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
186 Severity : enhancement
188 Description: plain/inodebits lock performance improvement
189 Details : Grouping plain/inodebits in granted list by their request modes
190 and bits policy, thus improving the performance of search through
193 ------------------------------------------------------------------------------
195 TBD Cluster File Systems, Inc. <info@clusterfs.com>
197 * Support for kernels:
198 2.6.9-42.0.3EL (RHEL 4)
200 2.4.21-47.0.1.EL (RHEL 3)
201 2.6.12.6 vanilla (kernel.org)
202 * Recommended e2fsprogs version: 1.39.cfs2-0
207 Description: make O_SYNC working on 2.6 kernels
208 Details : 2.6 kernels use different method for mark pages for write,
209 so need add a code to lustre for O_SYNC work.
214 Description: Failure to close file and release space on NFS
215 Details : Put inode details into lock acquired in ll_intent_file_open.
216 Use mdc_intent_lock in ll_intent_open to properly
217 detect all kind of errors unhandled by mdc_enqueue
222 Description: proc file read during shutdown sometimes raced obd removal,
224 Details : Add lock to prevent obd access after proc file removal.
227 Frequency : Only for files larger than 4GB on 32-bit clients.
229 Description: improperly doing page alignment of locks
230 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
231 PAGE_*. Make CFS_PAGE_MASK 64bit long.
236 Description: RPCs being resent when they shouldn't be
237 Details : Some RPCs that should not be resent are being resent. This
238 can cause inconsistencies in the RPC state machine. Do not
239 resend such requests.
241 ------------------------------------------------------------------------------
243 TBD Cluster File Systems, Inc. <info@clusterfs.com>
245 * Support for kernels:
246 2.6.9-42.0.3EL (RHEL 4)
248 2.4.21-40.0.1.EL (RHEL 3)
249 2.6.12.6 vanilla (kernel.org)
255 Description: "went back in time" messages on mds failover
256 Details : The greatest transno may be lost when the current operation
257 finishes with an error (transno==0) and the client's last_rcvd
258 record is over-written. Save the greatest transno in the
259 mds_last_transno for this case.
262 Frequency : always for specific kernels and striping counts
264 Description: client may get "Matching packet too big" without ACL support
265 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
266 when trying to access files in certain configurations. The
267 clients should in fact be denied when mounting because they do
271 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
273 Description: many clients connecting with IO in progress causes connect timeouts
274 Details : Avoid synchronous journal commits to avoid delays caused by many
275 clients connecting/disconnecting when bulk IO is in progress.
276 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
277 OST_IO_PORTAL to avoid delays behind potentially many pending
281 Frequency : occasionally with multiple writers to a single file
283 Description: shared writes to file may result in wrong size reported by stat()
284 Details : Allow growing of kms when extent lock is cancelled
287 Frequency : always with random mmap IO to multi-striped file
289 Description: mmap write might be lost if we are writing to a 'hole' in stripe
290 Details : Only if the hole is at the end of OST object so that kms is too
291 small. Fix is to increase kms accordingly in ll_nopage.
294 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
296 Description: writes to a missing object would leak memory on the OST
297 Details : If there is an inconsistency between the MDS and OST filesystems,
298 such that the MDS references an object that doesn't exist, writes
299 to that object will leak memory due to incorrect cleanup in the
300 error handling path, eventually running out of memory on the OST.
305 Description: Creating too long symlink causes lustre errors
306 Details : Check symlink and name lengths before sending requests to MDS.
309 Frequency : only if flock is enabled (not on by default)
311 Description: posix locks not released on fd closure on 2.6.9+
312 Details : We failed to add posix locks to list of inode locks on 2.6.9+
313 kernels, this caused such locks not to be released on fd close and
314 then assertions on fs unmount about still used locks.
317 Frequency : MDS failover only, very rarely
319 Description: clients may get ASSERTION(granted_lock != NULL)
320 Details : When request was taking a long time, and a client was resending
321 a getattr by name lock request. The were multiple lock
322 requests with the same client lock handle and
323 mds_getattr_name->fixup_handle_for_resent_request found one
324 of the lock handles but later failed with
325 ASSERTION(granted_lock != NULL).
330 Description: handle->h_buffer_credits > 0, assertion failure
331 Details : h_buffer_credits is zero after truncate, causing assertion
332 failure. This patch extends the transaction or creates a new
336 Frequency : NFS re-export or patchless client
337 Bugzilla : 11179, 10796
338 Description: Crash on NFS re-export node (__d_move)
339 Details : We do not want to hash the dentry if we don't have a lock.
340 But if this dentry is later used in d_move, we'd hit uninitialised
341 list head d_hash, so we just do this to init d_hash field but
342 leave dentry unhashed.
345 Frequency : NFS re-export or patchless client
347 Description: NFS exports has problem with symbolic link
348 Details : lustre client didn't properly install dentry when re-exported
349 to NFS or running patchless client.
352 Frequency : NFS re-export or patchless client
354 Description: Various nfs/patchless fixes.
355 Details : fixes reuse disconected alias for lookup process - this fixes
356 warning "find_exported_dentry: npd != pd", fix permission
357 error with open files at nfs.
360 Frequency : occasional
362 Description: Crash on NFS re-export node
363 Details : call clear_page on wrong pointer triggered oops in
364 generic_mapping_read().
367 Frequency : rarely, using O_DIRECT IO
369 Description: unaligned directio crashes client with LASSERT
370 Details : check for unaligned buffers before trying any requests.
373 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
375 Description: stale data returned from RAID cache
376 Details : If only a small amount of IO is done to the RAID device before
377 reading it again it is possible to get stale data from the RAID
378 cache instead of reading it from disk.
381 Frequency : depends on arch, kernel and compiler version, always on sles10
384 Description: recursive or deep enough symlinks cause stack overflow
385 Details : getting rid of large stack-allocated variable in
389 Frequency : depends on hardware
391 Description: lustre write performance loss in the SLES10 kernel
392 Details : the performance loss is caused by using of write barriers in the
393 ext3 code. The SLES10 kernel turns barrier support on by
394 default. The fix is to undo that change for ldiskfs.
396 ------------------------------------------------------------------------------
398 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
400 * Support for kernels:
401 2.6.9-42.0.3EL (RHEL 4)
403 2.4.21-47.0.1.EL (RHEL 3)
404 2.6.12.6 vanilla (kernel.org)
408 Frequency : quota enabled and large files being deleted
410 Description: releasing more than 4GB of quota at once hangs OST
411 Details : If a user deletes more than 4GB of files on a single OST it
412 will cause the OST to spin in an infinite loop. Release
413 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
418 Description: statfs data retrieved from /proc may be stale or zero
419 Details : When reading per-device statfs data from /proc, in the
420 {kbytes,files}_{total,free,avail} files, it may appear
421 as zero or be out of date.
424 Frequency : always, for aggregate stripe size over 4GB
426 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
427 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
428 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
431 Frequency : always if "lfs find" used on a local file/directory
433 Description: "lfs find" segfaults if used on a local file/directory
434 Details : The case where a directory component was not specified wasn't
435 handled correctly. Handle this properly.
438 Frequency : always on ppc64
440 Description: the write to an ext3 filesystem mounted with mballoc got stuck
441 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
442 perform endianness conversion.
445 Frequency : rarely (truncate to non-zero file size after write under load)
446 Bugzilla : 10730, 10687
447 Description: Files padded with zeros to next 4K multiple
448 Details : With filesystems mounted using the "extents" option (2.6 kernels)
449 it is possible that files that are truncated to a non-zero size
450 immediately after being written are filled with zero bytes beyond
451 the truncated size. No file data is lost.
453 Severity : enhancement
454 Frequency : liblustre only
456 Description: Allow recovery/failover for liblustre clients.
457 Details : liblustre clients were unaware of failover configurations until
460 Severity : enhancement
462 Description: user file locks should fail when not mounting with flock option
463 Details : Set up an error-returning stub in ll_file_operations.lock field
464 to prevent incorrect behaviour when client is mounted without
465 flock option. Also, set up properly f_op->flock field for
469 Frequency : always on ia64
471 Description: "lfs df" loops on printing out MDS statfs information
472 Details : The obd_ioctl_data was not initialized and in some systems
473 this caused a failure during the ioctl that did not return
474 an error. Initialize the struct and return an error on failure.
477 Frequency : SLES 9 only
479 Description: Error of copying files with lustre special EAs as root
480 Details : Client side always return success for setxattr call for lustre
481 special xattr (currently only "trusted.lov").
484 Frequency : rarely on clusters with both ia64+i386 clients
486 Description: ia64+i686 clients doing shared IO on the same file may LBUG
487 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
488 clients are doing concurrent writes to the same file it is
489 possible that the ia64 clients may LASSERT because the OST
490 extent locks are not PAGE_SIZE aligned. Ensure that grown
491 locks are always aligned on the request boundary.
494 Frequency : specific use, occasional
496 Description: Overwriting in use executable truncates on-disk binary image
497 Details : If one node attempts to overwrite an executable in use by
498 another node, we now correctly return ETXTBSY instead of
504 Description: chmod on Lustre root is propagated to other clients
505 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
506 invalid by the follow_mount time.
511 Description: Race in 'instant cancel' lock handling could lead to such locks
512 never to be granted in case of SMP MDS
513 Details : Do not destroy not yet granted but cbpending locks in
517 Frequency : replay/resend of open
519 Description: non null lock assetion failure in mds_intent_policy
520 Details : Trying to replay/resend lockless open requests resulted in
521 mds_open() returning 0 with no lock. Now it sets a flag if
522 a lock is going to be returned.
524 Severity : enhancement
526 Description: Checksum enhancements
527 Details : New checksum enhancements allow for resending RPCs that failed
530 Severity : enhancement
532 Description: Tunables on number of dirty pages in cacche
533 Details : Allow to set limit on number of dirty pages cached.
538 Description: client crash on unmount - lock still has references
539 Details : In some error handling cases it was possible to leak a lock
540 reference on a client while accessing a file. This was not
541 harmful to the client during operation, but would cause the
542 client to crash when the filesystem is unmounted.
545 Frequency : specific case, rare
547 Description: ETXTBSY on mds though file not in use by client
548 Details : ETXTBSY is no longer incorrectly returned when attempting to
549 chmod or chown a directory that the user previously tried to
550 execute or a currently-executing binary.
553 Frequency : extremely rare except on liblustre-based clients
555 Description: Lustre space not freed when files are deleted
556 Details : Clean up open-unlinked files after client eviction. Previously
557 the unlink was skipped and the files remained as orphans.
562 Description: OST failure "would be an LBUG" in waiting_locks_callback()
563 Details : In some cases it was possible to send a blocking callback to
564 a client doing a glimpse, even though that client didn't get
565 a lock granted. When the glimpse lock is cancelled on the OST
566 the freed lock is left on the waiting list and corrupted the list.
569 Frequency : all core dumps
571 Description: Broke core dumps to lustre
572 Details : Negative dentry may be unhashed if parent does not have UPDATE
573 lock, but some callers, e.g. do_coredump, expect dentry to be
574 hashed after successful create, hash it in ll_create_it.
576 ------------------------------------------------------------------------------
578 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
580 * Support for kernels:
581 2.6.9-42.0.2.EL (RHEL 4)
583 2.4.21-40.EL (RHEL 3)
584 2.6.12.6 vanilla (kernel.org)
588 Frequency : always on RHEL 3
590 Description: Number of open files grows over time
591 Details : The number of open files grows over time, whether or not
592 Lustre is started. This was due to a filp leak introduced
593 by one of our kernel patches.
595 ------------------------------------------------------------------------------
597 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
599 * Support for kernels:
602 2.4.21-40.EL (RHEL 3)
603 2.6.12.6 vanilla (kernel.org)
608 Bugzilla : 5719, 9635, 9792, 9684
609 Description: OST (or MDS) trips assertions in (re)connection under heavy load
610 Details : If a server is under heavy load and cannot reply to new
611 connection requests before the client resends the (re)connect,
612 the connection handling code can behave badly if two service
613 threads are concurrently handing separate (re)connections from
614 the same client. Add better locking to the connection handling
615 code, and ensure that only a single connection will be processed
616 for a given client UUID, even if the lock is dropped.
618 Severity : enhancement
620 Description: add TCP zero-copy support to kernel
621 Details : Add support to the kernel TCP stack to allow zero-copy bulk
622 sends if the hardware supports scatter-gather and checksumming.
623 This allows socklnd to do client-write and server-read more
624 efficiently and reduce CPU utilization from skbuf copying.
627 Frequency : only if NFS exporting from client
629 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
630 Details : If setting attributes on a file created under NFS that had
631 never been opened it would be possible to oops the client
632 if the file had no objects.
636 Bugzilla : 9326, 10402, 10897
637 Description: client crash in ptlrpcd_wake() thread when sending async RPC
638 Details : It is possible that ptlrpcd_wake() dereferences a freed async
639 RPC. In rare cases the ptlrpcd thread alread processed the RPC
640 before ptlrpcd_wake() was called and the request was freed.
643 Frequency : always for liblustre
645 Description: liblustre client does MDS+OSTs setattr RPC for each write
646 Details : When doing a write from a liblustre client, the client
647 incorrectly issued an RPC to the MDS and each OST the file was
648 striped over in order to update the timestamps. When writing
649 with small chunks and many clients this could overwhelm the MDS
650 with RPCs. In all cases it would slow down the write because
651 these RPCs are unnecessary.
653 Severity : enhancement
655 Description: allow number of MDS service threads to be changed at module load
656 Details : It is now possible to change the number of MDS service threads
657 running. Adding "options mds mds_num_threads={N}" to the MDS's
658 /etc/modprobe.conf will set the number of threads for the next
659 time Lustre is restarted (assuming the "mds" module is also
660 reloaded at that time). The default number of threads will
661 stay the same, 32 for most systems.
666 Description: OST crash if filesystem is unformatted or corrupt
667 Details : If an OST is started on a device that has never been formatted
668 or if the filesystem is corrupt and cannot even mount then the
669 error handling cleanup routines would dereference a NULL pointer.
674 Description: NULL pointer deref in llap_from_page.
675 Details : get_cache_page_nowait can return a page with NULL (or otherwise
676 incorrect) mapping if the page was truncated/reclaimed while it was
677 searched for. Check for this condition and skip such pages when
678 doing readahead. Introduce extra check to llap_from_page() to
679 verify page->mapping->host is non-NULL (so page is not anonymous).
682 Frequency : Sometimes when using sys_sendfile
684 Description: "page not covered by a lock" warnings from ll_readpage
685 Details : sendfile called ll_readpage without right page locks present.
686 Now we introduced ll_file_sendfile that does necessary locking
687 around call to generic_file_sendfile() much like we do in
691 Frequency : with certain MDS communication failures at client mount time
693 Description: NULL pointer deref after failed client mount
694 Details : a client connection request may delayed by the network layer
695 and not be sent until after the PTLRPC layer has timed out the
696 request. If the client fails the mount immediately it will try
697 to clean up before the network times out the request. Add a
698 reference from the request import to the obd device and delay
699 the cleanup until the network drops the request.
702 Frequency : occasionally during client (re)connect
704 Description: assertion failure during client (re)connect
705 Details : processing a client connection request may be delayed by the
706 client or server longer than the client connect timeout. This
707 causes the client to resend the connection request. If the
708 original connection request is replied in this interval, the
709 client may trip an assertion failure in ptlrpc_connect_interpret()
710 which thought it would be the only running connect process.
713 Frequency : only with obd_echo servers and clients that are rebooted
715 Description: kernel BUG accessing uninitialized data structure
716 Details : When running an obd_echo server it did not start the ping_evictor
717 thread, and when a client was evicted an uninitialized data
718 structure was accessed. Start the ping_evictor in the RPC
719 service startup instead of the OBD startup.
721 Severity : enhancement
722 Bugzilla : 10193 (patchless)
723 Description: Remove dependency on various unexported kernel interfaces.
724 Details : No longer need reparent_to_init, exit_mm, exit_files,
725 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
728 Frequency : rare (only users of deprecated and unsupported LDAP config)
730 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
731 Details : LDAP apparently contains 'lustreName' attributes instead of
732 'name'. A simple remapping of the name is sufficient.
735 Frequency : rare (only with non-default dump_on_timeout debug enabled)
737 Description: waiting_locks_callback trips kernel BUG if client is evicted
738 Details : Running with the dump_on_timeout debug flag turned on makes
739 it possible that the waiting_locks_callback() can try to dump
740 the Lustre kernel debug logs from an interrupt handler. Defer
741 this log dumping to the expired_lock_main() thread.
743 Severity : enhancement
745 Description: Support NFS exporting on 2.6 kernels.
746 Details : Implement non-rawops metadata methods for NFS server to use without
747 changing NFS server code.
750 Frequency : very rare (synthetic metadata workload only)
752 Description: two racing renames might cause an MDS thread to deadlock
753 Details : Running the "racer" program may cause one MDS thread to rename
754 a file from being the source of a rename to being the target of
755 a rename at exactly the same time that another thread is doing
756 so, and the second thread has already enqueued these locks after
757 doing a lookup of the target and is trying to relock them in
758 order. Ensure that we don't try to re-lock the same resource.
761 Frequency : only very large systems with liblustre clients
763 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
764 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
765 all OSTs in parallel. This allows the eviction of stale liblustre
766 clients to proceed much faster than if they were done in series,
767 and also offers similar improvements for other set_info RPCs.
772 Description: excessive CPU usage during initial read phase on client
773 Details : During the initial read phase on a client, it would agressively
774 retry readahead on the file, consuming too much CPU and impacting
775 performance (since 1.4.5.8). Improve the readahead algorithm
776 to avoid this, and also improve some other common cases (read
777 of small files in particular, where "small" is files smaller than
778 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
783 Description: MDS crash when receiving packet with unknown intent.
784 Details : Do not LBUG in unknown intent case, just return -EFAULT
786 Severity : enhancement
787 Bugzilla : 9293, 9385
788 Description: MDS RPCs are serialised on client. This is unnecessary for some.
789 Details : Do not serialize getattr (non-intent version) and statfs.
792 Frequency : occasional, when OST network is overloaded/intermittent
794 Description: client evicted by OST after bulk IO timeout
795 Details : If a client sends a bulk IO request (read or write) the OST
796 may evict the client if it is unresposive to its data GET/PUT
797 request. This is incorrect if the network is overloaded (takes
798 too long to transfer the RPC data) or dropped the OST GET/PUT
799 request. There is no need to evict the client at all, since
800 the pinger and/or lock callbacks will handle this, and the
801 client can restart the bulk request.
804 Frequency : Always when mmapping file with no objects
806 Description: client crashes when mmapping file with no objects
807 Details : Check that we actually have objects in a file before doing any
808 operations on objects in ll_vm_open, ll_vm_close and
814 Description: Request leak when working with deleted CWD
815 Details : Introduce advanced request refcount tracking for requests
816 referenced from lustre intent.
818 Severity : Enhancement
820 Description: Cache open file handles on client.
821 Details : MDS now will return special lock along with openhandle, if
822 requested and client is allowed to hold openhandle, even if unused,
823 until such a lock is revoked. Helps NFS a lot, since NFS is opening
824 closing files for every read/write openration.
826 Severity : Enhancement
828 Description: Cache open negative dentries on client when possible.
829 Details : Guard negative dentries with UPDATE lock on parent dir, drop
830 negative dentries on lock revocation.
835 Description: Remounting a client read-only wasn't possible with a zconf mount
836 Details : It wasn't possible to remount a client read-only with llmount.
838 Severity : enhancement
839 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
840 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
841 MPICH is included to add Lustre-specific ADIO interfaces.
842 This is based closely on the UFS ADIO layer and only differs
843 in file creation, in order to allow the OST striping to be set.
844 This is user-contributed code and not supported by CFS.
849 Description: extended inode attributes (immutable, append-only) work improperly
850 when 2.4 and 2.6 kernels are used on client/server or vice versa
851 Details : Introduce kernel-independent values for these flags.
853 Severity : enhancement
856 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
857 Details : Many of the /proc/ tunables can only be tuned at a megabyte
858 granularity. Now, Fractional MB granularity is be supported,
859 this is very useful for low memory system.
861 Severity : enhancement
863 Description: Getattr by fid
864 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
865 locks, avoids extra getattr rpc requests to MDS, allows '/' to
866 have locks and avoids getattr rpc requests for it on every stat.
869 Frequency : Always, for filesystems larger than 2TB
871 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
872 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
873 filesystems larger than 2TB to be mounted with mballoc enabled.
876 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
878 Description: filesystem corruption for non-standard kernels and very large OSTs
879 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
880 filesystems larger than 2TB is mounted then the kernel will
881 silently corrupt the start of the filesystem. CONFIG_LBD is
882 enabled for all CFS-supported kernels, but the possibility of
883 this happening with a modified kernel config exists.
885 Severity : enhancement
887 Description: add client O_DIRECT support for 2.6 kernels
888 Details : It is now possible to do O_DIRECT reads and writes to files
889 in the Lustre client mountpoint on 2.6 kernel clients.
891 Severity : enhancement
893 Description: parallel glimpse, setattr, statfs, punch, destroy requests
894 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
895 parallel, not waiting for response from every OST before sending
896 a rpc to the next OST.
901 Description: setattr vs write race when updating file timestamps
902 Details : Client processes that update a file timestamp into the past
903 right after writing to the file (e.g. tar) it is possible that
904 the updated file modification time can be reset to the current
905 time due to a race between processing the setattr and write RPC.
907 Severity : enhancement
909 Description: Bring 'lfs find' closer in line with regular Linux find.
910 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
911 -print0 options and obtains all the needed info through the lustre
914 Severity : enhancement
916 Description: support up to 1024 configured devices on one node
917 Details : change obd_dev array from statically allocated to dynamically
918 allocated structs as they are first used to reduce memory usage
923 Description: Flush dirty partially truncated pages during truncate
924 Details : Immediatelly flush partially truncated pages in filter_setattr,
925 this way we completely avoid having any pages in page cache on OST
926 and can retire ugly workarounds during writes to flush such pages.
931 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
932 Details : protect inode from truncation within vfs_unlink() context
933 just take a reference before calling vfs_unlink() and release it
934 when parent's i_sem is free.
937 Frequency : always, if extents are used on OSTs
939 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
940 Details : OSTs using the extents format would not zero the high 16 bits of
941 the index physical block number. This is not a problem for any
942 OST filesystems smaller than 16TB, and no kernels support ext3
943 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
944 new/modified files) and can be fixed for existing filesystems
945 with e2fsprogs-1.39-cfs1.
950 Description: import connection selection may be incorrect if timer wraps
951 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
952 import connections to be ignored if the 32-bit jiffies counter
953 wraps. Use a 64-bit jiffies counter.
956 Frequency : during server recovery
958 Description: crash after server is denying duplicate export
959 Details : If clients are resending connect requests to the server, the
960 server refuses to allow a client to connect multiple times.
961 Fixed a bug in the handling of this case.
964 Frequency : very large clusters immediately after boot
966 Description: LNET request buffers exhausted under heavy short-term load
967 Details : If a large number of client requests are generated on a service
968 that has previously never seen so many requests it is possible
969 that the request buffer growth cannot keep up with the spike in
970 demand. Instead of dropping incoming requests, they are held in
971 the LND until the RPC service can accept more requests.
974 Frequency : Sometimes during replay
976 Description: Assertion failure in ll_local_open after replay.
977 Details : If replay happened on an open request reply before we were able
978 to set replay handler, reply will become not swabbed tripping the
979 assertion in ll_local_open. Now we set the handler right after
980 recognising of open request
983 Frequency : very rare
985 Description: kernel reports "badness in vsnprintf"
986 Details : Reading from the "recovery_status" /proc file in small chunks
987 may cause a negative length in lprocfs_obd_rd_recovery_status()
988 call to vsnprintf() (which is otherwise harmless). Exit early
989 if there is no more space in the output buffer.
991 Severity : enhancement
993 Description: clear OBD RPC statistics by writing to them
994 Details : It is now possible to clear the OBD RPC statistics by writing
1000 Description: Client mtime is not the same on different clients after utimes
1001 Details : In some cases, the client was using the utimes() syscall on
1002 a file cached on another node. The clients now validate the
1003 ctime from the MDS + OSTs to determine which one is right.
1008 Description: Inability to activate failout mode
1009 Details : lconf script incorrectly assumed that in python string's numeric
1010 value is used in comparisons.
1013 Frequency : always with multiple stripes per file
1015 Description: Inefficient object allocation for mutli-stripe files
1016 Details : When selecting which OSTs to stripe files over, for files with
1017 a stripe count that divides evenly into the number of OSTs,
1018 the MDS is always picking the same starting OST for each file.
1019 Return the OST selection heuristic to the original design.
1024 Description: mount failures may take full timeout to return an error
1025 Details : Under some heavy load conditions it is possible that a
1026 failed mount can wait for the full obd_timeout interval,
1027 possibly several minutes, before reporting an error.
1028 Instead return an error as soon as the status is known.
1030 ------------------------------------------------------------------------------
1032 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
1034 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
1035 INTEROPERATE with older versions automatically. Please read the
1036 user documentation before upgrading any part of a live system.
1037 * WARNING: Lustre networking configuration changes are required with
1038 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
1041 * Support for kernels:
1042 2.6.9-22.0.2.EL (RHEL 4)
1043 2.6.5-7.244 (SLES 9)
1044 2.6.12.6 vanilla (kernel.org)
1047 Severity : enhancement
1048 Bugzilla : 7981/8208
1049 Description: Introduced Lustre Networking (LNET)
1050 Details : LNET is new networking infrastructure for Lustre, it includes
1051 a reorganized network configuration mode (see the user
1052 documentation for full details) as well as support for routing
1053 between different network fabrics. Lustre Networking Devices
1054 (LNDs) for the supported network fabrics have also been
1055 created for this new infrastructure.
1057 Severity : enhancement
1058 Description: Introduced Access control lists
1059 Details : clients can set ACLs on files and directories in order to have
1060 more fine-grained permissions than the standard Unix UGO+RWX.
1061 The MDS must be started with the "-o acl" mount option.
1063 Severity : enhancement
1064 Description: Introduced filesystem quotas
1065 Details : Administrators may now establish per-user quotas on the
1068 Severity : enhancement
1070 Description: Configuration change for the XT3
1071 The PTLLND is now used to run Lustre over Portals on the XT3
1072 The configure option(s) --with-cray-portals are no longer used.
1073 Rather --with-portals=<path-to-portals-includes> is used to
1074 enable building on the XT3. In addition to enable XT3 specific
1075 features the option --enable-cray-xt3 must be used.
1080 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
1081 Details : A race between allocating a new llcd and re-getting the llcd_lock
1082 allowed another thread to grab newly-allocated llcd.
1084 Severity : enhancement
1086 Description: 2.6 OST async journal commit and locking fix to improve performance
1087 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
1088 2.6 kernels are now async as they already were in 2.4 kernels so
1089 that they can commit concurrently with the network bulk transfer.
1090 For block-allocated files the filter allocation semaphore is held
1091 to avoid filesystem fragmentation during allocation. BKL lock
1092 removed for 2.6 xattr operations where it is no longer needed.
1097 Description: lconf incorrectly determined whether two IP networks could talk
1098 Details : In some more complicated routing and multiple-network
1099 configurations, lconf will avoid trying to make a network
1100 connection to a disjoint part of the IP space. It was doing the
1101 math incorrectly for one set of cases.
1106 Description: Fix for potential infinite loop processing records in an llog.
1107 Details : If an llog record is corrupted/zeroed, it is possible to loop
1108 forever in llog_process(). Validate the llog record length
1109 and skip the remainder of the block on error.
1112 Frequency : occasional (liblustre only)
1114 Description: liblustre could not open files whose last component is a symlink
1115 Details : sysio_path_walk() would incorrectly pass the open intent to
1116 intermediate path components.
1119 Frequency : rare (liblustre only with non-standard tuning)
1120 Bugzilla : 7201 (7350)
1121 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
1122 Details : llu_lookup_finish_locks() tries to set lock data on a lock
1123 after it has been released, only do this for referenced locks
1125 Severity : enhancement
1127 Description: specifying an (invalid) directory default stripe_size of -1
1128 would reset the directory default striping
1129 Details : stripe_size -1 was used internally to signal directory stripe
1130 removal, now use "all default" to signal dir stripe removal
1131 as a directory striping of "all default" is not useful
1134 Frequency : common for large clusters running liblustre clients
1136 Description: doing an ls when liblustre clients are running is slow
1137 Details : sending a glimpse AST to a liblustre client waits for every AST
1138 to time out, as liblustre clients will not respond. Since they
1139 cannot cache data we refresh the OST lock LVB from disk instead.
1141 Severity : enhancement
1143 Description: doing an ls at the same time as file IO can be slow
1144 Details : enqueue and other "small" requests can be blocked behind many
1145 large IO requests. Create a new OST IO portal for non-IO
1146 requests so they can be processed faster.
1149 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
1151 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
1152 created file with mode 000
1153 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
1154 allow an ftruncate() as a non-root user to a file with mode 000.
1155 Lustre now respects this flag to disable mode checks when
1156 truncating a file owned by the user
1159 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
1162 Description: Revoking locks from clients that went dead or catatonic might take
1164 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
1165 cancellation of such locks instant on servers without waiting for
1166 any reply from clients. Clients drops these locks when cancel
1167 notification from server is received without replying.
1170 Frequency : liblustre-only, when liblustre client dies or becomes busy
1172 Description: Doing ls on Linux clients can take a long time with active
1174 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
1175 granting such locks to it in the first place if possible. Locks
1176 are taken by proxy on the OST during the read or write and
1177 dropped immediately afterward. Add connect flags handling, do
1178 not grant locks to liblustre clients for glimpse ASTs.
1180 Severity : enhancement
1182 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
1183 Details : Existing read-ahead algorithm is tuned for the case of streamlined
1184 sequential reads and behaves badly with applications doing random
1185 reads. Improve it by reading ahead at least read region, and
1186 avoiding excessive large RPC for small reads.
1188 Severity : enhancement
1190 Description: Creating more than 1000 files for a single job may cause a load
1191 imbalance on the OSTs if there are also a large number of OSTs.
1192 Details : qos_prep_create() uses an OST index reseed value that is an
1193 even multiple of the number of available OSTs so that if the
1194 reseed happens in the middle of the object allocation it will
1195 still utilize the OSTs as uniformly as possible.
1200 Description: OST or MDS may oops in ping_evictor_main()
1201 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
1202 but doesn't restart at beginning of obd_exports_timed list
1205 Severity : enhancement
1207 Description: improve by-nid export eviction on the MDS and OST
1208 Details : allow multiple exports with the same NID to be evicted at one
1209 time without re-searching the exports list.
1212 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
1214 Description: MDS may oops in groups_free()
1215 Details : in rare race conditions a newly allocated group_info struct is
1216 freed again, and this can be NULL. The 2.4 compatibility code
1217 for groups_free() checked for a NULL pointer, but 2.6 did not.
1220 Frequency : common for liblustre clients doing little filesystem IO
1221 Bugzilla : 9352, 7313
1222 Description: server may evict liblustre clients accessing contended locks
1223 Details : if a client is granted a lock or receives a completion AST
1224 with a blocking AST already set it would not reply to the AST
1225 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
1228 Frequency : lfs setstripe, only systems with more than 160 OSTs
1230 Description: unable to set striping with a starting offset beyond OST 160
1231 Details : llapi_create_file() incorrectly limited the starting stripe
1232 index to the maximum single-file stripe count.
1235 Frequency : LDAP users only
1237 Description: lconf did not handle in-kernel recovery with LDAP properly
1238 Details : lconf/LustreDB get_refs() is searching the wrong namespace
1240 Severity : enhancement
1242 Description: bind OST threads to NUMA nodes to improve performance
1243 Details : all OST threads are uniformly bound to CPUs on a single NUMA
1244 node and do their allocations there to localize memory access
1246 Severity : enhancement
1248 Description: llmount can determine client NID directly from Myrinet (GM)
1249 Details : the client NID code from gmnalnid was moved directly into
1250 llmount, removing the need to use this or specifying the
1251 client NID explicitly when mounting GM clients with zeroconf
1254 Frequency : if client is started with down MDS
1256 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
1257 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
1258 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
1259 ptlrpc_import_delayed_req() because we didn't check if the
1260 request was interrupted, and we also didn't break out of the
1261 event loop if there was no timeout
1266 Description: data loss during non-page-aligned writes to a single file from
1267 both multiple nodes and multiple threads on one node at same time
1268 Details : updates to KMS and lsm weren't protected by common lock. Resulting
1269 inconsistency led to false short-reads, that were cached and later
1270 used by ->prepare_write() to fill in partially written page,
1271 leading to data loss.
1274 Frequency : always, if lconf --abort_recovery used
1276 Description: lconf --abort_recovery fails with 'Operation not supported'
1277 Details : lconf was attempting to abort recovery on the MDT device and not
1280 Severity : enhancement
1282 Description: remove cleanup logs
1283 Details : replace lconf-generated cleanup logs with lustre internal
1284 cleanup routines. Eliminates the need for client-cleanup and
1287 Severity : enhancement
1289 Description: add support for EAs (user and system) on lustre filesystems
1290 Details : it is now possible to store extended attributes in the Lustre
1291 client filesystem, and with the user_xattr mount option it
1292 is possible to allow users to store EAs on their files also
1294 Severity : enhancement
1296 Description: Add possibility (config option) to show minimal available OST free
1298 Details : When compiled with --enable-mindf configure option, statfs(2)
1299 (and so, df) will return least minimal free space available from
1300 all OSTs as amount of free space on FS, instead of summary of
1301 free spaces of all OSTs.
1303 Severity : enhancement
1305 Description: do not expand extent locks acquired on OST-side
1306 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
1307 by server: they are not cached anyway.
1310 Frequency : when mmap is used/binaries executed from Lustre
1312 Description: Unmmap pages before throwing them away from read cache.
1313 Details : llap_shrink cache now attempts to unmap pages before discarding
1314 them (if unmapping failed - do not discard). SLES9 kernel has
1315 extra checks that trigger if this unmapping is not done first.
1320 Description: lconf didn't resolve symlinks before checking to see whether a
1321 given mountpoint was already in use
1324 Frequency : when migrating failover services
1325 Bugzilla : 6395, 9514
1326 Description: When migrating a subset of services from a node (e.g. failback
1327 from a failover service node) the remaining services would
1328 time out and evict clients.
1329 Details : lconf --force (implied by --failover) sets the global obd_timeout
1330 to 5 seconds in order to quickly disconnect, but this caused
1331 other RPCs to time out too quickly. Do not change the global
1332 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
1334 Severity : enhancement
1335 Frequency : if MDS is started with down OST
1336 Bugzilla : 9439,5706
1337 Description: Allow startup/shutdown of an MDS without depending on the
1338 availability of the OSTs.
1339 Details : Asynchronously call mds_lov_synchronize during MDS startup.
1340 Add appropriate locking and lov-osc refcounts for safe
1341 cleaning. Add osc abort_inflight calls in case the
1342 synchronize never started.
1345 Frequency : occasional (Cray XT3 only)
1347 Description: root not authorized to access files in CRAY_PORTALS environment
1348 Details : The client process capabilities were not honoured on the MDS in
1349 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
1350 been accessed by an authorized user then root was able to access
1351 the file on the local client also. The root user capabilities
1352 are now allowed on the MDS, as this environment has secure UID.
1355 Frequency : occasional
1357 Description: ldiskfs "too long searching" message happens too often
1358 Details : A debugging message (otherwise harmless) prints too often on
1359 the OST console. This has been reduced to only happen when
1360 there are fragmentation problems on the filesystem.
1365 Description: Division by zero in statfs when all OSCs are inactive
1366 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
1367 lov_statfs divides by value returned by lov_get_stripecnt().
1371 Bugzilla : 9489, 3273
1372 Description: First write from each client to each OST was only 4kB in size,
1373 to initialize client writeback cache, which caused sub-optimal
1374 RPCs and poor layout on disk for the first writen file.
1375 Details : Clients now request an initial cache grant at (re)connect time
1376 and so that they can start streaming writes to the cache right
1377 away and always do full-sized RPCs if there is enough data.
1378 If the OST is rebooted the client also re-establishes its grant
1379 so that client cached writes will be honoured under the grant.
1384 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
1385 Details : Now I/O RPCs go to different portal number and (presumably) fast
1386 lock requests (and glimses) and other RPCs get their own service
1387 threads pool that should be able to service those RPCs
1390 Severity : enhancement
1392 Description: Ability to exchange lustre version between client and servers and
1393 issue warnings at client side if client is too old. Also for
1394 liblustre clients there is ability to refuse connection of too old
1396 Details : New 'version' field is added to connect data structure that is
1397 filled with version info. That info is later checked by server and
1401 Frequency : rare, liblustre only.
1402 Bugzilla : 9296, 9581
1403 Description: Two simultaneous writes from liblustre at offset within same page
1404 might proceed at the same time overwriting eachother with stale
1406 Details : I/O lock withing llu_file_prwv was released too early, before data
1407 actually was hitting the wire. Extended lock-holding time until
1408 server acknowledges receiving data.
1411 Frequency : extremely rare. Never observed in practice.
1413 Description: avoid generating lustre_handle cookie of 0.
1414 Details : class_handle_hash() generates handle cookies by incrementing
1415 global counter, and can hit 0 occasionaly (this is unlikely, but
1416 not impossible, because initial value of cookie counter is
1417 selected randonly). Value of 0 is used as a sentinel meaning
1418 "unassigned handle" --- avoid it. Also coalesce two critical
1419 sections in this function into one.
1421 Severity : enhancement
1423 Description: allow liblustre clients to delegate truncate locking to OST
1424 Details : To avoid overhead of locking, liblustre client instructs OST to
1425 take extent lock in ost_punch() on client's behalf. New connection
1426 flag is added to handle backward compatibility.
1428 Severity : enhancement
1429 Bugzilla : 4928, 7341, 9758
1430 Description: allow number of OST service threads to be specified
1431 Details : a module parameter allows the number of OST service threads
1432 to be specified via "options ost ost_num_threads={N}" in the
1433 OSS's /etc/modules.conf or /etc/modprobe.conf.
1437 Bugzilla : 6146, 9635, 9895
1438 Description: servers crash with bad pointer in target_handle_connect()
1439 Details : In rare cases when a client is reconnecting it was possible that
1440 the connection request was the last reference for that export.
1441 We would temporarily drop the export reference and get a new
1442 one, but this may have been the last reference and the export
1443 was just destroyed. Get new reference before dropping old one.
1445 Severity : enhancement
1446 Frequency : if client is started with failover MDS
1448 Description: Allow multiple MDS hostnames in the mount command
1449 Details : Try to read the configuration from all specified MDS
1450 hostnames during a client mount in case the "primary"
1453 Severity : enhancement
1455 Description: Stop sending data to evicted clients as soon as possible.
1456 Details : Check if the client we are about to send or are sending data to
1457 was evicted already. (Check is done every second of waiting,
1458 for which l_wait_event interface was extended to allow checking
1459 of exit condition at specified intervals).
1462 Frequency : rare, normally only when NFS exporting is done from client
1464 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
1466 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
1469 Frequency : always, if srand() is called before liblustre initialization
1471 Description: Liblustre uses system PRNG disturbing its usage by user application
1472 Details : Introduce internal to lustre fast and high-quality PRNG for
1473 lustre usage and make liblustre and some other places in generic
1474 lustre code to use it.
1476 Severity : enhancement
1477 Bugzilla : 9477, 9557, 9870
1478 Description: Verify that the MDS configuration logs are updated when xml is
1479 Details : Check if the .xml configuration logs are newer than the config
1480 logs stored on the MDS and report an error if this is the case.
1481 Request --write-conf, or allow starting with --old_conf.
1483 Severity : enhancement
1485 Description: Handle symlinks in the path when checking if Lustre is mounted.
1486 Details : Resolve intermediate symlinks when checking if a client has
1487 mounted a filesystem to avoid duplicate client mounts.
1492 Description: lconf can hit an error exception but still return success.
1493 Details : The lconf command catches the Command error exception at the top
1494 level script context and will exit with the associated exit
1495 status, but doesn't ensure that this exit status is non-zero.
1500 Description: failure of ptlrpc thread startup can cause oops
1501 Details : Starting a ptlrpc service thread can fail if there are a large
1502 number of threads or the server memory is very fragmented.
1503 Handle this without oopsing.
1506 Frequency : always, only if liblustre and non-default acceptor port was used
1508 Description: liblustre cannot connect to servers with non-default acceptor port
1509 Details : tcpnal_set_default_params() was not called and was therefore
1510 ignoring the environment varaible TCPNAL_PORT, as well as other
1511 TCPNAL_ environment variables
1516 Description: two objects could be created on the same OST for a single file
1517 Details : If an OST is down, in some cases it was possible to create two
1518 objects on a single OST for a single file. No problems other
1519 than potential performance impact and spurious error messages.
1523 Bugzilla : 5681, 9562
1524 Description: Client may oops in ll_unhash_aliases
1525 Details : Client dcache may become inconsistent in race condition.
1526 In some cases "getcwd" can fail if the current directory is
1532 Description: Inode refcounting problems in NFS export code
1533 Details : link_raw functions used to call d_instantiate without obtaining
1534 extra inode reference first.
1538 Bugzilla : 9942, 9903
1539 Description: Referencing freed requests leading to crash, memleaks with NFS.
1540 Details : We used to require that call to ll_revalidate_it was always
1541 followed by ll_lookup_it. Also with revalidate_special() it is
1542 possible to call ll_revalidate_it() twice for the same dentry
1543 even if first occurence returned success. This fix changes semantic
1544 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
1545 reference on a request referred from the intent.
1546 ll_intent_release() then releases such a request.
1549 Frequency : rare, normally benchmark loads only
1551 Description: unlinked inodes were kept in memory on the client
1552 Details : If a client is repeatedly creating and unlinking files it
1553 can accumulate a lot of stale inodes in the inode slab cache.
1554 If there is no other client load running this can cause the
1555 client node to run out of memory. Instead flush old inodes
1556 from client cache that have the same inode number as a new inode.
1559 Frequency : SLES9 2.6.5 kernel and long filenames only
1560 Bugzilla : 9969, 10379
1561 Description: utime reports stale NFS file handle
1562 Details : SLES9 uses out-of-dentry names in some cases, which confused
1563 the lustre dentry revalidation. Change it to always use the
1567 Frequency : rare, unless heavy write-truncate concurrency is continuous
1568 Bugzilla : 4180, 6984, 7171, 9963, 9331
1569 Description: OST becomes very slow and/or deadlocked during object unlink
1570 Details : filter_destroy() was holding onto the parent directory lock
1571 while truncating+unlinking objects. For very large objects this
1572 may block other threads for a long time and slow overall OST
1573 responsiveness. It may also be possible to get a lock ordering
1574 deadlock in this case, or run out of journal credits because of
1575 the combined truncate+unlink. Solution is to do object truncate
1576 first in one transaction without parent lock, and then do the
1577 final unlink in a new transaction with the parent lock. This
1578 reduces the lock hold time dramatically.
1581 Frequency : rare, 2.4 kernels only
1583 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
1584 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
1585 before unmounting their filesystem in order to avoid deadlock.
1586 The kernel_locked() function in 2.4 kernels only checks whether
1587 the kernel lock is held, not whether it is this process that is
1588 holding it as 2.6 kernels do.
1593 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
1594 Details : The client ptlrpc code may be trying to reconnect to a down
1595 server before a previous connection attempt has timed out.
1596 Increase the reconnect interval to be longer than the connection
1597 timeout interval to avoid sending duplicate connections to
1601 Frequency : echo_client brw_test command
1603 Description: fix echo_client to work with OST preallocated code
1604 Details : OST preallocation code (5137) didn't take echo_client IO path
1605 into account: echo_client calls filter methods outside of any
1606 OST thread and, hence, there is no per-thread preallocated
1607 pages and buffers to use. Solution: hijack pga pages for IO. As
1608 a byproduct, this avoids unnecessary data copying.
1612 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
1613 Description: Client can oops in mdc_commit_close() after open replay
1614 Details : It was possible for the MDS to return an open request with no
1615 transaction number in mds_finish_transno() if the client was
1616 evicted, but without actually returning an error. Clients
1617 would later try to replay that open and may trip an assertion
1618 Simplify the client close codepath, and always return an error
1619 from the MDS in case the open is not successful.
1622 Frequency : rare, 2.6 OSTs only
1624 Description: OST may deadlock under high load on fragmented files
1625 Details : If there was a heavy load and highly-fragmented OST filesystems
1626 it was possible to have all the OST threads deadlock waiting on
1627 allocation of biovecs, because the biovecs were not released
1628 until the entire RPC IO was completed. Instead, release biovecs
1629 as soon as they are complete to ensure forward IO progress.
1631 Severity : enhancement
1633 Description: Support for specifying external journal device at mount
1634 Details : If an OST or MDS device is formatted with an external journal
1635 device, this device major/minor is stored in the ext3 superblock
1636 and may not be valid for failover. Allow detecting and
1637 specifying the external journal at mount time.
1642 Description: Mounting an MDS with pending unlinked files may cause oops
1643 Details : target_finish_recovery() calls mds_postrecov() which returned
1644 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
1645 considers this an error and immediately begins cleaning up the
1646 lov, just after starting the mds_lov process
1648 Severity : enhancement
1650 Description: Implement 'lfs df' to report actual free space on per-OST basis
1651 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
1652 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
1653 usage of objects; '-h' to report in human readable format.
1655 ------------------------------------------------------------------------------
1657 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
1664 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
1665 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
1666 at startup during orphan unlinking. Instead, skip buddy update
1667 before setup, as it will be regenerated after recovery is complete.
1672 Description: If an OST is inactive, its locks might reference stale inodes.
1673 Details : lov_change_cbdata() must iterate over all namespaces, even if
1674 they are inactive to clear inode references from the lock.
1676 Severity : enhancement
1677 Frequency : occasional, if non-standard max_dirty_mb used
1679 Description: Client will block write RPCs if not enough grant
1680 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
1681 then the client will block writes while waiting for another RPC
1682 to complete instead of consuming its dirty limit. With change
1683 we get improved performance when max_dirty_mb is small.
1685 Severity : enhancement
1686 Bugzilla : 3389, 6253
1687 Description: Add support for supplementary groups on the MDS.
1688 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
1689 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
1690 lookups for user supplementary groups into a cache.
1694 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
1695 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
1696 for liblustre O_EXCL is enforced by MDS.
1700 Description: readdir never returns NULL in liblustre.
1701 Details : Corrected llu_iop_getdirentries logic, to return offset of next
1702 dentry in struct dirent.
1706 Frequency : liblustre only, depends on application IO pattern
1707 Description: liblustre clients evicted if not contacting servers
1708 Details : Don't put liblustre clients into the ping_evictor list, so
1709 they will not be evicted by the pinger ever.
1711 Severity : enhancement
1713 Description: Add ability to evict clients by NID from MDS.
1714 Details : By echoing "nid:$NID" string into
1715 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
1716 $NID would be instantly evicted from this MDS and from all active
1717 OSTs connected to it.
1721 Description: Do not query file size twice, somewhat slowing stat(2) calls.
1722 Details : lookup_it_finish() used to query file size from OSTs that was not
1727 Description: service threads change working directory to that of init
1728 Details : Starting lustre service threads may pin the working directory
1729 of the parent thread, making that filesystem busy. Threads
1730 now change to the working directory of init to avoid this.
1734 Frequency : during shutdown only
1735 Description: shutdown with a failed MDS or OST can cause unmount to hang
1736 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
1741 Frequency : 2.6 only
1742 Description: chmod/chown may include an extra supplementary group
1743 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
1744 supplementary group and if none is specified this is used.
1747 Bugzilla : 5479 (6816)
1749 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
1750 Details : If lookup is in progress on a file that is unlinked we might try
1751 to revalidate the inode and fail in revalidate after lookup is
1752 complete and ll_file_open() enqueues the open again but
1753 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
1756 Frequency : always, if lconf --abort_recovery used
1758 Description: lconf --abort_recovery fails with 'Operation not supported'
1759 Details : lconf was attempting to abort recovery on the MDT device and not
1762 ------------------------------------------------------------------------------
1764 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
1769 Frequency : rare (only unsupported configurations with a node running as an
1771 Bugzilla : 6514, 5137
1772 Description: Mounting a Lustre file system on a node running as an OST could
1774 Details : OSTs now preallocates memory needed to write out data at
1775 startup, instead of when needed, to avoid having to
1776 allocate memory in possibly low memory situations.
1777 Specifically, if the file system is mounted on on OST,
1778 memory pressure could force it to try to write out data,
1779 which it needed to allocate memory to do. Due to the low
1780 memory, it would be unable to do so and the node would
1781 become unresponsive.
1783 Severity : enhancement
1785 Description: Addition of lconf --service command line option
1786 Details : lconf now accepts a '--service <arg>' option, which is
1787 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
1789 Severity : enhancement
1791 Description: Failover mode is now the default for OSTs.
1792 Details : By default, OSTs will now run in failover mode. To return to
1793 the old behaviour, add '--failout' to the lmc line for OSTs.
1795 Severity : enhancement
1797 Description: Health checks are now provided for MDS and OSTs
1798 Details : Additional detailed health check information on MSD and OSTs
1799 is now provided through the procfs health_check value.
1802 Frequency : occasional, depends on IO load
1804 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
1805 after numerous create/delete cycles
1806 Details : The ext3 inode allocation policy would not allocate new inodes
1807 very well on the OSTs because there are no new directories
1808 being created. Instead we look for groups with free space if
1809 the parent directories are nearly full.
1814 Description: Network or server problems during mount may cause partially
1815 mounted clients instead of returning an error.
1816 Details : The config llog parsing code may overwrite the error return
1817 code during mount error handling, returning success instead
1823 Description: MDS can fail to allocate large reply buffers
1824 Details : After long uptimes the MDS can fail to allocate large reply
1825 buffers (e.g. zconf client mount config records) due to memory
1826 fragmentation or consumption by the buffer cache. Preallocate
1827 some large reply buffers so that these replies can be sent even
1828 under memory pressure.
1832 Frequency : rare (liblustre)
1833 Description: fsx running with liblustre complained that using truncate() to
1834 extend the file doesn't work. This patch corrects that issue.
1835 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
1836 ATTR_SIZE and lsm use in llu_setattr_raw.
1840 Frequency : rare, only 2.6 kernels
1841 Description: Unusual file access patterns on the MDS may result in inode
1842 data being lost in very rare circumstances.
1843 Details : Bad interaction between the ea-in-inode patch and the "no-read"
1844 code in the 2.6 kernel caused the inode and/or EA data not to
1845 be read from disk, causing single-file corruption.
1849 Frequency : rare, only 2.6 filesystems using extents
1850 Description: Heavy concurrent write and delete load may cause data corruption.
1851 Details : It was possible under high-load situations to have an extent
1852 metadata block in the block device cache from a just-unlinked
1853 file overwrite a newly-allocated data block. We now unmap any
1854 metadata buffers that alias just-allocated data blocks.
1858 Frequency : filesystems with default stripe_count larger than 77
1859 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
1860 Details : lconf specifies an inode size of 4096 bytes when the default
1861 stripe_count is larger than 77. This conflicts with the default
1862 inode density of 1 per 4096 bytes. Allocate smaller inodes in
1863 this case to avoid pinning too much memory for large EAs.
1865 ------------------------------------------------------------------------------
1867 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
1872 Frequency : rare (extremely heavy IO load with hundreds of clients)
1874 Description: Client is evicted, gets IO error writing to file
1875 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
1876 caused clients to be evicted for AST timeouts. The fixes in
1877 bug 5192 mean we no longer need to have such short AST timeouts
1878 so ldlm_timeout has been increased.
1881 Frequency : occasional during --force or --failover shutdown under load
1882 Bugzilla : 5949, 4834
1883 Description: Server oops/LBUG if stopped with --force or --failover under load
1884 Details : a collection of import/export refcount and cleanup ordering
1885 issues fixed for safer force cleanup
1888 Frequency : only filesystems larger than 120 OSTs
1889 Bugzilla : 5990, 6223
1890 Description: lfs getstripe would oops on a very large filesystem
1891 Details : lov_getconfig used kfree on vmalloc'd memory
1894 Frequency : only filesystems exporting via NFS to Solaris 10 clients
1895 Bugzilla : 6242, 6243
1896 Description: reading from files that had been truncated to a non-zero size
1897 but never opened returned no data
1898 Details : ll_file_read() reads zeros from no-object files to EOF
1903 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
1904 Details : There's little harm in aborting MDS/OSS recovery and letting it
1905 try again, so I removed the LASSERT and return an error instead.
1907 Severity : enhancement
1909 Description: New debugging infrastructure for tracking down data corruption
1910 Details : The I/O checksum code was replaced to: (a) control it at runtime,
1911 (b) cover more of the client-side code path, and (c) try to narrow
1912 down where problems occurred
1916 Bugzilla : 3819, 4364, 4397, 6313
1917 Description: Racing close and eviction MDS could cause assertion in mds_close
1918 Details : It was possible to get multiple mfd references during close and
1919 client eviction, leading to one thread referencing a freed mfd.
1921 Severity: : enhancement
1922 Bugzilla : 3262, 6359
1923 Description: Attempts to reconnect to servers are now more aggressive.
1924 Details : This builds on the enhanced upcall-less recovery that was added
1925 in 1.4.2. When trying to reconnect to servers, clients will
1926 now try each server in the failover group every 10 seconds. By
1927 default, clients would previously try one server every 25 seconds.
1932 Description: After recovery, certain operations trigger a failed
1933 assertion on a client.
1934 Details : Failing over an mds, using lconf -d --failover, while a
1935 client was doing a readdir() call would cause the client to
1936 LBUG after recovery completed and the readdir() was resent.
1938 Severity : enhancement
1940 Description: Default groups are now added by lconf
1941 Details : You can now run lconf --group <servicename> without having to
1942 manually add groups with lmc.
1945 Frequency : occasional
1947 Description: Nodes with an elan id of 0 trigger a failed assertion
1950 Frequency : always when accessing e.g. tty/console device nodes
1952 Description: tty and some other devices nodes cannot be used on lustre
1953 Details : file's private_data field is used by device data and lustre
1954 values in there got lost. New field was added to struct file to
1955 store fs-specific private data.
1958 Frequency : when exporting Lustre via NFS
1960 Description: NFSD failed occasionally when looking up a path component
1961 Details : NFSD is looking up ".." which was broken in ext3 directories
1962 that had grown large enough to become hashed.
1965 Frequency : Clusters with multiple interfaces not on the same subnet
1967 Description: Nodes will repeatedly try to reconnect to an interface which it
1968 cannot reach and report an error to the log.
1969 Details : Extra peer list entries will be created by lconf with some peers
1970 unreachable. lconf now validates the peer before adding it.
1973 Frequency : Only if a default stripe is set on the filesystem root.
1975 Description: Setting a default stripe on the filesystem root prevented the
1976 filesystem from being remounted.
1977 Details : The client was sending extra request flags in the root getattr
1978 request and did not allocate a reply buffer for the dir EA.
1981 Frequency : occasional, higher if lots of files are accessed by one client
1982 Bugzilla : 6159, 6097
1983 Description: Client trips assertion regarding lsm mismatch/magic
1984 Details : While revalidating inodes the VFS looks up inodes with ifind()
1985 and in rare cases can find an inode that is being freed.
1986 The ll_test_inode() code will free the lsm during ifind()
1987 when it finds an existing inode and then the VFS later attaches
1988 this free lsm to a new inode.
1992 Bugzilla : 6422, 7030
1993 Description: MDS deadlock between mkdir and client eviction
1994 Details : Creating a new file via mkdir or mknod (starting a transaction
1995 and getting the ns lock) can deadlock with client eviction
1996 (gets ns lock and trying to finish a synchronous transaction).
1999 Frequency : occasional
2000 Description: While starting a server, the fsfilt_ext3 module could not be
2002 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
2003 kernels. Previously, lconf would still use the ext3 name
2004 when trying to load modules. Now, it will correctly use
2005 ext3 on 2.4 and ldiskfs on 2.6.
2007 Severity : enhancement
2008 Description: The default stripe count has been changed to 1
2009 Details : The interpretation of the default stripe count (0, to lfs
2010 or lmc) has been changed to mean striping across a single
2011 OST, rather than all available. For general usage we have
2012 found a stripe count of 1 or 2 works best.
2014 Severity : enhancement
2015 Description: Add support for compiling against Cray portals.
2016 Details : Conditional compiling for some areas that are different
2020 Frequency : occasional
2021 Bugzilla : 6409, 6834
2022 Description: Creating files with an explicit stripe count may lead to
2023 a failed assertion on the MDS
2024 Details : If some OSTs are full or unavailable, creating files may
2025 trigger a failed assertion on the MDS. Now, Lustre will
2026 try to use other servers or return an error to the
2030 Frequency : occasional
2032 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
2033 caused lock timeout during readahead (since 1.4.2).
2034 Details : Processes doing readahead might match a lock that hasn't been
2035 granted yet if there are overlapping and conflicting lock
2036 requests. The readahead process waits on ungranted lock
2037 (original lock is CBPENDING), while OST waits for that process
2038 to cancel CBPENDING read lock and eventually evicts client.
2040 Severity : enhancement
2042 Description: Initial enabling of flock support for clients
2043 Details : Implements fcntl advisory locking and file status functions.
2044 This feature is provided as an optional mount flag (default
2045 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
2046 locking are implemented yet, and those that are are not guaranteed
2047 to be completely correct in production environments.
2048 mount -t lustre -o [flock|noflock] ...
2051 Frequency : occasional
2053 Description: OSTs running 2.4 kernels but with extents enabled might trip an
2054 assertion in the ext3 JBD (journaling) layer.
2055 Details : The b_committed_data struct is protected by the big kernel lock
2056 in 2.4 kernels, serializing journal_commit_transaction() and
2057 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
2058 access to this struct. In 2.6 kernels there is finer grained
2059 locking to improve SMP performance of the JBD layer.
2063 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
2065 -----------------------------------------------------------------------------
2067 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
2069 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
2070 versions of Lustre. Please update all servers and clients to
2071 version 1.4.2 or later at the same time. You must also run
2072 "lconf --write-conf {config}.xml" on the MDS while it is stopped
2073 to update the configuration logs.
2075 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
2076 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
2077 - swab llog records of type '0' so we get proper header size/idx (5861)
2078 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
2079 - fix rename of one directory over another leaking an inode (5953)
2080 - avoid SetPageDirty on 2.6 (5981)
2081 - don't re-add just-being-destroyed locks to the waiting list (5653)
2082 - when creating new directories, inherit the parent's custom
2083 striping settings if present parent (3048)
2084 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
2085 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
2086 - don't hold client locks on temporary worklist from l_lru (5666)
2087 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
2088 - automatically evict dead clients (5921)
2089 - Update file size properly in create+truncate+fstat case (6196)
2090 - Do not unhash mountpoint dentries, do not allow removal of
2092 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
2093 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
2094 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
2095 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
2096 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
2097 - Correctly handle reads of files with no objects (6243)
2098 - lctl recover will also mark a device active if deactivate used (5933)
2100 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
2101 - allow --write-conf on an MDS with different nettype than client (5619)
2102 - don't write config llogs to MDS for mounts not from that MDS (5617)
2103 - lconf should create multiple TCP connections from a client (5201)
2104 - init scripts are now turned off by default; run chkconfig --on
2105 lustre and chkconfig --on lustrefs to use them
2106 - upcalls are no longer needed for clients to recover to failover
2108 - add --abort-recovery option to lconf to abort recovery on device
2110 - add support for an arbitrary number of OSTs (3026)
2111 - Quota support protocol changes.
2112 - forward compatibility changes to wire structs (6007)
2113 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
2114 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
2115 - improved "lustre status" script
2116 - initialize blocksize for non-regular files (6062)
2117 - added --disable-server and --disable-client configure options (5782)
2118 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
2119 - Vanilla 2.4.29 support
2120 - increase maximum number of obd devices to 520 (6242)
2121 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
2122 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
2123 SLES 9 2.6.5 kernels
2124 - limit stripes per file to 160 (the maximum EA size) (6093)
2126 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2129 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
2130 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
2131 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
2132 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
2133 - small ext3 extents cleanups and fixes (5733)
2134 - improved mballoc code, several small races and bugs fixed (5733, 5638)
2135 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
2136 - avoid needless client->OST connect, fix handle mismatch (5317)
2137 - fix DLM error path that led to out-of-sync client, long delays (5779)
2138 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
2139 - fix several locking issues related to i_size (5492,5624,5654,5672)
2140 - don't move pending lock onto export if it is already evicted (5683)
2141 - fix kernel oops when creating .foo in unlinked directory (5548)
2142 - fix deadlock in obdfilter statistics vs. object create (5811)
2143 - use time_{before,after} to avoid timer jiffies wrap (5882)
2144 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
2145 - Do not leak request if server was not able to process it (5154)
2146 - If mds_open unable to find parent dir, make that negative lookup(5154)
2147 - don't create new directories with extent-mapping (5909, 5936)
2149 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
2150 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
2151 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
2152 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
2153 - support for RHEL4 kernel on i686 (5773)
2154 - provide error messages when incompatible logs are encountered (5898)
2156 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
2157 * version 1.4.0.10 (1.4.1 release candidate 1)
2159 - don't keep a lock reference when lock is not granted (4238)
2160 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2161 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
2162 - drop import inflight refcount on signal_completed_replay error (5255)
2163 - unlock page after async write error during send (3677)
2164 - handle missing objects in filter_preprw_read properly (5265)
2165 - no transno return for symlink open, don't save no-trasno open (3440)
2166 - don't try to complete elan receive that already failed (4012)
2167 - free RPC server reply state on error (5406)
2168 - clean up thread from ptlrpc_start_thread() on error (5160)
2169 - readahead could read extra page into cache that wasn't ejected (5388)
2170 - prevent races in class_attach/setup/cleanup/detach (5260)
2171 - don't dereference de->d_inode after l_dput of de (5458)
2172 - use "int" for stripe value returned from lock_to_stripe (5544)
2173 - mballoc allocation and error-checking fixes in 2.6 (5504)
2174 - block device patches to fix I/O request sizes in 2.6 (5482)
2175 - look up hostnames for IB nals (5602)
2176 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
2177 - don't start multiple acceptors for the same port (5277)
2178 - fix incorrect LASSERT in mds_getattr_name (5635)
2179 - export a proc file for general "ping" checking (5628)
2180 - fix "lfs check" to not block when the MDS is down (5628)
2182 - service request history (4965)
2183 - put {ll,lov,osc}_async_page structs in a single slab (4699)
2184 - create an "evict_client" /proc entry on OSTs, like the MDS has
2185 - fix mount usage message, return errors per mount(8) (5168)
2186 - change grep [] to grep "[]" in tests so they work in more UMLs
2187 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
2188 - remove ancient LOV_MAGIC_V0 EA support (5047)
2189 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
2190 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
2191 - allow readahead allocations to fail when low on memory (5383)
2192 - mmap locking landed again, after considerable improvement (2828)
2193 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
2195 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
2198 - send OST transaction number in read/write reply to free req (4966)
2199 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2200 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2201 - the watchdog thread now runs as interruptible (5246)
2202 - flock/lockf fixes (but it's still disabled, pending 5135)
2203 - don't use EXT3 constants in llite code (5094)
2204 - memory shortage at startup could cause assertion (5176)
2206 - reorganization of lov code
2207 - single portals codebase
2209 - add extents/mballoc support (5025)
2210 - direct I/O reads in the obdfilter (4048)
2211 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
2213 tbd Cluster File Systems, Inc. <info@clusterfs.com>
2216 - send OST transaction number in read/write reply to free req (4966)
2217 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
2218 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
2219 - the watchdog thread now runs as interruptible (5246)
2220 - handle missing objects in filter_preprw_read properly (5265)
2221 - unsafe list practices (rarely) led to infinite eviction loop (4908)
2222 - drop import inflight refcount on signal_completed_replay error (5255)
2223 - unlock page after async write error during send (3677)
2224 - return original error code on reconstructed replies (3761)
2225 - no transno return for symlink open, don't save no-trasno open (3440)
2227 - add pid to ldlm debugging output (4922)
2228 - bump the watchdog timeouts -- we can't handle 30sec yet
2229 - extra debugging for orphan dentry/inode bug (5259)
2231 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
2234 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
2235 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
2236 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
2237 - fix NULL dereference in /proc/sys/portals/routes (4827)
2238 - allow failed mdc_close() operations to be interrupted (4561)
2239 - stop precreate on OST before MDS would time out on it (4778)
2240 - don't send partial-page writes before EOF from client (4410)
2241 - discard client grant for sub-page writes on large-page clients (4520)
2242 - don't free dentries not owned by NFS code, check generation (4806)
2243 - fix lsm leak if mds_create_objects() fails (4801)
2244 - limit debug_daemon file size, always print CERROR messages (4789)
2245 - use transno after validating reply (3892)
2246 - process timed out requests if import state changes (3754)
2247 - update mtime on OST during writes, return in glimpse (4829)
2248 - add mkfsoptions to LDAP (4679)
2249 - use ->max_readahead method instead of zapping global ra (5039)
2250 - don't interrupt __l_wait_event() during strace
2252 - add software watchdogs to catch hung threads quickly (4941)
2253 - make lustrefs init script start after nfs is mounted
2254 - fix CWARN/ERROR duplication (4930)
2255 - return async write errors to application if possible (2248)
2256 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
2257 - print NAL number in %x format (4645)
2258 - update barely-supported suse-2.4.21-171 series (4842)
2259 - support for sles 9 %post scripts
2260 - support for building 2.6 kernel-source packages
2261 - support for sles km_* packages
2263 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
2266 - ignore -ENOENT errors in osc_destroy (3639)
2267 - notify osc create thread that OSC is being cleaned up (4600)
2268 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
2269 - reconstruct ost_handle() like mds_handle() (4657)
2270 - create a new thread to do import eviction to avoid deadlock (3969)
2271 - let lconf resolve symlinked-to devices (4629)
2272 - don't unlink "objects" from directory with default EA (4554)
2273 - hold socknal file ref over connect in case target is down (4394)
2274 - allow more than 32000 subdirectories in a single directory (3244)
2275 - fix blocks count for O_DIRECT writes (3751)
2276 - OST returns ENOSPC from object create when no space left (4539)
2277 - don't send truncate RPC if file size isn't changing (4410)
2278 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
2279 - bind to privileged port in socknal and tcpnal (3689)
2281 - rate limit CERROR/CWARN console message to avoid overload (4519)
2282 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
2283 - basic mmap support (3918)
2284 - kernel patch series update from b1_4 (4711)
2286 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
2289 - avoid crash during MDS cleanup with OST shut down (2775)
2290 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
2291 - don't use bad inodes on the MDS (3744)
2292 - dynamic object preallocation to improve recovery speed (4236)
2293 - don't hold spinlock over lock dumping or change debug flags (4401)
2294 - don't zero obd_dev when it is force cleaned (3651)
2295 - print grants to console if they go negative (4431)
2296 - "lctl deactivate" will stop automatic recovery attempts (3406)
2297 - look for existing locks in ldlm_handle_enqueue() (3764)
2298 - don't resolve lock handle twice in recovery avoiding race (4401)
2299 - revalidate should check working dir is a directory (4134)
2301 - don't always mark "slow" obdfilter messages as errors (4418)
2303 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
2306 - don't close LustreDB during write_conf until it is done (3860)
2307 - fix typo in lconf for_each_profile (3821)
2308 - allow dumping logs from multiple threads at one time (3820)
2309 - don't allow multiple threads in OSC recovery (3812)
2310 - fix debug_size parameters (3864)
2311 - fix mds_postrecov to initialize import for llog ctxt (3121)
2312 - replace config semaphore with spinlock (3306)
2313 - be sure to send a reply for a CANCEL rpc with bad export (3863)
2314 - don't allow enqueue to complete on a destroyed export (3822)
2315 - down write_lock before checking llog header bitmap (3825)
2316 - recover from lock replay timeout (3764)
2317 - up llog sem before sending rpc (3652)
2318 - reduce ns lock hold times when setting kms (3267)
2319 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
2320 - fix NULL deref and obd_dev leak on setup error (3312)
2321 - replace some LBUG about llog ops with error handling (3841)
2322 - don't match INVALID dentries from d_lookup and spin (3784)
2323 - hold dcache_lock while marking dentries INVALID and hashing (4255)
2324 - fix invalid assertion in ptlrpc_set_wait (3880)
2326 - add libwrap support for the TCP acceptor (3996)
2327 - add /proc/sys/portals/routes for non-root route listing (3994)
2328 - allow setting MDS UUID in .xml (2580)
2329 - print the stack of a process that LBUGs (4228)
2331 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
2334 - don't cleanup request in ll_file_open() on failed MDS open (3430)
2335 - make sure to unset replay flag from failed open requests (3440)
2336 - if default stripe count is 0, use OST count for inode size (3636)
2337 - update parent mtime/ctime on client for create/unlink (2611)
2338 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
2339 - free recovery state on server during a forced cleanup (3571)
2340 - unregister_reply for resent reqs (3063)
2341 - loop back devices mounting and status check on 2.6 (3563)
2342 - fix resource-creation race that can provoke i_size == 0 (3513)
2343 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
2344 - more debugging for page-accounting assertion (3746)
2345 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
2346 - avoid deadlock after precreation failure (3758)
2347 - fix race and lock order deadlock in orphan handling (3450, 3750)
2348 - add validity checks when grabbing inodes from l_ast_data (3599)
2350 - add /proc/.../recovery_status to obdfilter (3428)
2351 - lightweight CDEBUG infrastructure, debug daemon (3668)
2352 - change default OSC RPC parameters to be better on small clusters
2353 - turn off OST read cache for files smaller than 32MB
2354 - install man pages and include them in rpms (3100)
2355 - add new init script for (un)mounting lustre filesystems (2593)
2356 - run chkconfig in %post for init scripts (3701)
2357 - drop scimac NAL (unmaintained)
2359 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
2362 - clean kiobufs before and after use (3485)
2363 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
2364 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
2365 - change init script to fail more clearly if not run as root (1528)
2366 - allow clients to reconnect during replay (1742)
2367 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
2368 - don't do DNS lookups on NIDs too small for IP addresses (3442)
2369 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
2370 - fix cond_resched (3554)
2371 - only evict unfinished clients after recovery (3515)
2372 - allow bulk resend, prevent data loss (3570)
2373 - dynamic ptlrpc request buffer allocation (2102)
2374 - don't allow unlinking open directory if it isn't empty (2904)
2375 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
2376 - remove extraneous obd dereference causing LASSERT failure (3334)
2377 - don't use get_cycles() when creating temp. files on the mds (3156)
2378 - hold i_sem when setting i_size in ll_extent_lock() (3564)
2379 - handle EEXIST for set-stripe, set proper directory name (3336)
2381 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
2382 - fix ksocknal_fmb_callback() error messages (2918)
2384 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
2387 - don't copy lvb into (possibly NULL) reply on error (2983)
2388 - don't deref dentry after dput, don't free lvb on error (2922)
2389 - use the kms to determine writeback rpc length (2947)
2390 - increment oti_logcookies when osc is inactive (2948)
2391 - update client's i_blocks count via lvb messages (2543)
2392 - handle intent open/close of special files properly (1557)
2393 - mount MDS with errors=remount-ro, like obdfilter (2009)
2394 - initialize lock handle to avoid ASSERT on error cleanup (3057)
2395 - don't use cancelling-locks' kms values (2947)
2396 - use highest lock extent for kms, not last one (2925)
2397 - don't dereference ERR_PTR() dentry in error handling path (3107)
2398 - fix thread race in portals_debug_dumplog() (3122)
2399 - create lprocfs device entries at setup instead of at attach (1519)
2400 - common AST error handler, don't evict client on completion race (3145)
2401 - zero nameidata in detach_mnt in 2.6 (3118)
2402 - verify d_inode after revalidate_special is valid in 2.6 (3116)
2403 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
2404 - initialize RPC timeout timer earlier for 2.6 (3219)
2405 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
2406 - print nal/nid for unknown nid (3258)
2407 - additional checks for oscc recovery before doing precreate (3284)
2408 - fix ll_extent_lock() error return code for 64-bit systems (3043)
2409 - don't crash in mdc_close for bad permissions on open (3285)
2410 - zero i_rdev for non-device files (3147)
2411 - clear page->private before handing to FS, better assertion (3119)
2412 - tune the read pipeline (3236)
2413 - fix incorrect decref of invalidated dentry (2350)
2414 - provide read-ahead stats and refine rpc in flight stats (3328)
2415 - don't hold journal transaction open across create RPC (3313)
2416 - update atime on MDS at close time (3265)
2417 - close LDAP connection when recovering to avoid server load (3315)
2418 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
2419 - don't leak open file on MDS after open resend (3325)
2420 - serialize filter_precreate and filter_destroy_precreated (3329)
2421 - loop device shouldn't call sync_dev() for nul device (3092)
2422 - clear page cache after eviction (2766)
2423 - resynchronize MDS->OST in background (2824)
2424 - refuse to mount the same filesystem twice on same mountpoint (3394)
2425 - allow llmount to create routes for mounting behind routers (3320)
2426 - push lock cancellation to blocking thread for glimpse ASTs (3409)
2427 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
2428 - fix rare problem with rename on htree directories (3417)
2430 - allow default OST striping configuration per directory (1414)
2431 - fix compilation for qswnal for 2.6 kernels (3125)
2432 - increase maximum number of MDS request buffers for large systems
2433 - change liblustreapi to be useful for external progs like lfsck (3098)
2434 - increase local configuration timeout for slow disks (3353)
2435 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
2437 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
2440 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
2441 - don't overwrite extent policy data in reply if lock was blocked (2901)
2442 - drop filter export grants atomically with removal from device (2663)
2443 - del obd_self_export from work_list in class_disconnect_exports (2908)
2444 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
2445 - swab reply message in mdc_close, other PPC fixes (2464)
2446 - fix destroying of named logs (2325)
2447 - overwrite old logs when running lconf --write_conf (2264)
2448 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
2449 - fix race in target_handle_connect (2898)
2450 - mds_reint_create() should take same inode create lock (2926)
2451 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
2452 - don't close files for self_export to avoid uninitialized obd (2936)
2453 - allow MDS with the same name as client node (2939)
2454 - hold dentry reference for closed log files for unlink (2325)
2455 - reserve space for all logs during transactions (2059)
2456 - don't evict page beyond end of stripe extent (2925)
2457 - don't oops on a deleted current working directory (2399)
2458 - handle hard links to targets without a parent properly (2517)
2459 - don't dereference NULL lock when racing during eviction (2867)
2460 - don't grow lock extents when lots of conflicting locks (2919)
2462 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
2465 - account for cache space usage on clients to avoid data loss (974)
2466 - lfsck support in lustre kernel code (2349)
2467 - reduce journal credits needed for BRW writes (2370)
2468 - orphan handling to avoid losing space on client/server crashes
2469 - ptlrpcd can be blocked, stopping ALL progress (2477)
2470 - use lock value blocks to assist in proper KMS, faster stat (1021)
2471 - takes i_sem instead of DLM locks internally on obdfilter (2720)
2472 - recovery for initial connections (2355)
2473 - fixes for mds_cleanup_orphans (1934)
2474 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
2475 - block all file creations until orphan recovery completes (1901)
2476 - client remove rq_connection from request struct (2423)
2477 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
2478 - recovery timer race (2670)
2479 - mdc_close recovey bug (2532)
2480 - ptlrpc cleanup bug (2710)
2481 - mds timeout on local locks (2588)
2482 - namespace lock held during RPCs (2431)
2483 - handle interrupted sync write properly (2503)
2484 - don't try to handle a message that hasn't been replied to (2699)
2485 - client assert failure during cleanup after abort recovery (2701)
2486 - leak mdc device after failed mount (2712)
2487 - ptlrpc_check_set allows timedout requests to complete (2714)
2488 - wait for inflight reqs when ptlrpcd finishes (2710)
2489 - make sure unregistered services are removed from the srv_list
2490 - reset bulk XID's when resending them (caught by 1138 test)
2491 - unregister_bulk after timeout
2492 - fix lconf error (2694)
2493 - handle write after unfinished setstripe, stripe-only getstripe (2388)
2494 - readahead locks pages, leaves pending causing memory pressure (2673)
2495 - increase OST request buffers to 4096 on large machines (2729)
