent MD
983 - proper Lustre versions added (336, 389)
984 - fix memory leak for create error case (398)
985 - fix LOV locking bug that would get cli/srv out of sync
986 - fix echo client over LOV (409)
987 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
988 - fix double-O_EXCL intent crash (424)
989 - avoid sending multiple lock CANCELs (352)
991 - MDS can do multi-client recovery (modulo bugs in new code)
993 - many updates, edits, cleanups
995 2002-11-18 Phil Schwan <phil@clusterfs.com>
998 - fix null d_it dereference (346)
999 - fix full OST/dbench hang (333)
1000 - fix permission problem with file removal (286)
1001 - fix removal of OSCs from LOV when they fail
1002 - fix NULL deref during bulk timeout (214)
1003 - fix problems related to multiple filesystems on one MDS (241)
1004 - fixed serious subtle metadata locking bugs
1005 - free locks on clients when inodes are removed due to memory
1007 - fix inode pointer in lock data (285)
1008 - partial support for multiple MDS on a single host (241)
1009 - data locks weren't cancelled at clear_inode time (290, 311)
1010 - intent locks could lead to unbounded lock growth (205)
1011 - added a maximum lock count, an LRU list, and a flusher
1012 - fix multiple rename (365)
1013 - properly abstracted the echo client
1014 - OSC locked 1 byte too many; fixed
1015 - rewrote brw callback code:
1016 - fixed recovery bugs related to LOVs (306)
1017 - fixed too-many-pages-in-one-write crash (191)
1018 - fixed (again) crash in sync_io_timeout (214)
1019 - probably fixed callback-related race (385)
1021 - Add capability to MDS protocol
1022 - LDLM cancellations and callbacks on different portals
1024 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
1027 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
1028 - defer open object destruction to close time (601981, 138)
1029 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
1030 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
1031 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
1032 - don't call obd_setattr in ll_file_release for destroyed objects
1033 * protocol change to lustre_msg: move |version| and add |flags|
1034 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
1035 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
1036 * added replay of create, unlink, link and rename operations during
1037 MDS failover; recovery should be much more robust now
1038 * remove failed OSCs from LOVs (only lov_create uses this so far)
1039 * the lustre-HOWTO was brought (more) up to date (582544)
1041 2002-10-23 Phil Schwan <phil@clusterfs.com>
1044 - in-use dentries weren't being reused properly (617851)
1045 - prevent multiple LDLM setup (599178)
1046 - fix LOV size calculations for truncate (617853)
1047 - fix client handling of MDS intent errors (POSIX)
1048 - fix permission bug in lovstripe.c test (624321)
1049 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
1050 - truncate past end of file could corrupt data
1051 - proper cleanup after timeouts, crashes, etc (592524, 550815)
1052 - a race in recovery could return ETIMEDOUT to apps (623947)
1053 - building outside the source directory was fixed
1054 * the lustre-HOWTO was brought (more) up to date (582544)
1055 * major progress was made on recovery functionality
1057 2002-10-10 Phil Schwan <phil@clusterfs.com>
1060 - recovery deadlock fix
1061 - rm -rf causes LBUG fix (617817)
1062 - file open by multiple tasks fix (618962)
1063 - directory permissions bugs (602707 and 620007)
1064 - journal_stop fixed with locking (611313)
1065 - O_APPEND failures resolved (618273, perhaps 614459)
1066 - lconf PATH fix (619770)
1067 - IA64 build fix (621450)
1068 - RPC buffer sizes scale with amount of memory
1070 2002-10-01 Phil Schwan <phil@clusterfs.com>
1073 - locks would be cancelled without throwing away data pages,
1074 resulting in inconsistent data (605627)
1075 - inode attributes were not always being refreshed (605627, 612449)
1076 - lconf now continues to cleanup after lctl reports an error
1077 - MDS now enforces user permissions (602707)
1078 - lprocfs cleanup fixed, but not yet enabled (614157)
1079 - fixed infinite server hang, should a client not respond to an AST
1080 - avoid going into recovery if user calls readlink() with a buffer
1081 that's too small (613941)
