3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-220.17.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 2.6.32-220.17.1.el6 (RHEL6)
9 2.6.18-238.19.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
11 * Recommended e2fsprogs version: 1.41.90.wc4
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 Severity : enhancement
20 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
22 Severity : enhancement
24 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
26 Severity : enhancement
28 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
30 --------------------------------------------------------------------------------
32 04-30-2012 Whamcloud, Inc.
34 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
35 * Server support for kernels:
36 2.6.32-220.4.2.el6 (RHEL6)
37 * Client support for unpatched kernels:
38 2.6.32-220.4.2.el6 (RHEL6)
40 2.6.18-238.19.1.el5 (RHEL5)
41 2.6.16.60-0.69.1 (SLES10)
42 * Recommended e2fsprogs version: 1.41.90.wc4
43 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
44 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
45 stack overflow. For more information, please refer to bugzilla 17630.
46 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
47 the re-exporting nodes to be mounted with "32bitapi" mount option
49 Severity : enhancement
51 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
53 Severity : enhancement
55 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
57 Severity : enhancement
59 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
61 Severity : enhancement
63 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
65 --------------------------------------------------------------------------------
67 09-30-2011 Whamcloud, Inc.
69 * Server support for kernels:
70 2.6.18-238.19.1.el5 (RHEL5)
71 2.6.32-131.6.1.el6 (RHEL6)
72 * Client support for unpatched kernels:
73 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
74 2.6.18-238.19.1.el5 (RHEL5)
75 2.6.32-131.6.1.el6 (RHEL6)
76 2.6.29.4-167.fc11 (FC11)
77 2.6.16.60-0.69.1 (SLES10)
79 * Recommended e2fsprogs version: 1.41.90.wc3
80 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
81 Lustre filesystem with 4K stack may cause a stack overflow. For more
82 information, please refer to bugzilla 17630.
83 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
84 mounted with "32bitapi" mount option
86 Severity : enhancement
87 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
89 Severity : enhancement
90 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
92 Severity : enhancement
93 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
96 Frequency : only with SLES10
98 Description: Use OFED "KMP" provided by Novell
99 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
100 Lustre is now built against this package. That means you need to
101 install the ofed-kmp package from Novell for the patchless client.
102 Note that the ofed-kmp that Novell ships may not exactly match the
103 kernel version but should still be compatible.
105 Severity : enhancement
107 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
108 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
110 --------------------------------------------------------------------------------
112 2010-07-15 Oracle, Inc.
114 * Server support for kernels:
115 2.6.18-164.11.1.el5 (RHEL 5)
116 2.6.18-164.11.1.0.1.el5 (OEL 5)
117 * Client support for unpatched kernels:
118 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
119 2.6.18-164.11.1.el5 (RHEL 5),
120 2.6.18-164.11.1.0.1.el5 (OEL 5)
121 2.6.16.60-0.42.8 (SLES 10),
123 2.6.29.4-167.fc11 (FC11)
124 * Recommended e2fsprogs version: 1.41.10-sun2
126 Severity : enhancement
128 Description: Distribute lfs_migrate script with lustre
129 Details : The lfs_migrate mentioned in the manual is now included in the
130 lustre tree in lustre/scripts and will distributed in lustre RPMs.
132 Severity : enhancement
134 Description: Failover nid list added to import proc file.
136 Severity : enhancement
138 Description: "som_preview" mount option required to use SOM
139 Details : The clients will need to use som_preview mount option
143 Description: Lustre does not do 1MB IOs to HW RAID
144 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
146 Severity : enhancement
148 Description: lfs quota output is not very convenient for awk/sed-parsing
149 Details : Some positions in lfs quota output table could be empty or
150 non-empty which made it hard to parse it with scripts, now
151 a dash is put instead of space where there is not supposed
156 Description: Update to OFED 1.5.1
157 Details : OFED 1.5.1 will be used in environments that don't use
158 vendor supplied OFED.
162 Description: Remove lnet. prefix from lctl params display
163 Details : Also renamed lnet.memused to lnet_memused to
168 Description: Use conf_param -d to remove permanant settings.
172 Description: Add list_param -R
173 Details : lctl list_param now supports recursive mode.
176 Bugzilla : 21829, 22312
177 Description: Fix llobdstat.
178 Details : 1. Now updating the write_bytes counter
179 2. Port script fixes from 1.8.
183 Description: decrease the usage of memory on clients.
184 Details : 1. On clients, recycle dentries and inodes unused.
185 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
190 Description: llobdstat fix and enhancement
191 Details : add a counter to set a limit to how many samples will be returned
192 fix a wildcard in the path to limit to obdfilter stats only
196 Description: lfs check only allowed for root
197 Details : Code cleanups and added a regresion test of non-root use of
202 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
203 Details : Fix hr->hr_index overload.
207 Description: Fix FIEMAP functionality.
211 Description: FC11 patchless client support added
215 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
216 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
220 Description: New acc-sm script and various test-framework enhancments.
224 Description: Previously landed patch was reverted.
228 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
232 Description: Add ldiskfs maxdirsize mount option.
236 Description: Fix mkfs.lustre support for 16TB LUNs.
240 Description: Print mmp_check_interval
241 Details : print mmp_check_interval and make it possible to abort
242 mount operation in case it takes too long.
246 Description: Add cascading_rw to lustre mpi tests.
250 Description: Store SOM data in EA.
251 Details : Also several SOM related cleanups.
255 Description: Add refcount to md_open_data
259 Description: Use refcount on lov_request_set to prevent memory corruption.
263 Description: Statahead: use parent dir's i_mutex for synchronization.
267 Description: loadgen improvements.
271 Description: Enable peer health detection in ksocklnd and ko2iblnd.
275 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
279 Description: Fix race on obd_devcie:md_stats
283 Description: Fix SLES 11 client builds.
284 Details : Resolved uninitalized pointer gcc warnings.
288 Description: Allow non-root access for "lfs check".
289 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
293 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
295 Severity : enhancement
297 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
298 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
302 Description: Sanity test_27p fails.
303 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
304 Sanity tests 27 were cleaned up and improved.
308 Description: Unify req format on client/servers
309 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
310 buffers and swabbers approach..
315 Description: drop join file stuff.
317 Severity : enhancement
319 Description: Add 2.6.27 and SLES11 (patchless client) support
323 Description: Fix in ptlrpc_expire_one_request() to print the signed time
326 Severity : enhancement
328 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
331 Severity : enhancement
333 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
335 Severity : enhancement
337 Description: correctly handle big reply message.
338 Details : send LNet event if reply is bigger then buffer and adjust this buffer
343 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
345 Severity : enhancement
347 Description: Add support for OEL5.
351 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
353 Severity : enhancement
355 Description: "weak-modules" support
356 Details : Implement "weak-modules" support which enables kernel modules
357 to be used with any kernel that implements the same kABI. In
358 order to achieve this modules are now installed in
359 /lib/modules/$(uname -r)/updates/kernel on all distributions.
363 Description: client could not reconnect to OST because of an active request.
364 Details : The client now retries to reconnect to the same server, if a
365 connect request failed with EBUSY or -EAGAIN. Test case 35b
366 conf-sanity.sh has been created.
370 Description: Don't allow make backward step on assignin osc next id.
371 Details : race between allocation next id and ll_sync thread can be cause
372 of set wrong osc next id and can be kill valid ost objects.
376 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
380 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
384 Description: Error handling in osc_statfs_interpret() has been improved.
385 Details : Check in osc_statfs_interpret() for EBADR.
389 Description: Do not send statfs() requests to OSTs disabled by administrator.
390 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
392 Severity : enhancement
394 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
396 Severity : enhancement
398 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
400 Severity : enhancement
402 Description: Enhancement for lfs(1) command to use numeric uid/gid.
406 Description: Improved cleanup procedure for conf-sanity test_46a.
410 Description: Do not start bulk transfer, if client tried to reconnect.
411 Details : Do not start bulk transfer, if client tried to reconnect,
412 but failed due to active rpcs. Check for exp_abort_active_req
413 has been implemented.
417 Description: Prevent inconsistences between linux and lustre mount structures.
418 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
419 Make the sleep interruptible.
421 Severity : enhancement
423 Description: provide server to client comms path
424 Details : server to client communications path via new LDLM_SET_INFO rpc
428 Description: truncate starts GFP_FS allocation under transaction causing deadlock
429 Details : ldiskfs_truncate calls grab_cache_page which may start page
430 allocation under an open transaction. This may lead to
431 calling prune_icache with consequent lustre reentrance.
435 Description: send correctly lsm on open replay
436 Details : MDS is trust to LSM size on replay open, but client can set wrong size
441 Description: enable client interface failover
442 Details : When a child reconnects from another NID, properly update export
443 nid hash position and ldlm reverse import.
445 Severity : enhancement
447 Description: Communicate OST degraded/readonly state via statfs to MDS
448 Details : Flags in the statfs returned from OSTs indicate whether the
449 OST is in a degraded RAID state, or if the filesystem has
450 turned read-only after a filesystem error is detected.
455 Description: don't panic if EPROTO was hit when reading symlink
456 Details : correctly handling request reference in error cases.
458 Severity : enhancement
460 Description: Add LustreNetLink, a kernel-userspace communcation path.
462 Severity : enhancement
464 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
469 Description: access to llog context before init.
470 Details : move handling CATALOGS file at osc layer and forbid access to llog
474 Frequency : rare, only when client is writing to OST on local node
476 Description: Avoid deadlock for local client writes
477 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
478 memory freeing context. This allows OST threads to set the
479 PF_MEMALLOC flag on task structures in order to allocate memory
480 from reserved pools and complete IO.
481 Use GFP_HIGHUSER for OST allocations for non-local client writes,
482 so that the OST threads generate memory pressure and allow
483 inactive pages to be reclaimed.
485 Severity : enhancement
487 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
492 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
493 Details : move ldlm namespace creation in setup phase to avoid grab
494 _lprocfs_lock with cli_sem held.
498 Description: Temporarily disable grant shrink.
499 Details : Disable the feature for debugging.
502 Frequency : only when formatting test filesystems
504 Description: Unable to run several mkfs.lustre on loop devices at the same time
505 Details : mkfs.lustre returns error 256 on the concurrent loop devices
506 formatting. The solution is to proper handle the error.
508 Severity : enhancement
510 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
512 Severity : enhancement
514 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
519 Description: resolve race between obd_disconnect and class_disconnect_exports
520 Details : if obd_disconnect will be called to already disconnected export he
521 forget release one reference and osc module can't unloaded.
523 Severity : enhancement
525 Description: Allow tuning service thread via /proc
526 Details : For each service a new
527 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
528 created that can be used to set min/max thread counts, and get the
529 current number of running threads.
534 Descriptoin: don't return error if have partially created objects for file.
535 Details : lov_update_create_set uses set->set_success as index for created
536 objects, so if some requests will be failed, they will have hole
537 at end of array and qos_shrink_lsm for allocate correct lsm.
539 Severity : enhancement
541 Description: Update OFED support to 1.4.1
546 Description: Reduce small size read RPC
547 Details : Set read-ahead limit for every file and only do read-ahead when
548 available read-ahead pages are bigger than 1M to avoid small size
551 Severity : enhancement
553 Description: move AT tunable parameters for more consistent usage
554 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
556 Severity : enhancement
558 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
559 Details : allow skip disconnected ost for send statfs request and hide error
565 Description: fix racy locking of mballoc block bitmaps causing BUG
566 Details : The locking of the mballoc buddy bitmap and the in-memory
567 block bitmap was using two different spin locks in some
568 cases. This made it possible to incorrectly access the
569 mballoc bitmap while another process was modifying it,
570 causing a sanity assertion to fail. While no on-disk corruption
571 was reported, there was some risk of this happening.
574 Frequency : rare, on llog test 6
576 Descriptoin: don't allow connect to already connected import
577 Details : allowing connect to already connected import is hide connecting problem.
582 Descriptoin: allow kill process which wait statahead result
583 Details : for some reasons 'ls' can stick in waiting result from statahead,
584 in this case need way for kill this process.
586 Severity : enhancement
588 Description: Add state history info file, enhance import info file
589 Details : Track import connection state changes in a new osc/mdc proc file;
590 add overview-type data to the osc/mdc import proc file.
592 Severity : enhancement
594 Description: MDS create should not wait for statfs RPC while holding DLM lock.
596 Severity : enhancement
598 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
603 Descriptoin: OST grant too much space to client even there are not enough space.
604 Details : Client will shrink its grant cache to OST if there are no write
605 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
606 this grant cache if there are already not enough avaible space
607 (left_space < total_clients * 32M).
610 Frequency : start MDS on uncleanly shutdowned MDS device
612 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
613 Details : stay in waiting mds<>ost recovery finished produce random bugs
614 due race between two ll_sync thread for one lov target. send
615 ACTIVATE event only if connect realy finished and import have
619 Frequency : rare, connect and disconnect target at same time
621 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
622 Details : don't call obd_disconnect under lov_lock. this long time
623 operation and can block ptlrpcd which answer to connect request.
628 Descriptoin: don't lose wakeup for imp_recovery_waitq
629 Details : recover_import_no_retry or invalidate_import and import_close can
630 both sleep on imp_recovery_waitq, but we was send only one wakeup
634 Frequency : always with long access acl
636 Descriptoin: mds can't pack reply with long acl.
637 Details : mds don't control size of acl but they limited by reint/getattr
640 Severity : enhancement
642 Description: Update to SLES10 kernel-2.6.16.60-0.33.
644 Severity : enhancement
646 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
649 Frequency : start MDS on uncleanly shutdowned MDS device
651 Descriptoin: aborting recovery hang on MDS
652 Details : don't throttle destroy RPCs for the MDT.
655 Frequency : on remount
657 Description: external journal device not working after the remount
658 Details : clear dev_rdonly flag for external journal devices in
664 Description: shutdown vs evict race
665 Details : client_disconnect_export vs connect request race.
666 if client will evicted at this time - we start invalidate
667 thread without referece to import and import can be freed
671 Frequency : rare, need acl's on inode.
673 Description: client can't handle ost additional correctly
674 Details : if ost was added after client connected to mds client can have
675 hit lnet_try_match_md ... to big messages to wide striped files.
676 in this case need teach client to handle config events about add
677 lov target and update client max ea size at that event.
679 Severity : enhancement
681 Description: Changelogs
682 Details : Changelogs are a lightweight mechanism to track filesystem
683 metadata and namespace changes. The changelog is recorded
684 permanently on the MDTs, and is periodically "consumed" / purged
685 when records are no longer needed.
687 Severity : enhancement
689 Description: compact fld format with extents
690 Details : Store range of seq rather than every seq in FLD. Seq
691 controller update FLD rather than clients. In Case of CMD, mdt0
692 has FLD, all other metadata server act as non persistent proxy
693 for FLD queries and cache fld entries in fld cache.
698 Description: don't skip ost target if they assigned to file
699 Details : Drop slow OSCs if we can, but not for requested start idx.
700 This means "if OSC is slow and it is not the requested
701 start OST, then it can be skipped, otherwise skip it only
702 if it is inactive/recovering/out-of-space.
706 Description: more cleanup in mds_lov
707 Details : not send LOV EA under replay, we can't know about they size at this
708 time. Don't allow client connect to mds before any ost connected,
709 for avoid problems with LOV EA size and returning EIO to client.
711 Severity : enhancement
713 Description: Interoperability at server side (Disk interoperability)
715 Severity : enhancement
717 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
719 Severity : enhancement
721 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
723 Severity : enhancement
725 Description: New client IO stack (CLIO).
727 Severity : enhancement
729 Description: Commit on sharing. Eliminate inter-client dependencies between
730 uncommitted transactions by doing transaction commits.
731 Thereby clients may recovery independently.
734 Frequency : Create a symlink file with a very long name
736 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
737 Details : If there is no extra space in the request for early cancels,
738 ldlm_req_handles_avail() returns 0 instead of a negative value.
740 Severity : enhancement
742 Description: Add /proc entry for import status
743 Details : The mdc, osc, and mgc import directories now have
744 an import directory that contains useful import data for debugging
747 Severity : enhancement
749 Description: Re-disable certain /proc logging
750 Details : Enable and disable client's offset_stats, extents_stats and
751 extents_stats_per_process stats logging on the fly.
754 Frequency : Only on FC kernels 2.6.22+
756 Description: oops in statahead
757 Details : Do not drop reference count for the dentry from VFS when lookup,
758 VFS will do that by itself.
760 Severity : enhancement
762 Description: Generic /proc file permissions
763 Details : Set /Proc file permissions in a more generic way to enable non-
764 root users operate on some /proc files.
768 Description: Hitting mdc_commit_close() ASSERTION
769 Details : Properly handle request reference release in
770 ll_release_openhandle().
774 Description: quota recovery deadlock during mds failover
775 Details : This patch includes att18982, att18236, att18237 in bz14840.
777 1. fix osts hang when mds does failover with quotaon
778 2. prevent watchdog storm when osts threads wait for the
783 Frequency : only patchless client
784 Description: add workaround for race between add/remove dentry from hash
786 Severity : enhancement
788 Description: Allow OST glimpses to return PW locks
792 Description: LBUG when llog conf file is full
793 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
797 Description: Prevent import from entering FULL state when server in recovery
801 Description: service mount cannot take device name with ":"
802 Details : Only when device name contains ":/" will mount treat it as
808 Description: replace ptlrpcd with the statahead thread to interpret the async
809 statahead RPC callback
813 Frequency : on recovery
814 Description: I/O failures after umount during fail back
815 Details : if client reconnected to restarted server we need join to recovery
816 instead of find server handler is changed and process self
817 eviction with cancel all locks.
819 Severity : enhancement
821 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
823 Severity : enhancement
825 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
827 Severity : enhancement
829 Description: Upcall on Lustre log has been dumped
830 Details : Allow for a user mode script to be called once a Lustre log has
831 been dumped. It passes the filename of the dumped log to the
832 script, the location of the script can be specified via
833 /proc/sys/lnet/debug_log_upcall.
838 Description: avoid idr_remove called for id which is not allocated.
839 Details : Move assigment s_dev for clustered nfs to end of initialization,
840 to avoid problem with error handling.
845 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
846 Details : When connection is reused this not moved from CONN_UNUSED_HASH
847 into CONN_USED_HASH and this prodice warning when put connection
848 again in unused hash.
851 Severity : enhancement
853 Description: File striping can now be set to use an arbitrary pool of OSTs.
855 Severity : enhancement
857 Description: Export bytes_read/bytes_write count on OSC/OST.
861 Description: Early reply size mismatch, MGC loses connection
862 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
863 the connect flags are properly negotiated.
868 Description: quotas are not honored with O_DIRECT
869 Details : all writes with the flag O_DIRECT will use grants which leads to
870 this problem. Now using OBD_BRW_SYNC to guard this.
874 Description: add quota statistics
875 Details : 1. sort out quota proc entries and proc code.
876 2. add quota statistics
878 Severity : enhancement
880 Description: enable quota support for HEAD.
884 Description: Properly propagate oinfo flags from lov to osc for statfs
885 Details : restore missing copy oi_flags to lov requests.
887 Severity : enhancement
889 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
890 : l_getgroups(8), lst(8), routerstat(8)
891 Details : included man pages for llobdstat(8), llstat(8),
892 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
894 Severity : enhancement
896 Description: configure's --enable-quota should check the
897 : kernel .config for CONFIG_QUOTA
898 Details : configure is terminated if --enable-quota is passed but
899 : no quota support is in kernel
903 Description: Remove portals compatibility
904 Details : Remove portals compatibility, not interoperable with releases
909 Description: Resolve device initialization race
910 Details : Prevent proc handler from accessing devices added to the
911 obd_devs array but yet be intialized.
913 Severity : enhancement
915 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
917 Severity : enhancement
919 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
924 Description: Using wrong pointer in osc_brw_prep_request
925 Details : Access to array[-1] can produce panic if kernel compiled with
926 CONFIG_PAGE_ALLOC enabled
930 Description: Client runs out of low memory
931 Details : Consider only lowmem when counting initial number of llap pages
935 Description: *optional* service tags registration
936 Details : if the "service tags" package is installed on a Lustre node
937 When the filesystem is mounted, a local-node service tag will
938 be created. See http://inventory.sun.com/ for more information
939 about the Service Tags asset management system.
943 Description: Kernel BUG tries to release flock
944 Details : Lustre does not destroy flock lock before last reference goes
945 away. So always drop flock locks when client is evicted and
946 perform unlock regardless of successfulness of speaking to MDS.
950 Description: add refcount protection for osc callbacks, avoid panic on shutdown
954 Description: sanity test 65a fails if stripecount of -1 is set
955 Details : handle -1 striping on filesystem in ll_dirstripe_verify
960 Description: ASSERTION(CheckWriteback(page,cmd)) failed
961 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
964 Severity : enhancement
966 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
970 Description: do not process already freed flock
971 Details : flock can possibly be freed by another thread before it reaches
972 to ldlm_flock_completion_ast.
976 Description: LBUG during stress test
977 Details : Need properly lock accesses the flock deadlock detection list.
981 Description: oops in page fault handler
982 Details : kernel page fault handler can return two special 'pages' in error
983 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
987 Description: timeout with invalidate import.
988 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
989 should be handled by ptlrpcd. This produce long age waiting and
990 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
992 Severity : enhancement
994 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
998 Description: enable MGS and MDT services start separately
999 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1000 a co-located MGS without starting the MGS, which is a complement
1001 to 'nosvc' mount option.
1005 Frequency : after recovery
1006 Description: precreate to many object's after del orphan.
1007 Details : del orphan st in oscc last_id == next_id and this triger growing
1008 count of precreated objects. Set flag LOW to skip increase count
1009 of precreated objects.
1013 Frequency : rare, on clear nid stats
1014 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1015 Details : when clean nid stats sometimes try destroy live entry,
1016 and this produce panic in free.
1020 Description: Stack overflow during MDS log replay
1021 ease stack pressure by using a thread dealing llog_process.
1025 Description: wait until IO finished before start new when do lock cancel.
1026 Details : VM protocol want old IO finished before start new, in this case
1027 need wait until PG_writeback is cleared until check dirty flag and
1028 call writepages in lock cancel callback.
1030 Severity : enhancement
1032 Description: using special macro for print time and cleanup in includes.
1036 Description: mds_mfd_close() ASSERTION(rc == 0)
1037 Details : In mds_mfd_close(), we need protect inode's writecount change
1038 within its orphan write semaphore to prevent possible races.
1042 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1046 Frequency : rare, on shutdown ost
1047 Description: don't hit live lock with umount ost.
1048 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1049 use shrink_dcache_sb instead.
1053 Description: don't panic with use echo client
1054 Details : echo client pass NULL as client nid pointer and this produce null
1055 pointer dereference.
1059 Description: fix build on ppc32
1060 Details : compile code with -m64 flag produce wrong object file for ppc32.
1064 Description: add message levels for liblustreapi
1068 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1069 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1070 will result in an -EAGAIN and in the exit call path we call
1071 alloc_rr() which will with increasing aggressiveness attempt to
1072 aquire precreated objects on the minimum number of required OSCs.
1076 Description: Use old size assignment to avoid deadlock
1077 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1078 the scheduling while holding a spinlock. We do not need locking
1079 for size in ll_update_inode() because size is only updated from
1080 the MDS for directories or files without objects, so there is no
1081 other place to do the update, and concurrent access to such inodes
1082 are protected by the inode lock.
1086 Description: resolve "_IOWR redefined" build error on SLES10
1090 Description: dump the memory debugging after all modules are unloaded to
1091 suppress false negative in conf_sanity test 39
1093 Severity : enhancement
1095 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1100 Description: fix SLES kernel versioning
1101 Details : the kernel version for our SLES 10 kernel did not include a "-"
1102 before the "smp" at the end. while this was not a problem in
1103 general, it did mean that software trying to use the kernel
1104 version to try to detect a vendor specific kernel would fail.
1105 this was most evident by the OFED build scripts.
1109 Description: Don't update lov_desc members until making sure they are valid
1110 Details : When updating lov_desc members via proc fs, need fix their
1111 validities before doing the real update.
1115 Description: don't put request into delay list while invalidate in flight.
1116 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1117 import in flight. this produce timeout for invalidate and sometimes
1118 can cause stale data.
1122 Frequency : on ppc only
1123 Description: not convert ost objects for directory because it's not exist.
1124 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1128 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1130 Severity : enhancement
1132 Description: Update to support 2.6.22.14 vanilla kernel.
1136 Frequency : rare, on recovery
1137 Description: read procfs can produce deadlock in some situation
1138 Details : Holding lprocfs lock which send rpc can produce block for destroy
1139 obd objects and this also block reconnect with -EALREADY. This
1140 isn't fix all lprocfs bugs - but make it rare.
1142 Severity : enhancement
1144 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1147 Frequency : frequent on X2 node
1149 Description: mdc_set_open_replay_data LBUG
1150 Details : Set replay data for requests that are eligible for replay.
1154 Description: lustre_mgs: operation 101 on unconnected MGS
1155 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1156 MGC, and late on MGC cannot successfully connect to MGS and a lot
1157 of the error messages complaining that MGS is not connected.
1161 Frequency : on network error
1162 Description: panic with double free request if network error
1163 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1164 but it's true only for synchronus enqueue, for async enqueue
1165 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1168 Severity : enhancement
1170 Description: client-side metadata stat-ahead during readdir(directory readahead)
1171 Details : perform client-side metadata stat-ahead when the client detects
1172 readdir and sequential stat of dir entries therein
1175 Frequency : on start mds
1177 Description: Implement get_info(last_id) in obdfilter.
1180 Frequency : occasional
1182 Description: Correctly check stale fid, not start epoch if ost not support SOM
1183 Details : open with flag O_CREATE need set old fid in op_fid3 because
1184 op_fid2 was overwritten with new generated fid, but mds can answer
1185 with one of these two fids and both is not stale. Setattr
1186 incorrectly started an epoch and assume will be called
1187 done_writing, but without SOM done_writing ever being called.
1190 Frequency : rare, depends on device drivers and load
1192 Description: MDS or OSS nodes crash due to stack overflow
1193 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1194 In some cases, in conjunction with device drivers that use a lot
1195 of stack the MDS (or possibly OSS) service threads could overflow
1196 the stack. One change which was identified to consume additional
1197 stack has been reworked to avoid the extra stack usage.
1200 Frequency : occasional
1202 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1203 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1204 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1205 oscc_has_objects and see OSC has no objests and call
1206 oscc_internal_create to resend create request.
1208 Severity : enhancement
1210 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1212 Severity : enhancement
1214 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1217 Frequency : very rare
1219 Description: Fix replay if there is an un-replied request and open
1220 Details : In some cases, older replay request will revert the
1221 mcd->mcd_last_xid on MDS which is used to record the client's
1222 latest sent request.
1224 Severity : enhancement
1226 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1228 Severity : enhancement
1230 Description: Add rhel5 support to HEAD.
1232 Serverity : enhancement
1234 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1239 Description: Don't allow skipping OSTs if index has been specified.
1240 Details : Don't allow skipping OSTs if index has been specified, make
1241 locking in internal create lots better.
1245 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1246 Details : access to bitfield in structure is always rounded to long
1247 and this produce problem with not atomic change any bit.
1251 Description: Lustre make rpms failed.
1252 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1253 builds Lustre rpms from tarball.
1256 Frequency : rare on shutdown ost
1258 Description: If llog cancel was not send before clean_exports phase, this can
1259 produce deadlock in llog code.
1260 Details : If llog thread has last reference to obd and call class_import_put
1261 this produce deadlock because llog_cleanup_commit_master wait when
1262 last llog_commit_thread exited, but this never success because was
1263 called from llog_commit_thread.
1267 Description: allow userland application know is lost one of stripes.
1268 Details : fill lvb_blocks with error code on ost and return it to
1269 application if error flag found.
1273 Description: NULL lov_tgts causing MDS oops
1274 Details : more safe checks for NULL lov_tgts for avoid oops.
1276 Severity : enhancement
1278 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1282 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1283 Details : If an OST is down the MDS will hang indefinitely in
1284 obd_statfs_rqset() waiting for the statfs data. While for
1285 MDS QOS usage of statfs, it should not stuck in waiting.
1287 Severity : enhancement
1289 Description: remote_acl support
1290 Details : Support ACL-based permission check for remote user.
1291 Support setfacl/getfacl for remote user with the utils
1292 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1293 the system "{s,g}etfacl" utils.
1295 Severity : enhancement
1297 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1299 Severity : enhancement
1301 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1305 Description: make mgs_setparam() handle fsname containing dash
1306 Details : fsname containing a dash does not work with lctl conf_param
1308 Severity : enhancement
1310 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1312 Severity : enhancement
1314 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1318 Frequency : rare, at shutdown
1319 Description: access already free / zero obd_namespace.
1320 Details : if client_disconnect_export was called without force flag set,
1321 and exist connect request in flight, this can produce access to
1322 NULL pointer (or already free pointer) when connect_interpret
1323 store ocd flags in obd_namespace.
1327 Frequency : only at startup
1328 Description: not alloc memory with spinlock held.
1329 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1330 if any spinlock held.
1332 Severity : enhancement
1334 Description: make lustre randomly fail allocating memory
1335 Details : Make lustre randomly failed allocating memory for testing purpose.
1337 Severity : enhancement
1339 Description: lost problems with lov objid file
1340 Details : Fixes some scability and access to not inited memory problems
1341 in work with lov objdid file.
1346 Description: lfs find does not continue on file error
1347 Details : Continue other files processing when a file/dir is absent.
1351 Description: Inconsistent usage of lustre_pack_reply()
1352 Details : Standardize the usage of lustre_pack_reply() such that it
1353 always generate a CERROR on failure.
1356 Frequency : occasional
1358 Description: MDS hang or stay in waiting lock
1359 Details : If client receive lock with CBPENDING flag ldlm need send lock
1360 cancel as separate rpc, to avoid situation when cancel request
1361 can't processed due all i/o threads stay in wait lock.
1365 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1369 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1373 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1377 Description: Update to RHEL4 latest kernel.
1379 Severity : enhancement
1381 Description: Build SLES10 patchless client fails
1382 Details : The configure was broken by run ./configure with
1383 --with-linux-obj=.... argument for patchless client. When the
1384 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1385 can't search header adequately. Use absolute path such as
1386 -I($LINUX)/include instead.
1390 Description: interrupt oig_wait produce painc on resend.
1391 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1392 private set, and this produce situation when rq_ptlrpcd_data not
1393 copyed to new allocated request and triggered LBUG on assert
1394 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1395 ptlrpcd set if request is changed and can be safety changed to use
1400 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1401 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1402 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1403 these macros and replace them with their difinitions instead.
1407 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1408 Details : Update which_patch & target file for SLES10 latest kernel.
1411 Description: Remove client patches from SLES 10 kernel.
1412 Details : This causes SLES 10 clients to behave as patchless clients
1413 even on a Lustre-patched (server) kernel.
1415 Severity : enhancement
1417 Description: self-adjustable client's lru lists
1418 Details : use adaptive algorithm for managing client cached locks lru
1419 lists according to current server load, other client's work
1420 pattern, memory activities, etc. Both, server and client
1421 side namespaces provide number of proc tunables for controlling
1424 Severity : enhancement
1426 Description: light-weight GSS support
1427 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1428 and reduce performance overhead.
1430 Severity : enhancement
1432 Description: Linux keyring support
1433 Details : Support using service of Linux keyring for Lustre GSS internal
1434 context refresh/cache mechanism.
1438 Description: Fix errors in lfs documentation
1439 Details : Fixes man pages
1443 Description: don't use GFP_* in generic Lustre code.
1444 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1448 Description: obdclass is limited by single OBD_ALLOC(idarray)
1449 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1453 Description: warnings with build patchless client with vanila 2.6.19 and up
1454 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1455 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1459 Frequency : Only for SLES
1460 Description: sanity_quota fail test_1
1461 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1462 file, which makes TSTID[2] inunique.
1466 Frequency : Only for SLES9
1467 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1468 Details : Update target/ChangeLog/which_patch .
1472 Frequency : Only for test_10 in sanity-quota.sh
1473 Description: a bug in quota test code
1474 Details : checking a test flag in a wrong place
1478 Frequency : Only for test_13 in sanity-quota.sh
1479 Description: a bug in quota test script
1480 Details : shouldn't check overfull allocated quota
1484 Description: jbd statistics
1485 Details : Port older jbd statistics patch for sles10
1487 Severity : enhancement
1489 Description: Add printing inode info into message about error in writepage.
1494 Description: reply_lock_interpret crash due to race with it and lock cancel.
1495 Details : Do not replay locks that are being cancelled. Do not reference
1496 locks by their address during replay, just by their handle.
1500 Frequency : When flocks are used.
1501 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1502 Details : Flock locks might destroy just granted lock if it could be merged
1503 with another existing flock, this is done in completion handler,
1504 so teach ldlm_cli_enquque_fini that this is a valid case for
1510 Description: Oops in read and write path when failing to allocate lock.
1511 Details : Check if lock allocation failed and return error back.
1515 Description: lstripe command fails for valid OST index
1516 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1517 instead of lov->desc.ld_active_tgt_count.
1521 Description: sanity.sh failed test 103
1522 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1527 Description: df doesn't work properly if diskfs blocksize != 4K
1528 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1532 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1533 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1535 Severity : enhancement
1537 Description: Adaptive timeouts
1538 Details : RPC timeouts adapt to changing server load and network
1539 conditions to reduce resend attempts and improve recovery time.
1543 Description: llapi_file_create() does not allow some changes
1544 Details : add llapi_file_open() that allows specifying the mode and
1545 open flags, and also returns an open file handle.
1549 Description: merge and cleanup kernel patches.
1550 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1554 Description: Add journal checksum support.(Kernel part)
1555 Details : The journal checksum feature adds two new flags i.e
1556 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1557 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1558 indicates that the commit block contains the checksum for
1559 the blocks described by the descriptor blocks. Now commit
1560 record can be sent to disk without waiting for descriptor
1561 blocks to be written to disk. This behavior is controlled
1562 using JBD2_FEATURE_ASYNC_COMMIT flag.
1566 Description: OSS needs mutliple precreate threads
1567 Details : Add ability to start more than one create thread per OSS.
1571 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1572 Details : execute lfs setstripe on client
1576 Description: mds_obd_create error creating tmp object
1577 Details : When the user sets quota on root, llog will be affected and can't
1578 create files and write files.
1581 Frequency : Always on ia64 patchless client, and possibly others.
1583 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1584 Details : This allows the patchless client to be loaded on architectures
1585 without this export.
1589 Description: RedHat Update kernel for RHEL5
1590 Details : Add the kernel config file for RHEL5.
1594 Description: RedHat Update kernel for RHEL5
1595 Details : Modify the kernel config file more closer RHEL5.
1599 Description: Build failure against Centos5 (RHEL5)
1600 Details : Define PAGE_SIZE when it isn't present.
1604 Description: test_7 on sanity-quota.sh can't be used on separate machines
1605 Details : add facet to handle it
1609 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1610 Details : with NFS, the anon dentry's parent was set to itself in
1611 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1612 corresponding dentry other than use rec->ur_name.
1614 Severity : enhancement
1616 Description: lfs setstripe enhancement
1617 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1621 Description: enable data checksumming by default
1622 Details : enable checksum by default, allow --disable-checksum
1623 configure option and "-o nochecksum" mount option. Checksums
1624 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1628 Description: lustre support for RHEL5
1629 Details : Add support for RHEL5.
1633 Description: Client eviction due to failover config
1634 Details : after a connection loss, the lustre client should attempt to
1635 reconnect to the last active server first before trying the
1636 other potential connections.
1640 Description: when mds and osts use different quota unit(32bit and 64bit),
1641 quota will be released repeatly.
1642 Details : void sending multiple quota reqs to mds, which will keep the
1643 status between the reqs.
1647 Description: osts not allocated evenly to files
1648 Details : change the condition to increase offset_idx
1651 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1653 Description: Data corruption for OSTs that are formatted larger than 2TB
1655 Details : When generating the bio request for lustre file writes the
1656 sector number would overflow a temporary variable before being
1657 used for the IO. The data reads correctly from Lustre (which
1658 will overflow in a similar manner) but other file data or
1659 filesystem metadata may be corrupted in some cases.
1663 Description: Tune the kernel for good SCSI performance.
1664 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1665 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1670 Description: rewrite ext2-derived code in obdclass/uuid.c
1671 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1672 from scratch preserving functionality.
1676 Description: "lfs find -obd UUID" prints directories
1677 Details : "lfs find -obd UUID" will return all directory names instead
1678 of just file names. It is incorrect because the directories
1679 do not reside on the OSTs.
1683 Description: use i_size_read and i_size_write in 2.6 port
1684 Details : replace inode->i_size access with i_size_read/write()
1687 Frequency : when removing large files
1689 Description: scheduling issue during removal of large Lustre files
1690 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1691 It causes scheduling issues when removing large files (17TB in the
1695 Frequency : only with liblustre clients on XT3
1697 Description: evictions taking too long
1698 Details : allow llrd to evict clients directly on OSTs
1701 Frequency : only on ppc
1703 Description: /proc/fs/lustre/devices broken on ppc
1704 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1705 We should make sure the type of 'index' is loff_t and then cast
1706 explicitly as needed below. Do not assign an explicitly cast
1710 Frequency : only for rhel5
1712 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1713 Details : Modify the target file & which_kernel.
1715 Severity : enhancement
1717 Description: omit set fsid for export NFS
1718 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1719 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1723 Frequency : Always, for kernels after 2.6.16
1724 Description: Fix warning idr_remove called for id=.. which is not allocated.
1725 Details : Last kernels save old s_dev before kill super and not allow
1726 to restore from callback - restore it before call kill_anon_super.
1730 Description: buffer overruns could theoretically occur
1731 Details : llapi_semantic_traverse() modifies the "path" argument by
1732 appending values to the end of the origin string, and a
1733 overrun may occur. Adding buffer overrun check in liblustreapi.
1737 Description: Fix error on 'ls .' at the top of the Lustre mount.
1738 Details : Don't revalidate dentry if it is a root dentry.
1742 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1743 Details : Modify vm-tunables-rhel4.patch.
1747 Description: Kernel config for 2.6.18-vanilla.
1748 Details : Modify targets/2.6-vanilla.target.in.
1749 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1750 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1751 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1752 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1756 Description: improve handling recoverable errors
1757 Details : if request processig with error which can be recoverable on server
1758 request should be resend, otherwise page released from cache and
1763 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1764 Details : Modify target file & which_patch.
1765 A flaw was found in the IA32 system call emulation provided
1766 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1767 value could be stored in the %RAX register, which could trigger an
1768 out-of-bounds system call table access. An untrusted local user
1769 could exploit this flaw to run code in the kernel
1770 (ie a root privilege escalation). (CVE-2007-4573).
1774 Description: O_DIRECT bypasses client statistics.
1775 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1776 read_ahead_stats were updated. copied the stats section from
1777 osc_send_oap_rpc() into async_internal().
1781 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1785 Description: change order of libsysio includes
1786 Details : '#include sysio.h' should always come before '#include xtio.h'
1790 Description: handle "serious error: objid * already exists" more gracefully
1791 Details : If LAST_ID value on disk is smaller than the objects existing in
1792 the O/0/d* directories, it indicates disk corruption and causes an
1793 LBUG(). If the object is 0-length, then we should use the existing
1794 object. This will help to avoid a full fsck in most cases.
1796 Severity : enhancement
1798 Description: adapt the lustre_config script to support the upgrade case
1799 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1800 server targets to 1.6 in parallel.
1804 Description: To avoid grant space > avaible space when the disk is almost
1805 full. Without this patch you might see the error "grant XXXX >
1806 available" or some LBUG about grant, when the disk is almost
1808 Details : In filter_check_grant, for non_grant cache write, we should
1809 check the left space by if (*left > ungranted + bytes), instead
1810 of (*left > ungranted), because only we are sure the left space
1811 is enough for another "bytes", then the ungrant space should be
1812 increase. In client, we should update cl_avail_grant only there
1813 is OBD_MD_FLGRANT in the reply.
1817 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1818 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1819 described in CVE-2007-4573. This problem could allow untrusted
1820 local users to gain root access.
1823 Frequency : when using O_DIRECT and quotas
1825 Description: Incorrect file ownership on O_DIRECT output files
1826 Details : block usage reported by 'lfs quota' does not take into account
1827 files that have been written with O_DIRECT.
1832 Description: touch file failed when fs is not full
1833 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1834 otherwise we can get ENOSP while fs is not full.
1838 Description: parallel lock callbacks
1839 Details : Instead of sending blocking and completion callbacks as separated
1840 requests, adding them to a set and sending in parallel.
1843 Frequency : only for Cray XT3
1844 Bugzilla : 12829/13455
1845 Description: Changing primary group doesn't change the group lustre assigns to
1847 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1848 overridden with the primary group provided by the group upcall,
1849 whereas the supplied fsgid can be trusted if it is in the list of
1850 supplementary groups returned by the group upcall.
1852 Severity : enhancement
1854 Description: Allow masking D_WARNING, D_ERROR messages from console
1855 Details : Console messages can now be disabled via lnet.printk.
1859 Description: User code with malformed file open parameter crashes client node
1860 Details : Before packing join_file req, all the related reference should be
1861 checked carefully in case some malformed flags cause fake
1862 join_file req on client.
1866 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1867 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1868 mode is equal to the granted mode. However, it should also check
1869 that LDLM_FL_CP_REQD is not set.
1873 Description: Heavy nfs access might result in deadlocks
1874 Details : After ELC code landed, it is now improper to enqueue any mds
1875 locks under och_sem, because enqueue might want to decide to
1876 cancel open locks for same inode we are holding och_sem for.
1880 Description: Client eviction while running blogbench
1881 Details : A lot of unlink operations with concurrent I/O can lead to a
1882 deadlock causing evictions. To address the problem, the number of
1883 oustanding OST_DESTROY requests is now throttled to
1884 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1885 callbacks are processed in priority.
1889 Description: enable ACLs on MDS by default
1890 Details : ACLs must be enabled on MDS by default.
1893 Frequency : PPC/PPC64 only
1895 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1896 Details : fix duplicated definitions between asm-ppc64/types.h and
1897 lustre_types.h on PPC.
1900 Frequency : PPC/PPC64 only
1902 Description: asm-ppc/segment.h does not exist
1903 Details : fix compile issue on PPC.
1907 Description: better handle error messages in extents code
1910 Frequency : RHEL4 only
1912 Description: mkfs is very slow on IA64/RHEL4
1913 Details : A performance regression has been discovered in the MPT Fusion
1914 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1915 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1916 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1918 Severity : enhancement
1920 Description: SNMP support enhancement
1921 Details : Adding total number of sampled request for an MDS node in snmp
1924 Severity : enhancement
1926 Description: Optimize ldlm waiting list processing for PR extent locks
1927 Details : When processing waiting list for read extent lock and meeting read
1928 lock that is same or wider to it that is not contended, skip
1929 processing rest of the list and immediatelly return current
1930 status of conflictness, since we are guaranteed there are no
1931 conflicting locks in the rest of the list.
1935 Description: Time out and refuse to reconnect
1936 Details : When the failover node is the primary node, it is possible
1937 to have two identical connections in imp_conn_list. We must
1938 compare not conn's pointers but NIDs, otherwise we can defeat
1939 connection throttling.
1943 Description: port llog fixes from b1_6 into HEAD
1944 Details : Port llog reference couting and some llog cleanups from b1_6
1945 (bug 10800) into HEAD, for protect from panic and access to already
1946 free llog structures.
1950 Description: Detect stride IO mode in read-ahead
1951 Details : When a client does stride read, read-ahead should detect that and
1952 read-ahead pages according to the detected stride pattern.
1956 Description: data checksumming impacts single node performance
1957 Details : add support for several checksum algorithm. Currently, only CRC32
1958 and Adler-32 are supported. The checksum type can be changed on
1959 the fly via /proc/fs/lustre/osc/*/checksum_type.
1963 Description: use adler32 for page checksums
1964 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1969 Description: build for x2 fails
1970 Details : fix compile issue on Cray systems.
1974 Description: Properly match for duplicate locks
1975 Details : Due to different lock order from skiplists code, we need to
1976 traverse entire list for now
1979 Frequency : only on PPC/SLES10
1981 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1982 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1983 be defined. Add a hack in mkfs_lustre.c to work around this compile
1988 Description: LASSERT on MDS when client holding flock lock dies
1989 Details : ldlm pool logic depends on number of granted locks equal to
1990 number of released locks which is not true for flock locks, so
1991 just exclude such locks from consideration.
1995 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1996 Details : Use fsfilt_sync() for both the whole filesystem sync and
1997 individual file sync to eliminate dangerous inode locking
1998 with I_LOCK that can lead to a deadlock.
2002 Description: performance in 1.6.3
2003 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2004 sure that 1MB requests are not fragmented by the block layer.
2006 Severity : enhancement
2008 Description: organize the server-side client stats on per-nid basis
2009 Details : Change the structure of stats under obdfilter and mds to
2017 The "uuid"s file would list the uuids of _active_ exports.
2018 And the clear entry is to clear all stats and stale nids.
2020 Severity : enhancement
2022 Description: eliminate client locks in face of contention
2023 Details : file contention detection and lockless i/o implementation
2024 for contended files.
2028 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2031 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2033 Description: MDS may lose file striping (and hence file data) in some cases
2034 Details : If there are additional extended attributes stored on the MDS,
2035 in particular ACLs, SELinux, or user attributes (if user_xattr
2036 is specified for the client mount options) then there is a risk
2037 of attribute loss. Additionally, the Lustre file striping
2038 needs to be larger than default (e.g. striped over all OSTs),
2039 and an additional attribute must be stored initially in the
2040 inode and then increase in size enough to be moved to the
2041 external attribute block (e.g. ACL growing in size) for file
2046 Description: skiplist implementation simplification
2047 Details : skiplists are used to group compatible locks on granted list
2048 that was implemented as tracking first and last lock of each
2049 lock group the patch changes that to using doubly linked lists
2053 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2054 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2055 because LASSERT on both the data supplied by a client, and the
2056 data on disk is dangerous and incorrect.
2058 Severity : enhancement
2060 Description: Slow truncate/writes to huge files at high offsets.
2061 Details : Directly associate cached pages to lock that protect those pages,
2062 this allows us to quickly find what pages to write and remove
2063 once lock callback is received.
2067 Description: more ldlm soft lockups
2068 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2069 starve other threads from the resource lock for a long time in
2070 case of long waiting queue, so change the debug level from
2071 D_OTHER to the less frequently used D_INFO.
2073 Severity : enhancement
2075 Description: add -gid, -group, -uid, -user options to lfs find
2079 Description: Hung threads in invalidate_inode_pages2_range
2080 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2081 once one is completed. As a result, some RPCs are stuck in the
2082 queue and are never sent.
2086 Description: filter threads hungs on waiting journal commit
2087 Details : Cleanup filter group llog code, then only filter group llog will
2088 be only created in the MDS/OST syncing process.
2092 Description: Procfs and llog threads access destoryed import sometimes.
2093 Details : Sync the import destoryed process with procfs and llog threads by
2094 the import refcount and semaphore.
2096 Severity : enhancement
2098 Description: openlock cache of b1_6 port to HEAD
2103 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2104 Details : If insertion of an extent fails, then discard the inode
2105 preallocation and free data blocks else it can lead to duplicate
2110 Description: don't always update ctime in ext3_xattr_set_handle()
2111 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2112 In some cases the ctime should not be updated, for example for
2113 2.0->1.8 compatibility it is necessary to delete an xattr and it
2114 should not update the ctime.
2118 Bugzilla : 15713/16362
2119 Description: Assertion in iopen_connect_dentry in 1.6.3
2120 Details : looking up an inode via iopen with the wrong generation number can
2121 populate the dcache with a disconneced dentry while the inode
2122 number is in the process of being reallocated. This causes an
2123 assertion failure in iopen since the inode's dentry list contains
2124 both a connected and disconnected dentry.
2128 Description: assertion failure in ldlm_handle2lock()
2129 Details : fix a race between class_handle_unhash() and class_handle2object()
2130 introduced in lustre 1.6.5 by bug 13622.
2135 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2136 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2141 Description: oops in sd_iostats_seq_show()
2142 Details : unloading/reloading the scsi low level driver triggers a kernel
2143 bug when trying to access the sd iostat file.
2148 Description: Kernel panics during QLogic driver reload
2149 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2150 patch, causing memory corruption.
2155 Description: journal_dev option does not work in b1_6
2156 Details : pass mount option during pre-mount.
2158 Severity : enhancement
2160 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2161 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2162 extent information of a file. It can be used to map logical blocks
2163 in a file to physical blocks in the block device.
2167 Description: LDLM soft lockups - improvement
2168 Details : It is be possible to send the lock handle along with each read
2169 or write request because the client is already doing a lock match
2170 itself so there isn't any reason the OST should have to re-do that
2174 Frequency : only on Cray X2
2176 Description: X2 build failures
2177 Details : fix build failures on Cray X2.
2181 Description: xid & resent requests
2182 Details : Initialize RPC XID from clock at startup (randomly if clock is
2185 Severity : enhancement
2187 Description: Add lustre_start utility to start or stop multiple Lustre servers
2192 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2193 Details : In case of memory pressure, list_del() can be called twice on
2194 req->rq_history_list, causing a kernel oops.
2198 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2199 Details : fix stack overflow in the distributed lock manager by defering
2200 export eviction after a failed AST to the elt thread instead of
2201 handling it in the dlm interpret routine.
2205 Description: Convert some comments to new format.
2206 Details : Update documenting comments to match doxygen conventions.
2210 Description: Grammar fixes.
2211 Details : A couple of trivial spelling fixes.
2215 Description: OSD_COUNTERS-mandatory
2216 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2217 to simplify the code.
2221 Description: simplify cmm_device freeing logic.
2222 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2226 Description: Add lockdep support to dt_object_operations locking interface.
2227 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2228 parameter indicating lock ordering. Update mdd code to use new
2233 Description: Introduce failloc constants for lockless IO tests.
2234 Details : Add two new failloc constants to test lockless IO. Only one of
2235 them in implemented---another is checked in yet to be landed
2240 Description: Add lockdep support for inode mutex.
2241 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2242 in the situations where Lustre has to lock more than one inode
2247 Description: Add optional invariants checking support.
2248 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2249 configure switch. This macro is to be used for consistency and
2250 sanity checks that are too expensive to be left in `production'
2255 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2256 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2261 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2262 Details : Use spin_lock_nested() in (the only) situation where more than
2263 one ldlm_lock is locked simultaneously. Also, fix possible
2264 dead-lock in ldlm_lock_change_resource() by enforcing particular
2269 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2270 Details : Instead of specifying each ldlm_lock call-back through separate
2271 parameter, wrap them into struct ldlm_callback_suite.
2275 Description: Kill join_lru obd method and its callers.
2276 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2277 in memory---a requirement implemented through ->o_join_lru() obd
2278 method in HEAD. Remove this method and its users.
2282 Description: Add asynchronous ldlm ENQUEUE completion handler.
2283 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2284 case that stock ldlm_completion_ast() cannot handle as it waits
2285 until lock is granted. Introduce new ldlm_completion_ast_async()
2286 for this. Also comment ldlm_completion_ast().
2290 Description: ldlm_error <-> errno conversion.
2291 Details : Add functions to map (rather arbitrary) between LDLM error codes
2292 and standard errno values. CLIO needs this to prevent LDLM specific
2293 constants from escaping ldlm and osc.
2297 Description: Kill unused ldlm_handle2lock_ns() function.
2298 Details : Kill unused ldlm_handle2lock_ns() function.
2302 Description: Add lu_ref support to ldlm_lock
2303 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2304 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2305 and ldlm_resource. LDLM interface has to be changed, because code that
2306 releases a reference on a lock, has to "know" what reference this is.
2307 In the most frequent case
2309 lock = ldlm_handle2lock(handle);
2311 LDLM_LOCK_PUT(lock);
2313 no changes are required. When any other reference (received _not_ from
2314 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2315 instead of LDLM_LOCK_PUT().
2317 Arguably, changes are pervasive, and interface requires some discipline
2318 for proper use. On the other hand, it was very instrumental in finding
2319 a few leaked lock references.
2323 Description: Add ldlm_lock_addref_try().
2324 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2325 attempts to addref a lock that might be being canceled
2330 Description: Add ldlm_weigh_callback().
2331 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2332 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2333 hard-coded `number of pages' logic.
2337 Description: Add lockdep annotations to llog code.
2338 Details : Use appropriately tagged _nested() locking calls in the places
2339 where llog takes more than one ->lgh_lock lock.
2343 Description: Add loi_kms_set().
2344 Details : Wrap kms updates into a helper function.
2348 Description: Constify instances of struct lsm_operations.
2349 Details : Constify instances of struct lsm_operations.
2353 Description: lu_conf support.
2354 Details : On a server, a file system object is uniquely identified
2355 by a fid, which is sufficient to locate and load all object
2356 state (inode). On a client, on the other hand, more data are
2357 necessary instantiate an object. Change lu_object_find() and
2358 friends to take additional `lu_conf' argument describing object.
2359 Typically this includes layout information.
2363 Description: lu_context fixes.
2364 Details : Introduce new lu_context functions that are needed on the client
2365 side, where some system threads (ptlrpcd) are shared by multiple
2366 modules, and so cannot be stopped during module shutdown.
2370 Description: Add start and stop methods to lu_device_type_operations.
2371 Details : Introduce two new methods in lu_device_type_operations, that are
2372 invoked when first instance of a given type is created and last one
2373 is destroyed respectively. This is need by CLIO.
2377 Description: Add lu_ref support to struct lu_device.
2378 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2379 track leaked references.
2383 Description: Introduce lu_kmem_descr.
2384 Details : lu_kmem_descr and its companion interface allow to create
2385 and destroy a number of kmem caches at once.
2389 Description: Fix lu_object finalization race.
2390 Details : Fix a race between lu_object_find() finding an object and its
2391 concurrent finalization. This race is (most likely) not possible
2392 on the server, but might happen on the client.
2396 Description: Introduce lu_ref interface.
2397 Details : lu_ref is a debugging module allowing to track references to
2398 a given object. It is quite cpu expensive, and has to be
2399 explicitly enabled with --enable-lu_ref. See usage description
2404 Description: Factor lu_site procfs stats into a separate function.
2405 Details : Separate lu_site stats printing code into a separate function
2406 that can be reused on a client.
2410 Description: Constify instances of struct {lu,dt,md}_device_operations.
2411 Details : Constify instances of struct {lu,dt,md}_device_operations.
2415 Description: Introduce struct md_site and move meta-data specific parts of
2416 struct lu_site here.
2417 Details : Move md-specific fields out of struct lu_site into special struct
2418 md_site, so that lu_site can be used on a client.
2422 Description: Kill mdd_lov_destroy().
2423 Details : Remove unused mdd code.
2427 Description: Add st_block checking to multistat.c.
2428 Details : Add st_block checking to multistat.c.
2432 Description: Add lu_ref support to struct obd_device.
2433 Details : Add lu_ref tracking to obd_device.
2437 Description: Kill obd_set_fail_loc().
2438 Details : Remove unused code.
2442 Description: Add special type for ptlrpc_request interpret functions.
2443 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2444 there. Actual usage will be in CLIO.
2448 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2449 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2450 doesn't work with lockdep.
2454 Description: Add rwv.c test program.
2455 Details : New testing program exercising readv(2) and writev(2) (Qian).
2459 Description: Add sendfile.c test program.
2460 Details : New testing program exercising sendfile(2) (Jay).
2464 Description: Ratelimit a message that can be very frequent.
2465 Details : Ratelimit a memory allocation failure message that can
2470 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2471 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2475 Description: Liblustre build fixes.
2476 Details : Liblustre build fixes.
2480 Description: libcfs: add cfs_{need,cond}_resched() interface.
2481 Details : libcfs: add cfs_{need,cond}_resched() definition and
2482 implementations for Linux, NT, and liblustre.
2484 Severity : enhancement
2486 Description: More exported tunables for mballoc
2487 Details : Add support for tunable preallocation window and new tunables for
2488 large/small requests
2492 Description: Detect corruption of block bitmap and checking for preallocations
2493 Details : Checks validity of on-disk block bitmap. Also it does better
2494 checking of number of applied preallocations. When corruption is
2495 found, it turns filesystem readonly to prevent further corruptions.
2499 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2500 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2501 window to 0. If it does hit cache miss, and read-pattern is still
2502 stride-io mode, does not reset the stride window, but also does not
2503 increase the stride window length in this case.
2507 Frequency : only for big-endian servers
2508 Description: Check if system is big-endian while mounting fs with extents feature
2509 Details : Mounting a filesystem with extents feature will fail on big-endian
2510 systems since ext3-based ldiskfs is not supported on big-endian
2511 systems. This can be over-riden with "bigendian_extents" mount option.
2513 Severity : enhancement
2515 Description: The root squash functionality
2516 Details : A security feature, which is to prevent users from being able
2517 to mount lustre on their desktop, run as root, and delete
2518 all of the files in the filesystem. The goal is accomplished by
2519 remapping user id (UID) and group id (GID) of the root user to
2520 a UID and GID specified by the system administartor via Lustre
2521 configuration management server (MGS). The functionality also
2522 allows to specify sets of clients for which the remapping does
2527 Description: Excessive recovery window
2528 Details : With AT enabled, the recovery window can be excessively long (6000+
2529 seconds). To address this problem, we no longer use
2530 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2531 timeout no longer depends on the service time, it is set to
2532 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2533 time via pb_service_time.
2537 Description: Watchdog triggered on MDS failover
2538 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2539 the OSTs know that the MDS "UUID" can be reused for the same export
2540 from a different NID, so we do not need to wait for the export to be
2544 Frequency : rare, only if using MMP with Linux RAID
2546 Description: MMP doesn't work with Linux RAID
2547 Details : While using HA for Lustre servers with Linux RAID, it is possible
2548 that MMP will not detect multiple mounts. To make this work we
2549 need to unplug the device queue in RAID when the MMP block is being
2550 written. Also while reading the MMP block, we should read it from
2551 disk and not the cached one.
2553 Severity : enhancement
2555 Description: open file using fid
2556 Details : A file can be opened using just its fid, like
2557 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2560 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2563 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2564 Details : add proper locking for creation of jbd_4k slab cache
2568 Description: MMP check in ext3_remount() fails without displaying any error
2569 Details : When multiple mount protection fails during remount, proper error
2572 Severity : enhancement
2574 Description: Allow stripe size to be up to 4G-64k
2575 Details : Fix math logic to allow large stripe sizes.
2579 Description: add check for >8TB ldiskfs filesystems
2580 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2581 Don't allow >8TB ldiskfs filesystems to be mounted without
2582 force_over_8tb mount option
2587 Descriptoin: OSS complains "too many blocks" and crashes
2588 Details : In filter_commitrw_write():
2589 1. lquota_chkquota();
2590 2. obdo_from_inode();
2591 3. lquota_pending_commit();
2592 "2." has changed uid and gid in this case, then caused
2593 lquota_chkquota() and lquota_pending_commit() to use
2594 different uid and gid. That is the root of the bug.
2598 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2599 Details : Detect and print out the 'stripe count = -1' case properly.
2603 Description: Out or order replies might be lost on replay
2604 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2605 request with tid smaller then one currently being added, add it
2606 to the start, not end of the list.
2608 --------------------------------------------------------------------------------
2610 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2612 * Support for kernels:
2613 2.6.5-7.283 (SLES 9),
2614 2.6.9-55.EL (RHEL 4),
2615 2.6.16.46-0.14 (SLES 10),
2616 2.6.18.8 vanilla (kernel.org)
2617 * Client support for unpatched kernels:
2618 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2619 2.6.16 - 2.6.22 vanilla (kernel.org)
2620 * Due to recently discovered recovery problems, we do not recommend
2621 using patchless RHEL 4 clients with this or any earlier release.
2622 * Recommended e2fsprogs version: 1.39.cfs8
2623 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2624 * Starting with this release, the ldiskfs backing filesystem required
2625 by Lustre is now in its own package, lustre-ldiskfs. This package
2626 should be installed. It is versioned separately from Lustre and
2627 may be released separately in future.
2631 Description: block reactivating mgc import until all deactivates complete
2632 Details : Fix race when failing back MDT/MGS to itself (testing)
2635 Frequency : at statup only
2637 Description: mds_lov_synchronize race leads to various problems
2638 Details : simultaneous MDT->OST connections at startup can cause the
2639 sync to abort, leaving the OSC in a bad state.
2641 Severity : enhancement
2643 Description: add optional extra BUILD_VERSION info
2644 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2645 to override the lustre version.
2648 Frequency : 2.6.18 servers only
2650 Description: ll_kern_mount() doesn't release the module reference
2651 Details : The ldiskfs module reference count never drops down to 0
2652 because ll_kern_mount() doesn't release the module reference.
2657 Description: server LBUG when using old ost_num_threads parameter
2658 Details : Accept the old ost_num_threads parameter but warn that it
2659 is deprecated, and fix an off-by-one error that caused an LBUG.
2664 Description: Transient SCSI error results in persistent IO issue
2665 Details : iobuf->dr_error is not reinitialized to 0 between two
2669 Frequency : sometimes when underlying device returns I/O errors
2671 Description: OSTs not going read-only during write failures
2672 Details : OSTs are not remounted read-only when the journal commit threads
2673 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2674 instead of the ext3 wrappers.
2678 Description: poor connect scaling with increasing client count
2679 Details : Don't run filter_grant_sanity_check for more than 100 exports
2680 to improve scaling for large numbers of clients.
2683 Frequency : SLES10 only
2685 Description: sanity-quota.sh quotacheck failed: rc = -22
2686 Details : Quotas cannot be enabled on SLES10.
2689 Frequency : liblustre clients only
2691 Description: getdirentries does not give error when run on compute nodes
2692 Details : getdirentries does not fail when the size specified as an argument
2693 is too small to contain at least one entry
2695 Severity : enhancement
2697 Description: Add LNET router traceability for debug purposes
2698 Details : If a checksum failure occurs with a router as part of the
2699 IO path, the NID of the last router that forwarded the bulk data
2700 is printed so it can be identified.
2705 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2706 Details : Due to a race condition, liblustre clients were occasionally
2707 evicted incorrectly.
2709 Severity : enhancement
2711 Description: lfs setstripe use optional parameters instead of postional
2714 Severity : enhancement
2716 Description: Nanosecond timestamp support for ldiskfs
2717 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2718 resolution timestamps. There is not yet support for this at
2719 the Lustre filesystem level.
2722 Frequency : during server recovery
2724 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2725 Details : request with rq_no_resend flag not awake l_wait_event if they get
2729 Frequency : nfs export on patchless client
2731 Description: connectathon hang when test nfs export over patchless client
2732 Details : Disconnected dentry cannot be found with lookup, so we do not need
2733 to unhash it or make it invalid
2736 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2737 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2738 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2743 Description: open req refcounting wrong on reconnect
2744 Details : If reconnect happened between getting open reply from server and
2745 call to mdc_set_replay_data in ll_file_open, we will schedule
2746 replay for unreferenced request that we are about to free.
2747 Subsequent close will crash in variety of ways.
2748 Check that request is still eligible for replay in
2749 mdc_set_replay_data().
2754 Description: disable writes to filesystem when reading health_check file
2755 Details : the default for reading the health_check proc file has changed
2756 to NOT do a journal transaction and write to disk, because this
2757 can cause reads of the /proc file to hang and block HA state
2758 checking on a healthy but otherwise heavily loaded system. It
2759 is possible to return to the previous behaviour during configure
2760 with --enable-health-write.
2762 Severity : enhancement
2764 Description: 64-bit inode version
2765 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2766 to the inode. This will be required for version-based recovery.
2771 Description: MDS fails to start if a duplicate client export is detected
2772 Details : in some rare cases it was possible for a client to connect to
2773 an MDS multiple times. Upon recovery the MDS would detect this
2774 and fail during startup. Handle this more gracefully.
2776 Severity : enhancement
2778 Description: Add -o localflock option to simulate old noflock
2780 Details : This will achieve local-only flock/fcntl locks
2786 Description: log_commit_thread vs filter_destroy race leads to crash
2787 Details : Take import reference before releasing llog record semaphore
2792 Description: Wrong request locking in request set processing
2793 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2794 imp_delayed_list, in this place should be used imp_lock.
2797 Frequency : when reconnection
2799 Description: Grant Leak when osc reconnect to OST
2800 Details : When osc reconnect ost, OST(filter) should check whether it
2801 should grant more space to client by comparing fed_grant and
2802 cl_avail_grant, and return the granted space to client instead
2803 of "new granted" space, because client will call osc_init_grant
2804 to update the client grant space info.
2807 Frequency : when client reconnect to OST
2809 Description: Grant Leak when osc do resend and replay bulk write
2810 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2811 bulk write request, because the grant info while be sync between
2812 OSC and OST when reconnect, and we should ignore the grant info
2813 these of resend/replay write req.
2818 Description: Grant space more than avaiable left space sometimes.
2819 Details : When then OST is about to be full, if two bulk writing from
2820 different clients came to OST. Accord the avaliable space of the
2821 OST, the first req should be permitted, and the second one
2822 should be denied by ENOSPC. But if the seconde arrived before
2823 the first one is commited. The OST might wrongly permit second
2824 writing, which will cause grant space > avaiable space.
2827 Frequency : when client is evicted
2829 Description: Grant might be wrongly erased when osc is evicted by OST
2830 Details : when the import is evicted by server, it will fork another
2831 thread ptlrpc_invalidate_import_thread to invalidate the
2832 import, where the grant will be set to 0. While the original
2833 thread will update the grant it got when connecting. So if
2834 the former happened latter, the grant will be wrongly errased
2835 because of this race.
2840 Description: Checking Stale with correct fid
2841 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2842 to check whether it is stale, because sometimes, we want the
2843 enqueue happened anyway, and op_data.fid2 will not be initialized.
2845 Severity : enhancement
2847 Description: update patchless client
2848 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2851 Frequency : only with 2.4 kernel
2853 Description: random memory corruption
2854 Details : size of struct ll_inode_info is to big for union inode.u and this
2855 can be cause of random memory corruption.
2860 Description: Memory leak in recovery
2861 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2862 It should also check obd_fail before fsfilt_start, otherwise if
2863 fsfilt_start return -EROFS,(failover mds during mds recovery).
2864 then the req will return with repmsg->transno = 0 and rc = EROFS.
2865 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2866 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2867 be freed no matter whether fsfilt_commit success or not.
2870 Frequency : only with huge count clients
2872 Description: Prevents from taking the superblock lock in llap_from_page for
2874 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2875 for llap_from_page call in ll_removepage prevents from taking the
2876 superblock lock for a soon died page.
2881 Description: Not check open intent error before release open handle
2882 Details : in some rare cases, the open intent error is not checked before
2883 release open handle, which may cause
2884 ASSERTION(open_req->rq_transno != 0), because it tries to release
2885 the failed open handle.
2890 Description: Set cat log bitmap only after create log success.
2891 Details : in some rare cases, the cat log bitmap is set too early. and it
2892 should be set only after create log success.
2896 Description: Accessing a block bevice can re-enable I/O when Lustre is
2897 tearing down a device.
2898 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2903 Description: service threads may hog cpus when there are a lot of requests
2904 Details : Insert cond_resched to give other threads a chance to use some CPU
2909 Description: the cat log was not initialized in recovery
2910 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2911 unlink log on mds will not be initialized until mds post
2912 recovery. And also in mds post recovery, the unlink log will
2913 initialization will be done asynchronausly, so there will be race
2914 between add unlink log and unlink log initialization.
2918 Description: brw_stats were being printed incorrectly
2919 Details : brw_stats were being printed as log2 but all of them were not
2920 recorded as log2. Also remove some code duplication arising from
2921 filter_tally_{read,write}.
2925 Frequency : rare, only in recovery.
2926 Description: ASSERTION(req->rq_type != LI_POISON) failed
2927 Details : imp_lock should be held while iterating over imp_sending_list for
2928 prevent destroy request after get timeout in ptlrpc_queue_wait.
2932 Description: replay-single.sh test 52 fails
2933 Details : A lock's skiplist need to be cleanup when it being unlinked
2934 from its resource list.
2938 Description: Short directio read returns full requested size rather than
2940 Details : Direct I/O operations should return actual amount of bytes
2941 transferred rather than requested size.
2943 Severity : enhancement
2945 Description: metadata RPC reduction (e.g. for rm performance)
2946 Details : decrease the amount of synchronous RPC between clients and servers
2947 by canceling conflicing lock before the operation on the client
2948 and packing thier handles into the main operation RPC to server.
2950 Severity : enhancement
2952 Description: Async OSC create to avoid the blocking unnecessarily.
2953 Details : If a OST has no remain object, system will block on the creating
2954 when need to create a new object on this OST. Now, ways use
2955 pre-created objects when available, instead of blocking on an
2956 empty osc while others are not empty. If we must block, we block
2957 for the shortest possible period of time.
2961 Description: improve handling recoverable errors
2962 Details : if request processig with error which can be recoverable on server
2963 request should be resend, otherwise page released from cache and
2966 Severity : enhancement
2968 Description: refine locking for avoid write wrong info into lov_objid file
2969 Details : fix possible races with add new target and write/update data in
2972 --------------------------------------------------------------------------------
2974 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2979 Frequency : on some architectures
2981 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2982 Details : Uninitialized flags sometimes cause configuration commands to
2986 Frequency : patchless clients only
2988 Description: missing __iget() symbol export
2989 Details : The __iget() symbol export is missing. To avoid the need for
2990 this on patchless clients the deathrow inode reaper is turned
2991 off, and we depend on the VM to clean up old inodes. This
2992 dependency was during via the fix for bug 12181.
2994 --------------------------------------------------------------------------------
2996 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2998 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2999 INTEROPERATE with older versions automatically. In many cases a
3000 special upgrade step is needed. Please read the
3001 user documentation before upgrading any part of a live system.
3002 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3003 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3004 * WARNING: Lustre configuration and startup changes are required with
3005 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3009 Severity : enhancement
3011 Description: MountConf
3012 Details : Lustre configuration is now managed via mkfs and mount
3013 commands instead of lmc and lconf. New obd types (MGS, MGC)
3014 are added for dynamic configuration management. See
3015 https://mail.clusterfs.com/wikis/lustre/MountConf for
3018 Severity : enhancement
3020 Description: dynamic OST addition
3021 Details : OSTs can now be added to a live filesystem
3023 Severity : enhancement
3025 Description: startup order invariance
3026 Details : MDTs and OSTs can be started in any order. Clients only
3027 require the MDT to complete startup.
3029 Severity : enhancement
3031 Description: parallel, asynchronous orphan cleanup
3032 Details : orphan cleanup is now performed in separate threads for each
3033 OST, allowing parallel non-blocking operation.
3035 Severity : enhancement
3037 Description: optimized stripe assignment
3038 Details : stripe assignments are now made based on ost space available,
3039 ost previous usage, and OSS previous usage, in order to try
3040 to optimize storage space and networking resources.
3042 Severity : enhancement
3044 Description: Permanently set tunables
3045 Details : All writable /proc/fs/lustre tunables can now be permanently
3046 set on a per-server basis, at mkfs time or on a live system.
3048 Severity : enhancement
3050 Description: Lustre message v2
3051 Details : Add lustre message format v2.
3053 Severity : enhancement
3055 Description: client OST exclusion list
3056 Details : Clients can be started with a list of OSTs that should be
3057 declared "inactive" for known non-responsive OSTs.
3061 Description: SPEC SFS validation failure on NFS v2 over lustre.
3062 Details : Changes the blocksize for regular files to be 2x RPC size,
3063 and not depend on stripe size.
3065 Severity : enhancement
3067 Description: Multiple MD RPCs in flight.
3068 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3069 To avoid overly-overloading MDS, introduce a limit on number of
3070 MDS RPCs in flight for a single client and add /proc controls
3071 to adjust this limit.
3073 Severity : enhancement
3075 Description: client read/write statistics
3076 Details : Add client read/write call usage stats for performance
3077 analysis of user processes.
3078 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3079 file access. extents_stats shows chunk size distribution.
3080 extents_stats_per_process show chunk size distribution per
3083 Severity : enhancement
3085 Description: mds statistics
3086 Details : Add detailed mds operations statistics in
3087 /proc/fs/lustre/mds/*/stats.
3091 Description: Failure of copying files with lustre special EAs.
3092 Details : Client side always return success for setxattr call for lustre
3093 special xattr (currently only "trusted.lov").
3098 Description: Refcount LNET uuids
3099 Details : The global LNET uuid list grew linearly with every startup;
3100 refcount repeated list entries instead of always adding to
3103 Severity : enhancement
3105 Description: Dynamic service threads
3106 Details : Within a small range, start extra service threads
3107 automatically when the request queue builds up.
3110 Frequency : mixed-endian client/server environments
3112 Description: mixed-endian crashes
3113 Details : The new msg_v2 system had some failures in mixed-endian
3116 Severity : enhancement
3118 Description: Easy OST removal
3119 Details : OSTs can be permanently deactivated with e.g. 'lctl
3120 conf_param lustre-OST0001.osc.active=0'
3122 Severity : enhancement
3124 Description: MGS proc entries
3125 Details : Added basic proc entries for the MGS showing what filesystems
3128 Severity : enhancement
3130 Description: provide MGS failover
3131 Details : Added config lock reacquisition after MGS server failover.
3133 Severity : enhancement
3135 Description: add Linux 2.4 support
3136 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3140 Description: a large application tries to do I/O to the same resource and dies
3141 in the middle of it.
3142 Details : Check the req->rq_arrival time after the call to
3143 ost_brw_lock_get(), but before we do anything about
3144 processing it & sending the BULK transfer request. This
3145 should help move old stale pending locks off the queue as
3146 quickly as obd_timeout.
3149 Frequency : when an incorrect nid is specified during startup
3151 Description: ptlrpc connect to non-existant node causes kernel crash
3152 Details : LNET can't be re-entered from an event callback, which
3153 happened when we expire a message after the export has been
3154 cleaned up. Instead, hand the zombie cleanup off to another
3157 Severity : enhancement
3159 Description: plain/inodebits lock performance improvement
3160 Details : Grouping plain/inodebits in granted list by their request modes
3161 and bits policy, thus improving the performance of search through
3165 Frequency : only if OST filesystem is corrupted
3167 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3168 Details : for a short time RPCs with bulk IO are in the replay list,
3169 but replay of bulk IOs is unimplemented. If the OST filesystem
3170 is corrupted due to disk cache incoherency and then replay is
3171 started it is possible to trip an assertion. Avoid putting
3172 committed RPCs into the replay list at all to avoid this issue.
3175 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3177 Description: System hang on startup
3178 Details : This bug allowed the liblustre (e.g. catamount) client to
3179 return to the app before handling all startup RPCs. This
3180 could leave the node unresponsive to lustre network traffic
3181 and manifested as a server ptllnd timeout.
3183 Severity : enhancement
3185 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3186 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3187 boolean to control whether to print peer debug info when a
3188 client's RPC times out.
3191 Frequency : only for kernels with patches from Lustre below 1.4.3
3193 Description: Remove old rdonly API
3194 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3197 Frequency : only for devices with external journals
3199 Description: Set external device read-only also
3200 Details : During a commanded failover stop, we set the disk device
3201 read-only while the server shuts down. We now also set any
3202 external journal device read-only at the same time.
3205 Frequency : when upgrading from 1.4 while trying to change parameters
3207 Description: The wrong (new) MDC name was used when setting parameters for
3208 upgraded MDT's. Also allows changing of OSC (and MDC)
3209 parameters if --writeconf is specified at tunefs upgrade time.
3212 Frequency : when setting specific ost indicies
3214 Description: QOS code breaks on skipped indicies
3215 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3216 created with --index need not be sequential.
3218 Severity : enhancement
3220 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3221 Details : The uninit_groups feature works in conjunction with the kernel
3222 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3223 up the pass1 processing of e2fsck. This is a read-only feature
3224 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3225 mount filesystems that have had this feature enabled.
3227 Severity : enhancement
3229 Description: Improve multi-block allocation algorithm to avoid fragmentation
3230 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3231 allocation locality and avoid filesystem fragmentation.
3233 ------------------------------------------------------------------------------
3235 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3237 * Support for kernels:
3238 2.4.21-47.0.1.EL (RHEL 3)
3239 2.6.5-7.283 (SLES 9)
3240 2.6.9-42.0.10.EL (RHEL 4)
3241 2.6.12.6 vanilla (kernel.org)
3242 2.6.16.27-0.9 (SLES 10)
3243 * Recommended e2fsprogs version: 1.39.cfs5
3245 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3249 Frequency : occasional, depends on client load and configuration
3250 Bugzilla : 12181, 12203
3251 Description: data loss for recently-modified files
3253 Details : In some cases it is possible that recently written or created
3254 files may not be written to disk in a timely manner (this should
3255 normally be within 30s unless client IO load is very high).
3256 The problem appears as zero-length files or files that are a
3257 multiple of 1MB in size after a client crash or client eviction
3258 that are missing data at the end of the file.
3260 This problem is more likely to be hit on clients where files are
3261 repeatedly created and unlinked in the same directory, clients
3262 have a large amount of RAM, have many CPUs, the filesystem has
3263 many OSTs, the clients are rebooted frequently, and/or the files
3264 are not accessed by other nodes after being written.
3266 The presence of the problem can be detected by looking at
3267 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3268 smaller than the second (nr_unused) then dirty files will not
3269 be flushed automatically to disk. "sync; sleep 10" should be
3270 run several times on the node before unmounting it to update
3271 Lustre (this is also safe to run on nodes without this problem).
3273 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3274 that can cause this same problem, so customers using that kernel
3275 also need to update the kernel in addition to Lustre. In order
3276 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3278 It is normal that files written just before a client crash (less
3279 than 30s) may not yet have been flushed to disk, even for local
3283 Frequency : frequent on thin XT3 nodes
3285 Description: UUID collision on thin XT3 Linux nodes
3286 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3287 randomly, since we relied on an insufficiently-seeded PRNG.
3292 Description: OSS hangs after "All ost request buffers busy"
3293 Details : A deadlock between quota and journal operations caused OSS
3294 hangs after printing "All ost request buffers busy."
3297 Frequency : always on liblustre builds
3299 Description: Cleanup compiler warnings on liblustre
3302 Frequency : always on liblustre builds on XT3
3304 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3308 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3309 > 32000 subdirectories
3310 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3311 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3312 subdirectory count crosses 32000. This will aid e2fsck to
3313 correctly handle more than 32000 subdirectories.
3316 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3318 Description: System hang on startup
3319 Details : This bug allowed the liblustre (e.g. catamount) client to
3320 return to the app before handling all startup RPCs. This
3321 could leave the node unresponsive to lustre network traffic
3322 and manifested as a server ptllnd timeout.
3324 Severity : enhancement
3326 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3327 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3328 boolean to control whether to print peer debug info when a
3329 client's RPC times out.
3334 Description: make O_SYNC working on 2.6 kernels
3335 Details : 2.6 kernels use different method for mark pages for write,
3336 so need add a code to lustre for O_SYNC work.
3341 Description: Failure to close file and release space on NFS
3342 Details : Put inode details into lock acquired in ll_intent_file_open.
3343 Use mdc_intent_lock in ll_intent_open to properly
3344 detect all kind of errors unhandled by mdc_enqueue.
3349 Description: proc file read during shutdown sometimes raced obd removal,
3351 Details : Add lock to prevent obd access after proc file removal.
3354 Frequency : Only for files larger than 4GB on 32-bit clients.
3356 Description: improperly doing page alignment of locks
3357 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3358 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3363 Description: RPCs being resent when they shouldn't be
3364 Details : Some RPCs that should not be resent are being resent. This
3365 can cause inconsistencies in the RPC state machine. Do not
3366 resend such requests.
3369 Frequency : rare, only with NFS export
3371 Description: Crash on NFS re-export node
3372 Details : under very unusual load conditions an assertion is hit in
3373 ll_intent_file_open()
3376 Frequency : only if OST filesystem is corrupted
3378 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3379 Details : for a short time RPCs with bulk IO are in the replay list,
3380 but replay of bulk IOs is unimplemented. If the OST filesystem
3381 is corrupted due to disk cache incoherency and then replay is
3382 started it is possible to trip an assertion. Avoid putting
3383 committed RPCs into the replay list at all to avoid this issue.
3388 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3389 Details : Large single O_DIRECT read and write calls can fail to allocate
3390 a sufficiently large buffer to process the request. In case of
3391 allocation failure the allocation is retried with a smaller
3392 buffer and broken into smaller requests.
3394 Severity : enhancement
3396 Description: Add -o localflock option to simulate old noflock behaviour.
3397 Details : This will achieve local-only flock/fcntl locks coherentness.
3402 Description: versioning check is incomplete
3403 Details : Checking the version difference of client vs. server, report
3404 error if the gap is too big.
3409 Description: add support PG_writeback bit
3410 Details : add support for PG_writeback bit for Lustre, for more carefull
3411 work with page cache in 2.6 kernel. This also fix some deadlocks
3412 and remove hack for work O_SYNC with 2.6 kernel.
3414 Severity : enhancement
3416 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3417 Details : The uninit_groups feature works in conjunction with the kernel
3418 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3419 up the pass1 processing of e2fsck. This is a read-only feature
3420 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3421 mount filesystems that have had this feature enabled.
3423 Severity : enhancement
3425 Description: Improve multi-block allocation algorithm to avoid fragmentation
3426 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3427 allocation locality and avoid filesystem fragmentation.
3429 ------------------------------------------------------------------------------
3431 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3433 * Support for kernels:
3434 2.6.9-42.0.3.EL (RHEL 4)
3435 2.6.5-7.276 (SLES 9)
3436 2.4.21-47.0.1.EL (RHEL 3)
3437 2.6.12.6 vanilla (kernel.org)
3438 2.6.16.21-0.8 (SLES10)
3439 * Recommended e2fsprogs version: 1.39.cfs2-0
3441 * The backwards-compatible /proc/sys/portals symlink has been removed
3442 in this release. Before upgrading, please ensure that you change
3443 any configuration scripts or /etc/sysctl.conf files that access
3444 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3445 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3446 in advance of the upgrade on any system running Lustre 1.4.6 or
3447 newer, since /proc/sys/lnet was added in that version.
3448 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3452 Frequency : only when quota is used
3454 Description: avoid scanning export list for quota master
3455 Details : Change the algorithms to avoid scanning export list in order
3456 to improve the efficiency.
3459 Frequency : MDS failover only, very rarely
3461 Description: "went back in time" messages on mds failover
3462 Details : The greatest transno may be lost when the current operation
3463 finishes with an error (transno==0) and the client's last_rcvd
3464 record is over-written. Save the greatest transno in the
3465 mds_last_transno for this case.
3468 Frequency : always for specific kernels and striping counts
3470 Description: client may get "Matching packet too big" without ACL support
3471 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3472 when trying to access files in certain configurations. The
3473 clients should in fact be denied when mounting because they do
3474 not understand ACLs.
3477 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3479 Description: many clients connecting with IO in progress causes connect timeouts
3480 Details : Avoid synchronous journal commits to avoid delays caused by many
3481 clients connecting/disconnecting when bulk IO is in progress.
3482 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3483 OST_IO_PORTAL to avoid delays behind potentially many pending
3487 Frequency : occasionally with multiple writers to a single file
3489 Description: shared writes to file may result in wrong size reported by stat()
3490 Details : Allow growing of kms when extent lock is cancelled
3493 Frequency : always with random mmap IO to multi-striped file
3495 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3496 Details : Only if the hole is at the end of OST object so that kms is too
3497 small. Fix is to increase kms accordingly in ll_nopage.
3500 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3502 Description: writes to a missing object would leak memory on the OST
3503 Details : If there is an inconsistency between the MDS and OST filesystems,
3504 such that the MDS references an object that doesn't exist, writes
3505 to that object will leak memory due to incorrect cleanup in the
3506 error handling path, eventually running out of memory on the OST.
3511 Description: Creating too long symlink causes lustre errors
3512 Details : Check symlink and name lengths before sending requests to MDS.
3515 Frequency : only if flock is enabled (not on by default)
3517 Description: posix locks not released on fd closure on 2.6.9+
3518 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3519 kernels, this caused such locks not to be released on fd close and
3520 then assertions on fs unmount about still used locks.
3523 Frequency : MDS failover only, very rarely
3525 Description: clients may get ASSERTION(granted_lock != NULL)
3526 Details : When request was taking a long time, and a client was resending
3527 a getattr by name lock request. The were multiple lock requests
3528 with the same client lock handle and
3529 mds_getattr_name->fixup_handle_for_resent_request found one of the
3530 lock handles but later failed with ASSERTION(granted_lock != NULL).
3535 Description: handle->h_buffer_credits > 0, assertion failure
3536 Details : h_buffer_credits is zero after truncate, causing assertion
3537 failure. This patch extends the transaction or creates a new
3541 Frequency : NFS re-export or patchless client
3542 Bugzilla : 11179, 10796
3543 Description: Crash on NFS re-export node (__d_move)
3544 Details : We do not want to hash the dentry if we don't have a lock.
3545 But if this dentry is later used in d_move, we'd hit uninitialised
3546 list head d_hash, so we just do this to init d_hash field but
3547 leave dentry unhashed.
3550 Frequency : NFS re-export or patchless client
3552 Description: NFS exports has problem with symbolic link
3553 Details : lustre client didn't properly install dentry when re-exported
3554 to NFS or running patchless client.
3557 Frequency : NFS re-export or patchless client
3559 Description: Various nfs/patchless fixes.
3560 Details : fixes reuse disconected alias for lookup process - this fixes
3561 warning "find_exported_dentry: npd != pd",
3562 fix permission error with open files at nfs.
3563 fix apply umask when do revalidate.
3566 Frequency : occasional
3568 Description: Crash on NFS re-export node
3569 Details : calling clear_page() on the wrong pointer triggered oops in
3570 generic_mapping_read().
3573 Frequency : rarely, using O_DIRECT IO
3575 Description: unaligned directio crashes client with LASSERT
3576 Details : check for unaligned buffers before trying any requests.
3579 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3581 Description: stale data returned from RAID cache
3582 Details : If only a small amount of IO is done to the RAID device before
3583 reading it again it is possible to get stale data from the RAID
3584 cache instead of reading it from disk.
3587 Frequency : always for sles10 kernel
3589 Description: sles10 support
3590 Details : ll_follow_link: compile fixes and using of nd_set_link
3591 under newer kernels.
3594 Frequency : depends on arch, kernel and compiler version, always on sles10
3597 Description: recursive or deep enough symlinks cause stack overflow
3598 Details : getting rid of large stack-allocated variable in
3602 Frequency : depends on hardware
3604 Description: lustre write performance loss in the SLES10 kernel
3605 Details : the performance loss is caused by using of write barriers in the
3606 ext3 code. The SLES10 kernel turns barrier support on by
3607 default. The fix is to undo that change for ldiskfs.
3609 ------------------------------------------------------------------------------
3611 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3613 * Support for kernels:
3614 2.6.9-42.0.3EL (RHEL 4)
3615 2.6.5-7.276 (SLES 9)
3616 2.4.21-47.0.1.EL (RHEL 3)
3617 2.6.12.6 vanilla (kernel.org)
3621 Frequency : quota enabled and large files being deleted
3623 Description: releasing more than 4GB of quota at once hangs OST
3624 Details : If a user deletes more than 4GB of files on a single OST it
3625 will cause the OST to spin in an infinite loop. Release
3626 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3631 Description: statfs data retrieved from /proc may be stale or zero
3632 Details : When reading per-device statfs data from /proc, in the
3633 {kbytes,files}_{total,free,avail} files, it may appear
3634 as zero or be out of date.
3637 Frequency : systems with MD RAID1 external journal devices
3639 Description: lconf's call to blkid is confused by RAID1 journal devices
3640 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3641 of returning all block devices that match the journal UUID.
3644 Frequency : always, for aggregate stripe size over 4GB
3646 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3647 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3648 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3651 Frequency : always if "lfs find" used on a local file/directory
3653 Description: "lfs find" segfaults if used on a local file/directory
3654 Details : The case where a directory component was not specified wasn't
3655 handled correctly. Handle this properly.
3658 Frequency : always on ppc64
3660 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3661 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3662 perform endianness conversion.
3665 Frequency : rarely (truncate to non-zero file size after write under load)
3666 Bugzilla : 10730, 10687
3667 Description: Files padded with zeros to next 4K multiple
3668 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3669 it is possible that files that are truncated to a non-zero size
3670 immediately after being written are filled with zero bytes beyond
3671 the truncated size. No file data is lost.
3673 Severity : enhancement
3675 Description: Allow recovery/failover for liblustre clients.
3676 Details : liblustre clients were unaware of failover configurations until
3679 Severity : enhancement
3681 Description: user file locks should fail when not mounting with flock option
3682 Details : Set up an error-returning stub in ll_file_operations.lock field
3683 to prevent incorrect behaviour when client is mounted without
3684 flock option. Also, set up properly f_op->flock field for
3688 Frequency : always on ia64
3690 Description: "lfs df" loops on printing out MDS statfs information
3691 Details : The obd_ioctl_data was not initialized and in some systems
3692 this caused a failure during the ioctl that did not return
3693 an error. Initialize the struct and return an error on failure.
3696 Frequency : SLES 9 only
3698 Description: Error of copying files with lustre special EAs as root
3699 Details : Client side always return success for setxattr call for lustre
3700 special xattr (currently only "trusted.lov").
3703 Frequency : rarely on clusters with both ia64+i386 clients
3705 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3706 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3707 clients are doing concurrent writes to the same file it is
3708 possible that the ia64 clients may LASSERT because the OST
3709 extent locks are not PAGE_SIZE aligned. Ensure that grown
3710 locks are always aligned on the request boundary.
3713 Frequency : specific use, occasional
3715 Description: Overwriting in use executable truncates on-disk binary image
3716 Details : If one node attempts to overwrite an executable in use by
3717 another node, we now correctly return ETXTBSY instead of
3718 truncating the file.