2496 - fix up permission of existing directories in simple_mkdir (2661)
2497 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
2498 - don't assign transno to errored transactions (2742)
2499 - don't delete objects on OST if given a bogus objid from MDS (2751)
2500 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
2501 - if rq_no_resend, then timeout request after recovery (2432)
2502 - fix MDS llog_logid record size, 64-bit array alignment (2733)
2503 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
2504 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
2505 - MDS recovery shouldn't delete objects that it has given out (2730)
2506 - if enqueue arrives after completion, don't clobber LVB (2819)
2507 - don't unlock pages twice when trigger_group_io returns error (2814)
2508 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
2509 - don't write pages to disk if there was an error (1450)
2510 - don't ping imports that have recovery disabled (2676)
2511 - take buffered bytes into account when balancing socknal conn (2817)
2512 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
2513 - reconnect unlink llog connection after MDS reconnects to OST (2816)
2514 - remove little-endian swabbing of llog records (1987)
2515 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
2516 - retry reposting request buffers if they fail (1191)
2517 - grow extent at grant time to avoid granting a revoked lock (2809)
2518 - lock revoke doesn't evict page if covered by a second lock (2765)
2519 - disable VM readahead to avoid reading outside lock extents (2805)
2521 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
2522 - updated kernel patches for hp-2.4.20 kernel (2681)
2524 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
2527 - fix truncated write corruption (2366)
2528 - fix for failed assertion in iopen_connect_dentry (1792,2517)
2530 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
2531 - ensure len(uuid) < 37 in lmc (1171)
2532 - fix ia64 OOPS in llog_test (2255)
2533 - zero end of page at obdfilter for partial page writes (2648)
2534 - don't leave stale dentries around after renames (bug 2428)
2535 - fix timeouts when evicting a client with a single lock held (2642)
2536 - set deadline for the initial HELLO message to drain (2634)
2537 - print out dotted-quad IP addresses in the socknal (2302)
2539 - additional debugging for MDS client eviction problem (2443)
2540 - fix mkfsoptions support for osts (2603, 2604)
2542 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
2545 - add series for the vanilla 2.6.0 kernel
2546 - add series for the vanilla 2.4.24 kernel
2547 - add series for a cray x86/64 UL kernel drop
2548 - fix xattr patches for the vanilla 2.4.19 series
2550 - generate true UUIDs in lmc (1171)
2551 - have portals stack dumping break in UML (2466)
2552 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
2553 - allocate with _NOFS in ldlm to avoid deadlock (1933)
2554 - wake callback waiting threads on client eviction (2460)
2555 - Add --ptldebug and --subsystem to lmc (1719)
2556 - update assertion to allow safe interrupt allocation
2557 - set rq_no_resend for cancel requests (2432)
2558 - recalculate ptlrpcd timeout after resend (2494)
2559 - call vfs_rmdir when removing pending directories (2368)
2560 - fix renaming a file to itself (2429)
2561 - lmc creates a default one-stripe lov (2454)
2562 - expand procfs space to handle large clusters (2326)
2563 - increase UML stack to avoid overflow
2564 - update lconf's list of debug and subsystem masks
2565 - fix lfs find --obd (2510)
2566 - /proc tunable for disabling filter read caching (2591)
2567 - stop rpm packages from altering slapd.conf (2301)
2568 - disable nagle in the socknal under 0conf (2578)
2569 - choose mds inode size based on stripe count (2572)
2570 - fix kernel-source rpm problems (2516)
2572 - add --disable-doc to avoid pdf generation (2421)
2573 - update documentation, tests, type-os, comments
2574 - avoid format warnings on ia64
2575 - remove the TOE NAL
2576 - tiny code cleanups by removing unused fields
2578 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
2581 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
2582 - remove the most blatant lies from BUILDING file (2371)
2583 - change default debug level to reasonable production setting
2584 - reduce client side cache size to reduce cache flush time
2585 - reduce max RPCs in flight to avoid unnecessary file fragmentation
2586 - make TCP zerocopy and pinger support enabled by default (2476)
2587 - sync writes completed after process exits caused crashes (2319)
2588 - maintain correct mount count on the MDS (2356)
2589 - backout 1557, because 2316 wasn't really fixed
2590 - better file I/O statistics gathering in /proc
2591 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
2592 - another kernel patch to fix zero-copy TCP function export
2593 - don't take duplicate lock when processing re-sent getattr (2420)
2594 - lctl uses obd_self_export instead of creating new conn (2353)
2595 - MDS/OST recovery case which requires object creation asserted (2425)
2596 - move lfs from /usr/sbin to /usr/bin in packages
2597 - fix race between mds_client_add and mds_client_free (2417)
2598 - use kmalloc instead of slabs in portals (2430)
2599 - don't create duplicate records when a failover MDS is present (2442)
2600 - remove unnecessary mount age check (2332)
2601 - don't remove directory inodes from locks prematurely (2451)
2602 - don't break if MDS service name is the same as hostname (2103)
2603 - fix races in client write RPC generation when cache full (2482)
2605 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
2608 - remove now-unused request->rq_obd (278)
2609 - if an allocation fails, print out how much memory we've used (1933)
2610 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
2611 - add the "configurable stack size" patch to most series files (1256)
2612 - ability to write large log records, for 100+ OST configs (2306)
2613 - fix NULL deref when filter_prep fails (2314)
2614 - fix operator precedence error in filter_sync
2615 - dynamic allocation of socknal TX descriptors (2315)
2616 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
2617 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
2618 - ensure that CWARN messages in Portals always get to the syslog
2619 - __init/__exit are not for prototype decls (ldlm_init/exit)
2620 - x86-64 compile warning fixes
2621 - fix gateway LMC keyword conflict (2318)
2622 - fix MDS lock inversions in getattr/reint paths (1844)
2623 - fix a rare lock re-ordering bug, which caused deadlock (2322)
2624 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
2625 - DLM race condition prevented some lock evictions (2328)
2626 - ENOMEM detection and retry on socknal sends (2230)
2627 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
2628 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
2629 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
2630 - better error messages when a client is rejected during recovery (1505)
2631 - avoid cancelling locks which were never granted, after failure (2330)
2632 - fix i_sem/journal inversion in mds_client_add (2333)
2633 - fix truncate/getattr lock cycle deadlock (2334)
2634 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
2635 - fix two rare exit paths which could leak an l_lock() ref (2321)
2636 - fencepost error in MDS/OST orphan recovery (2226)
2637 - make log record alignment 8 bytes (1988)
2638 - lstripe now fails when requested offset > ost_count (2237)
2639 - ensure that all kernel series have a complete list.h (1607)
2640 - fix crashes in special-file operations (2316)
2641 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
2643 - allow configurable automake binary, for testing new versions
2644 - small update to the lfs documentation
2646 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
2648 * fix negative export reference count in fsfilt_sync (2312)
2650 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
2651 * release candidate 0.9.1
2653 - orphans are moved into the PENDING directory for possible recovery
2654 - replayed opens now open by fid for orphan/rename safety (1042)
2655 - last close of an orphan inode generates a transno (683)
2656 - chdir() and mount() now pin the directory entry (1020)
2657 - avoid CERROR in normal ll_setattr_raw() error case (1500)
2658 - discard very old requests without processing them (1502)
2659 - remove some common, well-understood CERRORs (1505)
2660 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
2661 - clear "grant" flags in OST replies until OST grant code lands (1644)
2662 - fix read performance by not clobbering i_blksize on client (1598)
2663 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
2664 - make LRU size a /proc tunable, clears locks when reduced (707)
2665 - fix some lprocfs rot that prevented ptlbd from loading (1732)
2666 - server locks take references on exports now (1558)
2667 - build fixes for 2.4.20-rh trees (1663)
2668 - return an error from lov_create if all OSCs are inactive (1751)
2669 - fix import levels when a reconnect happens without a timeout (1597)
2670 - exit early from mds_open if we get a lookup error (1749)
2671 - partial page read at EOF wouldn't wait for disk before sending (1642)
2672 - avoid NULL deref in obdfilter when reading page past EOF (1592)
2673 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
2674 - fix LBUG in ll_it_open_error with rc = -2 (1861)
2675 - write/truncate lock inversion (1639)
2676 - Don't auto-load obdclass, portals modules during cleanup (1495)
2677 - fix timestamps from jumping to "now" (1763)
2678 - extra journal assertions (1648)
2679 - add an extra multiunlink test (1771)
2680 - fix read_record/write_record API (1776)
2681 - fix leak of offset_extent, possible incorrect i_size later (1772)
2682 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
2683 - Debugging for the kqswnal_get_idle_tx problems (1820)
2684 - Allow recovery to be attempted multiple times (1536)
2685 - Write out MDS last_rcvd file after it is first created (1600)
2686 - Fix tx_descriptor leak in failed transmit situations (1827)
2687 - ext3 journaling fixes for assertion failure after IO error (1871)
2688 - class_export_put() on freed export after completion AST error (1896)
2689 - Fix revalidate looping in VFS (1322)
2690 - Don't access a freed export during MDS_REINT timeout (1521)
2691 - Add open-unlink recovery support on the MDS (1673,1764)
2692 - Return an error if no MDS data was read from last_rcvd (1946)
2693 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
2694 - Don't setup a disk device that doesn't match exported UUID (317)
2695 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
2696 - avoid committing NULL handle in force close
2697 - local.sh is now a one-stripe LOV configuration
2698 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
2699 - fix ext3 htree duplicate directory entry corruption (1516)
2700 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
2701 - update to new LOV EA format (2097)
2702 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
2704 2003-06-15 Phil Schwan <phil@clusterfs.com>
2707 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
2708 - per-import/export recovery handling (958, 931, 959)
2709 - multiple last-rcvd slots, for serving multiple FSes (949)
2710 - connections are again shared between multiple imp/exports (963, 964)
2711 - "umount -f" would hang if any requests needed to be sent (393, 978)
2712 - avoid pinning large req buffer by copying for queued messages (989)
2713 - add "uuid" to "lctl device" command to help upcalls (991)
2714 - "open" RPCs with transnos would confuse recovery counters (1037)
2715 - do proper endian conversion of all wire messages (288, 340, 891)
2716 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
2717 - call waiting locks callback from LDLM recovery thread (1127, 1151)
2718 - fix ptlrpc_connection leak in target_handle_connect (1174)
2719 - fix import refcounting bug in OST and MDS cleanup (1134)
2720 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
2721 - fix very unlikely obd_types race condition (501)
2722 - remove osc_open hack for echo_client (1187)
2723 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
2724 - a failure in read_inode2 leads to deadlock (1139)
2725 - cancel ack-locks as soon as transaction is committed (1072)
2726 - fix major leaks and crashes in the bulk I/O path (937, 1057)
2727 - make sure to commitrw after any preprw to avoid deadlock (1162)
2728 - failing to execute a file in a lustre FS would lock inode (1203)
2729 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
2730 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
2731 - obd_uuid2tgt would walk off the end of the list (1255)
2732 - on IA64 the osc would give portals incorrect bulk size (1258)
2733 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
2734 - fix lock inversion caused by new llite matching code (1282)
2735 - limit the number of dirty pages on a client to 10MB (1286)
2736 - timed out locks were not being corrected cancelled (1289)
2737 - fix O_DIRECT above 4GB on IA-32 (1292)
2738 * major user-visible changes
2739 - fail out/fail over policy now controlled by the upcall (993)
2741 - add OBD_PING to check server availability and failure (954)
2742 - lustre messages are now sent in sending host order (288, 340, 891)
2743 - add eadatalen to MDS getattr reply (340)
2744 - OST read replies may contain second buffer, with per-page status (593)
2746 2003-03-11 Phil Schwan <phil@clusterfs.com>
2749 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
2750 - failed open()s could cause deadlock; fixed (867, 869)
2751 - stop cancelling OST locks when files are closed (481)
2752 - overlapping XID spaces caused network corruption (851, 853)
2753 - fix unsafe fsfilt counter arithmetic; change to atomic_t
2754 - setattr_raw added, to do single-RPC, server-side setattrs
2755 - lmc/lconf syntax change for OST UUIDs
2756 - fix crashy race condition between ptlrpc_free_req and osc_close
2757 - don't use request in mdc_enqueue if we hit a timeout (889)
2758 - don't set the inode i_size for regular files from the MDS (896)
2759 - handle out of order completion AST (842)
2760 - don't LBUG if a lock request times out after receiving AST (913)
2761 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
2762 - if a bad lock AST arrives, send an error instead of dropping entirely
2763 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
2764 - fix leak in bulk IO when only partially completed (899, 900, 926)
2765 - fix O_DIRECT for ia64 (55)
2766 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
2767 - C-z after timeout could hang a process forever; fixed (977)
2769 - client-side I/O cache (678, 924, 929, 941, 970)
2771 - READPAGE and SETATTRs which don't take server-side locks get
2774 2003-02-11 Phil Schwan <phil@clusterfs.com>
2777 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
2778 - Fix fsfilt_extN_readpage() to read a full page of directory
2779 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
2780 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
2781 - Fix imbalanced LOV object allocation and out-of-bound access (469)
2782 - Most intent operations were removed, in favour of a new RPC mode
2783 that does a single RPC to the server and bypasses most of the VFS
2784 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
2785 - Aggressively cancel local locks on DLM servers
2786 - mds_reint_unlink sends EA to the client if it's the last nlink.