1082 - AST RPCs no longer require replies (614867) -- this may be changed
1083 - don't crash server if client sends an IOV that's too big (611336)
1084 - fixed lock conversion deadlock (611892)
1085 - fixed the following of symlinks (614622)
1086 * recovery: the server can remove locks from a client that dies, other
1087 clients can make progress
1088 * more extN patch fixes
1089 * compile-time configurable ptlrpc buffer allocations
1091 - collaborative read cache document
1092 - Lustre Lite Performance CDR document-in-progress
1094 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
1097 - fix typo in patch-2.4.18
1099 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
1102 - clear ptlrpc request each time in handle_incoming_request()
1103 - unlink of files now destroys the object on the OST
1105 2002-09-19 Peter Braam <braam@clusterfs.com>
1107 * add hard link support
1108 * change obdfile creation method
1109 * kernel patch changed
1111 2002-09-19 Peter Braam <braam@clusterfs.com>
1114 - stack overflow bug in extN fixed
1116 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
1118 * documentation updates
1119 - add man pages for config tools
1120 - update tests/README to describe testing with new config tools
1121 - finish metadata API descriptions
1122 * bug fixes and cleanups
1123 - statfs workaround for 16TB limit
1124 - LOV stripe allocation improved, can stripe on subset of OSTs
1125 - LOV file size/IO offset was wrong for files > 4GB in size
1126 - object EA data was being dropped, caused files to be unreadable
1127 - memory overflow with non-LOV OST caused memory corruption
1128 - fixed regression tests to work with new config tools, obdfilter
1129 - fixed bug when directory size became larger than 1 block
1130 - fixed bug (for single client case) when PWD was deleted
1131 - invalidate local directory pages when doing intent-based ops
1132 - avoid LDLM oops when lock callback contained bad data
1134 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
1136 * documentation updates
1137 * bug fixes and cleanups
1138 - configuration tools
1141 - 64-bit compile warnings
1142 - 64-bit internal statfs data
1144 * test_brw on persistent OST devices
1146 * lprocfs (disabled)
1148 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
1150 * documentation updates
1151 * bug fixes and cleanups
1152 * configuration tools
1154 2002-08-30 Peter J. Braam <braam@clusterfs.com>
1157 * many small fixes to 0_5_4
1158 * io/network handling
1159 * thinkos in MDS operations
1161 2002-08-24 Peter J. Braam <braam@clusterfs.com>
1164 * crucial basic fixes to 0.5.3
1165 * IOR, Iozone work over Elan
1168 2002-08-07 Phil Schwan <phil@clusterfs.com>
1169 * version 0_5_3, our first alpha
1170 * we use the new Portals iovs
1171 * documentation updates
1172 * bug fixes and cleanups
1173 * small changes in the DLM wire protocol
1175 2002-07-25 Peter J. Braam <braam@clusterfs.com>
1176 * version 0_5_1 with some initial stability,
1177 * locking on MD and file I/O.
1178 * documentation updates
1179 * several bug fixes since 0.5.0
1180 * small changes in wire protocol
1182 2002-07-18 Phil Schwan <phil@clusterfs.com>
1184 * delivered as Lustre Light Alpha
1185 * fixed a crash after handling invalid MDS requests
1186 * fixed directory pages for architectures with non-4k pages sizes
1188 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
1189 * release version v0_4_4
1190 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
1192 * Fixes a number of interruption problems with OST operations.
1193 * Update documentation for portals header changes
1194 * Move all wire protocol structs/defines to lustre_idl.h
1195 * Fixes symlink length bug.
1196 * Add tcpdump to repository.
1198 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
1199 * release version v0_4_3
1200 * Fixes statfs for inodes on extN.
1201 * Fixes bug in runtests which would delete /etc/hosts.
1202 * Use 64-bit object IDs wherever possible (not into VFS though)
1203 Remove ost_get_info, which is unused by lustre, and out of date.
1205 2002-07-03 Peter Braam <braam@clusterfs.com>
1206 * release version v0_4_2 Fixes a lookup error (type not passed)
1207 * move forward to head of Portals
1208 * move forward to latest Lustre kernel
1210 2002-06-25 Peter Braam <braam@clusterfs.com>
1211 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob