1 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
4 INTEROPERATE with older versions automatically. In many cases a
5 special upgrade step is needed. Please read the
6 user documentation before upgrading any part of a live system.
7 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
8 version will not interoperate with 1.6 betas before beta7 (1.5.97).
9 * WARNING: Lustre configuration and startup changes are required with
10 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
12 * Recommended e2fsprogs version: 1.39.cfs2-0
15 Severity : enhancement
17 Description: MountConf
18 Details : Lustre configuration is now managed via mkfs and mount
19 commands instead of lmc and lconf. New obd types (MGS, MGC)
20 are added for dynamic configuration management. See
21 https://mail.clusterfs.com/wikis/lustre/MountConf for
24 Severity : enhancement
26 Description: dynamic OST addition
27 Details : OSTs can now be added to a live filesystem
29 Severity : enhancement
31 Description: startup order invariance
32 Details : MDTs and OSTs can be started in any order. Clients only
33 require the MDT to complete startup.
35 Severity : enhancement
37 Description: parallel, asynchronous orphan cleanup
38 Details : orphan cleanup is now performed in separate threads for each
39 OST, allowing parallel non-blocking operation.
41 Severity : enhancement
43 Description: optimized stripe assignment
44 Details : stripe assignments are now made based on ost space available,
45 ost previous usage, and OSS previous usage, in order to try
46 to optimize storage space and networking resources.
48 Severity : enhancement
50 Description: Permanently set tunables
51 Details : All writable /proc/fs/lustre tunables can now be permanently
52 set on a per-server basis, at mkfs time or on a live
55 Severity : enhancement
57 Description: Lustre message v2
58 Details : Add lustre message format v2.
60 Severity : enhancement
62 Description: client OST exclusion list
63 Details : Clients can be started with a list of OSTs that should be
64 declared "inactive" for known non-responsive OSTs.
67 Frequency : SFS test only (otherwise harmless)
69 Description: SPEC SFS validation failure on NFS v2 over lustre.
70 Details : Changes the blocksize for regular files to be 2x RPC size,
71 and not depend on stripe size.
73 Severity : enhancement
75 Description: Multiple MD RPCs in flight.
76 Details : Further unserialise some read-only MDS RPCs - learn about intents.
77 To avoid overly-overloading MDS, introduce a limit on number of
78 MDS RPCs in flight for a single client and add /proc controls
81 Severity : enhancement
83 Description: client read/write statistics
84 Details : Add client read/write call usage stats for performance
85 analysis of user processes.
86 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
87 file access. extents_stats shows chunk size distribution.
88 extents_stats_per_process show chunk size distribution per
91 Severity : enhancement
93 Description: per-client statistics on server
94 Details : Add ldlm and operations statistics for each client in
95 /proc/fs/lustre/mds|obdfilter/*/exports/
97 Severity : enhancement
99 Description: mds statistics
100 Details : Add detailed mds operations statistics in
101 /proc/fs/lustre/mds/*/stats
103 Severity : enhancement
105 Description: VFS operations stats
106 Details : Add client VFS call stats, trackable by pid, ppid, or gid
107 /proc/fs/lustre/llite/*/vfs_ops_stats
108 /proc/fs/lustre/llite/*/track_[pid|ppid|gid]
113 Description: Fix client-side osc byte counters
114 Details : The osc read/write byte counters in
115 /proc/fs/lustre/osc/*/stats are now working
118 Frequency : always as root on SLES
120 Description: Failure of copying files with lustre special EAs.
121 Details : Client side always return success for setxattr call for lustre
122 special xattr (currently only "trusted.lov").
127 Description: Refcount LNET uuids
128 Details : The global LNET uuid list grew linearly with every startup;
129 refcount repeated list entries instead of always adding to
132 Severity : enhancement
134 Description: Dynamic service threads
135 Details : Within a small range, start extra service threads
136 automatically when the request queue builds up.
139 Frequency : mixed-endian client/server environments
141 Description: mixed-endian crashes
142 Details : The new msg_v2 system had some failures in mixed-endian
145 Severity : enhancement
147 Description: Easy OST removal
148 Details : OSTs can be permanently deactivated with e.g. 'lctl
149 conf_param lustre-OST0001.osc.active=0'
151 Severity : enhancement
153 Description: MGS proc entries
154 Details : Added basic proc entries for the MGS showing what filesystems
157 Severity : enhancement
159 Description: provide MGS failover
160 Details : Added config lock reacquisition after MGS server failover.
162 Severity : enhancement
164 Description: add Linux 2.4 support
165 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
169 Description: a large application tries to do I/O to the same resource and dies
171 Details : Check the req->rq_arrival time after the call to
172 ost_brw_lock_get(), but before we do anything about
173 processing it & sending the BULK transfer request. This
174 should help move old stale pending locks off the queue as
175 quickly as obd_timeout.
178 Frequency : when an incorrect nid is specified during startup
180 Description: ptlrpc connect to non-existant node causes kernel crash
181 Details : LNET can't be re-entered from an event callback, which
182 happened when we expire a message after the export has been
183 cleaned up. Instead, hand the zombie cleanup off to another
188 Description: make O_SYNC working on 2.6 kernels
189 Details : 2.6 kernels use different method for mark pages for write,
190 so need add a code to lustre for O_SYNC work.
195 Description: Failure to close file and release space on NFS
196 Details : Put inode details into lock acquired in ll_intent_file_open.
197 Use mdc_intent_lock in ll_intent_open to properly
198 detect all kind of errors unhandled by mdc_enqueue
200 Severity : enhancement
202 Description: plain/inodebits lock performance improvement
203 Details : Grouping plain/inodebits in granted list by their request modes
204 and bits policy, thus improving the performance of search through
210 Description: proc file read during shutdown sometimes raced obd removal,
212 Details : Add lock to prevent obd access after proc file removal
216 Description: improperly doing page alignment of locks
217 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
218 PAGE_*. Make CFS_PAGE_MASK 64bit long.
220 ------------------------------------------------------------------------------
222 TBD Cluster File Systems, Inc. <info@clusterfs.com>
224 * Support for kernels:
225 2.6.9-42.0.3EL (RHEL 4)
227 2.4.21-40.0.1.EL (RHEL 3)
228 2.6.12.6 vanilla (kernel.org)
234 Description: "went back in time" messages on mds failover
235 Details : The greatest transno may be lost when the current operation
236 finishes with an error (transno==0) and the client's last_rcvd
237 record is over-written. Save the greatest transno in the
238 mds_last_transno for this case.
241 Frequency : always for specific kernels and striping counts
243 Description: client may get "Matching packet too big" without ACL support
244 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
245 when trying to access files in certain configurations. The
246 clients should in fact be denied when mounting because they do
250 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
252 Description: many clients connecting with IO in progress causes connect timeouts
253 Details : Avoid synchronous journal commits to avoid delays caused by many
254 clients connecting/disconnecting when bulk IO is in progress.
255 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
256 OST_IO_PORTAL to avoid delays behind potentially many pending
260 Frequency : occasionally with multiple writers to a single file
262 Description: shared writes to file may result in wrong size reported by stat()
263 Details : Allow growing of kms when extent lock is cancelled
266 Frequency : always with random mmap IO to multi-striped file
268 Description: mmap write might be lost if we are writing to a 'hole' in stripe
269 Details : Only if the hole is at the end of OST object so that kms is too
270 small. Fix is to increase kms accordingly in ll_nopage.
273 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
275 Description: writes to a missing object would leak memory on the OST
276 Details : If there is an inconsistency between the MDS and OST filesystems,
277 such that the MDS references an object that doesn't exist, writes
278 to that object will leak memory due to incorrect cleanup in the
279 error handling path, eventually running out of memory on the OST.
284 Description: Creating too long symlink causes lustre errors
285 Details : Check symlink and name lengths before sending requests to MDS.
288 Frequency : only if flock is enabled (not on by default)
290 Description: posix locks not released on fd closure on 2.6.9+
291 Details : We failed to add posix locks to list of inode locks on 2.6.9+
292 kernels, this caused such locks not to be released on fd close and
293 then assertions on fs unmount about still used locks.
296 Frequency : MDS failover only, very rarely
298 Description: clients may get ASSERTION(granted_lock != NULL)
299 Details : When request was taking a long time, and a client was resending
300 a getattr by name lock request. The were multiple lock
301 requests with the same client lock handle and
302 mds_getattr_name->fixup_handle_for_resent_request found one
303 of the lock handles but later failed with
304 ASSERTION(granted_lock != NULL).
309 Description: handle->h_buffer_credits > 0, assertion failure
310 Details : h_buffer_credits is zero after truncate, causing assertion
311 failure. This patch extends the transaction or creates a new
315 Frequency : NFS re-export or patchless client
316 Bugzilla : 11179, 10796
317 Description: Crash on NFS re-export node (__d_move)
318 Details : We do not want to hash the dentry if we don't have a lock.
319 But if this dentry is later used in d_move, we'd hit uninitialised
320 list head d_hash, so we just do this to init d_hash field but
321 leave dentry unhashed.
324 Frequency : NFS re-export or patchless client
326 Description: NFS exports has problem with symbolic link
327 Details : lustre client didn't properly install dentry when re-exported
328 to NFS or running patchless client.
331 Frequency : NFS re-export or patchless client
333 Description: Various nfs/patchless fixes.
334 Details : fixes reuse disconected alias for lookup process - this fixes
335 warning "find_exported_dentry: npd != pd", fix permission
336 error with open files at nfs.
339 Frequency : occasional
341 Description: Crash on NFS re-export node
342 Details : call clear_page on wrong pointer triggered oops in
343 generic_mapping_read().
346 Frequency : rarely, using O_DIRECT IO
348 Description: unaligned directio crashes client with LASSERT
349 Details : check for unaligned buffers before trying any requests.
352 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
354 Description: stale data returned from RAID cache
355 Details : If only a small amount of IO is done to the RAID device before
356 reading it again it is possible to get stale data from the RAID
357 cache instead of reading it from disk.
360 Frequency : depends on arch, kernel and compiler version, always on sles10
363 Description: recursive or deep enough symlinks cause stack overflow
364 Details : getting rid of large stack-allocated variable in
368 Frequency : depends on hardware
370 Description: lustre write performance loss in the SLES10 kernel
371 Details : the performance loss is caused by using of write barriers in the
372 ext3 code. The SLES10 kernel turns barrier support on by
373 default. The fix is to undo that change for ldiskfs.
375 ------------------------------------------------------------------------------
377 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
379 * Support for kernels:
380 2.6.9-42.0.3EL (RHEL 4)
382 2.4.21-47.0.1.EL (RHEL 3)
383 2.6.12.6 vanilla (kernel.org)
387 Frequency : quota enabled and large files being deleted
389 Description: releasing more than 4GB of quota at once hangs OST
390 Details : If a user deletes more than 4GB of files on a single OST it
391 will cause the OST to spin in an infinite loop. Release
392 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
397 Description: statfs data retrieved from /proc may be stale or zero
398 Details : When reading per-device statfs data from /proc, in the
399 {kbytes,files}_{total,free,avail} files, it may appear
400 as zero or be out of date.
403 Frequency : always, for aggregate stripe size over 4GB
405 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
406 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
407 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
410 Frequency : always if "lfs find" used on a local file/directory
412 Description: "lfs find" segfaults if used on a local file/directory
413 Details : The case where a directory component was not specified wasn't
414 handled correctly. Handle this properly.
417 Frequency : always on ppc64
419 Description: the write to an ext3 filesystem mounted with mballoc got stuck
420 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
421 perform endianness conversion.
424 Frequency : rarely (truncate to non-zero file size after write under load)
425 Bugzilla : 10730, 10687
426 Description: Files padded with zeros to next 4K multiple
427 Details : With filesystems mounted using the "extents" option (2.6 kernels)
428 it is possible that files that are truncated to a non-zero size
429 immediately after being written are filled with zero bytes beyond
430 the truncated size. No file data is lost.
432 Severity : enhancement
433 Frequency : liblustre only
435 Description: Allow recovery/failover for liblustre clients.
436 Details : liblustre clients were unaware of failover configurations until
439 Severity : enhancement
441 Description: user file locks should fail when not mounting with flock option
442 Details : Set up an error-returning stub in ll_file_operations.lock field
443 to prevent incorrect behaviour when client is mounted without
444 flock option. Also, set up properly f_op->flock field for
448 Frequency : always on ia64
450 Description: "lfs df" loops on printing out MDS statfs information
451 Details : The obd_ioctl_data was not initialized and in some systems
452 this caused a failure during the ioctl that did not return
453 an error. Initialize the struct and return an error on failure.
456 Frequency : SLES 9 only
458 Description: Error of copying files with lustre special EAs as root
459 Details : Client side always return success for setxattr call for lustre
460 special xattr (currently only "trusted.lov").
463 Frequency : rarely on clusters with both ia64+i386 clients
465 Description: ia64+i686 clients doing shared IO on the same file may LBUG
466 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
467 clients are doing concurrent writes to the same file it is
468 possible that the ia64 clients may LASSERT because the OST
469 extent locks are not PAGE_SIZE aligned. Ensure that grown
470 locks are always aligned on the request boundary.
473 Frequency : specific use, occasional
475 Description: Overwriting in use executable truncates on-disk binary image
476 Details : If one node attempts to overwrite an executable in use by
477 another node, we now correctly return ETXTBSY instead of
483 Description: chmod on Lustre root is propagated to other clients
484 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
485 invalid by the follow_mount time.
490 Description: Race in 'instant cancel' lock handling could lead to such locks
491 never to be granted in case of SMP MDS
492 Details : Do not destroy not yet granted but cbpending locks in
496 Frequency : replay/resend of open
498 Description: non null lock assetion failure in mds_intent_policy
499 Details : Trying to replay/resend lockless open requests resulted in
500 mds_open() returning 0 with no lock. Now it sets a flag if
501 a lock is going to be returned.
503 Severity : enhancement
505 Description: Checksum enhancements
506 Details : New checksum enhancements allow for resending RPCs that failed
509 Severity : enhancement
511 Description: Tunables on number of dirty pages in cacche
512 Details : Allow to set limit on number of dirty pages cached.
517 Description: client crash on unmount - lock still has references
518 Details : In some error handling cases it was possible to leak a lock
519 reference on a client while accessing a file. This was not
520 harmful to the client during operation, but would cause the
521 client to crash when the filesystem is unmounted.
524 Frequency : specific case, rare
526 Description: ETXTBSY on mds though file not in use by client
527 Details : ETXTBSY is no longer incorrectly returned when attempting to
528 chmod or chown a directory that the user previously tried to
529 execute or a currently-executing binary.
532 Frequency : extremely rare except on liblustre-based clients
534 Description: Lustre space not freed when files are deleted
535 Details : Clean up open-unlinked files after client eviction. Previously
536 the unlink was skipped and the files remained as orphans.
541 Description: OST failure "would be an LBUG" in waiting_locks_callback()
542 Details : In some cases it was possible to send a blocking callback to
543 a client doing a glimpse, even though that client didn't get
544 a lock granted. When the glimpse lock is cancelled on the OST
545 the freed lock is left on the waiting list and corrupted the list.
548 Frequency : all core dumps
550 Description: Broke core dumps to lustre
551 Details : Negative dentry may be unhashed if parent does not have UPDATE
552 lock, but some callers, e.g. do_coredump, expect dentry to be
553 hashed after successful create, hash it in ll_create_it.
555 ------------------------------------------------------------------------------
557 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
559 * Support for kernels:
560 2.6.9-42.0.2.EL (RHEL 4)
562 2.4.21-40.EL (RHEL 3)
563 2.6.12.6 vanilla (kernel.org)
567 Frequency : always on RHEL 3
569 Description: Number of open files grows over time
570 Details : The number of open files grows over time, whether or not
571 Lustre is started. This was due to a filp leak introduced
572 by one of our kernel patches.
574 ------------------------------------------------------------------------------
576 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
578 * Support for kernels:
581 2.4.21-40.EL (RHEL 3)
582 2.6.12.6 vanilla (kernel.org)
587 Bugzilla : 5719, 9635, 9792, 9684
588 Description: OST (or MDS) trips assertions in (re)connection under heavy load
589 Details : If a server is under heavy load and cannot reply to new
590 connection requests before the client resends the (re)connect,
591 the connection handling code can behave badly if two service
592 threads are concurrently handing separate (re)connections from
593 the same client. Add better locking to the connection handling
594 code, and ensure that only a single connection will be processed
595 for a given client UUID, even if the lock is dropped.
597 Severity : enhancement
599 Description: add TCP zero-copy support to kernel
600 Details : Add support to the kernel TCP stack to allow zero-copy bulk
601 sends if the hardware supports scatter-gather and checksumming.
602 This allows socklnd to do client-write and server-read more
603 efficiently and reduce CPU utilization from skbuf copying.
606 Frequency : only if NFS exporting from client
608 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
609 Details : If setting attributes on a file created under NFS that had
610 never been opened it would be possible to oops the client
611 if the file had no objects.
615 Bugzilla : 9326, 10402, 10897
616 Description: client crash in ptlrpcd_wake() thread when sending async RPC
617 Details : It is possible that ptlrpcd_wake() dereferences a freed async
618 RPC. In rare cases the ptlrpcd thread alread processed the RPC
619 before ptlrpcd_wake() was called and the request was freed.
622 Frequency : always for liblustre
624 Description: liblustre client does MDS+OSTs setattr RPC for each write
625 Details : When doing a write from a liblustre client, the client
626 incorrectly issued an RPC to the MDS and each OST the file was
627 striped over in order to update the timestamps. When writing
628 with small chunks and many clients this could overwhelm the MDS
629 with RPCs. In all cases it would slow down the write because
630 these RPCs are unnecessary.
632 Severity : enhancement
634 Description: allow number of MDS service threads to be changed at module load
635 Details : It is now possible to change the number of MDS service threads
636 running. Adding "options mds mds_num_threads={N}" to the MDS's
637 /etc/modprobe.conf will set the number of threads for the next
638 time Lustre is restarted (assuming the "mds" module is also
639 reloaded at that time). The default number of threads will
640 stay the same, 32 for most systems.
645 Description: OST crash if filesystem is unformatted or corrupt
646 Details : If an OST is started on a device that has never been formatted
647 or if the filesystem is corrupt and cannot even mount then the
648 error handling cleanup routines would dereference a NULL pointer.
653 Description: NULL pointer deref in llap_from_page.
654 Details : get_cache_page_nowait can return a page with NULL (or otherwise
655 incorrect) mapping if the page was truncated/reclaimed while it was
656 searched for. Check for this condition and skip such pages when
657 doing readahead. Introduce extra check to llap_from_page() to
658 verify page->mapping->host is non-NULL (so page is not anonymous).
661 Frequency : Sometimes when using sys_sendfile
663 Description: "page not covered by a lock" warnings from ll_readpage
664 Details : sendfile called ll_readpage without right page locks present.
665 Now we introduced ll_file_sendfile that does necessary locking
666 around call to generic_file_sendfile() much like we do in
670 Frequency : with certain MDS communication failures at client mount time
672 Description: NULL pointer deref after failed client mount
673 Details : a client connection request may delayed by the network layer
674 and not be sent until after the PTLRPC layer has timed out the
675 request. If the client fails the mount immediately it will try
676 to clean up before the network times out the request. Add a
677 reference from the request import to the obd device and delay
678 the cleanup until the network drops the request.
681 Frequency : occasionally during client (re)connect
683 Description: assertion failure during client (re)connect
684 Details : processing a client connection request may be delayed by the
685 client or server longer than the client connect timeout. This
686 causes the client to resend the connection request. If the
687 original connection request is replied in this interval, the
688 client may trip an assertion failure in ptlrpc_connect_interpret()
689 which thought it would be the only running connect process.
692 Frequency : only with obd_echo servers and clients that are rebooted
694 Description: kernel BUG accessing uninitialized data structure
695 Details : When running an obd_echo server it did not start the ping_evictor
696 thread, and when a client was evicted an uninitialized data
697 structure was accessed. Start the ping_evictor in the RPC
698 service startup instead of the OBD startup.
700 Severity : enhancement
701 Bugzilla : 10193 (patchless)
702 Description: Remove dependency on various unexported kernel interfaces.
703 Details : No longer need reparent_to_init, exit_mm, exit_files,
704 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
707 Frequency : rare (only users of deprecated and unsupported LDAP config)
709 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
710 Details : LDAP apparently contains 'lustreName' attributes instead of
711 'name'. A simple remapping of the name is sufficient.
714 Frequency : rare (only with non-default dump_on_timeout debug enabled)
716 Description: waiting_locks_callback trips kernel BUG if client is evicted
717 Details : Running with the dump_on_timeout debug flag turned on makes
718 it possible that the waiting_locks_callback() can try to dump
719 the Lustre kernel debug logs from an interrupt handler. Defer
720 this log dumping to the expired_lock_main() thread.
722 Severity : enhancement
724 Description: Support NFS exporting on 2.6 kernels.
725 Details : Implement non-rawops metadata methods for NFS server to use without
726 changing NFS server code.
729 Frequency : very rare (synthetic metadata workload only)
731 Description: two racing renames might cause an MDS thread to deadlock
732 Details : Running the "racer" program may cause one MDS thread to rename
733 a file from being the source of a rename to being the target of
734 a rename at exactly the same time that another thread is doing
735 so, and the second thread has already enqueued these locks after
736 doing a lookup of the target and is trying to relock them in
737 order. Ensure that we don't try to re-lock the same resource.
740 Frequency : only very large systems with liblustre clients
742 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
743 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
744 all OSTs in parallel. This allows the eviction of stale liblustre
745 clients to proceed much faster than if they were done in series,
746 and also offers similar improvements for other set_info RPCs.
751 Description: excessive CPU usage during initial read phase on client
752 Details : During the initial read phase on a client, it would agressively
753 retry readahead on the file, consuming too much CPU and impacting
754 performance (since 1.4.5.8). Improve the readahead algorithm
755 to avoid this, and also improve some other common cases (read
756 of small files in particular, where "small" is files smaller than
757 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
762 Description: MDS crash when receiving packet with unknown intent.
763 Details : Do not LBUG in unknown intent case, just return -EFAULT
765 Severity : enhancement
766 Bugzilla : 9293, 9385
767 Description: MDS RPCs are serialised on client. This is unnecessary for some.
768 Details : Do not serialize getattr (non-intent version) and statfs.
771 Frequency : occasional, when OST network is overloaded/intermittent
773 Description: client evicted by OST after bulk IO timeout
774 Details : If a client sends a bulk IO request (read or write) the OST
775 may evict the client if it is unresposive to its data GET/PUT
776 request. This is incorrect if the network is overloaded (takes
777 too long to transfer the RPC data) or dropped the OST GET/PUT
778 request. There is no need to evict the client at all, since
779 the pinger and/or lock callbacks will handle this, and the
780 client can restart the bulk request.
783 Frequency : Always when mmapping file with no objects
785 Description: client crashes when mmapping file with no objects
786 Details : Check that we actually have objects in a file before doing any
787 operations on objects in ll_vm_open, ll_vm_close and
793 Description: Request leak when working with deleted CWD
794 Details : Introduce advanced request refcount tracking for requests
795 referenced from lustre intent.
797 Severity : Enhancement
799 Description: Cache open file handles on client.
800 Details : MDS now will return special lock along with openhandle, if
801 requested and client is allowed to hold openhandle, even if unused,
802 until such a lock is revoked. Helps NFS a lot, since NFS is opening
803 closing files for every read/write openration.
805 Severity : Enhancement
807 Description: Cache open negative dentries on client when possible.
808 Details : Guard negative dentries with UPDATE lock on parent dir, drop
809 negative dentries on lock revocation.
814 Description: Remounting a client read-only wasn't possible with a zconf mount
815 Details : It wasn't possible to remount a client read-only with llmount.
817 Severity : enhancement
818 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
819 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
820 MPICH is included to add Lustre-specific ADIO interfaces.
821 This is based closely on the UFS ADIO layer and only differs
822 in file creation, in order to allow the OST striping to be set.
823 This is user-contributed code and not supported by CFS.
828 Description: extended inode attributes (immutable, append-only) work improperly
829 when 2.4 and 2.6 kernels are used on client/server or vice versa
830 Details : Introduce kernel-independent values for these flags.
832 Severity : enhancement
835 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
836 Details : Many of the /proc/ tunables can only be tuned at a megabyte
837 granularity. Now, Fractional MB granularity is be supported,
838 this is very useful for low memory system.
840 Severity : enhancement
842 Description: Getattr by fid
843 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
844 locks, avoids extra getattr rpc requests to MDS, allows '/' to
845 have locks and avoids getattr rpc requests for it on every stat.
848 Frequency : Always, for filesystems larger than 2TB
850 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
851 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
852 filesystems larger than 2TB to be mounted with mballoc enabled.
855 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
857 Description: filesystem corruption for non-standard kernels and very large OSTs
858 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
859 filesystems larger than 2TB is mounted then the kernel will
860 silently corrupt the start of the filesystem. CONFIG_LBD is
861 enabled for all CFS-supported kernels, but the possibility of
862 this happening with a modified kernel config exists.
864 Severity : enhancement
866 Description: add client O_DIRECT support for 2.6 kernels
867 Details : It is now possible to do O_DIRECT reads and writes to files
868 in the Lustre client mountpoint on 2.6 kernel clients.
870 Severity : enhancement
872 Description: parallel glimpse, setattr, statfs, punch, destroy requests
873 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
874 parallel, not waiting for response from every OST before sending
875 a rpc to the next OST.
880 Description: setattr vs write race when updating file timestamps
881 Details : Client processes that update a file timestamp into the past
882 right after writing to the file (e.g. tar) it is possible that
883 the updated file modification time can be reset to the current
884 time due to a race between processing the setattr and write RPC.
886 Severity : enhancement
888 Description: Bring 'lfs find' closer in line with regular Linux find.
889 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
890 -print0 options and obtains all the needed info through the lustre
893 Severity : enhancement
895 Description: support up to 1024 configured devices on one node
896 Details : change obd_dev array from statically allocated to dynamically
897 allocated structs as they are first used to reduce memory usage
902 Description: Flush dirty partially truncated pages during truncate
903 Details : Immediatelly flush partially truncated pages in filter_setattr,
904 this way we completely avoid having any pages in page cache on OST
905 and can retire ugly workarounds during writes to flush such pages.
910 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
911 Details : protect inode from truncation within vfs_unlink() context
912 just take a reference before calling vfs_unlink() and release it
913 when parent's i_sem is free.