3720 Severity : enhancement
3722 Description: Async OSC create to avoid the blocking unnecessarily.
3723 Details : If a OST has no remain object, system will block on the creating
3724 when need to create a new object on this OST. Now, ways use
3725 pre-created objects when available, instead of blocking on an
3726 empty osc while others are not empty. If we must block, we block
3727 for the shortest possible period of time.
3732 Description: chmod on Lustre root is propagated to other clients
3733 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3734 invalid by the follow_mount time.
3737 Frequency : liblustre clients only
3739 Description: Race in 'instant cancel' lock handling could lead to such locks
3740 never to be granted in case of SMP MDS
3741 Details : Do not destroy not yet granted but cbpending locks in
3745 Frequency : replay/resend of open
3747 Description: non null lock assetion failure in mds_intent_policy
3748 Details : Trying to replay/resend lockless open requests resulted in
3749 mds_open() returning 0 with no lock. Now it sets a flag if
3750 a lock is going to be returned.
3752 Severity : enhancement
3754 Description: Checksum enhancements
3755 Details : New checksum enhancements allow for resending RPCs that failed
3758 Severity : enhancement
3760 Description: Tunables on number of dirty pages in cacche
3761 Details : Allow to set limit on number of dirty pages cached.
3766 Description: client crash on unmount - lock still has references
3767 Details : In some error handling cases it was possible to leak a lock
3768 reference on a client while accessing a file. This was not
3769 harmful to the client during operation, but would cause the
3770 client to crash when the filesystem is unmounted.
3773 Frequency : specific case, rare
3775 Description: ETXTBSY on mds though file not in use by client
3776 Details : ETXTBSY is no longer incorrectly returned when attempting to
3777 chmod or chown a directory that the user previously tried to
3778 execute or a currently-executing binary.
3781 Frequency : extremely rare except on liblustre-based clients
3783 Description: Lustre space not freed when files are deleted
3784 Details : Clean up open-unlinked files after client eviction. Previously
3785 the unlink was skipped and the files remained as orphans.
3790 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3791 Details : In some cases it was possible to send a blocking callback to
3792 a client doing a glimpse, even though that client didn't get
3793 a lock granted. When the glimpse lock is cancelled on the OST
3794 the freed lock is left on the waiting list and corrupted the list.
3797 Frequency : all core dumps
3799 Description: Broke core dumps to lustre
3800 Details : Negative dentry may be unhashed if parent does not have UPDATE
3801 lock, but some callers, e.g. do_coredump, expect dentry to be
3802 hashed after successful create, hash it in ll_create_it.
3804 ------------------------------------------------------------------------------
3806 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3808 * Support for kernels:
3809 2.6.9-42.0.2.EL (RHEL 4)
3810 2.6.5-7.276 (SLES 9)
3811 2.4.21-40.EL (RHEL 3)
3812 2.6.12.6 vanilla (kernel.org)
3816 Frequency : always on RHEL 3
3818 Description: Number of open files grows over time
3819 Details : The number of open files grows over time, whether or not
3820 Lustre is started. This was due to a filp leak introduced
3821 by one of our kernel patches.
3823 ------------------------------------------------------------------------------
3825 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3827 * Support for kernels:
3828 2.6.9-42.EL (RHEL 4)
3829 2.6.5-7.276 (SLES 9)
3830 2.4.21-40.EL (RHEL 3)
3831 2.6.12.6 vanilla (kernel.org)
3836 Bugzilla : 5719, 9635, 9792, 9684
3837 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3838 Details : If a server is under heavy load and cannot reply to new
3839 connection requests before the client resends the (re)connect,
3840 the connection handling code can behave badly if two service
3841 threads are concurrently handing separate (re)connections from
3842 the same client. Add better locking to the connection handling
3843 code, and ensure that only a single connection will be processed
3844 for a given client UUID, even if the lock is dropped.
3846 Severity : enhancement
3848 Description: add TCP zero-copy support to kernel
3849 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3850 sends if the hardware supports scatter-gather and checksumming.
3851 This allows socklnd to do client-write and server-read more
3852 efficiently and reduce CPU utilization from skbuf copying.
3855 Frequency : only if NFS exporting from client
3857 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3858 Details : If setting attributes on a file created under NFS that had
3859 never been opened it would be possible to oops the client
3860 if the file had no objects.
3863 Frequency : always for liblustre
3865 Description: liblustre client does MDS+OSTs setattr RPC for each write
3866 Details : When doing a write from a liblustre client, the client
3867 incorrectly issued an RPC to the MDS and each OST the file was
3868 striped over in order to update the timestamps. When writing
3869 with small chunks and many clients this could overwhelm the MDS
3870 with RPCs. In all cases it would slow down the write because
3871 these RPCs are unnecessary.
3873 Severity : enhancement
3875 Description: allow number of MDS service threads to be changed at module load
3876 Details : It is now possible to change the number of MDS service threads
3877 running. Adding "options mds mds_num_threads={N}" to the MDS's
3878 /etc/modprobe.conf will set the number of threads for the next
3879 time Lustre is restarted (assuming the "mds" module is also
3880 reloaded at that time). The default number of threads will
3881 stay the same, 32 for most systems.
3886 Description: OST crash if filesystem is unformatted or corrupt
3887 Details : If an OST is started on a device that has never been formatted
3888 or if the filesystem is corrupt and cannot even mount then the
3889 error handling cleanup routines would dereference a NULL pointer.
3894 Description: NULL pointer deref in llap_from_page.
3895 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3896 incorrect) mapping if the page was truncated/reclaimed while it was
3897 searched for. Check for this condition and skip such pages when
3898 doing readahead. Introduce extra check to llap_from_page() to
3899 verify page->mapping->host is non-NULL (so page is not anonymous).
3902 Frequency : Sometimes when using sys_sendfile
3904 Description: "page not covered by a lock" warnings from ll_readpage
3905 Details : sendfile called ll_readpage without right page locks present.
3906 Now we introduced ll_file_sendfile that does necessary locking
3907 around call to generic_file_sendfile() much like we do in
3911 Frequency : with certain MDS communication failures at client mount time
3913 Description: NULL pointer deref after failed client mount
3914 Details : a client connection request may delayed by the network layer
3915 and not be sent until after the PTLRPC layer has timed out the
3916 request. If the client fails the mount immediately it will try
3917 to clean up before the network times out the request. Add a
3918 reference from the request import to the obd device and delay
3919 the cleanup until the network drops the request.
3922 Frequency : occasionally during client (re)connect
3924 Description: assertion failure during client (re)connect
3925 Details : processing a client connection request may be delayed by the
3926 client or server longer than the client connect timeout. This
3927 causes the client to resend the connection request. If the
3928 original connection request is replied in this interval, the
3929 client may trip an assertion failure in ptlrpc_connect_interpret()
3930 which thought it would be the only running connect process.
3933 Frequency : only with obd_echo servers and clients that are rebooted
3935 Description: kernel BUG accessing uninitialized data structure
3936 Details : When running an obd_echo server it did not start the ping_evictor
3937 thread, and when a client was evicted an uninitialized data
3938 structure was accessed. Start the ping_evictor in the RPC
3939 service startup instead of the OBD startup.
3941 Severity : enhancement
3942 Bugzilla : 10193 (patchless)
3943 Description: Remove dependency on various unexported kernel interfaces.
3944 Details : No longer need reparent_to_init, exit_mm, exit_files,
3945 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3948 Frequency : rare (only users of deprecated and unsupported LDAP config)
3950 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3951 Details : LDAP apparently contains 'lustreName' attributes instead of
3952 'name'. A simple remapping of the name is sufficient.
3955 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3957 Description: waiting_locks_callback trips kernel BUG if client is evicted
3958 Details : Running with the dump_on_timeout debug flag turned on makes
3959 it possible that the waiting_locks_callback() can try to dump
3960 the Lustre kernel debug logs from an interrupt handler. Defer
3961 this log dumping to the expired_lock_main() thread.
3963 Severity : enhancement
3965 Description: Support NFS exporting on 2.6 kernels.
3966 Details : Implement non-rawops metadata methods for NFS server to use without
3967 changing NFS server code.
3970 Frequency : very rare (synthetic metadata workload only)
3972 Description: two racing renames might cause an MDS thread to deadlock
3973 Details : Running the "racer" program may cause one MDS thread to rename
3974 a file from being the source of a rename to being the target of
3975 a rename at exactly the same time that another thread is doing
3976 so, and the second thread has already enqueued these locks after
3977 doing a lookup of the target and is trying to relock them in
3978 order. Ensure that we don't try to re-lock the same resource.
3981 Frequency : only very large systems with liblustre clients
3983 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3984 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3985 all OSTs in parallel. This allows the eviction of stale liblustre
3986 clients to proceed much faster than if they were done in series,
3987 and also offers similar improvements for other set_info RPCs.
3992 Description: excessive CPU usage during initial read phase on client
3993 Details : During the initial read phase on a client, it would agressively
3994 retry readahead on the file, consuming too much CPU and impacting
3995 performance (since 1.4.5.8). Improve the readahead algorithm
3996 to avoid this, and also improve some other common cases (read
3997 of small files in particular, where "small" is files smaller than
3998 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4003 Description: MDS crash when receiving packet with unknown intent.
4004 Details : Do not LBUG in unknown intent case, just return -EFAULT
4006 Severity : enhancement
4007 Bugzilla : 9293, 9385
4008 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4009 Details : Do not serialize getattr (non-intent version) and statfs.
4012 Frequency : occasional, when OST network is overloaded/intermittent
4014 Description: client evicted by OST after bulk IO timeout
4015 Details : If a client sends a bulk IO request (read or write) the OST
4016 may evict the client if it is unresposive to its data GET/PUT
4017 request. This is incorrect if the network is overloaded (takes
4018 too long to transfer the RPC data) or dropped the OST GET/PUT
4019 request. There is no need to evict the client at all, since
4020 the pinger and/or lock callbacks will handle this, and the
4021 client can restart the bulk request.
4024 Frequency : Always when mmapping file with no objects
4026 Description: client crashes when mmapping file with no objects
4027 Details : Check that we actually have objects in a file before doing any
4028 operations on objects in ll_vm_open, ll_vm_close and
4034 Description: Request leak when working with deleted CWD
4035 Details : Introduce advanced request refcount tracking for requests
4036 referenced from lustre intent.
4038 Severity : Enhancement
4040 Description: Cache open file handles on client.
4041 Details : MDS now will return special lock along with openhandle, if
4042 requested and client is allowed to hold openhandle, even if unused,
4043 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4044 closing files for every read/write openration.
4046 Severity : Enhancement
4048 Description: Cache open negative dentries on client when possible.
4049 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4050 negative dentries on lock revocation.
4055 Description: Remounting a client read-only wasn't possible with a zconf mount
4056 Details : It wasn't possible to remount a client read-only with llmount.
4058 Severity : enhancement
4059 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4060 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4061 MPICH is included to add Lustre-specific ADIO interfaces.
4062 This is based closely on the UFS ADIO layer and only differs
4063 in file creation, in order to allow the OST striping to be set.
4064 This is user-contributed code and not supported by CFS.
4069 Description: extended inode attributes (immutable, append-only) work improperly
4070 when 2.4 and 2.6 kernels are used on client/server or vice versa
4071 Details : Introduce kernel-independent values for these flags.
4073 Severity : enhancement
4076 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4077 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4078 granularity. Now, Fractional MB granularity is be supported,
4079 this is very useful for low memory system.
4081 Severity : enhancement
4083 Description: Getattr by fid
4084 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4085 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4086 have locks and avoids getattr rpc requests for it on every stat.
4089 Frequency : Always, for filesystems larger than 2TB
4091 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4092 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4093 filesystems larger than 2TB to be mounted with mballoc enabled.
4096 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4098 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4099 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4100 filesystems larger than 2TB is mounted then the kernel will
4101 silently corrupt the start of the filesystem. CONFIG_LBD is
4102 enabled for all CFS-supported kernels, but the possibility of
4103 this happening with a modified kernel config exists.
4105 Severity : enhancement
4107 Description: add client O_DIRECT support for 2.6 kernels
4108 Details : It is now possible to do O_DIRECT reads and writes to files
4109 in the Lustre client mountpoint on 2.6 kernel clients.
4111 Severity : enhancement
4113 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4114 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4115 parallel, not waiting for response from every OST before sending
4116 a rpc to the next OST.
4121 Description: setattr vs write race when updating file timestamps
4122 Details : Client processes that update a file timestamp into the past
4123 right after writing to the file (e.g. tar) it is possible that
4124 the updated file modification time can be reset to the current
4125 time due to a race between processing the setattr and write RPC.
4127 Severity : enhancement
4129 Description: Bring 'lfs find' closer in line with regular Linux find.
4130 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4131 -print0 options and obtains all the needed info through the lustre
4134 Severity : enhancement
4136 Description: support up to 1024 configured devices on one node
4137 Details : change obd_dev array from statically allocated to dynamically
4138 allocated structs as they are first used to reduce memory usage
4143 Description: Flush dirty partially truncated pages during truncate
4144 Details : Immediatelly flush partially truncated pages in filter_setattr,
4145 this way we completely avoid having any pages in page cache on OST
4146 and can retire ugly workarounds during writes to flush such pages.
4151 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4152 Details : protect inode from truncation within vfs_unlink() context
4153 just take a reference before calling vfs_unlink() and release it
4154 when parent's i_sem is free.
4159 Description: last_id value checked outside lock on OST caused LASSERT failure
4160 Details : If there were multiple MDS->OST object precreate requests in
4161 flight, it was possible that the OST's last object id was checked
4162 outside a lock and incorrectly tripped an assertion. Move checks
4163 inside locks, and discard old precreate requests.
4166 Frequency : always, if extents are used on OSTs
4168 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4169 Details : OSTs using the extents format would not zero the high 16 bits of
4170 the index physical block number. This is not a problem for any
4171 OST filesystems smaller than 16TB, and no kernels support ext3
4172 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4173 new/modified files) and can be fixed for existing filesystems
4174 with e2fsprogs-1.39-cfs1.
4179 Description: import connection selection may be incorrect if timer wraps
4180 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4181 import connections to be ignored if the 32-bit jiffies counter
4182 wraps. Use a 64-bit jiffies counter.
4185 Frequency : very large clusters immediately after boot
4187 Description: LNET request buffers exhausted under heavy short-term load
4188 Details : If a large number of client requests are generated on a service
4189 that has previously never seen so many requests it is possible
4190 that the request buffer growth cannot keep up with the spike in
4191 demand. Instead of dropping incoming requests, they are held in
4192 the LND until the RPC service can accept more requests.
4195 Frequency : Sometimes during replay
4197 Description: Assertion failure in ll_local_open after replay.
4198 Details : If replay happened on an open request reply before we were able
4199 to set replay handler, reply will become not swabbed tripping the
4200 assertion in ll_local_open. Now we set the handler right after
4201 recognising of open request
4204 Frequency : very rare
4206 Description: kernel reports "badness in vsnprintf"
4207 Details : Reading from the "recovery_status" /proc file in small chunks
4208 may cause a negative length in lprocfs_obd_rd_recovery_status()
4209 call to vsnprintf() (which is otherwise harmless). Exit early
4210 if there is no more space in the output buffer.
4212 Severity : enhancement
4214 Description: clear OBD RPC statistics by writing to them
4215 Details : It is now possible to clear the OBD RPC statistics by writing
4216 to the "stats" file.
4221 Description: Client mtime is not the same on different clients after utimes
4222 Details : In some cases, the client was using the utimes() syscall on
4223 a file cached on another node. The clients now validate the
4224 ctime from the MDS + OSTs to determine which one is right.
4229 Description: Inability to activate failout mode
4230 Details : lconf script incorrectly assumed that in pythong string's numeric
4231 value is used in comparisons.
4234 Frequency : always with multiple stripes per file
4236 Description: Inefficient object allocation for mutli-stripe files
4237 Details : When selecting which OSTs to stripe files over, for files with
4238 a stripe count that divides evenly into the number of OSTs,
4239 the MDS is always picking the same starting OST for each file.
4240 Return the OST selection heuristic to the original design.
4245 Description: mount failures may take full timeout to return an error
4246 Details : Under some heavy load conditions it is possible that a
4247 failed mount can wait for the full obd_timeout interval,
4248 possibly several minutes, before reporting an error.
4249 Instead return an error as soon as the status is known.
4251 Frequency : quota enabled and large files being deleted
4253 Description: releasing more than 4GB of quota at once hangs OST
4254 Details : If a user deletes more than 4GB of files on a single OST it
4255 will cause the OST to spin in an infinite loop. Release
4256 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4261 Description: statfs data retrieved from /proc may be stale or zero
4262 Details : When reading per-device statfs data from /proc, in the
4263 {kbytes,files}_{total,free,avail} files, it may appear
4264 as zero or be out of date.
4267 Frequency : systems with MD RAID1 external journal devices
4269 Description: lconf's call to blkid is confused by RAID1 journal devices
4270 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4271 of returning all block devices that match the journal UUID.
4274 Frequency : always, for aggregate stripe size over 4GB
4276 Description: assertion fails when trying to use 4GB stripe size
4277 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4278 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4281 Frequency : always on ppc64
4283 Description: the first write on an ext3 filesystem with mballoc got stuck
4284 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4285 perform endianness conversion.
4287 ------------------------------------------------------------------------------
4289 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4291 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4292 INTEROPERATE with older versions automatically. Please read the
4293 user documentation before upgrading any part of a live system.
4294 * WARNING: Lustre networking configuration changes are required with
4295 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4298 * Support for kernels:
4299 2.6.9-22.0.2.EL (RHEL 4)
4300 2.6.5-7.244 (SLES 9)
4301 2.6.12.6 vanilla (kernel.org)
4304 Severity : enhancement
4305 Bugzilla : 7981/8208
4306 Description: Introduced Lustre Networking (LNET)
4307 Details : LNET is new networking infrastructure for Lustre, it includes
4308 a reorganized network configuration mode (see the user
4309 documentation for full details) as well as support for routing
4310 between different network fabrics. Lustre Networking Devices
4311 (LNDs) for the supported network fabrics have also been
4312 created for this new infrastructure.
4314 Severity : enhancement
4315 Description: Introduced Access control lists
4316 Details : clients can set ACLs on files and directories in order to have
4317 more fine-grained permissions than the standard Unix UGO+RWX.
4318 The MDS must be started with the "-o acl" mount option.
4320 Severity : enhancement
4321 Description: Introduced filesystem quotas
4322 Details : Administrators may now establish per-user quotas on the
4325 Severity : enhancement
4327 Description: Configuration change for the XT3
4328 The PTLLND is now used to run Lustre over Portals on the XT3
4329 The configure option(s) --with-cray-portals are no longer used.
4330 Rather --with-portals=<path-to-portals-includes> is used to
4331 enable building on the XT3. In addition to enable XT3 specific
4332 features the option --enable-cray-xt3 must be used.
4337 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4338 Details : A race between allocating a new llcd and re-getting the llcd_lock
4339 allowed another thread to grab newly-allocated llcd.
4341 Severity : enhancement
4343 Description: 2.6 OST async journal commit and locking fix to improve performance
4344 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4345 2.6 kernels are now async as they already were in 2.4 kernels so
4346 that they can commit concurrently with the network bulk transfer.
4347 For block-allocated files the filter allocation semaphore is held
4348 to avoid filesystem fragmentation during allocation. BKL lock
4349 removed for 2.6 xattr operations where it is no longer needed.
4354 Description: lconf incorrectly determined whether two IP networks could talk
4355 Details : In some more complicated routing and multiple-network
4356 configurations, lconf will avoid trying to make a network
4357 connection to a disjoint part of the IP space. It was doing the
4358 math incorrectly for one set of cases.
4363 Description: Fix for potential infinite loop processing records in an llog.
4364 Details : If an llog record is corrupted/zeroed, it is possible to loop
4365 forever in llog_process(). Validate the llog record length
4366 and skip the remainder of the block on error.
4369 Frequency : occasional (liblustre only)
4371 Description: liblustre could not open files whose last component is a symlink
4372 Details : sysio_path_walk() would incorrectly pass the open intent to
4373 intermediate path components.
4376 Frequency : rare (liblustre only with non-standard tuning)
4377 Bugzilla : 7201 (7350)
4378 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4379 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4380 after it has been released, only do this for referenced locks
4382 Severity : enhancement
4384 Description: specifying an (invalid) directory default stripe_size of -1
4385 would reset the directory default striping
4386 Details : stripe_size -1 was used internally to signal directory stripe
4387 removal, now use "all default" to signal dir stripe removal
4388 as a directory striping of "all default" is not useful
4391 Frequency : common for large clusters running liblustre clients
4393 Description: doing an ls when liblustre clients are running is slow
4394 Details : sending a glimpse AST to a liblustre client waits for every AST
4395 to time out, as liblustre clients will not respond. Since they
4396 cannot cache data we refresh the OST lock LVB from disk instead.
4398 Severity : enhancement
4400 Description: doing an ls at the same time as file IO can be slow
4401 Details : enqueue and other "small" requests can be blocked behind many
4402 large IO requests. Create a new OST IO portal for non-IO
4403 requests so they can be processed faster.
4406 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4408 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4409 created file with mode 000
4410 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4411 allow an ftruncate() as a non-root user to a file with mode 000.
4412 Lustre now respects this flag to disable mode checks when
4413 truncating a file owned by the user
4416 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4419 Description: Revoking locks from clients that went dead or catatonic might take
4421 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4422 cancellation of such locks instant on servers without waiting for
4423 any reply from clients. Clients drops these locks when cancel
4424 notification from server is received without replying.
4427 Frequency : liblustre-only, when liblustre client dies or becomes busy
4429 Description: Doing ls on Linux clients can take a long time with active
4431 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4432 granting such locks to it in the first place if possible. Locks
4433 are taken by proxy on the OST during the read or write and
4434 dropped immediately afterward. Add connect flags handling, do
4435 not grant locks to liblustre clients for glimpse ASTs.
4437 Severity : enhancement
4439 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4440 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4441 sequential reads and behaves badly with applications doing random
4442 reads. Improve it by reading ahead at least read region, and
4443 avoiding excessive large RPC for small reads.
4445 Severity : enhancement
4447 Description: Creating more than 1000 files for a single job may cause a load
4448 imbalance on the OSTs if there are also a large number of OSTs.
4449 Details : qos_prep_create() uses an OST index reseed value that is an
4450 even multiple of the number of available OSTs so that if the
4451 reseed happens in the middle of the object allocation it will
4452 still utilize the OSTs as uniformly as possible.
4457 Description: OST or MDS may oops in ping_evictor_main()
4458 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4459 but doesn't restart at beginning of obd_exports_timed list
4462 Severity : enhancement
4464 Description: improve by-nid export eviction on the MDS and OST
4465 Details : allow multiple exports with the same NID to be evicted at one
4466 time without re-searching the exports list.
4469 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4471 Description: MDS may oops in groups_free()
4472 Details : in rare race conditions a newly allocated group_info struct is
4473 freed again, and this can be NULL. The 2.4 compatibility code
4474 for groups_free() checked for a NULL pointer, but 2.6 did not.
4477 Frequency : common for liblustre clients doing little filesystem IO
4478 Bugzilla : 9352, 7313
4479 Description: server may evict liblustre clients accessing contended locks
4480 Details : if a client is granted a lock or receives a completion AST
4481 with a blocking AST already set it would not reply to the AST
4482 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4485 Frequency : lfs setstripe, only systems with more than 160 OSTs
4487 Description: unable to set striping with a starting offset beyond OST 160
4488 Details : llapi_create_file() incorrectly limited the starting stripe
4489 index to the maximum single-file stripe count.
4492 Frequency : LDAP users only
4494 Description: lconf did not handle in-kernel recovery with LDAP properly
4495 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4497 Severity : enhancement
4499 Description: bind OST threads to NUMA nodes to improve performance
4500 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4501 node and do their allocations there to localize memory access
4503 Severity : enhancement
4505 Description: llmount can determine client NID directly from Myrinet (GM)
4506 Details : the client NID code from gmnalnid was moved directly into
4507 llmount, removing the need to use this or specifying the
4508 client NID explicitly when mounting GM clients with zeroconf
4511 Frequency : if client is started with down MDS
4513 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4514 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4515 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4516 ptlrpc_import_delayed_req() because we didn't check if the
4517 request was interrupted, and we also didn't break out of the
4518 event loop if there was no timeout
4523 Description: data loss during non-page-aligned writes to a single file from
4524 both multiple nodes and multiple threads on one node at same time
4525 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4526 inconsistency led to false short-reads, that were cached and later
4527 used by ->prepare_write() to fill in partially written page,
4528 leading to data loss.
4531 Frequency : always, if lconf --abort_recovery used
4533 Description: lconf --abort_recovery fails with 'Operation not supported'
4534 Details : lconf was attempting to abort recovery on the MDT device and not
4537 Severity : enhancement
4539 Description: remove cleanup logs
4540 Details : replace lconf-generated cleanup logs with lustre internal
4541 cleanup routines. Eliminates the need for client-cleanup and
4544 Severity : enhancement
4546 Description: add support for EAs (user and system) on lustre filesystems
4547 Details : it is now possible to store extended attributes in the Lustre
4548 client filesystem, and with the user_xattr mount option it
4549 is possible to allow users to store EAs on their files also
4551 Severity : enhancement
4553 Description: Add possibility (config option) to show minimal available OST free
4555 Details : When compiled with --enable-mindf configure option, statfs(2)
4556 (and so, df) will return least minimal free space available from
4557 all OSTs as amount of free space on FS, instead of summary of
4558 free spaces of all OSTs.
4560 Severity : enhancement
4562 Description: do not expand extent locks acquired on OST-side
4563 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4564 by server: they are not cached anyway.
4567 Frequency : when mmap is used/binaries executed from Lustre
4569 Description: Unmmap pages before throwing them away from read cache.
4570 Details : llap_shrink cache now attempts to unmap pages before discarding
4571 them (if unmapping failed - do not discard). SLES9 kernel has
4572 extra checks that trigger if this unmapping is not done first.
4577 Description: lconf didn't resolve symlinks before checking to see whether a
4578 given mountpoint was already in use
4581 Frequency : when migrating failover services
4582 Bugzilla : 6395, 9514
4583 Description: When migrating a subset of services from a node (e.g. failback
4584 from a failover service node) the remaining services would
4585 time out and evict clients.
4586 Details : lconf --force (implied by --failover) sets the global obd_timeout
4587 to 5 seconds in order to quickly disconnect, but this caused
4588 other RPCs to time out too quickly. Do not change the global
4589 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4591 Severity : enhancement
4592 Frequency : if MDS is started with down OST
4593 Bugzilla : 9439,5706
4594 Description: Allow startup/shutdown of an MDS without depending on the
4595 availability of the OSTs.
4596 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4597 Add appropriate locking and lov-osc refcounts for safe
4598 cleaning. Add osc abort_inflight calls in case the
4599 synchronize never started.
4602 Frequency : occasional (Cray XT3 only)
4604 Description: root not authorized to access files in CRAY_PORTALS environment
4605 Details : The client process capabilities were not honoured on the MDS in
4606 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4607 been accessed by an authorized user then root was able to access
4608 the file on the local client also. The root user capabilities
4609 are now allowed on the MDS, as this environment has secure UID.
4612 Frequency : occasional
4614 Description: ldiskfs "too long searching" message happens too often
4615 Details : A debugging message (otherwise harmless) prints too often on
4616 the OST console. This has been reduced to only happen when
4617 there are fragmentation problems on the filesystem.
4622 Description: Division by zero in statfs when all OSCs are inactive
4623 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4624 lov_statfs divides by value returned by lov_get_stripecnt().
4628 Bugzilla : 9489, 3273
4629 Description: First write from each client to each OST was only 4kB in size,
4630 to initialize client writeback cache, which caused sub-optimal
4631 RPCs and poor layout on disk for the first writen file.
4632 Details : Clients now request an initial cache grant at (re)connect time
4633 and so that they can start streaming writes to the cache right
4634 away and always do full-sized RPCs if there is enough data.
4635 If the OST is rebooted the client also re-establishes its grant
4636 so that client cached writes will be honoured under the grant.
4641 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4642 Details : Now I/O RPCs go to different portal number and (presumably) fast
4643 lock requests (and glimses) and other RPCs get their own service
4644 threads pool that should be able to service those RPCs
4647 Severity : enhancement
4649 Description: Ability to exchange lustre version between client and servers and
4650 issue warnings at client side if client is too old. Also for
4651 liblustre clients there is ability to refuse connection of too old
4653 Details : New 'version' field is added to connect data structure that is
4654 filled with version info. That info is later checked by server and
4658 Frequency : rare, liblustre only.
4659 Bugzilla : 9296, 9581
4660 Description: Two simultaneous writes from liblustre at offset within same page
4661 might proceed at the same time overwriting eachother with stale
4663 Details : I/O lock withing llu_file_prwv was released too early, before data
4664 actually was hitting the wire. Extended lock-holding time until
4665 server acknowledges receiving data.
4668 Frequency : extremely rare. Never observed in practice.
4670 Description: avoid generating lustre_handle cookie of 0.
4671 Details : class_handle_hash() generates handle cookies by incrementing
4672 global counter, and can hit 0 occasionaly (this is unlikely, but
4673 not impossible, because initial value of cookie counter is
4674 selected randonly). Value of 0 is used as a sentinel meaning
4675 "unassigned handle" --- avoid it. Also coalesce two critical
4676 sections in this function into one.
4678 Severity : enhancement
4680 Description: allow liblustre clients to delegate truncate locking to OST
4681 Details : To avoid overhead of locking, liblustre client instructs OST to
4682 take extent lock in ost_punch() on client's behalf. New connection
4683 flag is added to handle backward compatibility.
4685 Severity : enhancement
4686 Bugzilla : 4928, 7341, 9758
4687 Description: allow number of OST service threads to be specified
4688 Details : a module parameter allows the number of OST service threads
4689 to be specified via "options ost ost_num_threads={N}" in the
4690 OSS's /etc/modules.conf or /etc/modprobe.conf.
4694 Bugzilla : 6146, 9635, 9895
4695 Description: servers crash with bad pointer in target_handle_connect()
4696 Details : In rare cases when a client is reconnecting it was possible that
4697 the connection request was the last reference for that export.
4698 We would temporarily drop the export reference and get a new
4699 one, but this may have been the last reference and the export
4700 was just destroyed. Get new reference before dropping old one.
4702 Severity : enhancement
4703 Frequency : if client is started with failover MDS
4705 Description: Allow multiple MDS hostnames in the mount command
4706 Details : Try to read the configuration from all specified MDS
4707 hostnames during a client mount in case the "primary"
4710 Severity : enhancement
4712 Description: Stop sending data to evicted clients as soon as possible.
4713 Details : Check if the client we are about to send or are sending data to
4714 was evicted already. (Check is done every second of waiting,
4715 for which l_wait_event interface was extended to allow checking
4716 of exit condition at specified intervals).
4719 Frequency : rare, normally only when NFS exporting is done from client
4721 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4723 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4726 Frequency : always, if srand() is called before liblustre initialization
4728 Description: Liblustre uses system PRNG disturbing its usage by user application
4729 Details : Introduce internal to lustre fast and high-quality PRNG for
4730 lustre usage and make liblustre and some other places in generic
4731 lustre code to use it.
4733 Severity : enhancement
4734 Bugzilla : 9477, 9557, 9870
4735 Description: Verify that the MDS configuration logs are updated when xml is
4736 Details : Check if the .xml configuration logs are newer than the config
4737 logs stored on the MDS and report an error if this is the case.
4738 Request --write-conf, or allow starting with --old_conf.
4740 Severity : enhancement
4742 Description: Handle symlinks in the path when checking if Lustre is mounted.
4743 Details : Resolve intermediate symlinks when checking if a client has
4744 mounted a filesystem to avoid duplicate client mounts.
4749 Description: lconf can hit an error exception but still return success.
4750 Details : The lconf command catches the Command error exception at the top
4751 level script context and will exit with the associated exit
4752 status, but doesn't ensure that this exit status is non-zero.
4757 Description: failure of ptlrpc thread startup can cause oops
4758 Details : Starting a ptlrpc service thread can fail if there are a large
4759 number of threads or the server memory is very fragmented.
4760 Handle this without oopsing.
4763 Frequency : always, only if liblustre and non-default acceptor port was used
4765 Description: liblustre cannot connect to servers with non-default acceptor port
4766 Details : tcpnal_set_default_params() was not called and was therefore
4767 ignoring the environment varaible TCPNAL_PORT, as well as other
4768 TCPNAL_ environment variables
4773 Description: two objects could be created on the same OST for a single file
4774 Details : If an OST is down, in some cases it was possible to create two
4775 objects on a single OST for a single file. No problems other
4776 than potential performance impact and spurious error messages.
4780 Bugzilla : 5681, 9562
4781 Description: Client may oops in ll_unhash_aliases
4782 Details : Client dcache may become inconsistent in race condition.
4783 In some cases "getcwd" can fail if the current directory is
4789 Description: Inode refcounting problems in NFS export code
4790 Details : link_raw functions used to call d_instantiate without obtaining
4791 extra inode reference first.
4795 Bugzilla : 9942, 9903
4796 Description: Referencing freed requests leading to crash, memleaks with NFS.
4797 Details : We used to require that call to ll_revalidate_it was always
4798 followed by ll_lookup_it. Also with revalidate_special() it is
4799 possible to call ll_revalidate_it() twice for the same dentry
4800 even if first occurence returned success. This fix changes semantic
4801 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4802 reference on a request referred from the intent.
4803 ll_intent_release() then releases such a request.
4806 Frequency : rare, normally benchmark loads only
4808 Description: unlinked inodes were kept in memory on the client
4809 Details : If a client is repeatedly creating and unlinking files it
4810 can accumulate a lot of stale inodes in the inode slab cache.
4811 If there is no other client load running this can cause the
4812 client node to run out of memory. Instead flush old inodes
4813 from client cache that have the same inode number as a new inode.
4816 Frequency : SLES9 2.6.5 kernel and long filenames only
4817 Bugzilla : 9969, 10379
4818 Description: utime reports stale NFS file handle
4819 Details : SLES9 uses out-of-dentry names in some cases, which confused
4820 the lustre dentry revalidation. Change it to always use the
4824 Frequency : rare, unless heavy write-truncate concurrency is continuous
4825 Bugzilla : 4180, 6984, 7171, 9963, 9331
4826 Description: OST becomes very slow and/or deadlocked during object unlink
4827 Details : filter_destroy() was holding onto the parent directory lock
4828 while truncating+unlinking objects. For very large objects this
4829 may block other threads for a long time and slow overall OST
4830 responsiveness. It may also be possible to get a lock ordering
4831 deadlock in this case, or run out of journal credits because of
4832 the combined truncate+unlink. Solution is to do object truncate
4833 first in one transaction without parent lock, and then do the
4834 final unlink in a new transaction with the parent lock. This
4835 reduces the lock hold time dramatically.
4838 Frequency : rare, 2.4 kernels only
4840 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4841 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4842 before unmounting their filesystem in order to avoid deadlock.
4843 The kernel_locked() function in 2.4 kernels only checks whether
4844 the kernel lock is held, not whether it is this process that is
4845 holding it as 2.6 kernels do.
4850 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4851 Details : The client ptlrpc code may be trying to reconnect to a down
4852 server before a previous connection attempt has timed out.
4853 Increase the reconnect interval to be longer than the connection
4854 timeout interval to avoid sending duplicate connections to
4858 Frequency : echo_client brw_test command
4860 Description: fix echo_client to work with OST preallocated code
4861 Details : OST preallocation code (5137) didn't take echo_client IO path
4862 into account: echo_client calls filter methods outside of any
4863 OST thread and, hence, there is no per-thread preallocated
4864 pages and buffers to use. Solution: hijack pga pages for IO. As
4865 a byproduct, this avoids unnecessary data copying.
4869 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4870 Description: Client can oops in mdc_commit_close() after open replay
4871 Details : It was possible for the MDS to return an open request with no
4872 transaction number in mds_finish_transno() if the client was
4873 evicted, but without actually returning an error. Clients
4874 would later try to replay that open and may trip an assertion
4875 Simplify the client close codepath, and always return an error
4876 from the MDS in case the open is not successful.
4879 Frequency : rare, 2.6 OSTs only
4881 Description: OST may deadlock under high load on fragmented files
4882 Details : If there was a heavy load and highly-fragmented OST filesystems
4883 it was possible to have all the OST threads deadlock waiting on
4884 allocation of biovecs, because the biovecs were not released
4885 until the entire RPC IO was completed. Instead, release biovecs
4886 as soon as they are complete to ensure forward IO progress.
4888 Severity : enhancement
4890 Description: Support for specifying external journal device at mount
4891 Details : If an OST or MDS device is formatted with an external journal
4892 device, this device major/minor is stored in the ext3 superblock
4893 and may not be valid for failover. Allow detecting and
4894 specifying the external journal at mount time.
4899 Description: Mounting an MDS with pending unlinked files may cause oops
4900 Details : target_finish_recovery() calls mds_postrecov() which returned
4901 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4902 considers this an error and immediately begins cleaning up the
4903 lov, just after starting the mds_lov process
4905 Severity : enhancement
4907 Description: Implement 'lfs df' to report actual free space on per-OST basis
4908 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4909 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4910 usage of objects; '-h' to report in human readable format.
4912 ------------------------------------------------------------------------------
4914 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4921 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4922 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4923 at startup during orphan unlinking. Instead, skip buddy update
4924 before setup, as it will be regenerated after recovery is complete.
4929 Description: If an OST is inactive, its locks might reference stale inodes.
4930 Details : lov_change_cbdata() must iterate over all namespaces, even if
4931 they are inactive to clear inode references from the lock.
4933 Severity : enhancement
4934 Frequency : occasional, if non-standard max_dirty_mb used
4936 Description: Client will block write RPCs if not enough grant
4937 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4938 then the client will block writes while waiting for another RPC
4939 to complete instead of consuming its dirty limit. With change
4940 we get improved performance when max_dirty_mb is small.
4942 Severity : enhancement
4943 Bugzilla : 3389, 6253
4944 Description: Add support for supplementary groups on the MDS.
4945 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4946 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4947 lookups for user supplementary groups into a cache.
4951 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4952 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4953 for liblustre O_EXCL is enforced by MDS.
4957 Description: readdir never returns NULL in liblustre.
4958 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4959 dentry in struct dirent.
4963 Frequency : liblustre only, depends on application IO pattern
4964 Description: liblustre clients evicted if not contacting servers
4965 Details : Don't put liblustre clients into the ping_evictor list, so
4966 they will not be evicted by the pinger ever.
4968 Severity : enhancement
4970 Description: Add ability to evict clients by NID from MDS.
4971 Details : By echoing "nid:$NID" string into
4972 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4973 $NID would be instantly evicted from this MDS and from all active
4974 OSTs connected to it.
4978 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4979 Details : lookup_it_finish() used to query file size from OSTs that was not
4984 Description: service threads change working directory to that of init
4985 Details : Starting lustre service threads may pin the working directory
4986 of the parent thread, making that filesystem busy. Threads
4987 now change to the working directory of init to avoid this.
4991 Frequency : during shutdown only
4992 Description: shutdown with a failed MDS or OST can cause unmount to hang
4993 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4998 Frequency : 2.6 only
4999 Description: chmod/chown may include an extra supplementary group
5000 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5001 supplementary group and if none is specified this is used.
5004 Bugzilla : 5479 (6816)
5006 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5007 Details : If lookup is in progress on a file that is unlinked we might try
5008 to revalidate the inode and fail in revalidate after lookup is
5009 complete and ll_file_open() enqueues the open again but
5010 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5013 Frequency : always, if lconf --abort_recovery used
5015 Description: lconf --abort_recovery fails with 'Operation not supported'
5016 Details : lconf was attempting to abort recovery on the MDT device and not
5019 ------------------------------------------------------------------------------
5021 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5026 Frequency : rare (only unsupported configurations with a node running as an
5028 Bugzilla : 6514, 5137
5029 Description: Mounting a Lustre file system on a node running as an OST could
5031 Details : OSTs now preallocates memory needed to write out data at
5032 startup, instead of when needed, to avoid having to
5033 allocate memory in possibly low memory situations.
5034 Specifically, if the file system is mounted on on OST,
5035 memory pressure could force it to try to write out data,
5036 which it needed to allocate memory to do. Due to the low
5037 memory, it would be unable to do so and the node would
5038 become unresponsive.
5040 Severity : enhancement
5042 Description: Addition of lconf --service command line option
5043 Details : lconf now accepts a '--service <arg>' option, which is
5044 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5046 Severity : enhancement
5048 Description: Failover mode is now the default for OSTs.
5049 Details : By default, OSTs will now run in failover mode. To return to
5050 the old behaviour, add '--failout' to the lmc line for OSTs.
5052 Severity : enhancement
5054 Description: Health checks are now provided for MDS and OSTs
5055 Details : Additional detailed health check information on MSD and OSTs
5056 is now provided through the procfs health_check value.
5059 Frequency : occasional, depends on IO load
5061 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5062 after numerous create/delete cycles
5063 Details : The ext3 inode allocation policy would not allocate new inodes
5064 very well on the OSTs because there are no new directories
5065 being created. Instead we look for groups with free space if
5066 the parent directories are nearly full.
5071 Description: Network or server problems during mount may cause partially
5072 mounted clients instead of returning an error.
5073 Details : The config llog parsing code may overwrite the error return
5074 code during mount error handling, returning success instead
5080 Description: MDS can fail to allocate large reply buffers
5081 Details : After long uptimes the MDS can fail to allocate large reply
5082 buffers (e.g. zconf client mount config records) due to memory
5083 fragmentation or consumption by the buffer cache. Preallocate
5084 some large reply buffers so that these replies can be sent even
5085 under memory pressure.
5089 Frequency : rare (liblustre)
5090 Description: fsx running with liblustre complained that using truncate() to
5091 extend the file doesn't work. This patch corrects that issue.
5092 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5093 ATTR_SIZE and lsm use in llu_setattr_raw.
5097 Frequency : rare, only 2.6 kernels
5098 Description: Unusual file access patterns on the MDS may result in inode
5099 data being lost in very rare circumstances.
5100 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5101 code in the 2.6 kernel caused the inode and/or EA data not to
5102 be read from disk, causing single-file corruption.
5106 Frequency : rare, only 2.6 filesystems using extents
5107 Description: Heavy concurrent write and delete load may cause data corruption.
5108 Details : It was possible under high-load situations to have an extent
5109 metadata block in the block device cache from a just-unlinked
5110 file overwrite a newly-allocated data block. We now unmap any
5111 metadata buffers that alias just-allocated data blocks.
5115 Frequency : filesystems with default stripe_count larger than 77
5116 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5117 Details : lconf specifies an inode size of 4096 bytes when the default
5118 stripe_count is larger than 77. This conflicts with the default
5119 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5120 this case to avoid pinning too much memory for large EAs.
5122 ------------------------------------------------------------------------------
5124 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5129 Frequency : rare (extremely heavy IO load with hundreds of clients)
5131 Description: Client is evicted, gets IO error writing to file
5132 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5133 caused clients to be evicted for AST timeouts. The fixes in
5134 bug 5192 mean we no longer need to have such short AST timeouts
5135 so ldlm_timeout has been increased.
5138 Frequency : occasional during --force or --failover shutdown under load
5139 Bugzilla : 5949, 4834
5140 Description: Server oops/LBUG if stopped with --force or --failover under load
5141 Details : a collection of import/export refcount and cleanup ordering
5142 issues fixed for safer force cleanup
5145 Frequency : only filesystems larger than 120 OSTs
5146 Bugzilla : 5990, 6223
5147 Description: lfs getstripe would oops on a very large filesystem
5148 Details : lov_getconfig used kfree on vmalloc'd memory
5151 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5152 Bugzilla : 6242, 6243
5153 Description: reading from files that had been truncated to a non-zero size
5154 but never opened returned no data
5155 Details : ll_file_read() reads zeros from no-object files to EOF
5160 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5161 Details : There's little harm in aborting MDS/OSS recovery and letting it
5162 try again, so I removed the LASSERT and return an error instead.
5164 Severity : enhancement
5166 Description: New debugging infrastructure for tracking down data corruption
5167 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5168 (b) cover more of the client-side code path, and (c) try to narrow
5169 down where problems occurred
5173 Bugzilla : 3819, 4364, 4397, 6313
5174 Description: Racing close and eviction MDS could cause assertion in mds_close
5175 Details : It was possible to get multiple mfd references during close and
5176 client eviction, leading to one thread referencing a freed mfd.
5178 Severity: : enhancement
5179 Bugzilla : 3262, 6359
5180 Description: Attempts to reconnect to servers are now more aggressive.
5181 Details : This builds on the enhanced upcall-less recovery that was added
5182 in 1.4.2. When trying to reconnect to servers, clients will
5183 now try each server in the failover group every 10 seconds. By
5184 default, clients would previously try one server every 25 seconds.
5189 Description: After recovery, certain operations trigger a failed
5190 assertion on a client.