2787 client uses that EA to unlink OST objects.
2788 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
2789 - recursive symlinks were fixed (439)
2790 - fixed NULL deref in DEBUG_REQ
2791 - filter_update_lastobjid no longer calls sync, which annoyed extN
2792 - fixed multi-client small-writes to a single file problem (445)
2793 - fixed mtime updates during file writes (607)
2794 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
2795 - fixed bug in obd_brw_read/write() (under guise of testing 367)
2796 - fixed Linux OST size reporting problem (444, 656)
2797 - OST now updates object mtime with writes or setattr (607, 619)
2798 - client verifies file size before zeroing page past EOF (445)
2799 - OST now writes last allocated objid to disk with allocation (108)
2800 - LOV on echo now works (409)
2802 - mds_reint_unlink sends a new buffer, with the EA included. this
2803 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
2804 set if a regular file was being unlinked, and it was the last link
2805 - use PtlGet from the target for bulk writes (315)
2806 - OST now updates object mtime with writes or setattr (607, 619)
2807 - LDLM now has a grant-time callback to revalidate locked items, if
2809 - Many MDS operations were reorganized to combat race conditions
2811 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
2812 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
2814 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
2817 - Fully reactivate OST imports after reconnection (512, others)
2818 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
2819 - More graceful error handling for truncating on dead OST (515)
2820 - Don't error out unless we're actually accessing dead stripes (474)
2821 - Fix garbage sizes when stripes are missing (410)
2822 - LRU counters were broken, causing constant lock purge (433, 432)
2823 - garbage on read from stripes with failed OSTs (441)
2824 - mark OSCs as active before reconnecting during recovery (438)
2825 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
2826 - lfind did not preserve OST order in output (443)
2827 - symlinks cause hung clients, incorrect data (439)
2828 - stop dereferencing request after dropping refcount (457)
2829 - don't LASSERT(spin_is_locked) on non-SMP (455)
2830 - fixes for many rename() bugs
2831 - fstat didn't correctly synchronize attributes (399)
2832 - server must handle lock cancellation during blocking AST prep (487)
2833 - bulk descriptors were free()d too soon (511)
2834 - fix paths in lconf, which would load incorrect modules (451, 507)
2835 - fix confusing lconf 'host not found' error message (386)
2836 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
2837 - fix race condition in mdc_blocking_ast() for inode access (526)
2838 - fix lov_unpackmd() unpacking wrong number of stripes (537)
2839 - fix lov_set_osc_active() marking wrong OSC inactive (440)
2840 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
2841 - fix multiple writes of stripe MD to MDS (358, maybe 519)
2842 - fix lstripe in several ways (kernel side) (527)
2843 - fix request leak in ldlm_cli_enqueue (262)
2844 - incorrect OSC was marked inactive after OST failure
2845 - call mds_fs_cleanup before unmounting filesystem (524)
2846 - fix races between taking ns_lock and ldlm_lock_change_resource
2847 - fix races updating LOV export open file list
2848 - fix lov_enqueue error path, avoid decref-ing bad lock handle
2849 - fix recovery NULL deref in ldlm_cli_cancel_unused
2850 - fix some DLM races by using new hash table for lock handles (419)
2851 - permit the client to specify desired inodes, at replay
2852 - duplicate requests when we queue them for replay reintegration
2853 - fix last_rcvd offset calculation
2854 - sync after each recovered transaction, so we always make progress
2855 - never, not always, ERESTART requests without transnos
2856 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
2858 - skip replay if the MDS says that the client is already connected
2859 - don't check for a recovery-enabled export to match lctl's UUID
2860 - don't INC_USE_COUNT for phantom exports
2861 - don't crash when cleaning up phantom exports (567)
2862 - don't double-finish or set replay data for errored mdc_open requests
2863 - abort requests when they time out, so we don't get old replies
2864 - send/receive replies for AST messages again
2865 - if the client says that it doesn't have the lock, cancel it on the
2867 - if we timeout during I/O, don't try to cancel an in-use lock; instead
2868 mark it as destroyed, it will all work out when decref is called
2869 - fix module use counts (22, 581)
2871 - ASTs now expect a reply (server cancels lock on error reply)
2873 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
2876 - fix many simultaneous client startup (392)
2877 - fix dentry->d_it clobbering
2878 - credentials weren't being shipped for readdir/getattr operations
2879 - remove invalid assertions triggered during some concurrent MD
2881 - proper Lustre versions added (336, 389)
2882 - fix memory leak for create error case (398)
2883 - fix LOV locking bug that would get cli/srv out of sync
2884 - fix echo client over LOV (409)
2885 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
2886 - fix double-O_EXCL intent crash (424)
2887 - avoid sending multiple lock CANCELs (352)
2889 - MDS can do multi-client recovery (modulo bugs in new code)
2891 - many updates, edits, cleanups
2893 2002-11-18 Phil Schwan <phil@clusterfs.com>
2896 - fix null d_it dereference (346)
2897 - fix full OST/dbench hang (333)
2898 - fix permission problem with file removal (286)
2899 - fix removal of OSCs from LOV when they fail
2900 - fix NULL deref during bulk timeout (214)
2901 - fix problems related to multiple filesystems on one MDS (241)
2902 - fixed serious subtle metadata locking bugs
2903 - free locks on clients when inodes are removed due to memory
2905 - fix inode pointer in lock data (285)
2906 - partial support for multiple MDS on a single host (241)
2907 - data locks weren't cancelled at clear_inode time (290, 311)
2908 - intent locks could lead to unbounded lock growth (205)
2909 - added a maximum lock count, an LRU list, and a flusher
2910 - fix multiple rename (365)
2911 - properly abstracted the echo client
2912 - OSC locked 1 byte too many; fixed
2913 - rewrote brw callback code:
2914 - fixed recovery bugs related to LOVs (306)
2915 - fixed too-many-pages-in-one-write crash (191)
2916 - fixed (again) crash in sync_io_timeout (214)
2917 - probably fixed callback-related race (385)
2919 - Add capability to MDS protocol
2920 - LDLM cancellations and callbacks on different portals
2922 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
2925 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
2926 - defer open object destruction to close time (601981, 138)
2927 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
2928 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
2929 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
2930 - don't call obd_setattr in ll_file_release for destroyed objects
2931 * protocol change to lustre_msg: move |version| and add |flags|
2932 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
2933 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
2934 * added replay of create, unlink, link and rename operations during
2935 MDS failover; recovery should be much more robust now
2936 * remove failed OSCs from LOVs (only lov_create uses this so far)
2937 * the lustre-HOWTO was brought (more) up to date (582544)
2939 2002-10-23 Phil Schwan <phil@clusterfs.com>
2942 - in-use dentries weren't being reused properly (617851)
2943 - prevent multiple LDLM setup (599178)
2944 - fix LOV size calculations for truncate (617853)
2945 - fix client handling of MDS intent errors (POSIX)
2946 - fix permission bug in lovstripe.c test (624321)
2947 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
2948 - truncate past end of file could corrupt data
2949 - proper cleanup after timeouts, crashes, etc (592524, 550815)
2950 - a race in recovery could return ETIMEDOUT to apps (623947)
2951 - building outside the source directory was fixed
2952 * the lustre-HOWTO was brought (more) up to date (582544)
2953 * major progress was made on recovery functionality
2955 2002-10-10 Phil Schwan <phil@clusterfs.com>
2958 - recovery deadlock fix
2959 - rm -rf causes LBUG fix (617817)
2960 - file open by multiple tasks fix (618962)
2961 - directory permissions bugs (602707 and 620007)
2962 - journal_stop fixed with locking (611313)
2963 - O_APPEND failures resolved (618273, perhaps 614459)
2964 - lconf PATH fix (619770)
2965 - IA64 build fix (621450)
2966 - RPC buffer sizes scale with amount of memory
2968 2002-10-01 Phil Schwan <phil@clusterfs.com>
2971 - locks would be cancelled without throwing away data pages,
2972 resulting in inconsistent data (605627)
2973 - inode attributes were not always being refreshed (605627, 612449)
2974 - lconf now continues to cleanup after lctl reports an error
2975 - MDS now enforces user permissions (602707)
2976 - lprocfs cleanup fixed, but not yet enabled (614157)
2977 - fixed infinite server hang, should a client not respond to an AST
2978 - avoid going into recovery if user calls readlink() with a buffer
2979 that's too small (613941)
2980 - AST RPCs no longer require replies (614867) -- this may be changed
2981 - don't crash server if client sends an IOV that's too big (611336)
2982 - fixed lock conversion deadlock (611892)
2983 - fixed the following of symlinks (614622)
2984 * recovery: the server can remove locks from a client that dies, other
2985 clients can make progress
2986 * more extN patch fixes
2987 * compile-time configurable ptlrpc buffer allocations
2989 - collaborative read cache document
2990 - Lustre Lite Performance CDR document-in-progress
2992 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
2995 - fix typo in patch-2.4.18
2997 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
3000 - clear ptlrpc request each time in handle_incoming_request()
3001 - unlink of files now destroys the object on the OST
3003 2002-09-19 Peter Braam <braam@clusterfs.com>
3005 * add hard link support
3006 * change obdfile creation method
3007 * kernel patch changed
3009 2002-09-19 Peter Braam <braam@clusterfs.com>
3012 - stack overflow bug in extN fixed
3014 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
3016 * documentation updates
3017 - add man pages for config tools
3018 - update tests/README to describe testing with new config tools
3019 - finish metadata API descriptions
3020 * bug fixes and cleanups
3021 - statfs workaround for 16TB limit
3022 - LOV stripe allocation improved, can stripe on subset of OSTs
3023 - LOV file size/IO offset was wrong for files > 4GB in size
3024 - object EA data was being dropped, caused files to be unreadable
3025 - memory overflow with non-LOV OST caused memory corruption
3026 - fixed regression tests to work with new config tools, obdfilter
3027 - fixed bug when directory size became larger than 1 block
3028 - fixed bug (for single client case) when PWD was deleted
3029 - invalidate local directory pages when doing intent-based ops
3030 - avoid LDLM oops when lock callback contained bad data
3032 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
3034 * documentation updates
3035 * bug fixes and cleanups
3036 - configuration tools
3039 - 64-bit compile warnings
3040 - 64-bit internal statfs data
3042 * test_brw on persistent OST devices
3044 * lprocfs (disabled)
3046 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
3048 * documentation updates
3049 * bug fixes and cleanups
3050 * configuration tools
3052 2002-08-30 Peter J. Braam <braam@clusterfs.com>
3055 * many small fixes to 0_5_4
3056 * io/network handling
3057 * thinkos in MDS operations
3059 2002-08-24 Peter J. Braam <braam@clusterfs.com>
3062 * crucial basic fixes to 0.5.3
3063 * IOR, Iozone work over Elan
3066 2002-08-07 Phil Schwan <phil@clusterfs.com>
3067 * version 0_5_3, our first alpha
3068 * we use the new Portals iovs
3069 * documentation updates
3070 * bug fixes and cleanups
3071 * small changes in the DLM wire protocol
3073 2002-07-25 Peter J. Braam <braam@clusterfs.com>
3074 * version 0_5_1 with some initial stability,
3075 * locking on MD and file I/O.
3076 * documentation updates
3077 * several bug fixes since 0.5.0
3078 * small changes in wire protocol
3080 2002-07-18 Phil Schwan <phil@clusterfs.com>
3082 * delivered as Lustre Light Alpha
3083 * fixed a crash after handling invalid MDS requests
3084 * fixed directory pages for architectures with non-4k pages sizes
3086 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
3087 * release version v0_4_4
3088 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
3090 * Fixes a number of interruption problems with OST operations.
3091 * Update documentation for portals header changes
3092 * Move all wire protocol structs/defines to lustre_idl.h
3093 * Fixes symlink length bug.
3094 * Add tcpdump to repository.
3096 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
3097 * release version v0_4_3
3098 * Fixes statfs for inodes on extN.
3099 * Fixes bug in runtests which would delete /etc/hosts.
3100 * Use 64-bit object IDs wherever possible (not into VFS though)
3101 Remove ost_get_info, which is unused by lustre, and out of date.
3103 2002-07-03 Peter Braam <braam@clusterfs.com>
3104 * release version v0_4_2 Fixes a lookup error (type not passed)
3105 * move forward to head of Portals
3106 * move forward to latest Lustre kernel
3108 2002-06-25 Peter Braam <braam@clusterfs.com>
3109 * release version v0_4_1. Hopefully stable on single node use.