916 Frequency : always, if extents are used on OSTs
918 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
919 Details : OSTs using the extents format would not zero the high 16 bits of
920 the index physical block number. This is not a problem for any
921 OST filesystems smaller than 16TB, and no kernels support ext3
922 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
923 new/modified files) and can be fixed for existing filesystems
924 with e2fsprogs-1.39-cfs1.
929 Description: import connection selection may be incorrect if timer wraps
930 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
931 import connections to be ignored if the 32-bit jiffies counter
932 wraps. Use a 64-bit jiffies counter.
935 Frequency : during server recovery
937 Description: crash after server is denying duplicate export
938 Details : If clients are resending connect requests to the server, the
939 server refuses to allow a client to connect multiple times.
940 Fixed a bug in the handling of this case.
943 Frequency : very large clusters immediately after boot
945 Description: LNET request buffers exhausted under heavy short-term load
946 Details : If a large number of client requests are generated on a service
947 that has previously never seen so many requests it is possible
948 that the request buffer growth cannot keep up with the spike in
949 demand. Instead of dropping incoming requests, they are held in
950 the LND until the RPC service can accept more requests.
953 Frequency : Sometimes during replay
955 Description: Assertion failure in ll_local_open after replay.
956 Details : If replay happened on an open request reply before we were able
957 to set replay handler, reply will become not swabbed tripping the
958 assertion in ll_local_open. Now we set the handler right after
959 recognising of open request
962 Frequency : very rare
964 Description: kernel reports "badness in vsnprintf"
965 Details : Reading from the "recovery_status" /proc file in small chunks
966 may cause a negative length in lprocfs_obd_rd_recovery_status()
967 call to vsnprintf() (which is otherwise harmless). Exit early
968 if there is no more space in the output buffer.
970 Severity : enhancement
972 Description: clear OBD RPC statistics by writing to them
973 Details : It is now possible to clear the OBD RPC statistics by writing
979 Description: Client mtime is not the same on different clients after utimes
980 Details : In some cases, the client was using the utimes() syscall on
981 a file cached on another node. The clients now validate the
982 ctime from the MDS + OSTs to determine which one is right.
987 Description: Inability to activate failout mode
988 Details : lconf script incorrectly assumed that in pythong string's numeric
989 value is used in comparisons.
992 Frequency : always with multiple stripes per file
994 Description: Inefficient object allocation for mutli-stripe files
995 Details : When selecting which OSTs to stripe files over, for files with
996 a stripe count that divides evenly into the number of OSTs,
997 the MDS is always picking the same starting OST for each file.
998 Return the OST selection heuristic to the original design.
1003 Description: mount failures may take full timeout to return an error
1004 Details : Under some heavy load conditions it is possible that a
1005 failed mount can wait for the full obd_timeout interval,
1006 possibly several minutes, before reporting an error.
1007 Instead return an error as soon as the status is known.
1009 ------------------------------------------------------------------------------
1011 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
1013 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1014 INTEROPERATE with older versions automatically. Please read the
1015 user documentation before upgrading any part of a live system.
1016 * WARNING: Lustre networking configuration changes are required with
1017 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1020 * Support for kernels:
1021 2.6.9-22.0.2.EL (RHEL 4)
1022 2.6.5-7.244 (SLES 9)
1023 2.6.12.6 vanilla (kernel.org)
1026 Severity : enhancement
1027 Bugzilla : 7981/8208
1028 Description: Introduced Lustre Networking (LNET)
1029 Details : LNET is new networking infrastructure for Lustre, it includes
1030 a reorganized network configuration mode (see the user
1031 documentation for full details) as well as support for routing
1032 between different network fabrics. Lustre Networking Devices
1033 (LNDs) for the supported network fabrics have also been
1034 created for this new infrastructure.
1036 Severity : enhancement
1037 Description: Introduced Access control lists
1038 Details : clients can set ACLs on files and directories in order to have
1039 more fine-grained permissions than the standard Unix UGO+RWX.
1040 The MDS must be started with the "-o acl" mount option.
1042 Severity : enhancement
1043 Description: Introduced filesystem quotas
1044 Details : Administrators may now establish per-user quotas on the
1047 Severity : enhancement
1049 Description: Configuration change for the XT3
1050 The PTLLND is now used to run Lustre over Portals on the XT3
1051 The configure option(s) --with-cray-portals are no longer used.
1052 Rather --with-portals=<path-to-portals-includes> is used to
1053 enable building on the XT3. In addition to enable XT3 specific
1054 features the option --enable-cray-xt3 must be used.
1059 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1060 Details : A race between allocating a new llcd and re-getting the llcd_lock
1061 allowed another thread to grab newly-allocated llcd.
1063 Severity : enhancement
1065 Description: 2.6 OST async journal commit and locking fix to improve performance
1066 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1067 2.6 kernels are now async as they already were in 2.4 kernels so
1068 that they can commit concurrently with the network bulk transfer.
1069 For block-allocated files the filter allocation semaphore is held
1070 to avoid filesystem fragmentation during allocation. BKL lock
1071 removed for 2.6 xattr operations where it is no longer needed.
1076 Description: lconf incorrectly determined whether two IP networks could talk
1077 Details : In some more complicated routing and multiple-network
1078 configurations, lconf will avoid trying to make a network
1079 connection to a disjoint part of the IP space. It was doing the
1080 math incorrectly for one set of cases.
1085 Description: Fix for potential infinite loop processing records in an llog.
1086 Details : If an llog record is corrupted/zeroed, it is possible to loop
1087 forever in llog_process(). Validate the llog record length
1088 and skip the remainder of the block on error.
1091 Frequency : occasional (liblustre only)
1093 Description: liblustre could not open files whose last component is a symlink
1094 Details : sysio_path_walk() would incorrectly pass the open intent to
1095 intermediate path components.
1098 Frequency : rare (liblustre only with non-standard tuning)
1099 Bugzilla : 7201 (7350)
1100 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1101 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1102 after it has been released, only do this for referenced locks
1104 Severity : enhancement
1106 Description: specifying an (invalid) directory default stripe_size of -1
1107 would reset the directory default striping
1108 Details : stripe_size -1 was used internally to signal directory stripe
1109 removal, now use "all default" to signal dir stripe removal
1110 as a directory striping of "all default" is not useful
1113 Frequency : common for large clusters running liblustre clients
1115 Description: doing an ls when liblustre clients are running is slow
1116 Details : sending a glimpse AST to a liblustre client waits for every AST
1117 to time out, as liblustre clients will not respond. Since they
1118 cannot cache data we refresh the OST lock LVB from disk instead.
1120 Severity : enhancement
1122 Description: doing an ls at the same time as file IO can be slow
1123 Details : enqueue and other "small" requests can be blocked behind many
1124 large IO requests. Create a new OST IO portal for non-IO
1125 requests so they can be processed faster.
1128 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1130 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1131 created file with mode 000
1132 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1133 allow an ftruncate() as a non-root user to a file with mode 000.
1134 Lustre now respects this flag to disable mode checks when
1135 truncating a file owned by the user
1138 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1141 Description: Revoking locks from clients that went dead or catatonic might take
1143 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1144 cancellation of such locks instant on servers without waiting for
1145 any reply from clients. Clients drops these locks when cancel
1146 notification from server is received without replying.
1149 Frequency : liblustre-only, when liblustre client dies or becomes busy
1151 Description: Doing ls on Linux clients can take a long time with active
1153 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1154 granting such locks to it in the first place if possible. Locks
1155 are taken by proxy on the OST during the read or write and
1156 dropped immediately afterward. Add connect flags handling, do
1157 not grant locks to liblustre clients for glimpse ASTs.
1159 Severity : enhancement
1161 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1162 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1163 sequential reads and behaves badly with applications doing random
1164 reads. Improve it by reading ahead at least read region, and
1165 avoiding excessive large RPC for small reads.
1167 Severity : enhancement
1169 Description: Creating more than 1000 files for a single job may cause a load
1170 imbalance on the OSTs if there are also a large number of OSTs.
1171 Details : qos_prep_create() uses an OST index reseed value that is an
1172 even multiple of the number of available OSTs so that if the
1173 reseed happens in the middle of the object allocation it will
1174 still utilize the OSTs as uniformly as possible.
1179 Description: OST or MDS may oops in ping_evictor_main()
1180 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1181 but doesn't restart at beginning of obd_exports_timed list
1184 Severity : enhancement
1186 Description: improve by-nid export eviction on the MDS and OST
1187 Details : allow multiple exports with the same NID to be evicted at one
1188 time without re-searching the exports list.
1191 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1193 Description: MDS may oops in groups_free()
1194 Details : in rare race conditions a newly allocated group_info struct is
1195 freed again, and this can be NULL. The 2.4 compatibility code
1196 for groups_free() checked for a NULL pointer, but 2.6 did not.
1199 Frequency : common for liblustre clients doing little filesystem IO
1200 Bugzilla : 9352, 7313
1201 Description: server may evict liblustre clients accessing contended locks
1202 Details : if a client is granted a lock or receives a completion AST
1203 with a blocking AST already set it would not reply to the AST
1204 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1207 Frequency : lfs setstripe, only systems with more than 160 OSTs
1209 Description: unable to set striping with a starting offset beyond OST 160
1210 Details : llapi_create_file() incorrectly limited the starting stripe
1211 index to the maximum single-file stripe count.
1214 Frequency : LDAP users only
1216 Description: lconf did not handle in-kernel recovery with LDAP properly
1217 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1219 Severity : enhancement
1221 Description: bind OST threads to NUMA nodes to improve performance
1222 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1223 node and do their allocations there to localize memory access
1225 Severity : enhancement
1227 Description: llmount can determine client NID directly from Myrinet (GM)
1228 Details : the client NID code from gmnalnid was moved directly into
1229 llmount, removing the need to use this or specifying the
1230 client NID explicitly when mounting GM clients with zeroconf
1233 Frequency : if client is started with down MDS
1235 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1236 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1237 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1238 ptlrpc_import_delayed_req() because we didn't check if the
1239 request was interrupted, and we also didn't break out of the
1240 event loop if there was no timeout
1245 Description: data loss during non-page-aligned writes to a single file from
1246 both multiple nodes and multiple threads on one node at same time
1247 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1248 inconsistency led to false short-reads, that were cached and later
1249 used by ->prepare_write() to fill in partially written page,
1250 leading to data loss.
1253 Frequency : always, if lconf --abort_recovery used
1255 Description: lconf --abort_recovery fails with 'Operation not supported'
1256 Details : lconf was attempting to abort recovery on the MDT device and not
1259 Severity : enhancement
1261 Description: remove cleanup logs
1262 Details : replace lconf-generated cleanup logs with lustre internal
1263 cleanup routines. Eliminates the need for client-cleanup and
1266 Severity : enhancement
1268 Description: add support for EAs (user and system) on lustre filesystems
1269 Details : it is now possible to store extended attributes in the Lustre
1270 client filesystem, and with the user_xattr mount option it
1271 is possible to allow users to store EAs on their files also
1273 Severity : enhancement
1275 Description: Add possibility (config option) to show minimal available OST free
1277 Details : When compiled with --enable-mindf configure option, statfs(2)
1278 (and so, df) will return least minimal free space available from
1279 all OSTs as amount of free space on FS, instead of summary of
1280 free spaces of all OSTs.
1282 Severity : enhancement
1284 Description: do not expand extent locks acquired on OST-side
1285 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
1286 by server: they are not cached anyway.
1289 Frequency : when mmap is used/binaries executed from Lustre
1291 Description: Unmmap pages before throwing them away from read cache.
1292 Details : llap_shrink cache now attempts to unmap pages before discarding
1293 them (if unmapping failed - do not discard). SLES9 kernel has
1294 extra checks that trigger if this unmapping is not done first.
1299 Description: lconf didn't resolve symlinks before checking to see whether a
1300 given mountpoint was already in use
1303 Frequency : when migrating failover services
1304 Bugzilla : 6395, 9514
1305 Description: When migrating a subset of services from a node (e.g. failback
1306 from a failover service node) the remaining services would
1307 time out and evict clients.
1308 Details : lconf --force (implied by --failover) sets the global obd_timeout
1309 to 5 seconds in order to quickly disconnect, but this caused
1310 other RPCs to time out too quickly. Do not change the global
1311 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
1313 Severity : enhancement
1314 Frequency : if MDS is started with down OST
1315 Bugzilla : 9439,5706
1316 Description: Allow startup/shutdown of an MDS without depending on the
1317 availability of the OSTs.
1318 Details : Asynchronously call mds_lov_synchronize during MDS startup.
1319 Add appropriate locking and lov-osc refcounts for safe
1320 cleaning. Add osc abort_inflight calls in case the
1321 synchronize never started.
1324 Frequency : occasional (Cray XT3 only)
1326 Description: root not authorized to access files in CRAY_PORTALS environment
1327 Details : The client process capabilities were not honoured on the MDS in
1328 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
1329 been accessed by an authorized user then root was able to access
1330 the file on the local client also. The root user capabilities
1331 are now allowed on the MDS, as this environment has secure UID.
1334 Frequency : occasional
1336 Description: ldiskfs "too long searching" message happens too often
1337 Details : A debugging message (otherwise harmless) prints too often on
1338 the OST console. This has been reduced to only happen when
1339 there are fragmentation problems on the filesystem.
1344 Description: Division by zero in statfs when all OSCs are inactive
1345 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
1346 lov_statfs divides by value returned by lov_get_stripecnt().
1350 Bugzilla : 9489, 3273
1351 Description: First write from each client to each OST was only 4kB in size,
1352 to initialize client writeback cache, which caused sub-optimal
1353 RPCs and poor layout on disk for the first writen file.
1354 Details : Clients now request an initial cache grant at (re)connect time
1355 and so that they can start streaming writes to the cache right
1356 away and always do full-sized RPCs if there is enough data.
1357 If the OST is rebooted the client also re-establishes its grant
1358 so that client cached writes will be honoured under the grant.
1363 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
1364 Details : Now I/O RPCs go to different portal number and (presumably) fast
1365 lock requests (and glimses) and other RPCs get their own service
1366 threads pool that should be able to service those RPCs
1369 Severity : enhancement
1371 Description: Ability to exchange lustre version between client and servers and
1372 issue warnings at client side if client is too old. Also for
1373 liblustre clients there is ability to refuse connection of too old
1375 Details : New 'version' field is added to connect data structure that is
1376 filled with version info. That info is later checked by server and
1380 Frequency : rare, liblustre only.
1381 Bugzilla : 9296, 9581
1382 Description: Two simultaneous writes from liblustre at offset within same page
1383 might proceed at the same time overwriting eachother with stale
1385 Details : I/O lock withing llu_file_prwv was released too early, before data
1386 actually was hitting the wire. Extended lock-holding time until
1387 server acknowledges receiving data.
1390 Frequency : extremely rare. Never observed in practice.
1392 Description: avoid generating lustre_handle cookie of 0.
1393 Details : class_handle_hash() generates handle cookies by incrementing
1394 global counter, and can hit 0 occasionaly (this is unlikely, but
1395 not impossible, because initial value of cookie counter is
1396 selected randonly). Value of 0 is used as a sentinel meaning
1397 "unassigned handle" --- avoid it. Also coalesce two critical
1398 sections in this function into one.
1400 Severity : enhancement
1402 Description: allow liblustre clients to delegate truncate locking to OST
1403 Details : To avoid overhead of locking, liblustre client instructs OST to
1404 take extent lock in ost_punch() on client's behalf. New connection
1405 flag is added to handle backward compatibility.
1407 Severity : enhancement
1408 Bugzilla : 4928, 7341, 9758
1409 Description: allow number of OST service threads to be specified
1410 Details : a module parameter allows the number of OST service threads
1411 to be specified via "options ost ost_num_threads={N}" in the
1412 OSS's /etc/modules.conf or /etc/modprobe.conf.
1416 Bugzilla : 6146, 9635, 9895
1417 Description: servers crash with bad pointer in target_handle_connect()
1418 Details : In rare cases when a client is reconnecting it was possible that
1419 the connection request was the last reference for that export.
1420 We would temporarily drop the export reference and get a new
1421 one, but this may have been the last reference and the export
1422 was just destroyed. Get new reference before dropping old one.
1424 Severity : enhancement
1425 Frequency : if client is started with failover MDS
1427 Description: Allow multiple MDS hostnames in the mount command
1428 Details : Try to read the configuration from all specified MDS
1429 hostnames during a client mount in case the "primary"
1432 Severity : enhancement
1434 Description: Stop sending data to evicted clients as soon as possible.
1435 Details : Check if the client we are about to send or are sending data to
1436 was evicted already. (Check is done every second of waiting,
1437 for which l_wait_event interface was extended to allow checking
1438 of exit condition at specified intervals).
1441 Frequency : rare, normally only when NFS exporting is done from client
1443 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
1445 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
1448 Frequency : always, if srand() is called before liblustre initialization
1450 Description: Liblustre uses system PRNG disturbing its usage by user application
1451 Details : Introduce internal to lustre fast and high-quality PRNG for
1452 lustre usage and make liblustre and some other places in generic
1453 lustre code to use it.
1455 Severity : enhancement
1456 Bugzilla : 9477, 9557, 9870
1457 Description: Verify that the MDS configuration logs are updated when xml is
1458 Details : Check if the .xml configuration logs are newer than the config
1459 logs stored on the MDS and report an error if this is the case.
1460 Request --write-conf, or allow starting with --old_conf.
1462 Severity : enhancement
1464 Description: Handle symlinks in the path when checking if Lustre is mounted.
1465 Details : Resolve intermediate symlinks when checking if a client has
1466 mounted a filesystem to avoid duplicate client mounts.
1471 Description: lconf can hit an error exception but still return success.
1472 Details : The lconf command catches the Command error exception at the top
1473 level script context and will exit with the associated exit
1474 status, but doesn't ensure that this exit status is non-zero.
1479 Description: failure of ptlrpc thread startup can cause oops
1480 Details : Starting a ptlrpc service thread can fail if there are a large
1481 number of threads or the server memory is very fragmented.
1482 Handle this without oopsing.
1485 Frequency : always, only if liblustre and non-default acceptor port was used
1487 Description: liblustre cannot connect to servers with non-default acceptor port
1488 Details : tcpnal_set_default_params() was not called and was therefore
1489 ignoring the environment varaible TCPNAL_PORT, as well as other
1490 TCPNAL_ environment variables
1495 Description: two objects could be created on the same OST for a single file
1496 Details : If an OST is down, in some cases it was possible to create two
1497 objects on a single OST for a single file. No problems other
1498 than potential performance impact and spurious error messages.
1502 Bugzilla : 5681, 9562
1503 Description: Client may oops in ll_unhash_aliases
1504 Details : Client dcache may become inconsistent in race condition.
1505 In some cases "getcwd" can fail if the current directory is
1511 Description: Inode refcounting problems in NFS export code
1512 Details : link_raw functions used to call d_instantiate without obtaining
1513 extra inode reference first.
1517 Bugzilla : 9942, 9903
1518 Description: Referencing freed requests leading to crash, memleaks with NFS.
1519 Details : We used to require that call to ll_revalidate_it was always
1520 followed by ll_lookup_it. Also with revalidate_special() it is
1521 possible to call ll_revalidate_it() twice for the same dentry
1522 even if first occurence returned success. This fix changes semantic
1523 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
1524 reference on a request referred from the intent.
1525 ll_intent_release() then releases such a request.
1528 Frequency : rare, normally benchmark loads only
1530 Description: unlinked inodes were kept in memory on the client
1531 Details : If a client is repeatedly creating and unlinking files it
1532 can accumulate a lot of stale inodes in the inode slab cache.
1533 If there is no other client load running this can cause the
1534 client node to run out of memory. Instead flush old inodes
1535 from client cache that have the same inode number as a new inode.
1538 Frequency : SLES9 2.6.5 kernel and long filenames only
1539 Bugzilla : 9969, 10379
1540 Description: utime reports stale NFS file handle
1541 Details : SLES9 uses out-of-dentry names in some cases, which confused
1542 the lustre dentry revalidation. Change it to always use the
1546 Frequency : rare, unless heavy write-truncate concurrency is continuous
1547 Bugzilla : 4180, 6984, 7171, 9963, 9331
1548 Description: OST becomes very slow and/or deadlocked during object unlink
1549 Details : filter_destroy() was holding onto the parent directory lock
1550 while truncating+unlinking objects. For very large objects this
1551 may block other threads for a long time and slow overall OST
1552 responsiveness. It may also be possible to get a lock ordering
1553 deadlock in this case, or run out of journal credits because of
1554 the combined truncate+unlink. Solution is to do object truncate
1555 first in one transaction without parent lock, and then do the
1556 final unlink in a new transaction with the parent lock. This
1557 reduces the lock hold time dramatically.
1560 Frequency : rare, 2.4 kernels only
1562 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
1563 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
1564 before unmounting their filesystem in order to avoid deadlock.
1565 The kernel_locked() function in 2.4 kernels only checks whether
1566 the kernel lock is held, not whether it is this process that is
1567 holding it as 2.6 kernels do.
1572 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
1573 Details : The client ptlrpc code may be trying to reconnect to a down
1574 server before a previous connection attempt has timed out.
1575 Increase the reconnect interval to be longer than the connection
1576 timeout interval to avoid sending duplicate connections to
1580 Frequency : echo_client brw_test command
1582 Description: fix echo_client to work with OST preallocated code
1583 Details : OST preallocation code (5137) didn't take echo_client IO path
1584 into account: echo_client calls filter methods outside of any
1585 OST thread and, hence, there is no per-thread preallocated
1586 pages and buffers to use. Solution: hijack pga pages for IO. As
1587 a byproduct, this avoids unnecessary data copying.
1591 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
1592 Description: Client can oops in mdc_commit_close() after open replay
1593 Details : It was possible for the MDS to return an open request with no
1594 transaction number in mds_finish_transno() if the client was
1595 evicted, but without actually returning an error. Clients
1596 would later try to replay that open and may trip an assertion
1597 Simplify the client close codepath, and always return an error
1598 from the MDS in case the open is not successful.
1601 Frequency : rare, 2.6 OSTs only
1603 Description: OST may deadlock under high load on fragmented files
1604 Details : If there was a heavy load and highly-fragmented OST filesystems
1605 it was possible to have all the OST threads deadlock waiting on
1606 allocation of biovecs, because the biovecs were not released
1607 until the entire RPC IO was completed. Instead, release biovecs
1608 as soon as they are complete to ensure forward IO progress.
1610 Severity : enhancement
1612 Description: Support for specifying external journal device at mount
1613 Details : If an OST or MDS device is formatted with an external journal
1614 device, this device major/minor is stored in the ext3 superblock
1615 and may not be valid for failover. Allow detecting and
1616 specifying the external journal at mount time.
1621 Description: Mounting an MDS with pending unlinked files may cause oops
1622 Details : target_finish_recovery() calls mds_postrecov() which returned
1623 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
1624 considers this an error and immediately begins cleaning up the
1625 lov, just after starting the mds_lov process
1627 Severity : enhancement
1629 Description: Implement 'lfs df' to report actual free space on per-OST basis
1630 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
1631 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
1632 usage of objects; '-h' to report in human readable format.
1634 ------------------------------------------------------------------------------
1636 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
1643 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
1644 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
1645 at startup during orphan unlinking. Instead, skip buddy update
1646 before setup, as it will be regenerated after recovery is complete.
1651 Description: If an OST is inactive, its locks might reference stale inodes.
1652 Details : lov_change_cbdata() must iterate over all namespaces, even if
1653 they are inactive to clear inode references from the lock.
1655 Severity : enhancement
1656 Frequency : occasional, if non-standard max_dirty_mb used
1658 Description: Client will block write RPCs if not enough grant
1659 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
1660 then the client will block writes while waiting for another RPC
1661 to complete instead of consuming its dirty limit. With change
1662 we get improved performance when max_dirty_mb is small.
1664 Severity : enhancement
1665 Bugzilla : 3389, 6253
1666 Description: Add support for supplementary groups on the MDS.
1667 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
1668 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
1669 lookups for user supplementary groups into a cache.
1673 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
1674 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
1675 for liblustre O_EXCL is enforced by MDS.
1679 Description: readdir never returns NULL in liblustre.
1680 Details : Corrected llu_iop_getdirentries logic, to return offset of next
1681 dentry in struct dirent.
1685 Frequency : liblustre only, depends on application IO pattern
1686 Description: liblustre clients evicted if not contacting servers
1687 Details : Don't put liblustre clients into the ping_evictor list, so
1688 they will not be evicted by the pinger ever.
1690 Severity : enhancement
1692 Description: Add ability to evict clients by NID from MDS.
1693 Details : By echoing "nid:$NID" string into
1694 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
1695 $NID would be instantly evicted from this MDS and from all active
1696 OSTs connected to it.
1700 Description: Do not query file size twice, somewhat slowing stat(2) calls.
1701 Details : lookup_it_finish() used to query file size from OSTs that was not
1706 Description: service threads change working directory to that of init
1707 Details : Starting lustre service threads may pin the working directory
1708 of the parent thread, making that filesystem busy. Threads
1709 now change to the working directory of init to avoid this.
1713 Frequency : during shutdown only
1714 Description: shutdown with a failed MDS or OST can cause unmount to hang
1715 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
1720 Frequency : 2.6 only
1721 Description: chmod/chown may include an extra supplementary group
1722 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
1723 supplementary group and if none is specified this is used.
1726 Bugzilla : 5479 (6816)
1728 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
1729 Details : If lookup is in progress on a file that is unlinked we might try
1730 to revalidate the inode and fail in revalidate after lookup is
1731 complete and ll_file_open() enqueues the open again but
1732 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
1735 Frequency : always, if lconf --abort_recovery used
1737 Description: lconf --abort_recovery fails with 'Operation not supported'
1738 Details : lconf was attempting to abort recovery on the MDT device and not
1741 ------------------------------------------------------------------------------
1743 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
1748 Frequency : rare (only unsupported configurations with a node running as an
1750 Bugzilla : 6514, 5137
1751 Description: Mounting a Lustre file system on a node running as an OST could
1753 Details : OSTs now preallocates memory needed to write out data at
1754 startup, instead of when needed, to avoid having to
1755 allocate memory in possibly low memory situations.
1756 Specifically, if the file system is mounted on on OST,
1757 memory pressure could force it to try to write out data,
1758 which it needed to allocate memory to do. Due to the low
1759 memory, it would be unable to do so and the node would
1760 become unresponsive.
1762 Severity : enhancement
1764 Description: Addition of lconf --service command line option
1765 Details : lconf now accepts a '--service <arg>' option, which is
1766 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
1768 Severity : enhancement
1770 Description: Failover mode is now the default for OSTs.
1771 Details : By default, OSTs will now run in failover mode. To return to
1772 the old behaviour, add '--failout' to the lmc line for OSTs.
1774 Severity : enhancement
1776 Description: Health checks are now provided for MDS and OSTs
1777 Details : Additional detailed health check information on MSD and OSTs
1778 is now provided through the procfs health_check value.
1781 Frequency : occasional, depends on IO load
1783 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
1784 after numerous create/delete cycles
1785 Details : The ext3 inode allocation policy would not allocate new inodes
1786 very well on the OSTs because there are no new directories
1787 being created. Instead we look for groups with free space if
1788 the parent directories are nearly full.
1793 Description: Network or server problems during mount may cause partially
1794 mounted clients instead of returning an error.
1795 Details : The config llog parsing code may overwrite the error return
1796 code during mount error handling, returning success instead
1802 Description: MDS can fail to allocate large reply buffers
1803 Details : After long uptimes the MDS can fail to allocate large reply
1804 buffers (e.g. zconf client mount config records) due to memory
1805 fragmentation or consumption by the buffer cache. Preallocate
1806 some large reply buffers so that these replies can be sent even
1807 under memory pressure.
1811 Frequency : rare (liblustre)
1812 Description: fsx running with liblustre complained that using truncate() to
1813 extend the file doesn't work. This patch corrects that issue.
1814 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
1815 ATTR_SIZE and lsm use in llu_setattr_raw.
1819 Frequency : rare, only 2.6 kernels
1820 Description: Unusual file access patterns on the MDS may result in inode
1821 data being lost in very rare circumstances.
1822 Details : Bad interaction between the ea-in-inode patch and the "no-read"
1823 code in the 2.6 kernel caused the inode and/or EA data not to
1824 be read from disk, causing single-file corruption.
1828 Frequency : rare, only 2.6 filesystems using extents
1829 Description: Heavy concurrent write and delete load may cause data corruption.
1830 Details : It was possible under high-load situations to have an extent
1831 metadata block in the block device cache from a just-unlinked
1832 file overwrite a newly-allocated data block. We now unmap any
1833 metadata buffers that alias just-allocated data blocks.
1837 Frequency : filesystems with default stripe_count larger than 77
1838 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
1839 Details : lconf specifies an inode size of 4096 bytes when the default
1840 stripe_count is larger than 77. This conflicts with the default
1841 inode density of 1 per 4096 bytes. Allocate smaller inodes in
1842 this case to avoid pinning too much memory for large EAs.
1844 ------------------------------------------------------------------------------
1846 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
1851 Frequency : rare (extremely heavy IO load with hundreds of clients)
1853 Description: Client is evicted, gets IO error writing to file
1854 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
1855 caused clients to be evicted for AST timeouts. The fixes in
1856 bug 5192 mean we no longer need to have such short AST timeouts
1857 so ldlm_timeout has been increased.
1860 Frequency : occasional during --force or --failover shutdown under load
1861 Bugzilla : 5949, 4834
1862 Description: Server oops/LBUG if stopped with --force or --failover under load
1863 Details : a collection of import/export refcount and cleanup ordering
1864 issues fixed for safer force cleanup
1867 Frequency : only filesystems larger than 120 OSTs
1868 Bugzilla : 5990, 6223
1869 Description: lfs getstripe would oops on a very large filesystem
1870 Details : lov_getconfig used kfree on vmalloc'd memory
1873 Frequency : only filesystems exporting via NFS to Solaris 10 clients
1874 Bugzilla : 6242, 6243
1875 Description: reading from files that had been truncated to a non-zero size
1876 but never opened returned no data
1877 Details : ll_file_read() reads zeros from no-object files to EOF
1882 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
1883 Details : There's little harm in aborting MDS/OSS recovery and letting it
1884 try again, so I removed the LASSERT and return an error instead.
1886 Severity : enhancement
1888 Description: New debugging infrastructure for tracking down data corruption
1889 Details : The I/O checksum code was replaced to: (a) control it at runtime,
1890 (b) cover more of the client-side code path, and (c) try to narrow
1891 down where problems occurred
1895 Bugzilla : 3819, 4364, 4397, 6313
1896 Description: Racing close and eviction MDS could cause assertion in mds_close
1897 Details : It was possible to get multiple mfd references during close and
1898 client eviction, leading to one thread referencing a freed mfd.
1900 Severity: : enhancement
1901 Bugzilla : 3262, 6359
1902 Description: Attempts to reconnect to servers are now more aggressive.
1903 Details : This builds on the enhanced upcall-less recovery that was added
1904 in 1.4.2. When trying to reconnect to servers, clients will
1905 now try each server in the failover group every 10 seconds. By
1906 default, clients would previously try one server every 25 seconds.
1911 Description: After recovery, certain operations trigger a failed
1912 assertion on a client.
1913 Details : Failing over an mds, using lconf -d --failover, while a
1914 client was doing a readdir() call would cause the client to
1915 LBUG after recovery completed and the readdir() was resent.
1917 Severity : enhancement
1919 Description: Default groups are now added by lconf
1920 Details : You can now run lconf --group <servicename> without having to
1921 manually add groups with lmc.
1924 Frequency : occasional
1926 Description: Nodes with an elan id of 0 trigger a failed assertion
1929 Frequency : always when accessing e.g. tty/console device nodes
1931 Description: tty and some other devices nodes cannot be used on lustre
1932 Details : file's private_data field is used by device data and lustre
1933 values in there got lost. New field was added to struct file to
1934 store fs-specific private data.
1937 Frequency : when exporting Lustre via NFS
1939 Description: NFSD failed occasionally when looking up a path component
1940 Details : NFSD is looking up ".." which was broken in ext3 directories
1941 that had grown large enough to become hashed.
1944 Frequency : Clusters with multiple interfaces not on the same subnet
1946 Description: Nodes will repeatedly try to reconnect to an interface which it
1947 cannot reach and report an error to the log.
1948 Details : Extra peer list entries will be created by lconf with some peers
1949 unreachable. lconf now validates the peer before adding it.
1952 Frequency : Only if a default stripe is set on the filesystem root.
1954 Description: Setting a default stripe on the filesystem root prevented the
1955 filesystem from being remounted.
1956 Details : The client was sending extra request flags in the root getattr
1957 request and did not allocate a reply buffer for the dir EA.
1960 Frequency : occasional, higher if lots of files are accessed by one client
1961 Bugzilla : 6159, 6097
1962 Description: Client trips assertion regarding lsm mismatch/magic
1963 Details : While revalidating inodes the VFS looks up inodes with ifind()
1964 and in rare cases can find an inode that is being freed.
1965 The ll_test_inode() code will free the lsm during ifind()
1966 when it finds an existing inode and then the VFS later attaches
1967 this free lsm to a new inode.
1971 Bugzilla : 6422, 7030
1972 Description: MDS deadlock between mkdir and client eviction
1973 Details : Creating a new file via mkdir or mknod (starting a transaction
1974 and getting the ns lock) can deadlock with client eviction
1975 (gets ns lock and trying to finish a synchronous transaction).
1978 Frequency : occasional
1979 Description: While starting a server, the fsfilt_ext3 module could not be
1981 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
1982 kernels. Previously, lconf would still use the ext3 name
1983 when trying to load modules. Now, it will correctly use
1984 ext3 on 2.4 and ldiskfs on 2.6.
1986 Severity : enhancement
1987 Description: The default stripe count has been changed to 1
1988 Details : The interpretation of the default stripe count (0, to lfs
1989 or lmc) has been changed to mean striping across a single
1990 OST, rather than all available. For general usage we have
1991 found a stripe count of 1 or 2 works best.
1993 Severity : enhancement
1994 Description: Add support for compiling against Cray portals.
1995 Details : Conditional compiling for some areas that are different
1999 Frequency : occasional
2000 Bugzilla : 6409, 6834
2001 Description: Creating files with an explicit stripe count may lead to
2002 a failed assertion on the MDS
2003 Details : If some OSTs are full or unavailable, creating files may
2004 trigger a failed assertion on the MDS. Now, Lustre will
2005 try to use other servers or return an error to the
2009 Frequency : occasional
2011 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2012 caused lock timeout during readahead (since 1.4.2).
2013 Details : Processes doing readahead might match a lock that hasn't been
2014 granted yet if there are overlapping and conflicting lock
2015 requests. The readahead process waits on ungranted lock
2016 (original lock is CBPENDING), while OST waits for that process
2017 to cancel CBPENDING read lock and eventually evicts client.
2019 Severity : enhancement
2021 Description: Initial enabling of flock support for clients
2022 Details : Implements fcntl advisory locking and file status functions.
2023 This feature is provided as an optional mount flag (default
2024 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2025 locking are implemented yet, and those that are are not guaranteed
2026 to be completely correct in production environments.
2027 mount -t lustre -o [flock|noflock] ...
2030 Frequency : occasional
2032 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2033 assertion in the ext3 JBD (journaling) layer.
2034 Details : The b_committed_data struct is protected by the big kernel lock
2035 in 2.4 kernels, serializing journal_commit_transaction() and
2036 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2037 access to this struct. In 2.6 kernels there is finer grained
2038 locking to improve SMP performance of the JBD layer.
2042 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2044 -----------------------------------------------------------------------------
2046 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2048 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2049 versions of Lustre. Please update all servers and clients to
2050 version 1.4.2 or later at the same time. You must also run
2051 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2052 to update the configuration logs.
2054 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2055 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2056 - swab llog records of type '0' so we get proper header size/idx (5861)
2057 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2058 - fix rename of one directory over another leaking an inode (5953)
2059 - avoid SetPageDirty on 2.6 (5981)
2060 - don't re-add just-being-destroyed locks to the waiting list (5653)
2061 - when creating new directories, inherit the parent's custom
2062 striping settings if present parent (3048)
2063 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2064 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2065 - don't hold client locks on temporary worklist from l_lru (5666)
2066 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2067 - automatically evict dead clients (5921)
2068 - Update file size properly in create+truncate+fstat case (6196)
2069 - Do not unhash mountpoint dentries, do not allow removal of
2071 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2072 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2073 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2074 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2075 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2076 - Correctly handle reads of files with no objects (6243)
2077 - lctl recover will also mark a device active if deactivate used (5933)
2079 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2080 - allow --write-conf on an MDS with different nettype than client (5619)
2081 - don't write config llogs to MDS for mounts not from that MDS (5617)
2082 - lconf should create multiple TCP connections from a client (5201)
2083 - init scripts are now turned off by default; run chkconfig --on
2084 lustre and chkconfig --on lustrefs to use them
2085 - upcalls are no longer needed for clients to recover to failover
2087 - add --abort-recovery option to lconf to abort recovery on device
2089 - add support for an arbitrary number of OSTs (3026)
2090 - Quota support protocol changes.