5191 Details : Failing over an mds, using lconf -d --failover, while a
5192 client was doing a readdir() call would cause the client to
5193 LBUG after recovery completed and the readdir() was resent.
5195 Severity : enhancement
5197 Description: Default groups are now added by lconf
5198 Details : You can now run lconf --group <servicename> without having to
5199 manually add groups with lmc.
5202 Frequency : occasional
5204 Description: Nodes with an elan id of 0 trigger a failed assertion
5207 Frequency : always when accessing e.g. tty/console device nodes
5209 Description: tty and some other devices nodes cannot be used on lustre
5210 Details : file's private_data field is used by device data and lustre
5211 values in there got lost. New field was added to struct file to
5212 store fs-specific private data.
5215 Frequency : when exporting Lustre via NFS
5217 Description: NFSD failed occasionally when looking up a path component
5218 Details : NFSD is looking up ".." which was broken in ext3 directories
5219 that had grown large enough to become hashed.
5222 Frequency : Clusters with multiple interfaces not on the same subnet
5224 Description: Nodes will repeatedly try to reconnect to an interface which it
5225 cannot reach and report an error to the log.
5226 Details : Extra peer list entries will be created by lconf with some peers
5227 unreachable. lconf now validates the peer before adding it.
5230 Frequency : Only if a default stripe is set on the filesystem root.
5232 Description: Setting a default stripe on the filesystem root prevented the
5233 filesystem from being remounted.
5234 Details : The client was sending extra request flags in the root getattr
5235 request and did not allocate a reply buffer for the dir EA.
5238 Frequency : occasional, higher if lots of files are accessed by one client
5239 Bugzilla : 6159, 6097
5240 Description: Client trips assertion regarding lsm mismatch/magic
5241 Details : While revalidating inodes the VFS looks up inodes with ifind()
5242 and in rare cases can find an inode that is being freed.
5243 The ll_test_inode() code will free the lsm during ifind()
5244 when it finds an existing inode and then the VFS later attaches
5245 this free lsm to a new inode.
5249 Bugzilla : 6422, 7030
5250 Description: MDS deadlock between mkdir and client eviction
5251 Details : Creating a new file via mkdir or mknod (starting a transaction
5252 and getting the ns lock) can deadlock with client eviction
5253 (gets ns lock and trying to finish a synchronous transaction).
5256 Frequency : occasional
5257 Description: While starting a server, the fsfilt_ext3 module could not be
5259 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5260 kernels. Previously, lconf would still use the ext3 name
5261 when trying to load modules. Now, it will correctly use
5262 ext3 on 2.4 and ldiskfs on 2.6.
5264 Severity : enhancement
5265 Description: The default stripe count has been changed to 1
5266 Details : The interpretation of the default stripe count (0, to lfs
5267 or lmc) has been changed to mean striping across a single
5268 OST, rather than all available. For general usage we have
5269 found a stripe count of 1 or 2 works best.
5271 Severity : enhancement
5272 Description: Add support for compiling against Cray portals.
5273 Details : Conditional compiling for some areas that are different
5277 Frequency : occasional
5278 Bugzilla : 6409, 6834
5279 Description: Creating files with an explicit stripe count may lead to
5280 a failed assertion on the MDS
5281 Details : If some OSTs are full or unavailable, creating files may
5282 trigger a failed assertion on the MDS. Now, Lustre will
5283 try to use other servers or return an error to the
5287 Frequency : occasional
5289 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5290 caused lock timeout during readahead (since 1.4.2).
5291 Details : Processes doing readahead might match a lock that hasn't been
5292 granted yet if there are overlapping and conflicting lock
5293 requests. The readahead process waits on ungranted lock
5294 (original lock is CBPENDING), while OST waits for that process
5295 to cancel CBPENDING read lock and eventually evicts client.
5297 Severity : enhancement
5299 Description: Initial enabling of flock support for clients
5300 Details : Implements fcntl advisory locking and file status functions.
5301 This feature is provided as an optional mount flag (default
5302 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5303 locking are implemented yet, and those that are are not guaranteed
5304 to be completely correct in production environments.
5305 mount -t lustre -o [flock|noflock] ...
5308 Frequency : occasional
5310 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5311 assertion in the ext3 JBD (journaling) layer.
5312 Details : The b_committed_data struct is protected by the big kernel lock
5313 in 2.4 kernels, serializing journal_commit_transaction() and
5314 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5315 access to this struct. In 2.6 kernels there is finer grained
5316 locking to improve SMP performance of the JBD layer.
5320 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5325 Description: merge and cleanup kernel patches.
5328 -----------------------------------------------------------------------------
5330 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5332 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5333 versions of Lustre. Please update all servers and clients to
5334 version 1.4.2 or later at the same time. You must also run
5335 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5336 to update the configuration logs.
5338 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5339 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5340 - swab llog records of type '0' so we get proper header size/idx (5861)
5341 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5342 - fix rename of one directory over another leaking an inode (5953)
5343 - avoid SetPageDirty on 2.6 (5981)
5344 - don't re-add just-being-destroyed locks to the waiting list (5653)
5345 - when creating new directories, inherit the parent's custom
5346 striping settings if present parent (3048)
5347 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5348 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5349 - don't hold client locks on temporary worklist from l_lru (5666)
5350 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5351 - automatically evict dead clients (5921)
5352 - Update file size properly in create+truncate+fstat case (6196)
5353 - Do not unhash mountpoint dentries, do not allow removal of
5355 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5356 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5357 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5358 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5359 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5360 - Correctly handle reads of files with no objects (6243)
5361 - lctl recover will also mark a device active if deactivate used (5933)
5363 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5364 - allow --write-conf on an MDS with different nettype than client (5619)
5365 - don't write config llogs to MDS for mounts not from that MDS (5617)
5366 - lconf should create multiple TCP connections from a client (5201)
5367 - init scripts are now turned off by default; run chkconfig --on
5368 lustre and chkconfig --on lustrefs to use them
5369 - upcalls are no longer needed for clients to recover to failover
5371 - add --abort-recovery option to lconf to abort recovery on device
5373 - add support for an arbitrary number of OSTs (3026)
5374 - Quota support protocol changes.
5375 - forward compatibility changes to wire structs (6007)
5376 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5377 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5378 - improved "lustre status" script
5379 - initialize blocksize for non-regular files (6062)
5380 - added --disable-server and --disable-client configure options (5782)
5381 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5382 - Vanilla 2.4.29 support
5383 - increase maximum number of obd devices to 520 (6242)
5384 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5385 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5386 SLES 9 2.6.5 kernels
5387 - limit stripes per file to 160 (the maximum EA size) (6093)
5389 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5392 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5393 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5394 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5395 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5396 - small ext3 extents cleanups and fixes (5733)
5397 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5398 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5399 - avoid needless client->OST connect, fix handle mismatch (5317)
5400 - fix DLM error path that led to out-of-sync client, long delays (5779)
5401 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5402 - fix several locking issues related to i_size (5492,5624,5654,5672)
5403 - don't move pending lock onto export if it is already evicted (5683)
5404 - fix kernel oops when creating .foo in unlinked directory (5548)
5405 - fix deadlock in obdfilter statistics vs. object create (5811)
5406 - use time_{before,after} to avoid timer jiffies wrap (5882)
5407 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5408 - Do not leak request if server was not able to process it (5154)
5409 - If mds_open unable to find parent dir, make that negative lookup(5154)
5410 - don't create new directories with extent-mapping (5909, 5936)
5412 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5413 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5414 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5415 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5416 - support for RHEL4 kernel on i686 (5773)
5417 - provide error messages when incompatible logs are encountered (5898)
5419 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5420 * version 1.4.0.10 (1.4.1 release candidate 1)
5422 - don't keep a lock reference when lock is not granted (4238)
5423 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5424 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5425 - drop import inflight refcount on signal_completed_replay error (5255)
5426 - unlock page after async write error during send (3677)
5427 - handle missing objects in filter_preprw_read properly (5265)
5428 - no transno return for symlink open, don't save no-trasno open (3440)
5429 - don't try to complete elan receive that already failed (4012)
5430 - free RPC server reply state on error (5406)
5431 - clean up thread from ptlrpc_start_thread() on error (5160)
5432 - readahead could read extra page into cache that wasn't ejected (5388)
5433 - prevent races in class_attach/setup/cleanup/detach (5260)
5434 - don't dereference de->d_inode after l_dput of de (5458)
5435 - use "int" for stripe value returned from lock_to_stripe (5544)
5436 - mballoc allocation and error-checking fixes in 2.6 (5504)
5437 - block device patches to fix I/O request sizes in 2.6 (5482)
5438 - look up hostnames for IB nals (5602)
5439 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5440 - don't start multiple acceptors for the same port (5277)
5441 - fix incorrect LASSERT in mds_getattr_name (5635)
5442 - export a proc file for general "ping" checking (5628)
5443 - fix "lfs check" to not block when the MDS is down (5628)
5445 - service request history (4965)
5446 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5447 - create an "evict_client" /proc entry on OSTs, like the MDS has
5448 - fix mount usage message, return errors per mount(8) (5168)
5449 - change grep [] to grep "[]" in tests so they work in more UMLs
5450 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5451 - remove ancient LOV_MAGIC_V0 EA support (5047)
5452 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5453 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5454 - allow readahead allocations to fail when low on memory (5383)
5455 - mmap locking landed again, after considerable improvement (2828)
5456 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5458 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5461 - send OST transaction number in read/write reply to free req (4966)
5462 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5463 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5464 - the watchdog thread now runs as interruptible (5246)
5465 - flock/lockf fixes (but it's still disabled, pending 5135)
5466 - don't use EXT3 constants in llite code (5094)
5467 - memory shortage at startup could cause assertion (5176)
5469 - reorganization of lov code
5470 - single portals codebase
5472 - add extents/mballoc support (5025)
5473 - direct I/O reads in the obdfilter (4048)
5474 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5476 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5479 - send OST transaction number in read/write reply to free req (4966)
5480 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5481 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5482 - the watchdog thread now runs as interruptible (5246)
5483 - handle missing objects in filter_preprw_read properly (5265)
5484 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5485 - drop import inflight refcount on signal_completed_replay error (5255)
5486 - unlock page after async write error during send (3677)
5487 - return original error code on reconstructed replies (3761)
5488 - no transno return for symlink open, don't save no-trasno open (3440)
5490 - add pid to ldlm debugging output (4922)
5491 - bump the watchdog timeouts -- we can't handle 30sec yet
5492 - extra debugging for orphan dentry/inode bug (5259)
5494 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5497 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5498 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5499 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5500 - fix NULL dereference in /proc/sys/portals/routes (4827)
5501 - allow failed mdc_close() operations to be interrupted (4561)
5502 - stop precreate on OST before MDS would time out on it (4778)
5503 - don't send partial-page writes before EOF from client (4410)
5504 - discard client grant for sub-page writes on large-page clients (4520)
5505 - don't free dentries not owned by NFS code, check generation (4806)
5506 - fix lsm leak if mds_create_objects() fails (4801)
5507 - limit debug_daemon file size, always print CERROR messages (4789)
5508 - use transno after validating reply (3892)
5509 - process timed out requests if import state changes (3754)
5510 - update mtime on OST during writes, return in glimpse (4829)
5511 - add mkfsoptions to LDAP (4679)
5512 - use ->max_readahead method instead of zapping global ra (5039)
5513 - don't interrupt __l_wait_event() during strace
5515 - add software watchdogs to catch hung threads quickly (4941)
5516 - make lustrefs init script start after nfs is mounted
5517 - fix CWARN/ERROR duplication (4930)
5518 - return async write errors to application if possible (2248)
5519 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5520 - print NAL number in %x format (4645)
5521 - update barely-supported suse-2.4.21-171 series (4842)
5522 - support for sles 9 %post scripts
5523 - support for building 2.6 kernel-source packages
5524 - support for sles km_* packages
5526 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5529 - ignore -ENOENT errors in osc_destroy (3639)
5530 - notify osc create thread that OSC is being cleaned up (4600)
5531 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5532 - reconstruct ost_handle() like mds_handle() (4657)
5533 - create a new thread to do import eviction to avoid deadlock (3969)
5534 - let lconf resolve symlinked-to devices (4629)
5535 - don't unlink "objects" from directory with default EA (4554)
5536 - hold socknal file ref over connect in case target is down (4394)
5537 - allow more than 32000 subdirectories in a single directory (3244)
5538 - fix blocks count for O_DIRECT writes (3751)
5539 - OST returns ENOSPC from object create when no space left (4539)
5540 - don't send truncate RPC if file size isn't changing (4410)
5541 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5542 - bind to privileged port in socknal and tcpnal (3689)
5544 - rate limit CERROR/CWARN console message to avoid overload (4519)
5545 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5546 - basic mmap support (3918)
5547 - kernel patch series update from b1_4 (4711)
5549 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5552 - avoid crash during MDS cleanup with OST shut down (2775)
5553 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5554 - don't use bad inodes on the MDS (3744)
5555 - dynamic object preallocation to improve recovery speed (4236)
5556 - don't hold spinlock over lock dumping or change debug flags (4401)
5557 - don't zero obd_dev when it is force cleaned (3651)
5558 - print grants to console if they go negative (4431)
5559 - "lctl deactivate" will stop automatic recovery attempts (3406)
5560 - look for existing locks in ldlm_handle_enqueue() (3764)
5561 - don't resolve lock handle twice in recovery avoiding race (4401)
5562 - revalidate should check working dir is a directory (4134)
5564 - don't always mark "slow" obdfilter messages as errors (4418)
5566 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5569 - don't close LustreDB during write_conf until it is done (3860)
5570 - fix typo in lconf for_each_profile (3821)
5571 - allow dumping logs from multiple threads at one time (3820)
5572 - don't allow multiple threads in OSC recovery (3812)
5573 - fix debug_size parameters (3864)
5574 - fix mds_postrecov to initialize import for llog ctxt (3121)
5575 - replace config semaphore with spinlock (3306)
5576 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5577 - don't allow enqueue to complete on a destroyed export (3822)
5578 - down write_lock before checking llog header bitmap (3825)
5579 - recover from lock replay timeout (3764)
5580 - up llog sem before sending rpc (3652)
5581 - reduce ns lock hold times when setting kms (3267)
5582 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5583 - fix NULL deref and obd_dev leak on setup error (3312)
5584 - replace some LBUG about llog ops with error handling (3841)
5585 - don't match INVALID dentries from d_lookup and spin (3784)
5586 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5587 - fix invalid assertion in ptlrpc_set_wait (3880)
5589 - add libwrap support for the TCP acceptor (3996)
5590 - add /proc/sys/portals/routes for non-root route listing (3994)
5591 - allow setting MDS UUID in .xml (2580)
5592 - print the stack of a process that LBUGs (4228)
5594 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5597 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5598 - make sure to unset replay flag from failed open requests (3440)
5599 - if default stripe count is 0, use OST count for inode size (3636)
5600 - update parent mtime/ctime on client for create/unlink (2611)
5601 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5602 - free recovery state on server during a forced cleanup (3571)
5603 - unregister_reply for resent reqs (3063)
5604 - loop back devices mounting and status check on 2.6 (3563)
5605 - fix resource-creation race that can provoke i_size == 0 (3513)
5606 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5607 - more debugging for page-accounting assertion (3746)
5608 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5609 - avoid deadlock after precreation failure (3758)
5610 - fix race and lock order deadlock in orphan handling (3450, 3750)
5611 - add validity checks when grabbing inodes from l_ast_data (3599)
5613 - add /proc/.../recovery_status to obdfilter (3428)
5614 - lightweight CDEBUG infrastructure, debug daemon (3668)
5615 - change default OSC RPC parameters to be better on small clusters
5616 - turn off OST read cache for files smaller than 32MB
5617 - install man pages and include them in rpms (3100)
5618 - add new init script for (un)mounting lustre filesystems (2593)
5619 - run chkconfig in %post for init scripts (3701)
5620 - drop scimac NAL (unmaintained)
5622 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5625 - clean kiobufs before and after use (3485)
5626 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5627 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5628 - change init script to fail more clearly if not run as root (1528)
5629 - allow clients to reconnect during replay (1742)
5630 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5631 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5632 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5633 - fix cond_resched (3554)
5634 - only evict unfinished clients after recovery (3515)
5635 - allow bulk resend, prevent data loss (3570)
5636 - dynamic ptlrpc request buffer allocation (2102)
5637 - don't allow unlinking open directory if it isn't empty (2904)
5638 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5639 - remove extraneous obd dereference causing LASSERT failure (3334)
5640 - don't use get_cycles() when creating temp. files on the mds (3156)
5641 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5642 - handle EEXIST for set-stripe, set proper directory name (3336)
5644 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5645 - fix ksocknal_fmb_callback() error messages (2918)
5647 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5650 - don't copy lvb into (possibly NULL) reply on error (2983)
5651 - don't deref dentry after dput, don't free lvb on error (2922)
5652 - use the kms to determine writeback rpc length (2947)
5653 - increment oti_logcookies when osc is inactive (2948)
5654 - update client's i_blocks count via lvb messages (2543)
5655 - handle intent open/close of special files properly (1557)
5656 - mount MDS with errors=remount-ro, like obdfilter (2009)
5657 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5658 - don't use cancelling-locks' kms values (2947)
5659 - use highest lock extent for kms, not last one (2925)
5660 - don't dereference ERR_PTR() dentry in error handling path (3107)
5661 - fix thread race in portals_debug_dumplog() (3122)
5662 - create lprocfs device entries at setup instead of at attach (1519)
5663 - common AST error handler, don't evict client on completion race (3145)
5664 - zero nameidata in detach_mnt in 2.6 (3118)
5665 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5666 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5667 - initialize RPC timeout timer earlier for 2.6 (3219)
5668 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5669 - print nal/nid for unknown nid (3258)
5670 - additional checks for oscc recovery before doing precreate (3284)
5671 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5672 - don't crash in mdc_close for bad permissions on open (3285)
5673 - zero i_rdev for non-device files (3147)
5674 - clear page->private before handing to FS, better assertion (3119)
5675 - tune the read pipeline (3236)
5676 - fix incorrect decref of invalidated dentry (2350)
5677 - provide read-ahead stats and refine rpc in flight stats (3328)
5678 - don't hold journal transaction open across create RPC (3313)
5679 - update atime on MDS at close time (3265)
5680 - close LDAP connection when recovering to avoid server load (3315)
5681 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5682 - don't leak open file on MDS after open resend (3325)
5683 - serialize filter_precreate and filter_destroy_precreated (3329)
5684 - loop device shouldn't call sync_dev() for nul device (3092)
5685 - clear page cache after eviction (2766)
5686 - resynchronize MDS->OST in background (2824)
5687 - refuse to mount the same filesystem twice on same mountpoint (3394)
5688 - allow llmount to create routes for mounting behind routers (3320)
5689 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5690 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5691 - fix rare problem with rename on htree directories (3417)
5693 - allow default OST striping configuration per directory (1414)
5694 - fix compilation for qswnal for 2.6 kernels (3125)
5695 - increase maximum number of MDS request buffers for large systems
5696 - change liblustreapi to be useful for external progs like lfsck (3098)
5697 - increase local configuration timeout for slow disks (3353)
5698 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5700 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5703 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5704 - don't overwrite extent policy data in reply if lock was blocked (2901)
5705 - drop filter export grants atomically with removal from device (2663)
5706 - del obd_self_export from work_list in class_disconnect_exports (2908)
5707 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5708 - swab reply message in mdc_close, other PPC fixes (2464)
5709 - fix destroying of named logs (2325)
5710 - overwrite old logs when running lconf --write_conf (2264)
5711 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5712 - fix race in target_handle_connect (2898)
5713 - mds_reint_create() should take same inode create lock (2926)
5714 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5715 - don't close files for self_export to avoid uninitialized obd (2936)
5716 - allow MDS with the same name as client node (2939)
5717 - hold dentry reference for closed log files for unlink (2325)
5718 - reserve space for all logs during transactions (2059)
5719 - don't evict page beyond end of stripe extent (2925)
5720 - don't oops on a deleted current working directory (2399)
5721 - handle hard links to targets without a parent properly (2517)
5722 - don't dereference NULL lock when racing during eviction (2867)
5723 - don't grow lock extents when lots of conflicting locks (2919)
5725 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5728 - account for cache space usage on clients to avoid data loss (974)
5729 - lfsck support in lustre kernel code (2349)
5730 - reduce journal credits needed for BRW writes (2370)
5731 - orphan handling to avoid losing space on client/server crashes
5732 - ptlrpcd can be blocked, stopping ALL progress (2477)
5733 - use lock value blocks to assist in proper KMS, faster stat (1021)
5734 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5735 - recovery for initial connections (2355)
5736 - fixes for mds_cleanup_orphans (1934)
5737 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5738 - block all file creations until orphan recovery completes (1901)
5739 - client remove rq_connection from request struct (2423)
5740 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5741 - recovery timer race (2670)
5742 - mdc_close recovey bug (2532)
5743 - ptlrpc cleanup bug (2710)
5744 - mds timeout on local locks (2588)
5745 - namespace lock held during RPCs (2431)
5746 - handle interrupted sync write properly (2503)
5747 - don't try to handle a message that hasn't been replied to (2699)
5748 - client assert failure during cleanup after abort recovery (2701)
5749 - leak mdc device after failed mount (2712)
5750 - ptlrpc_check_set allows timedout requests to complete (2714)
5751 - wait for inflight reqs when ptlrpcd finishes (2710)
5752 - make sure unregistered services are removed from the srv_list
5753 - reset bulk XID's when resending them (caught by 1138 test)
5754 - unregister_bulk after timeout
5755 - fix lconf error (2694)
5756 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5757 - readahead locks pages, leaves pending causing memory pressure (2673)
5758 - increase OST request buffers to 4096 on large machines (2729)
5759 - fix up permission of existing directories in simple_mkdir (2661)
5760 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5761 - don't assign transno to errored transactions (2742)
5762 - don't delete objects on OST if given a bogus objid from MDS (2751)
5763 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5764 - if rq_no_resend, then timeout request after recovery (2432)
5765 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5766 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5767 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5768 - MDS recovery shouldn't delete objects that it has given out (2730)