2091 - forward compatibility changes to wire structs (6007)
2092 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2093 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2094 - improved "lustre status" script
2095 - initialize blocksize for non-regular files (6062)
2096 - added --disable-server and --disable-client configure options (5782)
2097 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2098 - Vanilla 2.4.29 support
2099 - increase maximum number of obd devices to 520 (6242)
2100 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2101 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2102 SLES 9 2.6.5 kernels
2103 - limit stripes per file to 160 (the maximum EA size) (6093)
2105 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2108 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2109 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2110 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2111 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2112 - small ext3 extents cleanups and fixes (5733)
2113 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2114 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2115 - avoid needless client->OST connect, fix handle mismatch (5317)
2116 - fix DLM error path that led to out-of-sync client, long delays (5779)
2117 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2118 - fix several locking issues related to i_size (5492,5624,5654,5672)
2119 - don't move pending lock onto export if it is already evicted (5683)
2120 - fix kernel oops when creating .foo in unlinked directory (5548)
2121 - fix deadlock in obdfilter statistics vs. object create (5811)
2122 - use time_{before,after} to avoid timer jiffies wrap (5882)
2123 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2124 - Do not leak request if server was not able to process it (5154)
2125 - If mds_open unable to find parent dir, make that negative lookup(5154)
2126 - don't create new directories with extent-mapping (5909, 5936)
2128 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2129 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2130 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2131 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2132 - support for RHEL4 kernel on i686 (5773)
2133 - provide error messages when incompatible logs are encountered (5898)
2135 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2136 * version 1.4.0.10 (1.4.1 release candidate 1)
2138 - don't keep a lock reference when lock is not granted (4238)
2139 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2140 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2141 - drop import inflight refcount on signal_completed_replay error (5255)
2142 - unlock page after async write error during send (3677)
2143 - handle missing objects in filter_preprw_read properly (5265)
2144 - no transno return for symlink open, don't save no-trasno open (3440)
2145 - don't try to complete elan receive that already failed (4012)
2146 - free RPC server reply state on error (5406)
2147 - clean up thread from ptlrpc_start_thread() on error (5160)
2148 - readahead could read extra page into cache that wasn't ejected (5388)
2149 - prevent races in class_attach/setup/cleanup/detach (5260)
2150 - don't dereference de->d_inode after l_dput of de (5458)
2151 - use "int" for stripe value returned from lock_to_stripe (5544)
2152 - mballoc allocation and error-checking fixes in 2.6 (5504)
2153 - block device patches to fix I/O request sizes in 2.6 (5482)
2154 - look up hostnames for IB nals (5602)
2155 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2156 - don't start multiple acceptors for the same port (5277)
2157 - fix incorrect LASSERT in mds_getattr_name (5635)
2158 - export a proc file for general "ping" checking (5628)
2159 - fix "lfs check" to not block when the MDS is down (5628)
2161 - service request history (4965)
2162 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2163 - create an "evict_client" /proc entry on OSTs, like the MDS has
2164 - fix mount usage message, return errors per mount(8) (5168)
2165 - change grep [] to grep "[]" in tests so they work in more UMLs
2166 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2167 - remove ancient LOV_MAGIC_V0 EA support (5047)
2168 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2169 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2170 - allow readahead allocations to fail when low on memory (5383)
2171 - mmap locking landed again, after considerable improvement (2828)
2172 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2174 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2177 - send OST transaction number in read/write reply to free req (4966)
2178 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2179 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2180 - the watchdog thread now runs as interruptible (5246)
2181 - flock/lockf fixes (but it's still disabled, pending 5135)
2182 - don't use EXT3 constants in llite code (5094)
2183 - memory shortage at startup could cause assertion (5176)
2185 - reorganization of lov code
2186 - single portals codebase
2188 - add extents/mballoc support (5025)
2189 - direct I/O reads in the obdfilter (4048)
2190 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2192 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2195 - send OST transaction number in read/write reply to free req (4966)
2196 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2197 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2198 - the watchdog thread now runs as interruptible (5246)
2199 - handle missing objects in filter_preprw_read properly (5265)
2200 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2201 - drop import inflight refcount on signal_completed_replay error (5255)
2202 - unlock page after async write error during send (3677)
2203 - return original error code on reconstructed replies (3761)
2204 - no transno return for symlink open, don't save no-trasno open (3440)
2206 - add pid to ldlm debugging output (4922)
2207 - bump the watchdog timeouts -- we can't handle 30sec yet
2208 - extra debugging for orphan dentry/inode bug (5259)
2210 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2213 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2214 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2215 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2216 - fix NULL dereference in /proc/sys/portals/routes (4827)
2217 - allow failed mdc_close() operations to be interrupted (4561)
2218 - stop precreate on OST before MDS would time out on it (4778)
2219 - don't send partial-page writes before EOF from client (4410)
2220 - discard client grant for sub-page writes on large-page clients (4520)
2221 - don't free dentries not owned by NFS code, check generation (4806)
2222 - fix lsm leak if mds_create_objects() fails (4801)
2223 - limit debug_daemon file size, always print CERROR messages (4789)
2224 - use transno after validating reply (3892)
2225 - process timed out requests if import state changes (3754)
2226 - update mtime on OST during writes, return in glimpse (4829)
2227 - add mkfsoptions to LDAP (4679)
2228 - use ->max_readahead method instead of zapping global ra (5039)
2229 - don't interrupt __l_wait_event() during strace
2231 - add software watchdogs to catch hung threads quickly (4941)
2232 - make lustrefs init script start after nfs is mounted
2233 - fix CWARN/ERROR duplication (4930)
2234 - return async write errors to application if possible (2248)
2235 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2236 - print NAL number in %x format (4645)
2237 - update barely-supported suse-2.4.21-171 series (4842)
2238 - support for sles 9 %post scripts
2239 - support for building 2.6 kernel-source packages
2240 - support for sles km_* packages
2242 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2245 - ignore -ENOENT errors in osc_destroy (3639)
2246 - notify osc create thread that OSC is being cleaned up (4600)
2247 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2248 - reconstruct ost_handle() like mds_handle() (4657)
2249 - create a new thread to do import eviction to avoid deadlock (3969)
2250 - let lconf resolve symlinked-to devices (4629)
2251 - don't unlink "objects" from directory with default EA (4554)
2252 - hold socknal file ref over connect in case target is down (4394)
2253 - allow more than 32000 subdirectories in a single directory (3244)
2254 - fix blocks count for O_DIRECT writes (3751)
2255 - OST returns ENOSPC from object create when no space left (4539)
2256 - don't send truncate RPC if file size isn't changing (4410)
2257 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
2258 - bind to privileged port in socknal and tcpnal (3689)
2260 - rate limit CERROR/CWARN console message to avoid overload (4519)
2261 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
2262 - basic mmap support (3918)
2263 - kernel patch series update from b1_4 (4711)
2265 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
2268 - avoid crash during MDS cleanup with OST shut down (2775)
2269 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
2270 - don't use bad inodes on the MDS (3744)
2271 - dynamic object preallocation to improve recovery speed (4236)
2272 - don't hold spinlock over lock dumping or change debug flags (4401)
2273 - don't zero obd_dev when it is force cleaned (3651)
2274 - print grants to console if they go negative (4431)
2275 - "lctl deactivate" will stop automatic recovery attempts (3406)
2276 - look for existing locks in ldlm_handle_enqueue() (3764)
2277 - don't resolve lock handle twice in recovery avoiding race (4401)
2278 - revalidate should check working dir is a directory (4134)
2280 - don't always mark "slow" obdfilter messages as errors (4418)
2282 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
2285 - don't close LustreDB during write_conf until it is done (3860)
2286 - fix typo in lconf for_each_profile (3821)
2287 - allow dumping logs from multiple threads at one time (3820)
2288 - don't allow multiple threads in OSC recovery (3812)
2289 - fix debug_size parameters (3864)
2290 - fix mds_postrecov to initialize import for llog ctxt (3121)
2291 - replace config semaphore with spinlock (3306)
2292 - be sure to send a reply for a CANCEL rpc with bad export (3863)
2293 - don't allow enqueue to complete on a destroyed export (3822)
2294 - down write_lock before checking llog header bitmap (3825)
2295 - recover from lock replay timeout (3764)
2296 - up llog sem before sending rpc (3652)
2297 - reduce ns lock hold times when setting kms (3267)
2298 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
2299 - fix NULL deref and obd_dev leak on setup error (3312)
2300 - replace some LBUG about llog ops with error handling (3841)
2301 - don't match INVALID dentries from d_lookup and spin (3784)
2302 - hold dcache_lock while marking dentries INVALID and hashing (4255)
2303 - fix invalid assertion in ptlrpc_set_wait (3880)
2305 - add libwrap support for the TCP acceptor (3996)
2306 - add /proc/sys/portals/routes for non-root route listing (3994)
2307 - allow setting MDS UUID in .xml (2580)
2308 - print the stack of a process that LBUGs (4228)
2310 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
2313 - don't cleanup request in ll_file_open() on failed MDS open (3430)
2314 - make sure to unset replay flag from failed open requests (3440)
2315 - if default stripe count is 0, use OST count for inode size (3636)
2316 - update parent mtime/ctime on client for create/unlink (2611)
2317 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
2318 - free recovery state on server during a forced cleanup (3571)
2319 - unregister_reply for resent reqs (3063)
2320 - loop back devices mounting and status check on 2.6 (3563)
2321 - fix resource-creation race that can provoke i_size == 0 (3513)
2322 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
2323 - more debugging for page-accounting assertion (3746)
2324 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
2325 - avoid deadlock after precreation failure (3758)
2326 - fix race and lock order deadlock in orphan handling (3450, 3750)
2327 - add validity checks when grabbing inodes from l_ast_data (3599)
2329 - add /proc/.../recovery_status to obdfilter (3428)
2330 - lightweight CDEBUG infrastructure, debug daemon (3668)
2331 - change default OSC RPC parameters to be better on small clusters
2332 - turn off OST read cache for files smaller than 32MB
2333 - install man pages and include them in rpms (3100)
2334 - add new init script for (un)mounting lustre filesystems (2593)
2335 - run chkconfig in %post for init scripts (3701)
2336 - drop scimac NAL (unmaintained)
2338 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
2341 - clean kiobufs before and after use (3485)
2342 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
2343 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
2344 - change init script to fail more clearly if not run as root (1528)
2345 - allow clients to reconnect during replay (1742)
2346 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
2347 - don't do DNS lookups on NIDs too small for IP addresses (3442)
2348 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
2349 - fix cond_resched (3554)
2350 - only evict unfinished clients after recovery (3515)
2351 - allow bulk resend, prevent data loss (3570)
2352 - dynamic ptlrpc request buffer allocation (2102)
2353 - don't allow unlinking open directory if it isn't empty (2904)
2354 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
2355 - remove extraneous obd dereference causing LASSERT failure (3334)
2356 - don't use get_cycles() when creating temp. files on the mds (3156)
2357 - hold i_sem when setting i_size in ll_extent_lock() (3564)
2358 - handle EEXIST for set-stripe, set proper directory name (3336)
2360 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
2361 - fix ksocknal_fmb_callback() error messages (2918)
2363 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
2366 - don't copy lvb into (possibly NULL) reply on error (2983)
2367 - don't deref dentry after dput, don't free lvb on error (2922)
2368 - use the kms to determine writeback rpc length (2947)
2369 - increment oti_logcookies when osc is inactive (2948)
2370 - update client's i_blocks count via lvb messages (2543)
2371 - handle intent open/close of special files properly (1557)
2372 - mount MDS with errors=remount-ro, like obdfilter (2009)
2373 - initialize lock handle to avoid ASSERT on error cleanup (3057)
2374 - don't use cancelling-locks' kms values (2947)
2375 - use highest lock extent for kms, not last one (2925)
2376 - don't dereference ERR_PTR() dentry in error handling path (3107)
2377 - fix thread race in portals_debug_dumplog() (3122)
2378 - create lprocfs device entries at setup instead of at attach (1519)
2379 - common AST error handler, don't evict client on completion race (3145)
2380 - zero nameidata in detach_mnt in 2.6 (3118)
2381 - verify d_inode after revalidate_special is valid in 2.6 (3116)
2382 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
2383 - initialize RPC timeout timer earlier for 2.6 (3219)
2384 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
2385 - print nal/nid for unknown nid (3258)
2386 - additional checks for oscc recovery before doing precreate (3284)
2387 - fix ll_extent_lock() error return code for 64-bit systems (3043)
2388 - don't crash in mdc_close for bad permissions on open (3285)
2389 - zero i_rdev for non-device files (3147)
2390 - clear page->private before handing to FS, better assertion (3119)
2391 - tune the read pipeline (3236)
2392 - fix incorrect decref of invalidated dentry (2350)
2393 - provide read-ahead stats and refine rpc in flight stats (3328)
2394 - don't hold journal transaction open across create RPC (3313)
2395 - update atime on MDS at close time (3265)
2396 - close LDAP connection when recovering to avoid server load (3315)
2397 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
2398 - don't leak open file on MDS after open resend (3325)
2399 - serialize filter_precreate and filter_destroy_precreated (3329)
2400 - loop device shouldn't call sync_dev() for nul device (3092)
2401 - clear page cache after eviction (2766)
2402 - resynchronize MDS->OST in background (2824)
2403 - refuse to mount the same filesystem twice on same mountpoint (3394)
2404 - allow llmount to create routes for mounting behind routers (3320)
2405 - push lock cancellation to blocking thread for glimpse ASTs (3409)
2406 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
2407 - fix rare problem with rename on htree directories (3417)
2409 - allow default OST striping configuration per directory (1414)
2410 - fix compilation for qswnal for 2.6 kernels (3125)
2411 - increase maximum number of MDS request buffers for large systems
2412 - change liblustreapi to be useful for external progs like lfsck (3098)
2413 - increase local configuration timeout for slow disks (3353)
2414 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
2416 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2419 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
2420 - don't overwrite extent policy data in reply if lock was blocked (2901)
2421 - drop filter export grants atomically with removal from device (2663)
2422 - del obd_self_export from work_list in class_disconnect_exports (2908)
2423 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
2424 - swab reply message in mdc_close, other PPC fixes (2464)
2425 - fix destroying of named logs (2325)
2426 - overwrite old logs when running lconf --write_conf (2264)
2427 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
2428 - fix race in target_handle_connect (2898)
2429 - mds_reint_create() should take same inode create lock (2926)
2430 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
2431 - don't close files for self_export to avoid uninitialized obd (2936)
2432 - allow MDS with the same name as client node (2939)
2433 - hold dentry reference for closed log files for unlink (2325)
2434 - reserve space for all logs during transactions (2059)
2435 - don't evict page beyond end of stripe extent (2925)
2436 - don't oops on a deleted current working directory (2399)
2437 - handle hard links to targets without a parent properly (2517)
2438 - don't dereference NULL lock when racing during eviction (2867)
2439 - don't grow lock extents when lots of conflicting locks (2919)
2441 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
2444 - account for cache space usage on clients to avoid data loss (974)
2445 - lfsck support in lustre kernel code (2349)
2446 - reduce journal credits needed for BRW writes (2370)
2447 - orphan handling to avoid losing space on client/server crashes
2448 - ptlrpcd can be blocked, stopping ALL progress (2477)
2449 - use lock value blocks to assist in proper KMS, faster stat (1021)
2450 - takes i_sem instead of DLM locks internally on obdfilter (2720)
2451 - recovery for initial connections (2355)
2452 - fixes for mds_cleanup_orphans (1934)
2453 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
2454 - block all file creations until orphan recovery completes (1901)
2455 - client remove rq_connection from request struct (2423)
2456 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
2457 - recovery timer race (2670)
2458 - mdc_close recovey bug (2532)
2459 - ptlrpc cleanup bug (2710)
2460 - mds timeout on local locks (2588)
2461 - namespace lock held during RPCs (2431)
2462 - handle interrupted sync write properly (2503)
2463 - don't try to handle a message that hasn't been replied to (2699)
2464 - client assert failure during cleanup after abort recovery (2701)
2465 - leak mdc device after failed mount (2712)
2466 - ptlrpc_check_set allows timedout requests to complete (2714)
2467 - wait for inflight reqs when ptlrpcd finishes (2710)
2468 - make sure unregistered services are removed from the srv_list
2469 - reset bulk XID's when resending them (caught by 1138 test)
2470 - unregister_bulk after timeout
2471 - fix lconf error (2694)
2472 - handle write after unfinished setstripe, stripe-only getstripe (2388)
2473 - readahead locks pages, leaves pending causing memory pressure (2673)
2474 - increase OST request buffers to 4096 on large machines (2729)
2475 - fix up permission of existing directories in simple_mkdir (2661)
2476 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
2477 - don't assign transno to errored transactions (2742)
2478 - don't delete objects on OST if given a bogus objid from MDS (2751)
2479 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
2480 - if rq_no_resend, then timeout request after recovery (2432)
2481 - fix MDS llog_logid record size, 64-bit array alignment (2733)
2482 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
2483 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
2484 - MDS recovery shouldn't delete objects that it has given out (2730)
2485 - if enqueue arrives after completion, don't clobber LVB (2819)
2486 - don't unlock pages twice when trigger_group_io returns error (2814)
2487 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
2488 - don't write pages to disk if there was an error (1450)
2489 - don't ping imports that have recovery disabled (2676)
2490 - take buffered bytes into account when balancing socknal conn (2817)
2491 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
2492 - reconnect unlink llog connection after MDS reconnects to OST (2816)
2493 - remove little-endian swabbing of llog records (1987)
2494 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
2495 - retry reposting request buffers if they fail (1191)
2496 - grow extent at grant time to avoid granting a revoked lock (2809)
2497 - lock revoke doesn't evict page if covered by a second lock (2765)
2498 - disable VM readahead to avoid reading outside lock extents (2805)
2500 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
2501 - updated kernel patches for hp-2.4.20 kernel (2681)
2503 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
2506 - fix truncated write corruption (2366)
2507 - fix for failed assertion in iopen_connect_dentry (1792,2517)
2509 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
2510 - ensure len(uuid) < 37 in lmc (1171)
2511 - fix ia64 OOPS in llog_test (2255)
2512 - zero end of page at obdfilter for partial page writes (2648)
2513 - don't leave stale dentries around after renames (bug 2428)
2514 - fix timeouts when evicting a client with a single lock held (2642)
2515 - set deadline for the initial HELLO message to drain (2634)
2516 - print out dotted-quad IP addresses in the socknal (2302)
2518 - additional debugging for MDS client eviction problem (2443)
2519 - fix mkfsoptions support for osts (2603, 2604)
2521 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
2524 - add series for the vanilla 2.6.0 kernel
2525 - add series for the vanilla 2.4.24 kernel
2526 - add series for a cray x86/64 UL kernel drop
2527 - fix xattr patches for the vanilla 2.4.19 series
2529 - generate true UUIDs in lmc (1171)
2530 - have portals stack dumping break in UML (2466)
2531 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
2532 - allocate with _NOFS in ldlm to avoid deadlock (1933)
2533 - wake callback waiting threads on client eviction (2460)
2534 - Add --ptldebug and --subsystem to lmc (1719)
2535 - update assertion to allow safe interrupt allocation
2536 - set rq_no_resend for cancel requests (2432)
2537 - recalculate ptlrpcd timeout after resend (2494)
2538 - call vfs_rmdir when removing pending directories (2368)
2539 - fix renaming a file to itself (2429)
2540 - lmc creates a default one-stripe lov (2454)
2541 - expand procfs space to handle large clusters (2326)
2542 - increase UML stack to avoid overflow
2543 - update lconf's list of debug and subsystem masks
2544 - fix lfs find --obd (2510)
2545 - /proc tunable for disabling filter read caching (2591)
2546 - stop rpm packages from altering slapd.conf (2301)
2547 - disable nagle in the socknal under 0conf (2578)
2548 - choose mds inode size based on stripe count (2572)
2549 - fix kernel-source rpm problems (2516)
2551 - add --disable-doc to avoid pdf generation (2421)
2552 - update documentation, tests, type-os, comments
2553 - avoid format warnings on ia64
2554 - remove the TOE NAL
2555 - tiny code cleanups by removing unused fields
2557 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
2560 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
2561 - remove the most blatant lies from BUILDING file (2371)
2562 - change default debug level to reasonable production setting
2563 - reduce client side cache size to reduce cache flush time
2564 - reduce max RPCs in flight to avoid unnecessary file fragmentation
2565 - make TCP zerocopy and pinger support enabled by default (2476)
2566 - sync writes completed after process exits caused crashes (2319)
2567 - maintain correct mount count on the MDS (2356)
2568 - backout 1557, because 2316 wasn't really fixed
2569 - better file I/O statistics gathering in /proc
2570 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
2571 - another kernel patch to fix zero-copy TCP function export
2572 - don't take duplicate lock when processing re-sent getattr (2420)
2573 - lctl uses obd_self_export instead of creating new conn (2353)
2574 - MDS/OST recovery case which requires object creation asserted (2425)
2575 - move lfs from /usr/sbin to /usr/bin in packages
2576 - fix race between mds_client_add and mds_client_free (2417)
2577 - use kmalloc instead of slabs in portals (2430)
2578 - don't create duplicate records when a failover MDS is present (2442)
2579 - remove unnecessary mount age check (2332)
2580 - don't remove directory inodes from locks prematurely (2451)
2581 - don't break if MDS service name is the same as hostname (2103)
2582 - fix races in client write RPC generation when cache full (2482)
2584 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
2587 - remove now-unused request->rq_obd (278)
2588 - if an allocation fails, print out how much memory we've used (1933)
2589 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
2590 - add the "configurable stack size" patch to most series files (1256)
2591 - ability to write large log records, for 100+ OST configs (2306)
2592 - fix NULL deref when filter_prep fails (2314)
2593 - fix operator precedence error in filter_sync
2594 - dynamic allocation of socknal TX descriptors (2315)
2595 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
2596 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
2597 - ensure that CWARN messages in Portals always get to the syslog
2598 - __init/__exit are not for prototype decls (ldlm_init/exit)
2599 - x86-64 compile warning fixes
2600 - fix gateway LMC keyword conflict (2318)
2601 - fix MDS lock inversions in getattr/reint paths (1844)
2602 - fix a rare lock re-ordering bug, which caused deadlock (2322)
2603 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
2604 - DLM race condition prevented some lock evictions (2328)
2605 - ENOMEM detection and retry on socknal sends (2230)
2606 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
2607 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
2608 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
2609 - better error messages when a client is rejected during recovery (1505)
2610 - avoid cancelling locks which were never granted, after failure (2330)
2611 - fix i_sem/journal inversion in mds_client_add (2333)
2612 - fix truncate/getattr lock cycle deadlock (2334)
2613 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
2614 - fix two rare exit paths which could leak an l_lock() ref (2321)
2615 - fencepost error in MDS/OST orphan recovery (2226)
2616 - make log record alignment 8 bytes (1988)
2617 - lstripe now fails when requested offset > ost_count (2237)
2618 - ensure that all kernel series have a complete list.h (1607)
2619 - fix crashes in special-file operations (2316)
2620 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
2622 - allow configurable automake binary, for testing new versions
2623 - small update to the lfs documentation
2625 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
2627 * fix negative export reference count in fsfilt_sync (2312)
2629 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
2630 * release candidate 0.9.1
2632 - orphans are moved into the PENDING directory for possible recovery
2633 - replayed opens now open by fid for orphan/rename safety (1042)
2634 - last close of an orphan inode generates a transno (683)
2635 - chdir() and mount() now pin the directory entry (1020)
2636 - avoid CERROR in normal ll_setattr_raw() error case (1500)
2637 - discard very old requests without processing them (1502)
2638 - remove some common, well-understood CERRORs (1505)
2639 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
2640 - clear "grant" flags in OST replies until OST grant code lands (1644)
2641 - fix read performance by not clobbering i_blksize on client (1598)
2642 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
2643 - make LRU size a /proc tunable, clears locks when reduced (707)
2644 - fix some lprocfs rot that prevented ptlbd from loading (1732)
2645 - server locks take references on exports now (1558)
2646 - build fixes for 2.4.20-rh trees (1663)
2647 - return an error from lov_create if all OSCs are inactive (1751)
2648 - fix import levels when a reconnect happens without a timeout (1597)
2649 - exit early from mds_open if we get a lookup error (1749)
2650 - partial page read at EOF wouldn't wait for disk before sending (1642)
2651 - avoid NULL deref in obdfilter when reading page past EOF (1592)
2652 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
2653 - fix LBUG in ll_it_open_error with rc = -2 (1861)
2654 - write/truncate lock inversion (1639)
2655 - Don't auto-load obdclass, portals modules during cleanup (1495)
2656 - fix timestamps from jumping to "now" (1763)
2657 - extra journal assertions (1648)
2658 - add an extra multiunlink test (1771)
2659 - fix read_record/write_record API (1776)
2660 - fix leak of offset_extent, possible incorrect i_size later (1772)
2661 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
2662 - Debugging for the kqswnal_get_idle_tx problems (1820)
2663 - Allow recovery to be attempted multiple times (1536)
2664 - Write out MDS last_rcvd file after it is first created (1600)
2665 - Fix tx_descriptor leak in failed transmit situations (1827)
2666 - ext3 journaling fixes for assertion failure after IO error (1871)
2667 - class_export_put() on freed export after completion AST error (1896)
2668 - Fix revalidate looping in VFS (1322)
2669 - Don't access a freed export during MDS_REINT timeout (1521)
2670 - Add open-unlink recovery support on the MDS (1673,1764)
2671 - Return an error if no MDS data was read from last_rcvd (1946)
2672 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
2673 - Don't setup a disk device that doesn't match exported UUID (317)
2674 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
2675 - avoid committing NULL handle in force close
2676 - local.sh is now a one-stripe LOV configuration
2677 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
2678 - fix ext3 htree duplicate directory entry corruption (1516)
2679 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
2680 - update to new LOV EA format (2097)
2681 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
2683 2003-06-15 Phil Schwan <phil@clusterfs.com>
2686 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
2687 - per-import/export recovery handling (958, 931, 959)
2688 - multiple last-rcvd slots, for serving multiple FSes (949)
2689 - connections are again shared between multiple imp/exports (963, 964)
2690 - "umount -f" would hang if any requests needed to be sent (393, 978)
2691 - avoid pinning large req buffer by copying for queued messages (989)
2692 - add "uuid" to "lctl device" command to help upcalls (991)
2693 - "open" RPCs with transnos would confuse recovery counters (1037)
2694 - do proper endian conversion of all wire messages (288, 340, 891)
2695 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
2696 - call waiting locks callback from LDLM recovery thread (1127, 1151)
2697 - fix ptlrpc_connection leak in target_handle_connect (1174)
2698 - fix import refcounting bug in OST and MDS cleanup (1134)
2699 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
2700 - fix very unlikely obd_types race condition (501)
2701 - remove osc_open hack for echo_client (1187)
2702 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
2703 - a failure in read_inode2 leads to deadlock (1139)
2704 - cancel ack-locks as soon as transaction is committed (1072)
2705 - fix major leaks and crashes in the bulk I/O path (937, 1057)
2706 - make sure to commitrw after any preprw to avoid deadlock (1162)
2707 - failing to execute a file in a lustre FS would lock inode (1203)
2708 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
2709 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
2710 - obd_uuid2tgt would walk off the end of the list (1255)
2711 - on IA64 the osc would give portals incorrect bulk size (1258)
2712 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
2713 - fix lock inversion caused by new llite matching code (1282)
2714 - limit the number of dirty pages on a client to 10MB (1286)
2715 - timed out locks were not being corrected cancelled (1289)
2716 - fix O_DIRECT above 4GB on IA-32 (1292)
2717 * major user-visible changes
2718 - fail out/fail over policy now controlled by the upcall (993)
2720 - add OBD_PING to check server availability and failure (954)
2721 - lustre messages are now sent in sending host order (288, 340, 891)
2722 - add eadatalen to MDS getattr reply (340)
2723 - OST read replies may contain second buffer, with per-page status (593)
2725 2003-03-11 Phil Schwan <phil@clusterfs.com>
2728 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
2729 - failed open()s could cause deadlock; fixed (867, 869)
2730 - stop cancelling OST locks when files are closed (481)
2731 - overlapping XID spaces caused network corruption (851, 853)
2732 - fix unsafe fsfilt counter arithmetic; change to atomic_t
2733 - setattr_raw added, to do single-RPC, server-side setattrs
2734 - lmc/lconf syntax change for OST UUIDs
2735 - fix crashy race condition between ptlrpc_free_req and osc_close
2736 - don't use request in mdc_enqueue if we hit a timeout (889)
2737 - don't set the inode i_size for regular files from the MDS (896)
2738 - handle out of order completion AST (842)
2739 - don't LBUG if a lock request times out after receiving AST (913)
2740 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
2741 - if a bad lock AST arrives, send an error instead of dropping entirely
2742 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
2743 - fix leak in bulk IO when only partially completed (899, 900, 926)
2744 - fix O_DIRECT for ia64 (55)
2745 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
2746 - C-z after timeout could hang a process forever; fixed (977)
2748 - client-side I/O cache (678, 924, 929, 941, 970)
2750 - READPAGE and SETATTRs which don't take server-side locks get
2753 2003-02-11 Phil Schwan <phil@clusterfs.com>
2756 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
2757 - Fix fsfilt_extN_readpage() to read a full page of directory
2758 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
2759 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
2760 - Fix imbalanced LOV object allocation and out-of-bound access (469)
2761 - Most intent operations were removed, in favour of a new RPC mode
2762 that does a single RPC to the server and bypasses most of the VFS
2763 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
2764 - Aggressively cancel local locks on DLM servers
2765 - mds_reint_unlink sends EA to the client if it's the last nlink.