5769 - if enqueue arrives after completion, don't clobber LVB (2819)
5770 - don't unlock pages twice when trigger_group_io returns error (2814)
5771 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5772 - don't write pages to disk if there was an error (1450)
5773 - don't ping imports that have recovery disabled (2676)
5774 - take buffered bytes into account when balancing socknal conn (2817)
5775 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5776 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5777 - remove little-endian swabbing of llog records (1987)
5778 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5779 - retry reposting request buffers if they fail (1191)
5780 - grow extent at grant time to avoid granting a revoked lock (2809)
5781 - lock revoke doesn't evict page if covered by a second lock (2765)
5782 - disable VM readahead to avoid reading outside lock extents (2805)
5784 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5785 - updated kernel patches for hp-2.4.20 kernel (2681)
5787 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5790 - fix truncated write corruption (2366)
5791 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5793 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5794 - ensure len(uuid) < 37 in lmc (1171)
5795 - fix ia64 OOPS in llog_test (2255)
5796 - zero end of page at obdfilter for partial page writes (2648)
5797 - don't leave stale dentries around after renames (bug 2428)
5798 - fix timeouts when evicting a client with a single lock held (2642)
5799 - set deadline for the initial HELLO message to drain (2634)
5800 - print out dotted-quad IP addresses in the socknal (2302)
5802 - additional debugging for MDS client eviction problem (2443)
5803 - fix mkfsoptions support for osts (2603, 2604)
5805 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5808 - add series for the vanilla 2.6.0 kernel
5809 - add series for the vanilla 2.4.24 kernel
5810 - add series for a cray x86/64 UL kernel drop
5811 - fix xattr patches for the vanilla 2.4.19 series
5813 - generate true UUIDs in lmc (1171)
5814 - have portals stack dumping break in UML (2466)
5815 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5816 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5817 - wake callback waiting threads on client eviction (2460)
5818 - Add --ptldebug and --subsystem to lmc (1719)
5819 - update assertion to allow safe interrupt allocation
5820 - set rq_no_resend for cancel requests (2432)
5821 - recalculate ptlrpcd timeout after resend (2494)
5822 - call vfs_rmdir when removing pending directories (2368)
5823 - fix renaming a file to itself (2429)
5824 - lmc creates a default one-stripe lov (2454)
5825 - expand procfs space to handle large clusters (2326)
5826 - increase UML stack to avoid overflow
5827 - update lconf's list of debug and subsystem masks
5828 - fix lfs find --obd (2510)
5829 - /proc tunable for disabling filter read caching (2591)
5830 - stop rpm packages from altering slapd.conf (2301)
5831 - disable nagle in the socknal under 0conf (2578)
5832 - choose mds inode size based on stripe count (2572)
5833 - fix kernel-source rpm problems (2516)
5835 - add --disable-doc to avoid pdf generation (2421)
5836 - update documentation, tests, type-os, comments
5837 - avoid format warnings on ia64
5838 - remove the TOE NAL
5839 - tiny code cleanups by removing unused fields
5841 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5844 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5845 - remove the most blatant lies from BUILDING file (2371)
5846 - change default debug level to reasonable production setting
5847 - reduce client side cache size to reduce cache flush time
5848 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5849 - make TCP zerocopy and pinger support enabled by default (2476)
5850 - sync writes completed after process exits caused crashes (2319)
5851 - maintain correct mount count on the MDS (2356)
5852 - backout 1557, because 2316 wasn't really fixed
5853 - better file I/O statistics gathering in /proc
5854 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5855 - another kernel patch to fix zero-copy TCP function export
5856 - don't take duplicate lock when processing re-sent getattr (2420)
5857 - lctl uses obd_self_export instead of creating new conn (2353)
5858 - MDS/OST recovery case which requires object creation asserted (2425)
5859 - move lfs from /usr/sbin to /usr/bin in packages
5860 - fix race between mds_client_add and mds_client_free (2417)
5861 - use kmalloc instead of slabs in portals (2430)
5862 - don't create duplicate records when a failover MDS is present (2442)
5863 - remove unnecessary mount age check (2332)
5864 - don't remove directory inodes from locks prematurely (2451)
5865 - don't break if MDS service name is the same as hostname (2103)
5866 - fix races in client write RPC generation when cache full (2482)
5868 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5871 - remove now-unused request->rq_obd (278)
5872 - if an allocation fails, print out how much memory we've used (1933)
5873 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5874 - add the "configurable stack size" patch to most series files (1256)
5875 - ability to write large log records, for 100+ OST configs (2306)
5876 - fix NULL deref when filter_prep fails (2314)
5877 - fix operator precedence error in filter_sync
5878 - dynamic allocation of socknal TX descriptors (2315)
5879 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5880 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5881 - ensure that CWARN messages in Portals always get to the syslog
5882 - __init/__exit are not for prototype decls (ldlm_init/exit)
5883 - x86-64 compile warning fixes
5884 - fix gateway LMC keyword conflict (2318)
5885 - fix MDS lock inversions in getattr/reint paths (1844)
5886 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5887 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5888 - DLM race condition prevented some lock evictions (2328)
5889 - ENOMEM detection and retry on socknal sends (2230)
5890 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5891 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5892 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5893 - better error messages when a client is rejected during recovery (1505)
5894 - avoid cancelling locks which were never granted, after failure (2330)
5895 - fix i_sem/journal inversion in mds_client_add (2333)
5896 - fix truncate/getattr lock cycle deadlock (2334)
5897 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5898 - fix two rare exit paths which could leak an l_lock() ref (2321)
5899 - fencepost error in MDS/OST orphan recovery (2226)
5900 - make log record alignment 8 bytes (1988)
5901 - lstripe now fails when requested offset > ost_count (2237)
5902 - ensure that all kernel series have a complete list.h (1607)
5903 - fix crashes in special-file operations (2316)
5904 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5906 - allow configurable automake binary, for testing new versions
5907 - small update to the lfs documentation
5909 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5911 * fix negative export reference count in fsfilt_sync (2312)
5913 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5914 * release candidate 0.9.1
5916 - orphans are moved into the PENDING directory for possible recovery
5917 - replayed opens now open by fid for orphan/rename safety (1042)
5918 - last close of an orphan inode generates a transno (683)
5919 - chdir() and mount() now pin the directory entry (1020)
5920 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5921 - discard very old requests without processing them (1502)
5922 - remove some common, well-understood CERRORs (1505)
5923 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5924 - clear "grant" flags in OST replies until OST grant code lands (1644)
5925 - fix read performance by not clobbering i_blksize on client (1598)
5926 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5927 - make LRU size a /proc tunable, clears locks when reduced (707)
5928 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5929 - server locks take references on exports now (1558)
5930 - build fixes for 2.4.20-rh trees (1663)
5931 - return an error from lov_create if all OSCs are inactive (1751)
5932 - fix import levels when a reconnect happens without a timeout (1597)
5933 - exit early from mds_open if we get a lookup error (1749)
5934 - partial page read at EOF wouldn't wait for disk before sending (1642)
5935 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5936 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5937 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5938 - write/truncate lock inversion (1639)
5939 - Don't auto-load obdclass, portals modules during cleanup (1495)
5940 - fix timestamps from jumping to "now" (1763)
5941 - extra journal assertions (1648)
5942 - add an extra multiunlink test (1771)
5943 - fix read_record/write_record API (1776)
5944 - fix leak of offset_extent, possible incorrect i_size later (1772)
5945 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5946 - Debugging for the kqswnal_get_idle_tx problems (1820)
5947 - Allow recovery to be attempted multiple times (1536)
5948 - Write out MDS last_rcvd file after it is first created (1600)
5949 - Fix tx_descriptor leak in failed transmit situations (1827)
5950 - ext3 journaling fixes for assertion failure after IO error (1871)
5951 - class_export_put() on freed export after completion AST error (1896)
5952 - Fix revalidate looping in VFS (1322)
5953 - Don't access a freed export during MDS_REINT timeout (1521)
5954 - Add open-unlink recovery support on the MDS (1673,1764)
5955 - Return an error if no MDS data was read from last_rcvd (1946)
5956 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5957 - Don't setup a disk device that doesn't match exported UUID (317)
5958 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5959 - avoid committing NULL handle in force close
5960 - local.sh is now a one-stripe LOV configuration
5961 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5962 - fix ext3 htree duplicate directory entry corruption (1516)
5963 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5964 - update to new LOV EA format (2097)
5965 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5967 2003-06-15 Phil Schwan <phil@clusterfs.com>
5970 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5971 - per-import/export recovery handling (958, 931, 959)
5972 - multiple last-rcvd slots, for serving multiple FSes (949)
5973 - connections are again shared between multiple imp/exports (963, 964)
5974 - "umount -f" would hang if any requests needed to be sent (393, 978)
5975 - avoid pinning large req buffer by copying for queued messages (989)
5976 - add "uuid" to "lctl device" command to help upcalls (991)
5977 - "open" RPCs with transnos would confuse recovery counters (1037)
5978 - do proper endian conversion of all wire messages (288, 340, 891)
5979 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5980 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5981 - fix ptlrpc_connection leak in target_handle_connect (1174)
5982 - fix import refcounting bug in OST and MDS cleanup (1134)
5983 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5984 - fix very unlikely obd_types race condition (501)
5985 - remove osc_open hack for echo_client (1187)
5986 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5987 - a failure in read_inode2 leads to deadlock (1139)
5988 - cancel ack-locks as soon as transaction is committed (1072)
5989 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5990 - make sure to commitrw after any preprw to avoid deadlock (1162)
5991 - failing to execute a file in a lustre FS would lock inode (1203)
5992 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5993 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5994 - obd_uuid2tgt would walk off the end of the list (1255)
5995 - on IA64 the osc would give portals incorrect bulk size (1258)
5996 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5997 - fix lock inversion caused by new llite matching code (1282)
5998 - limit the number of dirty pages on a client to 10MB (1286)
5999 - timed out locks were not being corrected cancelled (1289)
6000 - fix O_DIRECT above 4GB on IA-32 (1292)
6001 * major user-visible changes
6002 - fail out/fail over policy now controlled by the upcall (993)
6004 - add OBD_PING to check server availability and failure (954)
6005 - lustre messages are now sent in sending host order (288, 340, 891)
6006 - add eadatalen to MDS getattr reply (340)
6007 - OST read replies may contain second buffer, with per-page status (593)
6009 2003-03-11 Phil Schwan <phil@clusterfs.com>
6012 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6013 - failed open()s could cause deadlock; fixed (867, 869)
6014 - stop cancelling OST locks when files are closed (481)
6015 - overlapping XID spaces caused network corruption (851, 853)
6016 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6017 - setattr_raw added, to do single-RPC, server-side setattrs
6018 - lmc/lconf syntax change for OST UUIDs
6019 - fix crashy race condition between ptlrpc_free_req and osc_close
6020 - don't use request in mdc_enqueue if we hit a timeout (889)
6021 - don't set the inode i_size for regular files from the MDS (896)
6022 - handle out of order completion AST (842)
6023 - don't LBUG if a lock request times out after receiving AST (913)
6024 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6025 - if a bad lock AST arrives, send an error instead of dropping entirely
6026 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6027 - fix leak in bulk IO when only partially completed (899, 900, 926)
6028 - fix O_DIRECT for ia64 (55)
6029 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6030 - C-z after timeout could hang a process forever; fixed (977)
6032 - client-side I/O cache (678, 924, 929, 941, 970)
6034 - READPAGE and SETATTRs which don't take server-side locks get
6037 2003-02-11 Phil Schwan <phil@clusterfs.com>
6040 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6041 - Fix fsfilt_extN_readpage() to read a full page of directory
6042 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6043 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6044 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6045 - Most intent operations were removed, in favour of a new RPC mode
6046 that does a single RPC to the server and bypasses most of the VFS
6047 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6048 - Aggressively cancel local locks on DLM servers
6049 - mds_reint_unlink sends EA to the client if it's the last nlink.
6050 client uses that EA to unlink OST objects.
6051 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6052 - recursive symlinks were fixed (439)
6053 - fixed NULL deref in DEBUG_REQ
6054 - filter_update_lastobjid no longer calls sync, which annoyed extN
6055 - fixed multi-client small-writes to a single file problem (445)
6056 - fixed mtime updates during file writes (607)
6057 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6058 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6059 - fixed Linux OST size reporting problem (444, 656)
6060 - OST now updates object mtime with writes or setattr (607, 619)
6061 - client verifies file size before zeroing page past EOF (445)
6062 - OST now writes last allocated objid to disk with allocation (108)
6063 - LOV on echo now works (409)
6065 - mds_reint_unlink sends a new buffer, with the EA included. this
6066 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6067 set if a regular file was being unlinked, and it was the last link
6068 - use PtlGet from the target for bulk writes (315)
6069 - OST now updates object mtime with writes or setattr (607, 619)
6070 - LDLM now has a grant-time callback to revalidate locked items, if
6072 - Many MDS operations were reorganized to combat race conditions
6074 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6075 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6077 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6080 - Fully reactivate OST imports after reconnection (512, others)
6081 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6082 - More graceful error handling for truncating on dead OST (515)
6083 - Don't error out unless we're actually accessing dead stripes (474)
6084 - Fix garbage sizes when stripes are missing (410)
6085 - LRU counters were broken, causing constant lock purge (433, 432)
6086 - garbage on read from stripes with failed OSTs (441)
6087 - mark OSCs as active before reconnecting during recovery (438)
6088 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6089 - lfind did not preserve OST order in output (443)
6090 - symlinks cause hung clients, incorrect data (439)
6091 - stop dereferencing request after dropping refcount (457)
6092 - don't LASSERT(spin_is_locked) on non-SMP (455)
6093 - fixes for many rename() bugs
6094 - fstat didn't correctly synchronize attributes (399)
6095 - server must handle lock cancellation during blocking AST prep (487)
6096 - bulk descriptors were free()d too soon (511)
6097 - fix paths in lconf, which would load incorrect modules (451, 507)
6098 - fix confusing lconf 'host not found' error message (386)
6099 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6100 - fix race condition in mdc_blocking_ast() for inode access (526)
6101 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6102 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6103 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6104 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6105 - fix lstripe in several ways (kernel side) (527)
6106 - fix request leak in ldlm_cli_enqueue (262)
6107 - incorrect OSC was marked inactive after OST failure
6108 - call mds_fs_cleanup before unmounting filesystem (524)
6109 - fix races between taking ns_lock and ldlm_lock_change_resource
6110 - fix races updating LOV export open file list
6111 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6112 - fix recovery NULL deref in ldlm_cli_cancel_unused
6113 - fix some DLM races by using new hash table for lock handles (419)
6114 - permit the client to specify desired inodes, at replay
6115 - duplicate requests when we queue them for replay reintegration
6116 - fix last_rcvd offset calculation
6117 - sync after each recovered transaction, so we always make progress
6118 - never, not always, ERESTART requests without transnos
6119 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6121 - skip replay if the MDS says that the client is already connected
6122 - don't check for a recovery-enabled export to match lctl's UUID
6123 - don't INC_USE_COUNT for phantom exports
6124 - don't crash when cleaning up phantom exports (567)
6125 - don't double-finish or set replay data for errored mdc_open requests
6126 - abort requests when they time out, so we don't get old replies
6127 - send/receive replies for AST messages again
6128 - if the client says that it doesn't have the lock, cancel it on the
6130 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6131 mark it as destroyed, it will all work out when decref is called
6132 - fix module use counts (22, 581)
6134 - ASTs now expect a reply (server cancels lock on error reply)
6136 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6139 - fix many simultaneous client startup (392)
6140 - fix dentry->d_it clobbering
6141 - credentials weren't being shipped for readdir/getattr operations
6142 - remove invalid assertions triggered during some concurrent MD
6144 - proper Lustre versions added (336, 389)
6145 - fix memory leak for create error case (398)
6146 - fix LOV locking bug that would get cli/srv out of sync
6147 - fix echo client over LOV (409)
6148 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6149 - fix double-O_EXCL intent crash (424)
6150 - avoid sending multiple lock CANCELs (352)
6152 - MDS can do multi-client recovery (modulo bugs in new code)
6154 - many updates, edits, cleanups
6156 2002-11-18 Phil Schwan <phil@clusterfs.com>
6159 - fix null d_it dereference (346)
6160 - fix full OST/dbench hang (333)
6161 - fix permission problem with file removal (286)
6162 - fix removal of OSCs from LOV when they fail
6163 - fix NULL deref during bulk timeout (214)
6164 - fix problems related to multiple filesystems on one MDS (241)
6165 - fixed serious subtle metadata locking bugs
6166 - free locks on clients when inodes are removed due to memory
6168 - fix inode pointer in lock data (285)
6169 - partial support for multiple MDS on a single host (241)
6170 - data locks weren't cancelled at clear_inode time (290, 311)
6171 - intent locks could lead to unbounded lock growth (205)
6172 - added a maximum lock count, an LRU list, and a flusher
6173 - fix multiple rename (365)
6174 - properly abstracted the echo client
6175 - OSC locked 1 byte too many; fixed
6176 - rewrote brw callback code:
6177 - fixed recovery bugs related to LOVs (306)
6178 - fixed too-many-pages-in-one-write crash (191)
6179 - fixed (again) crash in sync_io_timeout (214)
6180 - probably fixed callback-related race (385)
6182 - Add capability to MDS protocol
6183 - LDLM cancellations and callbacks on different portals
6185 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6188 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6189 - defer open object destruction to close time (601981, 138)
6190 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6191 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6192 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6193 - don't call obd_setattr in ll_file_release for destroyed objects
6194 * protocol change to lustre_msg: move |version| and add |flags|
6195 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6196 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6197 * added replay of create, unlink, link and rename operations during
6198 MDS failover; recovery should be much more robust now
6199 * remove failed OSCs from LOVs (only lov_create uses this so far)
6200 * the lustre-HOWTO was brought (more) up to date (582544)
6202 2002-10-23 Phil Schwan <phil@clusterfs.com>
6205 - in-use dentries weren't being reused properly (617851)
6206 - prevent multiple LDLM setup (599178)
6207 - fix LOV size calculations for truncate (617853)
6208 - fix client handling of MDS intent errors (POSIX)
6209 - fix permission bug in lovstripe.c test (624321)
6210 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6211 - truncate past end of file could corrupt data
6212 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6213 - a race in recovery could return ETIMEDOUT to apps (623947)
6214 - building outside the source directory was fixed
6215 * the lustre-HOWTO was brought (more) up to date (582544)
6216 * major progress was made on recovery functionality
6218 2002-10-10 Phil Schwan <phil@clusterfs.com>
6221 - recovery deadlock fix
6222 - rm -rf causes LBUG fix (617817)
6223 - file open by multiple tasks fix (618962)
6224 - directory permissions bugs (602707 and 620007)
6225 - journal_stop fixed with locking (611313)
6226 - O_APPEND failures resolved (618273, perhaps 614459)
6227 - lconf PATH fix (619770)
6228 - IA64 build fix (621450)
6229 - RPC buffer sizes scale with amount of memory
6231 2002-10-01 Phil Schwan <phil@clusterfs.com>
6234 - locks would be cancelled without throwing away data pages,
6235 resulting in inconsistent data (605627)
6236 - inode attributes were not always being refreshed (605627, 612449)
6237 - lconf now continues to cleanup after lctl reports an error
6238 - MDS now enforces user permissions (602707)
6239 - lprocfs cleanup fixed, but not yet enabled (614157)
6240 - fixed infinite server hang, should a client not respond to an AST
6241 - avoid going into recovery if user calls readlink() with a buffer
6242 that's too small (613941)
6243 - AST RPCs no longer require replies (614867) -- this may be changed
6244 - don't crash server if client sends an IOV that's too big (611336)
6245 - fixed lock conversion deadlock (611892)
6246 - fixed the following of symlinks (614622)
6247 * recovery: the server can remove locks from a client that dies, other
6248 clients can make progress
6249 * more extN patch fixes
6250 * compile-time configurable ptlrpc buffer allocations
6252 - collaborative read cache document
6253 - Lustre Lite Performance CDR document-in-progress
6255 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6258 - fix typo in patch-2.4.18
6260 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6263 - clear ptlrpc request each time in handle_incoming_request()
6264 - unlink of files now destroys the object on the OST
6266 2002-09-19 Peter Braam <braam@clusterfs.com>
6268 * add hard link support
6269 * change obdfile creation method
6270 * kernel patch changed
6272 2002-09-19 Peter Braam <braam@clusterfs.com>
6275 - stack overflow bug in extN fixed
6277 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6279 * documentation updates
6280 - add man pages for config tools
6281 - update tests/README to describe testing with new config tools
6282 - finish metadata API descriptions
6283 * bug fixes and cleanups
6284 - statfs workaround for 16TB limit
6285 - LOV stripe allocation improved, can stripe on subset of OSTs
6286 - LOV file size/IO offset was wrong for files > 4GB in size
6287 - object EA data was being dropped, caused files to be unreadable
6288 - memory overflow with non-LOV OST caused memory corruption
6289 - fixed regression tests to work with new config tools, obdfilter
6290 - fixed bug when directory size became larger than 1 block
6291 - fixed bug (for single client case) when PWD was deleted
6292 - invalidate local directory pages when doing intent-based ops
6293 - avoid LDLM oops when lock callback contained bad data
6295 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6297 * documentation updates
6298 * bug fixes and cleanups
6299 - configuration tools
6302 - 64-bit compile warnings
6303 - 64-bit internal statfs data
6305 * test_brw on persistent OST devices
6307 * lprocfs (disabled)
6309 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6311 * documentation updates
6312 * bug fixes and cleanups
6313 * configuration tools
6315 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6318 * many small fixes to 0_5_4
6319 * io/network handling
6320 * thinkos in MDS operations
6322 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6325 * crucial basic fixes to 0.5.3
6326 * IOR, Iozone work over Elan
6329 2002-08-07 Phil Schwan <phil@clusterfs.com>
6330 * version 0_5_3, our first alpha
6331 * we use the new Portals iovs
6332 * documentation updates
6333 * bug fixes and cleanups
6334 * small changes in the DLM wire protocol
6336 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6337 * version 0_5_1 with some initial stability,
6338 * locking on MD and file I/O.
6339 * documentation updates
6340 * several bug fixes since 0.5.0
6341 * small changes in wire protocol
6343 2002-07-18 Phil Schwan <phil@clusterfs.com>
6345 * delivered as Lustre Light Alpha
6346 * fixed a crash after handling invalid MDS requests
6347 * fixed directory pages for architectures with non-4k pages sizes
6349 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6350 * release version v0_4_4
6351 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6353 * Fixes a number of interruption problems with OST operations.
6354 * Update documentation for portals header changes
6355 * Move all wire protocol structs/defines to lustre_idl.h
6356 * Fixes symlink length bug.
6357 * Add tcpdump to repository.
6359 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6360 * release version v0_4_3
6361 * Fixes statfs for inodes on extN.
6362 * Fixes bug in runtests which would delete /etc/hosts.
6363 * Use 64-bit object IDs wherever possible (not into VFS though)
6364 Remove ost_get_info, which is unused by lustre, and out of date.
6366 2002-07-03 Peter Braam <braam@clusterfs.com>
6367 * release version v0_4_2 Fixes a lookup error (type not passed)
6368 * move forward to head of Portals
6369 * move forward to latest Lustre kernel
6371 2002-06-25 Peter Braam <braam@clusterfs.com>
6372 * release version v0_4_1. Hopefully stable on single node use.