2766 client uses that EA to unlink OST objects.
2767 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
2768 - recursive symlinks were fixed (439)
2769 - fixed NULL deref in DEBUG_REQ
2770 - filter_update_lastobjid no longer calls sync, which annoyed extN
2771 - fixed multi-client small-writes to a single file problem (445)
2772 - fixed mtime updates during file writes (607)
2773 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
2774 - fixed bug in obd_brw_read/write() (under guise of testing 367)
2775 - fixed Linux OST size reporting problem (444, 656)
2776 - OST now updates object mtime with writes or setattr (607, 619)
2777 - client verifies file size before zeroing page past EOF (445)
2778 - OST now writes last allocated objid to disk with allocation (108)
2779 - LOV on echo now works (409)
2781 - mds_reint_unlink sends a new buffer, with the EA included. this
2782 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
2783 set if a regular file was being unlinked, and it was the last link
2784 - use PtlGet from the target for bulk writes (315)
2785 - OST now updates object mtime with writes or setattr (607, 619)
2786 - LDLM now has a grant-time callback to revalidate locked items, if
2788 - Many MDS operations were reorganized to combat race conditions
2790 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
2791 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
2793 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
2796 - Fully reactivate OST imports after reconnection (512, others)
2797 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
2798 - More graceful error handling for truncating on dead OST (515)
2799 - Don't error out unless we're actually accessing dead stripes (474)
2800 - Fix garbage sizes when stripes are missing (410)
2801 - LRU counters were broken, causing constant lock purge (433, 432)
2802 - garbage on read from stripes with failed OSTs (441)
2803 - mark OSCs as active before reconnecting during recovery (438)
2804 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
2805 - lfind did not preserve OST order in output (443)
2806 - symlinks cause hung clients, incorrect data (439)
2807 - stop dereferencing request after dropping refcount (457)
2808 - don't LASSERT(spin_is_locked) on non-SMP (455)
2809 - fixes for many rename() bugs
2810 - fstat didn't correctly synchronize attributes (399)
2811 - server must handle lock cancellation during blocking AST prep (487)
2812 - bulk descriptors were free()d too soon (511)
2813 - fix paths in lconf, which would load incorrect modules (451, 507)
2814 - fix confusing lconf 'host not found' error message (386)
2815 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
2816 - fix race condition in mdc_blocking_ast() for inode access (526)
2817 - fix lov_unpackmd() unpacking wrong number of stripes (537)
2818 - fix lov_set_osc_active() marking wrong OSC inactive (440)
2819 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
2820 - fix multiple writes of stripe MD to MDS (358, maybe 519)
2821 - fix lstripe in several ways (kernel side) (527)
2822 - fix request leak in ldlm_cli_enqueue (262)
2823 - incorrect OSC was marked inactive after OST failure
2824 - call mds_fs_cleanup before unmounting filesystem (524)
2825 - fix races between taking ns_lock and ldlm_lock_change_resource
2826 - fix races updating LOV export open file list
2827 - fix lov_enqueue error path, avoid decref-ing bad lock handle
2828 - fix recovery NULL deref in ldlm_cli_cancel_unused
2829 - fix some DLM races by using new hash table for lock handles (419)
2830 - permit the client to specify desired inodes, at replay
2831 - duplicate requests when we queue them for replay reintegration
2832 - fix last_rcvd offset calculation
2833 - sync after each recovered transaction, so we always make progress
2834 - never, not always, ERESTART requests without transnos
2835 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
2837 - skip replay if the MDS says that the client is already connected
2838 - don't check for a recovery-enabled export to match lctl's UUID
2839 - don't INC_USE_COUNT for phantom exports
2840 - don't crash when cleaning up phantom exports (567)
2841 - don't double-finish or set replay data for errored mdc_open requests
2842 - abort requests when they time out, so we don't get old replies
2843 - send/receive replies for AST messages again
2844 - if the client says that it doesn't have the lock, cancel it on the
2846 - if we timeout during I/O, don't try to cancel an in-use lock; instead
2847 mark it as destroyed, it will all work out when decref is called
2848 - fix module use counts (22, 581)
2850 - ASTs now expect a reply (server cancels lock on error reply)
2852 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
2855 - fix many simultaneous client startup (392)
2856 - fix dentry->d_it clobbering
2857 - credentials weren't being shipped for readdir/getattr operations
2858 - remove invalid assertions triggered during some concurrent MD
2860 - proper Lustre versions added (336, 389)
2861 - fix memory leak for create error case (398)
2862 - fix LOV locking bug that would get cli/srv out of sync
2863 - fix echo client over LOV (409)
2864 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
2865 - fix double-O_EXCL intent crash (424)
2866 - avoid sending multiple lock CANCELs (352)
2868 - MDS can do multi-client recovery (modulo bugs in new code)
2870 - many updates, edits, cleanups
2872 2002-11-18 Phil Schwan <phil@clusterfs.com>
2875 - fix null d_it dereference (346)
2876 - fix full OST/dbench hang (333)
2877 - fix permission problem with file removal (286)
2878 - fix removal of OSCs from LOV when they fail
2879 - fix NULL deref during bulk timeout (214)
2880 - fix problems related to multiple filesystems on one MDS (241)
2881 - fixed serious subtle metadata locking bugs
2882 - free locks on clients when inodes are removed due to memory
2884 - fix inode pointer in lock data (285)
2885 - partial support for multiple MDS on a single host (241)
2886 - data locks weren't cancelled at clear_inode time (290, 311)
2887 - intent locks could lead to unbounded lock growth (205)
2888 - added a maximum lock count, an LRU list, and a flusher
2889 - fix multiple rename (365)
2890 - properly abstracted the echo client
2891 - OSC locked 1 byte too many; fixed
2892 - rewrote brw callback code:
2893 - fixed recovery bugs related to LOVs (306)
2894 - fixed too-many-pages-in-one-write crash (191)
2895 - fixed (again) crash in sync_io_timeout (214)
2896 - probably fixed callback-related race (385)
2898 - Add capability to MDS protocol
2899 - LDLM cancellations and callbacks on different portals
2901 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
2904 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
2905 - defer open object destruction to close time (601981, 138)
2906 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
2907 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
2908 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
2909 - don't call obd_setattr in ll_file_release for destroyed objects
2910 * protocol change to lustre_msg: move |version| and add |flags|
2911 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
2912 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
2913 * added replay of create, unlink, link and rename operations during
2914 MDS failover; recovery should be much more robust now
2915 * remove failed OSCs from LOVs (only lov_create uses this so far)
2916 * the lustre-HOWTO was brought (more) up to date (582544)
2918 2002-10-23 Phil Schwan <phil@clusterfs.com>
2921 - in-use dentries weren't being reused properly (617851)
2922 - prevent multiple LDLM setup (599178)
2923 - fix LOV size calculations for truncate (617853)
2924 - fix client handling of MDS intent errors (POSIX)
2925 - fix permission bug in lovstripe.c test (624321)
2926 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
2927 - truncate past end of file could corrupt data
2928 - proper cleanup after timeouts, crashes, etc (592524, 550815)
2929 - a race in recovery could return ETIMEDOUT to apps (623947)
2930 - building outside the source directory was fixed
2931 * the lustre-HOWTO was brought (more) up to date (582544)
2932 * major progress was made on recovery functionality
2934 2002-10-10 Phil Schwan <phil@clusterfs.com>
2937 - recovery deadlock fix
2938 - rm -rf causes LBUG fix (617817)
2939 - file open by multiple tasks fix (618962)
2940 - directory permissions bugs (602707 and 620007)
2941 - journal_stop fixed with locking (611313)
2942 - O_APPEND failures resolved (618273, perhaps 614459)
2943 - lconf PATH fix (619770)
2944 - IA64 build fix (621450)
2945 - RPC buffer sizes scale with amount of memory
2947 2002-10-01 Phil Schwan <phil@clusterfs.com>
2950 - locks would be cancelled without throwing away data pages,
2951 resulting in inconsistent data (605627)
2952 - inode attributes were not always being refreshed (605627, 612449)
2953 - lconf now continues to cleanup after lctl reports an error
2954 - MDS now enforces user permissions (602707)
2955 - lprocfs cleanup fixed, but not yet enabled (614157)
2956 - fixed infinite server hang, should a client not respond to an AST
2957 - avoid going into recovery if user calls readlink() with a buffer
2958 that's too small (613941)
2959 - AST RPCs no longer require replies (614867) -- this may be changed
2960 - don't crash server if client sends an IOV that's too big (611336)
2961 - fixed lock conversion deadlock (611892)
2962 - fixed the following of symlinks (614622)
2963 * recovery: the server can remove locks from a client that dies, other
2964 clients can make progress
2965 * more extN patch fixes
2966 * compile-time configurable ptlrpc buffer allocations
2968 - collaborative read cache document
2969 - Lustre Lite Performance CDR document-in-progress
2971 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
2974 - fix typo in patch-2.4.18
2976 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
2979 - clear ptlrpc request each time in handle_incoming_request()
2980 - unlink of files now destroys the object on the OST
2982 2002-09-19 Peter Braam <braam@clusterfs.com>
2984 * add hard link support
2985 * change obdfile creation method
2986 * kernel patch changed
2988 2002-09-19 Peter Braam <braam@clusterfs.com>
2991 - stack overflow bug in extN fixed
2993 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
2995 * documentation updates
2996 - add man pages for config tools
2997 - update tests/README to describe testing with new config tools
2998 - finish metadata API descriptions
2999 * bug fixes and cleanups
3000 - statfs workaround for 16TB limit
3001 - LOV stripe allocation improved, can stripe on subset of OSTs
3002 - LOV file size/IO offset was wrong for files > 4GB in size
3003 - object EA data was being dropped, caused files to be unreadable
3004 - memory overflow with non-LOV OST caused memory corruption
3005 - fixed regression tests to work with new config tools, obdfilter
3006 - fixed bug when directory size became larger than 1 block
3007 - fixed bug (for single client case) when PWD was deleted
3008 - invalidate local directory pages when doing intent-based ops
3009 - avoid LDLM oops when lock callback contained bad data
3011 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3013 * documentation updates
3014 * bug fixes and cleanups
3015 - configuration tools
3018 - 64-bit compile warnings
3019 - 64-bit internal statfs data
3021 * test_brw on persistent OST devices
3023 * lprocfs (disabled)
3025 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3027 * documentation updates
3028 * bug fixes and cleanups
3029 * configuration tools
3031 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3034 * many small fixes to 0_5_4
3035 * io/network handling
3036 * thinkos in MDS operations
3038 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3041 * crucial basic fixes to 0.5.3
3042 * IOR, Iozone work over Elan
3045 2002-08-07 Phil Schwan <phil@clusterfs.com>
3046 * version 0_5_3, our first alpha
3047 * we use the new Portals iovs
3048 * documentation updates
3049 * bug fixes and cleanups
3050 * small changes in the DLM wire protocol
3052 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3053 * version 0_5_1 with some initial stability,
3054 * locking on MD and file I/O.
3055 * documentation updates
3056 * several bug fixes since 0.5.0
3057 * small changes in wire protocol
3059 2002-07-18 Phil Schwan <phil@clusterfs.com>
3061 * delivered as Lustre Light Alpha
3062 * fixed a crash after handling invalid MDS requests
3063 * fixed directory pages for architectures with non-4k pages sizes
3065 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3066 * release version v0_4_4
3067 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3069 * Fixes a number of interruption problems with OST operations.
3070 * Update documentation for portals header changes
3071 * Move all wire protocol structs/defines to lustre_idl.h
3072 * Fixes symlink length bug.
3073 * Add tcpdump to repository.
3075 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3076 * release version v0_4_3
3077 * Fixes statfs for inodes on extN.
3078 * Fixes bug in runtests which would delete /etc/hosts.
3079 * Use 64-bit object IDs wherever possible (not into VFS though)
3080 Remove ost_get_info, which is unused by lustre, and out of date.
3082 2002-07-03 Peter Braam <braam@clusterfs.com>
3083 * release version v0_4_2 Fixes a lookup error (type not passed)
3084 * move forward to head of Portals
3085 * move forward to latest Lustre kernel
3087 2002-06-25 Peter Braam <braam@clusterfs.com>
3088 * release version v0_4_1. Hopefully stable on single node use.